Research Handbook of Financial Markets 180037531X, 9781800375314

Citation preview

Research Handbook of Financial Markets

Editor Refet S. Grkaynak Jonathan H. Wright

RESEARCH HANDBOOK OF FINANCIAL MARKETS

RESEARCH HANDBOOKS IN MONEY AND FINANCE The Research Handbooks in Money and Finance series presents a thorough analysis of recent scholarly developments in monetary and financial economics, forming an essential, authoritative and comprehensive reference guide to the field. Edited by esteemed international scholars, these Handbooks contain a wide range of specially-commissioned chapters covering the latest advances and research, and aim to be prestigious, high-quality works of lasting significance. Each Handbook consists of original contributions by an international team of scholars, and contributes to both the expansion of current debates and the development of future research. For a full list of Edward Elgar published titles, including the titles in this series, visit our website at www.e-elgar.com.

Research Handbook of Financial Markets Edited by

Refet S. Gürkaynak Professor of Economics, Bilkent University, Turkey and CEPR, CESIfo, and CFS

Jonathan H. Wright Professor of Economics, Johns Hopkins University, USA and NBER

RESEARCH HANDBOOKS IN MONEY AND FINANCE

Cheltenham, UK · Northampton, MA, USA

© Refet S. Gürkaynak and Jonathan H. Wright 2023 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA

A catalogue record for this book is available from the British Library Library of Congress Control Number: 2023933463

This book is available electronically in the Economics subject collection http://dx​.doi​.org​/10​.4337​/9781800375321

ISBN 978 1 80037 531 4 (cased) ISBN 978 1 80037 532 1 (eBook)

EEP BoX

Contents

vii

List of contributors Introduction to the Research Handbook of Financial Markets  Refet S. Gürkaynak and Jonathan H. Wright

1

PART I   CENTRAL BANKING 1

The Federal Reserve balance sheet Kristopher Dawsey, William B. English and Brian Sack

6

2

The balance sheet of the Eurosystem Oreste Tristani

33

3

The Bank of Japan’s balance sheet Kosuke Aoki

56

4

Central bank lending Brian Madigan and William Nelson

79

5

The workings of liquidity lines between central banks Saleem Bahaj and Ricardo Reis

102

PART II  INTERMEDIARIES 6 Banks Refet S. Gürkaynak, Jonathan H. Wright and Egon Zakrajšek

126

7

Non-bank financial intermediaries and financial stability Sirio Aramonte, Andreas Schrimpf and Hyun Song Shin

147

8

Government agencies: Fannie Mae and Freddie Mac Gillian Burgess, Wayne Passmore and Shane M. Sherlund

171

9

Money market funds Antoine Bouveret, Antoine Martin and Patrick E. McCabe

194

PART III   MONEY MARKETS 10

The federal funds market, pre- and post-2008 Eric T. Swanson

220

11

The repo market Benjamin Munyan

237

v

vi  Research handbook of financial markets

12

The foreign exchange market Alain Chaboud, Dagfinn Rime and Vladyslav Sushko

253

PART IV   CAPITAL MARKETS 13

The Treasury and when-issued markets J. Benson Durham and Roberto Perli

277

14

The municipal bond market Daniel Bergstresser

301

15

Mortgage-backed securities Andreas Fuster, David Lucca and James Vickery

331

16

Equity trading Caroline Fohlin

358

17

Sovereign debt Leonardo Martinez, Francisco Roch, Francisco Roldán and Jeromin Zettelmeyer

378

PART V  DERIVATIVE MARKETS 18

Interest rate swaps Bin Wei and Vivian Z. Yue

407

19

Credit default swaps Antulio N. Bomfim

429

20

Foreign exchange swaps and cross-currency swaps Angelo Ranaldo

451

21

Inflation hedging products Stefania D’Amico and Thomas B. King

470

22

Futures and options Refet S. Gürkaynak and Jonathan H. Wright

490

Index

509

Contributors

Kosuke Aoki, University of Tokyo Sirio Aramonte, Board of Governors of the Federal Reserve System Saleem Bahaj, University College London Daniel Bergstresser, Brandeis University Antulio N. Bomfim, Northern Trust Asset Management Antoine Bouveret, European Securities and Markets Authority Gillian Burgess, Board of Governors of the Federal Reserve System Alain Chaboud, Board of Governors of the Federal Reserve System Stefania D’Amico, Federal Reserve Bank of Chicago Kristopher Dawsey, The D. E. Shaw Group J. Benson Durham, Piper Sandler William B. English, Yale School of Management Caroline Fohlin, Emory University and CEPR Andreas Fuster, EPFL, Swiss Finance Institute and CEPR Refet S. Gürkaynak, Bilkent University, CEPR, CESIfo, and CFS Thomas B. King, Federal Reserve Bank of Chicago David Lucca, Jane Street Patrick E. McCabe, Board of Governors of the Federal Reserve System Brian Madigan, Georgetown University Antoine Martin, Federal Reserve Bank of New York Leonardo Martinez, International Monetary Fund Benjamin Munyan, Federal Reserve Bank of Dallas William Nelson, Bank Policy Institute and Georgetown University Wayne Passmore, Board of Governors of the Federal Reserve System Roberto Perli, Federal Reserve Bank of New York Angelo Ranaldo, University of St.Gallen Ricardo Reis, London School of Economics vii

viii  Research handbook of financial markets

Dagfinn Rime, BI Norwegian Business School Francisco Roch, International Monetary Fund and UTDT Francisco Roldán, International Monetary Fund Brian Sack, The D. E. Shaw Group Andreas Schrimpf, Bank for International Settlements Shane M. Sherlund, Board of Governors of the Federal Reserve System Hyun Song Shin, Bank for International Settlements Vladyslav Sushko, Bank for International Settlements Eric T. Swanson, University of California, Irvine and NBER Oreste Tristani, European Central Bank and CEPR James Vickery, Federal Reserve Bank of Philadelphia Bin Wei, Federal Reserve Bank of Atlanta Jonathan H. Wright, Johns Hopkins University and NBER Vivian Z. Yue, Emory University, Federal Reserve Bank of Atlanta, NBER, and CEPR Egon Zakrajšek, Bank for International Settlements and CEPR Jeromin Zettelmeyer, Bruegel

Introduction to the Research Handbook of Financial Markets Refet S. Gürkaynak and Jonathan H. Wright Financial markets have been important for millennia and have been a favorite subject of water cooler talk as well as a focus of research. Investors and regulators have also always tried to understand these markets and their relationship with each other better. Indeed, well-functioning financial markets are essential to a strong economy—a point made clear in the Global Financial Crisis of 2008–2009, when a breakdown in financial intermediation triggered an adverse feedback loop with the macroeconomy. In the period since, amid the European debt crisis and then the Covid shock, these markets have become even more important for investors, policymakers, and researchers alike. These are markets with myriad rules and conventions, where neither the mechanics of transactions nor the factors shaping prices are straightforward, often leading to instability and proneness to repeated crises that can trigger recessions. Indeed, excessive credit growth leading to financial instability is the cause of many severe economic downturns—a point highlighted by Hyman Minsky that has received more attention since his death. The view that finance was a “veil”, meaning that the financial cycle had no causal effect on the business cycle, used to be prevalent, but that is no longer the mainstream understanding. Now finance and macroeconomics are thought to be closely related. Only a joint analysis of finance and macroeconomics allows for studying financial crises that have been part of the economic landscape for centuries, as well as their welfare effects. This volume takes the perspective that finance is an integral part of macroeconomics. It brings together the foremost experts on various facets of financial markets to provide analyses of the market mechanics, key players, and asset prices. Each chapter explains the history and current state of the market it focuses on. In many cases, the history goes back a very long way indeed. These analytical chapters frame the market in question from a research perspective, discussing what we know about the market, what we learn from price formation, and what the open questions are. Although many of the chapters contain insights into asset pricing, this is not the sole focus. Market structure is also considered at some length. The chapters collected in this volume are resources for market participants, regulators, and researchers alike. For the uninitiated, they provide a deep first look into each market; for the expert, they suggest new ways of seeing these markets. All the chapters place special emphasis on recent developments such as high-frequency trading and regulatory changes since the Global Financial Crisis. Collectively, they afford a holistic understanding of financial markets themselves and the state of research into these fascinating mechanisms for human interaction. Financial markets are partly regulated by central banks and monetary policy forms an important input into price formation in any financial market. And central banking has taken a new turn after the turmoil of the past 15 years, with unconventional policies adopted by central banks leading to balance sheet sizes that were unthinkable before the Global Financial Crisis and the ensuing monetary expansion. These balance sheet actions have now become so commonplace that they are scarcely “unconventional” anymore. It is impossible to understand financial markets without understanding the behavior of central banks driving them. 1

2  Research handbook of financial markets

Part I of the volume is therefore dedicated to providing a deep understanding of major central banks and their policies. Chapter 1, by Kristopher Dawsey, William B. English, and Brian Sack, discusses the balance sheet of the Federal Reserve and argues that the Fed can and should manage its balance sheet in a way to provide macroeconomic and financial stability. Chapter 2, by Oreste Tristani, considers the balance sheet of the Eurosystem. The euro area has the distinct feature that there is no central fiscal authority: there is a monetary union, but not a fiscal union. This chapter evaluates the implications of this for the European Central Bank balance sheet. Chapter 3, by Kosuke Aoki, discusses the balance sheet of the Bank of Japan, which has some unique features, notably that it holds exchange rate funds and real estate investment trusts. Chapter 4, by Brian Madigan and William Nelson, studies the long history of central bank lending, how the nature of central bank lending has evolved over time, and the more recent lending policies that led central banks to have historically large balance sheets. This chapter includes a discussion of the potential drawbacks of the recent trend to larger central bank balance sheets. Chapter 5, by Saleem Bahaj and Ricardo Reis, discusses the liquidity lines between central banks that first emerged in the Global Financial Crisis and have become an important backstop to the international financial system. While financial markets as a whole are based on the intermediation of savings and credit, an imperfect classification of institutions and markets into somewhat more specific categories is possible. The non-central banking content of the volume is divided into four more parts comprising intermediaries, money markets, capital markets, and derivatives markets. Part II, on intermediaries, sheds light on four major types of institutions. Chapter 6, by the editors and Egon Zakrajšek, discusses the special role of banks. Banks are a potentially unstable form of financial intermediation, and yet they are ubiquitous—this chapter expounds on the reasons. The chapter on banks also discusses stablecoins and a future with central bank digital currencies, as well as providing an understanding of what banks are and how they function. Non-bank intermediation—an ever-larger share of financial intermediation—is discussed in Chapter 7, by Sirio Aramonte, Andreas Schrimpf, and Hyun Song Shin. This chapter emphasizes how systemic risks can be propagated by nonbank financial institutions. Chapter 8, by Gillian Burgess, Wayne Passmore, and Shane M. Sherlund, covers government agencies, Fannie Mae and Freddie Mac, which are important players in the US housing market. These used to be privately owned companies but have now fallen into US Treasury conservatorship. This part of the book is rounded up by Chapter 9 on money market funds, by Antoine Bouveret, Antoine Martin, and Patrick E. McCabe. Money market funds function similarly to banks and are similarly vulnerable to runs, but they are regulated quite differently and do not have deposit insurance. This chapter considers money market funds in both the US and elsewhere and emphasizes how frequent runs and crises in money market funds are. Part III of the volume is about money markets (not including money market funds, which we think of as intermediaries rather than money markets themselves). The three largest money markets are the interbank market, the repo market, and the foreign exchange market. The repo and interbank markets are both markets in short-term debt. In Chapter 10, Eric T. Swanson studies the US domestic market for uncollateralized interbank loans, known as the federal funds market. Supply of liquidity by the Federal Reserve into this market and the resulting price formation (the effective federal funds rate) have changed significantly after the Global Financial Crisis and so the pre- and post-crisis states of the federal funds market are discussed separately. In Chapter 11, Benjamin Munyan discusses the US repo market, showing how it is a key component of the plumbing of the financial system. Chapter 12, by Alain Chaboud,

Introduction  3

Dagfinn Rime, and Vladyslav Sushko, considers recent developments in the foreign exchange market. The foreign exchange market is relatively unregulated, compared to most bond and equity markets. Banks used to dominate the provision of liquidity in the foreign exchange market, but this role has recently been taken on by high-frequency trading firms. Part IV of the volume is on capital markets, where issuers sell securities to investors. The most important capital markets are the Treasury, municipal bond, mortgage-backed security, equity, and sovereign debt markets, which are covered in this part. These markets, with the exception of equities, are considered fixed income because the securities promise a fixed payment to the holder. Treasury securities that are issued by the US government are widely considered to be effectively free of default risk. This market is reviewed in Chapter 13 by J. Benson Durham and Roberto Perli. The chapter on Treasury markets considers both the primary market (Treasury auctions) and the secondary market and discusses the important but little-known when-issued market. This is a forward market where the right to a Treasury security can be traded after the auction announcement, but before the security has actually been issued. The chapter also considers the decomposition of Treasury yields into expectations of future short rates and term premia. Chapter 14, by Daniel Bergstresser, discusses the municipal bond market, where US state and local governments are able to borrow with somewhat distinct regulatory and tax treatments. This market features an enormous number of different issuers and the securities can be quite illiquid. Nonetheless, municipal bonds have a key role in financing infrastructure spending in the US. Chapter 15, by Andreas Fuster, David Lucca, and James Vickery, deals with US mortgage-backed securities, which are claims to the cash flow on bundles of mortgages. This is a market that is relatively recent—going back to 1968— but it is now a very large market. The chapter emphasizes agency mortgage-backed securities that were created by government agencies and carry their insurance against default risk. Mortgage-backed securities greatly expanded the investor base for mortgages. Chapter 16, by Caroline Fohlin, discusses equity markets including both history and recent developments such as the rise of algorithmic trading. This chapter also includes a discussion of equity pricing including newer topics such as machine learning. Although the equity market is smaller than the bond market in terms of market capitalization, it is uniquely central to both academic finance and the popular perception of financial markets. Chapter 17, by Leonardo Martinez, Francisco Roch, Francisco Roldán, and Jeromin Zettelmeyer, moves beyond the US Treasury market to consider international sovereign debt markets. It explores why sovereign debt is seen as risky in some countries and not others and discusses policy options to reduce sovereign debt risk. Part V of the book comprises chapters on the key derivatives markets. In Chapter 18, Bin Wei and Vivian Z. Yue describe interest rate swaps. The chapter considers types of interest rate swaps and how they are priced. Recently there has been the anomaly of “negative swap spreads” where fixed swap rates are lower than corresponding maturity Treasury yields, and the chapter devotes some space to discussing possible reasons for this anomaly. Chapter 19, by Antulio N. Bomfim, considers credit default swaps. These derivatives are useful to investors for hedging and can be used to reverse-engineer investor beliefs about the likelihood of default. But they were also a major contributor to the Global Financial Crisis and were famously referred to by Warren Buffet as “financial weapons of mass destruction”. Chapter 20, by Angelo Ranaldo, discusses foreign exchange and cross-currency swaps. Foreign exchange swaps are the most traded instrument in the foreign exchange market and this chapter examines the institutional framework and recent trends in this important financial market. This

4  Research handbook of financial markets

relates back to Chapter 5, which considered a foreign exchange swap facility but one that is put in place by central banks. Chapter 21, by Stefania D’Amico and Thomas B. King, deals with the surprisingly difficult problem of hedging inflation risk. The optimal hedge is argued to depend on the price index and horizon being considered by the investor. The chapter also makes the important point that whereas in the wake of the Great Inflation of the 1970s, investors were willing to pay a risk premium to hedge against inflation risk, deflation risk became a bigger concern after the Global Financial Crisis. Finally, Chapter 22, by the editors, considers the long history, market structure, pricing, and usage of futures and options and illustrates how they can be used as a rich source of information on investors’ beliefs. This volume can be read almost like a textbook on financial markets, from cover to cover. The chapters are ordered to facilitate such a reading. But each chapter is a standalone treatise on the market it studies and as such this is a reference book for practitioners and researchers alike. The chapters carefully map out the state of the market and research into that market, explicitly highlighting the open questions. The editors hope that readers of this volume will be motivated to seek answers to those.

Online appendices for chapters 7 and 13 are available on the companion website at: https://www.e-elgar. com/textbooks/gurkaynak.

PART I CENTRAL BANKING

1. The Federal Reserve balance sheet1 Kristopher Dawsey, William B. English and Brian Sack

1.1 INTRODUCTION1 The manner in which the Federal Reserve sets monetary policy and manages its balance sheet plays a critical role in financial markets. Its importance is perhaps most evident during periods of severe financial stress when the Federal Reserve (Fed) and other central banks have used their balance sheets to restore order to financial markets. But the presence of the Fed matters at all other times as well, as its asset holdings and its administered rates affect how financial markets operate and strongly influence the financial conditions experienced by all households and firms operating in the financial system in the United States and even globally.2 The Federal Reserve’s policy decisions affect financial markets in a number of ways. First, through its monetary policy decisions and its communications about those decisions, the Fed effectively sets short-term interest rates and influences longer-term interest rates that, in turn, drive the valuations of assets across markets (Gurkaynak, Sack, & Swanson, 2005). Second, since the Global Financial Crisis (GFC), the Federal Reserve has engaged in large-scale asset purchases aimed at having additional influence on longer-term yields and broader financial conditions (Bernanke, 2012). Finally, when managing crises, the Fed can use its balance sheet to support financial firms and the functioning of financial markets. For example, the Fed has provided loans to financial firms facing funding difficulties both directly and through swap lines with foreign central banks. In addition, the Fed has conducted purchases of assets to support market functioning, including the very large purchases of Treasuries and agency mortgage-backed securities (MBS) conducted in response to the Covid-19 pandemic. The liability side of the Fed’s balance sheet can also have significant effects on financial markets. Prior to the GFC, the Fed implemented monetary policy primarily by adjusting the supply of reserve balances held by banks. More recently, the level of reserve balances has increased greatly to accommodate the vast rise in Fed assets, and the Fed has launched an overnight repurchase program that has also become sizable. The Fed has used the payment of interest on reserves and on overnight reverse repurchase agreements to aid with the implementation of policy. The Fed’s liabilities also include the Treasury General Account, the account that the US Treasury Department uses to manage federal finances, which has also expanded greatly in size in recent years. A range of other institutions, including government agencies, designated financial market utilities, and foreign and international monetary authorities also hold funds in accounts at the Fed. 1 We thank the participants in the Research Handbook of Financial Markets Conference sponsored by Bilkent University and CEPR, and particularly our discussants, Brian Madigan and Bill Nelson, as well as the editors, for useful comments and suggestions. Sean Fulmer provided excellent research assistance. All remaining errors are ours. 2 Of course, the Fed also affects markets by establishing and enforcing regulations on banking institutions and other activities. This chapter focuses on the Fed’s balance-sheet-related actions. 6

The Federal Reserve balance sheet  7

In the next part of this chapter, we provide a description of the mechanics of the Federal Reserve balance sheet and a brief history of how it has changed over time. We then give a more detailed description of the Fed’s balance sheet prior to the GFC, followed by a discussion of the changes to the balance sheet during and after the crisis and during the pandemic, and the reasons for those changes. We go on to discuss the new policy implementation framework necessitated by the changes in the balance sheet. We then turn to the effects of lending programs and purchases of private assets on the balance sheet. And we end with some thoughts on the possible future evolution of the Fed’s balance sheet and the key decisions that the Fed will have to make regarding its size and composition.

1.2 THE MECHANICS OF THE FED’S BALANCE SHEET The size and structure of the Federal Reserve’s balance sheet are governed by a range of policy decisions. The largest influence on the balance sheet comes from the monetary policy decisions made by the Federal Open Market Committee (FOMC), which is the main monetary policymaking body of the Federal Reserve System. In recent years, the FOMC has set a target range for the overnight interest rate that it uses as its primary policy instrument (the federal funds rate) and has undertaken additional balance sheet actions (such as purchases of Treasury securities and agency MBS) to meet its economic objectives. However, the balance sheet is also affected by decisions by the Board of Governors, which sets the parameters for the Fed’s lending activities as well as the rate the Fed pays on reserve balances. A snapshot of the Fed’s balance sheet on a recent date is reported in Table 1.1. As can be seen, the Fed held nearly $9 trillion of assets at that time, which of course represents a remarkable size for any single portfolio. As with any balance sheet, these asset holdings were funded with a corresponding set of liabilities, which on this date totaled about $8.9 trillion. For the

Table 1.1  The Federal Reserve’s balance sheet, March 9, 2022 Assets

$ billions

Liabilities

$ billions

Treasury securities

5,753.4

Federal Reserve notes

2,209.4

Agency debt and agency mortgagebacked securities

2,693.7

Reverse repurchase agreements

1,786.2

Discount window loans

2.3

Reserves of banks

3,953.7

Section 13(3) loans

42.2

Treasury general account

609.4

Central bank liquidity swaps

0.3

Foreign official and other deposits

282.0

Foreign currency denominated assets

20.0

Other liabilities

28.9

Other assets

398.8

Total

8,910.7

Total

8,869.6

Notes:   Section 13(3) loans include the Commercial Paper Funding Facility II, the Corporate Credit Facilities, the Main Street lending program, the Municipal Liquidity Facility, the Term Asset-Backed Security Loan Facility II, and the Paycheck Protection Program Liquidity Facility. Figure is net of the Treasury contributions to credit facilities. For a list of sources of information on the Federal Reserve balance sheet, see Appendix 1.1. Source:   Federal Reserve, H.4.1. Statistical Release, March 10, 2022.

8  Research handbook of financial markets

discussion in the remainder of this chapter, it will be important to understand the mechanics of how these assets and liabilities end up on the Fed’s balance sheet. The largest component of assets on the balance sheet is the Fed’s securities portfolio, which is comprised almost entirely of US Treasury securities, agency debt, and agency-backed mortgage-backed securities.3 These assets are purchased in the market by the Open Market Desk (the Desk) at the Federal Reserve Bank of New York (the New York Fed), and the Treasury securities at most times are rolled over at Treasury auctions upon maturity. The authority to purchase these securities comes from Section 14 of the Federal Reserve Act, which grants the Fed permission to buy and sell in the open market “bonds and notes of the United States,… [and] any obligation which is a direct obligation of, or fully guaranteed as to principal and interest by, any agency of the United States.” These holdings of securities reflect two broad purposes, which are described in more detail later (in Sections 1.4 and 1.5). First, the Fed’s asset holdings grow over time to create the liquidity that a growing economy and financial system needs, as reflected in the liabilities on the Fed’s balance sheet (described next). Second, as noted earlier, the FOMC has used asset purchases as a policy instrument since the GFC. These securities are held in what is called the System Open Market Account (SOMA) portfolio, which simply represents the collective portfolio of the Fed’s Reserve Banks. The FOMC is able to assign the authority to conduct operations to any Reserve Bank, but it has for many decades chosen the New York Fed, given the market infrastructure and expertise that has been established by that Reserve Bank. In particular, at each policy meeting, the FOMC gives the New York Fed a directive for carrying out its policy decisions. The Desk has chosen to conduct most of its operations with a set of firms, called “primary dealers,” that meet a number of requirements in terms of their market-making activities and their participation in the Fed’s operations, although the set of counterparties has been expanded in recent years for some types of operations (as described in the following).4 The asset side of the balance sheet also includes several other items. One potentially important item is repurchase agreements (RPs) against the same types of domestic securities that the Fed holds outright. In a repurchase agreement, the Fed purchases a security from the primary dealer while simultaneously agreeing to sell it back to them at a future date, typically the next business day.5 This type of operation in effect provides a short-term loan to the counterparty using the underlying asset as collateral, thereby increasing liquidity in the market by crediting the reserve account of the counterparty or the counterparty’s bank. While there were

3 The Federal Reserve reports its securities holdings on an amortized cost basis, rather than a fair value basis. However, the Fed provides fair value figures on a quarterly basis in its “Reserve Bank Combined Quarterly Financial Report.” This report also provides information on unrealized gains and losses on securities holdings. The figures shown in the table are the face value of the securities held; the unamortized premiums and discounts are shown separately and not broken out by the type of security. “Agency” refers to US government-sponsored agencies such as Fannie Mae and Freddie Mac. 4 For more information, see Federal Reserve Bank of New York (2022a). 5 Somewhat confusingly, the Fed refers to these operations from the perspective of the counterparty. When a private firm does an RP, it is typically receiving funds and sending out the security – that is, it is funding the security. When the Fed does an RP, it is instead sending out funds and taking in the security – that is, it is conducting what is an RP for its counterparty.

The Federal Reserve balance sheet  9

essentially no RPs on the asset side of the balance sheet in the aforementioned snapshot, this has been a very large asset in the Fed’s portfolio at times. A second category of asset corresponds to credit that the Federal Reserve has provided to banks and, on occasion, to non-bank firms or other entities. This credit includes discount window loans to banks, authorized under Section 10 of the Federal Reserve Act, and other types of lending to non-bank entities, authorized under Section 13. These will be described in greater detail in Section 1.7, and further details are provided in the chapter in the volume by Madigan and Nelson. A third category of assets held is foreign assets, as the Fed maintains roughly $20 billion of assets denominated in Japanese yen and euros. This chapter does not discuss the purpose and management of these foreign asset holdings in detail. These assets have generally been rolled over for many years, with no active market operations that were intended to have an impact on FX markets.6 A final category of assets on the Fed’s balance sheet is central bank liquidity swaps. These are arrangements with foreign central banks under which US dollars are provided against foreign currency collateral. These transactions are authorized under Section 14 of the Federal Reserve Act. As described further in the following, the purpose of the swaps is to allow foreign central banks to on-lend the US dollars to financial firms in their jurisdictions and thereby ease strains in dollar funding markets abroad. As noted earlier, the assets held by the Fed are almost entirely matched by liabilities on its balance sheet.7 The major categories of Fed liabilities are currency held by the public, bank reserves, reverse repurchase agreements, and the Treasury’s account at the Fed.8 It may be easiest to start with reserve balances, which are the balances that banks have in their accounts held with the Fed.9 These accounts, which are assets of the banks, represent pure liquidity for banks, in that they can use them to make immediate payments over the Fedwire system. When the Fed purchases Treasury or other securities from a bank, it credits its account with the payment for the securities, thereby creating liquidity (bank reserves) while reducing the bank’s holdings of Treasury securities. If the purchases are from a market participant other than a bank, that participant will have un-invested balances from its payment from the Fed, which will generally end up (at least temporarily) as a deposit in a bank, leaving 6 The exception was that the Federal Reserve participated in a coordinated Group of Seven (G-7) intervention to sell Japanese yen in March 2011, in the wake of the earthquake and tsunami that struck Japan. For a description of activities surrounding the foreign asset holdings of the Fed, see the quarterly reports published by the New York Fed (for example, Federal Reserve Bank of New York, 2022b). 7 The gap between the assets and liabilities on the Fed’s balance sheet corresponds to the capital held by the Fed. Unlike some foreign central banks, the Federal Reserve holds little capital. In March 2022, Fed capital was only about $40 billion – roughly a half percent of Fed assets. While Fed earnings have been large in recent years (Federal Reserve, 2022), those earnings are almost entirely remitted to the Treasury each year. Indeed, Congress has on two occasions over the past decade required the Fed to transfer additional capital to the Treasury to help fund federal spending and has put limits on the Fed’s ability to accumulate capital. 8 In addition to those major liabilities, there is a variety of relatively small liabilities, including accounts maintained for various government agencies and clearing houses. We leave those aside in the discussion here. 9 All depository institutions – commercial banks, thrift institutions, and credit unions – are eligible to have accounts at the Fed. We use the term “banks” generically to refer to all depository institutions.

10  Research handbook of financial markets

the banking system with more reserves (reserves end up being the asset that banks hold that corresponds to the increased liability in the form of the deposit). If that were the only form of Fed-generated liquidity, it would be essentially a closed system in which all liquidity created by the Fed’s asset purchases would have to reside in the banking system (leaving currency aside for the moment). However, there are other Fed liabilities that provide more flexibility for the liquidity created by Fed purchases to be held in other forms. One important one is reverse repurchase agreements (RRPs), which are simply RPs done by the Fed in the opposite direction. That is, in an RRP, the Fed would sell securities to the counterparty while agreeing to buy them back on a later date, thereby draining bank reserves from the system. These operations allow the liquidity created by the Fed’s asset purchases to flow out of the banking system and into other institutions such as government-only money market mutual funds, since those institutions can hold RRPs but cannot hold reserves. In recent quarters, as the Fed’s securities holdings have reached new highs, RRPs have increased substantially. Another notable liability for the Fed is Federal Reserve notes (currency). Households and other private agents hold some amounts of currency for making various routine transactions in the economy. The demand for currency therefore tends to grow as the size of the economy (nominal GDP) expands. In addition, the high degree of credibility enjoyed by the US dollar contributes to foreign demand for US currency. In fact, a majority of US currency outstanding is estimated to be held outside the borders of the United States (Judson, 2017). Because currency is obtained from banks, rising demand for currency would reduce bank reserves, at least in the first instance, if the Fed took no action. The Fed therefore needs to purchase assets over time to create the liquidity that ends up being held in the form of currency. The final category of liability worth highlighting is the Treasury’s account, known as the Treasury General Account (TGA). This account is held at the Fed and is used by the Treasury to receive and make payments – that is, it is the account used for the majority of Treasury transactions. Given the volume of funds flowing in and out of the Treasury, managing the balance in this account and ensuring that it has sufficient funds is a complex process for the Treasury, and the Fed provides it with operational support for performing those functions. Of particular relevance for this chapter, balances held in the TGA are funds that are not available to the private markets. As such, an increase in the TGA balance drains liquidity from the financial system. The TGA used to be managed to maintain a very small and stable balance by shifting excess funds into accounts at depository institutions. But the Treasury has chosen to have TGA balances become much larger and more variable in recent years.10

10 The Treasury has seen this as a prudent shift to allow it to continue to make payments even if an unexpected event were to disrupt its inflows for several days, or to be ready for sizable outflows when legislative actions are pending. The shift in the Fed’s operating framework, discussed later in this chapter, accommodates such a shift in the Treasury’s cash management practices.

The Federal Reserve balance sheet  11

1.3 A HISTORICAL PERSPECTIVE ON THE FED’S BALANCE SHEET Before turning to what the Fed’s balance sheet and operating regime have looked like in recent decades, it is useful to look at a longer history. This history demonstrates that the Fed’s approach has varied considerably in the past, and it sheds some light on how the Fed arrived at its current regime. Over time, the variation in the Fed’s approach has been driven by the legal authority provided by Congress, the framework for implementing monetary policy, and the economic and political environment of the time. One of the primary purposes of the Federal Reserve as outlined in the Federal Reserve Act of 1913 was to “furnish an elastic currency.” While it is not clear how exactly to interpret this purpose in the context of modern financial markets, it at least suggests that the Fed would vary the size or composition of its balance sheet based on private demand for its liabilities. In the early years, the Fed primarily met this responsibility by rediscounting “real bills” for banks. These bills were receivables from commercial transactions and other commercial paper, which the Fed would convert to liquid money in the form of Federal Reserve notes or reserves of banks. Indeed, for the first decade and a half after its founding in 1914, the Federal Reserve’s main assets were gold – as the United States was on the gold standard during this time – and real bills (Chen & Gibson, 2017).11 An important turning point in the legal authority afforded to the Federal Reserve occurred in the 1932 Glass-Steagall Act. At this time, Treasury securities became eligible to serve as explicit backing for the Federal Reserve’s liabilities, instead of the earlier focus on gold and real bills. As explained by Friedman and Schwartz (1963), Congress had grown frustrated with the Federal Reserve’s contractionary monetary policies during the Great Depression and leaned heavily on the Federal Reserve to undertake expansionary open market operations in 1932. This resulted in a substantial expansion of the Fed’s holdings of government securities, as shown in Figure 1.1. According to Meltzer (2004), while initially conceived of as a temporary Depression-era response, this ended up as a permanent change, relaxing one of the main constraints on the Federal Reserve’s ability to expand the size of its balance sheet. Over time, government securities became the dominant asset backing the Fed’s liabilities, while gold backing requirements were diminished in several steps starting in 1945 before ultimately being brought to zero in 1965 (Ramage, 1968). Another shift in the size and composition of the Fed’s balance sheet came after the United States entered World War II. The Treasury Department requested that the Federal Reserve undertake open market operations in order to cap market interest rates on Treasury securities beginning in April 1942. This regime required considerable purchases of Treasury securities at times and persisted after the end of World War II (Chaurushiya & Kuttner, 2003). With the onset of the Korean War in 1950 and rising inflation associated with the war effort, the Federal Reserve became worried that maintaining the interest rate ceiling was inconsistent with achieving its monetary policy objectives. After the Fed-Treasury Accord of March 1951, explicit support to maintain the yield cap was gradually phased out over the remainder of the year. Perhaps seeking to more clearly demonstrate that there was no longer any target 11 At the time, the real bills doctrine was a popular theory. It held that monetary expansion that matched the expansion in economic activity as reflected in circulation of such commercial bills would not be inflationary.

12  Research handbook of financial markets

Percent of Nominal GDP

40.0%

30.0%

20.0%

10.0%

0.0% 1914

1924

1934

1944

1954

1964

1974

1984

1994

2004

2014

Year Assets

Securities

Loans and Discounts for Banks

Swap Lines

Gold and Gold Certificates

Loans and Discounts for Others

Other

Sources:   Asset data are from the Center for Financial Stability’s The Federal Reserve System’s Weekly Balance Sheet Since 1914 for 1914–2002; 2002–2020 data come from the Federal Reserve’s H.4.1 Statistical Release, both averaged annually. Annual nominal GDP data come from Johnston and Williamson (2020), The Annual Real and Nominal GDP for the United States, 1790–Present.

Figure 1.1  The Federal Reserve’s assets as a percent of nominal GDP, 1914–2021 for longer-term yields, the Federal Reserve implemented a “bills only” policy from 1953 to 1961, under which it restricted open market operations to Treasury bills only. This policy was abandoned in 1961 when higher bill yields were seen as desirable to stem the outflow of gold from the United States even though domestic economic considerations warranted low longerterm rates, and the Federal Reserve began buying longer-term securities again (Friedman & Schwartz, 1963).12 After the 1960s, the share of bills held generally fluctuated between onethird and one-half of the portfolio until the onset of the GFC. A further important change in the Federal Reserve’s authority came with the Interest Adjustment Act of 1966, which allowed for purchases of Federal agency debt for the first time. At first, the federal Reserve was reluctant to insert itself into this market, for fear of becoming entangled in Federal housing policy at the expense of monetary policy considerations, and initially, only repurchase agreements, not outright purchases of agency debt, were allowed (Haltom & Sharp, 2014). During the expansionary monetary policy period of the 1970s, the Federal Reserve acquired a significant quantity of agency debt until this practice was stopped by Chairman Volcker in 1981. The Federal Reserve’s portfolio of agency debt declined to zero by the early 2000s as these securities matured. As noted, the framework for implementing monetary policy was also a determinant of the Federal Reserve’s balance sheet. Early in the Federal Reserve’s history, monetary policy was 12 This was the original “Operation Twist,” a term later also used for the Maturity Extension Program (MEP) announced in 2011 (Swanson, 2011).

The Federal Reserve balance sheet  13

primarily adjusted via the discount rate, the rate at which the Fed purchased bankers acceptances, and reserve requirements, with an eye toward arresting stresses that arose from time to time in the banking system.13 In fact, as noted by Meltzer (2004), in the early years the balance sheet was too small to effect the monetary tightening that might have been needed during WWI, as the Fed had insufficient securities to sell in its portfolio and was reluctant to adjust discount rates at that time. While short-term market rates were seen as a helpful indicator for assessing money market conditions, they were not a target in and of themselves until Treasury bill rates were pegged during WWII.14 Despite acknowledging the relevance of the federal funds market, according to Meulendyke’s history of the Federal Reserve’s operating framework (1998), the Federal Reserve during the 1950s and 1960s was reluctant to adopt explicit targets for the federal funds rate due to the earlier negative experience with targeting bill yields. Instead, open market operations were used to adjust the balance sheet such that “free” reserve balances (that is, excess reserves less borrowed reserves) moved toward targeted levels. By the 1970s, the federal funds rate was seen as a useful intermediate target, and ranges for the federal funds rate were prescribed during intermeeting periods, within which fluctuations were allowed in order to try to hit reserves targets. These ranges were gradually narrowed over the course of the 1970s, from as wide as 1.5 percentage points early in the decade to as narrow as 25 basis points later in the decade. In 1979 explicit targeting of the federal funds rate was thrown out, and monetary aggregates were made the focus of policy. This was seen as the most direct way to ensure that the inflation problem at the time could be contained. Over the course of the 1980s as inflation was brought under control, targets for the federal funds rate were again re-emphasized – in part due to instability in the relationship between the monetary aggregates and income and interest rates – and by 1988 the FOMC was giving primary weight to the federal funds rate as its policy instrument. However, it was not until 1994 that the directional objective for the rate (i.e., higher or lower) was clearly communicated in a policy statement after FOMC meetings. In 1995, numerical targets started to be announced, in basically the same form they are communicated to the public today. Turning to the liability side of the balance sheet, for much of the Fed’s history reserve balances not only represented a liquid medium for interbank payments, but also a way of meeting reserve requirements – the liquidity buffer that needed to be held against deposits both to ensure banks could manage unexpected withdrawals and to facilitate the implementation of monetary policy. However, the extent to which these requirements were binding, and so encouraged trading of reserves among banks, varied significantly over time. For example, Goodfriend and Whelpley (1986) note that during the liquidity trap in the 1930s excess reserves were very abundant (Figure 1.2), and hence trading activity in the federal funds market dried up. In the 1940s, reserves remained effectively abundant due to the monetary 13 The discount window was initially not a single rate, but separate rates set independently by the Boards of the Reserve Banks, though subject to review and determination by the Board of Governors. 14 Goodfriend and Whelpley (1986) note that rates on overnight borrowing – conceptually very similar to the modern federal funds market – were quoted in the financial press as early as 1928. These rates were for two checks exchanged simultaneously, one drawn on the reserve account of the lending bank and one drawn on the clearinghouse account of the borrowing bank, the latter of which settled a day later than the former, hence the two taken together were a self-liquidating overnight loan of reserve balances.

14  Research handbook of financial markets

Percent of Nominal GDP

40.0%

30.0%

20.0%

10.0%

0.0% 1914

1924

1934

1944

1954

1964

1974

1984

1994

2004

2014

Year Liabilities

Currency

Required Reserves

TGA

Total Reserves (per-1929)

Excess Reserves

RRP

Other

Sources:   Liability data are from the Center for Financial Stability’s The Federal Reserve System’s Weekly Balance Sheet Since 1914 for 1914–2002; 2002–2021 data come from the Federal Reserve’s H.4.1 Statistical Release, both averaged annually. Annual nominal GDP data come from Johnston and Williamson (2022), The Annual Real and Nominal GDP for the United States, 1790–Present.

Figure 1.2  The Federal Reserve’s liabilities as a percent of nominal GDP, 1914–2021 regime needed to maintain the ceiling on Treasury bill rates. However, starting in the 1950s, as monetary policy tightened, reserve requirements became more binding, reserve scarcity grew, and trading in the federal funds market rose – continuing to be robust for most of the second half of the century (Anbil & Carlson, 2019). After the GFC, conditions moved back toward those seen in the 1930s, as large-scale asset purchases again boosted excess reserves to very high levels and trading in the federal funds market ebbed. In recent years reserve requirements were reduced significantly and increasingly became an afterthought in monetary and regulatory policy, as new prudential liquidity requirements were introduced that could be met with non-reserve assets (such as Treasury bills). In March 2020 the Federal Reserve officially dropped all reserve requirements to zero.

1.4 THE FED’S BALANCE SHEET LEADING UP TO THE GLOBAL FINANCIAL CRISIS In the period leading up to the GFC, the Fed had settled into a fairly straightforward balance sheet approach and operating regime for implementing monetary policy. In this section, we consider both its outright holdings of assets and its use of temporary open market operations to achieve its federal funds rate target over this period. Nearly all of the growth in the outright asset holdings on the Fed’s balance sheet over this period was driven by the expanding demand for currency by the general public, both in the United States and abroad. As noted earlier, it is natural for the demand for currency to increase

The Federal Reserve balance sheet  15

over time as nominal GDP grows. To accommodate that growth, the Fed expanded its holdings of Treasury securities over time. It did so by conducting periodic operations to purchase Treasury securities in the open market from the primary dealers. These purchases, which were referred to as “bill passes” or “coupon passes” depending on the securities purchased, were generally small in size and were designed to be market neutral, given their primary purpose. On average over the period from 2002 to 2006, the Open Market Desk at the New York Fed carried out an average of 36 bill or coupon passes per year, with an average size of just over $1 billion. The maturity structure of the Treasury securities held by the Fed was not notably different from that of the outstanding stock of debt, although the SOMA portfolio was skewed somewhat in the direction of shorter maturities. By purchasing securities only as needed to match the growth in currency, and by not being aggressive in terms of the duration risk held in the SOMA portfolio, the Fed was attempting to leave as little imprint on the market as possible with its balance sheet. The more active component of operations by the Fed involved the daily management of reserves by the Desk, which took place primarily through repurchase agreements (RPs). As noted earlier, RP operations serve to increase liquidity in the financial system, but they do so only temporarily, in a manner that automatically unwinds. In contrast, the coupon and bill passes described earlier also create liquidity, but they do so in a more permanent manner. Given that distinction, the uses of these tools can be understood by the needs that they were trying to address. Increases in currency demand are long-lasting, and the Desk therefore would meet currency demand through permanent increases in SOMA asset holdings. Shortrun increases in the demand for liquidity by the banking sector, or temporary fluctuations in the TGA, were instead better met through RPs, which could be allowed to unwind quickly if needed. The decision of how much liquidity to inject on any day through RPs was governed by the target for the federal funds rate. Over the period since 1997, the directive to the Desk has included an explicit target for the federal funds rate – the interest rate on uncollateralized overnight transactions between banks which has come to serve as the primary policy instrument of the FOMC. Moreover, as described earlier, the FOMC had begun explicitly referring to a federal funds rate target in 1995, and in its internal deliberations, it had implicitly prescribed a target for the federal funds rate for a number of years before that. To hit the federal funds rate target, the Desk had to make a daily decision, with input from the staff of the Board of Governors, about how much liquidity to create on any given day given the anticipated demand for reserves on the day. Pushing more liquidity into the system through larger RPs would typically lower the federal funds rate, as banks would have more liquidity to meet their needs and would therefore have less demand to borrow from the federal funds market. Similarly, injecting fewer reserves into the system would put upward pressure on the federal funds rate. The process of determining daily open market operations involved a forecast of the daily demand for reserves, which was a complicated process involving a number of esoteric details about payment flows. That said, the operating framework was highly effective, with the federal funds rate trading very close to target on most days (see Figure 1.3). The overall manner in which the Open Market Desk at the New York Fed chose to operate over this period was to set the outright holdings in SOMA at a level that would, by itself, leave a moderate but persistent reserve deficiency. This reserve deficiency would then be made up by conducting RP operations to generate the amount of liquidity needed to hit the federal funds rate target. In fact, over the period from 2002 to 2006, the Desk typically maintained

16  Research handbook of financial markets

Source:   Federal Reserve Economic Data (FRED).

Figure 1.3  Federal funds rate and target federal funds rate, 2000–2007 between $20 billion and $40 billion of RPs outstanding. This allowed them to take the level of reserves up or down quickly as needed to hit the federal funds rate target. This system ran with relatively low aggregate reserve balances over this period. Since reserves earned no interest, banks had an incentive to economize on their reserve balances. As a consequence, the Fed could effect a nontrivial influence on the federal funds rate through relatively small variations (typically just several billion dollars) of reserves, giving it the control needed to hit the federal funds rate target. This operational system for controlling overnight rates is often called a “corridor” system since the supply of reserves provided determined the federal funds rate within a range – or corridor – above the rate paid on reserves (then zero) and below the rate charged on loans of reserves (the discount rate).15

1.5 THE FED’S BALANCE SHEET SINCE THE GLOBAL FINANCIAL CRISIS Once the GFC hit, the Fed’s balance sheet was transformed in important ways. The Fed ended up holding a much larger volume of assets in its portfolio, and it shifted its operating regime to a so-called “floor” system. The evolution of the Fed’s balance sheet began in late 2007 and over the first half of 2008, well before the intense pressures created by the Lehman Brothers bankruptcy in September 2008. The Fed over that period launched a series of credit and liquidity facilities to address pressures in funding markets, as described in the next section and in the chapter by Madigan and Nelson in this volume. Those facilities, along with measures related to the collapse of 15 We are using the term “corridor system” flexibly here. Some commentators argue that a corridor system requires an above-zero rate paid on reserves to provide a floor for the system. We are considering the zero rate paid on reserves as the bottom of the corridor. The key aspect of the corridor system that we are highlighting, which holds under either definition, is that it requires active use of open market operations to achieve the supply of reserves that intersects the demand for reserves at the target rate.

The Federal Reserve balance sheet  17

Bear Stearns in March 2008, resulted in an expansion in the Fed’s holdings of instruments beyond the Treasury securities that had been typically held in the SOMA. This liquidity, if left unchecked, would have put significant downward pressure on the federal funds rate in the open market, pushing it below the target established by the FOMC. In an effort to maintain the federal funds rate near target, the Desk decided to shed some of the Federal Reserve’s other asset holdings by running off maturing Treasury bills (that is, letting them mature without replacement). By doing so, it was trying to keep bank reserves in a range that allowed the target funds rate to be hit.16 This process of “sterilizing” the reserve injections coming from the credit facilities would soon be overwhelmed, though, by the massive expansion of the balance sheet that took place after the failure of Lehman Brothers. In the immediate aftermath of the Lehman failure, the use of the credit facilities expanded rapidly and new ones were launched. The volume of lending across all Fed facilities surged to nearly $1.9 trillion, far outstripping the Fed’s ability to shed other assets or otherwise drain the reserves created, as shown in Figure 1.4. Moreover, the Fed also embarked on a large asset purchase program – the first step into a policy approach that has become known as quantitative easing or QE.17 The QE programs began with the announcement of $600 billion of purchases of agency debt and agency MBS in late November 2008. But the program quickly morphed into a more extensive program involving the purchase of over $1.7 trillion in Treasury debt, agency debt, and agency MBS that would last through March 2010. This program was later followed by a $600 billion second round of QE (QE2) that began in 2010, a $667 maturity extension program (MEP) that began in 2011, and a third round of asset purchases (QE3) that began in 2012 and would eventually increase the Fed’s security holdings by another $1.6 trillion (Table 1.2). Based on all of those programs, the Fed ended up purchasing about $3.75 trillion of Treasury debt, agency debt, and agency MBS, on net, over the course of the GFC and the sluggish recovery that followed it. Since these assets were longer-term in nature (in contrast to the lending programs that were launched), these programs represented a fundamental transformation of the Federal Reserve’s balance sheet – one that would carry forward into the new operating framework for the Fed. Purchases on this scale were a significant operational challenge for the Desk. Treasury securities were purchased over an internal system called FedTrade that let dealers make competitive offers to the Fed to ensure that the transactions occurred at fair market prices. While this was the standard operational system used by the Desk, it had never been employed to conduct outright purchases with that intensity. In addition, the Fed had never previously bought mortgage-backed securities, and hence it had to rely on external managers to conduct such purchases under close supervision and scrutiny by the New York Fed, at least for a time until it developed its own capabilities to do so (including eventually expanding FedTrade to include that asset class).

16 In addition, the Fed and Treasury established the Supplementary Financing Program, under which the Treasury would issue Treasury bills beyond those needed to finance the government and hold the funds in its account at the Fed. The effect of such issues was to drain reserves from the banking system and help support the Fed’s policy implementation (Treasury Department, 2008). 17 The Fed initially made a distinction between quantitative easing, which had previously been implemented by the Bank of Japan and involved large increases in reserves intended to encourage bank lending, and large-scale asset purchases, under which the Fed purchased large volumes of longerterm securities with the aim of lowering longer-term interest rates. However, the two terms are now essentially used interchangeably in public discussions.

18  Research handbook of financial markets Assets

Treasury Securities

Discount Window Credit

Swap Lines

Agency Securities

Section 13(3) Loans

Other

40.0% 30.0%

Percent of GDP

20.0% 10.0% 0.0% 0.0% 10.0% 20.0% 30.0% 40.0%

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Date Liabilities

Currency

Reserve Balances

RRPs

TGA

Other

Sources:   Balance sheet data are quarterly averages of weekly data from the Federal Reserve’s H.4.1. Statistical Release. Quarterly nominal GDP data, seasonally adjusted, comes from the Bureau of Economic Analysis.

Figure 1.4  The Federal Reserve’s assets and liabilities as a percent of nominal GDP, 2005–2021 In all of these cases, the Desk had to set out a strategy determining which securities to purchase and the procedures for making those selections, within the broad policy parameters established by the FOMC. And it had to meet the operational challenges of implementing those purchases. The Desk had moved to conducting less than two outright operations a month of modest size, to an average of more than three operations a week in Treasury and agency debt with an average size of over $3 billion per operation, and nearly continuous purchases of MBS with an average of over $11 billion per week. The purpose of the asset purchases was to make financial conditions more accommodative. In the Fed’s view, the primary channel for achieving this was by removing risk from the market, which was expected to lead the market to reprice the premiums for that risk through a “portfolio balance channel.” Since duration was the primary risk contained in the assets that the Fed was purchasing, it was thought that asset purchases would lower longer-term interest rates by reducing the term premium, with favorable knock-on effects to a wider range of asset prices. As Ben Bernanke put it, the Fed’s purchases of Treasury and agency securities “likely both reduced the yields on those securities and also pushed investors into holding other assets with similar characteristics, such as credit risk and duration. For example, some investors who sold MBS to the Fed may have replaced them in their portfolios with longer-term, high-quality corporate bonds, depressing the yields on those assets as well” (Bernanke, 2010).

The Federal Reserve balance sheet  19

Table 1.2  Federal Reserve asset purchase programs Program

Dates of operations

Modal purchase pace

Announcement date(s)

Size

QE1

November 25, 2008; increased March 18, 2009

$500bn agency MBS and $100bn agency debt Nov 2008– initially; expanded to $300bn Treasuries, $175bn Mar 2010 agency debt, and $1.25tn agency MBS

~$100bn/mo

QE2

November 3, 2010

$600bn longer-term Treasuries

Nov 2010– Jun 2011

~$75bn/mo

Maturity Extension Program

September 21, 2011; extended June 20, 2012

$400bn longer-term Treasuries initially; expanded to $667bn

Sep 2011– Dec 2012

$45bn/mo

QE3

Agency MBS purchases announced in September 13, 2012; Treasury $790bn Treasuries and purchases announced $823bn agency MBS in December 2012 as continuation of Maturity Extension Program

Sep 2012– Oct 2014

$85bn/mo ($45bn Treasuries, $40bn agency MBS)

Mar 2020– Mar 2022

Very rapid pace earlier in the program; settled to $120bn/mo ($80bn Treasuries, $40bn agency MBS)

Covidera QE program

March 15, 2020

$3.2tn Treasuries, $1.3tn agency MBS

Sources:  Federal Reserve Bank of New York (http://www.newyorkfed.org/markets/programs-archive/largescale-asset-purchases). Federal Reserve Board of Governors, H.4.1. Statistical Release, various dates. Authors' calculations.

Even though the intended effects on financial conditions were associated with the asset side of the Fed’s balance sheet, this expansion in asset holdings required a similar expansion of the liability side of the Fed’s balance sheet, which created some meaningful operational challenges for the Fed. Most importantly, it meant that the supply of reserves over this period increased by more than $2.5 trillion, making it impossible for the Desk to control overnight interest rates in the same manner as it had prior to the crisis. This situation necessitated a shift in the operating procedure of the Fed, which in turn required a change in the Federal Reserve Act. The Federal Reserve Board had long argued that Congress should provide it with the authority to pay interest on reserve balances. This was seen as desirable for several reasons (see Kohn, 2004). One reason given was that this authority could help the Fed implement monetary policy, since the rate paid on reserves would tend to put a floor under market rates. In 2006, Congress passed the Financial Services Regulatory Relief Act of 2006, which authorized the Federal Reserve to pay interest on balances held by banks but, for budgetary reasons, only made the authority effective in 2011. However, the Emergency Economic Stabilization Act of 2008, passed after the failure of Lehman Brothers, moved up the effective date to October 1, allowing the Fed to begin paying interest on reserve

20  Research handbook of financial markets

balances (IORB) to banks at a rate set by the Board of Governors. This administered rate became a key component of the Fed’s operational framework, effectively moving the Fed, at least for a time, to a floor system for managing its target rate. The idea underlying the new system (described in more detail in the next section) was that market rates would not be able to fall too far below the IORB rate, since banks could always leave their excess funds at the Fed and earn that rate. While this system generally operated as intended, the floor on market rates proved soggy, with overnight interest rates trading notably below the IORB rate in the fall of 2008. The system relied on arbitrage performed by banks to pull market rates towards the IORB rate: if borrowing rates were below the IORB rate, banks could borrow funds in the market and then hold them at the Fed, earning the spread between them in a risk-free transaction. However, banks had various regulatory costs related to their balance sheet size that limited their willingness to do such transactions, and hence the wedge between market rates and the IORB rate was larger than expected. This issue became inconsequential once the FOMC, in December 2008, decided to take the target policy rate all the way to its effective lower bound by cutting the target for the federal funds rate to a range of 0 to 25 basis points (zero is a harder floor on the federal funds rate for several technical reasons). However, the issue still lingered in the background, and, when the time came to consider raising rates, the Fed needed to decide on what tools would be used to implement policy. A key innovation on this front was the decision to use RRPs in a more expansive manner. In particular, the FOMC ultimately decided to allow the Desk to implement RRPs as a fixed-rate, full-allotment overnight facility – that is, allowing participants to invest large amounts in the program at the offering rate. In effect, this extended the ability to pay interest on overnight balances at the Fed to a wider set of counterparties, only in the form of RRPs rather than bank reserves. This facility proved highly effective at placing a floor on overnight market interest rates – particularly for overnight repo rates – since this facility reached many of the institutions with excess funds that would have otherwise invested them in RRPs or federal funds transactions. Once the economy had recovered sufficiently from the GFC, and interest rates had been raised well above the effective lower bound, the FOMC decided to begin reducing the size of the Fed’s balance sheet. In a set of principles published in June 2017, the FOMC said that it would “gradually reduce the Federal Reserve’s securities holdings by decreasing its reinvestment of the principal payments it receives from securities” (FOMC, 2017). To achieve this outcome, the Desk began to allow maturing securities to run off its books without reinvestment, up to caps on the monthly amount of run-off. The caps started out at a low level but were gradually raised until the allowed runoff reached $50 billion a month. Over the period from 2017 to 2019, the amount of securities held by the Fed declined steadily. However, the cumulative decline over this period ended up being only about $650 billion, as currency demand had continued to grow, and as the new regulatory environment for banks encouraged them to hold much higher levels of reserves than had been the case prior to the GFC. Indeed, the strength of the demand for reserves proved surprising to the Federal Reserve, as the decline in reserves to around $1.5 trillion in September 2019 led to unexpected pressure in money markets, particularly the market for RPs. In response, the Federal Reserve added reserves, initially through repo operations, and subsequently through outright purchases of Treasury bills (FOMC, 2019).

The Federal Reserve balance sheet  21

More recently, in response to the Covid shock to the economy, the Fed again turned to asset purchases and emergency lending programs, which have expanded its asset holdings. In March 2020, when it became clear that the pandemic was likely to have very adverse effects on the US and global economies, many investors tried to reduce their positions to limit their risk and build liquidity. However, broker-dealers were unable to handle the flow of transactions, and prices in many fixed-income markets declined notably amid disorderly market conditions. To improve functioning in the Treasury and agency MBS markets, the Fed announced on March 15 that it would purchase $500 billion in Treasury securities and $200 billion in agency mortgage-backed securities over “the coming months.” Unlike the Treasury purchases undertaken after the GFC, these purchases were made across the yield curve, without a focus on longer-term securities. Conditions remained badly strained, and on March 23, 2020, the Fed stated that it would purchase the securities “in the amounts needed to support smooth market functioning and effective transmission of monetary policy to broader financial conditions.” Overall, the Fed purchased a total of about $2.1 trillion of securities over the period through May 2020. As market conditions improved, it reduced the pace of purchases. In June 2020, the Fed said that it would continue to purchase $80 billion of Treasury securities and $40 billion of agency MBS per month “to sustain smooth market functioning.” In December 2020, it indicated that it would continue to make purchases at that pace, but effectively transitioned those purchases into an ongoing QE program aimed at supporting the economic recovery. Those purchases continued until October 2021, and then were gradually tapered to zero by March 2022, with total purchases reaching about $4.5 trillion.18 The result of these purchases is that the Fed’s balance sheet became massive, as indicated in Table 1.1. Securities holdings in the SOMA portfolio reached roughly $8.5 trillion by early 2022, leaving nearly $4 trillion of reserves in the banking system and nearly $2 trillion in RRPs outstanding.

1.6 THE POST-CRISIS POLICY IMPLEMENTATION FRAMEWORK As noted earlier, the expansion of the Fed’s balance sheet became intertwined with its decision to move to a new operating framework. As the economy continued to recover over the course of 2013 and 2014, the Fed began planning for how it would raise the federal funds rate when doing so became appropriate. Given the size of the balance sheet, the Fed would have to use a floor system at least initially, and it developed various tools that could be used to control rates. The first was the IORB rate, which had been used since the fall of 2008 to help put a floor under rates offered by banks. The second was the fixed-rate overnight reverse repurchase (ON RRP) program noted earlier, which would allow a range of RP investors to place funds at the Fed.19 In addition to these tools, the Fed tested auctions of term deposits to banks and of

18 See the post-FOMC statements for the March 15, June 9–10, and December 15–16, 2020 meetings. 19 The Desk began to build a more effective approach for conducting reverse repurchase agreements in large size as early as 2009, which involved a number of operational changes to reach a wider set of counterparties. It tested an ON RRP program with a broad set of counterparties – including primary dealers, government agencies, and money market mutual funds – in 2013 and 2014.

22  Research handbook of financial markets 2.50 2.25 2.00

Percent

1.75 1.50 1.25 1.00 0.75 0.50 0.25 0.00 2015-01-01

2016-01-01

2017-01-01

Overnight Bank Funding Rate Secured Overnight Financing Rate

2018-01-01

2019-01-01

Effective Federal Funds Rate Target Range

Note:   Excludes December 31, 2018, as money market rates are volatile at year-end. The OBFR is often identical to the effective federal funds rate. Source:   Federal Reserve Economic Data (FRED).

Figure 1.5  Money market rates and federal funds target range, 2015–2018 term RRPs to the ON RRP counterparties as ways to drain reserves and potentially help put upward pressure on money market rates. After considerable discussion in the spring of 2014, the FOMC announced in June that it would use the IORB rate as its primary tool when the time came to raise rates and that it would use the ON RRP program, and potentially other supplementary tools, to help ensure the effective implementation of monetary policy. The two administered rates would be set at levels intended to keep the federal funds rate in its target range. Initially, the FOMC imposed a cap on the size of the ON RRP program out of concern that a very large program could have adverse effects on the evolution of money markets. However, at the time of liftoff in late 2015, the FOMC removed the cap on the size of the ON RRP program to ensure that limits on the size of the program did not adversely affect policy implementation. In December 2015 the FOMC announced a 25 basis-point increase in the Federal funds rate; the new tools proved effective, and money market rates rose in line with the new target range despite the very large Fed balance sheet (Figure 1.5).20 Moreover, the ON RRP program proved to be of limited size at the time, and the cap on the size of the program was never reinstated.21

20 See Anderson et al. (2016) for a discussion. 21 The Fed also imposed a cap on the size of any individual counterparty’s ON RRP bids in order to limit the risk of strategic behavior by the participants. Those caps have been increased as the size of the program has risen in recent quarters.

The Federal Reserve balance sheet  23

Source:   Keister (2012).

Figure 1.6  Floor and corridor systems for policy implementation While the floor system worked well at the time of liftoff, the Fed still needed to decide on the policy implementation framework that it would use over the longer run, once there was time to reduce the size of the balance sheet. Broadly, there were two options. First, the Fed could aim to return to a corridor system similar to that used prior to the GFC. In that case, the IORB rate (and perhaps the ON RRP rate) would provide the floor of the corridor, the discount rate would provide a rough ceiling for the corridor, and the actual federal funds rate would be in the middle of the corridor, with its level determined by the supply of and demand for reserves (Figure 1.6, left-hand side).22 In this option, the Desk would conduct regular – likely daily – open market operations to ensure that the federal funds rate traded near target. The second option was to continue to use a floor system, with a very large supply of reserves ensuring that the federal funds rate trades near the floor set by the IORB rate and the ON RRP rate (Figure 1.6, right-hand side). With this option, the FOMC could make decisions about its balance sheet independent of its decision on where to place the target federal funds rate. In March 2019, the FOMC decided that its long-run policy implementation framework would be a floor system, using the administered rates on reserves and ON RRPs to keep shortterm market rates near the desired target.23 That system had become familiar over the period 22 As noted in the following and in the chapter by Madigan and Nelson, the role of the discount rate as the ceiling of the corridor is hampered by the stigma associated with banks’ use of the discount window. 23 Formally, the IORB rate is set by the Board of Governors, while the rate on ON RRPs is set by the FOMC. In practice, the two rates are set at the levels required to achieve the FOMC’s desired target range for the federal funds rate. The set of decisions required to implement the monetary policy decisions of the FOMC are reported in the Implementation Note published with the FOMC’s post-meeting statement.

24  Research handbook of financial markets

following the GFC and appeared to provide good control over the federal funds rate and other short-term market rates. A return to a corridor system was seen as potentially making policy implementation more difficult because of the need to adjust the supply of reserves on a daily basis in response to changes in demand that would be difficult to forecast, particularly given the changes in markets and bank regulations since the GFC. A floor system would also give the Committee control over short-term interest rates even if, as in 2008–2009, the supply of reserves was boosted substantially by emergency liquidity provision or asset purchases. Another advantage of the floor system was that the high level of reserves could support financial stability and limit operational risks by ensuring a high level of liquidity in the system at all times, potentially limiting the need to provide additional liquidity on an emergency basis. In addition to deciding on the implementation framework, the Committee also needed to agree on what interest rate or rates it would target. As a result of the high level of reserves and the ability to earn interest on them, the federal funds market became relatively small and somewhat idiosyncratic over the period after the GFC. The main lenders were the Federal Home Loan Banks, which had accounts at the Fed but were not eligible to receive interest, leading them to lend their overnight funds to banks in order to get a return on them. An alternative interest rate to target was the Overnight Bank Funding Rate (OBFR), which covered a broader set of lenders and so might provide a more robust measure of banks’ unsecured funding costs, particularly in times of market stress. Alternatively, a secured rate, such as the Secured Overnight Funding Rate (SOFR) could be used, as funding for banks and other financial firms had shifted toward secured markets over recent years, perhaps making the repo rate more important than unsecured funding rates for influencing broader financial conditions. However, the federal funds rate had been an effective tool for communication about the Fed’s policy intentions for a long time, and no change was made to the policy rate.24 In 2021, the Fed decided to add two additional features to its operating framework – a standing RP facility for primary dealers and banks (the SRF), and a standing RP facility for foreign official institutions (called the FIMA repo facility). Use of these facilities would have the effect of adding liquidity (held as bank reserves or overnight RRPs) automatically whenever overnight interest rates reach the rates set on them. Nevertheless, it is still appropriate to think of the Fed as operating in a floor system, one in which it maintains considerable amounts of liquidity in the system through its outright asset holdings and relies primarily on the IORB rate and the ON RRP facility to influence overnight market interest rates. The standing repo facilities will act as a guardrail to automatically respond when unexpected developments or episodes of market stress put upward pressure on overnight interest rates.

1.7 LENDING AND PRIVATE ASSET PROGRAMS In addition to the large-scale asset purchases undertaken to provide additional monetary accommodation at the lower bound, the Federal Reserve balance sheet has been deployed to address financial crises more directly by providing large volumes of liquidity and making purchases to support private asset markets. The Fed’s emergency lending programs undertaken in response to the GFC and the pandemic both left significant imprints on its balance sheet 24 See the FOMC Minutes for November and December 2018 and January and March 2019 for summaries of the Committee’s discussion of the policy implementation framework.

The Federal Reserve balance sheet  25

for a time. At their peak at the end of 2008, the Fed’s lending programs totaled nearly $1.9 trillion, divided roughly equally between credit provided directly to banks, credit provided to nonbanks, and swap lines with foreign central banks (to fund credit provided to banks in other jurisdictions). These emergency programs accounted for more than 80 percent of all Fed assets at the time (English & Mosser, 2020, table 2.1). Emergency lending programs surged again following the outbreak of the pandemic in 2020, but they remained more modest in size, peaking at around $700 billion, with the bulk accounted for by the swap lines with foreign central banks. Federal Reserve lending and other related activities are authorized under several sections of the Federal Reserve Act. First, the Federal Reserve can provide credit to banks under its traditional discount window authority in Section 10 of the Federal Reserve Act.25 In ordinary times, such credit is available to banks on a short-term basis – generally overnight. During the GFC and also in response to the pandemic, the Federal Reserve eased the terms on discount window credit to encourage its use. In addition, during the GFC the Fed provided large amounts of discount window credit to banks in regular auctions under the Term Auction Facility (English & Mosser, 2020). Discount window credit (including discount window credit extended through the Term Auction Facility) peaked at more than $500 billion during the GFC, but only reached $50 billion in the spring of 2020. Second, the Fed has broad lending authority in “unusual and exigent circumstances” under Section 13(3) of the Federal Reserve Act. During the GFC the Fed used this authority to provide liquidity to key institutions, including primary dealers and troubled systemically important firms, as well as to support important financial markets, including the commercial paper market and markets for asset-backed securities.26 Section 13(3) lending peaked at more than $700 billion in late 2008.27 In response to the pandemic, the Fed restarted virtually all of its crisis-era 13(3) programs and added newly developed programs to support lending to businesses and state and local governments. Nonetheless, total lending under the 13(3) programs during the pandemic peaked at less than $300 billion, well below the totals reached in 2008. Third, Section 14 of the Federal Reserve Act gives the Federal Reserve the authority to provide liquidity to broker-dealers and others through open market operations. During the GFC, the Fed provided credit to broker-dealers under this authority through term repo operations against Treasury, agency debt, and agency MBS collateral (the so-called Single-Tranche Repo Program).28 However, that program was superseded by the Term Securities Lending Program (TSLF) and so never exceeded $80 billion (English & Mosser, 2020). Following the start of the pandemic in the spring of 2020, the Fed undertook large repo operations to provide liquidity to primary dealers. These operations peaked at more than $400 billion before falling back. Finally, the Federal Reserve can use its open market authority under Section 14 of the Federal Reserve Act to engage in currency swaps with foreign central banks. Under these swaps, the Fed temporarily exchanges dollars for foreign currency; this can be very important 25 See the detailed discussion of the discount window in the chapter by Madigan and Nelson. 26 For a more detailed discussion of these programs, see Logan, Nelson, and Parkinson (2020) and the Madigan and Nelson chapter in this volume. 27 For more details, see English and Mosser 2020, especially table 2.1 and notes. 28 Ordinarily, in the Fed’s RP operations, there are separate rates established for Treasury, agency debt, and agency MBS collateral. In a single-tranche operation, all types of collateral are accepted at the same rate. Such operations can ease pressures on dealer firms by allowing them to obtain financing without undue concern about the collateral they will have to provide.

26  Research handbook of financial markets

in times of market stress because the foreign central banks can use the dollars provided to make dollar loans to financial firms in their jurisdictions. This mechanism ends up being crucial during periods of financial stress, given the global nature of dollar-denominated activity and the scope for dollar funding needs to arise in a number of different parts of the world. During the GFC, draws on swap lines by foreign central banks peaked at close to $600 billion; during the pandemic, such draws were also large, touching about $450 billion at the end of May 2020. The Fed’s emergency lending is intended to provide liquidity to support firms and markets during a crisis. Because the maturity of the credit is generally short and the interest rates are often elevated compared to market rates in normal times when conditions in funding and other markets recover, the lending programs tend to wind down endogenously as the loans are repaid. Thus, the sharp expansions in the size of the Fed balance sheet caused by emergency lending during crises reverse relatively quickly, in contrast to the effects of large-scale asset purchases, which are long-lived. For example, by the end of 2010, the GFC lending programs had declined by more than 90 percent and accounted for only about 6 percent of total Fed assets (Federal Reserve, 2010). Similarly, as the financial market effects of the pandemic waned, the Fed’s loan programs decreased in size fairly rapidly. Currently, the programs total less than $50 billion – about half a percent of Fed assets (see Table 1.1). It is important to note that the effect of an emergency program on total Fed assets may not be a useful measure of its effectiveness. For example, the Term Securities Lending Facility implemented in 2008 provided liquidity to primary dealers by swapping lower-quality collateral for Treasury securities, which the recipients could then use to raise funds in the repo market. As a result, the facility provided liquidity without an increase in Fed assets.29 Moreover, a commitment by the Federal Reserve to stand ready to purchase a class of assets may well be sufficient to improve market functioning without a significant increase in the size of the Fed’s balance sheet. The pandemic programs aimed at the corporate bond market and the municipal securities market are good examples. Following the outbreak of Covid in the spring of 2020, the corporate and municipal bond markets were badly disrupted, with risk spreads widening sharply and trading conditions impaired.30 The Federal Reserve established the Primary Market Corporate Credit Facility (PMCCF) and the Secondary Market Corporate Credit Facility (SMCCF) in late March to provide support to the corporate bond market. The PMCCF was available to purchase bonds from investment-grade and selected near-investment-grade firms unable to issue in the primary market, and the SMCCF purchased a range of similarly rated corporate bonds in the secondary market. However, the announcement of the programs triggered a rapid improvement in market sentiment, even though they were not yet operational. As a consequence, the PMCCF was never utilized, and total purchases by 29 The intention of this structure was to provide liquidity to the primary dealers without increasing the supply of reserves and so affecting the implementation of the Fed’s interest rate policy under the policy implementation framework of the time. 30 In addition, during the GFC the Federal Reserve, the Treasury, and the FDIC took part in asset guarantee programs to support Citigroup and Bank of America. Under the terms of these agreements, the Treasury provided capital to the institutions, the FDIC agreed to cover losses on a pool of illiquid assets above a specified threshold, and the Fed committed to lend a fixed amount against the pool if losses exceeded a higher threshold. The programs appeared to shore up confidence in the two firms, and the Federal Reserve was never called upon to lend under them.

The Federal Reserve balance sheet  27

the SMCCF totaled only $14 billion – a small fraction of the capacity of the program. The introduction of the Municipal Liquidity Facility (MLF) in April 2020 similarly contributed to substantial improvement in the functioning of municipal securities markets, while total Fed purchases of municipal securities totaled only about $6 billion.

1.8 THE OUTLOOK FOR THE FED’S BALANCE SHEET As a result of the very large size of its balance sheet and its critical role in a range of markets, the Federal Reserve has a tremendous influence on US financial developments. However, unlike a private investor aiming to maximize profits, the size and composition of the Fed’s balance sheet reflect the policy objectives of the Federal Reserve and the legal constraints under which it operates. In recent years, with two deep recessions driving the federal funds rate to its effective lower bound, the Fed has fostered its statutory objectives of maximum employment and stable prices by conducting large-scale purchases of Treasury securities and agency MBS. The result has been an extraordinary increase in Fed assets. On the liability side of the balance sheet, reserve balances have increased massively, and the Fed introduced an overnight reverse repo facility that has also risen to a very high level. In addition to the effects of these purchases, emergency liquidity provision also boosted the size of the Fed’s balance sheet, though those effects waned relatively quickly as the crises ebbed. The operating regime of the Federal Reserve is likely to keep the balance sheet sizable, in order to provide a sufficient amount of liquidity to keep the floor system operating in an efficient manner. However, even in that case, the balance sheet certainly does not need to be as large as it was in early 2022. A key question going forward, therefore, is how the Fed should “normalize” the balance sheet as it withdraws policy accommodation. The FOMC released a set of principles for reducing the size of the Fed’s balance sheet at its meeting in January 2022 (FOMC, 2022). Those principles largely followed the approach for policy normalization that was adopted after the Global Financial Crisis. In particular, the strategy involves first raising rates from the effective lower bound, and then reducing the Fed’s asset holdings primarily by not reinvesting some portion of the principal payments received on the securities in the Fed’s portfolio. This approach would result in a gradual normalization of the size of the Fed’s balance sheet over a period of several years. The principles also noted that, in the longer run, the FOMC intends to hold primarily Treasury securities in the SOMA portfolio. Achieving that outcome would require either reinvesting principal payments received on MBS in Treasury securities once the long-run size of the balance sheet has been reached, or outright sales of MBS at some stage. Given the size of the balance sheet, this policy strategy involves some risk of the Fed not completing the normalization process before the next recession. If that recession also pushes the federal funds rate to its effective lower bound, then the Fed could end up restarting asset purchases to provide additional accommodation. The result could be a ratchet effect, in which the Fed’s balance sheet gets significantly larger in each recession and never returns to its previous size (even as a share of GDP) before the next recession. In addition, the everincreasing amount of interest rate risk generated by larger holdings of longer-term securities funded by higher levels of interest-bearing reserves would presumably result in significant losses for the Fed at some stage. While not an economic problem, since the Federal Reserve

28  Research handbook of financial markets

could still implement appropriate monetary policy, such losses could be a political problem for the Fed.31 In making longer-run projections for the Fed’s balance sheet, it is also important to consider whether it will be influenced by major financial innovations or other structural changes to financial markets. One potentially important consideration in that regard is whether the Fed will choose to introduce some form of central bank digital currency (CBDC).32 At this time, Fed policymakers are still considering the possible benefits, costs, and risks of a Fed CBDC, and hence the outcome is uncertain.33 On the one hand, Brainard (2021) provides a generally positive assessment, pointing to several possible benefits of a CBDC, including increased payment system efficiency and stability, greater competition for incumbent financial services providers, gains in financial inclusion, lower cross-border transactions costs, and preservation of public access to central bank money in a world where payments are increasingly electronic. On the other hand, Waller (2021) argues that private sector innovation, such as stablecoins, could be a more effective source of efficiency gains and increased competition for current intermediaries; the introduction of low-cost bank accounts would likely be a simpler and more effective spur to financial inclusion; ongoing efforts to improve cross border payment efficiency are likely to prove effective; and central bank provision of retail payments is likely to be less efficient over time than competitive private sector provision. If in the end the Fed does decide to launch a CBDC and the demand for it proves substantial, the CBDC could greatly increase the size of the Fed’s “normal” balance sheet, thereby trimming the cumulative reduction needed to reach that level (or eliminating it altogether). In effect, this shift would affect the Fed’s balance sheet similarly to a large increase in the demand for non-digital currency, and the Fed would have to hold assets to accommodate those balances. Another key issue that could affect the Fed’s balance sheet over the longer term is whether it could be used to address new goals or objectives, such as limiting climate change. Central banks could, for example, adjust their collateral rules at the discount window to favor securities and loans related to green activities, or even purchase bonds issued to finance investments 31 See Hall and Reis (2015) for a theoretical discussion of the limited economic effects of central bank losses. Given the Fed’s accounting – as noted earlier, the Fed doesn’t mark its securities to market – losses would likely only occur if the interest payments on reserves rose above the interest earned on Fed assets. Ordinarily, the Fed transfers its net profits to the Treasury. If there were losses, those transfers would cease, and under the Fed’s accounting rules it would accumulate a “deferred asset” equal to the cumulative size of its losses, keeping Fed capital positive. Once earnings became positive again, the Fed would retain earnings over time sufficient to offset the earlier losses and eliminate the deferred asset before resuming transfers to the Treasury. Presumably Congress and the Treasury would be unhappy about the effect of the loss of Fed transfers on the budget, but unless losses exceeded all future Fed profits, the Fed could continue to operate without any technical problem. See Carpenter et al. (2015) for a more detailed description of how this accounting would work and the Financial Accounting Manual for Federal Reserve Banks (2021) for the full story. See various editions of the Open Market Operations annual report from the New York Fed for projections of Fed income and for some calibration of the changes in income under alternative scenarios for interest rates. 32 For a recent discussion of the issues, see BIS (2021). Private sector development of stablecoins could also have important effects on the Fed’s balance sheet, particularly if they were offered through banks that would in turn hold reserves. 33 Introduction of a CBDC would raise a number of difficult questions regarding the appropriate technology to use, the implications for users’ privacy, and the possible disintermediation of the banking system. See Federal Reserve (2022) and the references therein for a discussion.

The Federal Reserve balance sheet  29

in such activities.34 However, these suggestions raise two significant issues. First, in the case of the Federal Reserve, such steps would go beyond the Fed’s current monetary policy mandate.35 The Federal Reserve Act calls for the Fed to conduct monetary policy to promote “maximum employment, stable prices, and moderate long-term interest rates.” Thus, the Fed’s mandate does not include addressing climate change or other possible objectives.36 Second, it is not clear that the Fed’s tools would be all that helpful in achieving these objectives. Except in unusual and exigent circumstances, the Fed cannot purchase corporate securities of any sort, so it would need to be given authority to purchase securities or provide loans financing green investments, perhaps at concessional rates.37 However, such programs would be essentially fiscal policy. Indeed, fiscal policy could be a far more powerful tool, allowing the government to provide large subsidies for particular types of investment or make such investments itself.38 Overall, the discussion in this chapter of the Fed’s balance sheet decisions to date and the potential future course of the balance sheet highlights three broad points. First, the Fed’s balance sheet is tremendously important in terms of its influence on overall financial conditions and the functioning of financial markets. Second, the evolution of the Fed’s balance sheet is driven by the policy objectives of the Fed, the economic and financial shocks to which the Fed must respond, and the legal constraints on Fed activities. Finally, the Fed should manage its balance sheet in a flexible manner so that it can be used to address the key issues affecting financial markets at any particular time.

REFERENCES Anbil, S., & Carlson, M. (2019). The re-emergence of the Federal Reserve funds market in the 1950s. FEDS Notes 2019-03-22. Board of Governors of the Federal Reserve System, March. Anderson, A. G., Ihrig, J. E., Meade, E. E., & Weinbach, G. C. (2016). What happened in money markets after the Fed’s December rate increase? FEDS Notes 2016-02-22. Board of Governors of the Federal Reserve System, February. Baer, G. (2021). Central Bank digital currencies: Costs, benefits and major implications for the U.S. Economic System. Bank Policy Institute, Staff Working Paper, April 7. Bernanke, B. S. (2010). The economic outlook and monetary policy. Speech at the Federal Reserve Bank of Kansas City Economic Symposium, Jackson Hole, WY, August 27.

34 See Network for Greening the Financial System (2021) for a discussion of targeted credit operations and collateral rules. See Varoufakis (2021) for using the Fed to finance green investment. 35 Of course, it is appropriate for the Fed to use its supervisory activities to ensure that banks are measuring and managing climate risks appropriately. See Brainard (2020). 36 Other central banks may not be similarly constrained. For example, the ECB has a primary mandate of price stability, but its secondary mandate is to “support the general economic policies in the [European] Union with a view to contributing to the achievement of the objectives of the Union” (Ioannidis et al., 2021). That broad mandate could allow the ECB to use its tools to support EU policies to address climate change. 37 At this time the United States doesn’t issue green sovereign securities such as the UK’s green gilts, which presumably would be clearly within the Fed’s mandate to purchase in normal times, if they existed. 38 Similar arguments may apply to other new objectives that have been proposed for the Fed. For example, some have suggested that the Fed should use its tools to foster a more equal distribution of income. But beyond the objective of maximum employment, it is not clear that the Fed’s tools are likely to be very effective. The fiscal authorities’ ability to tax and transfer seems a much more potent tool.

30  Research handbook of financial markets

Bernanke, B. S. (2012). Monetary policy since the onset of the crisis. Speech at the Federal Reserve Bank of Kansas City Economic Symposium, Jackson Hole, WY, August 31. BIS. (2021). CBDCs: An opportunity for the monetary system. Chapter III in the BIS Annual Economic Report. Brainard, L. (2020). Strengthening the financial system to meet the challenge of climate change. Speech at The Financial System & Climate Change: A Regulatory Imperative, hosted by the Center for American Progress, December 18. Brainard, L. (2021). Private money and Central Bank money as payments go digital: An update on CBDCs. Speech at the Consensus by CoinDesk 2021 Conference, May 24. Carpenter, S., Ihrig, J., Klee, E., Quinn, D., & Boote, A. (2015). The Federal Reserve’s balance sheet and earnings: A primer and projections. International Journal of Central Banking, 11, 237–283. Chaurushiya, R., & Kuttner, K. (2003). Targeting the yield curve: The experience of the Federal Reserve, 1942–51. Memorandum distributed to the Federal Open Market Committee. Chen, J., & Gibson, A. (2017). Insights from the Federal Reserve’s weekly balance sheet, 1914–1941. Johns Hopkins Institute for Applied Economics, Global Health, and Study of Business Enterprise, Studies in Applied Economics No. 73. English, W. B., & Mosser, P. C. (2020). The use and effectiveness of conventional liquidity tools early in the financial crisis. Chapter 2 in Bernanke, B. S, T. F. Geithner, & H. M. Paulson (Eds.), First responders. Yale University Press. English, W. B., & Liang, N. (2020). Designing the main street lending program: Challenges and options. Journal of Financial Crises, 2, 1–40. Federal Open Market Committee. (2017). FOMC issues addendum to the policy normalization principles and plans. Press Release, June 14. Federal Open Market Committee. (2019). Minutes of the Federal Open Market Committee, October 29–30, 2019. (Portion on the minutes of the October 4 video conference.) Federal Open Market Committee. (2022). Minutes of the Federal Open Market Committee, January 25–26, 2022. Federal Reserve. (2010). H.4.1 statistical release. December 30, 2010. Federal Reserve. (2021). Financial accounting manual for Federal Reserve Banks. Federal Reserve. Federal Reserve. (2022). Federal Reserve Board announces Federal Reserve Bank income and expense data and transfers to the Treasury for 2021. Press Release, January 14. Federal Reserve Bank of New York. (2022a). Primary dealers. https://www​.newyorkfed​.org​/markets​/ primarydealers, downloaded February 26, 2022. Federal Reserve Bank of New York. (2022b). Treasury and Federal Reserve foreign exchange operations, October to December 2021. February 10. Friedman, M., & Schwartz, A. J. (1963). A monetary history of the United States, 1867–1960. Princeton University Press. Goodfriend, M., & Whelpley, W. (1986). Federal funds: Instrument of Federal Reserve policy. Economic Review, 72(5), September/October 1986. Gurkaynak, R. S., Sack, B., and Swanson, E. (2005). The sensitivity of long-term interest rates to economic news: Evidence and implications for macroeconomic models. American Economic Review, 95, 425–436. Hall, R. E., & Reis, R. (2015). Maintaining central-bank financial stability under new-style central banking. Working Paper No. 15109. The Hoover Institution, Stanford University. Haltom, R., & Sharp, R. (2014). The first time the Fed bought GSE debt. Federal Reserve Bank of Richmond Economic Brief. No 14-04. Ioannidis, M., Murphy, S. J. H., & Zilioli, C. (2021). The mandate of the ECB: Legal considerations in the ECB’s monetary policy strategy review. ECB Occasional Papers No. 276, September. Judson, R. (2017). The death of cash? Not so fast: Demand for U.S. currency at home and abroad, 1990–2016. In International Cash Conference 2017 – War on Cash: Is There a Future for Cash? Deutsche Bundesbank. Keister, T. (2012). Corridors and floors in monetary policy. Liberty Street Economics, April 4. Kohn, D. L. (2004). Regulatory reform proposals: Testimony before the Committee on Banking, Housing, and Urban Affairs, U.S. Senate, June 22, 2004. Board of Governors of the Federal Reserve System (U.S.).

The Federal Reserve balance sheet  31

Logan, L., Nelson, W., & Parkinson, P. (2020). The Fed’s novel lender of last resort programs. Chapter 3 in Bernanke, B. S., T. F. Geithner, & H. M. Paulson (Eds.), First responders. Yale University Press. Madigan, B., & Nelson, W. (2002). Proposed revisions to the Federal Reserve’s discount window lending programs. Federal Reserve Bulletin, 88, 313–319. Madigan, B., & Nelson, W. (2023). Central bank lending. Chapter 4 in Gürkaynak, R. S. and J. H. Wright, Research Handbook of Financial Markets. Edward Elgar Publishing. Markets Group of the Federal Reserve Bank of New York (various years). Open market operations. A report prepared for the Federal Open Market Committee. Meltzer, A. (2004). A history of the Federal Reserve, Vol. 1: 1913–1951. University of Chicago Press. Meulendyke, A.-M. (1998). U.S. monetary policy and financial markets. Federal Reserve Bank of New York. Network for Greening the Financial System. (2021). Adapting central bank operations to a hotter world: Reviewing some options. March 2021. Ramage, J. C. (1968). The gold cover. Federal Reserve Bank of Richmond Economic Quarterly, July 1968. Swanson, E. T. (2011). Let’s twist again: A high-frequency event-study analysis of operation twist and its implications for QE2. Brookings Papers on Economic Activity, Spring. Treasury Department. (2008). Treasury announces supplementary financing program. Press Release, September 17. Varoufakis, Y. (2021). A progressive monetary policy is the only alternative, blog post. Project Syndicate, October 28. Waller, C. J. (2021). CBDC: A solution in search of a problem? Speech at the American Enterprise Institute, Washington, DC, August 5.

32  Research handbook of financial markets

APPENDIX 1.1: SOURCES OF INFORMATION ON THE FEDERAL RESERVE BALANCE SHEET The following are useful sources of information on the Fed’s balance sheet: 1. By statute, the Fed publishes its balance sheet, as well as the balance sheets of each Reserve Bank, on a weekly basis on the H.4.1 Statistical Release, Factors Affecting Reserve Balances (https://www​.federalreserve​.gov​/releases​/ h41/). 2. Quarterly balance sheet information is published in the Quarterly Report on Fed Balance Sheet Developments (https://www​.federalreserve​.gov​/monetarypolicy​/quarterly​-balance​sheet​-developments​-report​.htm), with more detail, including the market value of Fed securities holdings, provided in the Federal Reserve Banks Quarterly Financial Reports (https://www​.federalreserve​.gov​/aboutthefed ​/combined​- quarterly​-financial​-reports​unaudited​.htm). 3. The Fed provides annual audited financial statements as well in the Federal Reserve System Audited Annual Financial Statements (https://www​.federalreserve​.gov​/aboutthefed​/audited​-annual​-financial​-statements​.htm). 4. Additional information on the Fed’s securities holdings is available in the Open Market Operations annual report prepared by the staff at the Desk at the New York Fed (https:// www​.newyorkfed​.org​/medialibrary​/media​/markets​/omo​/omo2020​-pdf​.pdf). 5. The results of auctions to purchase or sell securities are reported immediately on the website of the Federal Reserve Bank of New York (https://www​.newyorkfed​.org​/markets​/ domestic​-market​-operations). 6. Transaction-level data on asset purchases and sales, including the names of individual counterparties, are published on the NY Fed website with a two-year lag (for example, https://www​.newyorkfed​.org​/markets​/omo​_transaction​_data). 7. Transaction-level data on discount window loans are also available with a two-year lag, on the website of the Board of Governors (https://www​.federalreserve​.gov​/regreform​/discount​-window​.htm). 8. Transaction-level data on emergency lending programs are published with a lag on the Board website (for example, https://www​.federalreserve​.gov​/regreform​/transaction​-data​ .htm). 9. The accounting approach taken by the Fed is reported in the Financial Accounting Manual for Federal Reserve Banks (https://www​.federalreserve​.gov​/aboutthefed ​/files​/ bst​ fina​ccou​ntin​gmanual​.pdf). The manual details, among other things, the approach taken for consolidating the accounts of the 12 Reserve Banks as well as the special purpose vehicles used to manage many of the Fed’s emergency lending programs.

2. The balance sheet of the Eurosystem Oreste Tristani1

2.1 INTRODUCTION1 This chapter discusses the balance sheet of the Eurosystem, the authority responsible for monetary policy in the euro area and consisting of the European Central Bank (ECB) and the National Central Banks (NCBs) of the 19 member states that have adopted the euro.2 The balance sheet of the Eurosystem is a consolidated financial statement that presents the assets and liabilities of the ECB and of the NCBs vis-à-vis third parties.3 The annual consolidated balance sheet is published each year, together with the annual accounts of the ECB, along with a commentary on the developments of the main balance sheet items. Weekly updates of the balance sheet are available online every Tuesday. This chapter provides an overview of the evolution of the Eurosystem’s balance sheet since 1999, the start of the Economic and Monetary Union (EMU). Up until the onset of the global financial crisis of 2007–2008 (henceforth the crisis), developments in the balance sheet were of little relevance to gauge the monetary policy stance in the euro area. The policy stance was summarized by the level of a short-term interest rate – the instrument of monetary policy. In turn, the level of the short-term interest was set through liquidity operations producing small variations in the size and composition of the Eurosystem’s balance sheet. In the aftermath of the crisis, the monetary policy toolkit was expanded to include many new, or unconventional, monetary policy instruments: from market-making-of-last-resort operations in the early phases of the crisis; to large-scale asset-purchases, a.k.a. quantitative easing (QE), once the short-term nominal interest rate reached the effective lower bound. Monetary policy measures based on unconventional instruments led to, on the one hand, a considerable expansion of the Eurosystem’s balance sheet, on the other hand, a change in its composition, as new types of securities and lending operations with banks appeared on the asset side. As a result, the balance sheet increasingly gained relevance, both in policy discussions and in public debate. Not surprisingly, unconventional policy measures are also referred to as “balance sheet policies”.

1 The views expressed are personal and do not necessarily represent those of the European Central Bank. I wish to thank Kosuke Aoki, Alessandro Calza, Marco Corsi, Refet Gürkaynak, Polychronis Karakitsos, Luc Laeven, Wolfgang Lemke, Fernando Monar, Jonathan Wright, and participants in the Research Handbook of Financial Markets Online Conference for useful comments and suggestions on an early draft. I also benefited from excellent research assistance from Annachiara Tanzarella and Chiara Vergeat. 2 The Eurosystem is governed by the decision-making bodies of the ECB: the Governing Council and the Executive Board. It therefore customary to attribute monetary policy decisions for the euro area to “the ECB” (an abbreviation of “the Governing Council of the ECB”). 3 The balance sheet is consolidated in the sense that intra-Eurosystem claims and liabilities are netted out and therefore not visible. The ECB also publishes on a monthly basis a breakdown of how the ECB and the NCBs contribute to the balance sheet of the Eurosystem. 33

34  Research handbook of financial markets

In many respects, the ECB’s experience is comparable to that of other central banks in advanced economies. However, in the ECB case, there are also a few distinctive features stemming from the multi-country nature and the institutional set-up of EMU. The implications of these features for the Eurosystem’s balance sheet, including their impact on the implementation of QE, are also discussed in this chapter. Unconventional monetary policy measures gave rise to a vast empirical literature assessing their effects on financial and economic conditions. This assessment is challenging for at least two reasons. First, as with all policy interventions, one needs to disentangle the effects of the intervention itself from the effects of the economic and financial developments to which the intervention is responding. Second, many different unconventional measures were implemented over this period, often simultaneously, making it especially difficult to pinpoint the independent impact of each one of them. The chapter reviews the existing empirical literature and provides updated evidence on the effects of QE on asset prices. All in all, balance sheet policies do appear to have produced significant and persistent effects on financial conditions. There was also an impact on bank lending, even if, at times, at the cost of producing adverse effects on incentives. Much remains to be understood on the transmission channels of QE to inflation and the real economy. The rest of the chapter is organized as follows. Section 2.2 provides an account of developments in the Eurosystem’s balance sheet since the establishment of EMU on 1 January 1999. Key implications of the institutional setup of EMU for unconventional policies and for the balance sheet are highlighted in Section 2.3. Section 2.4 provides a brief summary of the theoretical mechanisms that have been proposed to understand the effectiveness of balance sheet policies and Section 2.5 reviews the empirical evidence regarding their effectiveness. Section 2.6 offers a few concluding remarks and discusses possible directions for future research.

2.2 THE BALANCE SHEET OF THE EUROSYSTEM This section provides an overview of the evolution of the Eurosystem’s balance sheet since the start of EMU.4 The overview is organized in three sub-periods: the pre-crisis years, in which the Eurosystem balance sheet was largely a side-show for monetary policy; the years from 2007 until 2014, where the balance sheet grew mostly as a result of liquidity-providing operations with banks; and the post-2014 period, in which the ECB deployed large-scale asset purchases. The main focus of this section is on the balance sheet items relevant from a monetary policy perspective. These are also the items whose size changed the most over the past two decades. The section discusses in more detail developments on the assets side of the balance sheet, while developments on the liabilities side are described more succinctly.

4 See Hartmann and Smets (2018) and Rostagno et al. (2019) for broader accounts of the ECB experience in its first 20 years.

The balance sheet of the Eurosystem  35

2.2.1 The Balance Sheet before the Financial Crisis Until 2006 the balance sheet remained roughly stable around levels close to 12% of euro area GDP – see Figure 2.1. “Lending to credit institutions” was the main asset item related to monetary policy. This item includes the open market operations in euros conducted by the Eurosystem to steer shortterm interest rates, manage the liquidity situation in the financial market, and signal the stance of monetary policy. In the early 2000s, open market operations did not include any outright transactions. They were exclusively conducted by means of reverse transactions whereby the Eurosystem provides credit to financial institutions in the form of collateralized loans. The

70 50 Assets:

% of euro area GDP

30 10 -10

Liabilities:

-30 -50

19

9 20 9 0 20 0 01 20 0 20 2 0 20 3 0 20 4 05 20 0 20 6 0 20 7 08 20 0 20 9 1 20 0 1 20 1 12 20 1 20 3 1 20 4 15 20 1 20 6 1 20 7 18 20 1 20 9 20 20 21

-70

Gold and foreign currency assets

Banknotes in circulation

Lending to credit institutions

Liabilities to credit institutions

Securities held for monetary policy purposes

Revaluation accounts, capital and reserves

Other assets

Other liabilities

Notes:   The figure groups the items of the consolidated balance sheet of the Eurosystem as follows. Assets: “Gold and foreign currency assets”: sum of A1, A2 and A3; “Lending to credit institutions”: A5; “Securities held for monetary policy purposes”: A7.1; “Other assets”: A4; “Other claims on euro area credit institutions denominated in euro”: A6; “Other securities”: A7.2; “General government debt denominated in euro”: A8; “Other assets”: A9. Liabilities: “Banknotes in circulation”: L1; “Liabilities credit institutions”: L2; “Revaluation account, capital and reserves”: sum of L11 and L12; “Other liabilities”: sum of L3, L4 , L5, L6, L7, L8, L9 and L10. Source:   ECB (see https://www​.ecb​.europa​.eu​/pub​/annual​/balance​/html​/all​_balance​_sheets​.en​.html).

Figure 2.1  The balance sheet of the Eurosystem

36  Research handbook of financial markets

main type of open market operation was the main refinancing operation (MRO), which provides liquidity with a maturity of one week in exchange for eligible collateral.5 The main refinancing operations are executed in the form of tender procedures, in which counterparties bid either the amount of money they wish to receive at an interest rate specified in advance by the ECB (fixed-rate tenders), or both the amount of money and the interest rates they are willing to pay (variable-rate tenders). In either case, the ECB decides the aggregate amount of liquidity to be provided, or “allotted”, to banks. If all bids are satisfied in full, tenders are said to be based on a full-allotment procedure. The operational framework is complemented by additional standing facilities, which can be used by banks to respond to idiosyncratic liquidity shocks: the deposit facility, which banks may use to make overnight deposits with the Eurosystem (at a rate below the MRO rate); and the marginal lending facility, which offers overnight credit to banks from the Eurosystem (at a penalty rate). The two facilities define an interest rate corridor around the MRO rate. Before the crisis, ECB lending to credit institutions hovered around 4% of euro area GDP – Figure 2.1. Total liquidity supply was determined by the structural liquidity needs of the banking sector as a whole, mostly due to the obligation to satisfy the minimum reserve requirement. The banking system as a whole held only a very small amount of excess reserves since idiosyncratic liquidity shocks could be dealt with either by borrowing and lending on the interbank market or through the averaging mechanism for required reserves. The interest rate on the main refinancing operations, or MRO rate, was the main signal of the ECB’s monetary policy stance. Except for some volatility at the end of the reserve-averaging period,6 the overnight interbank rate would typically remain very close to the MRO rate (ECB, 2002). The remaining items on the asset side of the balance sheet mainly include gold, foreign assets, and euro-denominated investments. On the liability side, the main item related to monetary policy before the crisis is “Liabilities to credit institutions”. Until 2006, they were mainly composed of reserve deposits to cover the minimum reserve requirement. To avoid imposing a burden on the banking system, banks’ required reserve balances with the Eurosystem are remunerated at the MRO rate, so that they do not carry an opportunity cost. Hence, changes in MRO rates can take place without implications for the demand for reserves – and therefore without requiring changes in the size of the Eurosystem’s balance sheet.7 The additional items on the liabilities side of the balance sheet are the stock of banknotes in circulation and items such as government deposits and residual items over which the Eurosystem has little or no control. Finally, liabilities include the Eurosystem’s capital and reserves, as well as revaluation accounts.

5 Initially, the MRO had a maturity of two weeks. It was shortened to one week in March 2004. 6 The Eurosystem’s minimum reserve requirement includes an averaging provision. Compliance with the reserve requirement is determined on the basis of a bank’s average daily reserve holdings over a pre-determined maintenance period, which is defined taking into account the calendar for Governing Council meetings. 7 See also Friedman and Kuttner (2010).

The balance sheet of the Eurosystem  37

2.2.2 The Balance Sheet between 2007 and 2014 After many years of stability, the balance sheet of the Eurosystem expanded considerably following the crisis. In an unprecedented growth spurt, it essentially doubled in size over two years reaching almost 22% of euro area GDP at the end of 2008. Following two years of stabilization around these higher levels, the balance sheet expanded again up to 30% of GDP in 2012, before experiencing a moderate contraction in the following two years, down to 22% of GDP in 2014. In the initial stages of the crisis, unconventional monetary policy measures largely consisted of the provision of liquidity to prevent the spillover of financial market stress to the real economy. These measures can be understood in the spirit of the traditional lender-of-lastresort function of central banks, i.e. aimed to prevent money market disruptions from spreading further and creating the conditions for a generalized banking panic (Garcia-de-Andoain et al., 2016). They were designed to foster the smooth transmission of changes in MRO rates to the real economy, rather than providing an independent monetary policy stimulus; De Fiore and Tristani (2019) provide a stylized characterization. The ECB has described the expansion of the balance sheet in this period as passive since the expansion was a by-product of monetary policy decisions, not its intended (intermediate) objective (ECB, 2015b). 2.2.2.1 Longer-term refinancing operations (LTROs) In the initial phases of the crisis that started in August 2007, the new measures mainly included a higher share of liquidity provision via three-month and later six-month operations to reduce funding uncertainty for banks. Beyond resulting in a lengthening of the average maturity of outstanding liquidity, this led to an increase in the stock of liquidity from EUR 451 billion in 2006 to EUR 637 billion in 2007 – see Figure 2.2.8 With the intensification of the crisis following the default of Lehman Brothers in September 2008, the ECB began conducting its refinancing operations as fixed-rate full-allotment tenders, thus satisfying in full banks’ demand for liquidity. This implied an increased intermediation role of the Eurosystem within the banking sector and a gradual increase in the size of the Eurosystem’s balance sheet to EUR 860 billion (or 9% of euro area GDP) in 2008. In addition, banks’ liquidity demand progressively shifted towards longer maturities. As shown in Figure 2.2, in 2008 LTROs replaced MROs as the main source of central bank liquidity for financial institutions – a switch that would prove to be persistent. Over time, funding uncertainty for banks led to a progressive lengthening of the maturity of LTROs. Already in June 2009, the ECB had started implementing one-year operations. With the intensification of the sovereign debt crisis, three-year operations were introduced at the end of 2011 to address renewed impairments in bank funding. The outstanding stock of liquidity reached EUR 1,126 billion in 2012 (over 11% of euro area GDP).

8 Refinancing via one-month and three-month operations also occurred before the financial crisis, but in lower amounts. For example, three-month operations amounted to about 20% of overall liquidity provided by the ECB over the course of 2004. At the end of 2007 this share had risen to approximately 50% of overall liquidity.

38  Research handbook of financial markets 2,202

2,100

MROs

1,800

TLTROs

LTROs 1,793

Other

EUR billion

1,500 1,200

1,126

900

860

864

637

600

630

269

00

204

228

02

249

01

547

99

300

298

345

406

764

752

750

559

596

734 624

451

04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21

03

20

20

20

20

20

19

0

Notes:   Breakdown of item A5 of the consolidated balance sheet of the Eurosystem (“Lending to euro area credit institutions related to monetary policy operations denominated in euro”). Legend: “MROs”: main refinancing operations; “LTROs”: longer-term refinancing operations; “TLTROs”: targeted longer-term refinancing operations; “Other”: pandemic emergency longer-term refinancing operation, fine-tuning reverse operations, structural reverse operations, marginal lending facility and credits related to margin calls. Source:   ECB (see https://www​.ecb​.europa​.eu​/mopo​/implement​/omo​/html​/index​.en​.html).

Figure 2.2  Lending to credit institutions 2.2.2.2 Covered bond purchase programmes (CBPPs) and securities markets programme (SMP) At the same time, the ECB started making more active use of its balance sheet through outright purchases of assets in specific financial market segments. In May 2009 the ECB decided to launch a first programme of purchases of euro-denominated covered bonds, i.e. bank bonds characterized by extremely low counterparty risk (CBPP). A second covered bond purchase programme was introduced in November 2011 (CBPP2). In May 2010 the securities markets programme (SMP) was launched to purchase government bonds in jurisdictions affected by the sovereign bond crisis. Overall, these three programmes had a rather limited impact on the ECB balance sheet during this period. Against a total balance sheet size of EUR 2,963 billion, the combined stock of securities held under CBPP1, CBPP2, and SMP peaked at EUR 277 billion in 2012 – see Figure 2.3. Covered bonds purchases were resumed as of October 2014 under a third programme (CBPP3). The total stock of covered bonds on the Eurosystem’s balance sheet under the CBPP3 peaked at EUR 301 billion in 2021. The SMP was terminated on 6 September 2012, but the securities purchased under this programme are being held to maturity. On the same day, the Governing Council decided to initiate the outright monetary transactions (OMT) programme. This programme has had no impact on the Eurosystem balance sheet since no bond purchases have been conducted.

The balance sheet of the Eurosystem  39 5,000 4,500

SMP CSPP+ABSPP

4,000

3,694

PSPP+PEPP

3,500

EUR billion

4,713

CBPP

3,000 2,651

2,500

2,632

2,386

2,000 1,654

1,500 1,000

803

500 0

29

135

274

277

236

217

2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021

Notes:   Breakdown of item A7.1 of the consolidated balance sheet of the Eurosystem (“Securities held for monetary policy purposes”). End-of-month book values at amortized cost. Legend: “CBPP”: covered bond purchase programme; “SMP”: securities markets programme; “CSPP+ABSPP”: Corporate sector purchase programme and asset-backed securities purchase programme; “PSPP+PEPP”: public sector purchase programme and pandemic emergency purchase programme (of both public and private assets). Source:   ECB (see https://www​.ecb​.europa​.eu​/mopo​/pdf​/APP​_breakdown​_history​.csv, https://www​.ecb​.europa​.eu​/ mopo​/pdf​/SMP​_breakdown​_history​.csv, https://www​.ecb​.europa​.eu​/mopo​/pdf​/CBPP2​_breakdown​_history​.csv and https://www​.ecb​.europa​.eu​/mopo​/pdf​/PEPP​_purchase​_history​.csv).

Figure 2.3  Securities held for monetary policy purposes 2.2.2.3 Liabilities The increase in Eurosystem’s lending to credit institutions went along with a rise in the level of excess liquidity in the system – see Figure 2.1. Excess liquidity is defined as the sum of the excess reserves held by credit institutions on current accounts with the Eurosystem, and the recourse to the deposit facility. In contrast to required reserves, excess liquidity carries an opportunity cost. Until May 2014 it was either not remunerated if held on current accounts or remunerated at the rate on the deposit facility. Between 2007 and 2012 liabilities to credit institutions increased almost threefold to a peak of over 9% of euro area GDP, but they fell again in the following two years when part of the LTROs was repaid.

40  Research handbook of financial markets

2.2.3 The Balance Sheet since 2014 The Eurosystem’s balance sheet started expanding again after 2014, up to a level equal to 40% of euro area GDP in 2018. The ECB introduced a new form of liquidity-providing operations and announced a new outright asset purchase programme (APP) of both private and public securities. These new measures were aimed at providing additional monetary accommodation. Asset purchases were accompanied by explicit quantity targets, which marked the shift to a more active use of the balance sheet. After becoming marginally smaller in 2019, the Eurosystem balance sheet shot up to a new unprecedented level of almost 70% of euro area GDP in 2021, as renewed longer-term refinancing operations and outright purchases played a crucial role in the response to the COVID-19 pandemic. 2.2.3.1 ABS purchase programme (ABSPP) and corporate sector purchase programme (CSPP) The first credit easing measures aimed at stimulating the lackluster recovery in 2014 were purchases of asset-backed securities and, as already discussed, covered bonds. The asset-backed securities purchase programme, or ABSPP, was directly aimed at stimulating new credit to the economy, as well as helping banks to diversify their funding sources and to stimulate the issuance of new securities. While amounts purchased under the ABSPP have been on an increasing trend ever since 2014, the overall contribution of this programme to the Eurosystem balance sheet remains negligible. Starting in June 2016, the Eurosystem’s asset purchases were extended to include the corporate sector purchase programme (CSPP) with the aim to further strengthen the pass-through of the monetary accommodation to the financing conditions of the real economy. The programme targets bonds issued by non-financial corporations, including small and mediumsized corporations. By 2021 the combined ABSPP and CSPP portfolios accounted for about EUR 340 billion – see Figure 2.3. 2.2.3.2 Public sector purchase programme (PSPP) The largest purchase programme implemented by the ECB is the PSPP, which was launched in January 2015. The programme targets investment-grade bonds issued by the public sector (i.e. euro area governments and agencies, or international and supranational institutions). It is characterized by precise guidelines in terms of the monthly pace of purchases and their crosscountry distribution. Notable aspects of this programme are discussed further in Section 2.3. The portfolio of government bonds purchased under the PSPP quickly rose from EUR 491 billion in 2015 to EUR 2.1 trillion in 2018 – see Figure 2.3. After stabilizing in 2019, the PSPP portfolio increased again, as a consequence of the COVID-19 pandemic. At the end of 2021 it stood at approximately EUR 2.5 trillion. 2.2.3.3 The pandemic emergency purchase programme (PEPP) In early 2020 the COVID-19 pandemic increased the level of macroeconomic uncertainty and triggered sharp stock market losses. The ensuing flight-to-safety process threatened to produce renewed sovereign bond market tensions across euro-area jurisdictions. The pandemic emergency purchase programme, or PEPP, was designed to counter these adverse dynamics through purchases of private and public sector securities conducted in a flexible manner. Net

The balance sheet of the Eurosystem  41

purchases under this programme were discontinued at the end of March 2022, when public sector securities purchased under the PEPP amounted to EUR 1,644 billion. 2.2.3.4 Targeted longer-term refinancing operations (TLTROs) and pandemic emergency longer-term refinancing operations (PELTROs) A second response to the COVID-19 pandemic was a form of liquidity support under a range of different programmes. The lion’s share was played by targeted longer-term refinancing operations, or TLTROs, which offered banks an amount of three-year funding linked to their loans to non-financial corporations and households. Two TLTROs had already been activated before the COVID crisis, in 2014 and in 2016. A third TLTRO programme launched in 2019 was recalibrated in April 2020 to include particularly favourable interest rate conditions during the period from June 2020 to June 2021.9 In June 2020 this resulted in the largest amount allotted in any single lending operation, equal to EUR 1,3 trillion. Lending to credit institutions, which had oscillated between roughly EUR 550 billion and EUR 750 billion after 2014, jumped up again in 2021 to reach the level of EUR 2.2 trillion – see Figure 2.2. Additional liquidity support was provided through a series of non-targeted pandemic emergency longer-term refinancing operations (PELTROs). While also helping to ensure sufficient liquidity in response to the pandemic, PELTROs played a less conspicuous influence on the Eurosystem’s balance sheet. 2.2.3.5 Liabilities Quantitative easing caused a new, unprecedented rise in the level of excess central bank liquidity in the banking system. Compared to a level of EUR 367 billion at the end of 2014, liabilities to credit institutions related to monetary policy operations in euros increased to EUR 768 billion in 2015, EUR 1.3 trillion in 2016, and EUR 1,9 trillion in 2017. After remaining stable around this level for two years, liquidity soared again as a result of pandemic-induced measures and reached an unprecedented value equal to 35% of euro area GDP in 2021 – see Figure 2.1. The deposit facility rate, which had been cut to zero in July 2011, was lowered to –0.10% in June 2014. The remuneration of excess reserves (previously zero) was also made equal to the deposit facility rate. At the same time, the MRO rate had reached the 0.15% mark, hence the opportunity cost of excess liquidity was contained, at least compared to the pre-crisis era in which it was equal to one percentage point. The abundant level of excess liquidity implied a de facto change of the operational framework to a floor system, in which market rates were influenced by the rate on the deposit facility, rather than the MRO rate. Cutting the deposit facility rate to negative territory led to a reduction in market rates and additional monetary policy accommodation. In the following years, the deposit facility rate would be cut further, reaching the –0.5% mark in September 2019.10   9 The interest can be as low as 50 basis points below the average interest rate on the deposit facility (and not higher than –1%) over the period from June 2020 to June 2022, and as low as the average interest rate on the deposit facility during the rest of the life of the respective TLTRO. 10 For banks with a large retail deposit base and holding large amounts of excess liquidity, the negative rate on the deposit facility represents an opportunity cost. In October 2019 the ECB introduced a two-tier remuneration system for excess reserves. A pre-defined multiple of banks’ required reserves would be exempt from the negative remuneration. The remaining amount of excess reserve would continue to be remunerated at the deposit facility rate.

42  Research handbook of financial markets

2.3 THE BALANCE SHEET AND THE MULTI-COUNTRY NATURE OF EMU The ECB’s experience with balance sheet policies is in many ways comparable to those of the Bank of Japan and the Federal Reserve, described elsewhere in this handbook. It is, however, also characterized by distinguishing features arising from the multi-country nature of the EMU, i.e. the lack of a central fiscal authority. One immediate implication of this feature is the coexistence of many sovereign bond issuers in the euro area. When setting up the PSPP, the ECB had to select not only the overall amount of bonds to purchase but also their cross-country distribution. Various options were available. From the perspective of maximizing the efficiency of the programme – i.e. its ability to lower yields for a given amount of purchase – an allocation tilted towards high-yield bonds could have been desirable. A focus on AAA securities would have been preferable to minimize the Eurosystem’s risk exposure. The ECB’s choice was to go for an intermediate solution, i.e. to purchase from each government a share of bonds corresponding to the respective NCB’s share in the ECB capital.11 The lack of a central fiscal authority also implies that no truly safe asset exists in the euro area. Some commentators argue that only government bonds in an economy with a central fiscal authority are truly safe because they will be redeemable with certainty against central bank money in extreme circumstances of fiscal stress. In other words, it is taken for granted that the central bank will purchase government debt in the primary market in extreme circumstances, even if it cannot be forced to do so in normal times. For the Eurosystem, this option is forbidden by the EU Treaty, which prohibits purchases of public debts on the primary market in any circumstance. Hence all assets on the balance sheet of the Eurosystem are subject to default risk, albeit to different degrees. It goes without saying that taking on balance sheet risks does not represent a problem per se for central banks, which do not pursue profit-maximizing objectives. Moreover, central banks are large players, whose large-scale purchases can, while increasing their own exposure to a certain risk, also reduce the amount of risk in the market. With reference to the euro area, Caballero et al. (2020) find that lender-of-last-resort operations and asset purchases tended to produce negatively correlated risks during the sovereign debt crisis, between 2010 and 2012. The paper concludes that, at times of malfunctioning markets, additional lending or asset exposures can increase the overall risk borne by the central bank less than proportionally. Nevertheless, risk-mitigating measures have been implemented by the Eurosystem to minimize the probability of losses occurring on its balance sheet. With regard to lending operations, such measures were already in place before the crisis. Lending operations with financial institutions are conducted against eligible collateral – i.e., collateral that satisfies minimum credit quality requirements and minimizes the risk of credit losses for the Eurosystem.12 With the launch of the APP, the ECB introduced additional risk-management measures, including eligibility requirements and exposure management – see ECB (2015a) for a detailed description. More specifically, exposure management involves the definition of limits for each of the purchase programmes. For private assets, due diligence procedures are also in place, which 11 NCBs’ shares are proportional to the relative size of the respective countries, measured in terms of population and GDP. 12 Over the years, the collateral framework has evolved, partly to ensure the availability of collateral for euro area banks – see Bindseil et al. (2017).

The balance sheet of the Eurosystem  43

can result in lower risk limits for riskier assets. For government bonds, an issuer share limit of 33% (on a stock basis) is applied.13 The Governing Council also decided that, in contrast to other monetary policy operations, the risks arising from most of the government bond purchases under the APP would not be shared between all NCBs.14 As in the case of other jurisdictions, purchases of long-term government bonds also increased the Eurosystem’s exposure to interest rate risk, i.e. the risk of improved economic conditions leading to a faster-than-expected increase in short-term interest rates. If bonds are accounted for at amortized cost, as is the case for the Eurosystem, no losses would be recorded following a reduction in the market value of previously purchased long-term bonds. The maturity mismatch between assets and liabilities, however, implies that higher policy interest rates would drive up the cost of reserves, while interest income on assets would remain virtually unchanged. Risk-management measures can reduce the adverse consequences of asset impairments, but not eliminate them altogether. Potential losses incurred by the Eurosystem in its monetary policy operations would lead to a reduction of the net income paid by NCBs to their respective governments.15 If significant, losses may prevent the distribution of net income to national governments altogether over a certain time period. Such developments may be asymmetric across countries.16 Compared to a single country with single fiscal and monetary authorities, in a monetary union, there is a higher risk that balance sheet losses become politically charged. An indicative example is the heated debate on the alleged fiscal implications of NCBs’ TARGET2 balances, which was sparked by the accumulation of large creditor (and debtor) TARGET2 positions at various points in time over the past two decades (see Perotti, 2020 for a recent overview of, and new contribution to, this debate).17 13 There is additionally an issue share limit initially set set at 25% and later increased to 33% subject to a case-by-case verification. The objective of the issue share limit is to avoid creating a situation in which the Eurosystem would have a blocking minority for the purposes of collective action clauses. 14 The PSPP targets both EU national sovereign bonds and EU supranational bonds. The risk-shared part of the PSPP is equal to 20% of the total, of which 10% (originally 8%) is made of national bonds and 10% (originally 12%) of supranational bonds. 15 NCBs also receive a share of ECB profits. Pursuant to Article 33 of the Statute, an amount to be determined by the Governing Council, which may not exceed 20% of the net profit of the ECB, can be transferred to the general reserve fund, subject to a limit equal to 100% of the capital. The remaining net profit must be distributed to NCBs in proportion to their paid-up shares. See Bunea et al. (2016) for a more detailed discussion of central bank profit distribution. 16 One can also entertain a theoretical scenario in which the whole, expected present discounted value of remittances of a central bank to the treasury is negative (see, for example, Sims, 2003; Reis, 2013a; Del Negro & Sims, 2015; Benigno & Nisticó, 2020). The likelihood of this scenario is however difficult to quantify in practice, given the need to compute the present discounted value of seignorage. Independent estimates accounting for the first wave of PSPP purchases suggest that the risk of very large balance sheet losses for the Eurosystem is remote (Hall & Reis, 2015). 17 At the end of each business day, the TARGET2 system records all bilateral transactions in central bank money among NCBs. For example, when the NCB of country A buys a bond issued by country A’s government from a financial institution located in country B, reserves will be credited on the financial institution’s current account. The payment must come from the NCB in country B, because that’s where the financial institution holds its reserves account. The NCB in country B therefore receives a TARGET2 claim for the corresponding amount, while the NCB in country A books a TARGET2 liability.

44  Research handbook of financial markets

2.4 THE BALANCE SHEET AS AN INSTRUMENT OF MONETARY POLICY IN THEORY Before reviewing the empirical evidence on the impact of balance sheet policies, it is useful to recall the theoretical channels that have been proposed to explain their effectiveness. A prominent view in the academic literature posits that balance sheet policies can only produce effects if they are understood as signals that the future path of policy interest rates will be lower than previously intended – a signalling channel.18 Due to increasing uncertainty over economic conditions in the more and more distant future, signalling is a more plausible transmission channel at short- and medium-term horizons. An additional mechanism, typically referred to as a portfolio balance channel, may account for the effects of large-scale asset purchases on longer-term yields. The starting observation is that assets exposed to different types of risk are not perfect substitutes for investors. As a result, central-bank purchases of risky assets can create scarcity and lead to a reduction in their yield. In turn, the lower yield will induce investors to rebalance their portfolios towards other assets, and the yield on the other assets will also be reduced. The broad-based decline in yields will produce more accommodative financial conditions and stimulate both the real economy and inflation. A few dynamic models have been put forward to capture the portfolio balance channel. For example, in Gertler and Karadi (2011) and Gertler and Kiyotaki (2010), imperfect asset substitutability arises because investors are leveraged-constrained financial institutions, such as banks. Their stochastic discount factor in the model can be written as L tB,t +1 = L t ,t +1 (1 - q + qlft +1 ) (2.1)



where L t ,t +1, is the households’ discount factor, ft denotes financial institutions’ leverage (assets over net worth), and 0 < q £ 1 and l > 0 are model parameters.19 Equation (2.1) shows that the higher leverage, the higher the financial institutions’ discount factor, i.e. their effective risk aversion. It also shows that, as in standard macroeconomic models, the households’ 18 Monetary theory has established a famous irrelevance result for a “pure” form of quantitative easing – see Eggertsson and Woodford (2003) and Wallace (1981). The simple intuition is that an increase in central bank reserves to finance an increase in central-bank holdings of government securities is equivalent to a swap of two highly substitutable liabilities of the public sector. The swap should produce no effects on the price of the security. 19 A variant of equation (2.1) is derived in Bocola (2016). Equation (2.1) is based on Correia et al. (2021), which shows that the value of the bank can be written recursively as a function of net worth, Z t , and the bank’s assets, Stb . At the beginning of period t, it can be written as

(

)

(

)

V Stb , Z t = (1 - q ) Et L t ,t +1Z t +1 + qEt L t ,t +1V Stb+1, Z t +1



where θ is the probability for the bank to to remain in business. Correia et al. also shows that in equilibrium V Stb , Z t = ut Stb + ht Z t for time-varying coefficients ut and ht . Using this result, and assuming that the incentive constraint is always binding, the value of the bank can be equivalently written as

(



)

æ ht +1 ö V Stb , Z t = Et L t ,t +1 ç 1 - q + lq ÷ Z t +1 l - ut +1 ø è

(

)

The balance sheet of the Eurosystem  45

discount factor L t ,t +1 is not affected by large-scale asset purchases. Large-scale asset purchases are, however, consequential for financial institutions if they reduce leverage and thus relax leverage constraints. This increases financial institutions’ risk tolerance, inducing them to rebalance their portfolios towards other assets, including new loans. Non-negligible effects of central-bank asset purchases on bond prices are also produced in the Vayanos and Vila (2021) model, which focuses on the term structure of interest rates and how it is affected by the preference of certain investors for specific maturity segments (a “preferred-habitat” view). Equation (2.1) highlights a few distinctive features of the portfolio balance channel. ●

●

●

●

To begin with, through their impact on financial institutions’ stochastic discount factor, central-bank asset purchases will tend to affect primarily risk premia, rather than expectations of future interest rates. Because of its impact on financial institutions’ stochastic discount factor, QE should have an impact on the price of all risky securities held by those institutions, not only those directly purchased by the central bank. This observation is important for empirical studies on the effects of QE. The effectiveness of QE hinges on targeting assets which cause a tightening of financial institutions’ leverage constraint (or assets that are highly desirable for certain investors in a preferred habitat framework). The intrinsic source of riskiness – be it interest rate risk for long-term bonds or counterparty risk for private credit claims – is immaterial for QE effectiveness. The potency of asset purchases is state-dependent. Effects should be more pronounced, the tighter financial institutions’ leverage constraints. This suggests that QE will be more effective, for example, when the financial sector is poorly capitalized.

2.5 THE EMPIRICAL EFFECTIVENESS OF THE BALANCE SHEET AS AN INSTRUMENT OF MONETARY POLICY A burgeoning literature tries to assess the empirical effectiveness of the various unconventional programmes implemented by the ECB. The majority of these papers have been written by central bank researchers, who, especially for analyses of credit conditions, could exploit confidential information.20 The rest of this section reviews this literature with a focus on papers studying the euro area evidence. For cross-country surveys of the literature on the effects of balance-sheet policies, see Dell’Ariccia et al. (2018) and Rossi (2021). The type of effectiveness discussed here focuses on the ability of lending programmes and asset purchase programmes to ease the overall stance of monetary policy in the euro area. Lending programmes also had a dimension of market-making of last resort, which was especially important in the early phases of the financial crisis. This dimension is not discussed in where l is a parameter indexing the financial friction. Hence, the bank discounts future profæ ht +1 ö its using the discount rate L t ,t +1 ç 1 - q + lq ÷ . This can be rewritten as in equation (2.1) l - ut +1 ø è because it can be shown that bank leverage ft = ht / ( l - ut ) . 20 Fabo et al. (2021) provide a critical review of these papers.

46  Research handbook of financial markets

detail, given that there is a broader consensus on the effectiveness of emergency provision of liquidity at times of crisis. 2.5.1 Evidence on Programmes of Long-Term Liquidity Provision 2.5.1.1 Effects on asset prices and bank lending Any asset price reaction to the announcement of these programmes can be a gauge of their macroeconomic effects. Krishnamurthy et al. (2018) look at the effects of the LTRO (as well as the SMP and OMT programmes) on sovereign yields in Italy, Spain, and Portugal. The paper relies on an event-study analysis and finds generally small effects both on sovereign yields and on stock returns – see also Szczerbowicz (2015).21 A different strand of the literature relies on disaggregated, or bank-level, loan data to analyse the pass-through of the provision of central-bank liquidity to lending. These papers focus on the particular transmission channel of monetary policy that was the immediate target of ECB liquidity-providing operations: the bank lending channel. The studies rely on crosssectional differences to identify the causal effects of monetary policy. They use micro-level datasets on banks and firms to control for firm-specific characteristics that may affect their demand for credit, the endogeneity of banks’ demand for LTRO liquidity to their own characteristics, and the non-random assignment of bank-firm lending relationships. Most studies focus on the impact of three-year LTROs on individual euro area countries – Garcia-Posada and Marchetti (2016) for Spain, Andrade et al. (2019) for France, Carpinelli and Crosignani (2021) for Italy, and Crosignani et al. (2020), Alves et al. (2021), and Jasova et al. (2021) for Portugal. In all countries, the results show that the three-year LTROs and especially the first operation implemented in December 2011 led to a moderate increase in loan supply to firms. The effect on bank lending was stronger for banks that were ex-ante more financially constrained and heterogeneous across firms of different sizes. These studies also show that, in countries hit by the sovereign debt crisis, banks used part of the central bank liquidity to fund purchases of domestic sovereign debt. This effect, which is indicative of bank risk-taking, appears to have been stronger for weakly capitalized and publicly owned banks (Acharya & Steffen, 2015; Drechsler et al., 2016; Altavilla et al., 2017; Ongena et al., 2019). A few recent papers have applied this methodology to analyse TLTROs (Albertazzi et al., 2021; Andreeva & García-Posada, 2021; de Haan et al., 2021). No evidence of increased risktaking is uncovered in the case of TLTROs, underscoring the benefits of having conditionality attached to long-term refinancing operations. 2.5.1.2 Macroeconomic effects A few studies of the macro impact of the early balance sheet policies rely on identified VARs (Boeckx et al., 2017; Boeckx et al., 2020; Burriel & Galesi, 2018; Darracq-Paries & De Santis, 2015; Giannone et al., 2012; and Lenza et al., 2010). Macroeconomic data force researchers to work at a quarterly frequency, which makes the identification of structural shocks challenging. In most cases, relatively strong identifying assumptions must be imposed. 21 Various papers have analysed the impact of the SMP – see De Pooter et al. (2018), Eser and Schwaab (2016), Ghysels et al. (2016), and Jäger and Grigoriadis (2017). They tend to find significant impact effects.

The balance sheet of the Eurosystem  47

Even stronger assumptions are imposed when studying the impact of balance-sheet policies through structural models. In the models, impairments in the interbank market lead to a credit crunch, since banks that have good investment opportunities may lack funding and be unable to obtain it from other banks. In this situation, Cahn et  al. (2017) and Quint and Tristani (2018) find that liquidity injections by the ECB produced substantial benefits, especially on aggregate investment. Bocola (2016) focuses on the extent to which LTROs were effective in reducing the adverse impact of the sovereign bonds crisis and finds small average effects on credit provision and on output. 2.5.2 Evidence on Large-Scale Asset Purchases 2.5.2.1 Effects on asset prices Many studies of QE have focused on its impact on asset prices using event studies. Given that they can be observed at high frequency, asset prices allow researchers to more easily identify the causal effects of QE upon announcement.22 For example, for the ECB announcement of January 2015 (of a QE programme amounting to about 10% of euro-area GDP), Altavilla et al. (2021) find an impact effect on ten-year sovereign bond yields of around 65 basis points – see also Urbschat and Watzka (2020) and Gnewuch (2022). However, any change in asset prices may be very short-lived and, as a result, produce no impact on the real economy. Wright (2012) is the first paper to analyse the persistence of asset price effects using a structural VAR estimated on daily US data. In the euro area, Andrade et al. (2016) provide early empirical evidence on this matter, based on the approach that identifies shocks using external instruments (Gertler & Karadi, 2015). The instrument for QE shocks is high-frequency changes in the five-year Bund yield on APP announcement dates. Altavilla et al. (2019) also analyse this question based on a related approach. In this case, QE shocks are instrumented with a suitably rotated statistical factor derived from high-frequency changes in many asset prices, as in Swanson (2021). Both Andrade et al. (2016) and Altavilla et al. (2019) tend to find persistent effects of QE in the euro area.23 Figure 2.4 updates the results in Andrade et al. (2016) and Altavilla et al. (2019) based on a sample extended to include more recent data.24 The figure compares the results obtained with two alternative identification approaches for QE shocks on APP announcement dates: high-frequency changes in the five-year Bund yield (denoted by “QE Dates” in the figure); and high-frequency changes in the Swanson (2021) statistical factor (denoted by “QE Factor” in

22 A limitation of the high-frequency identification approach is that it fails to capture some of the effects of a given policy measure, if that measure is anticipated by financial markets. This was at least partly the case for some of the ECB announcements on QE (see De Santis, 2020). 23 See also Rogers et al. (2014), which analyses the effects on bond yields, stock prices and exchange rates of the measures implemented until early 2014 by the ECB (as well as by the Federal Reserve, the Bank of England, and the Bank of Japan). 24 The instrument regression is performed over the period from 9 January 2014 to 12 December 2019. A total of 12 QE announcement dates are identified over this period. The high-frequency changes in asset prices around the ECB Press conference are taken from the dataset constructed by Altavilla et al. (2019). F-statistics of the first stage regressions of the instruments on the residual of the ten-year OIS rate are sufficiently high to suggest that instruments are valid. The VAR based on daily data is estimated with one lag from 19 July 2005 to 30 December 2019 and includes nine variables – see the note to the figure.

48

50

50

days

100 150 200 250 300 350

-0.5

0

0.5

50

days

100 150 200 250 300 350

GovALL/BUND 1

100 150 200 250 300 350

BEIR10Y

50

EURUSD

days

-0.4

-0.35

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

0.05

days

100 150 200 250 300 350

OIS10Y

-2

-1.5

-1

-0.5

0

0.5

1

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

50

50

days

100 150 200 250 300 350

EURIBOR1M

days

100 150 200 250 300 350

QE Dates QE Factor

EUROSTOXX

Figure 2.4  Impulse responses to a QE shock

Notes:   This figure plots the impulse responses (in days) of a monetary policy surprise consisting of a 100 basis point easing of the ten-year OIS rate in the euro area. The two identification strategies are described in Section 2.5.2.1. The specification is a VAR(l) including the following variables: ten-year OIS yields (OIS10Y); the euro-dollar exchange rate (EURUSD); the stock market index (EUROSTOXX); the ten-year break-even inflation rate (BEIR10Y); the spread between the average yield on ten-year euro area government bonds and the yield on the ten-year Bund (Gov ALL/BUND); the one-month euribor (EURIBOR1M). Additionally, the VAR includes the spread between the ten-year Bund yield and the ten-year OIS rate, the two-year OIS yield and the index of financial stocks. Dashed lines indicate 90% bootstrapped confidence intervals. Source:   author’s calculations.

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

-2

-1.5

-1

-0.5

0

0.5

percentage change percentage change

percentage change

percentage change

percentage change percentage change

The balance sheet of the Eurosystem  49

the figure). The impact effect of an identified QE shock is normalized to produce a reduction of 1% in the ten-year OIS rate. The figure confirms that QE reduces ten-year OIS rates quite persistently. The impulse response function of this variable is significant many months after the shock (either four or 12 months depending on the identification approach). Figure 2.4 also shows that the response of the short-term interest rate is insignificant, suggesting that the reduction in the long-term interest rates is mainly due to a decrease in risk premia. Eser et al. (2019) and Lemke and Werner (2020) reach similar conclusions based on versions of affine term-structure models. An additional effect of QE emerging from Figure 2.4 is a persistent depreciation of the euro against the dollar. This result is in line with those of Dedola et al. (2021), which finds that a 20% expansion of the Eurosystem balance sheet, relative to the Federal Reserve’s, causes a persistent euro depreciation of about 7%. Finally, Figure 2.4 suggests that QE leads to a significant increase in equity prices. Results are less robust across identification approaches for break-even rates and for the spread between the average five-year euro area government bond yield and the ten-year Bund yield. A few studies have investigated the impact of QE through a detailed analysis of the securities held by the various sectors of the economy. Koijen et al. (2017, 2021) focus on the three initial quarters of the ECB’s large-scale asset purchases and document that non-euro area investors, followed by euro-area banks, sold most of the bonds purchased over this period by the Eurosystem. Based on an estimated demand schedule for euro-area government bonds, the paper suggests that government bond yields decreased by 65 basis points on average. Albertazzi et al. (2020) and Paludkiewicz (2021) adopt a related approach to characterize the portfolio balance channel in different euro area countries. Their results indicate that, in countries vulnerable to the sovereign debt crisis, sectors with larger QE-related valuation gains reallocated their portfolios toward riskier corporate debt; in other countries, QE led to higher growth of bank lending to non-financial corporations. Te Kaat et al. (2021) focuses on portfolio rebalancing from bonds to houses. It argues that this channel can account for a large share of the impact of QE on regional growth differentials in Germany. 2.5.2.2 Macroeconomic and distributional effects Papers based on DSGE models produce marginally different results depending on whether QE purchases are considered in isolation (as in Andrade et  al., 2016; Sahuc, 2019; Burlon et al., 2017; Cova et al., 2019), or jointly with other programmes (such as TLTROs and negative interest rates, as in Mouabbi & Sahuc, 2019). Results are also affected by the underlying assumption as to how long monetary policy interest rates are kept at the effective lower bound. By and large, the peak expansionary effects of a programme of purchases equal to 10% of euro area GDP are found to be of the order of magnitude of 1% for GDP and 0.5% for inflation. Rather than using a fully-specified structural model, some authors adopt a two-step approach (Hutchinson & Smets, 2017; Rostagno et al., 2021). In the first step, they analyse the impact of central bank asset purchases on long-term yields. In the second step, the estimated yield impact is passed through a macro model to gauge the impact on real activity and inflation. Their results are comparable to those of pure DSGE models. A different type of multi-step approach has also been used to estimate the possible distributional effects of balance-sheet policies (Casiraghi et al., 2018; Lenza & Slacalek, 2018). These results suggest that households at the bottom of the income scale enjoyed large benefits from

50  Research handbook of financial markets

those measures, thanks to the employment gains produced by the economic stimulus. The impact of the programmes on wealth inequality appears to be negligible.

2.6 CONCLUDING REMARKS In the aftermath of the crisis, the ECB adopted unconventional monetary policy measures that led to a large expansion of the balance sheet of the Eurosystem. Over ten years have now gone past, yet the balance sheet remains an important indicator of the monetary policy stance in the euro area. Extended liquidity-providing operations and large-scale asset purchases continue to be part of the ECB’s toolkit. Regarding the effectiveness of balance sheet policies as tools of monetary policy accommodation in the euro area, a number of broad results of the empirical literature reviewed in this chapter appear to be robust and can be considered persuasive. First, balance sheet policies that acted through the banking system succeeded in stimulating lending to the private sector, either through a reduction in lending rates or through an increase in lending volumes. Second, large-scale asset purchases produced significant and persistent effects on long-term yields and on the exchange rate. Taken together, these two pieces of evidence are suggestive of expansionary effects also on the macroeconomy and on inflation. Third, the impact of QE on bond yields is strong at long maturities, which is indicative of effects through risk premia, rather than only through forward guidance. This appears consistent with a sizable role for the portfolio balance channel. Many questions, however, remain open. The conditions ensuring QE effectiveness on asset prices remain imperfectly understood. A notable feature of the euro area experience is that the bulk of QE purchases was implemented as of 2015, hence in relatively calm financial market conditions. This is in contrast to the US, where the first QE programme was announced in November 2008, in the midst of the crisis. Although a direct cross-country comparison is difficult, these differences in the financial environment do not appear to correspond to markedly different impacts on bond prices. QE effectiveness does not appear to hinge on the existence of financial market disruptions. Understanding the transmission of QE to the real economy poses additional challenges. A key question is whether a given change in bond yields will produce the same macroeconomic effects, independently of it being caused by QE or by conventional tools (current and expected future short-term rates). The evidence of a relatively large impact of QE on longer-maturity yields and on risk premia suggests that its transmission channels may be different from those of conventional monetary policy. For example, the traditional intertemporal substitution channel, which is a dominant part of the transmission mechanism of interest rate shocks in aggregate macro models, may play a more limited role for QE. Other channels operating through changes in asset prices, including the exchange rate, and incomes may be more relevant.25 Understanding the transmission mechanism of QE is also crucial to assess its potential side effects and its relative efficiency vis-à-vis other unconventional measures, such as negative interest rates. Side effects of QE could be produced on asset prices: recent papers have 25 At the same time, the role of the intertemporal substitution channel for the transmission of conventional monetary policy in aggregate macro models is likely to be excessive, as pointed out by the so-called HANK literature (see e.g. Kaplan et al., 2018).

The balance sheet of the Eurosystem  51

highlighted that the asset scarcity created by some purchase programmes can have adverse spillovers on the price of collateral and on certain repo rates (Corradin & Maddaloni, 2020; Ballensiefen et  al., 2021). However, negative interest rates can also create side effects for banks, due to their depressing net interest income. ECB (2021) provides a first encompassing assessment of the costs and benefits and concludes that negative side-effects have been minimized in the euro area, thanks to a flexible implementation of the various programs. An assessment of the benefits and costs of QE is also important to determine the optimal pace of “exit” – that is how actively and how fast the stock of purchases under the APP should be reduced, once conditions are ripe. Many considerations may affect this assessment, including environmental concerns (Papoutsi et al., 2021) and distributional consequences. These questions are especially pressing in an environment of rising inflation and higher short-term interest rates. Finally, the optimal size of the central bank balance sheet under normal conditions remains a subject of debate. The benchmark perspective in the macro literature is based on the “Friedman rule”, prescribing that the balance sheet of the central bank should remain large and respond elastically to financial institutions’ demand for liquidity (Reis, 2013b; Woodford, 2003). Other authors, however, have pointed out the benefits of maintaining tighter control on the size of the balance sheet, for example, due to macroprudential concerns (Kashyap & Stein, 2012). In the euro area, political considerations may favour a leaner balance sheet, as illustrated by the debate on TARGET2 balances. The ideal composition of the Eurosystem’s balance sheet under normal conditions is also a subject of debate. A specific question is whether all the assets in the APP portfolios should eventually be sold, so as to return to the pre-crisis situation in which no securities were held outright for monetary policy purposes. A new factor in this debate is the possible development of a “digital euro”, i.e. a Eurosystem liability offered in digital form to the general public for retail payments (ECB, 2020).

REFERENCES Acharya, V. V., & Steffen, S. (2015). The greatest carry trade ever? Understanding eurozone bank risks. Journal of Financial Economics, 115, 215–236. Albertazzi, U., Becker, B., & Boucinha, M. (2020). Portfolio rebalancing and the transmission of largescale asset programmes: Evidence from the euro area. Journal of Financial Intermediation, 17, 100896. Albertazzi, U., Nobili, A., & Signoretti, F. M. (2021). The bank lending channel of conventional and unconventional monetary policy. Journal of Money, Credit and Banking, 53, 261–299. Altavilla, C., Brugnolini, L., Gürkaynak, R. S., Motto, R., & Ragusa, G. (2019). Measuring euro area monetary policy. Journal of Monetary Economics, 108, 162–179. Altavilla, C., Canova, F., & Ciccarelli, M. (2020). Mending the broken link: Heterogeneous bank lending rates and monetary policy pass-through. Journal of Monetary Economics, 110, 81–98. Altavilla, C., Carboni, G., & Motto, R. (2021). Asset purchase programs and financial markets: Lessons from the euro area. International Journal of Central Banking, 17, 1–48. Altavilla, C., Pagano, M., & Simonelli, S. (2017). Bank exposures and sovereign stress transmission. Review of Finance, 21, 2103–2139. Alves, N., Bonfim, D., & Soares, C. (2021). Surviving the perfect storm: The role of the lender of last resort. Journal of Financial Intermediation, 47, 100918. Andrade, P., Cahn, C., Fraisse, H., & Mésonnier, J.-S. (2019). Can the provision of long-term liquidity help to avoid a credit crunch? Evidence from the eurosystem’s LTRO. Journal of the European Economic Association, 17, 1070–1106.

52  Research handbook of financial markets

Andrade, P., Breckenfelder, J., De Fiore, F., Karadi, P., & Tristani, O. (2016). The ECB’s asset purchase programme: An early assessment. ECB Working Paper No. 1956, September. Andreeva, D. C., & Garca-Posada, M. (2021). The impact of the ECB’s targeted long-term refinancing operations on banks’ lending policies: The role of competition. Journal of Banking & Finance, 122, 105992. Ballensiefen, B., Ranaldo, A., & Winterberg, H. (2021). Monetary policy disconnect. University of St. Gallen. Benigno, P., & Nisticò, S. (2020). Non-neutrality of open-market operations. American Economic Journal: Macroeconomics, 12, 175–226. Bindseil, U., Corsi, M., Sahel, B., & Visser, A. (2017). The eurosystem collateral framework explained. ECB Occasional Paper No. 189, May. Bocola, L. (2016). The pass-through of sovereign risk. Journal of Political Economy, 124, 879–926. Boeckx, J., De Sola Perea, M., & Peersman, G. (2020). The transmission mechanism of credit support policies in the euro area. European Economic Review, 124, 1–35. Boeckx, J., Dossche, M., & Peersman, G. (2017). Effectiveness and transmission of the ECB’s balance sheet policies. International Journal of Central Banking, 13, 297–333. Bunea, D., Karakitsos, P., Merriman, N., & Studener, W. (2016). Profit distribution and loss coverage rules for central banks. ECB Occasional Paper No. 169, April. Burlon, L., Gerali, A., Notarpietro, A., & Pisani, M. (2017). Macroeconomic effectiveness of nonstandard monetary policy and early exit. A model-based evaluation. International Finance, 20, 155–173. Burriel, P., & Galesi, A. (2018). Uncovering the heterogeneous effects of ECB unconventional monetary policies across euro area countries. European Economic Review, 101, 210–229. Caballero, D., Lucas, A., Schwaab, B., & Zhang, X. (2020). Risk endogeneity at the lender/investor-oflast-resort. Journal of Monetary Economics, 116, 283–297. Cahn, C., Matheron, J., & Sahuc, J.-G. (2017). Assessing the macroeconomic effects of LTROs during the great recession. Journal of Money, Credit and Banking, 49, 1443–1482. Carpinelli, L., & Crosignani, M. (2021). The design and transmission of central bank liquidity provisions. Journal of Financial Economics, 141, 27–47. Casiraghi, M., Gaiotti, E., Rodano, L., & Secchi, A. (2018). A “reverse Robin Hood”? The distributional implications of non-standard monetary policy for Italian households. Journal of International Money and Finance, 85, 215–235. Corradin, S., & Maddaloni, A. (2020). The importance of being special: Repo markets during the crisis. Journal of Financial Economics, 137, 392–429. Correia, I., De Fiore, F., Teles, P., & Tristani, O. (2021). Credit subsidies. Journal of Monetary Economics, 118, 2–14. Cova, P., Pagano, P., & Pisani, M. (2019). Domestic and international effects of the eurosystem expanded asset purchase programme: A structural model-based analysis. IMF Economic Review, 67, 315–348. Crosignani, M., Faria-e-Castro, M., & Fonseca, L. (2020). The (unintended?) consequences of the largest liquidity injection ever. Journal of Monetary Economics, 112, 97–112. Darracq-Paries, M., & De Santis, R. (2015). A non-standard monetary policy shock: The ECB’s 3-year LTROs and the shift in credit supply. Journal of International Money and Finance, 54, 1–34. Dedola, L., Georgiadis, G., Gräb, J., & Mehl, A. (2021). Does a big bazooka matter? Quantitative easing policies and exchange rates. Journal Monetary Economics, 117, 489–506. Dell’Ariccia, G., Rabanal, P., & Sandri, D. (2018). Unconventional monetary policies in the euro area, Japan, and the United Kingdom. Journal of Economic Perspectives, 32, 147–172. Del Negro, M., & Sims, C. A. (2015). When does a central bank’s balance sheet require fiscal support? Journal of Monetary Economics, 73, 1–19. Drechsler, I., Drechsel, T., Marques-Ibanez, D., & Schnabl, P. (2016). Who borrows from the lender of last resort? Journal of Finance, 71, 1933–1974. De Fiore, F., & Tristani, O. (2019). (Un)conventional policy and the effective lower bound. Journal of Economic Dynamics & Control, 106, 103696. De Pooter, M., Martin, R. F., & Pruitt, S. (2018). The liquidity effects of official bond market intervention. Journal of Financial and Quantitative Analysis, 53, 243–268.

The balance sheet of the Eurosystem  53

De Santis, R. A. (2020). Impact of the asset purchase programme on euro area government bond yields using market news. Economic Modelling, 86, 192–209. De Santis, R. A., & Holm-Hadulla, F. (2020). Flow effects of central bank asset purchases on sovereign bond prices: Evidence from a natural experiment. Journal of Money, Credit and Banking, 52, 1468–1491. ECB. (2002). The liquidity management of the ECB. Monthly Bulletin, May. ECB. (2011). The implementation of monetary policy in the euro area. Official Journal of the European Union, 20 September. https://www​.ecb​.europa​.eu​/pub​/pdf​/other​/gendoc201109en​.pdf. ECB. (2015a). The financial risk management of the Eurosystem’s monetary policy operations. July. ECB. (2015b). The role of the central bank balance sheet in monetary policy. Economic Bulletin, Issue 4, 61–77. ECB. (2020). Report on a digital euro, October. ECB. (2021). Assessing the efficacy, efficiency and potential side effects of the ECB’s monetary policy instruments since 2014. Occasional Paper No. 278. Eggertsson, G. B., & Woodford, M. (2003). The zero bound on interest rates and optimal monetary policy. Brookings Papers on Economic Activity, No. 1, 139–211. Eser, F., & Schwaab, B. (2016). Evaluating the impact of unconventional monetary policy measures: Empirical evidence from the ECB’s Securities Markets Programme. Journal of Financial Economics, 119, 147–167. Eser, F., Lemke, W., Nyholm, K., Radde, S., & Vladu, A. (2019). Tracing the impact of the ECB’s asset purchase programme on the yield curve. ECB Working Paper No. 2293. Fabo, B., Jančoková, M., Kempf, E., & Pástor, L. (2021). Fifty shades of QE: Comparing findings of central bankers and academics. Journal of Monetary Economics, 120, 1–20. Fahr, S., Motto, R., Rostagno, M., Smets, F., & Tristani, O. (2013). A monetary policy strategy in good and bad times: Lessons from the recent past. Economic Policy, 28, 243–288. Friedman, B. M., & Kuttner, K. N. (2010). Implementation of monetary policy: How do central banks set interest rates? In B. Friedman & M. Woodford (Eds.), Handbook of monetary economics 3, Ch. 24 (pp. 1345–1438), Elsevier. Garcia-de-Andoain, C., Heidera, F., Hoerova, M., & Manganelli, S. (2016). Lending-of-last-resort is as lending-of-last-resort does: Central bank liquidity provision and interbank market functioning in the euro area. Journal of Financial Intermediation, 28, 32–47. Garca-Posada, M., & Marchetti, M. (2016). The bank lending channel of unconventional monetary policy: The impact of the VLTROs on credit supply in Spain. Economic Modelling, 58, 427–441. Gertler, M., & Karadi, P. (2011). A model of unconventional monetary policy. Journal of Monetary Economics, 58, 17–34. Gertler, M., & Karadi, P. (2015). Monetary policy surprises, credit costs, and economic activity. American Economic Journal: Macroeconomics, 7, 44–76. Gertler, M., & Kiyotaki, N. (2010). Financial intermediation and credit policy in business cycle analysis. In B. Friedman & M. Woodford (Eds.), Handbook of monetary economics 3, Ch. 11 (pp. 547–599), Elsevier. Ghysels, E., Idier, J., Manganelli, S., & Vergote, O. (2016). A high-frequency assessment of the ECB Securities Markets Programme. Journal of the European Economic Association, 15, 218–243. Giannone, D., Lenza, M., Pill, H., & Reichlin, L. (2012). The ECB and the Interbank Market. Economic Journal, 122, F467–F486. Gnewuch, M. (2022). Spillover effects of sovereign debt-based quantitative easing in the euro area. European Economic Review, 145, 104133. de Haan, L., Holton, S., & van den End, J. W. (2021). The impact of central bank liquidity support on banks’ sovereign exposures. Applied Economics, 53, 1788–1806. Hall, R. E., & Reis, R. (2015). Maintaining central-bank solvency under new-style central banking. NBER Working Paper 21173. Hartmann, P., & Smets, F. (2018). The European Central Bank’s monetary policy during its first 20 years. Brookings Papers on Economic Activity, Fall, 1–118. Hutchinson, J., & Smets, F. (2017). Monetary policy in uncertain times: ECB monetary policy since June 2014. Manchester School, 85, e1–e15.

54  Research handbook of financial markets

Jasova, M., Mendicino, C., & Supera, D. (2021). Policy uncertainty, lender of last resort and the real economy. Journal of Monetary Economics, 118, 381–398. Jäger, J., & Grigoriadis, T. (2017). The effectiveness of the ECB’s unconventional monetary policy: Comparative evidence from crisis and non-crisis Euro-area countries. Journal of International Money and Finance, 78, 21–43. Te Kaat, D. M., Ma, C., & Rebucci, A. (2021). Real effects of the ECB’s quantitative easing: A housing portfolio channel. Mimeo, JHU Carey Business School. Kaplan, G., Moll, B., & Violante, G. L. (2018). Monetary policy according to HANK. American Economic Review, 108, 697–743. Kashyap, A. K., & Stein, J. C. (2012). The optimal conduct of monetary policy with interest on reserves. American Economic Journal: Macroeconomics, 4, 266–282. Koijen, R. S. J., Koulischer, F., Nguyen, B., & Yogo, M. (2017). Euro-area quantitative easing and portfolio rebalancing. American Economic Review P&P, 107, 621–627. Koijen, R. S. J., Koulischer, F., Nguyen, B., & Yogo, M. (2021). Inspecting the mechanism of quantitative easing in the euro area. Journal of Financial Economics, 140, 1–20. Krishnamurthy, A., Nagel, S., & Vissing-Jorgensen, A. (2018). ECB policies involving government bond purchases: Impact and channels. Review of Finance, 22, 1–44. Kühl, M. (2018). The effects of government bond purchases on leverage constraints of banks and nonfinancial firms. International Journal of Central Banking, 14, 93–161. Lemke, W., & Werner, T. (2020). Dissecting long-term bund yields in the run-up to the ECB’s public sector purchase programme. Journal of Banking and Finance, 111, 105682. Lenza, M., Pill, H., & Reichlin, L. (2010). Monetary policy in exceptional times. Economic Policy, 25, 295–339. Lenza, M., & Slacalek, J. (2018). How does monetary policy affect income and wealth inequality? Evidence from quantitative easing in the euro area. ECB Working Paper No. 2190, October. Mouabbi, S., & Sahuc, J.-G. (2019). Evaluating the macroeconomic effects of the ECB’s unconventional monetary policies. Journal of Money, Credit and Banking, 51, 831–858. Ongena, S., Popov, A., & Van Horen, N. (2019). The invisible hand of the government: Moral Suasion during the European sovereign debt crisis. American Economic Journal: Macroeconomics, 11, 346–379. Paludkiewicz, K. (2021). Unconventional monetary policy, bank lending, and security holdings: The yield-induced portfolio-rebalancing channel. Journal of Financial and Quantitative Analysis, 56, 531–568. Papoutsi, M., Piazzesi, M., & Schneider, M. (2021). How unconventional is green monetary policy?. Stanford University. Perotti, R. (2020). Understanding the German criticism of the target system and the role of central bank capital. CEPR Discussion Paper DP15067, July. Quint, D., & Tristani, O. (2018). Liquidity provision as a monetary policy tool: The ECB’s non-standard measures after the financial crisis. Journal of International Money and Finance, 80, 15–34. Reis, R. (2009). Interpreting the unconventional U.S. monetary policy of 2007–09. Brookings Papers on Economic Activity, 119–165. Reis, R. (2013a). The mystique surrounding the central bank’s balance sheet, applied to the European crisis. American Economic Review Papers & Proceedings, 103, 135–140. Reis, R. (2013b). Central bank design. Journal of Economic Perspectives, 27, 17–44. Reis, R. (2019). Can the central bank alleviate fiscal burdens? In D. Mayes, P. Siklos, & J.-E. Strum (Eds.), The economics of central banking, Oxford University Press handbooks in economics. Oxford University Press. Rogers, J. H., Scotti, C., & Wright, J. H. (2014). Evaluating asset-market effects of unconventional monetary policy: A multi-country review. Economic Policy, 29, 749–799. Rossi, B. (2021). Identifying and estimating the effects of unconventional monetary policy: How to do it and what have we learned? Econometrics Journal, 24, C1–C32. Rostagno, M., Altavilla, C., Carboni, G., Lemke, W., Motto, R., Saint Guilhem, A., & Yiangou, J. (2019). A tale of two decades: The ECB’s monetary policy at 20. ECB Working Paper No. 2346.

The balance sheet of the Eurosystem  55

Rostagno, M., Altavilla, C., Carboni, G., Lemke, W., Motto, R., & Saint Guilhem, A. (2021). Combining rates, forward guidance and asset purchases: Identification and impacts of the ECB’s unconventional policies. ECB Working Paper No. 2564. Sahuc, J.-G. (2019). The ECB’s asset purchase programme: A model-based evaluation. Economics Letters, 145, 136–140. Sims, C. A. (2003). Fiscal aspects of central bank independence. http://sims​.princeton​.edu​/yftp​/ Munich​/CBInd​.pdf. Swanson, E. T. (2021). Measuring the effects of federal reserve forward guidance and asset purchases on financial markets. Journal of Monetary Economics, 118, 32–53. Szczerbowicz, U. (2015). The ECB unconventional monetary policies: Have they lowered market borrowing costs for banks and governments? International Journal of Central Banking, 11, 91–127. Urbschat, F., & Watzka, S. (2020). Quantitative easing in the euro area – An event study approach. Quarterly Review of Economics and Finance, 77, 14–36. Vayanos,D., & Vila, J.-L. (2021). A preferred-habitat model of the term structure of interest rates. Econometrica, 89, 77–112. Wallace, N. (1981). A Modigliani–Miller theorem for open-market operations. American Economic Review, 71, 267–274. Woodford, M. (2003). Interest and prices. Princeton University Press. Wright, J. H. (2012). What does monetary policy do to long-term interest rates at the zero lower bound? Economic Journal, 122, F447–F466.

3. The Bank of Japan’s balance sheet Kosuke Aoki1

3.1 INTRODUCTION1 This chapter reviews developments in the Bank of Japan (BOJ)’s policies involving its balance sheet and empirical studies on those policies. Since the late 1990s, the BOJ has introduced a range of unconventional monetary policy measures that have changed the size and composition of its balance sheet. Having lowered its policy rate – the overall call rate – to 0.5% in 1995 and to virtually zero in 1999, the BOJ encountered the zero lower bound on interest rates much earlier than other central banks. This can be seen in Figure 3.1, which shows the BOJ’s call rate from 1986 to 2021 and the policy rates of the European Central Bank (ECB) and the Federal Reserve Bank (FRB) for comparison. As will be described in more detail in Section 3.2, the Japanese economy experienced an asset price bubble during the second half of the 1980s and the burst of this bubble in the early 1990s as well as a domestic financial crisis in 1997. In response to the strong deflationary pressure from the financial side of the economy, the BOJ introduced a zero interest rate policy in February 1999, followed by the first quantitative easing (QE) policy in March 2001. Since then, the BOJ has introduced a variety of unconventional monetary policy measures that use its balance sheet. These include the purchase of stocks owned by commercial banks in 2002, the Comprehensive Easing of 2010, and the Quantitative and Qualitative Easing (QQE) of 2013 and its subsequent enhanced versions. In 2020, in response to the economic contraction due to the outbreak of COVID-19, the BOJ further enhanced QQE. Each of these measures led to an increase in the BOJ’s balance sheet. To provide a sense of the extent to which the BOJ’s balance sheet has grown, Figure 3.2 shows the size of the central bank balance sheet relative to GDP for the BOJ, ECB, and FRB. The figure shows that the BOJ’s balance sheet started to expand much earlier and that, relative to GDP, the BOJ’s balance sheet has grown much larger than those of the ECB and FRB, reaching 130.4% of GDP as of the end of the year 2020. Figure 3.3 shows the inflation rates of Japan, the Euro area, and the United States. Despite its prolonged aggressive monetary easing, the BOJ has not been able to achieve its inflation target of 2%. This raises the question of whether the BOJ’s policies have been effective in increasing inflation and output. This chapter is organised as follows. Section 3.2 describes the Japanese economy and the BOJ’s policies since the 1980s. Next, Section 3.3 reviews the empirical evidence on the effects of the BOJ’s policies. Finally, Section 3.4 concludes and discusses avenues for future research.

1 I would like to thank Refet Gurkaynak, Junko Koeda, Ricardo Reis, Oreste Tristani, and Toshitaka Sekine for helpful comments. I am also grateful to Kazuki Matsuo, Shunji Suzuki, and Miho Tanaka for their excellent research assistance. Finally, financial support from the JSPS Grant-in-Aid for Scientific Research No. 18H05217 is gratefully acknowledged. 56

The Bank of Japan’s balance sheet  57 16%

BOJ

ECB

FRB

14% 12% 10% 8% 6% 4% 2% 0% -2%

86

88

90

92

94

96

98

00

02

04

06

08

10

12

14

16

18

20

22

Notes:   Japan: overnight call rate; ECB: main refinancing operations rate; FRB: Federal funds rate. Source:   BOJ, ECB, and FRB.

Figure 3.1  Policy rate of the BOJ, ECB, and FRB

3.2 THE JAPANESE ECONOMY AND THE BOJ’s POLICIES SINCE THE LATE 1980s The evolution of the BOJ’s balance sheet reflects its policies to achieve its mission of ensuring price stability and the stability of the financial system. To understand the BOJ’s policies, it is important to know that inflation in Japan has been considerably lower than in other major economies not only in recent years but since the 1980s, as is shown in Figure 3.3. Therefore, it is useful to briefly review developments in the Japanese economy and the BOJ’s policies since the 1980s. The 1980s were a decade characterized by financial deregulation as well as the Plaza Accord2 as well as the rise of Japan’s asset price bubble, which collapsed in the early 1990s. The early stages of financial deregulation in Japan took place in the first half of the 1980s (see Hoshi and Kashyap, 2001 for details). During this stage, borrowers such as firms started to obtain financing other than from banks, while banks – thanks to retail deposits – had funds to lend but remained stuck in doing business in the traditional way. This led banks to increase their lending toward the real estate sector. Meanwhile, the Plaza Accord signed in September 1985 attempted to correct the over-valuation of the US dollar. This resulted in a sharp appreciation of the Japanese yen, which caused a recession and put downward pressure on inflation. Looking again at Figure 3.3 depicting inflation in Japan, the Euro area, and the United States shows that in Japan inflation has rarely 2 The Plaza Accord was a joint agreement among France, West Germany, Japan, the United Kingdom, and the United States to bring about a depreciation of the US dollar against major currencies. It was signed on 22 September 1985.

58  Research handbook of financial markets 140%

BOJ

ECB

FRB

120% 100% 80% 60% 40% 20% 0%

99

00

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

16

17

18

19

20

Source:   BOJ, ECB, FRB, JP.Cabinet Office, Eurostat, U.S. Bureau of Economic Analysis.

Figure 3.2  Central bank balance sheet size relative to GDP: BOJ, ECB, and FRB

15%

Japan

the US

the Euro area

10%

5%

0%

-5%

80

82

84

86

88

90

92

94

96

98

00

02

04

06

08

10

12

14

16

18

20

Note:  For Japan, the figure shows the year-on-year rate of change in the Consumer Price Index (all items), for the Euro area the Harmonised Index of Consumer Prices (all items), and for the United States the Consumer Price Index for All Urban Consumers (all items). No adjustments for the effect of changes in taxes on the price level are made. In Japan, a consumption tax of 3% was introduced in 1987. It was raised from 3 to 5% in 1997, 8% in 2014, and 10% in 2019. Part of the changes in Japan’s price level in these years reflects the changes in the tax rate. Source:   Ministry of Internal Affairs and Communications (Japan), Eurostat, U.S Bureau of Labor Statistics.

Figure 3.3  Inflation rates of Japan, the Euro area, and the United States (year-on-year)

The Bank of Japan’s balance sheet  59 15%

Japan

the Euro area

the US

10% 5% 0% -5% -10% -15%

80

82

84

86

88

90

92

94

96

98

00

02

04

06

08

10

12

14

16

18

20

Source:   JP.Cabinet Office, Eurostat, U.S. Bureau of Economic Analysis.

Figure 3.4  Growth rate of real output (yoy) exceeded 2% since the 1980s. This likely has contributed to the low and stable inflation expectations of Japanese households and firms.3 In response to the deflationary pressure caused by the appreciation of the yen, the BOJ cut its policy rate.​ The financial deregulation and the low interest rates led to increases in land prices and stock prices. Figure 3.5 shows that land and stock prices started increasing in the mid-1980s, culminating in a speculative bubble. The bubble collapsed in the early 1990s and was followed by a long and severe recession and the banking crisis of 1997. Two policies have been cited as triggering the collapse of the bubble. The first is the imposition of aggregate lending restrictions on property-related lending in March 1990 by the Ministry of Finance. The second is the increase in the policy rate by the BOJ. The BOJ started raising the rate in May 1989 and continued to do so until August 1990. However, once the bubble collapsed and it became clear that Japan was heading for recession, the BOJ started decreasing the policy rate in July 1991 and continued doing so until the official discount rate fell to 0.5% in September 1995. Japan’s policy rate can be regarded to have hit the effective lower bound at this time, although the BOJ reduced the policy rate even further when it introduced the zero interest rate policy on 12 February 1999.4 After briefly exiting from the zero interest policy, the BOJ introduced its first quantitative easing (QE) policy on 19 March 2001.5 The BOJ changed its operating target from the overnight call rate to the outstanding balance of banks’ current accounts at the BOJ. Current account balances at the BOJ 3 See Diamond, Watanabe, and Watanabe (2020) for an analysis of how Japanese households’ past experience of inflation affects their inflation expectations. 4 ​ https://www.boj.or.jp/en/mopo/mpmdeci/mpr_1999/k990212c.htm. The Bank announced that it would aim to guide the call rate to move around 0.15%, and subsequently induce a further decline. 5 ​https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2001/k010319a.htm

60

5 4 3 2 1 0

5

4

3

2

1

0

Figure 3.5  Japan’s stock price and land price

Note:   All prices are normalised at 1 in 1980. Source:   Ministry of Land, Infrastructure, Transport, and Tourism, Nikkei Inc.

6

6

80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 18 20

7

7

9 8

Nikkei225

8

9 Commercial

Combined

80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 18 20

Residential

The Bank of Japan’s balance sheet  61

increased throughout the period this policy was in operation until the BOJ exited from QE on 9 March 2006.6 At the peak (at the end of December 2005), current account balances stood at 155.6 trillion yen. During the first round of QE, the BOJ expanded its balance sheet mainly by purchasing conventional assets (namely, government bonds with shorter maturities). During this period, the BOJ also introduced two other important policy measures. First, the BOJ introduced forward guidance for the first time in April 1999.7 At the time, the then Governor Masaru Hayami stated: “I think that the Bank will maintain the zero interest rate policy until deflationary concerns are dispelled”.8 Since then, the BOJ has updated its guidance repeatedly to increase the clarity and effectiveness of its communication with the public. Second, in October 2002, the BOJ decided to purchase stocks from commercial banks.9 This can be regarded as one of the earliest examples of “credit easing policy” by a central bank. At the time, the negative feedback loop between declining stock prices and the health of bank balance sheets was perceived as a serious problem. By reducing commercial banks’ exposure to stock market risks, the BOJ attempted to reduce the adverse effects of stock price fluctuations on commercial banks’ balance sheets and to mitigate this negative feedback loop. Economic conditions improved and deflation abated around 2005, and the BOJ exited from the first round of QE on 9 March 2006. At the same time, the BOJ announced “The Bank’s Thinking on Price Stability”.10 This announcement clarifies what the BOJ sought to achieve in the long run. Following the exit from QE, the BOJ’s balance sheet shrank fairly quickly. On 14 July 2006, the BOJ increased the call rate to 0.25%, followed by another increase to 0.5% on 21 February 2007. As in other countries, the Global Financial Crisis of 2007–2008 caused a serious recession in Japan. In addition to the decrease in aggregate demand, the sharp appreciation of the Japanese yen put downward pressure on the Japanese economy. This appreciation was largely caused by a large decrease in the interest rate differentials between Japan and other countries such as the United States and countries of the Eurozone. As is shown in Figure 3.1, while the other two central banks decreased their policy rates aggressively during 2007 and 2008, there was almost no room for the BOJ to cut interest rates. The BOJ decreased the call rate to 0.3% and then to 0.1% in 2008. It then introduced “Comprehensive Monetary Easing” on 5 October 2010,11 the main purpose of which was to decrease long-term interest rates. It consisted of three measures. The first was a reduction in the policy rate to 0–0.1%. Second, the BOJ strengthened its forward guidance by clarifying the time horizon. Specifically, the BOJ announced that it would maintain its virtually zero interest rate policy until price stability was in sight, and clarified that it would base its judgement of whether price stability had been achieved on its “understanding of medium- to long-term price stability”. Third, the BOJ established an Asset Purchase Programme, under   6 ​https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2006/k060309.htm   7 At the time, the BOJ referred to the policy as a commitment with a “Policy Duration Effect”.   8 The original announcement at the press conference on the 13 April 1999 is in Japanese. For the background to this announcement, see the “Minutes of the Monetary Policy Meeting” on 9 April 1999. https://www​.boj​.or​.jp​/en​/mopo​/mpmsche​_minu​/minu​_1999​/g990409​.htm/   9 ​https://www.boj.or.jp/en/finsys/spp/fss0210b.htm. This decision was made at a regular board meeting, not at a Monetary Policy Meeting. It was regarded as a policy measure to ensure the stability of the financial system rather than a monetary policy measure. 10 https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2006/mpo0603a.htm 11 https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2010/k101005.pdf

62  Research handbook of financial markets

which it would purchase various financial assets such as Japanese government bonds (JGBs), commercial paper, corporate bonds, exchange-traded funds (ETFs), and Japan real estate investment trusts (J-REITs). The size of the programme started at 35 trillion yen and was subsequently increased to some 80 trillion yen before the programme was abolished at the introduction of QQE on April 2013. On 22 January 2013, the BOJ introduced its “price stability target” of 2%.12 In addition, it introduced the “open-ended asset purchasing method”, under which it purchases financial assets without setting any termination date. This implied that the size of the Asset Purchase Programme would be increased by about 10 trillion yen in 2014. Finally, together with the government, the BOJ released the “Joint Statement of the Government and the Bank of Japan on Overcoming Deflation and Achieving Sustainable Economic Growth”. The BOJ introduced Quantitative and Qualitative Easing (QQE) on 4 April 2013.13 This was regarded as one of the “three arrows” of the so-called Abenomics policies pursued under the Abe administration.14 The BOJ announced the following: “The BOJ will achieve the price stability target of 2% in terms of the year-on-year rate of change in the consumer price index (CPI) at the earliest possible time, with a time horizon of about two years” (to see the statement, go to the URL in footnote 5). The “quantitative” aspect of QQE consists of the increase in the size of the BOJ’s balance sheet. When the BOJ announced QQE, it also switched its operating target from the call rate to the monetary base. It further announced that it would increase the monetary base at an annual pace of about 60–70 trillion yen and that it would purchase JGBs at an annual pace of about 50 trillion yen. The “qualitative” aspect of QQE consists of a change in the composition of the BOJ’s assets that the BOJ would purchase. It announced that it would increase the average remaining maturity of its JGB purchases from about three years to about seven years. In order to lower risk premiums on assets, the BOJ decided to purchase ETFs and J-REITs at an annual pace of one trillion yen and 30 billion yen respectively. Since then, the BOJ has extended QQE policy measures in several ways. On 31 October 2014, it announced an “expansion” of QQE,15 accelerating the annual increase in the monetary base, increasing asset purchases, and extending the average remaining maturity of JGB purchases. Specifically, it extended the average remaining maturity of JGB purchases to about 7–10 years and increased purchases of ETF and J-REITs. On 29 January 2016, the BOJ introduced negative interest rates,16 and on 21 September 2016, it introduced a new framework called “QQE with Yield Curve Control”.17 At the same time, the BOJ published a “Comprehensive Assessment” of its monetary policies with the aim of examining why it had failed to achieve its price stability target of 2%. The main document is Bank of Japan (2016), supplemented by four research papers: Nishino et al. (2016), Fujiwara et al. (2016), Kan, Kishaba, and Tsuruga (2016), and Kawamoto and Nakahama (2017). Their main conclusion is that while QQE was effective in decreasing long-term interest rates and stimulating aggregate demand, inflation failed to achieve the BOJ’s target of 2% because 12 ​https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2013/k130122a.pdf 13 ​https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2013/k130404a.pdf 14 Abenomics attempted to raise the inflation rate and to promote economic growth through (1) expansionary monetary policy, (2) flexible fiscal policy, and (3) policies to enhance growth. 15 https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2014/k141031a.pdf 16 https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2016/k160129a.pdf 17 https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2016/k160921a.pdf

The Bank of Japan’s balance sheet  63

of the sluggish nature of inflation expectations. They argue that inflation expectations were “adaptive” in the sense that inflation expectations react to past realisations of inflation rather than being guided by the BOJ’s inflation target. A key feature of QQE with Yield Curve Control (YCC) is its emphasis on the attempt to control long-term interest rates rather than focusing on the amount of JGB purchases. In particular, the BOJ aims at holding the ten-year JGB yield at 0%.18 In addition to the traditional “competitive auction method”, the BOJ introduced a “fix-rate method” under which it purchases a fixed or unlimited amount at purchasing yields set by the Bank. When YCC was introduced in September 2016, the BOJ announced that [i]n case of a spike in interest rates, the BOJ stands ready to conduct fixed-rate JGB purchase operations – for example, those with regard to 10-year and 20-year JGB yields – in order to prevent the yield curve from deviating substantially from the current levels.19

Yield Curve Control has been quite effective in controlling ten-year JGB yields. For example, by investigating JGB futures prices, Shioji (2021) shows that since the introduction of the YCC fiscal policy-related news does not have a significant impact on the JGB market. The BOJ also strengthened its forward guidance policy by introducing an “inflation-overshooting commitment”. Specifically, it announced that “[t]he Bank will continue expanding the monetary base until the year-on-year rate of increase in the observed CPI (all items less fresh food) exceeds the price stability target of 2 percent and stays above the target in a stable manner”. This statement aims at raising inflation expectations. On 16 March 2020, the BOJ further enhanced further monetary easing in response to the outbreak of COVID​-19.20 Policy measures include a further increase in purchases of JGBs, ETFs, and J-REITs, as well as the provision of loans against corporate debt as collateral. The BOJ also increased purchases of commercial papers and corporate debt. On 19 March 2021, the BOJ announced “Further Effective and Sustainable Monetary Easing”.21 This policy announcement was accompanied by the “Assessment for Further Effective and Sustainable Easing”,22 which is a follow-up analysis of the “Comprehensive Assessment” that is published in 2016. Bank of Japan (2021) is the main document, and it is accompanied by three research papers by Kawamoto et al. (2021), Kawamoto, Nakajima, and Mikami (2021), and Adachi, Hiraki, and Kitamura (2021).

18 In July 2018, the BOJ announced that it would allow yields to fluctuate to some extent in line with economic conditions. At the press conference after the policy meeting, the Governor Kuroda replied that the yield may fluctuate around plus and minus 0.2 percentage points from the target level. Link to the policy statement: https://www​.boj​.or​.jp​/en​/mopo​/mpmdeci​/state​_2018​/ k180731a​ .htm/​# nt01. Link to the press conference transcript (in Japanese): https://www​.boj​.or​.jp​/announcements​/press​/ kaiken​_2018​/ kk180801a​.htm/ 19 ​https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2016/rel160921b.pdf. This statement refers to the BOJ’s purchases of JGBs in September 2016. In a subsequent statement released on 30 September 2016, the BOJ announced that it would conduct such auctions “as needed, such as when the level of the yield curve changes substantially”. For the latter statement, see https://www.boj.or.jp/en/mopo/ mpmdeci/mpr_2016/rel160930c.pdf 20 ​https:/​/www​.boj​.or​.jp​/en​/mopo​/mpmdeci​/state​_2020​/ ​k200316b​.htm/ 21 https://www​.boj​.or​.jp​/en​/mopo​/mpmdeci​/state​_2021​/ k210319a​.htm/ 22 https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2021/k210319c.pdf

64  Research handbook of financial markets 800 600 400 200 0 -200

-600 -800

Currency in circulaon Reserves balances Bills sold & Reverse repo Other liabilies & net assets

ETFs JGBs TBs Short-tem funds-supplying operaons Foreign currency assets Credit facilities & equities excluding ETFs LLR lending Other assets

-400

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

16

17

18

19

20

21

Source:   Bank of Japan.

Figure 3.6  The BOJ’s balance sheet composition As a result of this sequence of aggressive monetary easing policies, the BOJ’s balance sheet increased to almost 130% of GDP, as shown in Figure 3.2. Figure 3.6 shows the composition of the BOJ’s balance sheet from 2001 to 2021, while Table 3.1 shows the balance sheet of the BOJ as of 31 January 2022.23 The table shows that JGBs account for 70.6% of the BOJ’s assets. The effects of the JGB purchase policy on a variety of aspects such as interest rates and macroeconomy are reviewed in Sections 3.3.3 and 3.3.5. ETF holdings have been increasing since 2010 and accounted for 5.0% in January 2022. The effects of the ETF purchase policy are reviewed in Section 3.3.2. Following the outbreak of COVID-19, funds supplying operations increased markedly, reflecting the Bank’s effort to support corporate financing. As a result, loans under such operations now account for 19.8% of the BOJ’s assets. Turning to the BOJ’s liabilities, the amount of currency in circulation has been very stable. The expansion of the BOJ’s balance sheet therefore has come about through an increase in reserve balances, which account for 74.3% of liabilities. Reserve balances are divided into three tiers. According to the monetary policy statement on 29 January 2016 (see footnote 15 for the URL), this three-tier system was introduced in order to prevent an excessive decrease in financial institutions’ profits that could weaken their functioning as financial intermediaries. The three-tier system consists of basic balances, macro add-on balances, and policy-rate balances, with a positive interest rate applied to basic balances, a zero interest rate applied to macro add-on balances, and a negative interest rate applied to policy-rate balances. As of 23 The most recent data can be found on the BOJ’s website: https://www​.boj​.or​.jp​/en​/statistics​/boj​/ other​/acmai​/index​.htm/. Time series data are available here: https://www​.stat​-search​.boj​.or​.jp​/ssi​/ cgi​-bin ​/famecgi2​?cgi=$nme_a000_en&lstSelection=BS01

The Bank of Japan’s balance sheet  65

Table 3.1  The BOJ's balance sheet as of January 2022 Assets

Liabilities and Net Assets Billion yen

JGBs

512,139 (70.63%)

Treasury discounted bills

11,089 (1.53%)

Loans through short-term funds-supplying operations

143,724 (19.82%)

Foreign currency assets

7,681 (1.06%)

Credit facilities and equities excluding ETFs

12,685 (1.75%)

ETFs

36,404 (5.02%)

Other assets Total

Billion yen Currency in circulation 119,069 (16.42%) Reserves balances

539,074 (74.34%)

Bills sold and reverse repo

24 (0.00%)

1,422 (0.20%)

Other liabilities and net assets

66,977 (9.24%)

725,144

Total

725,144

Source:   Bank of Japan.

January 2021, a negative interest rate of minus 0.1% is applied to the policy-rate balance. The BOJ periodically reviews the way that these balances are calculated.24

3.3 EMPIRICAL STUDIES ON THE BOJ’s BALANCE SHEET POLICY As shown in Section 3.2, the BOJ has adopted a wide variety of conventional and unconventional monetary policy measures, including forward guidance (adopted in 1999), Quantitative Easing (2001), Comprehensive Monetary Easing (2010), the adoption of an inflation target (2013), Quantitative and Qualitative Easing (2013), the adoption of a negative interest rate (2016), the introduction of Yield Curve Control (2016), and so on. However, despite these policies, Japan’s inflation rate has remained below the BOJ’s target of 2%. This raises the question of whether these policies have been ineffective or whether inflation would have been even lower in their absence, and whether they at least had other positive effects on the economy. Since this chapter is concerned with the BOJ’s balance sheet, the literature review in this section concentrates on empirical studies that have examined policies that affect the size and composition of the BOJ’s balance sheet. We review the literature focusing on the three channels through which balance sheet policy affects the economy highlighted by Krishnamurthy and Vissing-Jorgensen (2013): the signalling channel, the capital constraints channel, and the scarcity (or, more broadly, portfolio rebalancing) channel.

24 See, for example, the BOJ’s “Market Operations in Fiscal 2020” for details. https://www​.boj​.or​.jp​/ en​/research​/ brp​/mor​/data​/mor210909​.pdf

66  Research handbook of financial markets

The signalling channel refers to the fact that in addition to any direct effects of asset purchases, a central bank’s balance sheet policy also affects long-term interest rates and financial variables indirectly by changing market participants’ expectations about the future course of short-term policy rates. Therefore, one aspect of balance sheet policy is that it works as a forward guidance policy with regard to short-term interest rates. The capital constraints channel refers to the impact that central bank asset purchases have on risk premiums on financial assets. As such, the capital constraints channel is most relevant when financial markets are in distress and risk premiums are high because investors (such as financial intermediaries) are financially constrained. Under such circumstances, abnormal returns may not be arbitraged away, and central bank intervention can raise asset prices and improve economic efficiency. Any effects that the BOJ’s ETF purchases may have had would be via this channel. Finally, the scarcity channel relates to central bank purchases of long-term government bonds and is closely linked to the preferred habitat theory (Vayanos and Vila, 2021; Krishnamurthy and Vissing-Jorgensen, 2012). Long-term government bonds are special assets in the sense that they are very safe and liquid. Certain investors, such as pension funds and insurance companies, have special demand for long-term safe assets. When the supply of such assets is limited, government bonds carry a convenience yield. Long-term government bond purchases by a central bank reduce the supply of such bonds to the public, increasing their convenience yield (so that the interest rate declines). The effects of the BOJ’s purchases of JGBs would be via this channel. Surveys of the BOJ’s quantitative easing in the early period from 2001 to 2006 are provided by Ugai (2007), Ueda (2012), and Shiratsuka (2010). These surveys suggest that there is clear evidence for the signalling channel. However, there is no clear evidence that QE during this period had significantly positive effects on inflation and output. 3.3.1 Balance Sheet Expansion and Expectations Management The academic literature since the early 2000s has highlighted the importance of managing private sector expectations (see, e.g., Jung, Teranishi, and Watanabe, 2005; Eggertsson and Woodford, 2003). One question therefore is whether the BOJ’s policy has changed private sector expectations about interest rates and/or inflation. Regarding the effects on short-term interest rates, the literature shows that there is clear evidence that balance sheet policies have changed market expectations about future shortterm interest rates, as reviewed by Ugai (2007) and Ueda (2012).25 For example, investigating the effects of the BOJ’s commitment to the zero interest rate policy on interest rates using a macro-finance approach, Oda and Ueda (2007) show that the commitment lowered longterm interest rates — both the expectations component and the risk premium component. The authors also investigate whether the increase in the BOJ’s balance sheet had any effects on market participants’ expectations with regard to long-term interest rates. They find that the

25 Signalling also affects asset prices such as stock prices and exchange rates. Right after Shinzo Abe was elected as Prime Minister, stock prices and exchange rates reacted strongly. The response of stock prices and exchange rates has been documented and examined by Ueda (2002), Ito (2014), and Fukuda (2015).

The Bank of Japan’s balance sheet  67

BOJ’s balance sheet expansion was perceived as indicating a greater willingness to keep the call rate at zero. However, the effects on inflation expectations have been shown to be modest even if the period of analysis is extended to more recent periods that include QQE. For instance, examining developments in inflation expectations from the introduction of QQE until 2016, Nishino et al. (2016) find that such developments in inflation expectations can be divided into three phases. The first phase is the period from April 2013 until summer 2014, during which inflation expectations rose. The second phase consists of the period from summer 2014 to summer 2015, during which inflation expectations stayed flat. Finally, the third phase covers the period from summer 2015 until autumn 2016 (the time their study was written), during which inflation expectations weakened. They argue that inflation expectations failed to increase to the BOJ’s inflation target of 2% because of adverse shocks and the adaptive nature of Japan’s inflation expectations. Meanwhile, Romer (2013) has argued that policy regime shifts play an important role in changing inflation expectations. She suggests that the introduction of QQE might be comparable to the regime change brought about by Franklin Roosevelt in the 1930s that ended the Great Depression. Fujiwara, Nakazono, and Ueda (2015) and Michelis and Iacoviello (2016) investigate whether the introduction of QQE led to the perception of a “policy regime shift” that drastically changed the public’s expectations. However, both studies come to the conclusion that this was not the case. Other studies examining the impact on inflation expectations include that by Fujiwara, Nakazono, and Ueda (2015), who use the QUICK survey system that asks market participants about their expectations regarding the future course of interest rates and inflation expectations. They show that the increase in inflation expectations at the onset of Abenomics was small. The authors also examined whether the perception of the BOJ’s monetary policy stance changed. They find that this was not the case. They argue that market participants already expected that short-term interest rates would be kept at zero even before the introduction of QQE, and there was little room for changing market expectations about the future course of the BOJ’s policy rate. Meanwhile, Michelis and Iacoviello (2016) use a vector autoregression (VAR) approach to examine the impact on expectations. Using a long-run identification restriction, they identify inflation target shocks and estimate impulse response functions of output and inflation to a shock. They conclude that while QQE did increase inflation expectations it was insufficient to generate a regime shift. Similarly, Hattori et al. (2021) find that private sector inflation forecasts increased less than the BOJ’s forecasts after the introduction of the inflation target, confirming how difficult it is to raise inflation expectations. Other studies highlighting the weak response of inflation expectations to policy changes in Japan include those by Shintani and Soma (2020) and Van den End and Pattipeilohy (2017). Using a panel dataset of forecasts by professional forecasters, Shintani and Soma (2020) found that the introduction of QQE in March 2013 increased inflation expectations to only around 1%, below the BOJ’s 2% target. Meanwhile, Van den End and Pattipeilohy (2017) investigate the effect of balance sheet policies on inflation expectations in Japan, the United States, the United Kingdom, and the Euro area. Specifically, they investigate whether a shock to the size and composition of central banks’ balance sheets changes inflation expectations. They found some positive (but weak) signalling effects on inflation expectations in Japan. To summarise, the literature suggests that the BOJ’s policy did change market expectations with regard to interest rates, but that these effects on inflation expectations were modest. Why

68  Research handbook of financial markets 40,000 35,000

(billions of JPY)

30,000 25,000 20,000 15,000 10,000 5,000 0

10

11

12

13

14

15

16

17

18

19

20

21

Source:   Bank of Japan.

Figure 3.7  The BOJ’s purchases of ETF this is the case and, linked to this, the formation of inflation expectations in Japan, is an area that continues to warrant research. 3.3.2 Effects of ETF Purchases A unique feature of the BOJ’s balance sheet policy is that the BOJ purchases ETFs and J-REITs. When the BOJ introduced this policy measure on 5 October 2010, it explained that the aim of this policy measure was to reduce various risk premiums.26 Since then, the BOJ has steadily increased the size of this asset purchase programme. As of March 2021, the maximum amounts of annual ETF and J-REIT purchases were set to about 12 trillion and 180 billion yen, respectively. Figure 3.7 shows the volume of the BOJ’s ETF purchases, while Figure 3.8 shows the share of the BOJ’s holding of ETFs out of the total outstanding value of investment trusts. Figure 3.8 shows that the BOJ’s purchases mean that it now holds a substantial share of ETFs available in the market. As mentioned earlier, possible channels through which such ETF purchases have an effect are the scarcity channel and the capital constraints channel. For instance, using a differencein-differences approach, Harada and Okimoto (2021) compare the difference in the performances of stocks that the BOJ purchased and those that it did not. Regarding stocks in the Nikkei 225 as the treatment group and those not in the Nikkei 225 as the control group, they find that the returns of stocks in the Nikkei 225 were significantly higher than those of stocks not in the Nikkei 225. Meanwhile, using an event study approach, Barbon and Gianinazzi (2019) focus on two events in which the BOJ announced major expansions of its ETF purchases, one in 2014 and one in 2016. They show that the BOJ’s ETF purchases had a positive 26 ​https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2010/k101005.pdf

The Bank of Japan’s balance sheet  69 80% 70% 60% 50% 40% 30% 20% 10% 0% 10

11

12

13

14

15

16

17

18

19

20

21

Source:   Bank of Japan, Investment Trust Association, Japan.

Figure 3.8  Share of ETF purchased by the BOJ and persistent impact on stock prices, which is consistent with the prediction of their theoretical portfolio rebalancing model. Based on their findings, the authors argue that the purchases of ETFs tracking the Nikkei 225 generates pricing distortions because the weights in the Nikkei 225 are based on prices rather than on market capitalisation. The capital constraint channel implies that policy effects may be state dependent. For example, the effects may be stronger during market downturns because investors may be subject to funding constraints. Shirota (2018) and Adachi, Hiraki, and Kitamura (2021) find that this is indeed the case. Another study of interest in this context is that by Charoenwong, Morck, and Wiwattanakantang (2021), who not only investigate the effects of the BOJ’s ETF purchases on stock prices but also how the policy changes the behaviour of firms whose stocks are included in the EFTs that the BOJ purchased. They found that while firms hold more cash when the BOJ purchases their shares, they do not increase their investment in tangible assets. Based on this finding, the authors argue that the objective of the ETF purchases, that is, to stimulate firm investment by firms, has not been achieved. Furthermore, closer investigation reveals that the BOJ’s ETF purchases stimulate capital investment by firms with few valuable investment opportunities or with weak corporate governance. They consequently argue that the policy may have harmed economic efficiency. Many studies of the BOJ’s ETF purchases policy assume that the BOJ’s purchases are exogenous. However, investigating the BOJ’s purchasing behaviour using intra-day data, Hattori and Yoshida (2020) show that the BOJ follows a contrarian purchasing rule. Specifically, the BOJ purchases ETFs following a negative stock return over the previous night and during the morning market.

70  Research handbook of financial markets

To summarise, the literature finds that the BOJ’s ETF purchases policy has significantly positive effects on stock prices by reducing risk premiums. This is what the BOJ intends to achieve. However, so far, there is little evidence that this policy has stimulated investment by firms. More research is needed on this issue. 3.3.3 Effects of JGB Purchases Until the Global Financial Crisis, the signalling channel was regarded to be the main transmission channel of the BOJ’s JGB purchase policy. This likely is because of the following two reasons. First, in the early 2000s, the theoretical literature on preferred habitat, which is considered a theoretical platform for the analysis of central bank quantitative easing policies consisting of government bond purchases, was not yet well developed, although it has a long history which goes back to Modigliani and Sutch (1966). The New Keynesian policy analysis of the early 2000s, as highlighted by Eggertsson and Woodford (2003), emphasised the importance of expectations management through forward guidance policies and cast doubt on the efficacy of quantitative easing policies. Second, as highlighted by Ugai (2007), early empirical studies on the first QE from 2000 to 2006 find limited evidence on the “portfolio rebalancing” channel. Note that, during this first QE of 2000–2006, the BOJ expanded its balance sheet by purchasing mainly JGBs of short maturities. When the nominal interest rates on short-term JGBs are zero, money and short-term JGBs are perfect substitutes. Note that, in order for the portfolio rebalancing channel to work, money and assets that a central bank purchases need to be imperfect substitutes to financial institutions. Therefore, it may not be surprising that the BOJ’s short-term JGB purchases did not generate any portfolio rebalancing by financial institutions. However, after the Global Financial Crisis, the situation changed. The BOJ and other central banks have increased purchases of government bonds with longer maturities, and a modern version of the preferred habitat theory has emerged (see, e.g., Andrés, López-Salido, and Nelson, 2004; Chen, Cúrdia, and Ferrero, 2012; and Vayanos and Vila, 2021). As a result, there is a growing body of empirical studies of the BOJ’s JGB purchases policy based on this modern theoretical framework. One of the channels through which the BOJ’s JGB purchase policy has decreased long-term interest rates is by decreasing the net supply of JGBs available to the private sector. For example, to examine how the net supply of JGBs affects the term structure of and risk premiums on long-term JGBs, Fukunaga, Kato, and Koeda (2015) construct a database on the amount of outstanding of JGBs by holder and remaining maturity, and construct measures of net supply. They show that, consistent with the predictions of the preferred habitat model of Vayanos and Vila (2021), a decrease in the net supply of JGBs flattens the yield curve and reduces risk premiums. Meanwhile, compiling the data on the maturity structure of the JGB markets and using a state space representation of a discrete version of the model of Vayanos and Vila (2021), Koeda and Kimura (2022) show that the BOJ’s QQE has pushed down the ten-year bond yield by some 100 basis points. 3.3.4 Effects on Bank Lending Banks are perhaps the entities most affected by the BOJ’s balance sheet policies. Banks hold current account balances at the BOJ, and their current account balances are directly linked to the BOJ’s policy stance. This means that QE is likely to affect bank lending.

The Bank of Japan’s balance sheet  71

To investigate the impact of QE on bank lending, Bowman et al. (2015), focusing on the period from 2000 to 2009, use bank-level data to examine the effect of the BOJ’s injections of liquidity into the interbank market on bank lending. An advantage of using bank-level data is that it enables the authors to use heterogeneity across banks to examine whether the policy effects are stronger for liquidity-constrained banks than unconstrained banks. They found a positive and significant effect of the BOJ’s liquidity injections on bank lending, and that the effect is stronger for liquidity-constrained banks. However, the size of the effect is modest, and the effect had evaporated by 2005. Similar results are obtained by Shioji (2019). Using a much longer sample, from the 1970s to 2017, he is able to compare banks’ behaviour in the period when the nominal interest rate was zero (after 1999) and the period when it was positive (before 1999). He found that the effect of excess reserves on bank loans was stronger in the early 2000s. Further, he did not find any significant effect on bank loans of the QQE in place since 2013. Finally, he found that the policy effects are stronger for banks with weaker balance sheets. Meanwhile, Hosono and Miyakawa (2014) and Ono et al. (2016) use firm-bank matched data to examine the effects of monetary policy on loan supply, enabling them to rigorously identify loan supply and demand by controlling for time-varying unobserved firm heterogeneity through the use of firm-year fixed effects. Employing various dummy variables accounting for changes in monetary policy, Hosono and Miyakawa (2014) examine how bank loan supply reacts to monetary policy changes. On the other hand, Ono et al. (2016) investigate how loan supply reacts to unanticipated declines in long-term interest rates. Both studies find that the BOJ’s monetary easing has had an effect on bank loan supply. Overall, the literature suggests that the BOJ’s JGB purchases had an effect on bank lending. Consistent with the capital constraint channel, the effect is stronger when banks are in distress. A remaining question that warrants research in the future is whether the BOJ’s JGB purchases through their impact on bank loans had any effect on investment by firms. 3.3.5 Effects on the Macroeconomy This section reviews empirical studies investigating the overall macroeconomic effects of the BOJ’s balance sheet policies. A useful review of developments in the Japanese economy three years after the introduction of Abenomics and QQE is provided by Hausman and Wieland (2015). They show that QQE decreased interest rates, weakened the Japanese yen, and raised stock prices. However, inflation expectations did not rise to a level of 2%, the BOJ’s price stability target. While inflation did rise, the real effects, such as the effects on consumption and export, remained weak.27 3.3.5.1 Studies using a time-series approach Many empirical studies on the macroeconomic effects of the BOJ’s balance sheet policy use time-series models. An advantage of using time-series techniques such as VAR models is that this allows researchers to investigate the dynamic effects of policies on variables of interest without having to make too many assumptions.

27 Also see Hausman, Unayama, and Wieland (2021) for a follow-up analysis.

72  Research handbook of financial markets

Perhaps one of the earliest studies using this approach is that by Honda, Kuroki, and Tachibana (2013) who examine the first QE in 2001–2006.28 They use a simple VAR model in which the BOJ’s current account balance is regarded as its policy instrument. Monetary policy shocks are identified by using a standard Choleski decomposition. Studies using a similar approach are those by Kimura and Nakajima (2016), who use banks’ reserves at the BOJ as the policy variable, and Schenkelberg and Watzka (2013), who use a sign restriction to identify a QE shock. All three studies find that the effects of the BOJ’s balance sheet policy on inflation are weak. In contrast, Hanisch (2017) finds that the effects on inflation are strong, while the effects on output are weak and transitory.29 Some studies investigate how different policies at different times had different effects. For instance, using a smooth-transition vector auto-regression model (STVAR) that allows possible regime changes depending on the aggressiveness of monetary easing, Miyao and Okimoto (2020) examine how the BOJ’s monetary policy stance changed over the period from 2001 to 2015. They consider two regimes: a non-aggressive monetary easing regime and an aggressive monetary easing regime. They obtain the following results. The period between August 2003 and March 2006 and the period after May 2011 are classified as an aggressive regime. This implies that not only QQE but also the later period of Comprehensive Monetary Easing is regarded as aggressive. It is interesting that they find the early QE (August 2003–March 2006) should be classified as aggressive, despite the fact that the BOJ at that time mainly bought short-term JGBs. Their impulse response exercises show that the effects of a shock to the monetary base on inflation and output are stronger and more persistent under the aggressive regime than under the non-aggressive regime. Another study using a time-varying parameter VAR model is that by Michaelis and Watzka (2017). They regard the amount of bank reserves on the BOJ’s balance sheet as the policy instrument and regard a shock to those bank reserves as a monetary policy shock. They use a sign restriction to identify monetary policy shocks. Their observation period covers the years from 1996 to 2015, which include the first QE, Comprehensive Easing, and the early phase of QQE. They show that the effects on prices became stronger during later parts of the observation period, while the effects on output appear to have declined. Hayashi and Koeda (2019) and Koeda (2019) construct a regime-switching VAR model that explicitly takes a zero interest rate regime and a normal regime (in which the nominal interest rate is positive) into account. They model a Taylor-type interest rate rule subject to the zero bound. They further assume that in the normal regime, the BOJ’s policy instrument is the call rate. When the nominal interest rate hits the zero bound, the BOJ’s policy instrument is its supply of excess reserves. The conditions for an exit from QE are that the interest rate implied by the Taylor rule is positive and that inflation exceeds a certain threshold. This latter condition captures the BOJ’s forward guidance policy. Hayashi and Koeda (2019) use an observation period that ends in 2012 (i.e., before the introduction of QQE) and Koeda (2019) extends the observation period to 2016. They find that quantitative easing has expansionary effects. Another interesting finding of theirs is that the exit from quantitative easing can be expansionary. Another interesting question that has received research interest is how strong the effects of the BOJ’s balance sheet policies are compared with the effects of the policies of other central 28 This study was originally written in Japanese in 2007. 29 Hanisch (2017)’s observation period covers the years 1985–2014.

The Bank of Japan’s balance sheet  73

banks. Rogers et al. (2014), Gambacorta, Hofmann, and Peersman (2014), and Puonti (2019) conduct cross-country comparisons of central bank balance sheet policies. Since these studies use the same methodology for multiple countries, their analysis is informative regarding the degree of efficacy of the BOJ’s balance sheet policies relative to those of the other central banks. Rogers et al. (2014) examine the spill-over effects of the balance sheet policies on financial markets, focusing on the United States, the United Kingdom, the Euro area, and Japan. They define a monetary policy surprise as an intra-day change in government bond yields at the time of a policy announcement. They then estimate the pass-through of surprise changes in government bond yields to other asset prices and find that, similar to the policies of the other central banks, the BOJ’s policy reduces corporate bond yields. However, the policy effects of the BOJ’s policy seem to be smaller than those of the other central banks (Table 6 of Rogers et al., 2014). They further show that the BOJ’s policy has no significant effect on stock prices, while the other central banks’ policies do have significant effects.30 Meanwhile, Gambacorta, Hofmann, and Peersman (2014) use a country-panel VAR model to investigate the effects of central bank balance sheet policies for Canada, the Euro area, Japan, Norway, Sweden, Switzerland, the United Kingdom, and the United States for the period 2008–2011. The policy instrument is assumed to be the size of central bank assets. The results suggest that the responses of prices and output in Japan are weaker than in the other countries, This may reflect the fact that their model does not include exchange rates. Japan experienced a sharp appreciation of the yen caused by the sharp drop in the interestrate differentials (see Figure 3.1), putting downward pressure on output and prices, which is not captured in the authors’ VAR model. Along similar lines, Puonti (2019), focusing on the period 1995–2010, also finds that the effects of the BOJ’s policy are weaker than those of the FRB and the ECB. Overall, this strand of the literature finds that the BOJ’s policies were effective in decreasing nominal interest rates and in increasing output, but the effects on inflation are weak. Many of the studies in this strand of the literature regard innovations to a variable representing the size of the BOJ’s balance sheet (such as current account balances at the BOJ) as a policy shock. However, this is not always without problems, since the identification of policy shock of course depends on the identification assumptions. More recently, studies have used highfrequency financial data to identify monetary policy shocks. Studies employing this approach for Japan include Arai (2017) and Kubota and Shintani (2022). 3.3.5.2 Model-based approaches The previous subsection reviewed studies that use time-series approaches to examine the overall effects of the BOJ’s balance sheet policies. A possible disadvantage of such approaches is that they do not necessarily distinguish the transmission mechanisms at work. For example, the increase in the BOJ’s balance sheet has been driven by purchases of a wide range of assets, such as long-term JGBs, ETFs, and REITs, and lending against corporate debt. From a theoretical point of view, purchases of different kinds of assets have different effects on the economy. While VAR models can capture the effects of overall policies, they may not 30 However, when they restricted the observation period to the years 2013 and 2014, they found that the BOJ’s expansionary monetary policy significantly increased stock prices and the effect was economically large.

74  Research handbook of financial markets

necessarily be able to separately examine the effects of different policy measures. An alternative therefore is to construct economic models that can be estimated to assess the effects of a particular policy measure. An example is the study by Sudo and Tanaka (2021), who use a dynamic stochastic general equilibrium (DSGE) model to investigate the effects of long-term JGBs purchases on the macroeconomy. Following Chen, Cúrdia, and Ferrero (2012) and Arce et al. (2020), their model assumes a preferred habitat and market segmentation. Employing Japanese data to estimate the model parameters using Bayesian techniques, Sudo and Tanaka (2021) find that, as of the end of 2017, the BOJ’s JGB purchase policy reduced long-term bond yields by 50–100 basis points. Their impulse response exercises show that the peak response of inflation to a temporary increase in the BOJ’s JGB purchases equivalent to 10% of GDP is about ten basis points when the short-term nominal rate is kept constant. Output also increases by about 45 basis points. The size of the effects is larger than what Chen, Cúrdia, and Ferrero (2012) find for the US economy.31 Finally, we review the BOJ’s own assessment of its policy using its macroeconomic model (Q-JEM).32 Using Q-JEM, Kan, Kishaba, and Tsuruga (2016) examine the effects of QQE during 2013–2016, while Kawamoto et al. (2021) is a follow-up analysis for the period 2013–2020. The analysis in these two studies to some extent is a compromise between a reduced-form approach and a model-based approach. Q-JEM does not explicitly model the transmission mechanisms of unconventional monetary policy, such as the portfolio rebalancing channel, through which the BOJ’s asset purchases affect financial variables and economic activity. Kan, Kishaba, and Tsuruga (2016) and Kawamoto et  al. (2021) therefore first estimate the hypothetical paths of financial variables had QQE not been implemented and compute the corresponding counterfactual paths of variables such as output and inflation. Kawamoto et al. (2021), for example, estimate that had QQE not been implemented, long-term JGB yields would have been 100 basis points higher on average during the period. They also estimate that inflation expectations would have been some 50 basis points lower. These estimates imply that QQE has pushed down long-term real interest rates by some 150 basis points. They then plug these hypothetical financial variables into Q-JEM and compute the counterfactual path of the Japanese economy had QQE not been implemented. The authors conclude that QQE has pushed up the level of real GDP by 0.9–1.3 percentage points and the CPI inflation rate (excluding fresh food and energy) by 0.6–0.7 percentage points. The studies reviewed in this section find strong evidence suggesting that the BOJ’s policy had effects on the Japanese economy. An advantage of model-based approaches is that the transmission mechanism through which a particular policy affects the economy is explicitly considered. However, the policy assessments of such approaches crucially depend on the underlying assumptions, which may sometimes be very strong, making it difficult to judge how robust the conclusions of such assessments are. Nevertheless, constructing rich structural models that capture the key elements of an economy and incorporate the transmission

31 Based on a smaller estimate of the share of restricted agents and a smaller elasticity of the term premium with respect to the quantity of long-term debt, Chen, Cúrdia, and Ferrero (2012) conclude that the effects of long-term asset purchases are modest. 32 Q-JEM stands for Quarterly Japanese Economic Model. It is a large-scale semi-structural model of the Japanese economy. The description of the 2019 version of the model and its replication files can be found here: https://www​.boj​.or​.jp​/en​/research​/wps​_ rev​/wps​_2019​/wp19e07​.htm/ https://www​.boj​.or​.jp​/en ​/mopo​/outlook ​/index​.htm

The Bank of Japan’s balance sheet  75

mechanisms of balance sheet policies represents one important avenue for the analysis of such policies, including the balance sheet policies of the BOJ. 3.3.6 Effects of the BOJ’s Policy on Its Balance Sheet The BOJ’s balance sheet policy consists of adjustments to both its assets and liabilities. Yet, to date, there has been little academic research on how the BOJ’s current and past policies shape the future evolution of its balance sheet. A notable exception is a study by Fujiki and Tomura (2017), who conduct a simulation of the evolution of the BOJ’s balance sheet after a hypothetical exit from QQE using an approach developed by Hall and Reis (2015) and Carpenter et al. (2015). In their study written in 2016, they assume that the BOJ achieves its inflation target of 2% in March 2018 and gradually increases its policy rate to 2.75%. They further assume that the demand for currency shrinks in response to the increase in nominal interest rates. Their simulation shows that the Bank would run accounting losses for more than 15 years and the maximum annual losses amount to 1.4% of GDP. The authors highlight that the losses arise for the following reasons The first reason is the maturity transformation on the BOJ’s balance sheet. While the interest rates on long-term JGBs purchased as part of QQE are fixed at low levels, the interest paid on current account balances at the BOJ will rise as short-term interest rates rise, resulting in losses for the BOJ that will continue until current account balances have declined sufficiently with the redemption of JGBs. The second reason for losses is seigniorage losses, the size of which depends on the elasticity of the demand for banknotes with respect to nominal interest rates. Since banknotes do not bear interest, supplying banknotes generates seigniorage revenue in the form of savings in interest payments to the public. If the demand for cash shrinks significantly upon exit from QQE, seigniorage revenue shrinks, contributing to the BOJ’s losses. The authors furthermore highlight that the duration of QQE is crucial for the estimated level of losses. In the July 2021 issue of “Outlook for Economic Activity and Prices”, the BOJ forecasted that the CPI inflation rate would only rise to 0.9–1.1% in fiscal year 2023, which is below the Bank’s inflation target of 2%. The duration of QQE will be much longer than Fujiki and Tomura (2017) assumed, and the BOJ’s losses can be expected to be higher as a result.

3.4 CONCLUSION The BOJ has actively engaged in balance sheet policies over the past two decades. There is a growing body of research on the BOJ’s balance sheet policies, which was reviewed in Section 3.3. The review indicated that many studies find significant effects of the policies on financial variables such as long-term interest rates, stock prices, and exchange rates. However, studies in this strand of the literature tend to find that the effects on macroeconomic variables such as output and inflation have remained modest over the past two decades. As a result, inflation remains below the target of 2% as of the time of writing (early 2022). Meanwhile, a number of issues related to the BOJ’s balance sheet policy have not yet been examined in detail. For example, although there are a growing number of empirical studies on the BOJ’s purchases of ETFs, there is almost no academic research on the BOJ’s purchases of J-REITs. Further, research into the fiscal implications of the BOJ’s balance sheet policy is still in its infancy. Finally, understanding why inflation in Japan has been so low for so long despite the BOJ’s aggressive monetary easing still remains an important research topic.

76  Research handbook of financial markets

REFERENCES Adachi, K., Hiraki, K., & Kitamura, T. (2021). Supplementary paper series for the “assessment” (1): The effects of the Bank of Japan’s ETF purchases on risk premia in the stock markets. Bank of Japan Working Paper Series No.21-E-3:1–35. Andrés, J., López-Salido, J. D., & Nelson, E. (2004). Tobin’s imperfect asset substitution in optimizing general equilibrium. Journal of Money, Credit, and Banking, 36(4), 665–690. Arai, N. (2017). The effects of monetary policy announcements at the zero lower bound. International Journal of Central Banking, 13(2), 159–196. Arce, Ó., Nuño, G., Thaler, D., & Thomas, C. (2020). A large central bank balance sheet? Floor vs corridor systems in a new Keynesian environment. Journal of Monetary Economics, 114, 350–367. Bank of Japan. (2016). Comprehensive assessment:developments in economic activity and prices as well as policy effects since the introduction of quantitative and qualitative monetary easing (QQE) the background. September 21st, 2016. Bank of Japan. (2021). Assessment for further effective and sustainable monetary easing. March 19th, 2021. Barbon, A., & Gianinazzi, V. (2019). Quantitative easing and equity prices: Evidence from the ETF program of the Bank of Japan. Review of Asset Pricing Studies, 9(2), 210–255. Bowman, D., Cai, F., Davies, S., & Kamin, S. (2015). Quantitative easing and bank lending: Evidence from Japan. Journal of International Money and Finance, 57, 15–30. Carpenter, S., Ihrig, J., Klee, E., Quinn, D., & Boote, A. (2015). The Federal Reserve’s balance sheet and earnings: A primer and projections. International Journal of Central Banking, 11(2), 237–283. Charoenwong, B., Morck, R., & Wiwattanakantang, Y. (2021). Bank of Japan equity purchases: The (Non-)effects of extreme quantitative easing. Review of Finance, 25(3), 713–743. Chen, H., Cúrdia, V., & Ferrero, A. (2012). The macroeconomic effects of large-scale asset purchase programmes. Economic Journal, 122(564), 289–315. Diamond, J., Watanabe, K., & Watanabe, T. (2020). The formation of consumer inflation expectations: New evidence from Japan’s deflation experience. International Economic Review, 61(1), 241–281. De Michelis, A., & Iacoviello, M. 2016. Raising an inflation target: The Japanese experience with abenomics. European Economic Review, 88, 67–87. Eggertsson, G. B., & Woodford, M. (2003). The zero bound on interest rates and optimal monetary policy. Brookings Papers on Economic Activity, 1(1), 139–233. Fujiki, H., & Tomura, H. (2017). Fiscal cost to exit quantitative easing: The case of Japan. Japan and the World Economy, 42, 1–11. Fujiwara, I., Nakazono, Y., & Ueda, K. (2015). Policy regime change against chronic deflation? Policy option under a long-term liquidity trap. Journal of the Japanese and International Economies, 37, 59–81. Fujiwara, S., Iwasaki, Y., Muto, I., Nishizaki, K., & Sudo, N. (2016). Supplementary paper series for the “comprehensive assessment” (2): Developments in the natural rate of interest in Japan. Bank of Japan Review Series No.2016-E-12:1–8. Fukuda, S. (2015). Abenomics: Why was it so successful in changing market expectations? Journal of the Japanese and International Economies, 37, 1–20. Fukunaga, I., Kato, N., & Koeda, J. (2015). Maturity structure and supply factors in Japanese government bond markets. Monetary and Economic Studies, 33, 45–96. Gambacorta, L., Hofmann, B., & Peersman, G. (2014). The effectiveness of unconventional monetary policy at the zero lower bound: A cross-country analysis. Journal of Money, Credit, and Banking, 46(4), 615–642. Hall, R. E., & Reis, R. (2015). Maintaining central-bank financial stability under new-style central banking. NBER Working Paper Series (21173). Hanisch, M. (2017). The effectiveness of conventional and unconventional monetary policy: Evidence from a structural dynamic factor model for Japan. Journal of International Money and Finance, 70, 110–134. Harada, K., & Okimoto, T. (2021). The BOJ’s ETF purchases and its effects on Nikkei 225 stocks. International Review of Financial Analysis, 77, 1–11.

The Bank of Japan’s balance sheet  77

Hattori, M., Kong, S., Packer, F., & Sekine, T. (2021). The impact of regime change on the influence of the Central Bank’s inflation forecasts: Evidence from Japan’s shift to inflation targeting. International Journal of Central Banking, 17(4), 257–290. Hattori, T., & Yoshida, J. (2020). Bank of Japan as a contrarian stock investor: Large-scale ETF purchases. CREPE Discussion Paper Series No.70:1–12. Hausman, J. K., Unayama, T., & Wieland, J. F. (2021). Chap. 6. Abenomics, monetary policy, and consumption. In T. Hoshi & P. Y. Lipscy (Eds.), The political economy of the Abe government and Abenomics reforms (pp. 139–169). Cambridge: Cambridge University Press. Hausman, J. K., & Wieland, J. F. (2015). Overcoming the lost decades? Abenomics after three years. Brookings Papers on Economic Activity, 2(2), 385–413. Hayashi, F., & Koeda, J. (2019). Exiting from quantitative easing. Quantitative Economics, 10(3), 1069–1107. Honda, Y., Kuroki, Y., & Tachibana, M. (2013). An injection of base money at zero interest rates: Empirical evidence from the Japanese experience 2001–2006. Japanese Journal of Monetary and Financial Economics, 1(1), 1–24. Hoshi, T., & Kashyap, A. K. (2001). Corporate financing and governance in Japan: The road to the future. Cambridge, MA: MIT Press. Hosono, K., & Miyakawa, D. (2014). Business cycles, monetary policy, and bank lending: Identifying the bank balance sheet channel with firm-bank match-level loan data. RIETI Discussion Paper Series 14-E-026:1–34. Ito, T. (2014). We are all QE-Sians now. IMES Discussion Paper Series 2014-E-5:1–51. Jung, T., Teranishi, Y., & Watanabe, T. (2005). Optimal monetary policy at the zero-interest-rate bound. Journal of Money, Credit, and Banking, 37(5), 813–835. Kan, K., Kishaba, Y., & Tsuruga, T. (2016). Supplementary paper series for the “comprehensive assessment” (3): Policy effects since the introduction of quantitative and qualitative monetary easing (QQE) – Assessment based on the Bank of Japan’s large-scale macroeconomic model (Q-JEM). Bank of Japan Working Paper Series No.16-E-15:1–15. Kawamoto, T., & Nakahama, M. (2017). Supplementary paper series for the “comprehensive assessment” (4): Why did the BOJ not achieve the 2 percent inflation target with a time horizon of about two years?-Examination by time series analysis. Bank of Japan Working Paper Series No.17-E-10:1–14. Kawamoto, T., Nakajima, J., & Mikami, T. (2021a). Supplementary paper series for the “assessment” (3): Inflation-overshooting commitment: An analysis using a macroeconomic model. Bank of Japan Working Paper Series No.21-E-9:1–27. Kawamoto, T., Nakazawa, T., Kishaba, Y., Matsumura, K., & Nakajima, J. (2021b). Supplementary paper series for the “assessment” (2): Estimating effects of expansionary monetary policy since the introduction of quantitative and qualitative monetary easing (QQE) using the macroeconomic model (Q-JEM). Bank of Japan Working Paper Series No.21-E-4:1–18. Kimura, T., & Nakajima, J. (2016). Identifying conventional and unconventional monetary policy shocks: A latent threshold approach. B.E. Journal of Macroeconomics, 16(1), 277–300. Koeda, J. (2019). Macroeconomic effects of quantitative and qualitative monetary easing measures. Journal of the Japanese and International Economies, 52, 121–141. Koeda, J., & Kimura, Y. (2022). Government debt maturity and the term structure in Japan. SSRN 4015576: 1–36. Krishnamurthy, A., & Vissing-Jorgensen, A. (2012). The aggregate demand for treasury debt. Journal of Political Economy, 120(2), 233–267. Krishnamurthy, A., & Vissing-Jorgensen, A. (2013). The ins and outs of LSAPs. In Kansas city federal reserve symposium on global dimensions of unconventional monetary policy. Federal Reserve Bank of Kansas City, 57–111. Kubota, H., & Shintani, M. (2022). High-frequency identification of monetary policy shocks in Japan. Japanese Economic Review, 73(3), 483–513. Michaelis, H., & Watzka, S. (2017). Are there differences in the effectiveness of quantitative easing at the zero-lower-bound in Japan over time? Journal of International Money and Finance, 70, 204–233. Miyao, R., & Okimoto, T. (2020). Regime shifts in the effects of Japan’s unconventional monetary policies. Manchester School, 88(6), 749–772.

78  Research handbook of financial markets

Modigliani, F., & Sutch, R. (1966). Innovations in interest rate policy. American Economic Review, 56(1/2), 178–197. Nishino, K., Yamamoto, H., Kitahara, J., & Nagahata, T. (2016). Supplementary paper series for the “comprehensive assessment” (1): Developments in inflation expectations over the three years since the introduction of quantitative and qualitative monetary easing (QQE). Bank of Japan Review Series No.2016-E-13:1–7. Oda, N., & Ueda, K. (2007). The effects of the Bank of Japan’s zero interest rate commitment and quantitative monetary easing on the yield curve: A macro-finance approach. The Japanese Economic Review, 58(3), 303–328. Ono, A., Aoki, K., Nishioka, S., Shintani, K., & Yasui, Y. (2016). Long-term interest rates and bank loan supply: Evidence from firm-bank loan-level data. Bank of Japan Working Paper Series No.16-E-2:1–41. Puonti, P. (2019). Data-driven structural BVAR analysis of unconventional monetary policy. Journal of Macroeconomics, 61, 103131. Rogers, J. H., Scotti, C., Wright, J. H., Ellison, M., & Kara, H. (2014). Evaluating asset-market effects of unconventional monetary policy: A multi-country review. Economic Policy, 29(80), 749–799. Romer, C. D. (2013). It takes a regime shift: Recent developments in Japanese monetary policy through the lens of the Great Depression. In J. A. Parker & M. Woodford (Eds.), NBER Macroeconomics Annual 2013, chap. 6 (Vol. 28, pp. 383–400). University of Chicago Press. Schenkelberg, H., & Watzka, S. (2013). Real effects of quantitative easing at the zero lower bound: Structural VAR-based evidence from Japan. Journal of International Money and Finance, 33, 327–357. Shintani, M., & Soma, N. (2020). The effects of QQE on long-run inflation expectations in Japan. CARF Working Paper No.CARF-F-494:1–42. Shioji, E. (2019). Quantitative ‘flooding’ and bank lending: Evidence from 18 years of near-zero interest rate. Journal of the Japanese and International Economies, 52, 107–120. Shioji, E. (2021). Does the Japanese financial market believe in fiscal sustainability?: Analysis of the market for the JGB futures options. Public Policy Review, 17(2), 1–33. Shiratsuka, S. (2010). Size and composition of the Central Bank balance sheet: Revisiting Japan’s experience of the quantitative easing policy. Monetary and Economic Studies, 28, 79–105. Shirota, T. (2018). Evaluating the unconventional monetary policy in stock markets: A semi-parametric approach. Hokkaido University Discussion Paper, Series A No.2018-322:1–22. Sudo, N., & Tanaka, M. (2021). Quantifying stock and flow effects of QE. Journal of Money, Credit, and Banking, 53(7), 1719–1755. Ueda, K. (2002). The transmission mechanism of monetary policy near zero interest rates: The Japanese experience, 1998–2000. In L. Mahadeva & P. Sinclair (Eds.), Monetary transmission in diverse economies (pp. 127–136). Cambridge: Cambridge University Press. Ueda, K. (2012). Japan’s deflation and the Bank of Japan’s experience with nontraditional monetary policy. Journal of Money, Credit, and Banking, 44, 175–190. Ugai, H. (2007). Effects of the quantitative easing policy: A survey of empirical analyses. Monetary and Economic Studies, 25(1), 1–48. Van den End, J. W., & Pattipeilohy, C. (2017). Central Bank balance sheet policies and inflation expectations. Open Economies Review, 28(3), 499–522. Vayanos, D., & Vila, J.-L. (2021). A preferred–habitat model of the term structure of interest rates. Econometrica, 89(1), 77–112.

4. Central bank lending1 Brian Madigan and William Nelson

Central banks generally have responsibility for promoting an economy’s macroeconomic goals, such as price stability or low and stable inflation, high employment, and maximum sustainable economic growth. To do so, they typically use a number of tools such as open market operations, reserve requirements, adjustments to policy interest rates, and lending. This chapter focuses on the lending function of central banks, often referred to as the discount window. Reflecting the professional backgrounds of the authors, the article focuses on the U.S. experience.1

4.1 HISTORY AND RATIONALE Lending has always been regarded as a core function of central banks, but the nature of central bank lending and the role that it plays have evolved considerably over time and varied across economies. Early in the history of some of the first central banks, including Sveriges Riksbank and the Bank of England, which were established in the 17th century, lending to the sovereign to finance wars was of central importance (Bordo and Siklos, 2018, p. 26). In the first decades of the United States, enhancing the U.S. Treasury’s access to funds was part of the motivation for considering the establishment of the First and Second Banks of the United States.2 When the Federal Reserve (Fed) was eventually established in 1913, however, its central purpose was to reduce the frequency and severity of severe financial panics and resulting business cycles that the U.S. had experienced over the latter half of the nineteenth century by lending to banks rather than to the government.3 For a more recent example, the European Central Bank (ECB) was established for a number of purposes, including providing a common currency, conducting monetary policy for the Eurozone (specifically, to maintain price stability), promoting the smooth operation of payment systems, and contributing to prudential supervision of credit institutions and the stability of the financial system. The centerpiece of its operating framework is lending to banks.4 To this day, most central bank lending, although not in the United States, consists of ongoing lending to financial institutions rather than contingency funding. Central banks generally 1 The authors thank Ulrich Bindseil, Mark Carlson, William B. English, Refet S. Gürkaynak, and Jonathan Wright for helpful comments and suggestions and Jose Maria U. Tapia for excellent research assistance. 2 See Hill (2009a, 2009b). 3 The official title of the Federal Reserve Act of 1913 was “An Act To provide for the establishment of Federal reserve banks, to furnish an elastic currency, to afford means of rediscounting commercial paper, to establish a more effective supervision of banking in the United States, and for other purposes.” 4 Issing (2018, p. 471) states that “The main refinancing operation is the principal means whereby the ECB supplies the banking system with central bank money.” 79

80  Research handbook of financial markets

have assets that are at least equal to their monetary liabilities, and in many cases, such as the ECB, those assets consist in large part of loans to banks. In many economies, the principal practical alternative to an asset structure consisting primarily of loans to banks is a large portfolio of government securities purchased through open market operations.5 The Bundesbank judged that lending by rediscounting bank loans was preferable because it avoided the risks to price stability associated with the central bank in effect financing the government by buying government securities in the open market. Unlike securities repurchase operations, the Bundesbank has hitherto engaged in large-scale “outright'” open market operations in long-term securities only sporadically, and only in the secondary market, confining its activities to purchases and sales of public debt securities. Open market operations of this kind in debt securities have not yet resulted in the Bundesbank holding major amounts of public debt securities for any length of time. If the central bank had built up a sizeable portfolio of long-term government paper, this might easily have given rise to the suspicion that its primary aim was to facilitate the financing of public sector budget deficits. Moreover, the Bundesank has always applied very strict standards to intervention of this nature, since any heavier commitments by the central bank in the debt securities market might arouse the false impression that the movement of interest rates in that market was largely the outcome of open market operations, and hence was the Bundesbank’s direct responsibility. In actual fact, the long-term interest rate is determined predominantly by market factors such as interest rate movements abroad and inflation expectations, which the Bundesbank can influence only indirectly. (Deutsche Bundesbank, 1995, pp. 114–115)

As described in the following, central bank lending is also an important source of backup liquidity. In ordinary circumstances, because interbank payments typically clear through accounts at the central bank, central banks can provide loans late in the day, after interbank markets have essentially shut down, to banks with unexpected funding shortfalls. Sometimes those shortfalls are the result of idiosyncratic developments – the bank experienced an unexpected outflow of funds, or an expected payment did not arrive. Sometimes, the shortfalls arise because the market-wide demand for reserve balances that day ended up exceeding supply. In the latter case, central bank lending serves an important monetary policy purpose by injecting reserve balances when needed, thus limiting upward spikes in interbank lending rates. In rare cases, central bank lending facilitates an orderly failure by enabling a bank to continue to operate for a short period until it can be safely shut down or sold, often over a weekend. Central banks at times provide contingency funding to both banks and nonbanks in periods of widespread market disruption, such as during the Global Financial Crisis (GFC) and the Covid-19 pandemic, often in circumstances where market liquidity has evaporated. In the 19th century, Walter Bagehot (1896), in his widely cited Lombard Street, discussed the operation of banks, central banks, and the money market, their inherent instability (even of central banks in systems of commodity-backed money), and the rationale and appropriate principles for central bank lending. Bagehot argued The holders of the cash reserve must be ready not only to keep it for their own liabilities, but to advance it most freely for the liabilities of others. They must lend to merchants, to minor bankers, to “this man and that man,” whenever the security is good. In wild periods of alarm, one failure makes 5 In some economies, especially smaller economies, foreign-currency assets account for a significant proportion of total central bank assets. For example, the assets of the Swiss National Bank “currently consist almost exclusively of currency reserves” (Swiss National Bank, 2022).

Central bank lending 

81

many, and the best way to prevent the derivative failures is to arrest the primary failure which causes them. The way in which the panic of 1825 was stopped by advancing money has been described in so broad and graphic a way that the passage has become classical. “‘We lent it,” said Mr. Harman, on behalf of the Bank of England, “by every possible means and in modes we had never adopted before … in short by every possible means consistent with the safety of the Bank, and we were not on some occasions over nice.” (Bagehot, 1896, pp. 53–54)

Notably, Bagehot’s advice did not restrict lending to banks or even financial institutions. Tucker (2009, p. 3) summarized Bagehot’s central thesis as follows: “Bagehot’s famous dictum, in Lombard Street, was that, to avert panic, central banks should lend early and freely (ie without limit), to solvent firms, against good collateral, and at ‘high rates.’” In the modern era, the rationale for central banks standing ready to provide emergency liquidity was set out most starkly by Diamond and Dybvig (1983). The Diamond-Dybvig model can be viewed as a multiple-equilibrium game-theoretic model.6 In the model, banks fund illiquid assets with liquid (demand) deposits, potentially contributing to productive capital formation. If the behavior of depositors is such that they withdraw cash only as needed for routine liquidity demands, the productive equilibrium is achieved. Depositors recognize that, because of the illiquidity of the assets, banks will not be able to repay depositors fully if they all demand cash at the same time. As a result, another equilibrium exists: a run on the bank. This equilibrium is socially suboptimal because it can inhibit capital formation. The suboptimal equilibrium can be eliminated, or made less likely, if deposit insurance or a central bank lending facility is in place, assuaging depositors’ concerns about the bank’s ability to repay their deposits. Other researchers and policymakers (for example, Goodfriend & Lacker, 1999), however, have emphasized the inefficiencies that result from the moral hazard that can accompany central bank lending, a topic discussed in the following. Relatedly, as we discuss later, because a central bank is never liquidity constrained, when a solvent bank is at risk of a liquidity default, it is welfare-enhancing for the central bank to lend to the illiquid bank. Not only does such lending reduce the unnecessary social costs of bankruptcies but it also prevents the externalities associated with the bank disposing of its assets at firesale prices. Moreover, because the central bank is a risk-free counterparty, banks will be willing to provide a large amount of collateral to the central bank to secure a loan without fear that the collateral will not be returned. A substantial excess of estimated collateral value over loan proceeds is necessary when the financial condition of the bank or the value of the bank’s collateral is highly uncertain. Central bank lending is also an important tool for addressing market-wide strains. Often, the first response of the Federal Reserve to a substantial financial shock is to issue a public statement underscoring that the discount window is open and operating, and, for more severe and long-lived shocks, to ease terms on discount window loans. For example, the Federal Reserve issued such a statement after the 1987 stock market crash, after the September 11, 2001, terrorist attacks, during the GFC, and during the Covid-19 financial crisis. Such announcements and actions are intended in large part to encourage banks to continue to provide credit to others with confidence that backup funding is available if needed. Most notoriously, central banks may lend to individual institutions to prevent their immediate failure if a sudden failure would result in unacceptably high economic costs. Such 6 “Diamond–Dybvig model,” Wikipedia, 13-Dec-2022. [Online]. Available: https://en.wikipedia.org/ wiki/Diamond%E2%80%93Dybvig_model.

82  Research handbook of financial markets

institutions are referred to as “too big to fail” (TBTF). In March 2008, for example, the Federal Reserve lent on very short notice to facilitate the acquisition of Bear Stearns, a large broker-dealer, by JP Morgan Chase, a large bank holding company. Although the Fed was ultimately repaid in full, the market value of the collateral backing the loan fell at times below the outstanding balance.7 In some sense, “too big to fail” is a misnomer. The Fed has often lent to a small depository institution for a few days to prevent its disorderly failure until the Federal Deposit Insurance Corporation (FDIC) can complete a resolution transaction. As in the case of a failing large bank, such lending is based on the view that the economic costs of a disorderly failure are higher than the moral hazard costs (see the following) or financial risks of lending. The Federal Reserve’s main standing lending facility has traditionally been called the discount window. The Federal Reserve Act authorizes Federal Reserve Banks, among other things, to purchase “notes, drafts, and bills of exchange arising from actual commercial transactions” and, historically, lending typically took this form. To allow such credit to earn implicit interest, the amount paid in such transactions was less than the face value of the asset – in other words, the asset was purchased at a discount. This genesis of the terms “discount window” and “discount rate” had nothing to do with the fact that, for many years, the Federal Reserve set the discount rate below market interest rates such as the federal funds rate. (The federal funds rate is the interest rate on overnight uncollateralized loans between banks. It is a market-determined rate but is heavily influenced by Federal Reserve monetary policy decisions.) For decades, however, most Federal Reserve lending has taken the form of advances against collateral rather than discounting, partly for operational convenience.

4.2 ROLE IN MONETARY POLICY When the Federal Reserve System was created, it was envisioned that lending to banks would be the primary tool to achieve the objectives established by Congress (Eichengreen, 2018, p. 74). Over time, open market operations became a more actively used monetary policy tool than the discount window, but the discount window and discount rate both remained important aspects of the monetary policy framework. As discussed in the following, the Federal Reserve periodically revisited the objectives, issues, and rules surrounding the discount window. Figure 4.1 shows, from the Federal Reserve’s founding into 2021, the percentage of its assets that were loans to depository institutions and, separately, loans to other entities. Through the early 1960s, the discount rate was generally set above the federal funds rate. Anbil and Carlson (2019) note that the discount rate in the 1950s was typically set 50 basis points above market rates, and borrowing at the window served as a barometer for the Federal Reserve for tightness in the market for reserves. However, in response to increasing inflationary pressure in the mid-1960s, the Federal Reserve at times tightened monetary policy by reducing the supply of reserves to the banking system through open market operations, rather than by raising the more visible and politically sensitive discount rate (Nelson, 2000). The discount rate subsequently remained below the federal funds rate until 2003, when the Federal Reserve reformed several aspects of the discount window, as will be discussed. 7 For example, on June 30, 2010, the outstanding principal amount of the loan extended by the New York Fed was $28.8 billion and the fair value of the collateral was $28.5 billion. See the Fed’s H.4.1 statistical release “Factors Affecting Reserve Balances,” table 4, July 1, 2010. https://www​.federalreserve​.gov​/releases​/ h41​/20100701/

Central bank lending 

83

Source:   Federal Reserve H.4.1 release.

Figure 4.1   Federal Reserve loans as a percent of Federal Reserve assets During most of the 1970s, monetary policy was conducted primarily by using open market operations to keep the federal funds rate around a level that the Federal Open Market Committee (FOMC) judged was consistent with its statutory objectives and, more specifically, its monetary growth objectives.8 Changes in the discount rate played a secondary role, partly because the amount of discount window borrowing was a relatively small component of the Fed’s balance sheet. At the end of 1978, total loans extended by Reserve Banks were around $1 billion while their holdings of U.S. government securities, federal agency obligations, and bankers’ acceptances were about $119 billion (Board of Governors of the Federal Reserve System, 1978, p. 390, Table 1). In October 1979, the FOMC shifted its approach to monetary policy. High and rising inflation rates, and a tendency for money growth to exceed the FOMC’s targets, led to a view that more stringent control of money supply growth was necessary. The FOMC adopted an approach that involved short-run objectives for nonborrowed reserves (total banking system

8 The FOMC did not set formal numerical targets for money growth throughout the entire decade of the 1970s, but even early in the period the FOMC clearly had objectives for monetary growth. For example, the Record of Policy Actions for the FOMC meeting held on July 27, 1971, p. 7, stated that “The Committee decided that the achievement of more moderate growth in the monetary aggregates over the months ahead remained the appropriate objective of System open market operations.” By the mid-1970s, the FOMC was explicitly setting ranges of tolerance for money growth. The Federal Reserve Reform Act of 1977 required the Federal Reserve to report to the Congress its “objectives and plans with respect to the ranges of growth or diminution of monetary and credit aggregates.”

84  Research handbook of financial markets

reserve balances at the Federal Reserve plus vault cash applied to reserve requirements minus borrowings from the discount window) that were judged to be consistent with its money growth targets. If money growth exceeded its targets, required reserves increased as a result, boosting the demand for reserves. If the Federal Reserve maintained its target for nonborrowed reserves in the face of the increase in required reserves, and if banks’ aggregate demand for excess reserves was roughly unchanged, the banking system would need to borrow the additional reserves from the discount window. Because banks were reluctant to borrow from the discount window for reasons discussed in the following, they bid more aggressively for funds in the money markets, pushing up the federal funds rate and, through arbitrage, other money market rates relative to the discount rate. Over time the higher money market rates would reduce money demand both by increasing the opportunity cost of holding money and by slowing the growth of nominal income, spending, and wealth, thus bringing money growth back toward its target. In this period, the relationship between actual and targeted money supply growth played an important role in determining the aggregate amount of borrowings from the discount window. During the 1970s and early 1980s, deposit deregulation and financial innovation greatly expanded the range of available deposits and closely related financial instruments such as money market mutual funds. As a result, household and business cash management practices changed significantly, altering the characteristics of money demand. By October 1982, bivariate relationships between money supply growth, on the one hand, and macroeconomic measures such as nominal GDP growth and inflation, on the other, appeared to have loosened to such an extent that the FOMC decided to discontinue its procedure of controlling money supply growth through nonborrowed reserves targeting. In its place, the FOMC gradually reverted to close control over the federal funds rate, albeit for a number of years through an indirect method and initially with a wide (four percentage point) range for the federal funds rate. In particular, rather than focusing on control of nonborrowed reserves in the short run, the FOMC employed a procedure involving an assumption for borrowed reserves – the amount of discount window borrowing – judged consistent with a desired range for the federal funds rate and a given discount rate. For many years, in its operating directive to the Federal Reserve Bank of New York, the FOMC used the phrase “degree of pressure on reserve positions” as a synonym for the borrowing assumption (Lindsey, 2003, Chapter III). The calibration of the post-1982 approach implicitly depended on an assessment of the degree of banks’ reluctance to borrow from the discount window, which subsequent experience demonstrated could change appreciably. When the FOMC reached a policy decision to tighten monetary policy, it could conduct open market operations in such a way as to reduce the supply of nonborrowed reserves relative to the estimated demand for total reserves, forcing the banking system to borrow incremental amounts from the discount window. Given banks’ reluctance to borrow from the window, they would first bid up the federal funds rate in an attempt to borrow the reserves from other banks. Alternatively, if the Board of Governors of the Federal Reserve System (Board of Governors or Board) increased the discount rate in line with the increase in the federal funds rate range, the borrowing assumption could remain about unchanged. To the extent that banks’ reluctance to borrow from the window was reasonably stable, this reluctance induced a predictable relationship between the federal funds rate, the discount rate, and aggregate discount window borrowings, a relationship that was known as the borrowing function. For most of the period from late 1982 until around the turn of the decade, monetary policy implementation needed to take careful account of the borrowing

Central bank lending 

85

function, and shifts in the function needed to be offset through adjustments to nonborrowed reserves for a given discount rate and a desired range for the federal funds rate. In 1984, Continental Illinois National Bank and Trust Company experienced serious financial difficulties, in part as a result of a deterioration in the performance of loans to energysector firms that it held in its portfolio (Board of Governors, 1984). The publicity surrounding these developments led to a depositor run on Continental Illinois. As a result, the bank needed to borrow substantial amounts from the discount window. With anxieties about the health of large banks also heightened by Latin American debt problems, other banks apparently became concerned that they would be perceived as also in trouble. Consequently the aggregate reluctance to borrow from the discount window evidently increased, with the federal funds rate rising despite an unchanged borrowing assumption (which excluded borrowings by Continental) and an unchanged discount rate. From the point of view of monetary policy implementation, these developments were somewhat fortuitous, as they coincided with a need to tighten monetary policy in response to the evolving macroeconomic outlook (Lindsey, 2003, Chapter III). A somewhat similar development occurred in 1988–1989, a period of increasing stress in the banking system that particularly reflected difficulties of the thrift industry as well as troubled commercial real estate loans, but in this case related macroeconomic developments eventually called for an easing of monetary policy (Board of Governors staff, Monetary Policy Alternatives, December 9, 1988; February 3, 1989; March 24, 1989; February 2, 1990; March 23, 1990). Throughout the period from late 1982 to the early 1990s, the Federal Reserve largely implemented monetary policy using borrowed reserve targeting, together with signaling the stance of policy through open market operations, to achieve a target range for the federal funds rate that was not contemporaneously announced. By the early 1990s, partly as a result of increased instability of the borrowing function, the FOMC focused more closely and directly on the federal funds rate as its operating target. Over the next few years, the FOMC moved toward and ultimately adopted a procedure of immediately announcing changes in the target federal funds rate following FOMC decisions, and borrowed reserve targeting was discontinued.9 For the next decade, from the early 1990s through the early 2000s, the structure of the discount window was largely unchanged, and it generally played a secondary role in monetary policy formulation and implementation. An important exception was in September 2001, when severe disruptions to financial infrastructure resulting from the terrorist attacks on September 11 caused a large but temporary increase in borrowing from the discount window. This experience in which the discount window played an important role in an emergency situation was part of the motivation for the Federal Reserve’s restructuring of the discount window in 2003, discussed in the following.

9 Following the stock market crash of October 19, 1987, the FOMC immediately shifted away from its borrowed reserve approach and focused on controlling the federal funds rate more closely, but gradually resumed targeting borrowed reserves over the next six months. Again, in November 1989, the FOMC discontinued use of a borrowing-target approach to monetary policy following misinterpretations by market analysts that an open market operation signaled a change in the stance of monetary policy. The FOMC subsequently focused closely on achieving a federal funds rate target. Nonetheless, the FOMC modified its communication practices gradually over the next decade. In December 1999, the FOMC decided to begin to announce the federal funds rate target after each meeting (Lindsey 2003, Chapters IV–V).

86  Research handbook of financial markets

4.3 THE FEDERAL RESERVE’S REGULAR DISCOUNT WINDOW PROGRAMS The Federal Reserve Act assigns responsibility for discount window policy to the Board of Governors of the Federal Reserve System. The Board sets discount window policy formally through its Regulation A. For example, in that regulation, the Board establishes the System’s regular discount window programs, specifies discount rates, and sets forth procedures for complying with statutory requirements. Discount rates are established by Federal Reserve Banks “subject to the review and determination of the Board of Governors” (Federal Reserve Act Section 13(8)). Although Reserve Banks’ Boards of Directors play a formal role in the discount-rate process, this statutory provision gives the Board of Governors effective control over such rates.10 Notably, the FOMC, which includes all seven members of the Board as well as a rotating set of five of the twelve Federal Reserve Bank presidents and is the Federal Reserve’s primary monetary policymaking body, does not control the discount rate. Lending by the Federal Reserve takes two broad forms: discount window lending to banks and other depository institutions and lending to nonbanks (individuals, partnerships, and corporations that are not depository institutions) in unusual and exigent circumstances. Lending to nonbanks is discussed in the next section. Amounts outstanding under the main lending programs as well as emergency credit extended to nonbanks can be seen in Figure 4.2. Regular discount window lending consists of loans to depository institutions (DIs) – commercial banks, thrifts, and credit unions – extended under Section 10B of the Federal Reserve Act.11 Since 2003, when the Federal Reserve moved the discount rate from a belowmarket to an above-market configuration, there have been three types of regular discount window lending – primary, secondary, and seasonal credit. As noted previously, in normal times the Federal Reserve does not seek to be a regular source of ongoing funding to depository institutions through the discount window. The main discount window lending facility is primary credit. Primary credit is provided to financially sound banks at an above-market interest rate. Assessments of financial soundness are based largely on regulatory capital and examiner ratings. Primary credit is extended on a no-questions-asked basis and is normally provided at an overnight maturity. The primary credit rate is often referred to as “the discount rate” even though the separate secondary credit and seasonal credit interest rates are also discount rates. In normal times, a Federal Reserve Bank may provide primary credit to a small bank that encounters a temporary funding need or to a large bank if transitory tightness in money markets drives the federal funds rate above the primary credit rate. That latter situation has become rare under the Federal Reserve’s current monetary policy implementation framework under which the Fed, through open market operations, oversupplies reserve balances, pushing the federal funds rate down to or below the interest rate it pays on reserve balances (deposits of DIs at Federal Reserve Banks). In crisis 10 In a letter dated December 9, 1919, to Secretary of the Treasury Carter Glass, Acting Attorney General Alex. C. King wrote “I am of the opinion that the Federal Reserve Board has the right under the powers conferred by the Federal Reserve Act, to determine what rates of discount should be charged from time to time by a Federal reserve bank, and under their powers of review and supervision, to require such rates to be put into effect by such bank.” 11 Nonmember banks, thrift institutions, and credit unions were granted access to the discount window by the Monetary Control Act of 1980 (MCA). The MCA also subjected these institutions to reserve requirements and gave them access to Federal Reserve services.

Central bank lending 

87

Figure 4.2   Federal Reserve lending times, such as during the GFC and the Covid crisis of 2020–2021, many banks may borrow to replace funding temporarily unavailable because financial markets are disrupted. As noted, often the Federal Reserve’s first response to a crisis is to ease terms on discount window lending. For the first several years of the primary credit program, the primary credit rate was set 100 basis points above the target federal funds rate. The spread was reduced during the GFC and after the crisis was set at 50 basis points above the top of the FOMC’s target range for the federal funds rate until the Covid crisis began. If a DI that needs funding does not meet the financial soundness criteria for primary credit, the Federal Reserve Bank may provide secondary credit. Note that this article classifies secondary credit as regular discount lending, because, like primary credit, it can be extended to depository institutions and is authorized under the same section of the Federal Reserve Act, although it could also reasonably be classified as emergency credit. Secondary credit may be provided to a DI until it can return to market funding or as a bridge to resolution by the FDIC. Since the secondary credit program was established, the secondary credit rate has been set 50 basis points above the primary credit rate. Secondary credit is closely administered by the lending Reserve Bank – it is not provided on a no-questions-asked basis. The Federal Deposit Insurance Corporation Improvement Act (1991) established guidelines on the ability of the Fed to lend to a troubled institution. Under these guidelines, the Fed should not provide credit to an undercapitalized bank for more than 60 days in any 120-day period nor to a critically undercapitalized bank beyond the five days after it became critically undercapitalized. If the Fed lends beyond those periods, it is liable for the minimum of any resulting increased resolution costs of the FDIC or the interest it earned on the loan. While the pecuniary costs of such lending might be low, the Federal Reserve would probably be unwilling to lend beyond these Congressionally imposed guidelines.

88  Research handbook of financial markets

Seasonal credit is provided to small DIs that have a seasonal funding need such as banks in farming or resort communities. Seasonal credit is intended to make it unnecessary for such banks to hold large stocks of liquid assets to liquidate during peak lending periods, thereby allowing banks to provide more credit to the real economy. With market sources of funding available to banks of all sizes (at a minimum from correspondent banks), it has long been unclear that seasonal credit remains necessary, but political support for seasonal credit is strong. Seasonal credit is far and away the most commonly used type of discount window lending in terms of the number of loans, but because of the modest size of the borrowing institutions and their seasonal funding needs, the aggregate amount of seasonal credit outstanding tends to be small. The seasonal credit rate is set as the two-week average of the federal funds rate and the rate on three-month CDs and is reset every two weeks. Because the seasonal credit rate closely tracks market interest rates and because aggregate seasonal borrowing is low, seasonal credit plays no role in monetary policy. Operationally, discount window loans are extended by Federal Reserve Banks and are reflected as assets on the Banks’ balance sheets. Under Section 16 of the Federal Reserve Act, loans serve as collateral against Federal Reserve note liabilities (currency) in the same way as do securities acquired through open market operations and other Federal Reserve assets. In general, DIs borrow from the Federal Reserve Bank for the Federal Reserve District in which they are headquartered. An institution that wants to be prepared to borrow must have on file with its Reserve Bank the necessary authorizing resolutions and agreements. All discount window lending is fully collateralized. The Federal Reserve accepts a broad range of collateral, including loans as well as securities. Haircuts, which range from a few percent for Treasury securities to half or more for loans, are applied to pledged assets to determine lendable value. They are based on the Fed’s judgment about the loss it could sustain if the borrower defaulted and it needed to liquidate collateral, and so depend on both the liquidity and the volatility of the asset (Board of Governors, 2002, pp. 1–37). During the GFC, despite the rise in asset price volatility and the decline in liquidity, the Fed did not change its haircuts, seeking to avoid a procyclical tightening in its lending policy. DIs tend to maintain a pool of discount window collateral at the Federal Reserve rather than just pledging collateral when they need to borrow. As of February 26, 2020, $1.6 trillion in collateral (lendable value) was pledged to the Fed (Board of Governors, 2020, p. 13, footnote to Table 5). The Federal Reserve has not had a loss on a discount window loan since the 1940s. Under the Federal Reserve’s current monetary policy implementation framework in which reserve balances are oversupplied, discount window credit is very low in normal times because the banking system never ends the day short of reserve balances. Seasonal credit has varied each year between zero and about $300 million, while primary and secondary credit have remained near zero. However, during the 1970s, 80s, and 90s, banks borrowed much more frequently from the discount window, and the amount outstanding typically varied in the $1 billion to $4 billion range. As noted, discount window lending also rose sharply during the global financial crisis and the Covid-19 crisis, rising to $112 billion and $50 billion, respectively.

4.4 LENDING TO NONBANKS IN UNUSUAL AND EXIGENT CIRCUMSTANCES In unusual and exigent circumstances, the Federal Reserve can provide credit to non-DIs under the authority granted in Section 13(3) of the Federal Reserve Act (FRA). Such lending

Central bank lending 

89

requires the affirmative vote of at least five members of the Board of Governors and that the lending Federal Reserve Bank “obtain evidence that such participant in any program or facility with broad-based eligibility is unable to secure adequate credit accommodations from other banking institutions.” The Fed also may lend to nonbanks against Treasury or federal agency securities under Section 13(13) of the FRA. Although through the early 2000s, the Fed had not lent under these authorities since the Great Depression, it used its Section 13(3) authority extensively in the GFC and the Covid crisis. During the GFC, the Fed provided credit through broad-based facilities accessible by classes of institutions (for example, money market mutual funds) and to individual financial institutions, such as American International Group (AIG). Lending under Section 13(13) did not occur in the GFC or Covid crisis, although the Board authorized its use if needed to stabilize Fannie Mae and Freddie Mac during the GFC. Following the GFC, Congress required the Fed to henceforth extend credit under Section 13(3) only through broad-based facilities, only to solvent borrowers, and only with the approval of the Secretary of the Treasury. The law also required the Federal Reserve to disclose to Congress within seven days certain information on such lending and to publish information on such loans one year after the emergency credit facility is closed. The broad-based facilities that the Fed operated during the GFC and then again during the Covid crisis took two general forms: facilities designed to lend to entities that needed funding, and facilities designed to lend to intermediaries so that they would, in turn, lend to entities that needed funding. As an example of the first form, in the two crises, the Fed extended credit directly, through the Primary Dealer Credit Facility (PDCF), to primary government securities dealers that needed funds.12 For the second, in the two crises, the Fed also lent to financial institutions that did not themselves need the money, in order to finance the institutions’ purchases of commercial paper from money market mutual funds, many of which were in desperate need of liquidity. The Fed operated five broad-based facilities for lending to nonbanks during the GFC. Two programs were opened in March 2008 amidst increasing strains in money markets and the fallout from the financial difficulties of Bear Stearns: the Term Securities Lending Facility (TSLF) and the PDCF. Both facilities provided support to primary government securities dealers. The TSLF allowed dealers to exchange less liquid securities for Treasury securities at rates determined in periodic auctions. The PDCF lent to dealers using repo transactions against public and private securities at the primary credit rate. Three facilities were created in the wake of the Lehman bankruptcy in the fall of 2008. Under the Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility (AMLF), the Fed lent to commercial banks, bank holding companies, and primary dealers to fund their purchases of asset-backed commercial paper (ABCP) from money funds to get liquidity to the funds, which were experiencing heavy redemptions. The Commercial Paper Funding Facility (CPFF) purchased newly issued top-tier commercial paper from corporations that were experiencing difficulty rolling over maturing paper. And the Term Asset-Backed Securities Loan Facility (TALF) lent to investors to finance purchases of new and existing asset-backed securities (ABS) and commercial mortgage-backed securities. The Fed encountered various challenges in establishing lending programs for nonbanks in the GFC. For example, immediately after the Reserve Primary Fund broke the buck in 12 Primary dealers are entities that are designated by the Federal Reserve Bank of New York as its main counterparties for open market operations.

90  Research handbook of financial markets

September 2008, the Fed established the AMLF to provide liquidity to money funds that were experiencing massive withdrawals. Because the Fed could not purchase the ABCP that the money funds owned, it lent to banks and bank holding companies to finance purchases of the ABCP by those entities. However, in that period of extraordinarily heightened uncertainty, no bank would risk exposure to the possibility the ABCP would default. To make the program work, the Fed lent the full face value of the ABCP on a nonrecourse basis, meaning the borrowers faced no downside risk – if the ABCP went down in value, the borrowers could just hand the Fed the ABCP in lieu of repaying the loan. Another example is provided by developments in the ABS market, which was a key market for funding a broad range of economic activity. In the years before the GFC, many different types of ABS were issued, with the specifics of the securities very much dependent on institutional aspects of the underlying economic activity and financing instruments. Assessing the needs for financial support, whether the Federal Reserve could provide such support, and the appropriate role of the Treasury was very time-consuming, implying that programs for various types of ABS could only be announced over the course of several months. Subsequent analysis has generally found that the Fed’s broad-based emergency lending during the GFC was effective. Wiggins and Metrick (2016), Hrung and Seligman (2011), and Carlson and Macchiavelli (2020) found that the TSLF was helpful in relieving liquidity strains at primary dealers and pressures in repo markets. Duygan-Bump et al. (2013) found outflows were lower at money funds that held AMLF-eligible collateral and that spreads on AMLFeligible ABCP narrowed by more than yields on similar but ineligible securities. Similarly, Adrian et al. (2011) found that the CPFF improved conditions in the commercial paper (CP) market, and that spreads on CPFF-eligible CP narrowed significantly more than spreads on CP that the facility did not accept. A number of studies found that the TALF improved conditions in ABS markets. Ashcraft et al. (2010), Campbell et al. (2011), and Ashcraft et al. (2012) all found that the facility contributed to a sharp decline in yields on newly issued ABS and existing CMBS. In response to the financial market turmoil and severe negative macroeconomic shock caused by the onset of the Covid-19 pandemic in early 2020, the Fed immediately took many of the same lending actions it had taken in response to the GFC as well as some additional actions. As in 2007, the Fed reduced the interest rate and lengthened the term of primary credit loans. The Fed also opened nearly all of the broad-based facilities for nonbanks it had opened in 2008, including the PDCF, CPFF, and TALF. The Fed also opened a facility that was similar in structure to the AMLF but named the Money Market Mutual Fund Liquidity Facility (MMLF) because it was not limited to ABCP as collateral, instead accepting unsecured CP as well as other securities owned by money market mutual funds. In addition, the Fed opened the Municipal Liquidity Facility to purchase newly issued securities from states and local governments, the Main Street Lending Program to purchase participations in bank loans to medium-sized firms and nonprofits, the Primary and Secondary Market Corporate Credit Facilities to purchase corporate bonds directly from issuers or on the secondary markets, and the Paycheck Protection Program Liquidity Facility to lend to banks to finance their guaranteed loans to small businesses. Research on the effectiveness of the Fed’s lending in response to the Covid crisis has just begun. Clarida et al. (2021) provide a comprehensive overview of the Federal Reserve’s actions in response to the Covid-19 crisis and conclude that, despite modest uptake, the emergency lending facilities were important backstops that supported the continued flow of credit to the

Central bank lending 

91

real economy. Anbil et al. (2021) find that the Federal Reserve’s Paycheck Protection Program Liquidity Facility boosted lending under the Small Business Administration’s Paycheck Protection Program directly, and particularly in the case of large banks, indirectly by providing a conditional, backup source of liquidity. Lopez and Spiegel (2020) find that the direct effect was mainly on lending by small and medium-sized banks. Li et al. (2021) find that the provision of funding by the MMLF stabilized prime money funds.

4.5 REPO LENDING In general, central bank lending and open market operations are distinct policy instruments. However, repurchase agreements – an important policy implementation instrument for many central banks – have characteristics of both types of instruments. Under a repurchase agreement (repo), a security is purchased by the central bank in exchange for cash, and the transaction is reversed at a specified price the next day or at some other specified date in the future.13 At a central bank, repos are generally used to manage aggregate banking system reserves and achieve the short-run monetary policy operating objectives of the central bank. In the United States, the Federal Reserve allocates repo transactions using an auction mechanism. The Fed conducts repos under the authority provided by Section 14 of the Federal Reserve Act to buy and sell government securities in the open market. In fundamental economic terms, however, the contractual pairing of the purchase and sale under repurchase agreements as well as other terms of the transactions mean that repos closely resemble collateralized loans. In July 2021, the Fed established the Standing Repo Facility (SRF). Under the facility, each day the Fed auctions a large amount of repo financing with a minimum bid rate that is well above normal market interest rates. The Fed established the facility so that primary dealers and large commercial banks would be willing to lend into the repo market when repo rates were elevated with confidence that they would always be able to finance their lending at the SRF if necessary. The Fed also created the facility in part in hopes that there would be no stigma associated with borrowing from it, unlike the severe stigma associated with borrowing from the discount window, the subject of the next section.

4.6 RELUCTANCE TO BORROW, STIGMA, AND “LENDER OF LAST RESORT” This chapter intentionally avoids the use of the term “lender of last resort” (LOLR) as a synonym for central bank lending. The term gives a misleading impression of the circumstances under which most central bank lending takes place. In particular, when used as a synonym for central bank lending, it may give the sense that such lending typically takes place in a failing bank situation, when no private-sector counterparty is willing to provide funding to the bank. In reality, most central bank lending takes place in a relatively routine setting. As described earlier, looking across central banks and across time, most central bank loans, on an ongoing basis, are extended to depository institutions as a central bank asset maintained to provide the 13 Central banks also conduct reverse repurchase agreements under which they initially sell securities and subsequently repurchase the securities at a prespecified price.

92  Research handbook of financial markets

public currency and DIs required reserves. Even when credit is provided on a temporary basis, reasons for borrowing are usually mundane. For example, a bank may experience an unexpected funding need for any of a range of unremarkable reasons that relate directly to standard banking functions, such as a jump in loan demand by commercial borrowers or residential mortgage borrowers, or the withdrawal of a large deposit by a municipality to repay a maturing bond. Banks may also experience the need to borrow as a result of market-wide developments that may be quite temporary and unrelated to banks’ individual financial strength, such as when the money market tightens unexpectedly as a result of factors that affect aggregate availability of bank reserves – for instance, a sharp increase in the U.S. Treasury’s deposit balance at Federal Reserve Banks resulting from stronger-than-anticipated inflows of tax revenues. Banks may also borrow when the payment system is physically disrupted; borrowing rose sharply, for example, in response to a major blackout in the northeastern and midwestern United States in August 2003. Despite the routine nature of most lending, the Federal Reserve has always struggled with a stigma associated with borrowing from the discount window. The Fed provides discount window credit to serve two purposes: (1) to help maintain the intended stance of monetary policy by injecting reserve balances when money markets are tight; and (2) to provide a backup source of funding for individual banks. To accomplish the first objective, banks need to be willing to borrow from the discount window freely whenever interbank interest rates rise above the discount rate. However, a backup funding source is intended to be used sparingly. For example, decades ago W. Randolph Burgess wrote “Just as in the old days the bank which borrowed largely and continuously from its correspondents was looked upon with suspicion, so today there exists generally a feeling against large and continuous borrowing from a Federal Reserve Bank” (Burgess, 1946, p. 219).14 Indeed, the stigma associated with borrowing from the discount window is something that the Fed sought to cultivate and utilize for much of its history. A 1954 System report on the discount window (Federal Reserve System Committee on the Discount and Discount Rate Mechanisms, 1954, p. 11) notes that it was possible by the mid-Thirties to speak of an established tradition against member bank reliance on the discount facility as a supplement to its resources … Future discount policy … should build on the tradition as a keystone.

A 1971 report on the discount window (Board of Governors, 1971) describes how the Fed administered discount window borrowing to support the “tradition against borrowing.” Initial requests for credit were almost always granted, but Beyond this initial accommodation, the administrative process can, for purposes of analysis, be broken down into three consecutive stages: (1) surveillance of the borrowing bank; (2) a decision with respect to the “appropriateness” of the borrowing; and (3) in cases where an “inappropriate” decision is reached, the undertaking of “administrative counseling” or “discipline” aimed at securing repayment and “educating” the borrower in the appropriate use of the discount window. (Board of Governors, 1971, p. 42) 14 When The Reserve Banks and the Money Markets was first published in 1927, Burgess was an official at the Federal Reserve Bank of New York.

Central bank lending 

93

In a major reversal, in 1971 the Board revised significantly how it administered the discount window in an attempt to increase banks’ willingness to use the discount window. After engaging in an extensive review, a System committee concluded that A more liberal and convenient mechanism should enable individual member banks to adjust to changes in fund availability in a more orderly fashion and, in so doing, should lessen some of the causes of instability in financial markets without hampering overall monetary control. (Board of Governors, 1971, Volume 1, p. 3)

The committee had judged that a revision to Regulation A was necessary because Failure of the Federal Reserve to communicate clearly, consistently, and unambiguously with member banks regarding the availability of discount-window accommodation has caused many of these banks to view this as an uncertain source of credit. In addition, occasional Federal Reserve counsel as to what would be regarded as appropriate adjustments for borrowing banks has led many banks to regard the window as having too great a potential for interfering with bank management decisions. As a result, many banks having temporary needs for funds often make adjustments by more costly, less efficient avenues than that afforded through the discount window, sometimes to the detriment of adequate credit availability for their local communities. (Board of Governors, 1971, Volume 1, p. 9)

The committee also noted that discount window administration relying on “bank reluctance to borrow” had resulted in differences across the Reserve Bank Districts in how credit was provided (Board of Governors, 1971, p. 9). To make discount window administration more transparent and consistent across Districts, the Fed provided banks a “basic borrowing privilege … that formalized limited and temporary access to the window” (Board of Governors, 1971, Volume 1, p. 10). Still, some limits were necessary because, as discussed earlier, since the mid-1960s, the discount rate had been set below market rates. As noted, in 2003, the Board again revised Regulation A, importantly in an effort to reduce stigma. The Federal Reserve shifted the discount rate from a below-market rate to an abovemarket rate, added financial soundness criteria, and began providing funds on a no-questionsasked basis rather than requiring banks to attempt unsuccessfully to get funding elsewhere before borrowing from the Fed (Madigan & Nelson, 2002, pp. 313–319). The realignment of the discount rate relative to the federal funds rate removed the purely financial incentive to borrow and thus obviated most administrative review of borrowing requests, allowing credit generally to be extended on a no-questions-asked basis. The financial soundness criteria also contributed to the reduced need for administrative review. In addition, the soundness criteria should have helped reduce discount window stigma, as counterparties and market analysts would be expected to recognize that banks thought to have obtained primary credit at the discount window must have been judged in financially sound condition by the Federal Reserve. Prior to both the 1971 and 2003 revisions to Regulation A, the System reviewed the lending policies of other major central banks with particular interest in the extent of a reluctance to borrow. The 1971 report states that Only a few of the central banks surveyed administer the discount window on the assumption that commercial banks are reluctant to borrow … In most countries surveyed, commercial banks tend to

94  Research handbook of financial markets

regard access to the discount window as a right rather than a privilege (within applicable limitations such as quotas) even though the central bank normally has discretionary authority. (Board of Governors, 1971, pp. 181–182)

In discussing the proposed 2003 revision, Madigan and Nelson (2002, p. 317) state Discussions with staff members of these central banks indicate that Lombard facilities that are a part of monetary policy operations similar to those of the Federal Reserve have been very effective in setting an upper limit on market rates. (The effectiveness of lending facilities at other central banks in capping rates suggests that they were largely free of stigma.)

As described earlier, likely in part because of somewhat inconsistent messaging from the Federal Reserve on borrowing, banks are reluctant to borrow from the discount window, and banks with good access to money markets will only borrow if interest rates rise well above the discount rate. Before 2003, the Federal Reserve limited banks’ borrowings, reviewed the circumstances of borrowing requests, and potentially even referred instances of borrowing to bank examiners for further review. Historically, the discount window function was typically part of a Reserve Bank’s supervision department. In general, bank supervisors discouraged reliance on the discount window. Some limitation on borrowing by the Federal Reserve was necessary simply because the below-market discount rate gave banks a financial incentive to borrow that had to be offset through moral suasion, but the administration and supervision reduced banks’ willingness to borrow even in circumstances in which borrowing would have had valuable stabilizing effects on financial markets. Furthermore, even though borrowing is contemporaneously kept secret by the Federal Reserve, bankers are concerned that borrowing will be discerned by market participants or analysts, partly on the basis of their bidding behavior in money markets. Funding managers worry that borrowing will be seen by market participants, analysts, and examiners as an indication that their bank is suffering from liquidity problems, or internally by bank management as an indication of an error in the bank’s funding strategy or operations. These factors all appear to contribute to the reluctance of banks to borrow from the discount window. Although previously, details about discount window borrowing by individual depository institutions were kept secret permanently by the Fed, post-GFC legislation required it to publish, after a two-year lag, details of even routine lending to depository institutions, including borrower identity, collateral securing the loan, and the interest rate. As a result, banks may also be reluctant to borrow because they are concerned about the eventual publication of their borrowing. Historically, stigma has worsened when the banking system is weak. For example, Madigan and Nelson (2002, p. 315) report that the average amount of borrowing on days when the federal funds rate rose 25–200 basis points above the FOMC’s target rate fell from over $400 billion in 1989 to just over $40 billion in 1993, a period during which (as noted previously) several banks experienced financial difficulties, and then rose steadily over the 1990s as the banking system strengthened. Despite the 2003 overhaul of the discount window, stigma intensified sharply during the GFC, impairing the ability of the Federal Reserve to respond effectively to the reduced supply of term interbank lending using the primary credit facility alone. In response, the Fed lowered the primary credit rate relative to the federal funds rate and allowed longerterm borrowings. In addition, the Fed created a new discount window program – the Term Auction Facility (TAF) – in December 2007. The TAF provided depository institutions term

Central bank lending 

95

discount window loans at interest rates determined in regular auctions. The TAF was largely stigma-free and successfully supplemented bank term funding (Carlson & Rose, 2017). Nevertheless, because of its auction structure, the TAF could not accomplish two critical and related functions of central bank standing lending facilities that were necessary to increase banks’ willingness to lend term funds to each other: because of the several-day lag between auction and settlement, the TAF was not able to provide funds to a bank that needed funds quickly, and it was unable to reassure banks that they would be able to obtain funding on demand in the future if they needed it. These features were probably central to the absence of stigma from the TAF, as the delay in and uncertainty of obtaining TAF funding would suggest to outside observers that the borrower was not in extremis. This experience thus illustrates the difficulty of designing central bank lending facilities that simultaneously meet all objectives. At times, concerns about the stigma associated with borrowing from the central bank are dismissed on the view that if the institution needs the funds it will eventually borrow. However, credit facilities that lend to intermediaries that do not themselves need the funds, but rather so that the intermediaries will on-lend the funds or, as discussed above in the case of the AMLF, use the funds to finance asset purchases, will not function if the intermediaries are unwilling to borrow. Furthermore, such a view ignores the fact that, especially if multiple institutions are afflicted by the same stigma, their bidding for funds will tend to affect money market conditions in a way that undermines the achievement of the central bank’s objectives when lending. Even though at the time of this writing banks in the United States are strong, stigma associated with borrowing at the discount window is extremely elevated. In response to a Fed survey in March 2021 (Senior Financial Officer Survey), less than 10 percent of respondents indicated that they would borrow from the discount window “if other funding sources are too expensive” and over 60 percent indicated that they would borrow “only if other funding sources became less available due to market-wide conditions.” Notably, over 10 percent of respondents listed no conditions under which they would borrow. The continued stigma is likely due largely to the negative responses of the public, media, and Congress to Fed lending during the GFC, but also in part to the Fed’s continued mixed messages on borrowing. For example, even though the Fed encouraged banks to make use of primary credit and TAF credit during the GFC, in May 2019 the Fed proposed that foreign banking organizations be subject to toughened liquidity regulations, citing as a reason that “analysis using Federal Reserve Board data on Term Auction Facility usage in 2008 and 2009 finds that approximately 40 percent of foreign banking organizations borrowed from the facility during the financial crisis” (Board of Governors, 2019).

4.7 MORAL HAZARD AND TOO BIG TO FAIL One of the costs of a central bank providing backup liquidity support may be moral hazard. Specifically, if a bank facing liquidity difficulties is able to borrow from the central bank to repay its short-term creditors, its short-term creditors may not fully price in the risk of default. In reaction, banks may take on more liquidity risk by funding themselves to an even greater extent with short-term debt and investing more in illiquid assets. As a result, the likelihood that banks will need central bank liquidity support could rise further.

96  Research handbook of financial markets

Liquidity regulations are one way to control such moral hazard. Stein (2013) argues that by requiring banks to hold low-yielding assets, liquidity regulations provide an incentive for banks to increase their liquidity and reduce their need for emergency liquidity support. In contrast, an approach that was popular in Europe before the GFC was “constructive ambiguity” (see Freixas et al., 2000). Under constructive ambiguity, the central bank is studiously vague about whether it will provide liquidity support or not, with the intention of limiting moral hazard. Constructive ambiguity did not survive the GFC. Domanski et al. (2014) observed that once the financial system became sufficiently fragile, it was not credible that the central bank would withhold liquidity support. They concluded that “[a]gainst this backdrop, it is questionable whether constructive ambiguity is a viable policy option in the future.” There is now a widely held view that a central bank should rapidly provide generous and reliable liquidity support in response to a crisis. Indeed, one criticism of the Fed’s response to the GFC has been that it ramped up too slowly (see Bernanke et al., 2020, pp. 13–14). By contrast, as noted earlier, when the Covid crisis hit the Fed immediately rolled out its entire GFC arsenal as well as some additional artillery. The constructive disambiguity approach to central bank liquidity support largely sets aside moral hazard concerns or judges them to be addressed more appropriately through regulation and supervision. Going further, Carlson et al. (2015) question whether liquidity regulations are necessary if liquidity support is reliably available from the central bank, on the view that a central bank can essentially eliminate liquidity risk. They conclude that it is nevertheless important that a central bank refrain from lending to an insolvent firm and that liquidity reserves are valuable because they provide the central bank time to ensure that the potential borrower is solvent before lending. Moreover, as reflected in the findings reported above that the announcement of liquidity support can reduce illiquidity premiums materially, the value of a borrower’s assets and therefore the solvency of the borrower can be a function of expectations about whether the central bank will provide liquidity support. A related concern is that, by requiring collateral, a central bank loan to an institution that ends up failing could shift risk onto uncollateralized lenders and, if the borrower is an insured depository institution and the loss is large enough, onto the FDIC insurance fund. Such an outcome would also reduce the default risk facing short-term creditors of the bank at the expense of other creditors (or the FDIC). However, that possibility exists for any new collateralized borrowing, not just loans from the central bank, and the problem to be fixed, if there is one, is that other creditors and the FDIC are not taking proper steps to protect themselves. It is worth noting that identical risk-shifting occurs if the bank sells the assets that it is required to hold to meet its liquidity requirements (“high-quality liquid assets” or HQLA); good assets are sold to pay off short-term creditors, reducing the pool of creditors to absorb losses as well as the assets available to offset the loss. This conceptual discussion elides some practical but important reasons why central banks should, in reality, require sufficient collateral that any loan is essentially riskless based on the collateral alone. First, the central bank can only act as a reliable source of backup liquidity if it can specify clearly and in advance when it would lend. The pernicious dynamic in which a liquidity shock becomes a liquidity crisis because banks stop lending to each other out of fear of themselves coming up short is broken if banks can rely on liquidity from the central bank. Given that the financial condition of the borrower can be uncertain, especially at a time of market stress, the central bank in such circumstances will need to rely almost entirely on collateral, as opposed to the financial strength of the borrower, as the source of repayment. Because of the adverse selection problems discussed in Stiglitz and Weiss (1981), a central

Central bank lending 

97

bank cannot use the discount rate alone as a means to ensure that it is adequately compensated for risk, especially when it is lending to a bank under funding pressure. The central bank can only be a transparent, reliable backup source of liquidity when it understands and is compensated for the risk it is taking.15 In practice, when lending with incomplete information about the condition of the borrower and the value of the borrower’s assets, certainty about repayment can only be achieved by excess collateral sufficient to reduce risk to near zero. Another reason why central bank lending should be collateralized is that it may be appropriate for a central bank, as an independent government agency, to only make very low-risk loans. Restricting the central bank to essentially riskless loans reduces the scope for the central bank to engage in credit allocation; if the central bank makes risky loans, it is difficult to know the correct (unsubsidized) credit risk premiums. As argued by Tucker (2018), this difficulty makes such credit decisions inherently fiscal and so should usually be made by elected officials. As a practical compromise, he suggests that central banks be granted a “Fiscal Carve-Out” that establishes a “zone of constrained discretion, leaving them in control of their balance sheets within those bounds.” The carve-out needs to cover: the kind of assets [the central bank] can lend against; the kind of assets it can buy, in what circumstances, and for which of its purposes; whether those operations are even subject to consultation with the executive government or legislature; and how losses will be covered by the fiscal authority and communicated to the executive government and legislature. (Tucker, 2018, p. 488)

Moreover, some losses would be suffered even if loans were priced so that expected losses were equal to zero; it could be argued that the potential for realizing such losses makes extending them essentially fiscal decisions. Sufficient collateral with conservative haircuts, enough to reduce the credit risk of the loan essentially to zero, would eliminate this concern. Alternatively, the Treasury or finance ministry could provide sufficient capital or indemnification to make the loan essentially riskless to the central bank, but such arrangements raise complicated and difficult questions about central bank independence. For one, it is unclear why the Treasury should not finance the entire loan other than to obscure the amount of taxpayer funds being committed. That said, it should be recognized that lending against abundant collateral has costs as well as benefits. For example, by taking collateral the central bank may shift risks onto other creditors and the deposit insurance fund, which may be costly because it may be difficult for such entities to control fully for such risks. For another, by consuming collateral, the central bank loan may accelerate the demise of the borrower by reducing its capacity to borrow in the marketplace and providing an incentive for its other creditors to withdraw rather than have risk shifted onto them.16 While the foregoing concludes that abundant HQLA is desirable to enable an orderly failure and that abundant collateral pledged to the central bank is desirable to secure a reliable liquidity backstop, the reasoning does not suggest that both are necessary. As envisioned by King 15 Although Stiglitz and Weiss show that collateral does not solve the adverse selection problem, they do so making additional assumptions that are irrelevant for the situation being considered such as that wealthier individuals are less risk averse. 16 For example, the Fed created and authorized but never opened a credit facility in 2008 that would have lent directly to money market mutual funds. The funds asked the Fed not to open the facility because they feared that its existence would accelerate investor withdrawals.

98  Research handbook of financial markets

(2016), if a bank has collateral pledged to the central bank sufficient to secure funding to cover all contingencies, the bank is not exposed to liquidity risk. In the extreme, if, as King suggests, banks are required to pledge enough collateral to cover all of their liabilities, the central bank’s collateral haircuts amount to capital requirements (because capital equals assets minus liabilities) and the risk-shifting problem goes away entirely.

4.8 CONCLUDING COMMENTS Lending to banks has long been recognized as a key tool, and perhaps even the defining feature, of a central bank. However, central banks have diverged notably in their philosophies and frameworks for lending. Some central banks extend credit essentially continuously to banks, and such loans account for a large fraction of their assets in routine circumstances. Other central banks lend primarily on a contingency basis, where the contingency could be so routine as an unexpected loan request by a commercial customer or so significant as a systemic liquidity crisis. Central banks that lend primarily on a contingency basis typically have only a low level of loans outstanding and, because of the balance-sheet identity, must largely rely on other assets, such as government securities, as counterparts to their currency and reserves liabilities. For such central banks, open market operations tend to displace lending as the primary tool of monetary policy. In economies with highly developed financial markets, open market operations can be an efficient tool for controlling short-term interest rates, the standard operating target for monetary policy, as well as for influencing financial market conditions more broadly. Nonetheless, an important open question is whether the combination of heavy reliance on open market operations and a secondary role for central bank lending undermines the effectiveness of the discount window at crucial junctures. Indeed, some central banks have struggled with the ineffectiveness of the discount window in forestalling or subduing financial crises. Banks may be reluctant to borrow from the central bank and rightly perceive a stigma of borrowing. Banks’ reluctance to borrow may have a variety of causes, some of which should be under the control of the central bank – for example, the terms and conditions of lending – and some of which may not be – for example, the attitudes and practices of bank supervisors other than the central bank itself and laws enacted by the legislature. A potential drawback of the trend in recent years toward large central bank balance sheets and high levels of bank reserves is that banks may rarely need to borrow from the discount window in relatively tranquil periods and partly as a result will go to extreme lengths to avoid borrowing in financial crises. An important open issue for monetary economists and central bankers is how central bank lending can regain its stabilizing function in economies where it has been severely diminished by banks’ unwillingness to borrow. A closely related important issue is how borrowing from the central bank should be treated in bank liquidity requirements – as a source of liquidity or as something to be discouraged.

REFERENCES Adrian, T., Kimbrough, K., & Marchioni, D. (2011). The Federal Reserve’s commercial paper funding facility. Economic Policy Review, 17(1), May. Federal Reserve Bank of New York. Anbil, S., & Carlson, M. (2019). The re-emergence of the Federal Reserve funds market in the 1950. FEDS notes. Washington, DC: Board of Governors of the Federal Reserve System, March 22, 2019. https://doi​.org​/10​.17016​/2380​-7172​.2312.

Central bank lending 

99

Anbil, S., Carlson, M., & Styczynski, M.-F. (2021). The effect of the PPPLF on PPP lending by commercial banks. Finance and Economics Discussion Series 2021-030. Board of Governors of the Federal Reserve System. Ashcraft, A., Gârleanu, N., & Pedersen, L. (2010). Two monetary tools: Interest rates and haircuts. No 16337, NBER Working Papers. National Bureau of Economic Research, Inc. Retrieved from https:// EconPapers​.repec​.org​/ RePEc​:nbr​:nberwo​:16337. Ashcraft, A., Malz, A., & Pozsar, Z. (2012). The Federal Reserve’s term asset-backed securities loan facility. Economic Policy Review, Federal Reserve Bank of New York, 18(Nov), 29–66. Retrieved from https://ideas​.repec​.org​/a ​/fip​/fednep​/y2012inovp29​- 66nv​.18no​.3​.html. Bagehot, W. (1896 [1873]). Lombard street. Kagan Paul, Trench, Trubner & Co. Ltd. Bank of England. (2012). The funding for lending scheme. Quarterly Bulletin 2012 Q4. Retrieved from https://www​.bankofengland​.co​.uk​/quarterly​-bulletin​/2012​/q4​/the​-funding​-for​-lending​-scheme. Bank of England. (2018). The term funding scheme: Design, operations and impact. Quarterly Bulletin 2018 Q4. Retrieved from https://www​.bankofengland​.co​.uk​/quarterly​-bulletin​/2018​/2018​-q4​/the​term​-funding​-scheme​-design​-operation​-and​-impact. Bernanke, B. S., Geithner, T. F., & Paulson, H. M. (2020). First responders. Yale University Press. Board of Governors of the Federal Reserve System. (1971). Reappraisal of the Federal Reserve discount mechanism, Volume 1, p. 3. Board of Governors of the Federal Reserve System. (1978). Sixty-fifth annual report of the Board of Governors of the Federal Reserve System. Board of Governors of the Federal Reserve System. (1984). Monetary policy testimony, July. Board of Governors of the Federal Reserve System. (2002). Alternative instruments for open market and discount window operations, Federal Reserve System study group on alternative instruments for system operations, December 2002. Retrieved from https://www​.federalreserve​.gov​/boarddocs​/ surveys​/soma​/alt​_instrmnts​.pdf. Board of Governors of the Federal Reserve System. (2019). Prudential standards for large foreign banking organizations; revisions to proposed prudential standards for large domestic bank holding companies and savings and loan holding companies, Federal Register, pp. 21988–22036. Retrieved from https://www​.federalregister​.gov​/documents​/2019​/05​/15​/2019​- 07895​/prudential​-standards​-for​ -large​-foreign​-banking​-organizations​-revisions​-to​-proposed​-prudential. Board of Governors of the Federal Reserve System. (March 2020). Quarterly report on Federal Reserve balance sheet developments. Retrieved from https://www​.federalreserve​.gov​/monetarypolicy​/files​/ quarterly​_balance​_sheet​_developments​_ report​_202003​.pdf. Board of Governors of the Federal Reserve System. (March 2021). Senior financial officer survey. Retrieved from https://www​.federalreserve​.gov​/data ​/sfos​/march​-2021​-senior​-financial​-officer​-survey​ .htm. Board of Governors of the Federal Reserve System Staff. (Various years). Monetary policy alternatives. December 9, 1988; February 3, 1989; March 24, 1989; February 2, 1990. Retrieved from https://www​ .federalreserve​.gov​/monetarypolicy​/fomc​_ historical​_year​.htm. Bordo, M. D., & Siklos, P. L. (2018). Central banks: Evolution and innovation in historical perspective. In R. Edvinsson (Ed.), Sveriges riksbank and the history of central banking. Stockholms Universitet, Tor Jacobson, Sveriges Riksbank, Daniel Waldenström, Uppsala Universitet. Sweden: Cambridge University Press. Burgess, W. R. (1946 [1927]). The reserve banks and the money markets. Harper and Brothers. Campbell, S., Covitz, D., Nelson, W., & Pence, K. (2011). Securitization markets and central banking: An evaluation of the term asset-backed securities loan facility. Journal of Monetary Economics, 58(5), 518–531. Carlson, M., Duygan-Bump, B., & Nelson, W. (2015). Why do we need both liquidity regulations and a lender of last resort? A perspective from Federal Reserve lending during the 2007–09 U.S. Financial crisis. Finance and Economics Discussion Series 2015-011. Washington: Board of Governors of the Federal Reserve System. http://doi​.org​/10​.17016​/ FEDS​.2015​.011. Carlson, M., & Macchiavelli, M. (2020). Emergency loans and collateral upgrades: How broker-dealers used Federal Reserve credit during the 2008 financial crisis. Journal of Financial Economics, 137(3), 701–722. Carlson, M., & Rose, J. D. (2017). Stigma and the discount window. FEDS Notes 2017-12-19. Board of Governors of the Federal Reserve System.

100  Research handbook of financial markets

Clarida, R. H., Duygan-Bump, B., & Scotti, C. (2021). The COVID-19 crisis and the Federal Reserve’s policy response. Finance and Economics Discussion Series 2021-035. Washington: Board of Governors of the Federal Reserve System. https://doi​.org​/10​.17016​/ FEDS​.2021​.035. Deutsche Bundesbank. (1995). The monetary policy of the Bundesbank. Retrieved from https://www​ .bundesbank​.de​/resource​/ blob​/ 702946​/091​0 77e​3165​3280 ​48b7​3f63​7f6aab346​/mL ​/the ​-monetary​ -policy​-of​-the​-bundesbank​-data​.pdf. Diamond, D. W., & Dybvig, P. H. (1983). Bank runs, deposit insurance, and liquidity. Journal of Political Economy, 91(3). Domanski, D., Moessner, R., & Nelson, W. (2014). Central banks as lenders of last resort: Experiences during the 2007–10 crisis and lessons for the future. In Re-thinking the lender of last resort (Vol. 79, pp. 43–75). Bank for International Settlements. Retrieved from https://www​.bis​.org​/publ​/bppdf​/ bispap79​.pdf. Duygan-Bump, B., Parkinson, P., Rosengren, E., Suarez, G. A., & Willen, P. (2013). How effective were the Federal Reserve emergency liquidity facilities? Evidence from the asset-backed commercial paper money market mutual fund liquidity facility. Journal of Finance, 68(2), 715–737. Eichengreen, B. (2018). The two eras of central banking in the United States. In R. Edvinsson, T. Jacobson & Daniel Waldenström (eds.), Sveriges Riksbank and the History of Central Banking. Uppsala Universitet. Sweden: Cambridge University Press. Federal Open Market Committee. (1971). Record of policy actions. Retrieved from https://www​ .federalreserve​.gov​/monetarypolicy​/files​/fomcropa19710727​.pdf. Federal Reserve System Committee on the Discount and Discount Rate Mechanism. (1954). Report on the discount mechanism. Retrieved from https://fraser​.stlouisfed​.org​/files​/docs​/ historical​/federal​ %20reserve​%20history​/discountrate​_19540312​.pdf. Freixas, X., Giannini, C., Hoggarth, G., & Soussa, F. (2000). Lender of last resort: What have we learned since Bagehot? Journal of Financial Services Research, 18(1), 63–84. Goodfriend, M., & Lacker, J. (1999). Limited commitment and central bank lending, No. 99-02, Working Paper, Federal Reserve Bank of Richmond. Hill, A. T. (2009a and 2009b). The First Bank of The United States and the Second Bank of the United States. Federal Reserve History. Retrieved from https://www​.fed​eral​rese​r veh​istory​.org/. Hrung, W. B., & Seligman, J. S. (2011). Responses to the financial crisis, treasury debt, and the impact on short-term money markets January 1, 2011. FRB of New York Staff Report No. 481, Available at SSRN. https://ssrn​.com​/abstract​=1739777 or http://dx​.doi​.org​/10​.2139​/ssrn​.1739777. Issing, O. (2018). The European Central Bank. In R. Edvinsson, Tor Jacobson, & Daniel Waldenström (eds.), Sveriges Riksbank and the history of Central Banking. Uppsala Universitet. Sweden: Cambridge University Press. King, A. C. (1919). Letter to honorable Carter Glass, Secretary of the Treasury. Retrieved from https:// fraser​.stlouisfed​.org​/files​/docs​/meltzer​/ bogsub120919​.pdf. King, M. (2016). The end of alchemy. New York; London: W.W. Norton & Company. Li, L., Li, Y., Macchiavelli, M., & Zhou, X. A. (2021). Liquidity restrictions, runs, and central bank interventions: Evidence from money market funds (May 24, 2021). Available at SSRN. Retrieved from https://ssrn​.com​/abstract​=3607593 or http://dx​.doi​.org​/10​.2139​/ssrn​.3607593. Lindsey, D. E. (2003). A modern history of FOMC communication: 1975–2002. Federal Open Market Committee. Lopez, J. A., & Spiegel, M. M. (2020). Small business lending under the PPP and PPPLF programs. Federal Reserve Bank of San Francisco. Madigan, B., & Nelson, W. R. (2002). Proposed revision to the Federal Reserve’s discount window lending programs. Federal Reserve Bulletin. Board of Governors of the Federal Reserve System, 88(7), July. Nelson, B. [William R.] (2000). Transition to a below-market discount rate. Memorandum to Governor Meyer. Board of Governors of the Federal Reserve System. Stein, J. C. (2013). Liquidity regulation and central banking. Speech at “finding the right balance” 2013 Credit Markets Symposium sponsored by the Federal Reserve Bank of Richmond. Charlotte, NC. Retrieved from https://www​.federalreserve​.gov​/newsevents​/speech ​/stein20130419a​.htm. Stiglitz, J. E., & Weiss, A. (1981). Credit rationing in markets with imperfect information. American Economic Review, 71(3), 393–410. Retrieved from http://www​.jstor​.org​/stable​/1802787.

Central bank lending 

101

Swiss National Bank. (2022). Investment of assets. Retrieved from https://www​.snb​.ch​/en​/iabout​/assets. Tucker, P. (2009). The repertoire of official sector interventions in the financial system – Last resort lending, market-making, and capital. Remarks by Mr Paul Tucker, Deputy Governor, Financial Stability, of the Bank of England, at the Bank of Japan 2009 International Conference “Financial System and Monetary Policy: Implementation”, Bank of Japan, Tokyo, 27–28 May 2009. Tucker, P. (2018). Unelected power. The quest for legitimacy in central banking and the regulatory state. Princeton & Oxford: Princeton University Press. Wiggins, R., & Metrick, A. (2016). The Federal Reserve’s financial crisis response: The term assetbacked securities loan facility (February 1, 2016). Available at SSRN. Retrieved from https://ssrn​ .com​/abstract​=2723614 or http://dx​.doi​.org​/10​.2139​/ssrn​.2723614.

5. The workings of liquidity lines between central banks1 Saleem Bahaj and Ricardo Reis

5.1 INTRODUCTION1 A liquidity line between two central banks is an agreement to provide a collateralized loan of the currency issued by the source central bank to the recipient central bank. They are often structured as a foreign exchange (FX) swap, and so are commonly referred to as central bank swap lines, even though they are only superficially related to the FX swaps seen in private markets. In this swap, while the recipient central bank receives the foreign currency it desires, the source central bank gets as collateral a deposit with the recipient central bank, which it will almost never use, and which has questionable value as security. More recently, the liquidity lines have been structured as repurchase agreements, a more accurate representation of how they work. The agreements also offer substantially more protection to the source central bank. Historically, the proceeds of the loans were used in FX markets. More recently, they have been used to provide loans to banks in the jurisdiction of the recipient central bank. After spectacular growth since the financial crisis of 2007–2009, consolidated after the pandemic of 2020, today, the liquidity lines are one of the foundations of the international financial system, potentially involving larger amounts than the IMF could ever lend. This chapter provides an entryway for readers interested in understanding how these liquidity lines work. We start by briefly laying out their historical evolution since the 1960s. While liquidity lines predate this time, and their history could fill more than one long book, looking only at their use in the last 60 years already provides a rich picture of the diversity in their origins, goals, and uses. We then provide a thorough description of how the modern liquidity lines that are in operation in 2021 work. We provide descriptions of the relation between the two central banks, the use of the funds by the recipient central banks, and the steps needed to establish the line. Each involves many choices, and this has led to a very diverse set of arrangements in place today. The liquidity lines can perform three roles in the policy toolkit: to intervene in FX markets, to preserve financial stability, and to promote currency usage and trade credit. We describe each separately, as well as their interactions. Finally, we draw lessons for researchers, policymakers, and financial market participants to keep in mind in their future usage of these liquidity lines. Central bank liquidity lines are a pervasive tool with a rich history; they have been used by multiple central banks for different purposes and played a key part in the policy response to recent financial crises. With this chapter, we hope they will also become less mysterious.

1 We are grateful to Marina Feliciano and Borui Niklas Zhu for research assistance and to Refet Gürkaynak for comments. 102

The workings of liquidity lines between central banks  103

5.2 THE EVOLUTION OF THE LIQUIDITY LINES Liquidity lines between central banks have a long history. For instance, Flandreau (1997) describes a loan of silver from the Bank of England to the Banque de France in 1847 following bad harvests in France. Moreover, the classic gold standard that dominated the international monetary system in the 19th century required flows of specie back and forth across countries. These were smoothed out by bilateral credit between central banks that effectively pooled some of their gold reserves. The collapse of this form of cooperation between central banks was one of the contributors to the end of the regime after World War I (Eichengreen, 1996). While this is not the place to provide a detailed historical account of the liquidity lines, a brief historical review of the last few decades is useful to put the chapter into context. The modern history of liquidity lines starts with the last decade of the Bretton Woods era. Many of the contractual arrangements between central banks that we see today originated in this period. In 1960, the Federal Reserve (Fed) began swapping USD for CHF with the Swiss National Bank. This was followed in 1962 by a more ambitious program to establish bilateral swap lines with major counterparts in the rest of Europe, as well as Canada and Japan. These grew over the decade into a broader network, with the Bank for International Settlements (BIS) in a coordinating role. An important motivation for the liquidity lines was to preserve the fixed exchange rate regime, either by financing foreign exchange interventions or by using swaps as a substitute for transfers of gold. As the United States ran large current account deficits during the decade but was unwilling to transfer gold to other nations, it instead gave them USD through liquidity lines. Another important motivation of these lines was financial stability. The US Treasury had tight financial regulations, partly as a legacy of the Great Depression, and partly as a barrier to international capital flows that made it temporarily possible for the US to run large deficits and yet retain its central role in the system. This led to the growth of offshore USD credit markets, most notably the eurodollar market. The Fed used USD swap lines, coordinated by the BIS, to intervene in these markets (McCauley & Schenk, 2020). The usage of these USD liquidity lines peaked in 1974 at around $240bn (2017 prices). It followed the financial instability driven by the USD going off gold in 1971, the sharp increase in oil prices in 1973, and the resulting sudden rise in interest rates and inflation. With Bretton Woods over, and in the aftermath of a USD crisis in 1978, the liquidity lines stopped playing an important role. They remained open but rarely used, with constant nominal balances. The Fed formally ended them in 1998 (with the exception of lines to Canada and Mexico under NAFTA). Right at this time, elsewhere in the world, a second stage in the history of liquidity lines started with the Chiang-Mai initiative. In 1997, several South East Asian countries went through balance of payments crises. The desire to prevent future crises led many of the affected central banks to accumulate large reserves of assets denominated in USD (and other foreign currencies). In order to boost the effectiveness of their reserves, 14 central banks in East Asia formed a network of liquidity lines in 2000 that would share these reserves if one country needed them to intervene in the foreign exchange market. Originally, the initiative was a network of bilateral swap lines. However, in 2010, it became a multilateral swap line whereby participating central banks could swap their own currency for USD drawn from a pool generated from the combined reserves of the participants. While the goal of these lines was similar to those under Bretton Woods, the novelty was that the central bank of the currency being lent (the Fed) was not part of the network.

104  Research handbook of financial markets

The year 2001 saw the first of a different type of liquidity line. These lines were not motivated by exchange rate stabilization and focused instead on providing lender-of-last-resort facilities to foreign commercial banks. At the time, the global financial system was highly integrated, with the USD as the dominant global currency. Banks outside the US had substantial USD-denominated lending or holdings of USD securities (McGuire & von Peter, 2013; Aldasoro, Ehlers & Eren, 2019), relying on wholesale funding instead of having a significant USD deposit base. When the terrorist attacks of September 11, 2001, disrupted USD money markets in the United States, banks outside the country faced difficulties borrowing the dollars they needed. Unlike US banks, they did not have easy access to the Fed’s lending facilities. On September 12, the Fed opened a new swap line with the ECB, swapping USD for EUR; on September 13, it expanded its dormant swap line with the Bank of Canada; and on September 14, it introduced a new swap line with the Bank of England. In total, $90bn was made available for the recipient central banks to lend to commercial banks in their jurisdictions that had difficulty borrowing dollars, although only the ECB ended up making any drawings (Kos, 2001). These 2001 swap lines were short-lived, expiring after 30 days, once the initial financial disruption brought about by the terrorist attacks was over. However, the situation repeated itself in 2007, as the global financial crisis again disrupted USD money markets. Unlike in 2001, the situation did not stabilize in a couple of weeks. Over the course of 2007–2008, the Fed opened up 14 new bilateral swap lines and at the peak lent $583bn to foreign central banks in order to provide the dollars that could be lent to commercial banks in their jurisdictions (Obstfeld, Shambaugh & Taylor, 2009). These lines were, again, wound down as the crisis abated. By February 2010, they had all been discontinued. However, just three months later, in May 2010, the financial market repercussions of the Euro sovereign debt crisis prompted the Fed to reactivate its swap lines with the ECB, Bank of Japan, Bank of England, and the Swiss National Bank (SNB). After years of institutional inactivity, the Fed had found itself setting up liquidity lines to foreign central banks at relatively short notice for the third time in a decade. Repeated uses motivated a series of agreements over the course of 2010–2013 that culminated in a new stage of liquidity lines, that we refer to in this chapter as the standing swap line network: the four central banks that received swap lines in 2010, plus the Fed and the Bank of Canada, agreed to a network of reciprocal, permanent, standing swap arrangements of unlimited amounts. This network among these six major central banks has been the source of the vast majority of lending through central bank liquidity lines since the global financial crisis. The pattern of a crisis reforming the Fed’s liquidity line network continued when the Covid-19 pandemic disrupted dollar money markets in 2020. The terms of the standing swap line network were eased, nine new temporary swap lines were established and the Fed lent $449bn to counterparty central banks (Bahaj & Reis, 2020a; Aizenman, Ito & Pasricha, 2021). The liquidity line network extends beyond the Fed. While the USD is dominant globally, other currencies have a regional role around a regional financial center (Eichengreen & Lombardi, 2017). The central banks in these centers have created similar swap lines with their neighboring central banks. In Central and Eastern Europe, many mortgages and other bank loans are denominated in EUR or CHF, which causes funding problems for their banks during a financial crisis. The Swiss National Bank created CHF and EUR-denominated swap lines with the Polish National Bank, the Hungarian National Bank, and the ECB at the time of

The workings of liquidity lines between central banks  105

the global financial crisis (Andries, Fischer & Yesin, 2017), and the ECB set up a network of bilateral liquidity lines with European central banks outside the Eurosystem during the Covid pandemic (Albrizio et al., 2021). The Bank of Japan has established a small network of JPYdenominated swap lines in the Asia-Pacific region for similar reasons. A different type of network has its center in China and emerged in 2009 with the creation of the RMB swap lines. International trade is commonly invoiced in USD, inducing importers and exporters to get trade credit in USD as well (Bahaj & Reis, 2020b). The global financial crisis raised these borrowing costs (Bruno, Kim & Shin, 2018), motivating Chinese policymakers to try to internationalize the RMB and insulate their trading firms from future shocks to USD funding (Zhou, 2009, 2017). One of the key planks of this initiative was the establishment of 38 bilateral swap arrangements by the People’s Bank of China (PBoC) over the decade through 2020 (Garcia-Herreroa & Xia, 2015; Bahaj & Reis, 2020b). Rather than FX stabilization or providing a lender of last resort facility to banks, these swap lines are designed to provide RMB offshore that can be used for the purpose of trade settlement. Currently, this is the widest network of bilateral liquidity lines among all central banks, and the notional limit on drawings is comparable to the amounts drawn from the Fed’s liquidity lines during the global financial crisis. Reported drawings, however, have been relatively limited at around $10bn (Perks et al., 2021). The final innovation of note came with the 2020 pandemic and the creation of a new set of repo lines by the ECB and the Fed. These repurchase agreements differ from other liquidity lines in how the loan is structured but have similar purposes to the conventional foreign exchange swap lines. The Fed established its Foreign and International Monetary Authorities (FIMA) repo facility on March 31, 2020, which allows foreign central banks to borrow USD overnight against US treasuries so long as they are eligible for an account with the FRB New York. The ECB followed suit on June 25, 2020, with the Eurosystem repo facility for central banks (EUREP), which offers EUR loans to central banks against EUR-denominated debt issued by Euro-area governments. One important innovation with these repo facilities is that they are, in principle, available to a broader set of central banks than the more exclusive bilateral liquidity lines. They are arm’slength relationships, open to any central bank that has the government securities to pledge as collateral, with standardized take-it-or-leave-it terms that are equal for all. The terms are worse compared to a bilateral line, a point that the ECB explicitly made when setting up the EUREP facility (ECB, 2020). On top of FIMA and EUREP, the ECB also established bilateral repo lines with six other European central banks in 2020, but it has not disclosed the terms of these arrangements. To conclude, the network of agreements between central banks has evolved over the past half-century, typically in response to crises in financial markets that affected offshore borrowing costs. The Asian financial crisis, the global financial crisis, the Euro sovereign debt crisis, and the pandemic recession all came and went. They left behind a growing network of liquidity lines connecting most central banks. Figure 5.1 illustrates this evolution. The liquidity lines were near defunct during the 1990s, but they have grown in the 21st century to become, today, one of the three pillars of the international financial system (the other two being the IMF and regional financial agreements like the European Stability Mechanism). We now turn to how the lines are actually structured and operate before moving on to how they can be used to achieve their policy objectives.

106

Figure 5.1  The evolution of the global liquidity line network

Notes:   1970 – Federal Reserve swap lines only, bubble size reflects outstanding drawings as of end-1970 per McCauley and Schenk (2020). 2000 – bubble size reflects the sum of the notional limits of all swap lines available to country per Denbee, Jung and Paterno (2016). 2009 – bubble size reflects the sum of either the notional limit of all swap lines available to country or, if the line is unlimited, the historical drawings per Perks et al. (2021). 2020 – as 2009 but augmented to include ECB’s bilateral repo lines sourced from Albrizio et al. (2021).

The workings of liquidity lines between central banks  107

5.3 THE OPERATION OF THE LIQUIDITY LINES We focus on the arrangements that are currently in place (for the agreements established during the Bretton Woods era, see the review in McCauley and Schenk (2020)). The transparency of the agreements greatly differs across source central banks. The Fed is perhaps the most transparent. It has published the contracts that underpin the standing swap line network and it regularly reports the individual drawings from its lines at the central bank level.2 The FOMC discussions during the 2008 and 2011 crises that led to setting up the liquidity line policies are publicly available in the meeting transcriptions. From those, one can learn about the concerns and motivations behind these policies. The PBoC, which has the largest number of bilateral liquidity lines, only reveals the dates of the agreements, the quantities available to borrow and aggregate borrowings as annual snapshots published in its monetary policy report. In between these two cases, other source central banks provide varying degrees of detail. A complementary source of information comes from the recipient central banks. The liquidity lines are typically used to fund open market operations (OMOs) in source currency, and the details of these operations, including the amount lent, are often publicly disclosed. In most cases, the recipient just replicates the terms of the liquidity line when setting up its OMOs. When this is not the case, the terms of the OMO provide bounds, since a recipient central bank is unlikely to conduct an OMO at either a longer maturity or a lower interest rate than the liquidity line from the source central bank. From an operational standpoint, there are two legs to a liquidity line. The first is the structure of the agreement between the two central banks. It determines the risk that the source central bank exposes itself to by lending to the recipient central bank. This is a sovereign credit risk. The second leg is how the recipient central bank goes about using the money that it is lent. This leg is often more important for the transmission of policy to the financial system. We discuss each in turn. We then end the section with a summary of the process to set up a liquidity line. 5.3.1 The Agreement between the Two Central Banks Table 5.1 provides a summary of the key features of selected liquidity lines that have been established over the past decade. This is not a comprehensive list but is designed to give a set of examples of how different agreements are structured. Activating the liquidity line. The process for drawing from a liquidity line is as follows. The recipient central bank initiates the transaction by making a request to obtain a certain amount of source currency at a particular date for a specific maturity from the source central bank.3 The source central bank must then approve the request. If approved, the relevant funds and collateral are then deposited at the agreed transaction date. All the central banks in the standing swap line network require at least one day’s notice between initiation and the transaction

2 See  https://www​.newyorkfed​.org​/markets​/international​-market​- operations​/central​-bank​-swap ​arrangements (last accessed December 30, 2021). 3 Reserve sharing agreements, like the Chiang-Mai initiative, are different in that the currency lent is a reserve currency, typically USD, rather than the currency of the source central bank. The process still works in the same way, except the source central bank would need to liquidate reserves to fund the transaction rather than just create the money being lent.

108

Federal Reserve

Federal Reserve

European Central Bank

People’s Bank of China

Bank of Japan

Federal Reserve

Federal Reserve

European Central Bank

Swap line network

Swap line network

Swap line network

Bilateral

Bilateral

Bilateral

FIMA

Bilateral

National Bank of Romania

Hong Kong Monetary Authority

Norges Bank

Reserve Bank of Australia

Monetary Authority of Singapore

Bank of England

Bank of Japan

European Central Bank

Recipient

Repo

Repo

Swap

Swap

Swap

Swap

Swap

Swap

Type

4,500

60,0003

30,000

1,600,000

300,000

∞

∞

∞

Max. Borrowable Amount (Source Currency mil.)

0

1,400

5,400

0

Undisclosed

10

225,839

291,289

Max. Borrowed Amount (Source Currency mil.)

Undisclosed

Interest excess reserves + 25bp

USD OIS + 25bp

Undisclosed

Undisclosed

ECB repo rate + 25bp

USD OIS + 25bp

USD OIS + 25bp

Interest Rate

5/06/2020

31/03/20204

19/03/2020

18/03/2016

8/03/2013

30/11/2011

10/05/2010

10/05/2010

Date of Agreement1

–

14/05/2020

30/03/2020

–

Undisclosed

13/03/2019

20/05/2010

12/05/2010

Date of First Drawing

Euro area government debt, EUR denominated, haircuts undisclosed

US treasuries, haircuts applied “similar to discount window”

NoK equivalent to USD borrowed

AUD equivalent to JPY borrowed

SGD equivalent to RMB borrowed

GBP equivalent to EUR borrowed

JPY equivalent to USD borrowed

EUR equivalent to USD borrowed

Collateral

Undisclosed

Overnight 5

~3 months2

Undisclosed

~3 months2

88 days (greater upon agreement)

88 days (greater upon agreement)

88 days (greater upon agreement)

Maximum Maturity

No

No

No

Yes

Yes

Yes

Yes

Yes

Reciprocal

Notes:   Terms accurate as of publication date. 1 Excludes prior agreements that have since lapsed. 2 Maximum maturity of recipient central banks operations funded through the liquidity line. 3 HKMA has committed to only draw up to $10,000 million from the facility. 4 HKMA announced it would start using the FIMA repo facility on 22/04/2020. 5 FIMA facility is stated as being limited to overnight maturity. HKMA operates one-week dollar repos funded through FIMA, how this is reconciled is not disclosed.

Source

Framework

Table 5.1  Terms and conditions of selected liquidity lines

The workings of liquidity lines between central banks  109

taking place, with the exception of the Bank of Japan which requires two days due to the time difference.4 The request for activation must also align with the timing of the settlement cycle of operations funded through the liquidity line, which we discuss below. Collateral. The collateral that the recipient provides is the operational distinction between a swap line and a repo line. In a swap arrangement, the recipient central bank gives the source central bank a deposit of the recipient currency of the same value as the source currency borrowed. The spot exchange rate at initiation is typically used so that the loan is structured as an FX swap. In a repo line, the recipient bank pledges as collateral securities denominated in source currency (although nothing prevents other denominations from being used) subject to haircuts imposed by the source central bank. This collateral requirement raises the bar for access, as the recipient central bank cannot just issue the collateral that it needs. Interest rates. The interest rate on the loan is de jure set by the source central bank, although in practice, there will be a negotiation between institutions. Only the source central bank receives interest, as the recipient does not charge interest on any of the currency it provides as collateral (in the case of a swap). Typically, the interest rate is set as a spread over a policy rate or a benchmark market interest rate in line with the standard Bagehot (1873) principle that a central bank should lend at a penalty rate. As of 2020, this spread is 25bp (down from 100bp in 2007) for the Fed’s liquidity lines (and some other lines between major central banks). Anecdotes suggest that the PBoC charges higher spreads. Maturity. For swap line contracts in the public domain, upon mutual agreement, any maturity of loan is possible. In practice, the maximum duration seen in individual drawings is around three months, with maturities of overnight, one week, and one month being common.5 The maturity of the transaction partly reflects the aims of the loan. For example, the Monetary Authority of Singapore (MAS) has an RMB facility whereby it borrows RMB via its PBoC swap line for one week or one month if the funds are to be used for the purposes of stabilizing the offshore RMB market in Singapore, and for three months if the RMB is to be used to finance international trade. Reciprocity. With a swap line, a reciprocal arrangement means that either bank can be the source institution and the loan can go in either direction. There are occasions where reciprocity is relevant. One example is the bilateral swap line between the Bank of Korea and the PBoC where both have borrowed via the swap line to support local banks in supplying trade finance in RMB and KRW, respectively. However, most swap lines are established with the tacit understanding that a specific central bank is likely to play the role of the source institution most or all of the time. For example, since the Fed established its reciprocal swap arrangements with five other central banks in 2009, it has yet to indicate that it plans to use those facilities to borrow foreign currency. The FOMC transcripts suggest reciprocity was granted to signal a commitment to global financial stability. Since the mechanics of borrowing and lending via the swap line are near equivalent, having the arrangement be reciprocal has close to zero operational marginal cost. Limits. In the standing swap line network, the lending amounts are uncapped. Other existing agreements specify a limit to the total amount of loans. The recipient central bank then needs a system to ration access to source currency; typically, it uses variable-priced operations 4 Overnight swaps can be approved on the day so long as the request reaches the source central bank before 8am local time on the same day. 5 The Fed’s swap line contracts limit the maturity to 88 days unless both parties agree to relax this.

110  Research handbook of financial markets

(see the following discussion). In reciprocal arrangements, there is a limit in each direction and the ratio between them need not match market exchange rates. The financial flows. The funds are lent via the source central bank by crediting the recipient central bank’s account within the financial system of the source country. In the case of the USD liquidity lines, recipient central banks have a correspondent account at the Federal Reserve Bank of New York with access to the US payment system. Similar correspondent accounts exist at central banks in other jurisdictions. The PBoC’s liquidity lines are an exception in that the account will typically be at an offshore RMB clearing bank that provides international RMB payment services, rather than onshore in China. In the case of a swap line, the recipient central bank also creates a parallel correspondent account for the source central bank, which is then credited with the amount to serve as collateral. The money held on account at the recipient central bank is not used during the life of the swap; indeed, some swap line contracts actively forbid the use of the account. In the case of a repo line, the recipient instead sells the securities that serve as collateral to the source central bank. For example, in the FIMA facility, the recipient central bank sells treasuries to the Fed System’s Open Market Account. Upon maturity, these transactions are reversed. All correspondent accounts are debited with the same values as at initiation plus interest, and any securities are sold back. Default. To our knowledge, there has never been a publicly disclosed default on a central bank liquidity line. If a default were to occur, the basic assumption is that the source central bank would liquidate the collateral, convert it back to source currency, and then pursue the recipient central bank for any residual amount, including costs. The more likely outcome, however, is that, upon default, the loan would be rolled over while negotiations take place. Indeed, the contracts for the Fed’s swap lines with the other major central bank have no explicit provision for withdrawing the deposit from the recipient central bank in the event of a missed payment. Instead, contractually, any balance not repaid is repeatedly rolled over using an overnight swap at the same exchange and interest rate until the balance is cleared. However, the contract does allow the Fed to offset any missed payments against other sums that the Fed may owe to the recipient central bank. The recipient central bank will typically lend the money it receives to banks within its own jurisdiction in a collateralized operation. A default by the recipient central bank is likely, therefore, when it has been defaulted on by a commercial bank. The recipient central bank will need to either pursue the commercial bank or to liquidate the collateral it is holding. Any residual losses would need to be covered through the country’s foreign exchange reserves, and accessing them would likely require political approval. In the more extreme case where the recipient central bank is unable or unwilling to repay, the collateral available to the source central bank becomes relevant. In a repo line, the collateral is straightforward to access. With a swap line, the collateral is a deposit in an account of a central bank that is already in default. It is likely that a central bank that is unable to pay would be experiencing a balance of payments crisis. Therefore, the recipient currency would have depreciated sharply, diminishing the value of the collateral. Moreover, if the recipient central bank was unwilling to pay, it could unilaterally freeze the correspondent account. Of course, the reputational consequences of doing this would be severe, and it seems unlikely that a central bank would behave in this way. However, this illustrates that ultimately it is the central bank’s reputation for having a stable currency and honoring commitments that serves as security in a swap operation.

The workings of liquidity lines between central banks  111

5.3.2 The Recipient Central Bank’s Use of the Money Once the recipient central bank has access to the source currency, it is free to make whatever payments it wishes. If it sets up domestic lending facilities, it can choose who to lend to, at what maturity, and against which collateral. Across some of these dimensions, there is great homogeneity, while across some others there are significant differences. Table 5.2 summarizes the terms for the USD facilities funded by the Fed’s liquidity lines that were active in 2021, which we now discuss. Purpose. Most liquidity lines are agreements between the two central banks for a specific purpose. If the recipient uses the money for a different purpose, it risks the source central bank in the future no longer authorizing any further drawings, or imposing new contractual terms. One apparent case of misuse is various counterparties of the PBoC drawing RMB from the swap lines for the purpose of padding out official exchange reserves. For example, over the course of 2014–15, the central bank of Argentina (BCRA) borrowed RMB to buy USD in order to bolster its reserves, although none of the USD appears to have been spent (McDowell, 2019). This goes against the typically stated goal of the PBoC’s swap lines being for trade settlement (Georgiadis et al., 2021). However, the PBoC did not publicly protest the arrangement and some sources report that the BCRA’s move had the PBoC’s tacit approval (Tresor Economics, 2018). Given the opacity of the agreement (as with most other PBoC liquidity lines), it is difficult to say whether the contract has been revised and so whether the BCRA and other counterparty central banks will be able to do this again. The Argentinian case is illustrative of a broader trend of using liquidity lines to window-dress official exchange reserves. Borrowing via a swap line generates a foreign liability in domestic currency, but if the proceeds are in a reserve currency (or can be converted into one), then they count towards the country’s gross foreign exchange reserves in official statistics. (Repo lines do not suffer from this problem as the country pledges securities from its reserves as collateral.) Such a transaction does nothing to improve the country’s net financial position, and may even worsen it if the cost of borrowing from the swap line is onerous. Alongside Argentina, Pakistan, Egypt, and Turkey appear to have used their PBOC swap line in this manner. Eygpt and Turkey have likewise drawn on swap lines they have with central banks in gulf states. These transactions have led to discussions regarding a revision of the technical definition of official reserves, in order to prevent a swap line from being used for window dressing (IMF, 2017). Setting aside misuse of the arrangement, the majority of modern lines that do not have an FX stabilization motive are meant to enable the recipient central bank to provide credit in source currency to banks in its jurisdiction. The loans could be a bespoke (often confidential) arrangement between the central bank and a private counterparty. However, recipient central banks typically set up formal facilities where they offer credit in the currency of the source central bank via a market operation. There are many examples of this. All of the drawings from the Fed’s swap lines during the pandemic have been used for USD repo operations conducted by the recipient central banks, with a range of maturities of up to three months. The Bank of England has used its swap line with the ECB within the standing swap line network to offer EUR-denominated one-week repos to UK banks. The Bank of Korea, the Monetary Authority of Singapore, and the Hong Kong Monetary Authority (HKMA) all have RMB lending facilities backed by their swap line with the PBoC. The HKMA also has a USD facility that provides one-week USD repos to banks in Hong Kong that is funded by the repo line with the Fed, the FIMA.

112

Up to 602

4 or 8*** (84)

10

Banco de Mexico

Monetary Authority of Singapore

Hong Kong Monetary Authority

7 days

Weekly

Weekly/fortnightly

Quarterly4

84 days

7/28/84 days

Suspended3

Monthly

Weekly

Weekly

84 days

84 days

7 days

7 days

Weekly

Weekly

Frequency

Interest on excess reserves + 25bp

Tenor equivalent OIS + 25bp

3 month OIS + 25bp

3 month OIS + 25bp

3 month OIS + 25bp

1 week OIS + 25bp

1 week OIS + 25bp

1 week OIS + 25bp

1 week OIS + 25bp

Min. Bid Interest Rate

HKMA exchange fund bills and notes

SGD and G10 currency cash, debt securities and covered bonds rated at least BBB–. CNH cash

Same as for Banco de Mexico credit facilities; public and corporate debt securities

Same as for Danish National Bank DKK credit facilities; Danish government debt securities and Danish covered bonds

All assets under Riksbank collateral criteria; debt securities rated at least AA–, no ABS

Same as for SNB CHF repos; debt securities rated at least AA–, no ABS

All assets under BoE’s Sterling monetary collateral criteria; includes on-balance-sheet corporate loans, asset-backed securities and covered bonds

BoJ pooled collateral; includes debt securities rated either AA or A depending on type, some ABS, mortgage debt

All assets within Eurosystem collateral framework; includes non-marketable credit claims and peripheral Euro area governement debt

Eligible Collateral

None

Undisclosed

Undisclosed

6%

6%

None

For non-USD collateral: 6%, +2% for YEN, AUD, NZD and RMB denominated collateral. Extra haircuts added if mismatch exceeds $10bn

13%; 25% for (terminated) operations of more than one month in maturity

12%

Additional Haircuts

1 Notes:   Tenor sometimes adjusted due to holidays. 2 Limit on the swap line with the Federal Reserve. 3 Danish National Bank’s last dollar operation matured on March 12, 2021. No further operations have taken place since then. 4 Banco de Mexico carries out two 84-day dollar repo operations one week apart every quarter. 4 $4bn is the maximum allotted for seven- and 28-day operations. $8bn is the maximum allotted on the 84-day operations. The Bank of Korea and Norges Bank also carried out operations in 2020; these operations were not extended into 2021. Sources:   European Central Bank, Bank of Japan, Bank of England, Swiss National Bank, Sveriges Riksbank, Danish National Bank, Banco de Mexico, Monetary Authority of Singapore, Hong Kong Monetary Authority.

Up to 302

∞

Swiss National Bank

Danish National Bank

∞

Bank of England

Up to 602

∞

Bank of Japan

Sveriges Riksbank

7 days

∞

European Central Bank

7 days

Tenor1

Max. Allotted Amount ($bn)

Lending Institution

Table 5.2  Terms and conditions of USD operations funded through Federal Reserve liquidity lines (operational in 2021)

The workings of liquidity lines between central banks  113

Settlement, frequency, tenor and limits. The recipient central bank must coordinate the two legs of the swap line so that both the drawing and the OMO can be completed. Typically, bids for domestic operations are taken first. Knowing the demand for source currency, the recipient central bank can then initiate the drawing from the source central bank, so the funds are transferred to its account, and it can then settle the OMO. For the frequency and tenor of the operations, the four central banks within the standing swap line network coordinate the timing and maturity of their USD operations in order to prevent arbitrage opportunities for global banks operating in multiple jurisdictions. But the other central banks with a USD facility funded by a swap line with the Fed make different choices. Another choice that is similar across central banks is the list of eligible counterparties within their banking system. Typically, these are the same set of banks that have access to domestic currency lending facilities. In terms of amounts lent, the drawing limits imposed by the source central bank constrain the recipient’s discretion over the total amount it can lend. The ECB can draw an unlimited amount of USD from its swap line with the Fed, hence it lends USD using a fixed-price full allotment operation. In contrast, central banks with caps on drawings conduct operations of a fixed allotment size and a variable price (or, more precisely, an interest rate). Banks need to bid for the source currency, raising the cost of borrowing from the facility. Even so, in practice, it is nearly always the case that the limit is non-binding and so the total quantity drawn from the liquidity line is still demand driven. Collateral. The difference that stands out the most among institutions is the heterogeneity in the collateral that banks need to provide in order to access the lending facility. This is almost entirely a reflection of the heterogeneity in collateral regimes across central banks in general. Recipient central banks normally simply adopt the same eligibility criteria for their foreign currency operations as they do for the domestic currency ones. For instance, the HKMA limits collateral at its USD facility to Exchange Fund Bills and Notes. These are the HKD-denominated debt securities that the HKMA itself issues to serve as security in domestic monetary policy operations. Another example is the ECB, which applies the broad Eurosystem collateral criteria, including non-marketable bank loans, to Euro Area residents. However, recipient central banks sometimes demand additional haircuts on collateral, reflecting the additional of a foreign currency loan. The BoJ demands 13% more collateral in value terms for a one-week USD loan compared to a one-week JPY loan of the same initial value. The Bank of England, Riksbank, and Danish National Bank have extra haircuts starting at 6%. The SNB and the HKMA ask for no extra haircuts at all. It is the case, however, that these facilities all extend the criteria for eligible collateral relative to that of the Fed’s own standing repo facility, which is limited to US treasuries and government-sponsored agency securities. A historical parallel. The current arrangements of the USD liquidity lines have a historical parallel with the early operations of the Federal Reserve system. When they provide loans, the regional reserve banks effectively borrow (uncollateralized) from the system as a whole, similar to how European central banks use the TARGET II system today. But, in the 1920s and 1930s, the regional reserve banks individually determined the terms under which banks in their jurisdiction could access the lender of last resort facilities. This is just like what happens today with the recipient foreign central banks in the Fed’s current swap line network. The USD liquidity lines turn the Fed into a liquidity provider of last resort on the global scale, analogous to its founding mission at the national level.

114  Research handbook of financial markets

The experience of the Fed shows that such heterogeneity in lender-of-last-resort policies has economic consequences. Richardson and Troost (2009) show that during the great depression, banks located along the Mississippi in the more generous sixth Federal Reserve district had lower failure rates and were better able to maintain credit supply compared to those located in the non-interventionist eighth district. Whether heterogeneity in the lending facilities funded through liquidity lines has had similar economic effects is an open question for research. Differences among the recipient economies pose a severe challenge to identification. 5.3.3 Setting up a Liquidity Line Because of the different layers involved in a liquidity line, there is a difference between two central banks reaching an agreement and the line actually being in operation. There are three stages to setting up a liquidity line. First, the central banks agree in principle to engage in a lending relationship (reciprocal or not, swap or repo). Typically, the maximum amount that can be drawn will be specified and the relevant accounts required for the transactions will be set up. These agreements can lie dormant for months and years. In fact, most of the liquidity lines in Figure 5.1 are stuck at the end of this first stage. The second stage is triggered by the recipient central bank telling the source central bank that it wants to start drawing on the line. They then agree on the terms, such as the maturity and frequency of drawings and the interest rate. Note that since every drawing requires authorization, nothing stops the source central bank from reneging on the agreement and refusing the recipient’s request to start using the line. Once the terms are finalized, the third and final stage can take place. The recipient then sets up a lending facility in source currency, chooses its terms and publicizes it to market participants. The liquidity line is now ready to start operating. In a crisis, these three steps can be completed relatively quickly, but the process is not always instantaneous. For example, during the Covid pandemic, the Fed announced that it would set up new swap lines with three Scandinavian central banks (among others) on March 19, 2020, but the three recipients took until March 26 to take bids on their first operations, and these were not settled until March 30. These were the fastest among central banks that agreed to new USD liquidity lines in March 2020. Others were slower to establish facilities, and others have still not done so, presumably because they anticipated little demand. The fact that most liquidity lines are not fully operationalized does not mean they are ineffectual. The dormant lines still serve as insurance against future shocks to the supply of source currency credit to the banks in the recipient’s jurisdiction. There are some signs in the data of a positive insurance value by merely having a liquidity line announced (Aizenman, Ito & Pasricha, 2021). However, Bahaj and Reis (2022a, 2020a) have shown that actual drawings from the liquidity lines, even when their timing can be anticipated, have a large impact on asset prices in crisis times.

5.4 ECONOMIC CONSEQUENCES The setup of the liquidity lines implies that the source central bank outsources the risk management of the loans to foreign commercial banks, while still extending the umbrella of its liquidity facilities. This division of tasks seems natural. The recipient is better placed

The workings of liquidity lines between central banks  115

to monitor the counterparties in its jurisdiction and the quality of the collateral. But the source central bank is better placed to provide the underlying loan, as it is the issuer of the currency. The net effect is to provide source currency funds for banks in the recipient country. Why they need these funds in the first place is less well understood. In principle, banks could get the source currency from secured credit, unsecured loans, recipient currency loans that are synthesized into source currency using foreign exchange swaps, or (for large banks) internal capital markets across subsidiaries and branches in different jurisdictions. During financial crises, these alternatives can either become unavailable or are more expensive than the liquidity lines. Which one it is appears to depend on the line and the crisis. For instance, for the PBoC swap line, access to RMB may be hard given capital controls in China that prevent access to onshore RMB credit markets, and due to volatility of borrowing costs in the offshore money market in Hong Kong. Avdjiev, Eren and McGuire (2020) argue that, during the Covid crisis, the USD swap lines replaced a large contraction in the unsecured portion of the market. Ultimately, research to answer this question will likely have to use bank-level data on drawings, which as of yet is not made available. There are three primary policy aims behind providing the funds: financial stability, promoting international trade, and FX interventions. We discuss each in turn. 5.4.1 Liquidity Lines as a Financial Stability Tool The top panel of Figure 5.2 illustrates how a liquidity line, in this case, a swap line, can be used to stabilize the financial system in the recipient central bank’s jurisdiction, using the example of the ECB and the Fed. The figure shows the flows of funds and collateral that occur upon the ECB’s drawing on the swap line and lending to a Eurozone bank via a USDdenominated market operation. Initially, there must be a shock to USD-denominated credit markets. When those markets are working normally, liquidity lines would be unused because they charge a penalty rate. Whether that shock hits the banks directly or some other parts of the financial system, it must create some profitable opportunity for the Euro area banks that prompts them to use the line. The Euro area bank can then use these USD: (i) to cover withdrawals, (ii) to provide USDdenominated credit to clients (potentially via the FX swap market), or (iii) to purchase USDdenominated securities. In practice, there is evidence that all three usages are relevant. First, drawings from the Fed’s swap lines have coincided with situations when USD money markets are withdrawing funds from foreign banks (Avdjiev, Eren & McGuire, 2020). Second, setting up the swap lines was correlated with lower deviations from covered interest parity (CIP), which measures the cost of borrowing foreign currency offshore via the swap market (Goldberg, Kennedy & Miu, 2011; Baba & Packer, 2009). Bahaj and Reis (2022a) show that changes in the terms and availability of the swap line causally lower CIP. Ivashina, Scharfstein and Stein (2015) and Eguren-Martin, Busch and Reinhardt (2019) find that a reduction in CIP deviations leads to banks lending more in foreign currency. The lending can also spill over to other jurisdictions; for example, Yun (2021) finds that the Korean branches of banks that had access to the Fed’s swap line drew funding from their parents during the pandemic (relative to other branches and prior to the BOK activating its own USD swap line). Third, Bahaj and Reis (2022a) show that lowering the costs of the swap line leads recipient banks to purchase more USD-denominated corporate bonds.

116  Research handbook of financial markets

(a) As a Financial Stability Tool

(b) As a Tool to Encourage International Trade

(c) As a Tool for FX Interventions

Figure 5.2  The liquidity lines in their different uses

The workings of liquidity lines between central banks  117

This flow of credit can stabilize the financial system through a number of potential mechanisms. From the perspective of the ECB, the lines can prevent costly bank failures. The mere presence of a lender-of-last-resort facility can even head off self-fulfilling runs. Either way, it lowers the cost of borrowing for commercial banks and boosts their profitability, and thus their capacity to engage in financial intermediation. From the Fed’s perspective, liquidity prevents a fire sale of USD-denominated assets by Euro-area banks. It also potentially prevents spikes in key benchmark interest rates in wholesale USD funding markets driven by the spike in demand from Euro area banks (FOMC, 2007). Lastly, the provision of USD abroad may flow back to the domestic financial system, closing arbitrage opportunities and increasing the capacity of the financial system to intermediate between agents (Cetorelli & Goldberg, 2012). On the other side of the scale is moral hazard. The net effect is to insure banks against some of the downside risks of having assets in source currency. This provides an implicit subsidy to Euro area banks having activities in USD. To the extent that an excessive amount of this activity poses financial stability risks, this poses a tradeoff for the policy. The tradeoff is especially complicated since the two central banks have different objectives and different costs and benefits from the loans. Whether or not the facilities are designed optimally is still an open question in the literature. Altogether, as noted by Bahaj and Reis (2022a), the liquidity lines are a lender of last resort facility justified by liquidity crises, preventing runs and reducing fire sales. Their benefits and costs align with those of standard central bank lending facilities that have been studied at least since Bagehot. Promoting financial stability was the stated goal of most central banks when establishing their modern liquidity lines. 5.4.2 Liquidity Lines as a Tool to Encourage International Trade and Currency Usage The middle panel of Figure 5.2 illustrates how a liquidity line can support international trade using the example of the Bank of Korea borrowing RMB from the PboC for the purposes of trade finance. Most of the flows are similar to the top panel, with the added complication that Chinese capital controls require having an offshore RMB clearing bank to intermediate the transaction. In the figure, we have assumed for compactness that both the Bank of Korea and the commercial bank use the same RMB clearing bank for their RMB payments, but this does not need to be true in general. Now, the Korean bank that generates the drawing from the swap line lends the money to a Korean importer to purchase products from a Chinese exporter. The liquidity line caps the wholesale cost of providing RMB trade finance for the Korean banks. Since trade finance is just a specific form of financial intermediation, this could be seen as fitting within a broad financial stability objective. It is well known that instability in credit markets has a knock-on effect on international trade (Amiti & Weinstein, 2011). However, there are three reasons that lead us to separate trade finance from financial stability. First, within the context of financial stability, subsidizing source currency activity by recipient currency banks is seen as an undesirable side effect of providing a lender of last resort facility, as we just discussed. The opposite is true in the case of a line for trade purposes. The goal is to promote trade between the two jurisdictions, and the associated trade credit. There is some evidence that reaching a swap agreement with the PBoC is associated with the country having stronger trade linkages with China (Zhang et al., 2017).

118  Research handbook of financial markets

Second, the liquidity line can smooth frictions in the international payments system outside of crises. A typical cross-border payment, particularly in foreign currency, can pass through several different correspondent banks, leading to delays and transaction costs (BIS, 2021). The liquidity line plugs the Korean Bank directly into the RMB payment system, making transactions faster and cheaper, and lowering the cost of trade. Third, and combining the previous two reasons, trade sales and trade credit require payments, which will happen using the RMB. This directly increases the international use of this currency. Additionally, if the importer is buying inputs for a product that it plans to export, it has an incentive to price exports in source currency too in order to align prices with marginal cost (Gopinath, Itskhoki & Rigobon, 2010). In turn, if a firm is pricing exports in a particular currency it has an incentive to get credit for working capital in that currency as well (Bahaj & Reis, 2020b). Empirically, Bahaj and Reis (2020b), Song and Xia (2020) and Georgiadis et al. (2021) find that the PBoC’s swap lines are associated with increased use of the RMB for cross-border payments and for trade invoicing. Insofar as there are rewards to having an international currency, this provides a separate benefit from the liquidity lines. 5.4.3 Liquidity Lines as a Tool for FX Interventions Using a liquidity line to fund an FX intervention was the purpose of swap lines during the Bretton Woods era and persists today, perhaps most notably in the Chiang-Mai initiative. Some liquidity lines involving the major central banks would also potentially fall into this category. The swap line between the Banco de Mexico and the Fed was used to support the Peso during the 1982 Mexican crisis. The ECB has not indicated whether it would allow its swap lines to be used to defend the pegs of non-Euro-area central banks in the EU that peg to the EUR, but there has been speculation that they could be used in this way too, even if the stated objective of the lines is them being a financial stability tool. As an example, the bottom panel of Figure 5.2 shows an FX intervention by the Danish National Bank (DNB) funded through its swap line with the ECB. This is a hypothetical case; the DNB has not used the swap line in this way. Indeed, most of the FX interventions to defend its EUR peg in recent years have involved selling DKK, but the example is still illustrative. The ECB provides the loan of EUR which the DNB then uses to buy its currency in the FX spot market.6 This would directly increase demand for the recipient currency and raise its exchange rate. Naturally, this usually happens when the recipient is defending a currency peg to the source currency. Of course, once the loan has to be repaid to the source currency, this would put pressure on the exchange rate in the opposite direction. The hope (often unfounded) is that the central bank can deftly use the liquidity line to intervene in a way that lowers the volatility of the exchange rate and to deter any speculative attacks by increasing the time (and so cost) for speculators to sell the currency short. The FX stabilizing aspect can be more subtle than direct interventions. During the 1960s, the US ran a large and persistent current account deficit with European countries, which under the Bretton Woods rules would require a movement of gold from the US to Europe. As the US Treasury resisted this, the Fed instead provided USD, via the swap lines, to European central 6 The timing of these transactions does not need to align precisely; for example, the recipient central bank can sell source currency forward and tap the liquidity line only when the contract expires (McCauley & Schenk, 2020).

The workings of liquidity lines between central banks  119

banks, who would then hold them as reserves instead of drawing at the gold window. Since the swap contract would mature at some point, the Fed was offering the recipient central bank a guaranteed exchange rate for the duration of the swap, thus providing the incentive not to switch to gold (Bordo, Humpage & Schwartz, 2015). When the recipient needed to repay the swap line though, this could prompt a balance of payments crisis as happened in 1976, triggering an IMF program with the UK (McCauley & Schenk, 2020). 5.4.4 Intersections between the Three Tools If the liquidity lines help to make the financial system more stable, then exporters can count on the supply of trade credit promoting international trade. Moreover, as the swap lines put a ceiling on CIP deviations to help prevent fire sales and bank runs, this effect on the forward FX market will naturally spill over to the FX spot market and affect the exchange rate. A stable exchange rate may contribute to reducing the prevalence of sudden stops in capital flows across borders, and is one of the factors that lead to more invoicing in a currency, which in turn boosts demand for trade credit in that currency. In short, all three tools have clear interactions and spillovers with each other. The unifying thread, emphasized by Bahaj and Reis (2022a), is that all three uses of a liquidity line broadly fall under a lender of last resort function. The situations in which this backstop liquidity is more important is what distinguishes between the three. First, if there are large gross international investment positions denominated in source currency, the financial stability role dominates. In contrast, FX interventions typically arise from a negative net investment position. Finally, liquidity lines for trade settlement purposes support gross trade flows. Countries can easily have very different positions on gross investment, net investment, and gross trade.

5.5 LESSONS We conclude with some lessons for three sets of potential readers of this article: researchers, policymakers, and market participants. 5.5.1 Three Lessons for Applied Researchers Drawings are not a policy choice. It is tempting to use liquidity line drawings, or allotments at an operation, as a measure of the size of the policy intervention. This is incorrect. Drawings are endogenous, partly determined by demand for credit in source currency, rather than solely by changes in policy that shift the supply. Worse, the central banks that have been the heaviest users of liquidity lines conduct fixed-price, full allotment operations. Therefore, the supply curve of liquidity is perfectly elastic, and any variation in quantities is completely driven by changes in demand from banks. Even for the other central banks, where there is a limit to the size of the liquidity lines, these limits have not been a binding constraint of the quantity lent. To measure policy shocks associated with the liquidity lines, researchers can instead look for changes in the terms of the agreement between central banks (including interest rate, the limit and the maturity) or in the terms of the recipient central banks’ operations (including the frequency, the eligible collateral and any haircuts imposed).

120  Research handbook of financial markets

Effects should be more noticeable during crises, measured by CIP deviations. Liquidity lines are instruments of last resort. Banks use them as an outside option, as the penalty rate makes operations funded through a liquidity line unappealing most of the time. Therefore, their effects should be more noticeable during crises. Since research has shown that the direct effect of a liquidity line is to put a ceiling on CIP deviations (Bahaj & Reis, 2022a), these crises can be identified as times when these observed CIP deviations are close to the ceiling. If CIP is close to holding, we should expect the liquidity lines to have effects that are too small to be detected. Alternatively, if CIP deviations are close to the ceiling, but not at it, the liquidity lines can exert a large effect on economic decisions. Even if the constraint placed by the swap line does not bind, being close to the constraint will lower the distribution of likely outcomes in the near future. This can significantly reduce the ex-ante expected cost of borrowing in the source currency. A liquidity line can have large effects even if it is never drawn on. The liquidity lines are only one piece of international financial architecture. The liquidity lines involve central banks and, through them, give rise to capital flows of short maturities between banks across countries during times of stress in private money markets, with no direct link to other policies. The IMF is instead a multilateral organization that extends credit to sovereign nations lasting for many months, during crises involving balance of payments or sovereign debt and conditional on a package of policy reforms. Further, today there are many regional financial agreements, including development banks like the Asian Infrastructure Investment Bank, and intergovernmental organizations like the European Stability Mechanism. These extend credit across borders that can take many different shapes and are often more strongly tied to political goals and decisions. Together, these are three of the main legs of the current international financial system. Because they interact with each other (Gourinchas, Rey & Sauzet, 2019), researchers have to be especially careful when trying to study one in isolation from the others. In times of crisis, all three will be active, with many policy decisions responding to economic outcomes and the expected policies of others. Identification of individual effects can therefore be tricky, and models that can take into account these interactions may be essential. 5.5.2 Three Lessons for Policymakers Specify the terms of the liquidity lines ahead of time. While there is evidence that just having an agreement for a liquidity line can affect economic outcomes, it is still the case that the terms of most agreements lack detail, especially the terms on which the recipient central banks will lend the money. Fischer (1999) provides some compelling reasons for why clearly laying out the terms of any lender of last resort facility in advance is advantageous: this lowers the likelihood of self-fulfilling runs, enables prospective borrowers to take preemptive steps to access the facilities, and serves as a commitment device for lenders that may be exposed to political pressure. Consider setting up standing facilities. Most central banks operate a standing domestic currency lending facility (the discount window) where commercial banks can obtain an overnight loan at short notice. For central banks that have a liquidity line, very little stands in the way of offering a similar emergency overnight facility in the source currency (as opposed to only repo operations on a schedule and with weekly or monthly tenors). This is perhaps most relevant for central banks with a USD liquidity line with the Fed given the dominant role of the USD. Setting up a standing facility is possible since the Fed’s swap line contracts allow for

The workings of liquidity lines between central banks  121

short notice drawings if the maturity is overnight. The virtue of doing so is that implementation lags of actually using liquidity lines could be decreased. Spikes in CIP deviations and interactions with reporting requirements in financial regulations could be avoided (Bahaj & Reis, 2022a). Central banks in the standing swap line network moved in this direction during the pandemic by offering daily repo operations, but the frequency was dropped as the crisis abated. Make collateral regimes consistent across central banks. Within the USD network, some central banks lend only against government securities, while others lend against a broad range of illiquid loans, including household mortgages. Because there are many global banks with operations in multiple jurisdictions, this presents opportunities to arbitrage different lending facilities. For example, the collateral criteria needed to access the ECB’s dollar operations is much broader than to access the Fed’s own standing repo facility, encouraging global banks to use the liquidity lines to obtain USD for their US branches (Cetorelli & Goldberg, 2012). Arguably, from a US perspective, this puts foreign global banks at an advantage compared to domestic banks. Working in the opposite direction, the interest rate charged in the USD swap lines is currently priced 25bp above the Fed’s domestically focused repo facility. 5.5.3 Three Lessons for Financial Market Participants A liquidity line agreement is only one step in the process. As we described, there are multiple steps involved in channeling the proceeds of a liquidity line to commercial banks. Many of the swap line agreements shown in Figure 5.1 have not yet been operationalized. The experience of 2020 and the height of the pandemic is that this operationalization can happen relatively quickly, although with significant differences across countries. The system of liquidity lines is in flux and does not provide the certainty that domestic lending can. Moreover, there are important differences between announcing an agreement (often at unspecified terms), announcing terms or changes of terms, and actually conducting operations (Bahaj & Reis, 2020a; Aizenman, Ito & Pasricha, 2021). Details matter. Different liquidity lines differ in their purpose as well as in the details of their implementation. Many of these details may not matter for macroeconomic outcomes at quarterly frequencies. But for high-frequency variations in prices, they can be very important. An example comes from the settlement cycle of the operations. This cycle takes at least a day, during which bids are taken, amounts are allotted, and then the money is transferred to the account of the bidding bank. This delay causes spikes in offshore funding costs concentrated in the window between bids being taken in an operation and the next operation being settled, and means that allotted amounts are independent of shocks that occur on the day that settlement occurs (Bahaj & Reis, 2022a; Syrstad & Viswanath-Natraj, 2020). Identifying who is borrowing is hard. When the Fed provides a loan through its discount window, by law it has to soon reveal who the recipient was. This is often seen as a pitfall of loans of last resort. If it isolates which banks needed the funds at the penalty rate, it can trigger runs on them by other banks. Such stigma may lead banks to avoid borrowing from the central bank in the first place, thus limiting the effectiveness of the lender of last resort. With the introduction of the USD liquidity lines with foreign central banks, identifying the borrowing banks is much harder. Many foreign central banks only make public which banks borrowed from them many years later. Even today, it is not known which European banks borrowed as much as $285bn from the ECB at the peak of lending in 2008. Most large US banks can today

122  Research handbook of financial markets

obtain USD through their branches and subsidiaries abroad without making this known until much later, thus avoiding any stigma.

5.6 CONCLUSION This chapter provided an overview of the history, terms, and mechanisms behind central bank liquidity lines, ending with lessons for researchers, policymakers, and financial market participants. Research on this topic is still in its infancy, as the liquidity lines became much more prominent following the global financial crisis. After an initial wave of descriptive work, only recently have researchers written models that isolate the concrete channels through which the swap lines work, and used credible identification strategies to measure their causal effects. Today, we can already move well beyond the initial, vague justification for liquidity lines as a tool to “alleviate funding pressures”. Still, much work remains to be done. While we have highlighted some specific open questions in this chapter, in a companion paper, Bahaj and Reis (2022b), we lay out three broad sets of issues for future research. The first issue is how to write the contract that connects the two central banks, as well as the arrangement between the financial institutions and their domestic central bank. The current contracts were designed quickly, in crisis times. An optimal contract would have to balance the usual trade-offs in lender-of-last-resort policies, together with the involvement of two central banks that may differ in incentives, objectives, and constraints. A second issue is how the overall network should be structured. Should the international financial system combine different bilateral arrangements, with holes and indirect connections, or is a multilateral setup with broader coverage optimal? A third broad area of inquiry is the twoway interaction between a central bank providing a liquidity line and its currency being used internationally. Do the liquidity lines contribute toward a currency being used internationally, or do they instead mitigate some of the financial frictions that cause the international financial system to gravitate to having one or a small number of dominant currencies? These are important questions that require detailed theoretical and empirical analysis. Judging by how quickly central banks turned to the liquidity lines during the pandemic of 2020, the work is also urgent.

REFERENCES Aizenman, J., Ito, H., & Pasricha, G. K. (2021). Central Bank swap arrangements in the COVID-19 crisis. NBER Working Paper 28585. Albrizio, S., Kataryniuk, I., Molina, L., & Schäfer, J. (2021). ECB euro liquidity lines. Banco de Espana Documentos de Trabajo 2125. Aldasoro, I., Ehlers, T., & Eren, E. (2019). Global banks, dollar funding, and regulation. BIS Working Paper 708. Amiti, M., & Weinstein, D. E. (2011). Exports and financial shocks. Quarterly Journal of Economics, 126(4), 1841–1877. Andries, A. M., Fischer, A. M., & Yesin, P. (2017). The asymmetric effect of international swap lines on banks in emerging markets. Journal of Banking and Finance, 75, 215–234. Avdjiev, S., Eren, E., & McGuire, P. (2020). Dollar funding costs during the Covid-19 crisis through the lens of the FX swap market. BIS Bulletin, 1, 1–8.

The workings of liquidity lines between central banks  123

Baba, N., & Packer, F. (2009). Interpreting deviations from covered interest parity during the financial market turmoil of 2007–08. Journal of Banking and Finance, 33(11), 1953–1962. Bagehot, W. (1873). Lombard street: A description of the money market. Scribner: Armstrong & Company. Bahaj, S., & Reis, R. (2020a). Central Bank swap lines during the Covid-19 pandemic. Covid Economics, 2. Bahaj, S., & Reis, R. (2020b). Jumpstarting an international currency. CEPR discussion paper 14793. Bahaj, S., & Reis, R. (2022a). Central Bank swap lines: Evidence on the effect of the lender of last resort. Review of Economic Studies, 89(4), 1654–1693. Bahaj, S., & Reis, R. (2022b). The economics of liquidity lines between central banks. Annual Review of Financial Economics, 14, 57–74. BIS. (2021). Central Bank digital currencies for cross-border payments. Bank of International Settlements, IMF, World Bank, Report to the G20. Bordo, M. D., Humpage, O. F., & Schwartz, A. J. (2015). The evolution of the Federal Reserve swap lines since 1962. IMF Economic Review, 63(2), 353–372. Bruno, V., Kim, S.-J., & Shin, H. (2018). Exchange rates and the working capital channel of trade fluctuations. AEA Papers and Proceedings, 108, 531–536. Cetorelli, N., & Goldberg, L. S. (2012). Banking globalization and monetary transmission. Journal of Finance, 67(5), 1811–1843. Denbee, E., Jung, C., & Paterno, F. (2016). Stitching together the global financial safety net. Bank of England Financial Stability Paper 36. Eguren-Martin, F., Busch, M. O., & Reinhardt, D. (2019). Global banks and synthetic funding: The benefits of foreign relatives. Bank of England Staff Working Paper 762. Eichengreen, B. (1996). Golden fetters: The gold standard and the Great Depression, 1919–1939. Oxford: Oxford University Press. Eichengreen, B., & Lombardi, D. (2017). RMBI or RMBR? Is the renminbi destined to become a global or regional currency? Asian Economic Papers, 16(1), 35–59. European Central Bank. (2020). Eurosystem repo facility for centralbanks (EUREP) – FAQ. European Central Bank. Fischer, S. (1999). On the need for an international lender of last resort. Journal of Economic Perspectives, 13(4), 85–104. Flandreau, M. (1997). Central Bank cooperation in historical perspective: A sceptical view. Economic History Review, 50(4), 735–763. FOMC. (2007). Conference call of the Federal Open Market Committee on December 7, 2007. Garcia-Herreroa, A., & Xia, L. (2015). RMB bilateral swap agreements: How China chooses its partners? Asia-Pacific Journal of Accounting and Economics, 22(4), 368–383. Georgiadis, G., Le Mezo, H., Mehl, A., & Tille, C. (2021). Fundamentals vs. policies: Can the US dollar’s dominance in global trade be dented? ECB Working Paper 2574. Goldberg, L. S., Kennedy, C., & Miu, J. (2011). Central Bank dollar swap lines and overseas dollar funding costs. FRB New York Economic Policy Review, 3–20. Gopinath, G., Itskhoki, O., & Rigobon, R. (2010). Currency choice and exchange rate pass-through. American Economic Review, 100(1), 304–336. Gourinchas, P.-O., Rey, H., & Sauzet, M. (2019). The international monetary and financial system. Annual Review of Economics, 11(1), 859–893. IMF. (2017). The treatment of currency swaps between Central Banks: Egypt experience. International Monetary Fund, Thirtieth Meeting of the IMF Committee on Balance of Payments Statistics. Ivashina, V., Scharfstein, D. S., & Stein, J. C. (2015). Dollar funding and the lending behavior of global banks. Quarterly Journal of Economics, 130(3), 1241–1281. Kos, D. (2001). Treasury and Federal Reserve foreign exchange operations. Federal Reserve Bulletin, 87(12). McCauley, R. N., & Schenk, C. R. (2020). Central Bank swaps then and now: Swaps and dollar liquidity in the 1960s. BIS Working Paper 851. McDowell, D. (2019). The (ineffective) financial statecraft of China’s bilateral swap agreements. Development and Change, 50(1), 122–143. McGuire, P., & von Peter, G. (2013). The dollar shortage in global banking and the international policy response. International Finance, 15(2), 155–178.

124  Research handbook of financial markets

Obstfeld, M., Shambaugh, J. C., & Taylor, A. M. (2009). Financial instability, reserves, and Central Bank swap lines in the panic of 2008. American Economic Review Papers and Proceedings, 99(2), 480–486. Perks, M., Rao, Y., Shin, J., & Tokuoka, K. (2021). Evolution of bilateral swap lines. IMF Working Paper 2021/210. Richardson, G., & Troost, W. (2009). Monetary intervention mitigated banking panics during the Great Depression: Quasi‑experimental evidence from a Federal Reserve district border, 1929–1933. Journal of Political Economy, 117(6), 1031–1073. Song, K., & Xia, L. (2020). Bilateral swap agreement and renminbi settlement in cross-border trade. Economic and Political Studies, 8(3), 255–373. Syrstad, O., & Viswanath-Natraj, G. (2020). Price-Setting in the foreign exchange swap market: Evidence from order flow. Warwick Business School [Manuscript]. Tresor Economics. (2018). The global network of central bank swap lines. Ministere de L’Economie et des Finances [Republique francaise]. Yun, Y. (2021). International spillover of Central Bank swap lines - Evidence from the COVID-19 experience of Korea. Finance Research Letters, 102003. Zhang, F., Miaojie, Y., Jiantuo, Y., & Yang, J. (2017). The effect of RMB internationalization on belt and road initiative: Evidence from bilateral swap agreements. Emerging Markets Finance and Trade, 53(12), 2845–2857. Zhou, X. (2009). Reform the international monetary system. Essay by Dr Zhou Xiaochuan, Governor of the People’s Bank of China in BIS Review 41/2009. Zhou, X. (2017). Prospects of the Chinese economy: Broad-based growth Governor Zhou Xiaochuan’s Speech and Q&A at the 32rd G30 Annual International Banking Seminar.

PART II INTERMEDIARIES

6. Banks1 Refet S. Gürkaynak, Jonathan H. Wright and Egon Zakrajšek

6.1 INTRODUCTION1 Imagine an institution offering to buy one unit of wealth in return for a certificate, while promising to buy the certificate back at any time for at least one unit of wealth. If the certificate is an ownership share in a company, the institution is a mutual fund; if it is a demand deposit, then the institution is a bank; and if it is a digital token, the institution is a stablecoin platform. Mutual funds are covered elsewhere in this Handbook. In this chapter, we study banks, eventually relating them to stablecoins and central bank digital currencies. As with many other examples of financial markets and institutions, the origin of banking traces back to Ancient Babylon (Hoggson, 1926). Commercial banking developed in Florence in the late 14th and 15th centuries, with the Medici bank becoming a particularly large and well-known institution throughout Europe. The Medici bank made a number of important innovations in banking. At the time, the prohibition on charging interest—for theological reasons—was a major impediment to the development of a banking system; the Medici bank circumvented this by exchanging money and letters of credit at terms that built in unmistakable but implicit interest payments. The Medici bank also originated the concept of a bank holding company—a company that itself owns one or more banks—an organizational structure that can protect the overall company from the failure of one of its components. In the 17th century, goldsmiths in London and Amsterdam would keep gold in storage for their owners and in return issue receipts. These receipts became a means of payment in their own right, and goldsmiths evolved into banks, making loans as well as accepting deposits. Banks at this point issued their own currency (i.e., banknotes) that evolved out of gold receipts; it is only more recently that central banks have enjoyed a monopoly on currency creation.2 Early banks were very unstable, and there were frequent and widespread bank failures. In the United States during the 19th and early 20th centuries, these failures were made worse by the fact that the U.S. was slower than other industrialized countries to establish a central bank that could act as a lender of last resort during a banking panic. It was finally in 1913, in the aftermath of the banking panic of 1907, that the Federal Reserve System was founded.3 Initially tasked primarily with stabilizing the banking system, the Federal Reserve flunked its first big test by letting thousands of banks fail—resulting in a collapse of money supply—during the Great Depression (Friedman & Schwartz, 1963).

1 We thank Giulio Cornelli for outstanding research assistance and Şant Manukyan for comments. 2 Indeed, there are some places, like Scotland and Hong Kong, where private banks still can print their own notes, though subject to strict limitations. 3 There were two earlier short-lived U.S. central banks during the late 18th and early 19th centuries (Bordo & Haubrich, 2010). 126

Banks 

127

The severe economic downturn of the early 1930s and the associated financial distress brought new reforms to the Federal Reserve System, and a host of new rules and regulations were imposed on the U.S. commercial banking system. Most notable was the 1933 Banking (Glass-Steagall) Act, which imposed a separation between deposit-taking banks and producers of other financial services (i.e., investment banks), established deposit insurance, and prohibited banks from paying interest on deposit and checking accounts. Starting in the 1980s, many of these restrictions on banks’ activities were rolled back. By historical standards, however, the post-WWII period, up to the 2007–2009 global financial crisis (GFC), was a time of unusual stability for the U.S. commercial banking system. The basic business model of commercial banks is to accept deposits and make loans to households and businesses. Many large banks today also have investment-banking arms, which provide financial services to companies, including securities underwriting and syndication, as well as giving advice on merger and acquisition activities. There are also bank-affiliated broker-dealers that facilitate trading and investment in securities. Banks that combine these functions with traditional commercial banking are known as universal banks. In the U.S., the Glass-Steagall Act set up a firewall between commercial and investment banks; the Gramm-Leach-Bliley Act of 1999 repealed this rule, setting the stage for universal banking and further consolidation of the industry (Figure 6.1). In 1974, in response to serious dislocations in international currency and banking markets, most notably the failure of the privately-owned West German Herstatt Bank, the Basel Committee on Banking Supervision (BCBS) was established by the world’s major central banks. Charged with enhancing financial stability by improving the quality of banking supervision worldwide, the BCBS has to date formulated three major agreements—the so-called Basel accords—among signatory countries, which were implemented by the countries through national regulations. Begun in 2009, the Basel III accord substantially tightened bank regulation, especially for large and systemically important financial institutions. As a result of the accord, along with other reforms in individual countries, banks in A. FDIC-insured banks 12,000

41

51

61

71

81

91

01

11

21

Number

8,000

60 50

4,000

40

0

30 03

05

07

09

11

Top 5 banks

13

15

17

19

Percent of assets

B. Banking industry concentration

21

Top 10 banks

Note:   Panel A depicts the number of commercial banks since the inception of the Federal Deposit Insurance Corporation (FDIC), an independent agency created by the Glass-Steagall Act to insure bank deposits and examine and supervise banks. Panel B shows the fraction of the banking industry assets accounted for by the five and ten largest commercial banks over the past 20 years. Source:   FDIC and Federal Reserve Board.

Figure 6.1  Consolidation of the U.S. banking industry

128  Research handbook of financial markets A. Bank solvency risk

B. Return on assets

0.5

10

0.0

5 97

99

01

03

05

07

09

11

13

15

17

Percent

15

1.0 Z-score

20

–0.5 97

Median

99

01

03

05

07

09

11

13

15

17

Interquartile range

Note:   The solid line in Panel A depicts the cross-sectional median of Z-scores of banking systems for advanced economies, while the shaded band shows the corresponding interquartile (P75–P25) range. Panel B depicts the same information for the corresponding return on assets (ROA). Bank Z-score—a common measure of solvency risk— compares the buffer of a country’s banking system (capitalization and returns) with the volatility of those returns; it is estimated as [ROA + (equity/assets)]/σ(ROA), where σ(ROA) is the standard deviation of ROA. Higher Z-scores indicate lower risk. Source:   Global Financial Development Database, World Bank.

Figure 6.2  Bank solvency risk and profitability in advanced economies advanced economies are less leveraged and hence less risky than they were before the GFC (Figure 6.2). But their return on assets was also quite low for a time after the GFC. The role of banks in promoting economic development has been a subject of debate over the centuries, including among the founders of the United States. Alexander Hamilton saw banks as “[T]he happiest engines that ever were invented for advancing trade” (Cowen & Sylla, 2018). Thomas Jefferson, on the other hand, worried that “[B]anking establishments are more dangerous than standing armies” (Jefferson, 1816). Bagehot (1873) emphasized the role of the banking system in enabling new inventions, including those of the Industrial Revolution, to be financed. More recently, authors such as Levine and Zervos (1998) studied the effects of banking development on economic growth and capital accumulation and found evidence of a positive relationship. The literature on the relationship between banking development and growth is vast, and there are questions about reverse causality and possible diminishing returns to increasing banking intermediation. Nonetheless, economists are effectively unanimous that the existence of a healthy banking system is essential to a well-functioning economy. For example, the University of Chicago Initiative on Global Markets surveyed academic economists for their views on the statement that “[T]here is a social value to having institutions that issue liquid liabilities that are backed by illiquid assets.” Not a single respondent disagreed with the statement.4

4 See https://www​.igmchicago​.org​/surveys​/ liquidity/.

Banks 

129

6.2 THEORETICAL FOUNDATIONS OF BANK INTERMEDIATION In his seminal work that relates to most lender-borrower relationships, Townsend (1979) observes that the contract between the bank and its borrower makes the borrower shoulder the risk associated with the project’s uncertain return. Regardless of how successful the project is, there are only two outcomes: repayment of the loan at a predetermined interest rate or bankruptcy. Why is the borrower—who is likely to be at least as risk-averse as the lender— taking on all the risk associated with the project’s uncertain return? And why is the non-full repayment state equated with bankruptcy, whereby the borrower loses everything? Townsend (1979) shows that in the presence of asymmetric information, where only the borrower costlessly observes the outcome of a project, the optimal contract will look exactly like the contract we see between banks and their borrowers today. Subsequently, Mookherjee and Png (1989) called this the “standard debt contract.” To the extent that banks have lower costs of observing or verifying uncertain project outcomes, they have a comparative advantage in being lenders. Any relationship between borrowers and lenders is potentially subject to problems of adverse selection and moral hazard. Suppose that a borrower knows how risky the project that requires external financing is, but the lender does not. As the loan interest rate rises, only borrowers with high-risk projects will want to take out loans. Borrowers with low-risk projects will stay out of the loan market because they do not want to incur high interest costs as their projects are unlikely to fail, resulting in an adverse selection problem. (Stiglitz & Weiss, 1981). Moreover, once a borrower has taken out a loan, they may be less careful with the funds than they would be with their own funds, giving rise to a moral hazard problem. Leland and Pyle (1977) point out that a bank might be especially well equipped to mitigate these information problems. Because of economies of scale, a single institution may be willing to invest the resources needed to obtain information about the riskiness of a project and subsequently monitor the loan. Diamond (1984), however, shows that delegated monitoring without diversification does not work: if the bank makes a single loan, then in the event of the loan being liquidated, the bank itself will fail. Delegated monitoring and a well-diversified loan portfolio go together. Some countries, notably the U.S., have deep and active markets in private debt securities, so the provision of external financing is not necessarily centered only on the commercial banking system. Other countries, notably those in the euro area, rely much more on bankintermediated external finance (Figure 6.3). De Fiore and Uhlig (2011) tie these jurisdictional differences to there being more public information about corporations in the U.S. than in Europe. We discuss the role of banks as producers of information further in the following. The core function of a bank is making loans to businesses and households and accepting deposits. These deposits are short-term—typically demand deposits—whereas the loans are longer-term loans. Indeed, this is the key characteristic of banks across time and countries, which sets them apart from other types of financial intermediaries. Banks are illiquid institutions that by their very nature feature a mismatch between the maturity of their assets (longterm loans) and the maturity of their liabilities (short-term deposits). As emphasized by Diamond and Dybvig (1983), this is both a source of value but also of fragility in the banking system. Many projects that need to be financed with external funds take a long time to be completed (e.g., construction). Meanwhile, depositors may need access to their savings urgently and may keep too much liquidity if they need to self-insure against unexpected

130  Research handbook of financial markets

60 50

Percent

70

40 30 1997

1999

2001

2003

2005 DE

2007 FR

2009 GB

2011 JP

2013

2015

2017

2019

2021

US

Note:   For each specified country, the line shows the share of total private nonfinancial credit outstanding, accounted for by credit intermediated by commercial banks. Source:   Bank for International Settlements.

Figure 6.3  The importance of banks in financial intermediation liquidity needs. Banks can solve this problem by taking in deposits, holding some funds in reserve, and lending out the rest to finance illiquid projects. If only a small number of depositors who actually need their funds make early withdrawals, then the bank is able to cover these using its reserves, without the need for a fire-sale liquidation of assets, the projects it is financing. This system features perhaps the quintessential example of multiple equilibria: if depositors who do not have liquidity needs are not withdrawing their funds, a good equilibrium is attained. In that case, the bank is healthy, and projects are financed until their completion. However, if all depositors simultaneously ask to withdraw their funds, then neither the reserves nor the firesale value of the projects can cover the amount of deposits being withdrawn. In that case, it is rational for each depositor to want to withdraw their money while they can—regardless of whether they have an actual liquidity need for funds—given that the other depositors are withdrawing. The result is a bank-run equilibrium, whereby the bank fails, and projects are prematurely liquidated at fire-sale values, leading to losses for the depositors as well. To avoid the run equilibrium, banking has been subject to oversight and regulation for a long time. Indeed, the first central banks were formed for the express purpose of pooling liquid reserves and being able to withstand unanticipated surges in withdrawals. Banks in most countries now are subject to extensive regulation, in particular, to make sure that they hold sufficient liquid assets (liquidity ratio) and have enough capital (capital adequacy ratio) to both induce bankers to internalize risks and to allow banks to absorb losses due to non-performing loans.

6.3 BANKS AS LIQUIDITY PROVIDERS The model of Diamond and Dybvig (1983) involves both the deposit-taking and loan-making functions of a bank but does not propose any purpose of bank runs, featuring them as an unfortunate by-product. Calomiris and Kahn (1991) and Diamond and Rajan (2001) argue that demand deposits provide incentives for depositors to monitor bank lending in a way that other bank financing arrangements do not. An individual depositor, realizing that the bank with

Banks 

131

their funds has made poor lending decisions, can nonetheless recover their funds by being first in line. Hence, the very fragility of the banks’ funding structure also serves as a way of improving banks’ incentives and thus lowering their cost of capital. Starting with the 1933 Glass-Steagall Act, the U.S. has offered deposit insurance, subject to a ceiling on the amount of insured deposits, which has reduced the frequency of runs on commercial banks. However, uninsured depositors still have an incentive to run and sometimes do as in the case of Silicon Valley Bank in 2023. Before the GFC, many European countries had more limited deposit insurance schemes (Demirgüç-Kunt et al., 2015), but in response to severe strains of their banking systems, they quickly raised deposit ceilings in an effort to avoid bank runs. Nevertheless, significant differences in national legal regimes for dealing with bank failures within the euro area remain. The European Union’s goal of establishing a Banking Union with a European Deposit Insurance Scheme that provides a uniform level of protection for the same category of investors and depositors across participating member states is still a work in progress (Tümmler, 2022). While deposit insurance has greatly lessened the runs problem of Diamond and Dybvig (1983), such deposit-protection schemes lessen the incentive for depositors to monitor banks. At the same time, most policymakers would think it inappropriate and infeasible to expect retail customers to monitor their banks in this way. However, ceilings on the amount of insured deposits provide some incentive for the largest depositors to invest in differentiating good banks from poor ones. In principle, the banks’ two core functions, deposit taking and loan making, could be separated into two separate institutions: the so-called narrow banks that take deposits and invest them in very safe, liquid, short-term securities, such as U.S. Treasury bills (Pennacchi, 2012); and other institutions that make longer-term loans and operate more like a mutual fund or an investment bank. However, we see these two functions conjoined in banks. The most natural explanation for the co-existence of deposit-taking and loan-making within the same institution is that there are significant synergies between the two activities. As argued earlier, it is possible that deposits solve a principal-agent problem. But commercial banks also provide liquidity to customers with unpredictable liquidity needs; the same customer may have surplus funds today but may also need to borrow at some point in the future, hence the desire to have both a deposit and a loan commitment. Both deposits and loan commitments require the bank to have liquid asset holdings, making it efficient for the same institution to offer both (Kashyap et al., 2002). Moreover, Kashyap et al. (2002) show empirically that banks that rely more on deposits, as opposed to interbank lending to fund their activities, are able to grant more credit lines to their borrowers. Ippolito et al. (2016) similarly find that banks that use interbank funding more heavily offer fewer and smaller credit lines to financially constrained borrowers, adding to the evidence that deposits and credit lines are closely related. The introduction of deposit insurance has mitigated although not entirely solved the problem of bank runs. In fact, since the Great Depression, financial crises have generally resulted in inflows of deposits to the banking system. This gives banks the ability to provide insurance against market-wide liquidity shocks (Gatev & Strahan, 2006). An example that these authors highlight is banks providing backup lines of credit to issuers of commercial paper.5 In other 5 Commercial paper is a type of unsecured, short-term debt instrument issued by highly rated corporations, the proceeds of which are used for the financing of payroll, accounts payable and inventories, and meeting other short-term liabilities.

132  Research handbook of financial markets

words, banks are better able to provide this liquidity insurance than other financial intermediaries precisely because of the inflows they receive in times of widespread market turmoil. The theoretical emphasis on liquidity creation as one of the banks’ key functions has led to many attempts to measure this activity. Berger and Bouwman (2009) construct a measure of bank liquidity creation by comparing the liquidity of banks’ assets with that of their liabilities—liquidity is created when the bank has more illiquid assets than liquid liabilities. Overall, they find that U.S. banks over time created substantial and growing liquidity; they also document that bank capital has a positive effect on liquidity creation for large banks, but not for small banks. In related work, Berger et al. (2016) study the effects of regulatory interventions and bailouts on German banks’ liquidity creation, measured in the same way.6 Regulatory interventions are found to reduce liquidity creation, while bailouts have no significant effect. Berger and Sedunov (2017) document a significant positive relationship between bank liquidity creation and real output in a panel of U.S. states; they also find that this relationship is stronger for industries that are more bank-dependent (e.g., manufacturing) than those that are less bank-dependent (e.g., government, healthcare), results consistent with the role of banks as important catalysts of economic growth.

6.4 RELATIONSHIPS AND BANKING Loans account for a major portion of banks’ assets, and through interest income and funding expense, they account for a large share of banks’ revenues and costs (Figure 6.4). Loans can involve “relationship lending,” where banks have worked on acquiring “soft” information on informationally opaque borrowers through repeated transactions; or “transaction lending,” which views each loan as a stand-alone proposition, with the loan officer’s decision based largely on hard data such as tax returns or audited financial statements detailing business activities and the financial performance of a company (Boot, 2000). Much information about small businesses is thought to be soft and acquired by way of relationships. In effect, the relationship between the bank and the borrower gives the bank a form of inside information. Stein (2002) argues that a small bank size is efficient when soft information predominates, whereas organizational efficiencies make a large bank more efficient when loan decisions can be readily reduced to hard data. Many studies have found that the longer that a bank and borrower have had a relationship, the more credit is made available to the borrower and the better the ensuing terms of credit. The benefits of relationships are especially important for small firms, which lack access to arm’s length capital markets and often face binding credit constraints (Petersen & Rajan, 1994; Berger & Udell, 1995), but they also matter for large firms (Lummer & McConnell, 1989; Slovin et al., 1993). In the same spirit, Hoshi et al. (1990) find that Japanese companies with close ties to a single bank were less likely to be credit constrained. The centrality of relationship banking is also important to the transmission of shocks to the macroeconomy. A view of frictionless financial intermediation might imply that a reduction in the loan supply from one bank would quickly be offset by other banks, or indeed by non-bank lenders, as well as in capital markets. There are many studies showing that this is not the case. Slovin et al. (1993) examine the failure of Continental Illinois National Bank in 1984—the 6 The bailouts in question are either public or from bankers’ association insurance funds.

133

Banks  A. Composition of bank assets

B. Composition of bank liabilities 60

40

20

20

0 95 97 99 01 03 05 07 09 11 13 15 17 19 21 Other assets Total securities

Cash and due from banks non-earning assets Gross loans

USD trn

USD trn

40

60

0 95 97 99 01 03 05 07 09 11 13 15 17 19 21 Other liabilities Wholesale liabilities

Total customer deposits

C. Determinants of bank profitability

2.5 0.0 –2.5 –5.0

Percent of assets (a.r.)

5.0

–7.5 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 Operating profit Interest expense

Total non-interest operating income Non-interest expense

Interest income Other operating income/loss

Note:   Panel A shows the major components of banks’ assets, whereas Panel B shows the major components of banks’ liabilities. Panel C shows the decomposition of banks’ operating profits. The sample of banks, which changes over time, includes all Global Systemically Important Banks (G-SIBs) and Domestic Systemically Important Banks (D-SIBs) in advanced economies. For the period shown, this sample of banks accounts for the vast majority of industry assets in each advanced economy. The dollar amounts shown are deflated by the U.S. GDP price deflator (2012 = 100). Source:   Bank for International Settlements.

Figure 6.4  Bank assets, liabilities, and profits in advanced economies largest bank failure in U.S. history at the time—and find that firms with relationships with this bank had difficulty switching to other lenders. The research into the importance of whether and what type of bank a firm happened to have a relationship with took off with the GFC. Studies find that firms that had established relationships with banks were better able to obtain credit after the collapse of Lehman Brothers in the autumn of 2008, compared with firms that had no previously established relationships (Bolton et al., 2016; Banerjee et al., 2021). Paravisini et al. (2014) compare exports of the same product to the same destination by Peruvian firms during the GFC, an identification strategy that accounts for potential differences in demand faced by the different firms. The study finds that firms that had ex-ante relationships with financially stronger banks exported more than their counterparts whose ex-ante relationships were with financially weaker banks. All told,

134  Research handbook of financial markets

the results of these studies provide compelling evidence that lending relationships and the financial health of the bank in the relationship significantly affect the ability of firms to obtain credit. Relationship-based lending may have the advantage of enabling banks to offer loan contracts that reduce risk to the borrowers, most notably by smoothing loan rates over the course of a business cycle (Petersen & Rajan, 1995). Because core deposits are a much more stable source of funding compared to wholesale liabilities, they may allow a bank to reduce borrowers’ interest rate risk.7 Consistent with this hypothesis, Berlin and Mester (2015) find that loan markups increase less during economic downturns for banks with high levels of core deposits, compared to banks that rely more heavily on wholesale liabilities to fund their activities. The decades following the liberalization of the banking industry in the U.S. and Europe in the 1990s saw numerous domestic and cross-border bank mergers. At first glance, the resulting increase in industry concentration could potentially decrease the importance of relationship lending, which is especially relevant for small and medium-sized enterprises (SMEs). Berger et al. (1999) indeed find that when banks merge, the resulting larger bank tends to cut back on small business lending. This could owe to organizational inefficiencies in large banks, which make them unwilling to engage in what would otherwise be positive net present value propositions (Stein, 2002). On the other hand, a more concentrated banking sector might leave more scope for banks to invest in relationships with individual borrowers because they are less worried that another bank will jump in if the borrower turns out to be successful (Petersen & Rajan, 1995). Thus, the impact of bank concentration on relationship lending is ambiguous. Relatedly, the development of new data-driven technologies has greatly increased the importance of hard data in lending decisions, suggesting a diminished role for relationship banking (Boot et al., 2021). Frost et al. (2019) discuss the implications of technology in banking and financial intermediation. The emerging and rapidly developing technologies have the potential to further increase industry concentration, move financial intermediation out of the regulated banking sector, and make banking more transactional. These developments, however, do not necessarily imply the demise of relationship banking altogether. Rather, relationship banking may become a more niche activity for the most informationally opaque borrowers (Boot & Thakor, 2000). A potential downside to relationship banking is that the borrower is locked into a particular bank that is then effectively a monopolist and may offer uncompetitive terms—the so-called holdup problem (Boot, 2000). The problem can of course be solved by the firm having multiple banks, but that weakens the benefits of relationship banking (Ongena & Smith, 2001). Petersen and Rajan (1994) and Berger and Udell (1995) find that in the U.S. terms of lending improve with the duration of the lending relationship, which suggests that the holdup problem is not too severe. Degryse and Van Cayseele (2000), in contrast, document the opposite result for Europe. 7 Core deposits are the most stable source of funds for banks and include small-denomination time deposits, payment accounts, and checking accounts. Wholesale liabilities, in contrast, are funding sources that can be increased or decreased at the bank’s discretion to pay off its maturing deposits and fund new loans. They include, but are not limited to, negotiable large time deposits; Eurodollar and other Eurocurrency borrowings; repurchase agreements; and federal funds purchased. Unlike retail depositors whose core deposits are insured, the providers of wholesale liabilities pay close attention to the credit-risk profile of the bank to which they are providing these funds; moreover, the cost of these wholesale funds is tied closely to market interest rates.

Banks 

135

Degryse and Ongena (2005) study spatial price discrimination in bank business lending in Belgium. They find that loan rates are higher for banks that are geographically close to their borrowers and lower when competing banks are nearby. These results suggest that banks are indeed using their relationships to charge higher interest rates. As banking moves away from physical banks, spatial distance may play less of a role in determining banking relationships. 6.4.1 Effects of Globalization Over the past few decades, banking globalization has led to more diversified ownership, as well as increased foreign ownership, in the banking industry in host countries. As a result, the analysis of the role of geographical distance in borrower–lender relationships has expanded to also examine the role of international borders and the international transmission of financial shocks. Petersen and Rajan (2002) document that within the U.S., the geographic distance between banks and small firms has increased notably since the early 1970s, and the two parties communicate in more impersonal ways. At the same time, they find that relationship banking is still important because of improvements in lender technology. Berger et al. (2003) argue that the need for relationship lending is a constraint on the extent of globalization in banking. There are many studies showing that when a global bank suffers losses in one market, it cuts back on its lending in other markets. In an influential paper, Peek and Rosengren (2000) identify the effects of the Japanese banking crisis of the 1990s on U.S. economic activity by looking at relationships between U.S. firms and U.S. subsidiaries of Japanese banks. Popov and Van Horen (2014) study the lending of European banks from countries other than the five countries with immediate problems (Portugal, Italy, Ireland, Greece, and Spain), but who had sizeable holdings of sovereign bonds of those five countries, and find that the banks with relatively greater exposure to this risky sovereign debt pulled back their lending by substantially more. There is also considerable evidence that in periods of financial market turmoil banks cut their foreign lending disproportionately—in effect, they increase their home bias (Popov & Van Horen, 2014; De Haas & Van Horen, 2013; Giannetti & Laeven, 2012). This means that a high concentration of foreign banks in a country tends to increase the volatility of the domestic credit supply. De Haas and Van Horen (2013) study the pullback in foreign lending after the collapse of Lehman Brothers and find that it was concentrated in markets that were geographically distant from the bank’s home country and in markets where the bank had less experience and did not have a subsidiary. Albertazzi and Bottero (2014) likewise study the decline in lending by foreign banks in Italy following Lehman’s demise. Interestingly, they find that the decline in lending was not an across-the-board pullback. In particular, the decline was concentrated at branches of foreign banks rather than at foreign subsidiaries. Moreover, it was foreign banks that had a large local funding gap—that is, they had more local loans than local deposits—that curtailed their lending the most. All in all, given the importance of relationships in banking, global banks have the downside that foreign credit supply shocks may adversely affect lending and economic activity in the host country. On the other hand, they also act as a buffer against domestic shocks. For example, Bofondi et al. (2017) show that foreign banks in Italy partly offset the contraction in credit supply from domestic banks during the euro area sovereign debt crisis.

136  Research handbook of financial markets

Banking globalization also has important implications for the transmission of monetary policy and associated spillovers. Cetorelli and Goldberg (2012) find that U.S. multinational banks get liquidity from their foreign subsidiaries following a tightening of U.S. monetary policy, which partly insulates them from the effects of domestic monetary policy. At the same time, changes in the stance of U.S. monetary policy have a notable effect on domestic banks without global operations. All told, the empirical evidence demonstrates that global banks have active internal capital markets, through which these institutions contribute to the international propagation of monetary and other financial shocks.

6.5 BANKS AS PRODUCERS OF INFORMATION A crucial role that banks play in the economy is as producers of information (Leland & Pyle, 1977; Diamond, 1984). But as Fama (1985) observes, banks face regulatory costs that nonbank intermediaries do not.8 As a result, banks must add some value in order to survive in equilibrium. The argument in Fama (1985) is that this value lies primarily in avoiding duplication of monitoring costs, especially for small firms. James (1987) considers event-study evidence on the stock price response of firms to public announcements of bank credit agreements and debt issues. He finds that announcements of bank credit agreements generate significant abnormal positive two-day stock returns, whereas announcements of debt issues do not, a finding consistent with banks having a special role as producers of information. Lummer and McConnell (1989) also take an event-study approach, but they break out announcements of new loans, favorable existing loan revisions, and unfavorable existing loan revisions. They find significantly positive (negative) excess stock returns around favorable (unfavorable) loan revisions and no significant excess stock returns around new loan announcements. The interpretation of these results is that banks have a role as producers of information, which may also be arising through the ongoing monitoring process. Part of banks’ role in producing information comes from bundling lending with other activities. Banks provide checking account services for their borrowers, and this may enable them to extract information on the firm that helps them in lending decisions (Black, 1975; Fama, 1985; Mester et al., 2006). Syndicated loans are an innovation that became widespread in the 1990s. The idea is for many banks to join in a set of bilateral loans that are monitored jointly, so as to cut down on the monitoring and enforcement costs. The monitoring is delegated to one or more arranging banks. Naturally, this gives rise to a principal-agent problem. The problem can be mitigated by the arranger banks holding a large share of the syndicated loan. Focarelli, Pozzolo, and Casolaro (2008) find that the larger the share retained by the arranging bank, the lower the interest rate charged on the loan, and the bigger the jump in stock price of the borrowing company when the loan is announced. The arranging bank, however, faces a trade-off; retaining

8 The specific regulatory cost that was the focus of Fama (1985) was the requirement that banks must hold non-interest-bearing reserves in their accounts at the Fed. Since that time, large banks have for the most part circumvented reserve requirements and in March 2020 these requirements were abolished, and the Fed in any case now pays interest on reserves. Banks, however, continue to face other significant regulatory costs, especially compared with most other non-bank financial intermediaries.

Banks 

137

a larger share will lower the cost of funding, but will reduce the diversification benefits of the syndication (Ivashina, 2009). Another innovation in banking has been a move from the “originate-to-hold” to the “originate-to-distribute” model. In the latter case, the bank originates the loan but then sells it on to various third parties, either directly or more commonly by bundling a bunch of loans into various financial instruments, the so-called asset-backed securities. Begun in the U.S. in the 1980s with the issuance of mortgage-backed securities, the originate-to-distribute model allows banks to reduce the credit risk of their loan portfolios; diversify their funding sources thus economizing on overall funding costs; and in certain circumstances hold less regulatory capital. In addition, by moving loans off their balance sheets, the process frees up capital for new lending.9 While adopting the originate-to-distribute model may give banks greater balance sheet capacity to make new loans, it could also be a way in which banks economize on the costs of obtaining information about opaque borrowers. Unfortunately, this undercuts the role of banks as producers of information, as the banks have an incentive to sell loans where they know the borrower to be less creditworthy and/or to reduce the intensity of their subsequent monitoring efforts (Parlour & Plantin, 2008). In that regard, Berndt and Gupta (2009) compare the abnormal stock returns of firms whose loans were sold with those whose loans were not sold in the three years following the loan sale. They find that the equity returns of firms whose loans were sold underperformed by 9% per year. Further evidence of the adverse effects of the originate-to-distribute model on bank lending standards is provided by Keys et al. (2010) for the residential mortgage market and by Bord and Santos (2015) for the corporate loan market.

6.6 REAL EFFECTS OF BANK INTERMEDIATION Research efforts on financial–real interactions were hampered for a long time by the enormously influential Modigliani–Miller theorem (Modigliani & Miller, 1958). This theoretical result shows that under certain conditions—most notably, complete markets—different forms of external and internal financing are perfect substitutes. As a result, not only there is no value-maximizing financing mix for a firm, but the firm’s optimal liability structure is indeterminate. When applied to banks, a clearly articulated discussion of this “irrelevance” is provided by Fama (1980). In essence, banks in the Modigliani–Miller world produce convenience services and earn an appropriate fee but are not special in any other respect. Thus, loss of banking would have no real consequences, and banks play no role in business cycles other than being passive responders. Financial markets and in particular banks mattering for cyclical fluctuations thus require the Modigliani–Miller theorem to not hold. If banks are to play a role in the transmission of monetary policy, or in business cycles more generally, two conditions have to be satisfied. First, the bank loan supply should respond to monetary policy directly. And second, bank loans should not be a perfect substitute for other forms of external finance, at least for some firms. 9 Banks also earn sizable (noninterest) income from the sale of the loans, while the continued servicing of the underlying loans—that is, collecting interest and principal repayments and passing them on to the holders of the asset-backed securities—generates significant fee income.

138  Research handbook of financial markets

Whether financial conditions and bank health mattered for real economic outcomes was studied by Bernanke (1983) in regard to the Great Depression. In this seminal paper, Bernanke empirically argued that financial factors played an important role in prolonging and exacerbating the economic downturn. The major difficulty in empirical studies of what is now called the bank-lending channel is identifying exogenous changes in bank loan supply. The fact that bank loans are pro-cyclical is not surprising: loan demand falls in recessions and increases during booms (Figure 6.5). But during economic downturns, is the equilibrium loan amount declining only because of a fall in loan demand? Or are firms looking to borrow from banks but are unable to do so? Kashyap et al. (1993) (KSW) tried to answer these questions by looking at other forms of borrowing by firms during cyclical downturns, just as bank lending begins to decline. If the decline in bank loans is due to falling loan demand by firms, then all forms of borrowing should decline. If it is due to bank loan supply contracting while firms are looking to borrow, the firms’ issuance of other types of liabilities will increase as bank loans are declining. KSW find strong evidence that other forms of borrowing—especially the issuance of commercial paper—increase while bank loans are falling. This strongly suggests that it is the bank loan supply that is shrinking, and firms are trying to make up for the financing gap by ramping up other types of borrowing. Using more disaggregated data for the U.S. manufacturing sector, Gertler and Gilchrist (1994) show that cyclical fluctuations in the mix between commercial paper and bank loans documented at the aggregate level by KSW are primarily driven by differences in borrowing by small and large firms. In particular, Gertler and Gilchrist (1994) show that following a tightening of monetary policy, all types of borrowing by small firms decline, whereas primarily non-bank borrowing by large firms actually expands during the first couple of quarters following a contractionary monetary policy shock. These differences reflect the fact many firms have a countercyclical demand for short-term credit, as inventories build up and cash flows decline in response to a tightening of monetary policy or at business cycle turning points. If funds were available at prevailing market interest

90 80 70 1999

2001

2003

2005

2007

2009

United States

2011

2013

2015

2017

2019

Euro area

Note:   For each jurisdiction, the line show bank loans outstanding normalized by nominal GDP. Source:   Bank for International Settlements.

Figure 6.5  Bank lending around the GFC

2021

2007:Q4 = 100

100

Banks 

139

rates, all firms would presumably increase their short-term borrowing to counteract the effect of declining cash flows. It turns out, however, that only firms with relatively unimpeded access to credit are able to obtain the desired funds. Thus, a shift to tight money forces high-quality firms to tap the commercial paper market (Calomiris et al., 1995); firms with bank commitments draw down their lines of credit (Morgan, 1998); and the high-quality bank customers receive the funds obtained through the banking system’s liquidation of securities (Lang & Nakamura, 1995). Left holding the bag are the smaller, riskier, lower-valued bank customers, which, when shut out of the bank loan market, have no recourse but to liquidate inventories, cut capital expenditures, and lay off workers. Their reductions in spending and production exacerbate the downturn, leading to an even greater contraction in economic activity than before. For a bank lending channel to exist in monetary policy transmission, the first condition— namely, bank loan supply directly responding to monetary policy—should also be satisfied. Whether and why this may be the case has been a subject of extensive academic debate. In their seminal contribution, Bernanke and Blinder (1992) relate the bank lending channel to the mechanical effect of a fractional reserve banking system. The fact that reserve requirements were never all that binding for most U.S. commercial banks casts doubt on the empirical relevance of this channel.10 Banks, however, are also likely to suffer from asymmetric information and moral hazard problems when raising funds to finance their lending activities. Hence banks’ financial health may impact their ability to extend credit. As shown by English et al. (2018), an increase in market interest rates caused by a policy tightening induces an outflow of core deposits and a switch to wholesale liabilities, resulting in banks facing a more expensive funding mix. As a result, poorly capitalized banks that are unable to raise external funds cut back on their lending (Kashyap & Stein, 2000). Reductions in bank capital during economic downturns can also reduce lending activity. As economic growth slows and defaults rise, the quality of bank loan portfolios worsens. In response, banks seeking to shore up their capital or to meet regulatory capital requirements, tighten their credit standards and cut back on lending, an inward shift in loan supply that curtails spending of bank-dependent borrowers (Van den Heuvel, 2008; Bassett et al., 2014). Many studies have used cross-sectional differences in banks’ financial health to identify the partial-equilibrium effect of loan supply shocks on macroeconomic outcomes. ChodorowReich (2014), for example, finds that firms whose pre-crisis relationships were with banks that turned out to be relatively healthy during the GFC cut employment by significantly less than firms whose pre-crisis relationships were with banks that ex-post turned out to be in a relatively weak financial position; according to his estimates, the reduction in loan supply accounts for nearly one-half of the employment decline at U.S. small and medium-sized firms. Jiménez et al. (2017) and Cingano et al. (2016) similarly find that credit supply disruptions have quantitatively important effects on both employment and investment. In this literature, a natural concern is whether weaker firms would have had established relationships with weaker banks and so the causal effect of a reduction in loan supply on economic activity is not identified. However, the authors of these papers make every effort to control for other observable characteristics of firms and end up with a quite convincing identification. 10 After many years of irrelevance, in March 2020, the Fed eliminated reserve requirements for all depository institutions.

140  Research handbook of financial markets

Overall, the evidence for the presence of a bank lending channel is very strong and motivates the desire by governments to keep their banking systems healthy (as well as, more controversially, encouraging bank credit expansion before elections).

6.7 STABLECOINS AND CENTRAL BANK DIGITAL CURRENCIES With accelerating digitization of the global economy, the number and value of crypto assets has exploded over the past decade (Bank for International Settlements, 2022). The distributed ledger technology, blockchain, coins that are exchanged on these ledgers, such as Bitcoin and Ethereum, and other assets that traded on blockchain implementations, such as non-fungible tokens (NFTs), have proliferated. From the perspective of banks and the banking industry, a particularly salient class of crypto assets is the so-called stablecoins. Stablecoins start from the understanding that the assets traded on blockchains are not money in themselves. As such, they offer to fix the value of the stablecoin against the proper numeraire, the legal tender in the country. Many stablecoins aspire to have parity with the U.S. dollar, whereby the holder of such a stablecoin can exchange it with one dollar at any time. There are two broad types of stablecoins. The first is what is called algorithmic stablecoins, coins that are issued in pairs, one with a fixed value against the dollar and the other with a floating value. An algorithm is employed to execute an arbitrage trading strategy between the two coins so as to maintain the fixed coin’s one-for-one peg to the dollar. In particular, the algorithm sells the stablecoin and buys the floating coin if the value of the stablecoin against the dollar increases beyond parity; conversely, it sells the floating coin and buys the stable one if the stablecoin is trading below parity, enforcing a fixed value of unity for the stablecoin. This system to fix the value of the stablecoin is predicated on the assumption that the supply can be adjusted to maintain a stable value against the dollar at all times. It goes without saying that for two digital entries without any intrinsic value this is a strong assumption.11 In fact, in May 2022, TerraUSD—at that point the third largest stablecoin with a peak market capitalization of nearly $19 billion—collapsed spectacularly. Over the course of just a few days, its value dropped from $1 to just a few cents, as investors lost confidence in the sustainability of the system and rushed to redeem their funds. The second type of stablecoins are the so-called asset-backed ones, with Tether, USD Coin, and Binance USD being the most prominent examples. Managed by a centralized intermediary who controls the coins’ redemption and creation, and invests the underlying collateral into liquid, short-term assets (e.g., U.S. Treasury bills, short-term corporate debt, bank deposits), asset-backed stablecoins can be redeemed at any time for one dollar by liquidating these assets. Such a stablecoin is identical in its function to a deposit issued by a narrow bank, with a crucial distinction: the intermediary issuing the stablecoin is not regulated as a bank; it does not hold reserves at the central bank, does not have access to the discount window, and is not subject to liquidity or capital requirements. Moreover, investors typically do not know whether the composition of reserve assets is such that conversion at par in the case of largescale redemptions is guaranteed. This makes asset-backed stablecoins subject to runs—much 11 On the other hand, an economist from the gold standard era would feel similarly about an emerging market economy keeping the value of its currency constant by buying or selling U.S. dollars, which is also an exchange of two intrinsically worthless fiat currencies.

Banks 

141

like a bank that is not covered by deposit insurance—which has led to repeated calls for their regulation (Gensler, 2022). An obvious question here is why anyone would like to hold a digital token that can be exchanged for a dollar, rather than the dollar itself, perhaps in a bank account. One possible answer is cross-border payments, which are relatively easy with tokens, and slow and expensive if the transfer is done through banks.12 This convenience aside, so far there have been few obvious uses of stablecoins or any crypto “currencies.” Private currencies have circulated before, as exemplified by the original bank notes, but in an environment with a governmentmandated legal tender issued by a central bank, a viable private stablecoin has yet—and is unlikely—to take hold. One obvious way a stablecoin will circulate widely and in parallel with central bank money is if it is denominated in the national unit of account, which is a direct liability of the central bank. Such a digital payment instrument, a central bank digital currency (CBDC), would have parity with the existing currency by design—both are issued by the same central bank that commits to exchanging them at parity. While clearly feasible, whether CBDCs address any shortcomings in the financial system and what effects they may have in particular on banks are topics of current research. On the former, most central bank money is already digital, with the central bank issuing reserves as electronic entries and providing physical cash only on demand. Moreover, in many jurisdictions, most payment transactions involve the use of debit and credit cards or smartphone applications, with some establishments eschewing cash payments altogether. The latter question of what effect a CBDC will have on the banking industry depends on the form of the digital currency that may be implemented. In one possible implementation, the central bank is the keeper of the ledger for everyone, essentially allowing individual persons and firms to have accounts at the central bank.13 Such a retail CBDC represents a marked departure from the current system, in which only regulated depository institutions have access to the central bank’s balance sheet—individuals can own physical currency issued by the central bank and have accounts with depository institutions, but they do not have accounts with the central bank. This is an idea that goes back at least to Tobin (1985). Such a CBDC has the potential to provide access to banking services for those who do not have bank accounts, a feature that is especially attractive for low-income countries where a significant fraction of the population may be “unbanked.” Recent advances in distributed ledger technology, created in the context of cryptocurrencies, could be employed for central bank accounts, although given the centralized structure, this may not be needed. In the case of retail CBDC, the liquidity provision function (to depositors) of banks will shift to the central bank. This then begs the question of whether banking—without being able to issue demand deposits as liabilities—can survive as it has over the past half-millennium. In other words, if the narrow bank is taken out of the bank, can the remaining intermediary 12 The execution of international financial transactions and payments among banks generally takes place through the Society for Worldwide Interbank Financial Telecommunication (SWIFT), a Belgium-based cooperative of financial institutions throughout the world. When using SWIFT, cross-border payments are frequently routed through multiple banks before reaching their final destination, making their processing time significantly longer, as well as incurring additional costs, than those for domestic payments. 13 Although in many ways a substitute for paper money, such a CBDC would obviously lack the anonymity of physical cash.

142  Research handbook of financial markets

survive? Should it? By competing with bank deposits, CBDCs are going to reduce banks’ ability to fund their activities with low-cost, stable demand deposits. Worse still, the attractiveness of a CBDC, relative to any remaining bank deposits, would go up in crisis times, making bank runs more likely (Monnet et al., 2021). Meanwhile, the central bank is not going to make loans to nonfinancial companies and households on the asset side of its balance sheet. The central bank might extend credit to financial institutions, but this still increases its role in credit allocation in a way that the central bank (and the public) might not be comfortable with. Hence, a sudden switch from bank deposits to a CBDC could be very disruptive to financing investment. Keister and Sanches (2022) highlight a trade-off between the CBDC improving efficiency, but also crowding out bank deposits and thereby reducing investment. Fernández-Villaverde et al. (2021) consider a model in which the CBDC is immune to runs, but depositors internalize this feature, and the central bank becomes in effect a monopolist. Another way a CBDC can be structured is similar to the current system, in which only banks have accounts with the central bank, and all other agents have accounts with commercial banks (Auer & Böhme, 2020). This is the current structure, except that currency is replaced by wholesale CBDC. In this case, it is still banks that attract demand deposits, allowing narrow banking and project finance to co-exist. Banks understandably prefer this setup should a CBDC be issued, but the optimal mechanism design remains an open question. Suppose first that we consider the case where everyone can have an account at the central bank. In this case, if project finance continues as before—perhaps by banks that have evolved to shed narrow banking businesses or by other financial intermediaries with banks as we know them no longer in existence—then this will be efficiency-enhancing (neglecting privacy concerns). If, on the other hand, this leads to disintermediation because banks can no longer function and, as discussed earlier, their soft information cannot be replicated by other intermediaries, then this may have adverse real effects. Suppose instead that the path where only banks can have accounts at the central bank is taken. If banks are indeed special financial intermediaries that cannot exist without demand deposits, then this will be the more efficient form of CBDC. However, if narrow banking and project finance can be separated, but the CBDC system forces non-banks to only have accounts with commercial banks, then this would give unwarranted market power to those banks, which will be clearly welfare-reducing. Central banks are proceeding cautiously with the development and implementation of CBDCs that have the potential for major disruptions of the financial system. The Bahamas has already begun issuing a CBDC, and CBDCs may be introduced in larger advanced economies, with China conducting pilot studies for its CBDC, the e-CNY. Cecchetti and Schoenholtz (2021) worry that central banks will feel pressured to introduce CBDCs just because others are doing so. But central banks will likely limit the potential for moving money out of the banking system too suddenly. They could do so by limiting the amount that can be held in a CBDC; making the account non-interest bearing; and/or imposing fees when the account expands beyond a certain size. Keister and Sanches (2023) argue that a restricted form of CBDC is welfare-improving, while an unrestricted one is not. However, once a CBDC is in place, the central bank may come under pressure to relax restrictions. It may also prove difficult in practice to have a CBDC on a scale that is large, but not too large. While the pertinent issues are easy to articulate, in an environment where experimentation is not easy (one would not want to trigger a banking panic), coming up with robust

Banks 

143

welfare-enhancing implementations of CBDCs, to the extent that they address existing inefficiencies, requires much economic research. Many central banks and international financial institutions, including the Fed, Bank of England, the ECB, and the BIS, have working groups on these questions, and a number of them have issued white papers outlining digital monetary systems that will serve the public interest (European Central Bank, 2020; Federal Reserve Board, 2020; Bank for International Settlements, 2022). Last but not least, one quirk of CBDCs is their potential to be interest-bearing, which could improve the effectiveness of monetary policy to stabilize macroeconomic fluctuations. Cecchetti and Schoenholtz (2021) speculate that it might be politically difficult for central banks to pay interest on bank reserves but not on private accounts, should the accounts be with the central bank. In principle that interest rate could even be negative, overcoming the zero lower bound problem, although that is unlikely to be politically palatable. The politics and economics of a positive interest rate paid to holders of a CBDC are also interesting. How different would that be from paying interest on reserves? What are the fiscal implications? The feasibility, design, and desirability of CBDCs for economies with already functioning currencies and financial systems remain active areas of research.

6.8 CONCLUSIONS Banking is a remarkably durable form of financial intermediation. In this chapter, we summarized the origins of banks and how they have evolved over time, as well as their defining characteristics, which have remained surprisingly stable. Banks create liquidity by issuing short-term deposits and holding long-term loans. This core function in return makes these institutions illiquid themselves and open to runs. This is a very peculiar arrangement. What efficiency gains are created by merging deposit-taking and loan-making in the same institution, whether the financial stability costs are worth these, and why firms need bank intermediation for financing by households have been and continue to be important research and policy questions. We have presented the main arguments in the literature and covered the work on the real effects of bank intermediation (or lack thereof), finally relating banking to stablecoins and possible central bank digital currencies. In so doing, we hope to have presented a bird’s-eye view of the vast literature on banking, which has also laid bare the open questions on the topic.

REFERENCES Albertazzi, U., & Bottero, M. (2014). Foreign bank lending: Evidence from the global financial crisis. Journal of International Economics, 92, S22–S35. Auer, R., & Böhme, R. (2020). The technology of retail central bank digital currency. BIS Quarterly Review, March. Bagehot, W. (1873). Lombard street. Henry S. King & Co., London. Banerjee, R. N., Gambacorta, L., & Sette, E. (2021). The real effects of relationship lending. Journal of Financial Intermediation, 48, 1–14. Bank for International Settlements. (2022). The future monetary system. Annual Economic Report, Chapter III. Bassett, W. F., Chosak, M. B., Driscoll, J. C., & Zakrajšek, E. (2014). Changes in bank lending standards and the macroeconomy. Journal of Monetary Economics, 62(C), 23–40.

144  Research handbook of financial markets

Berger, A. N., & Bouwman, C. H. (2009). Bank liquidity creation. Review of Financial Studies, 22(9), 3779–3837. Berger, A. N., Bouwman, C., Kick, T., & Schaeck, K. (2016). Bank liquidity following regulatory interventions and capital support. Journal of Financial Intermediation, 26, 115–141. Berger, A. N., Dai, Q., Ongena, S., & Smith, D. C. (2003). To what extent will the banking industry be globalized? A study of bank nationality and reach in 20 European nations. Journal of Banking and Finance, 27(3), 383–415. Berger, A. N., Demsetz, R. S., & Strahan, P. E. (1999). The consolidation of the financial services industry: Causes, consequences and implications for the future. Journal of Banking and Finance, 23(2–4), 135–194. Berger, A. N., & Sedunov, J. (2017). Bank liquidity creation and real economic output. Journal of Banking and Finance, 81, 1–19. Berger, A. N., & Udell, G. F. (1995). Relationship lending and lines of credit in small firm finance. Journal of Business, 351–381. Berlin, M., & Mester, L. J. (2015). Deposits and relationship lending. Review of Financial Studies, 12(3), 579–607. Bernanke, B. S. (1983). Non-monetary effects of the financial crisis in the propagation of the Great Depression. American Economic Review, 73, 257–276. Bernanke, B. S., & Blinder, A. S. (1992). The federal funds rate and the channels of monetary transmission. American Economic Review, 82, 901–921. Berndt, A., & Gupta, A. (2009). Moral hazard and adverse selection in the originate-to-distribute model of bank credit. Journal of Monetary Economics, 56(5), 725–743. Black, F. (1975). Bank funds management in an efficient market. Journal of Financial Economics, 2(4), 323–339. Bofondi, M., Carpinelli, L., & Sette, E. (2017). Credit supply during a sovereign debt crisis. Journal of the European Economic Association, 16(3), 696–729. Bolton, P., Freixas, X., Gambacorta, L., & Mistrulli, P. E. (2016). Relationship and transaction lending in a crisis. Review of Financial Studies, 29(10), 2643–2676. Boot, A. W. (2000). Relationship lending: What do we know? Journal of Financial Intermediation, 9(1), 7–25. Boot, A. W., & Thakor, A. V. (2000). Can relationship banking survive competition? Journal of Finance, 55(2), 679–714. Boot, A., Hoffmann, P., Laeven, L., & Ratnovski, L. (2021). Fintech: What’s old, what’s new? Journal of Financial Stability, 53, 1–13. Bord, V. M., & Santos, J. A. (2015). Does securitization of corporate loans lead to riskier lending? Journal of Money, Credit, and Banking, 47(2–3), 415–444. Bordo, M. D., & Haubrich, J. G. (2010). Credit crises, money, and contractions: An historical overview. Journal of Monetary Economics, 57(1), 1–18. Calomiris, C. W., Himmelberg, C. P., & Wachtel, P. (1995). Commercial paper, corporate finance, and the business cycle: A microeconomic perspective. Carnegie-Rochester Conference Series on Public Policy, 42(1), 203–250. Calomiris, C. W., & Kahn, C. M. (1991). The role of demandable debt in structuring optimal banking arrangements. American Economic Review, 81, 497–513. Cecchetti, S., & Schoenholtz, K. (2021). Central Bank digital currency: The battle for the soul of the financial system. VoxEU​.or​g, 8 July. Cetorelli, N., & Goldberg, L. S. (2012). Banking globalization and monetary transmission. Journal of Finance, 67(5), 1811–1843. Chodorow-Reich, G. (2014). The employment effects of credit market disruptions: Firm-level evidence from the 2008–9 financial crisis. Quarterly Journal of Economics, 129(1), 1–59. Cingano, F., Manaresi, F., & Sette, E. (2016). Does credit crunch investment down? New evidence on the real effects of the bank-lending channel. Review of Financial Studies, 29(10), 2737–2773. Cowen, D., & Sylla, R. (2018). Alexander Hamilton on finance, credit, and debt. New York: Columbia University Press. De Fiore, F., & Uhlig, H. (2011). Bank finance versus bond finance. Journal of Money, Credit, and Banking, 43(7), 1399–1421.

Banks 

145

De Haas, R., & Van Horen, N. (2013). Running for the exit? International bank lending during a financial crisis. Review of Financial Studies, 26(1), 244–285. Degryse, H., & Ongena, S. (2005). Distance, lending relationships and competition. Journal of Finance, 60(1), 231–266. Degryse, H., & Van Cayseele, P. (2000). Relationship lending within a bank-based system: Evidence from European small business data. Journal of Financial Intermediation, 9(1), 90–109. Demirgüç-Kunt, A., Kane, E., & Laeven, L. (2015). Deposit insurance around the world: A comprehensive analysis and database. Journal of Financial Stability, 20, 155–183. Diamond, D. W. (1984). Financial intermediation and delegated monitoring. Review of Economic Studies, 51(3), 393–414. Diamond, D. W., & Dybvig, P. (1983). Bank runs, deposit insurance and liquidity. Journal of Political Economy, 91(3), 401–419. Diamond, D. W., & Rajan, R. G. (2001). Liquidity risk, liquidity creation and financial fragility: A theory of banking. Journal of Political Economy, 109(2), 287–327. English, W. B., Van den Heuvel, S. J., & Zakrajšek, E. (2018). Interest rate risk and bank equity valuations. Journal of Monetary Economics, 98(C), 80–97. European Central Bank. (2020). Report on a digital euro. Available at: https://www.ecb.europa.eu/pub/ pdf/other/Report_on_a_digital_euro~4d7268b458.en.pdf. Fama, E. F. (1980). Banking in the theory of finance. Journal of Monetary Economics, 6(1), 39–57. Fama, E. F. (1985). What’s different about banks? Journal of Monetary Economics, 15(1), 29–39. Federal Reserve Board. (2020). Money and payments: The U. S. Dollar in the age of digital transformation. Available at: https://www.federalreserve.gov/publications/files/money-and-payments-20220120.pdf Fernández-Villaverde, J., Sanches, D., Schilling, L., & Uhlig, H. (2021). Central Bank digital currency: Central banking for all? Review of Economic Dynamics, 41, 225–242. Focarelli, D., Pozzolo, A. F., & Casolaro, L. (2008). The pricing effect of certification on syndicated loans. Journal of Monetary Economics, 55(2), 335–349. Friedman, M., & Schwartz, A. J. (1963). A monetary history of the United States, 1867–1960. Princeton, NJ: Princeton University Press. Frost, J., Gambacorta, L., Huang, Y., Shin, H. S., & Zbinden, P. (2019). BigTech and the changing structure of financial intermediation. Economic Policy, 34(100), 761–799. Gatev, E., & Strahan, P. E. (2006). Banks’ advantage in hedging liquidity risk: Theory and evidence from the commercial paper market. Journal of Finance, 61(2), 867–892. Gensler, G. (2022). Prepared remarks of Gary Gensler on crypto markets. Penn Law Capital Markets Association Annual Conference. Gertler, M., & Gilchrist, S. (1994). Monetary policy, business cycles and the behavior of small manufacturing firms. Quarterly Journal of Economics, 109(2), 309–340. Giannetti, M., & Laeven, L. (2012). Flight home, flight abroad, and international credit cycles. American Economic Review, 10(3), 219–224. Hoggson, N. F. (1926). Banking through the ages. Dodd, Mean & Co. Hoshi, T., Kashyap, A., & Scharfstein, D. (1990). The role of banks in reducing the costs of financial distress in Japan. Journal of Financial Economics, 27(1), 67–88. Ippolito, F., Peydro, J.-L., Polo, A., & Sette, E. (2016). Double bank runs and liquidity risk management. Journal of Financial Economics, 122. Ivashina, V. (2009). Asymmetric information effects on loan spreads. Journal of Financial Economics, 92(2), 300–319. James, C. (1987). Some evidence on the uniqueness of bank loans. Journal of Financial Economics, 19(2), 217–236. Jefferson, T. (1816). Letter to John Taylor, 28 May. Jiménez, G., Ongena, S., Peydró, J.-L., & Saurina, J. (2017). Macroprudential policy, countercyclical bank capital buffers, and credit supply: Evidence from the Spanish dynamic provisioning experiments. Journal of Political Economy, 125(6), 2126–2177. Kashyap, A. K., Rajan, R. G., & Stein, J. C. (2002). Banks as liquidity providers: An explanantion for the coexistence of lending and deposit-taking. Journal of Finance, 57(1), 33–73. Kashyap, A. K., & Stein, J. C. (2000). What do a million observations on banks say about the transmission of monetary policy? American Economic Review, 90(3), 407–428.

146  Research handbook of financial markets

Kashyap, A. K., Stein, J. C., & Wilcox, D. W. (1993). Monetary policy and credit conditions: Evidence from the composition of external finance. American Economic Review, 83, 78–98. Keister, T., & Sanches, D. R. (2022). Should central banks issue digital currency? Review of Economic Studies, 90(1), 404–431. Keys, B. J., Mukherjee, T., Seru, A., & Vig, V. (2010). Did securitization lead to lax screening? Evidence from subprime loans. Quarterly Journal of Economics, 125(1), 307–362. Lang, W. W., & Nakamura, L. I. (1995). “Flight to quality” in banking and economic activity. Journal of Monetary Economics, 36(1), 145–164. Leland, H. E., & Pyle, D. H. (1977). Informational asymmetries, financial structure and financial intermediation. Journal of Finance, 32, 371–387. Levine, R., & Zervos, S. (1998). Stock markets, banks, and economic growth. American Economic Review, 537–558. Lummer, S. L., & McConnell, J. J. (1989). Further evidence on the bank lending process and the capitalmarket response to bank loan agreements. Journal of Financial Economics, 25(1), 99–122. Mester, L. J., Nakamura, L. I., & Renault, M. (2006). Transactions of the accounts and loan monitoring. Review of Financial Studies, 20, 529–556. Modigliani, F., & Miller, M. H. (1958). The cost of capital, corporation finance and the theory of investment. American Economic Review, 48, 261–297. Monnet, E., Riva, A., & Ungaro, S. (2021). Bank runs and central bank digital currency. VoxEU​.or​g, 3 May. Mookherjee, D., & Png, I. (1989). Optimal auditing, insurance, and redistribution. Quarterly Journal of Economics, 104(2), 399–415. Morgan, D. P. (1998). The credit effects of monetary policy: Evidence from using loan commitments. Journal of Money, Credit, and Banking, 30(1), 102–118. Ongena, S., & Smith, D. C. (2001). The duration of bank relationships. Journal of Financial Economics, 61(3), 449–475. Paravisini, D., Rappoport, V., Schnabl, P., & Wolfenzon, D. (2014). Dissecting the effect of credit supply on trade: Evidence from matched credit-export data. Review of Economic Studies, 82(1), 333–359. Parlour, C. A., & Plantin, G. (2008). Loan sales and relationship banking. The Journal of Finance, 63(3), 1291–1314. Peek, J., & Rosengren, E. S. (2000). Collateral damage: Effects of the Japanese bank crisis on real activity in the United States. American Economic Review, 90(1), 30–45. Pennacchi, G. (2012). Narrow banking. Annual Review of Financial Economics, 4(1), 141–159. Petersen, M. A., & Rajan, R. G. (1994). The benefits of lending relationships: Evidence from small business data. Journal of Finance, 49(1), 3–37. Petersen, M. A., & Rajan, R. G. (1995). The effect of credit market competition on lending relationships. Quarterly Journal of Economics, 110(2), 407–443. Petersen, M. A., & Rajan, R. G. (2002). Does distance still matter? The information revolution in small business lending. Journal of Finance, 57(6), 2533–2570. Popov, A., & Van Horen, N. (2014). Exporting sovereign stress: Evidence from syndicated bank lending during the euro area sovereign debt crisis. Review of Finance, 19(5), 1825–1866. Slovin, M. B., Sushka, M. E., & Polonchek, J. A. (1993). The value of bank durability: Borrowers as bank stakeholders. Journal of Finance, 48(1), 247–266. Stein, J. C. (2002). Information production and capital allocation: Hierarchical vs. decentralized firms. Journal of Finance, 57, 1891–1921. Stiglitz, J. E., & Weiss, A. (1981). Credit rationing in markets with imperfect information. American Economic Review, 71, 393–410. Tobin, J. (1985). Financial innovation and deregulation in perspective. Bank of Japan Monetary and Economic Studies, 3, 19–29. Townsend, R. M. (1979). Optimal contracts and competitive markets with costly state verification. Journal of Economic Theory, 21(2), 265–293. Tümmler, M. (2022). Completing banking union? The role of national deposit guarantee schemes in shifting member states’ preferences on the European deposit insurance scheme. Journal of Common Market Studies. https://doi​.org​/10​.1111​/jcms​.13318. Van den Heuvel, S. J. (2008). The welfare costs of bank capital regulation. Journal of Monetary Economics, 55(2), 298–320.

7. Non-bank financial intermediaries and financial stability1 Sirio Aramonte, Andreas Schrimpf and Hyun Song Shin

The ecosystem that supports financial intermediation has evolved markedly over time. Banks and their affiliated broker-dealers remain a vital component, but they are now part of a larger mosaic of institutions that route the flow of funds and facilitate trading. Especially important, from a financial stability perspective, has been the greater involvement of non-bank financial intermediaries (NBFIs).1 The NBFI landscape is vast and varied, covering a diverse set of players with a number of business models and subject to different regulatory regimes. The boundaries of which activities are the purview of an NBFI as opposed to a bank’s can at times be blurred, and so can be the delineations among NBFIs. Figure 7.1 provides a brief overview of the main NBFIs and of their relevance for financial stability. NBFIs also differ in how they are interconnected with other players in the system, especially with banks. In contrast to banks, NBFIs have historically not been granted access to statuary public backstops, with either no (or only limited) access to the central bank balance sheet. To keep our analysis focused, we restrict the coverage of this paper to NBFIs that matter the most for market liquidity (which refers to the ease with which a security can be traded) (Brunnermeier and Pedersen, 2009). We mainly consider non-banks that – through business models involving liquidity mismatches and/or the use of leverage – are most likely to contribute to liquidity imbalances that can endanger financial stability. These institutions typically handle intermediate debt, which is generally less liquid than equity. Practically, this means that we focus on entities such as principal trading firms, hedge funds and asset managers of various types (notably money market funds and open-ended bond funds) as well as central counterparties (CCPs).2 The growing role of non-banks and market-based intermediation over the past decade has been driven by various factors. Key elements include regulatory reforms that constrained 1 We are grateful to Refet Gürkaynak and Jonathan Wright (the editors), Matteo Aquilina, Fernando Avalos, Adrien d’Avernas, Christian Cabanilla, Steve Cecchetti, Stijn Claessens, Ben Cohen, Dobrislav Dobrev, Mathias Drehmann, Egemen Eren, Monika Piazzesi, Andrew Hauser, Wenqian Huang, Semyon Malamud, Antoine Martin, Jay Kahn, Ilhyock Shim, Nikola Tarashev, Karamfil Todorov, Quentin Vandeweyer, Goetz von Peter, Phil Wooldridge and seminar and conference participants at the Office of Financial Research, the Federal Reserve Board, the Bank for International Settlements (BIS), the Research Handbook of Financial Markets Conference and the Swedish House of Finance for helpful comments. This article represents the views of the authors and not those of the BIS or other members of its staff. Research support by Alan Villegas and Cornelius Nicolay is gratefully acknowledged. 2 The intermediaries we do not cover could be subsumed under shadow banking (e.g. securitisations) or market-based finance (e.g. pension funds and other long-horizon investors). See, inter alia, Pozsar et al. (2010) for a comprehensive review and Adrian (2017) for a taxonomy of shadow banking and market-based finance. 147

148

Central counterparties*

trading platforms*

Exchanges & electronic

Principal trading firms*

Broker-dealers*

Securitisations†

Mutual funds*†

Exchange-traded funds*

Hedge funds*

Sovereign wealth funds

Pension funds

Insurance companies

Intermediaries

Liquidity transformation (if open ended), possibly

Shares can be redeemed daily even if underlying assets are

contracts, netting and managing counterparty risk

They act as counterparties to holders of certain financial

contracts like derivatives

Marketplaces for trading securities and/or financial

holding minimal end-of-day inventories

High-frequency buyers and sellers in electronic markets,

trades. They often enable leverage for their clients

They use relationships or own inventory to facilitate client

notes with different seniority, including AAA-rated

They invest in various assets, possibly risky, and issue

illiquid (if open-ended, incl. money-market funds)

in initial margins, technical disruptions

Pro-cyclicality in market-wide leverage due to changes

risks) could affect broader financial markets

Technical disruptions (eg, due to operational or cyber

Pro-cyclicality in liquidity provision, intra-day leverage

Leverage, liquidity transformation

Credit-risk transformation

leverage

redemption mechanism)

Some liquidity transformation (limited by the

Shares trade in secondary markets and are generally redeemed in-kind only by selected intermediaries

redemption notices)

Leverage, some liquidity transformation (limited by

Possibly leverage

Some credit-risk transformation

Some leverage, some liquidity transformation

Main systemic risks

deployed through strategies that may involve arbitrage

Investors' capital is augmented with leverage and

focused on long-term illiquid assets

Vehicles managed by state-affiliated entities, often

public-market and private-market assets

Contributions by participants are invested in a mix of

various assets, often long-lived and illiquid

Premia collected from insured parties are invested in

Key characteristics from a financial-stability perpsective

Figure 7.1  Overview of main NBFIs and related financial-stability risks

Notes:  The table shows key NBFIs, together with a short description of their characteristics relevant to financial stability, together with the attending financial stability risks.

(*) asterisks indicate intermediaries that can affect imbalances in the demand and supply of financial market liquidity more directly, and that we focus on in this paper (†) entities engaged in elevated liquidity or credit-risk tranformation, such as most money-market funds or certain securitisations, are often considered shadow banks (eg, Adrian (2017))

infrastructures

Financial market

Market intermediaries

and asset managers

Institutional investors

Broad categories

Non-bank financial intermediaries and financial stability  149

the activities of banks and their affiliated broker-dealers, demographic changes and greater importance of capital markets in providing for retirement, as well as technological change and the pursuit of operational efficiencies. In some cases, it was also active policy choices that strengthened the role of certain NBFIs. Notably, this includes the promotion of financial infrastructures such as electronic trading platforms and the strong push to move activity towards CCPs, which was aimed at reducing opaqueness and addressing vulnerabilities in over-thecounter (OTC) markets. These developments, which have turned NBFIs into indispensable building blocks of the financial system, have also had a profound impact on the demand and supply of liquidity. The management of liquidity risk has arguably gained importance from a financial stability point of view. On the one hand, the NBFI sector itself has become a key source of spikes in liquidity demand, particularly from investment funds exposed to liquidity mismatches, such as money-market and bond funds. On the other hand, the supply of liquidity is no longer the exclusive domain of bank dealers alone, but it increasingly involves NBFIs as well. Cases in point are the activity of principal trading firms (PTFs) in electronic markets and the trading strategies of certain hedge funds. Yet, as several recent episodes have shown, liquidity provision by non-banks tends to be more opportunistic and more prone to evaporate at times of stress, with entities that generally provide liquidity suddenly turning into liquidity consumers. And, as broker-dealers have reassessed their business models and scaled back market-making activities, the supply of liquidity has become less responsive when the demand for it spikes. These structural shifts mean that liquidity imbalances have the potential to greatly affect prices and, in extreme cases, endanger financial stability. The “dash for cash” turmoil at the height of the Covid-19 crisis (when investors shifted away from risky assets to cash-like assets on a massive scale) painfully exposed such structural NBFI vulnerabilities and spillovers that affected other participants in the financial system. Ultimately, it was only central banks’ flexible use of their balance sheets that arrested the adverse feedback loops and helped to restore market functioning. Building on our analysis of the key changes in intermediation and their implications for liquidity imbalances, we lay out a stylised “accounting framework” for system-wide risk capacity. In this setting, an investor can take on a leveraged position through derivatives or by pledging the assets as collateral. However, the borrowing is subject to a margin that must be met by the investor’s own funds – that is, equity. The total amount of posted margin is bounded by the economic capital of the investor, which in turn is limited by the investor’s equity. In this way, the investor’s portfolio choice entails the allocation of scarce economic capital across assets. Within the framework, we derive two propositions. The first is that the debt capacity of an investor is increasing in the debt capacity of other investors. In this sense, debt capacity is recursive, and leverage enables greater leverage. Conversely, spikes in margins can lead to system-wide deleveraging. The second proposition is that deleveraging and the “dash for cash” go hand in hand, as a generalised increase in margins in the financial system leads both to deleveraging and to the re-allocation of economic capital away from assets with high margins toward cash-like assets with low margins. The deleveraging channel and the associated pecuniary externalities – i.e. externalities that operate through prices and risk measures based on prices – can be important for stress propagation, adding to the effect of other sources of systemic risk such as liquidity transformation. Importantly, stress can propagate in the system even in the absence of defaults. We use

150  Research handbook of financial markets

this risk accounting framework to provide a unifying perspective on the liquidity imbalances that rocked financial markets in March 2020, amid the uncertainty shock of the Covid-19 pandemic.

7.1 NBFIs AND THE EVOLUTION OF MARKET-BASED INTERMEDIATION By some estimates, NBFIs currently account for about 50% of global financing activities (FSB, 2020a). Back in the 1980s, banks funded about 30% of non-mortgage US corporate debt through loans, but this figure has fallen now to 10%.3 Consistent with the greater presence of market-based finance, bonds and commercial paper currently constitute the bulk of US corporate debt, at roughly 65%. Non-banks have always been the main investors in these securities, but their role has further expanded after the Great Financial Crisis (GFC). Mutual funds, insurance companies, and pension funds hold nearly 80% of corporate and foreign bonds as of 2020, with a pronounced increase for mutual funds. Similar trends have emerged internationally, and various types of asset managers, in particular, play an increasing role in financing the real economy. As the activities of NBFIs often involve significant mismatches in the liquidity of assets and liabilities, the scope for liquidity demand pressures has grown correspondingly. At the same time, the supply of liquidity by traditional intermediaries, i.e. broker-dealers, has not kept up with rising demand. Broker-dealers are institutions that facilitate trading by other investors.4 While they can be independent firms, they often form part of banking groups (“dealer banks”) and are subject to applicable capital and prudential regulations on a consolidated basis. The bond holdings of broker-dealers have shrunk after the GFC, even as the overall market expanded. This trend stands in sharp contrast with pre-GFC dynamics, when they played a crucial role in driving the shift from a bank-centric financial system towards a market-based one, as attested by the ten-fold expansion of their balance sheets between 1990 and 2008 and the corresponding increase in leverage (Figure 7.2). In general, higher leverage need not correspond to larger balance sheets (Figure 7.3, top panel), but broker-dealers clearly used debt to finance asset growth (Figure 7.3, bottom panel; see Adrian and Shin, 2014). As we highlight in the following, margins are a crucial source of fluctuations in the leverage of market intermediaries: for a fixed amount of own funds, a compression in margin requirements allows market participants to maintain a larger balance sheet. Margins tend to increase rapidly during periods of distress when volatility, their main driver, spikes. The rise in margins can, in turn, create knock-on effects impacting other players in the financial system and give rise to procyclicality (see in particular, BCBS, CPMI and IOSCO, 2021 and Section 7.4). Due to regulatory tightening and a market-driven reassessment of business models, broker-dealer balance sheets now have a significantly smaller heft in the financial system than pre-GFC. Leverage supporting these balance sheets has also come down significantly. Important shifts in dealers’ market-making business models have accompanied such trends (CGFS, 2014, 2016). The “principal” model, where dealer banks use balance sheet capacity to accommodate client trading demands, has given way to a model where they primarily match 3 Financial Accounts of the United States, Table L.103 4 Typical services include matching transactions between clients wishing to trade in opposite directions; accommodating customer trades using an inventory of securities financed with the dealers’ liabilities; providing clients with leverage; and clearing and collateral-management services.

Non-bank financial intermediaries and financial stability  151

Notes:   The size of the broker-dealer sector is illustrated using log-assets (top panel) and leverage (bottom panel). Shaded areas indicate the December 2007–June 2009 recession.

Figure 7.2  Evolution of assets and leverage of broker-dealers

clients wishing to trade in opposite directions (see e.g. Adrian, Boyarchenko and Shachar, 2017). One important consequence of the principal model’s retrenchment is that liquidity provision has moved increasingly outside of the broker-dealer sector, in favour of a broader set of players. Two types of entities stand out: principal trading firms (PTFs), which facilitate the redistribution of risk by buying and selling securities while keeping minimal inventories, and

152  Research handbook of financial markets

Three ways of increasing leverage Mode 1: Increased leverage due to equity buyback

A Assets

L Equity

A

L

Mode 2: Increased leverage due to fall in asset value

A

Equity Assets

Debt

Debt

L Equity

Assets Debt

A Assets

L

Equity Debt

Mode 3: Decline in margin

A

L

A

Equity

Equity

Assets Debt

L

Assets Debt

Broker-dealers’ balance sheet expansion driven one-to-one by increase in debt

Notes:   The top panel illustrates alternative ways in which leverage can be increased and balance sheets expanded or contracted. The bottom panel plots quarterly changes in broker-dealer assets against changes in their debt and equity. Source:   Data are from the Financial Accounts of the United States.

Figure 7.3  Leverage and changes to broker-dealers’ balance sheets

Non-bank financial intermediaries and financial stability  153

certain types of hedge funds, which effectively warehouse risks (Eren and Wooldridge, 2021). We briefly describe the salient characteristics of both intermediaries in the following. PTFs, which are more lightly regulated than broker-dealers, are sometimes referred to as the “new electronic market makers” (Menkveld, 2013), since many of them pursue passive market-making strategies in electronic markets. They hold relatively limited capital and trade on their own account, typically using automated high-frequency strategies (Markets Committee, 2018). PTFs and bank dealers compete on liquidity provision in financial markets, but there also tends to be a symbiotic relationship in that large broker-dealers usually act as prime brokers for PTFs. Prime brokerage enables firms such as PTFs to conduct trades with a group of predetermined third-party wholesale counterparties in the prime broker’s name and use its credit (see e.g. Schrimpf and Sushko, 2019; Treasury Markets Practices Group, 2019). PTFs first rose to prominence in exchange-traded equities and futures, but have subsequently made inroads into traditional OTC markets. They currently account for the bulk of trading volumes on electronic trading platforms for on-the-run US Treasury securities, futures contracts, equities, spot FX, as well as certain classes of derivatives.5 PTFs generate large amounts of short-lived orders, with very tight inventory control (Adrian et al., 2020). In normal times, PTFs help incorporate information into prices and distribute risk among market participants. In periods of stress, however, questions remain about their true risk-bearing capacity, since their business model involves minimal risk warehousing. In addition, PTFs trade anonymously on electronic markets, hence they have no client relationships at stake.6 Several recent episodes of market dysfunction indicate that PTFs tend to scale down liquidity provision when volatility spikes (see e.g. Dobrev and Meldrum, 2020; Aronovich, Dobrev and Meldrum, 2021 for an analysis of the evaporation of high-speed liquidity provision in the US Treasury market). Hedge funds are the second set of players to have gained prominence in liquidity supply, typically complementing broker-dealers. Large hedge funds have global operations and usually rely on multiple prime brokers, highlighting the close interconnections with systemically important banks. Hedge funds’ trading in fixed-income markets often involves exploiting small mispricings between similar instruments, such as cash bonds and futures contracts. To profit from these “relative-value” opportunities, hedge funds take significant leverage, often through repo borrowing that is facilitated by bank-affiliated broker-dealers. Under normal conditions, hedge funds’ activity adds to liquidity, but sudden deleveraging forcing an unwind of positions can reverberate through financial markets, turning opportunistic liquidity provision into large liquidity demand. One such case was the stress in US Treasury markets in March 2020, when the so-called Treasury cash-futures basis trade ground to a halt (see e.g. Barth and Kahn, 2020; Duffie, 2020; Hauser, 2020; Schrimpf, Shin and Sushko, 2020; Kruttli et al., 2021).7 Typically, the futures-implied bond price is higher, reflecting that a futures contract consumes only limited 5 In other important market segments, such as non-US government bond markets or corporate bond markets, PTFs have not made significant inroads yet. 6 In recent years, a few PTFs have gained ground in traditional OTC segments by building client relationships through trades in which they disclose their identities (Schrimpf and Sushko, 2019). 7 The cash-futures basis trade is just one example of many other relative-value trading strategies of hedge funds in fixed income markets. There are multiple other relative-value trades too, e.g. involving interest rate swaps rather than futures, or trades that benefit from mispricing of securities along the yield curve.

154  Research handbook of financial markets

balance-sheet capacity when it is entered into. In contrast, the equivalent cash bond entails balance-sheet usage. In the run-up to the turmoil in March 2020, hedge funds would purchase relatively illiquid off-the-run Treasury securities and sell Treasury futures. This trade, which accounted for up to 25% of dealers’ repo volumes in 2019, meant that hedge funds effectively warehoused the liquidity risk embedded in cash bonds (Barth and Kahn, 2021). Our overview of key changes in market-based intermediation would not be complete without emphasising the role of financial market infrastructures such as central counterparties (CCPs), exchanges and other platforms. We illustrate this in the flow chart of Figure 7.4, where market infrastructures are placed at the centre of the stylised market-intermediation ecosystem. While the rise of NBFIs has generally occurred in an evolutionary manner, the growth of CCPs owes to an active policy push to reform the notoriously opaque OTC derivatives markets, whose vulnerabilities had been exposed during the GFC (see Borio, Farag and Tarashev, 2020). The crucial role of CCPs in many of today’s markets is hard to overstate. CCPs act as intermediate agents that, thanks to netting, reduce overall counterparty risk between the holders of a contract. They also increase capital efficiency, since they generally reduce posted margins compared to uncleared transactions. In certain markets, including interest rate derivatives and, to a lesser extent, credit derivatives, CCPs currently support the majority of positions outstanding. As a result, CCPs’ risk management practices of their exposures to clearing members take centre stage from a financial stability standpoint (Huang, Menkveld and Yu,

Notes:   The figure depicts the structure of the market-intermediation ecosystem. It is organised as a flow chart, with the ultimate savers at the right and the ultimate borrowers at the left. Managing the households’ savings are various types of institutional investors (hedge funds, asset managers and money-market funds). Market intermediaries include broker-dealers and principal trading firms (the latter typically trading in financial markets through a prime brokerage relation). At the centre of the diagram are financial market infrastructures, i.e. exchanges, electronic trading platforms and central counterparties (CCPs). The lines connecting the different boxes represent financial flows between the various entities (e.g. repos and reverse repos, securities or derivatives transactions).

Figure 7.4  Stylised view of the market-intermediation ecosystem post-GFC

Non-bank financial intermediaries and financial stability  155

2021). In particular, they can sometimes exacerbate system-wide liquidity needs. Specifically, margins are currently set in order to manage the counterparty risk faced by CCPs, but margin fluctuations have broad repercussions that can affect the risk-taking capacity of the financial system as a whole – a central theme we will come back to repeatedly in this chapter (in particular, when discussing policy implications in Section 7.5). All in all, the greater role of NBFIs means that risk exposures are increasingly intermediated and held outside of the banking system.8 A key theme in what follows is that such structural changes have alleviated counterparty credit risk, but have rendered the financial system more vulnerable to large swings in liquidity imbalances, with potential effects on credit availability in the real economy. The reason is that the business models of NBFIs typically revolve around exploiting liquidity mismatches and, on net, provide liquidity in good times. During periods of financial turmoil, however, NBFIs often retrench and their liquidity supply can suddenly turn into substantial liquidity demand.

7.2 SYSTEMIC RISK IN NBFIs AND LIQUIDITY DEMAND NBFIs bring a range of benefits to the financial system and the economy. They increase the diversity of the ecosystem, improving market functioning especially when their trading motives are less correlated with those of other players. In particular, some NBFIs may pick up the slack when banks retrench from certain intermediation activities. PTFs and hedge funds have indeed helped deepen market efficiency and liquidity in good times, complementing the role of bank dealers. But, NBFIs can also contribute materially to systemic risk. Systemic risk refers to the possibility that disruptions to the activity of an intermediary could impose substantial costs, chiefly in the form of externalities, on other financial institutions or non-financial firms (Acharya et al., 2017). For instance, spikes in margins and haircuts reduce the risk-bearing capacity that NBFIs can sustain, potentially depressing prices and impairing market liquidity. Addressing systemic risk requires a macroprudential perspective. Focusing on investor protection or on the soundness of individual institutions, as regulations have done until recently, largely means issuing enough loss-absorbing liabilities to minimise default risk. From this microprudential vantage point, the main concern is managing the risks that stem from assets. An important feature of macroprudential policies is their focus on the liability side as much as the asset side because it is the interaction of the two that generates systemic risk (Morris and Shin, 2008). This can be exemplified by the case of a corporate bond mutual fund that offers daily redemptions even if its illiquid holdings may take much longer to sell. If liabilities were not redeemable on short notice, the systemic risk would be much reduced, irrespective of asset illiquidity. NBFIs are mostly linked to two fundamental drivers of systemic risk that can lead to heightened demand for liquidity at times of stress. The first is liquidity/maturity transformation, and

8 This does not mean, of course, that banks and their affiliated broker-dealers have entirely withdrawn from such activities. For instance, banks often supply liquidity to large clients via margin lending or collateral transformation.

156  Research handbook of financial markets

the second is leverage procyclicality.9 The overlap between the systemic risks that characterise banks and NBFIs indicates a close correspondence between some of their economic functions, even as important differences remain, such as the more encompassing nature of bank intermediation. In the remainder of this section, we provide further details on the mapping between NBFI activities and sources of systemic risk, focusing in particular on entities that can give rise to substantial liquidity demand during market stress. 7.2.1 Mutual Funds While certain mutual funds can take some leverage (Boguth and Simutin, 2018), their most relevant feature, from a systemic-risk perspective, is that they offer daily redemptions also when their investments are illiquid. Redemptions are generally honored at fair value, even if the corresponding asset sales incur a liquidity discount, which is borne by the remaining shareholders. This setup gives rise to a first-mover advantage: expectations that a large number of investors might sell create an incentive to be among the first to redeem, potentially leading to full-fledged runs and disorderly fire sales that heighten liquidity demand (Chen, Goldstein and Jiang, 2010; Goldstein, Jiang and Ng, 2017).10 Since the assets managed by funds holding illiquid securities increased markedly after the GFC, the contribution of mutual funds to possible liquidity disruptions has climbed sharply over time. In principle, mutual fund managers can use various strategies to avoid incurring a liquidity discount. In particular, they often hold a buffer of easily tradeable securities that can be sold to meet outflows and are bought back over time as assets are disposed of. Many funds adopt this strategy (Chernenko and Sundaram, 2016; Aramonte, Scotti and Zer, 2020), especially in tranquil times (Jiang, Li and Wang, 2020a). In volatile periods and when portfolios are very illiquid, however, managers tend to sell more assets than needed, so to increase buffers in the face of possibly prolonged outflows – a practice known as “cash hoarding” that we discuss in more detail in Section 7.4. 7.2.2 Money Market Funds Among mutual funds, money market funds (MMFs) hold very short-term assets and issue shares that can be redeemed daily. We cover MMFs separately here due to their crucial role in the functioning of short-term funding markets (notably in the US dollar) and their heightened systemic importance.11 MMFs investing in non-public debt (so-called prime funds in the US and LVNAV and VNAV funds in Europe) use the proceeds to invest in certificates of deposit (CD) and   9 Another source of systemic risk in NBFIs is credit risk transformation. This activity entails the issuance of liabilities with a substantially different risk profile than the underlying assets. In principle, investment vehicles can be designed so that, irrespective of asset quality, some investors bear little credit risk and others face high probability of loss. See Gorton and Metrick (2013) for a detailed discussion. 10  Funds can suspend redemptions to protect shareholders (Section 22(e)(3) of the Investment Company Act of 1940) and can alter the redemption value based on flows – a tool known as “Swing Pricing” (Lewrick and Schanz, 2017; Jin et al., 2021), but such actions carry significant stigma. 11 MMFs are an important source of US dollar funding for non-US banks, and disruptions to the MMF sector can have significant spillover effects on FX swap markets and funding conditions for international banks (Eren, Schrimpf and Sushko, 2020a).

Non-bank financial intermediaries and financial stability  157

commercial paper (CP), which lack a developed secondary market and are typically held to maturity. In contrast, government MMFs hold short-dated Treasury securities or reverse repos backed by Treasury collateral. Because of the latter, they are crucial for the functioning of USD repo markets and an important source of funding for broker-dealers and hedge funds. Investors have come to expect that MMF liabilities are broadly equivalent to cash. This perceived equivalence to money emerges from relatively low yields (Cipriani and La Spada, 2021), even if MMFs generally have an incentive to purchase comparatively risky securities in order to attract investor flows (Kacperczyk and Schnabl, 2013). To avoid “breaking the buck”, MMF sponsors have often supported their funds with their own resources during stress episodes (Brady, Anadu and Cooper, 2012). Holding assets with variable prices and issuing liabilities with approximately stable value means that MMFs engage in liquidity transformation. A number of MMFs experienced runs over time, particularly during the GFC (Schmidt, Timmermann and Wermers, 2016). To reduce run risk, US regulations limit the types of assets certain MMFs can invest in. Additionally, funds can consider restricting redemptions if certain asset-liquidity thresholds are breached. Nonetheless, some run-like dynamics were at play for MMFs in March 2020 (see e.g. Li et al., 2021; Anadu et al., 2021), as regulatory liquidity thresholds may have led some investors to pre-emptively redeem to avoid the consequences of a fund crossing those thresholds. 7.2.3 Exchange-Traded Funds Similarly to mutual funds, exchange-traded funds (ETFs) allow investors to gain exposure to illiquid assets. The redemption mechanism, however, is fundamentally different (Ben-David, Franzoni and Moussawi, 2018; Todorov, 2019). ETFs hold portfolios of securities financed with the issuance of shares that can be traded continuously but can only be redeemed by specialised intermediaries known as Authorised Participants (APs). Trading pressure from ETF investors can open a wedge between the price of ETF shares and the value of the underlying portfolio. Unless this gap reflects fundamental differences in price informativeness (Aramonte and Avalos, 2020), Aps engage in arbitrage that involves creating and selling ETF shares or redeeming previously purchased ETF shares. This mechanism implies that investors selling ETF shares bear any liquidity discount incurred when assets are disposed of, which in turn may mitigate incentives to redeem early (Shim and Todorov, 2021). ETFs can still affect underlying prices through certain channels (Ben-David, Franzoni and Moussawi, 2018), such as higher informational efficiency (Glosten, Nallareddy and Zou, 2021) and portfolio rebalancing (Todorov, 2019). However, the nature of their liabilities is arguably less prone to systemic issues than that of open-end mutual funds holding illiquid assets. 7.2.4 Hedge Funds The liabilities of hedge funds have two key characteristics that are relevant to our discussion. The first is that investors are typically subject to relatively long notice periods before they can redeem their interest. For about half of hedge funds’ assets, the notice must be submitted at least 90 days in advance, limiting issues arising from liquidity transformation.12 The second is that hedge funds can be highly leveraged, often with credit provided by prime brokers through repos and/or synthetically through the use of derivatives. Hedge funds have grown from niche

12 Securities and Exchange Commission, Private Funds Statistics, Second Calendar Quarter 2019.

158  Research handbook of financial markets

investment vehicles to a large sector that complements traditional intermediaries like mutual funds (Stulz, 2007). The ratio of total assets to net assets, a proxy for leverage, increased markedly between the mid-2010s and the early 2020s in the United States, with debt growth, driven to a substantial extent by higher repo funding, underpinning most of the rise in assets under management. That said, there is cross-sectional heterogeneity in the use of leverage depending on fund strategies, with those focused on relative-value trades in fixed-income markets being the most leveraged (Barth and Kahn, 2021). Hedge funds often achieve leverage through repos. They are collateralised short-term loans in the form of asset sales with the agreement to buy back later at a pre-set price. To protect against borrower default, repos involve a haircut, so that the amount the debtor can borrow is less than the value of the pledged securities. In principle, haircuts can increase when default risk rises, leading to pro-cyclical changes in leverage – even, potentially, when lending is backed by safe assets (Morris and Shin, 2008). Hedge funds can also achieve leverage through the use of derivatives, which allow taking exposures without fully funding positions. Margins are used to protect against the default of the derivative counterparty. However, margins can increase rapidly when volatility spikes, leading to a sharp decline in the amount of attainable leverage and hence in risk-taking capacity. 7.2.5 Looking Ahead: Decentralised Finance The main sources of financial instability – such as leverage and liquidity transformation – tend to stay remarkably similar in the course of time, even if they take different guises as financial innovation advances. A relevant example is stablecoins, which are crypto-assets that tie their value to that of fiat currencies such as the US dollar. Stablecoins increasingly play a role as vehicle currencies within the decentralised finance (DeFi) ecosystem and act as a store of value, especially in countries with a high history of inflation. Normally, stablecoins strive to maintain a fixed value relative to fiat currencies by investing in short-term and sometimes illiquid financial assets. Certain stablecoins tied to the US dollar hold significant amounts of CPs and CDs and have characteristics reminiscent of MMFs. Not unlike open-end regulated investment funds, stablecoins investing in illiquid securities can be a source of vulnerability in the financial system, not least due to their fragile structure as near-money substitutes (Aramonte, Huang and Schrimpf, 2021; IMF, 2021).

7.3 NBFIs AND THE PROPAGATION OF SYSTEMIC RISKS Some themes in the financial stability analysis of NBFIs discussed earlier, such as liquidity transformation and the role of leverage, share points in common with that of banks. But, there are also new dimensions arising from the importance of market prices and balance-sheet management by NBFIs. These distinctive elements matter for liquidity imbalances and especially for the propagation of systemic risk. In this section, we describe the core elements of a conceptual risk accounting framework that is useful to characterise a relatively less explored issue – fluctuations in leverage and how they propagate liquidity risk and amplify shocks in the financial system. The framework builds on Shin (2008) and is formally developed in the Online Appendix and in Aramonte, Schrimpf and Shin (2021).

Non-bank financial intermediaries and financial stability  159

A traditional perspective on systemic risk revolves around a “domino” model of cascading defaults. According to this model, if Bank A has borrowed from Bank B, while Bank B has borrowed from Bank C, and so on, then a shock to Bank A’s assets that leads to its default will hit Bank B as well. If the hit is big enough, Bank B’s solvency will be impaired, in which case Bank C would be hit, and so on further down the line. Insolvency is seen as the driver of systemic risk in the domino model. However, while insolvency often figures in systemic crises, it need not do so. Fluctuations in leverage working through shifts in risk-taking capacity can also be a potent channel of propagation of stress, especially in settings with market-based intermediation. Oftentimes, leveraged positions require posting margin using own funds (equity), so that the ratio of total exposure to margin corresponds to overall leverage. Attainable leverage is therefore the reciprocal of the size of the margin investors post to open their positions. Changes in margin (and the corresponding fluctuations in leverage) are reflected in the fluctuations in the balance sheet size of market participants and in the broader risk-taking capacity of the financial system. In this context, a sharp increase in margins, especially after a protracted period of thin margins, will tighten financial conditions for the system as a whole. While insolvencies may exacerbate the stress, they are not a necessary ingredient. Instead, pecuniary externalities – that is, spillovers that work through prices – can become potent channels through which stress can spread. The fact that financial stress can emanate from safe assets such as government bonds (Morris and Shin, 2008), as evident during the Covid-19 crisis (see Section 7.4), is also an important theme of our discussion. The organising idea of our risk accounting framework is that fluctuations in the risk capacity of market participants can be amplified by the actions of market participants themselves. The main building block is the risk budgeting decision of an investor who posts margins to establish leveraged positions, for instance in the context of collateralised borrowing or derivatives transactions. The investor is risk-neutral and maximises expected returns, subject to portfolio risk constraint expressed in the form of Value-at-Risk (VaR).13 Allocating economic capital across different assets entails a risk budgeting decision akin to a consumer choice problem over goods. Higher economic capital relaxes the VaR constraint and allows the investor to take on more risk. The main insights that come from our risk framework can be summarised in two propositions: 1. The first proposition is that the debt capacity of an investor is increasing in the debt capacity of other investors. In this sense, debt capacity is recursive, and leverage enables greater leverage. Conversely, a spike in margins can set off a generalised deleveraging that leads to system-wide spillovers, especially after a prolonged period of thin margins that facilitated risk-taking. 2. The second proposition is that the deleveraging channel of risk propagation can manifest as cash hoarding or a “dash for cash”. This is because a broad increase in margins across assets sets off a reallocation of scarce economic capital, whereby investors rebalance their portfolios towards less risky assets with low margin requirements, such as cash or 13 Intuitively, VaR is a given percentile of the profit-and-loss (PnL) experienced by an institution so that, any loss larger than VaR happens with some given small probability. Formally, for a Î ( 0,1) , VaR at level α is the smallest number X such that the probability that PnL = 740

 97

0.00

0.25

0.25

0.50

0.25

0.25

0.25

0.75

0.75

720–739

0.00

0.25

0.50

0.75

0.50

0.50

0.50

1.00

1.00

700–719

0.00

0.50

1.00

1.25

1.00

1.00

1.00

1.50

1.50

680–699

0.00

0.50

1.25

1.75

1.50

1.25

1.25

1.50

1.50

660–679

0.00

1.00

2.25

2.75

2.75

2.25

2.25

2.25

2.25

640–659

0.50

1.25

2.75

3.00

3.25

2.75

2.75

2.75

2.75

620–639

0.50

1.50

3.00

3.00

3.25

3.25

3.25

3.50

3.50

rioer . The federal funds rate also has clear downward spikes at the end of each month from 2015 to 2018, as U.S. branches of foreign banks pursue “window dressing” for their home regulators and reduce the amount of IOER arbitrage on their balance sheets at the end of the month (Banegas & Tase, 2020). By varying these three administered interest rates, the Fed has thus been able to adjust its target for rf*f without changing the total quantity of federal funds in the market. Thus, although the Federal Reserve continues to communicate monetary policy through a target for the federal funds rate, the way in which that target is implemented is very different from before. Prior to 2008, the Fed achieved its target for rf*f by varying the total quantity of federal funds in the market, shifting the supply and demand for fed funds loans left and right in Figure 10.1. Now, the Fed implements its target for rf*f by varying the administered interest rates ronrrp , rioer , and rd in Figure 10.4, without changing the quantity of reserves or shifting those supply and demand curves left or right. Afonso, Armenter, and Lester (2019) provide a detailed estimation of the supply and demand for federal funds loans in the pre- and post-2008 periods which takes into account each of these key interest rates and the search process between would-be federal funds borrowers and lenders. See also Ihrig, Senyuz, and Weinbach (2020). 10.4.2.4 Other short-term U.S. interest rates after 2008 Despite the dramatic changes in the federal funds market since 2008, the relationship between the equilibrium federal funds rate and other short-term interest rates in the U.S. has remained largely unchanged. Essentially all large U.S. financial institutions now hold enormous quantities of federal funds and earn the IOER rate on those reserves. If other short-term interest rates in the U.S. differed very much from the IOER rate, then arbitrage by these U.S. financial

232

2018

2019

2020

2021

2017

2018

2019

2020

2021

Figure 10.5  Relationship between the Fed’s administered rates and other short-term interest rates since 2015

Note:  Relationship between the Fed’s directly administered interest rates (discount rate, IOER rate, and ON RRP rate), the federal funds rate (panel a), and other shortterm interest rates (panel b) from 2015 to 2020. See text for details.

2016

discount rate IOER rate ON RRP rate 30-day A2/P2 rate 1m Treasury bill rate

(b) Administered Rates and Other Short-Term Interest Rates

0 2015

0.5

0.5

2017

1

1

2016

1.5

1.5

0 2015

2

2

3

3.5

4

2.5

discount rate IOER rate ON RRP rate federal funds rate

(a) Administered Rates and the Federal Funds Rate

2.5

3

3.5

4

The federal funds market, pre- and post-2008  233

institutions would drive that rate back toward rioer .14 Thus, by setting the IOER rate, the Fed effectively controls all short-term interest rates in the U.S. economy. This is illustrated in the right-hand panel of Figure 10.5, which depicts the relationship between the Fed’s directly administered interest rates and two representative short-term market rates: the one-month Treasury bill rate and the 30-day A2/P2 commercial paper rate. Note that both of these market rates are one-month rates rather than overnight and have different risk, liquidity, and tax characteristics than the federal funds rate, so these market rates are not necessarily bounded between ronrrp and rd the way the federal funds rate is. For example, the one-month Treasury bill is safer than a federal funds loan and has tax advantages, leading its yield to frequently fall below ronrrp , while 30-day A2/P2 commercial paper is riskier than overnight fed funds loans and sometimes trades above rd, especially in early 2020 at the onset of the Covid pandemic. Nevertheless, these short-term market interest rates track the Fed’s administered interest rates closely.

10.5 THE FUTURE OF THE FEDERAL FUNDS MARKET Going forward, the Federal Reserve has declared that it will continue to express monetary policy in terms of a target for the federal funds rate, and will implement this policy through the administered ON RRP offer rate, IOER rate, and discount rate, just as it has been doing for the past several years (Federal Reserve Board, 2019). Thus, we should expect the equilibrium in the federal funds market to remain consistent with Figures 10.4 and 10.5 for the foreseeable future. Despite the fact that the Fed will continue to communicate monetary policy in terms of the federal funds rate going forward, that interest rate is very different from what it was in the past. The market participants are very different (trading is now completely dominated by the GSEs and U.S. branches of foreign banks), the reasons for trading federal funds in the market are completely different (trading is now dominated by arbitrage between two of the Fed’s administered interest rates, the ON RRP rate and IOER rate), and the way the Fed implements the federal funds rate target is completely different (varying the administered rates ronrrp , rioer , and rd rather than the quantity of reserves). These changes are arguably as large and important as those during the period of reserves targeting under Paul Volcker from 1979 to 1982, discussed by Bernanke and Mihov (1998). It is thus somewhat surprising that the FOMC has decided to continue to put so much emphasis on an interest rate that is now little more than an arbitrage indicator, as opposed to communicating policy in terms of the IOER rate directly. Another interesting question going forward is whether the Fed will at some point set the federal funds rate target below zero. Several other central banks, including the Swiss National Bank, Swedish Riksbank, European Central Bank, Danish Nationalbank, and Bank of Japan have all set negative policy rate targets and maintained those targets for several years (Swanson, 2018). The Danish Nationalbank and Swiss National Bank, in particular, have set their policy rates as low as -0.75%. In Figure 10.4, the Federal Reserve could attain such an equilibrium by setting rioer and ronrrp less than zero—in other words, by charging institutions 14 Of course, this arbitrage is now subject to the same FDIC-imposed costs discussed above, which prevents arbitrage of very small differentials. Nevertheless, the point remains valid that other short-term U.S. interest rates cannot deviate very far from the IOER rate.

234  Research handbook of financial markets

a fee to hold federal funds deposits at the Fed. Although this gives the institutions an incentive to convert reserves into physical currency, the costs of holding physical currency (storage, transportation, security, risk of loss due to theft or accident, etc.) make this alternative unattractive for all but the smallest quantities of currency unless the IOER rate becomes substantially negative, appreciably less than the −0.75% seen in Europe so far (Swanson, 2018). Moreover, the convenience and efficiency of conducting interbank transactions via federal funds rather than physical currency are so large that there will always be a market for federal funds at the margin even at very negative interest rates. Once the Fed has set a negative IOER rate, the same arbitrage arguments as in the previous section apply to other short-term market rates, driving them negative as well, analogous to Figure 10.5(b). This has certainly been the case in Europe, where yields on government bonds have fallen below zero for maturities out to several years in some cases. To date, the Fed has equivocated regarding its views on the costs and benefits of a negative federal funds rate. In testimony before Congress in 2016, Fed Chair Yellen stated that “I’m not aware of anything that would prevent us from doing it,” but also that there were legal questions and technical issues with the Fed’s internal computer systems that would need to be resolved (Derby & Zumbrun, 2016). But with medium- and long-term interest rates near historic lows, there is a limit to the effectiveness of the Fed’s other unconventional monetary policy tools— forward guidance and large-scale asset purchases—that makes consideration of a negative federal funds rate a distinct possibility in the next crisis.

10.6 SUMMARY The federal funds market has changed dramatically since its inception in the 1920s. Traditionally, U.S. depository institutions used the federal funds market to meet their reserve requirements or to earn interest on excess reserves. Beginning in 2008, several major policy changes completely transformed the market: the Fed began paying interest on reserves, the Fed increased the total quantity of federal funds in the market by a factor of 60, and the FDIC began charging U.S. depository institutions a fee on total assets held rather than just deposits. As a result of these changes, U.S. depository institutions have now largely left the federal funds market, since they are almost never short of their reserve requirements, they can earn the IOER rate on reserve balances held at the Fed, and the FDIC fees make IOER arbitrage unprofitable for them except in rare cases. Essentially all federal funds lending is now done by the GSEs, who are ineligible to receive interest on reserves, and almost all federal funds borrowing is now done by U.S. branches of foreign banks, who are exempt from FDIC fees and can profitably conduct IOER arbitrage between the GSEs and the Fed. The Federal Reserve also implements its target for the federal funds rate very differently than it did before. Traditionally, the Fed would vary the total quantity of reserves to shift the supply and demand for federal funds loans in the market. Now, the Fed uses its directly administered interest rates—the ON RRP rate, the IOER rate, and the discount rate—to implement its target for the federal funds rate without changing the quantity of reserves. The federal funds market is thus not at all what it used to be, and the structural break is arguably as large and significant as that documented by Bernanke and Mihov (1998) for 1979–1982. Researchers thus need to be mindful of the break in structure and behavior of the federal funds market around 2008 that continues to this day.

The federal funds market, pre- and post-2008  235

Going forward, the Federal Reserve has declared its intention to continue communicating monetary policy in terms of a target for the federal funds rate and to continue operating in an environment of abundant reserves, as it has been doing since late 2008. An interesting open question is whether, in some future crisis, the Fed will set a negative value for the federal funds rate target as many other central banks around the world have done recently with their own short-term interest rate targets.

REFERENCES Afonso, G., Armenter, R., & Lester, B. (2019). A model of the federal funds market: Yesterday, today, and tomorrow. Review of Economic Dynamnics, 33, 177–204. Afonso, G., Kovner, A., & Schoar, A. (2011). Stressed, not frozen: The federal funds market in the financial crisis. Journal of Finance, 66(4), 1109–1139. Anbil, S., Carlson, M., Hanes, C., & Wheelock, D. (2021). A new daily federal funds rate series and history of the federal funds market, 1928–54. Federal Reserve Bank of St. Louis Review, First Quarter, 45–70. Banegas, A., & Tase, M. (2020). Reserve balances, the federal funds market, and arbitrage in the new regulatory framework. Journal of Banking and Finance, 118, 105893. Bech, M., & Klee, E. (2011). The mechanics of a graceful exit: Interest on reserves and segmentation in the federal funds market. Journal of Monetary Economics, 58(5), 415–431. Bernanke, B., & Mihov, I. (1998). Measuring monetary policy. Quarterly Journal of Economics, 113(3), 869–902. Carlson, M., & Rose, J. (2017, December 19). Stigma and the discount window. Federal Reserve Board .federalreserve​ .gov​ /econres​ /notes​ /feds​ [Website]. Retrieved October 22, 2021, from https://www​ -notes​/stigma​-and​-the​-discount​-window​-20171219​.htm. Craig, B., & Millington, S. (2017). The federal funds market since the financial crisis. Federal Reserve Bank of Cleveland Economic Commentary 2017–07. Derby, M., & Zumbrun, J. (2016, February 11). Fed nods to negative rates, hurdles and all. Wall Street Journal. Retrieved from https://www​.wsj​.com ​/articles​/fed​-nods​-to​-negative​-rates​-hurdles​-and​-all​1455126278. Federal Reserve Bank of New York. (1998). Open market operations during 1997. Federal Reserve Bank of New York [Website]. Retrieved January 24, 2022, from https://www​.newyorkfed​.org​/medialibrary​/ media​/markets​/omo​/omo97​.pdf. Federal Reserve Bank of New York. (2013, March). Federal funds and interest on reserves. Federal Reserve Bank of New York [Website]. Retrieved October 11, 2021, from https://www​.newyorkfed​.org​ /aboutthefed​/fedpoint​/fed15​.html. Federal Reserve Board. (2018, January 3). Overnight reverse repurchase agreement facility. Federal Reserve Board [Website]. Retrieved October 16, 2021, from https:/-0.9pt/www​.federalreserve​.gov​/ monetarypolicy​/overnight​-reverse​-repurchase​-agreements​.htm. Federal Reserve Board. (2019, February 22). Timelines of policy actions and communications: Policy normalization principles and plans. Federal Reserve Board [Website]. Retrieved October 16, 2021, from https://www​.federalreserve​.gov​/monetarypolicy​/timeline​-policy​-normalization​-principles​-and​plans​.htm. Federal Reserve Board. (2021, February 3). Reserve requirements. Federal Reserve Board [Website]. Retrieved October 6, 2021, from feder​​alres​​erve.​​gov​/m​​oneta​​r ypol​​icy​/r​​eserv​​​ereq.​​htm. Friedman, M., & Schwartz, A. J. (1963). A monetary history of the United States, 1867–1960. Princeton, NJ: Princeton University Press. Furfine, C. (1999). The microstructure of the federal funds market. Financial Markets, Institutions and Instruments, 8(5), 24–44. Gürkaynak, R., Sack, B., & Swanson, E. (2007). Market-based measures of monetary policy expectations. Journal of Business and Economic Statistics, 25(2), 201–212. Hamilton, J. (1996). The daily market for federal funds. Journal of Political Economy, 104(1), 26–56.

236  Research handbook of financial markets

Ihrig, J., Senyuz, Z., & Weinbach, G. (2020). The Fed’s ‘ample reserves’ approach to implementing monetary policy. Federal Reserve Board Finance and Economics Discussion Series 2020-022. Meulendyke, A.-M. (1998). U.S. Monetary policy and financial markets. New York: Federal Reserve Bank of New York. Naber, J., Sambasivam, R., & Styczynski, M.-F. (2017, January 4). Demand for voluntary balance requirements: The U.S. Experience with contractual clearing balances from 2000 to 2007. Federal Reserve Board [Website]. Retrieved October 6, 2021, from https://www​ .federalreserve​ .gov​ / econresdata ​/notes​/feds​-notes​/2017​/demand​-for​-voluntary​-balance​-requirements​-the​-us​-experience​with​-contractual​-clearing​-balances​-from​-2000​-to​-2007​-20170104​.html. Swanson, E. (2018). The Federal Reserve is not very constrained by the lower bound on nominal interest rates. Brookings Papers on Economic Activity, 2018(Fall), 555–571. Wheelock, D. (2013). The Fed’s formative years. Federal Reserve History [Website]. Retrieved October 6, 2021, from https://www​.fed​eral​rese​r veh​istory​.org.

11. The repo market Benjamin Munyan

11.1 INTRODUCTION AND HISTORY OF REPO The sale of a security with an agreement to repurchase, or a “repo” transaction for short, exists in a legal niche between securities and loans. One party is offering financial securities, another party offers cash. Sometimes the transaction is driven by a desire to obtain cash, and other times it is driven by a desire to obtain access to a particular security. Services such as triparty or central clearing exist in some repo markets to handle the valuation of collateral, counterparty credit risk, and settlement. Some repo agreements allow for “general collateral”, which can be swapped out for different securities as the cash borrower’s trading inventory changes. This flexibility is one reason that the repo market is a dominant channel in global short-term funding markets. 11.1.1 The Invention of Repo The Federal Reserve’s (“the Fed”) formation in December 1913 coincided with the outbreak of World War I just eight months later and led to the establishment of the first repo market in 1917. As banker to the United States Treasury, one of the Federal Reserve’s first challenges was to facilitate sales of war bonds and keep credit flowing to American businesses. Initially, this took the form of marketing committees and fundraising drives, which evolved to offering preferential lending terms to banks that bought more U.S. Treasury debt. As the war progressed, European allies shipped increasing amounts of gold to the U.S. to pay for weapons and supplies, increasing the supply of gold-backed U.S. dollars as well as the reserves of the Federal Reserve. With this increase in reserves, the Fed was able to purchase Treasury securities to influence interest rates, conducting its first “open market operations” and laying the groundwork for the Fed’s role in supporting financial stability. On November 28, 1917, the surprise introduction of a $0.02 wartime stamp tax on promissory notes caused a sudden shock to U.S. short-term funding markets. A fixed tax might be bearable for longer-term bills or loans but was exorbitant for banker’s acceptances (then the predominant form of commercial paper) and other short-term lending, which would re-incur the tax each time the loan was rolled over. In the words of then-Fed Governor William P.G. Harding: “this tax practically prohibits this form of short term borrowings”.1 While the Fed pressed for an exemption on this tax, their member banks were clamoring for a way to exit the market for banker’s acceptances and avoid the tax, threatening a market panic. The new legislation did contain an exemption stating that banks could offer their commercial paper for cash at the Federal Reserve without incurring an additional stamp tax on

1 December 1, 1917 letter by W.P.G. Harding to U.S. Treasury Secretary William McAdoo and Comptroller of the Currency John Skelton Williams. 237

238  Research handbook of financial markets

that transaction. The standard practice at the time would have been for the Fed to “discount” these securities, i.e. deduct the full interest upfront based on the remaining time to maturity of the commercial paper. Given that the discount window rate is designed as a penalty rate, this could have resulted in losses for banks—paying more in interest to the Fed than they earned on the commercial paper (even before the stamp tax). Instead, the Fed decided to come up with a new solution, which they called “resale agreements”. The Fed offered to buy banks’ commercial paper with a discount only for interest during the agreement period (initially intended to be 15 days or less) and then re-sell the securities back to banks who are contractually obligated to re-purchase them after the panic subsided, or alternatively, the Fed could buy these securities with no upfront discount and charge interest in the form of a higher price when they re-sell the security. Thus, the first repo transaction was born. Repos were an immediate success, providing $34 million out of the total $85 million in Fed liquidity to the U.S. banking system by the end of December 1918. 11.1.2 Growth of Repo Markets After WWI, the Fed continued to use repo to support the growth of commercial paper markets, even allowing nonbank dealers to participate. However, the repo market collapsed along with much of the financial system during the Great Depression. Then in 1951, the TreasuryFederal Reserve Accord was reached and the Fed was no longer required to support Treasury debt prices. This return of a free market brought higher trading activity in U.S. government bonds, bringing with it a need to finance trading inventory through a new interdealer repo market.2 The 1980s brought modernization and globalization to repo markets. Garbade (2006) discusses how the collapse of two dealers—Drysdale Government Securities and LombardWall—in 1982 caused repo dealers to recognize accrued interest (i.e. the “dirty price” of a bond) when offering cash in repo and led to Congress passing a law exempting repo from the “automatic stay” process of bankruptcy courts. In Europe, the U.K.’s de-regulatory “Big Bang” created a demand for cash financing via repo as London investment banks built up large bond trading positions. Elsewhere in Europe, the need for cheaper ways to borrow securities and growth in trading of Bunds and new Eurobonds (Matif’s “Notionel” product) created a growing opportunity for repo markets. A second “Big Bang” in 1996 created an open repo market for trading gilts in the U.K. Recent years have been marked by a return to central bank involvement in supporting repo functioning. In response to a sovereign debt and liquidity crisis, in 2011 the European Central Bank (ECB) created its Long-Term Refinancing Operation (LTRO), an essentially unlimited three-year repo facility offering a 1% rate (and a large haircut). In the U.S., the Fed created a Reverse Repo Facility in 2014, offering assets they had purchased through quantitative easing in exchange for cash from Money Market funds and other cash lenders, in order to put a floor on short-term interest rates. In 2020, the Covid pandemic caused central banks to intervene forcefully with new and existing tools, including a new Dollar Repo Facility between the Fed and other Central Banks. Then in June 2021, the Federal Reserve began to offer five basis points of interest in their Reverse Repurchase Facility, which coincided with a resurgence in 2 The U.S. government bond market and the U.S. Treasuries repo market are currently the largest in the world.

The repo market 

239

Source:   Federal Reserve Bank of New York.

Figure 11.1  Daily U.S. triparty repo transaction volumes ($) U.S. repo activity. In particular, the triparty segment of the U.S. repo market (discussed in section 11.5.1) grew from $1.25 trillion in daily volume at the start of 2021 to more than $2.85 trillion by the end of 2021 (Figure 11.1). The repo market remains a very large and active source of financing for the broader financial markets.​

11.2 STRUCTURE OF A REPO CONTRACT The key terms of a repo contract are its ●

●

●

Starting date: Also called settlement date, on-side date, or value date, this is the date at which securities and cash are exchanged. Repo transactions are typically negotiated on the same day they will start, but not always. Maturity: The date when the exchange of securities and cash is reversed. This can be specified (e.g. one day, 30 days, until stated maturity of the collateral, i.e. “Repo to Maturity”, etc.) or it can be “rolling”, where the contract is renewed each day but the collateral provider may change which particular CUSIP-level securities are exchanged.3 Rate: The rate of interest due at maturity, as a percentage of the cash initially exchanged. If the repo transaction is motivated by the counterparty seeking cash, this rate will be positive, but if the repo is driven by the counterparty seeking a specific security, this rate can

3 This is useful, for example, when a dealer is making markets in various securities and financing their fluctuating inventory levels using repo.

240  Research handbook of financial markets

●

●

be relatively small or even negative (i.e. the security is “special”). Interest is assessed under Number of Days standard money market rate conventions, i.e. effective rate = Rate * . 360 Description of acceptable collateral: A contractual specification of what securities will be exchanged. The acceptable collateral may be any security within a broad category (e.g. government bonds, or the subset of government bonds with no more than X years to maturity), or it may be as specific as one particular security, like the latest on-the-run ten-year U.S. Treasury bond. Haircut: The required over-collateralization of securities versus cash exchanged, or how much the collateral value exceeds the cash value (similar to initial margin). This protects the cash provider from losses if the collateral provider defaults. Haircuts are typically assessed at the initiation of the repo and can remain static while the repo is open. However, if the value of the securities is volatile (e.g. equities), the cash provider may require the securities to be market-to-market on a daily basis. If a drop in collateral value is material, the securities provider may be required to deliver additional securities to the cash provider (similar to variation margin), if this was specified during the negotiation of the repo.

During a repo transaction, the economic effects of ownership are retained with the collateral provider. If a bond pays a coupon or a stock pays a dividend, that cash is returned along with the security itself when the repo matures. However, if a corporate action such as a shareholder vote occurs during the repo, the cash provider (who now possesses the collateral) gets to vote. 11.2.1 Example of a Classic Repo Transaction On Wednesday, February 5, 2020, a dealer wants to borrow cash overnight to finance their securities inventory. They can offer a $10 million par value of five-year U.S. Treasuries to general collateral (“GC”) investors. On that date, those investors would have offered a rate of 1.57%. The clean price4 of the five-year on-the-run U.S. Treasury bond was 99.57813 on the repo start date, and the dirty price was 99.600795. In the U.S. general collateral market the haircut is zero, so the initial cash exchanged would be equal to the dirty price of the bonds. Thus, with a maturity of one day (overnight), our dealer would sell the $10 million par value Treasuries for $9,960,079.50. On Thursday, February 6, the repo matures and our dealer repurchases those same æ .0157*1 ö Treasuries by wiring the cash investor $9,960,079.50 *ç 1 + , or $9,960,513.87, and 360 ÷ø è receiving their five-year Treasuries. The total interest paid was $434.37, and the security was repurchased to close the repo at the same price it had been sold for to open the repo.

4 The clean price of a bond ignores accrued interest; the dirty price includes accrued interest. Both are quoted as a percentage of the bond’s par value, e.g. $10,000. Bonds are typically quoted in clean prices for trading purposes, with trade settlement and financing (e.g. repo financing) conducted using dirty prices. In general, interest accrues daily based on the market convention for that bond, which is actual/actual for U.S., U.K., and many European sovereign bonds. Thus a $1,000 par value U.S. Treasury bond with a coupon rate of 3.75% that paid its most recent coupon 65 days ago on August 15 (meaning its next coupon is paid semi-annually on February 15) would be valued with 65 ´ 0.5 = $6.6236 in addition to its “clean price” if it were accrued interest of .0375 ´ $1, 000 ´ 184 traded today.

The repo market 

241

11.2.2 Repo versus Reverse Repo In a repo transaction, one side is receiving cash in exchange for their securities (selling securities under agreement to repurchase), while their counterparty is doing the opposite (buying securities under agreement to re-sell). The former is sometimes called “cash-in” on the repo, while the latter would be “cash-out”. In standard industry terms, the “cash-in” leg is called a repo, while the “cash-out” leg is called a reverse repo. Somewhat idiosyncratically, central banks such as the Federal Reserve Bank of New York (“the Fed”) may adopt the opposite nomenclature of typical market participants. When the Fed is “cash-out” (purchasing securities with agreement to re-sell, and therefore transmitting cash to a dealer) they call this a repo. The Fed’s Reverse Repurchase Agreement facility calls their operations reverse repo even though they are “cash-in”, selling Fed securities holdings under agreement to repurchase.5 11.2.3 Rehypothecation In a process called rehypothecation lending, a client of a dealer offers securities in exchange for cash financing (sometimes through a repo, or other times simply through a prime brokerage margin account), and the dealer re-pledges those same securities for its own cash financing. This re-pledging can then be repeated by the next cash lender, and so on. While some amount of rehypothecation may help lower borrowing costs for the initial cash borrowers, the practice could be potentially problematic if one participant somewhere in the “chain of rehypothecation” defaulted.6 Indeed, Singh and Aitken (2009) show that after Lehman Brothers’ bankruptcy in 2008, the extent of rehypothecation declined substantially, as investment firms feared losing access to their collateral if their prime broker went bankrupt. 11.2.4 Default versus Fail-to-Deliver If a cash borrower in a repo fails to return the cash with interest at the repo maturity date, this is considered a default, and the cash lender may hope to recoup their losses by liquidating the collateral. However, if a cash lender (collateral borrower) in a repo fails to return the collateral at the repo maturity date, this is not considered a default but rather a fail-to-deliver (or “fail”). In a fail-to-deliver, the cash lender may pay a fee and the cash borrower must wait to unwind the repo. Alternatively, at the start of a repo transaction, a fail may occur when the cash borrower does not deliver the obligated securities (and therefore doesn’t get cash but must still pay the full amount of specified interest on the repo). While an uncommon occurrence, fail-to-delivers are usually triggered by some type of special collateral. For example, Fleming and Garbade (2004) and Fleming and Garbade (2005) discuss how heavy short interest in the on-the-run ten-year U.S. Treasury Note led to a surge in “strategic” fails: investors had expected the note to be re-opened through a new Treasury auction in September 2003 (increasing the supply of that security for short sellers and reducing its specialness), but when the Treasury announced it wouldn’t re-open the note, these investors simply failed to deliver on repos involving that note rather than try to locate these scarce securities. To remedy this situation, Garbade, Keane, Logan, Kirby, and Wolgemuth (2010) discuss how a “fails charge” was introduced to penalize fail-to-delivers and discourage the practice of strategic fails. 5 This facility is discussed more in section 11.7.3 6 Especially if the collateral had also been rehypothecated off to a foreign country!

242  Research handbook of financial markets

11.3 REPO PRICING When a repo transaction is being arranged, there are conceivably many different dimensions that could be negotiated: for example, cash borrowers could seek to optimize their financing by finding a repo cash lender who applies a lower haircut, or who is willing to finance a larger amount of lower-quality collateral. However, in practice, it is the repo rate that prices the transaction, and this rate is highly correlated with the general policy rates set by central bankers. Figure 11.2 shows how U.S. repo rates rose as the Federal Reserve tightened monetary policy beginning in 2016, as well as the reversal in rates back towards zero when the Fed reversed their tightening beginning in 2019 and the Covid-19 pandemic in early 2020. Because a repo is economically similar to a collateralized loan, the interest rate on a repo contract should be capped by the interest rate on (unsecured) interbank lending, i.e. Libor. Choudhry (2010) shows that the GC overnight rate tracks closely with the Libor overnight rate in the U.K. In the U.S. we can also see this relationship typically holds when we compare the GC overnight rate to the effective federal funds rate (see Figure 11.3, which uses a shorter time series to make the two rates more visible). However, especially in the 2018–2019 period of this figure (before both rates went to nearly zero), we notice the repo rate does periodically “spike” and exceed the federal funds rate, which violates our intuition. As we will discuss in this section, it can be very useful to think of repo as a baseline “cost of carry” for basis trading, futures arbitrage, or other essential trading activities that underpin financial markets. In practice, frictions such as specialness (see section 11.3.2), market segmentation (11.3.4), or responses to regulatory policy (11.7.1) may distort repo rates from their theoretical value.

Source:   Federal Reserve Bank of New York.

Figure 11.2  Daily U.S. triparty repo volume-weighted average rate (%)

The repo market 

243

Source:   Federal Reserve Bank of New York.

Figure 11.3  Effective federal funds rate versus U.S. triparty repo GC rate (%) 11.3.1 Implied Repo Rates A theoretical repo rate can be calculated using the relationship between bond futures and the price of underlying bonds. For a bond with spot price P and a futures contract on that bond priced F, a spot rate s and a forward yield f, and time to contract maturity T, the forward-spot parity relationship is T



æsö F = P ç ÷ (11.1) èfø

Re-expressing this equation to solve for the risk-free spot rate s gives

æFö s= fç ÷ èPø

(1/ T )

(11.2)

This rate is the implied repo rate from market prices. The implied repo rate can be calculated for different futures contracts and their underlying bonds to understand which assets or maturities should carry a higher rate as repo collateral.7 By the same intuition, the degree of contagion versus backwardation expressed in bond futures is a reflection of the cost of carry, which for bonds is the cost of financing that bond via repo.

7 More detailed accounts of calculating implied repo rates in U.S. Treasuries and European bond markets can be found in Burghardt and Belton (1994), Plona (1997), and Choudhry (2010).

244  Research handbook of financial markets

11.3.2 Specialness Duffie (1996) shows that the general collateral (GC) rate should be the upper bound for repo rates, but that individual repo rates can be significantly lower when the collateral is “special” for some reason. For example, on-the-run ten-year U.S. Treasury bonds typically trade “on special”, with a lower repo rate shortly after issuance due to their importance in the interest rate derivatives market. As another example, a particular stock or bond that traders want to sell short on a given day may be difficult to locate, so dealers can offer a “special” lower repo financing rate on that collateral and thereby secure it for their customers’ short trades.8 Thus, the search costs and trading frictions can embed themselves in repo prices, in a way that is theoretically separable from the base cost of cash financing. In general, if the price of a general collateral bond is P and the price of a bond on special is P’, the no-arbitrage relation between the general collateral rate R and the special repo rate r must be given by

P¢ = P

1+ r (11.3) 1+ R

The specialness of a bond can therefore be measured by the difference in overnight repo rates s = r - R for the special versus general collateral security. Jordan and Jordan (1997) empirically test this model and find strong support for it in the U.S. Treasury market. One useful implication of this model is that although Duffie (1996) argues the most liquid securities should be the most special, this specialness can be filtered out when constructing a yield curve (1 + R) . simply by multiplying each bond’s market price by its relative specialness 1 + r) 11.3.3 SOFR Following the 2008 financial crisis, investigations revealed that employees of some banks were routinely and systematically colluding to manipulate the Libor interest rate. In order to avoid future manipulation, the secured overnight funding rate (SOFR) was designed to use overnight repo market transacted interest rates and replace Libor in U.S. markets. Forward contracts on one-month and longer tenors pay out based on the realized overnight SOFR over the contract period and are used to determine the longer-term SOFR rates. These features should make SOFR a floating rate benchmark, which is less susceptible to manipulation. On November 30, 2020, the Federal Reserve announced Libor would be fully phased out in favor of SOFR by June 2023. 11.3.4 Market Segmentation Repo rates may also be lower due to market segmentation and relationships. As Anbil, Anderson, and Senyuz (2021) show, triparty repo market rates are persistently lower than GCF repo rates, and this spread rises during periods of market stress. Han and Nikolaou (2016) show that within the triparty repo market, relationships between cash investors and dealers matter, and stronger relationships lead to more stable trading volumes at lower repo rates. 8 Of course, the dealer earns a profit from facilitating their customers’ trades, which motivates the special rate they offer in repo.

The repo market 

245

11.4 VARIATIONS ON REPO CONTRACTS AND SUBSTITUTES FOR REPO 11.4.1 Variations on Repo Contracts While section 11.2 describes the standard features that define a repo contract, many variations exist within the repo market. Just as repurchase agreements were first invented due to an unusual financing need, repo has continued to evolve to suit the needs of different market participants. In a sell/buyback repo, the security is sold at an initial spot price, with an agreement to repurchase the security at a higher forward price. While no separate interest is collected, the difference in spot and forward prices for the bond is calculated to incorporate interest on the repo. This structure has been common in Italy (until 2017) and Spain, but because this type of repo is so simple it is also used in countries where no legal framework dealing with repo transactions exists. In a triparty repo, securities providers and cash providers interact through a clearing bank (in the U.S., Bank of NY Mellon is currently the only clearing bank, after JP Morgan Chase stopped clearing in 2017). This third party handles custody of the collateral, values the securities and applies specified haircuts, and settles the repo on their books. These administrative services are attractive to many cash-rich investors who want an easy way to access the repo markets, and this market segmentation is part of why triparty repo rates are often cheaper than other forms of repo. Cash received through triparty repo often flows into other repo markets. Sometimes repo dealers borrow cash in triparty just so they can lend it out in other repo markets and collect a spread. Other times, a dealer may finance their customer’s securities inventory by doing a bilateral repo with that customer, and then re-pledging (“rehypothecating”) those same securities in the triparty repo market in order to source the cash their customer needs. Within government securities dealers, the GCF (general collateral financing) repo market operated by the Depository Trust & Clearing Corporation (DTCC) in the U.S. offers triparty repo as a way to enhance liquidity in the U.S. government securities market. Trading within this market is anonymous and counterparty credit risk is guaranteed by the parent of DTCC (the Fixed Income Clearing Corporation). Settlement is handled on a daily basis after netting repo transactions across all parties, and cash borrowers can substitute collateral within a broad class of “general collateral” government securities as their securities holdings change throughout the day. Other types of repo include dollar rolls, used commonly in the U.S. as part of the process to create mortgage-backed securities, whole loan repo, which uses higher-yielding collateral such as mortgage pass-through securities or credit card loans, and repo-to-maturity, which simply matches the maturity of the repo to the maturity of the underlying collateral, reducing rollover risk for the cash borrower. 11.4.2 Substitutes for Repo Similar to repo agreements, securities lending offers an exchange of cash and securities. Typically, securities lending is collateral-driven, such as when an investor is trying to locate a stock they would like to sell short. Cash collateral above the amount of the security’s value is exchanged, and that cash may be invested by the securities lender to earn additional interest

246  Research handbook of financial markets

income. Additionally, a securities lending fee is charged to the securities borrower similar to a repo rate, with the rate depending on how “special” the security is. However, the fact that there is no actual sale of the securities and the fee is fixed makes it easier to monitor and administer these transactions and may explain many asset managers’ preference for lending securities rather than selling them under agreement to repurchase. As a derivative instrument, total return swaps can offer similar economic exposure as a repo without the need to pay out cash or trade the asset. To complete the swap, the asset holder or “beneficiary” of the swap agrees to pay the total return (interest and appreciation) of the underlying asset to their counterparty (the “guarantor”).9 In return, the guarantor pays the beneficiary a market interest rate such as Libor or SOFR, plus a spread. The economic effect of the total return swap is as if the guarantor now owned the asset, while paying a financing cost to the beneficiary. If the term of the total return swap is less than the maturity of the underlying asset, the guarantor earns an interest spread (carry) from the short-term versus long-term interest rate spread. They could have achieved this result by purchasing the asset and pledging it as collateral in a repo, but now there is no need to purchase the asset or hold it on the guarantor’s balance sheet.10

11.5 MARKETS FOR TRADING REPO 11.5.1 U.S. Repo Markets There are three main markets for trading repo in the U.S.: triparty, GCF repo, and bilateral. The triparty repo market is the most studied U.S. repo market, due to its importance as a cash source for repo markets, as well as the availability of data. Following the collapse of Bear Stearns in March 2008, the Fed began collecting daily transaction-level data on triparty repo, identifying the counterparties, maturities, pricing, amounts traded, and collateral. From this, studies such as Copeland, Martin, and Walker (2014) are able to show that while broad triparty repo market activity was stable during the financial crisis, there was a sharp decline in triparty cash financing of Lehman Brothers in the days before its September 2008 collapse. The current size of the triparty repo market in September 2021 is $3.2 trillion, having surged from just $2.3 trillion in April 2021 and a low of $1.5 trillion in 2016.11 Paddrik, Ramírez, and McCormick (2021) study intraday transaction data in the triparty market and find a regular clearing cycle, with most activity occurring between 8 and 9 AM. Those cash lenders who do wish to transact late in the day almost always end up with the Fed as their counterparty through its ON RRP program. This suggests that for some triparty cash lenders, the Fed’s repo market participation functions somewhat like a deposit account. Instead of using restricted regulatory data on repo transactions to study triparty repo, Krishnamurthy, Nagel, and Orlov (2014) use quarterly form N-CSR filings by Money Market   9 If there is a drop in the asset’s value during the swap, that would result in a payment from the guarantor to the beneficiary. 10 This can be useful for banks facing size-based capital requirements such as the supplementary leverage ratio, since the same notional position is now “off-balance sheet” and carries a smaller penalty. 11 This surge is due largely to the Fed’s expandsion of its ON RRP facility, which operates through the triparty repo market.

The repo market 

247

Mutual Funds (a major class of triparty repo cash lender) to document the repo activity of the 20 largest money market fund families during the 2007–2010 period surrounding the Global Financial Crisis. They find that their sample of repo transactions was fairly stable during this period, but the contraction in repo against private-sector collateral was significant enough to drive key dealer banks to emergency lending programs from the Federal Reserve. Similarly, Hu, Pan, and Wang (2021) are able to infer repo activity using the SEC’s Form N-MFP filings. These filings provide detailed information on repo counterparties, collateral, and pricing at a monthly frequency, and the authors use it to show the important role of key fund families, particularly Fidelity, in the financial stability of the triparty repo market. In the GCF repo market, Copeland, Davis, and Martin (2015) use regulatory data to examine and infer the strategies used by dealers. They find that dealers use GCF repo and triparty repo as substitutes, and find that the market mainly serves as a source of cash and a resource for dealers to manage their bond inventory (e.g. by simultaneously pledging U.S. Treasuries to the market in repo and receiving U.S. agency mortgage-backed securities in reverse repo). A key benefit of this study was to dispel concerns of “collateral upgrade” trades, where dealers had been suspected of holding excessively risky securities, but using repo markets to swap them for safer securities. Additionally, Boyarchenko et al. (2015) find that the supplementary leverage ratio (SLR) has reduced the size of the GCF repo market.12 The bilateral repo market conducts repo operations without the infrastructure of a thirdparty custodian or netting service. It is also the most opaque repo market, with volumes often inferred through primary dealer statistics published by the Federal Reserve Bank of New York rather than examining direct regulatory data. Copeland et al. (2014) use such data to estimate the size of the bilateral repo market as roughly equal to triparty repo, with U.S. Treasuries comprising approximately 67% of bilateral repo collateral. In a special data collection pilot program, Baklanova et al. (2016) were able to examine three one-day snapshots of nine large bank holding companies’ bilateral repo and securities lending activities over the course of three months. The authors found that the great majority of bilateral repo involved U.S. Treasuries as collateral, and approximately one-third of bilateral repo was for overnight maturities. Interestingly, the authors did find that among equity collateral there was a high degree of variation in repo rates, suggesting collateral “specialness” drove many of these transactions. 11.5.2 European Repo Markets Although the European repo markets became well-established in the 1980s, the 2008 financial crisis highlighted the importance of collateral in interbank lending and spurred short-term cash investors away from unsecured lending and into repo. In contrast to the relatively opaque repo markets of the U.S., the ECB’s Money Market Statistical Reporting (MMSR) Regulation dataset provides detailed public data on daily repo market activity in the Eurozone. These data show nearly 90% of repo is overnight in maturity, and more than 75% of repo is collateralized by government bonds. The main centers of repo activity in the Eurozone are Germany, Italy, France, and Spain. Mancini, Ranaldo, and Wrampelmeyer (2016) note that while the size of the European repo market is comparable to U.S. repo markets, most European repo transactions are cleared through a central counterparty (CCP). That market feature appears to 12 GCF repo volumes are down from approximately $1 trillion to $450 billion since 2012.

248  Research handbook of financial markets

explain why European interbank repo did not experience the types of disruptions (e.g. higher haircuts, rates, or shorter maturities) that had been documented in U.S. repo markets, and instead, CCP-based repo became investors’ destination during “risk-off” periods of liquidity hoarding. Corradin and Maddaloni (2020) have shown that in contrast to prior periods of funding stress, European Central Bank purchases of high-quality collateral have created ample funding liquidity but drove up the specialness of remaining collateral. Schaffner, Ranaldo, and Tsatsaronis (2019) find that obtaining specific securities rather than cash has begun to drive repo activity. The authors are able to discern this through the rise in “specific collateral” repo compared with the stagnation and decline in “general collateral” Euro repo activity. This has had the unintended effect of segmenting European repo markets by country, consistent with a “home bias” of repo traders for bonds issued by their domestic sovereign. The U.K. repo market, also called the “gilt repo” market, has been an open market since 1996, encouraging non-dealer participation. This openness encouraged market growth, and gilt repo grew steadily from ₤43 billion to over ₤400 billion by 2007. The market shrank by as much as 40% following the 2008 financial crisis and the Eurozone sovereign debt crisis and gilt repo rates but was back above ₤400 billion by 2018. However, during the 2008 crisis period itself, cash investor demand for gilt repo soared due to a flight to safety, driving rates on one-month repo more than 150 basis points below interbank (Libor) rates. Due to the open market structure design, both gilt repo and gilt stock lending operate essentially as one market. A trader that wants to short a security can source that security either through a gilt repo or a gilt stock loan, and some investors do both. Although participation is open, repos are conducted bilaterally and settled through the CREST settlement system—triparty repo is not a significant part of the market.

11.6 FRAGILITY IN REPO MARKETS AND SYSTEMIC RISK Since at least the 2008 financial crisis, academics and policymakers have recognized that repo markets may be fragile or vulnerable to sudden stress. Gorton and Metrick (2012) described the 2008 Financial Crisis as a panic equivalent to a bank run. Essentially, they argue repo markets created a downward “liquidity spiral”: concerns about the liquidity of bond markets led repo cash investors to raise haircuts, which forced asset sales by repo cash borrowers, leading to lower prices and still higher haircuts, and thus further reduction in repo financing and yet more asset sales, and so on. Copeland, Martin, and Walker (2014) and Krishnamurthy, Nagel, and Orlov (2014) offer a somewhat contrasting view. Using regulatory data on triparty repo and money market mutual fund filings, respectively, these authors show that this part of the repo market was rather stable during the crisis, and the “run” on repo in September 2008 was confined to just Lehman Brothers. Gorton, Metrick, and Ross (2020) then responded to this theme, showing that market survey evidence suggested bilateral repo was much larger than the triparty repo market in 2008, and thus studies showing the stability of triparty repo were missing the bigger picture. Investigating Flow of Funds data on net repo funding sources and “nonreporting cash pools”, they argue that net repo financing to U.S. banks and dealers may have fallen by more than half during the crisis.

The repo market 

249

Ennis (2011) and Martin, Skeie, and Thadden (2014) offer theoretical models to show that runs are possible in repo markets. Moreover, these papers demonstrate how technical details of different repo markets can influence their susceptibility to runs, and individual banks may be subject to repo runs if they are perceived to be weak: unprofitable, small, or excessively reliant on short-term funding. More recent papers have shown that even outside of a financial crisis, repo markets are critical to the smooth functioning of other markets. Macchiavelli and Zhou (2021) show that when dealers are unable to access their normal levels of repo funding (in this case due to a shock to money market mutual funds in 2016), market liquidity falls in the form of wider bid-ask spreads and realized transaction costs in corporate bonds. At the individual dealer level, those dealers with weaker relationships with money funds in triparty repo also have lower corporate bond market share and conduct more riskless-principal trades to avoid taking positions into inventory. Huh and Infante (2021) and Infante and Vardoulakis (2021) offer a theoretical model of the interlinkages between repo markets and bond market liquidity. These findings about runs, liquidity spirals, and knock-on effects in other markets inspired the “macroprudential regulation” approach espoused by Hanson, Kashyap, and Stein (2011). In this view, repo is important to the health of the overall financial system, and regulation should treat repo markets accordingly. In this spirit, much of the recent literature has examined how regulatory policy has affected the function and fragility of repo markets.

11.7 REGULATION AND CENTRAL BANK INTERACTIONS WITH REPO MARKETS 11.7.1 Bank Capital Requirements and Window Dressing As daily data on transaction volumes began to be collected in different repo markets, a clear seasonality pattern became evident around quarter- and year-ends. Because European banks report to regulators on a quarter-end “snapshot” basis, they are able to improve their reported leverage ratio by reducing their assets and thus their need for repo financing. After the new quarter starts, these banks can swell their balance sheets again and return to the repo market for financing. Although Munyan (2017) was the first to show this “window dressing” effect was driven by European banks, it has now been well documented across both U.S. and European repo markets by Kotidis and Van Horen (2018), Schaffner, Ranaldo, and Tsatsaronis (2019), and others. 11.7.2 Basel III Implementation Effects on Repo In addition to higher risk-based capital requirements for banks, the 2010 Basel III capital accords brought additional measures designed to combat bank leverage and liquidity risk exposure, as these were perceived to be key drivers of the 2008 Financial Crisis. Both the leverage ratio and the liquidity coverage ratio (LCR) reduced the attractiveness of repo to bank-affiliated dealers. As studied by Allahrakha, Cetina, and Munyan (2018), Kotidis and Van Horen (2018), Macchiavelli and Pettit (2021), and Ranaldo, Schaffner, and Vasios (2021), these policies had unintended consequences for repo market volume, bank dealer participation in repo markets, and liquidity in other markets such as corporate bonds.

250  Research handbook of financial markets

11.7.3 Recent Central Bank Innovations in Repo As central banks have purchased large quantities of assets to suppress interest rates (“quantitative easing” or QE), this has pushed rates to zero in some countries.13 Seeking to avoid negative rates while continuing QE, the Fed decided in 2014 to offer some of its now vast securities holdings as “reverse repo” to absorb excess short-term cash and put a floor under interest rates. Anbil and Senyuz (2018) show this RRP facility has acted as a shock absorber to changes in repo market demand. In mid-September 2019, the repo market experienced an unexpected shock as overnight rates suddenly spiked. Copeland, Duffie, and Yang (2021) show that this spike coincided with the reduction in aggregate reserves in the banking system, as a result of the Federal Reserve’s “balance-sheet normalization” which had begun in 2017. Figure 11.4 shows on the right axis the reserve balances of the 100 largest U.S. banks, and on the left axis the difference between the Secured Overnight Funding Rate (SOFR) and Interest on Excess Reserves (IOER) (i.e. a repo rate minus an alternative monetary policy-based rate from depositing excess reserves at the Fed). In mid-September 2019 there is a visible decline in bank reserves, and on September 17, 2019, the SOFR-IOER difference jumps more than 3.15%. They present evidence that in the environment of scarce reserves, some repo-active banks may have begun to hoard their reserves and delay intraday payments, perhaps precipitating the abrupt surge in repo rates in a

Source:   Copeland, Duffie, and Yang (2021), Figure 1.

Figure 11.4  Bank reserve balances and the September 2019 repo spike 13 In fact, rates have gone below zero in countries where the central bank also charged a fee on holding reserves.

The repo market 

251

run-like phenomenon. A similar event appears to have happened in March 2020, where bank reserves again declined and the SOFR-IOER spread rose to 0.44%, during the initial panic of the Covid-19 pandemic. In response to these disruptions, the Fed has re-evaluated its “balance-sheet normalization” strategy and moved instead to a focus on “ample reserves”, where the Fed raises interest rates through adjusting a “floor” policy rate, rather than by aggressively reducing its balance sheet size and draining reserves. More recently, in June 2021 the interest rate of the Reverse Repurchase Facility (RRP) (a Fed “floor” policy rate) was raised from zero to five basis points, and the size of this program surged from $20 billion in March 2021 to almost $1.5 trillion by the end of September. Just as it invented the repo in 1917, the Fed continues to innovate in order to support repo markets’ role in the broader financial system.

11.8 CONCLUSION Although there are certainly other ways to finance a position or obtain a security, the repo market persists. Born out of a short-term funding crisis, the repo market’s structure has evolved to become a key component of liquidity for government securities and other markets. As the U.S. Treasury Borrowing Advisory Committee noted in July 2013, repo is the “silently beating heart” of financial markets. The academic literature attests to that fact, with many papers showing both the externalities of proper repo market functioning, as well as the threat to financial stability from a repo market failure. The new repo facilities and programs, along with ever-changing regulations, offer new ways to study repo markets from a macroprudential perspective. Yet much of the market is still only partially observed, non-bank dealer participation in repo has risen due to enhanced bank regulation, and nontraditional financial institutions such as cryptocurrencies offer an increasing array of products that compete with traditional money markets. Therefore many opportunities remain to complete our picture of the modern repo market.

REFERENCES Allahrakha, M., Cetina, J., & Munyan, B. (2018). Do higher capital standards always reduce bank risk? The impact of the Basel leverage ratio on the US triparty repo market. Journal of Financial Intermediation, 34, 3–16. Anbil, S., Anderson, A. G., & Senyuz, Z. (2021). Are repo markets fragile? Evidence from September 2019. FEDS Working Paper. Anbil, S., & Senyuz, Z. (2018). The regulatory and monetary policy nexus in the repo market. FEDS Working Paper. Baklanova, V., Caglio, C., Cipriani, M., Copeland, A. et  al. (2016). The US bilateral repo market: Lessons from a new survey. OFR Brief Series, 16. Boyarchenko, N., Eisenbach, T. M., Shachar, O. et al. (2015). Have dealers’ strategies in the GCF repo market changed? Discussion paper. Federal Reserve Bank of New York. Burghardt, G. D., & Belton, T. M. (1994). The treasury bond basis: An in-depth analysis for hedgers, speculators, and arbitrageurs. New York: McGraw-Hill Companies. Choudhry, M. (2010). The repo handbook (2nd ed.) Woburn, MA: Butterworth-Heinemann. Copeland, A., Davis, I., LeSueur, E., Martin, A. et al. (2014). Lifting the veil on the US bilateral repo market. Discussion paper. Federal Reserve Bank of New York.

252  Research handbook of financial markets

Copeland, A., Duffie, D., & Yang, Y. (2021). Reserves were not so ample after all. Discussion paper. National Bureau of Economic Research. Copeland, A., Martin, A., & Walker, M. (2014). Repo runs: Evidence from the tri-party repo market. Journal of Finance, 69(6), 2343–2380. Copeland, A. M., Davis, I., & Martin, A. (2015). An empirical analysis of the GCF Repo® Service. Economic Policy Review, 25–37. Corradin, S., & Maddaloni, A. (2020). The importance of being special: Repo markets during the crisis. Journal of Financial Economics, 137(2), 392–429. Duffie, D. (1996). Special repo rates. Journal of Finance, 51(2), 493–526. Ennis, H. M. (2011). Strategic behavior in the tri-party repo market. FRB Richmond Economic Quarterly, 97, 389–413. Fleming, M. J., & Garbade, K. (2004). Repurchase agreements with negative interest rates. Available at SSRN 599907. Fleming, M. J., & Garbade, K. (2005). Explaining settlement fails. Current Issues in Economics and Finance, 11. Garbade, K., Keane, F. M., Logan, L., Kirby, A. S., & Wolgemuth, J. (2010). The introduction of the TMPG fails charge for US treasury securities. Economic Policy Review, 16. Garbade, K. D. (2006). The evolution of repo contracting conventions in the 1980s. Federal Reserve Bank of New York Economic Policy Review. Gorton, G., & Metrick, A. (2012). Securitized banking and the run on repo. Journal of Financial Economics, 104(3), 425–451. Gorton, G. B., Metrick, A., & Ross, C. P. (2020). Who ran on repo? AEA Papers and Proceedings, 110, 487–492. Han, S., & Nikolaou, K. (2016). Trading relationships in the OTC market for secured claims: Evidence from triparty repos. FEDS Working Paper. Hanson, S. G., Kashyap, A. K., & Stein, J. C. (2011). A macroprudential approach to financial regulation. Journal of Economic Perspectives, 25(1), 3–28. Hu, G. X., Pan, J., & Wang, J. (2021). Tri-party repo pricing. Journal of Financial and Quantitative Analysis, 56(1), 337–371. Huh, Y., & Infante, S. (2021). Bond market intermediation and the role of repo. Journal of Banking and Finance, 122, 105999. Infante, S., & Vardoulakis, A. P. (2021). Collateral runs. Review of Financial Studies, 34(6), 2949–2992. Jordan, B. D., & Jordan, S. D. (1997). Special repo rates: An empirical analysis. Journal of Finance, 52(5), 2051–2072. Kotidis, A., & Van Horen, N. (2018). Repo market functioning: The role of capital regulation. CEPR Discussion Paper No. DP13090. Krishnamurthy, A., Nagel, S., & Orlov, D. (2014). Sizing up repo. Journal of Finance, 69(6), 2381–2417. Macchiavelli, M., & Pettit, L. (2021). Liquidity regulation and financial intermediaries. Journal of Financial and Quantitative Analysis, 56(6), 2237–2271. Macchiavelli, M., & Zhou, X. (2021). Funding liquidity and market liquidity: The broker-dealer perspective. Management Science, 68(5), 3379–3398. Mancini, L., Ranaldo, A., & Wrampelmeyer, J. (2016). The euro interbank repo market. Review of Financial Studies, 29(7), 1747–1779. Martin, A., Skeie, D., & von Thadden, E.-L. (2014). Repo runs. Review of Financial Studies, 27(4), 957–989. Munyan, B. (2017). Regulatory arbitrage in repo markets. Office of Financial Research Working Paper. Paddrik, M. E., Ramírez, C. A., & McCormick, M. J. (2021). The dynamics of the US overnight triparty repo market. FEDS Notes, p. 2. Plona, C. (1997). The European bond basis: An in-depth analysis for hedgers, speculators & arbitrageurs. Irwin Professional Pub. Ranaldo, A., Schaffner, P., & Vasios, M. (2021). Regulatory effects on short-term interest rates. Journal of Financial Economics. Schaffner, P., Ranaldo, A., & Tsatsaronis, K. (2019). Euro repo market functioning: Collateral is king. BIS Quarterly Review, December. Singh, M., & Aitken, J. (2009). Deleveraging after Lehman–Evidence from reduced rehypothecation. IMF Working Paper.

12. The foreign exchange market1 Alain Chaboud, Dagfinn Rime and Vladyslav Sushko

Foreign exchange spot is the simplest asset class one can trade, yet it has the most complex trading environment. (Quote from an executive at a major FX liquidity provider)

12.1 INTRODUCTION1 The foreign exchange (FX) market, where the relative prices of the world’s currencies are determined, is essential for international transactions in goods, services and financial assets. In addition, FX is often viewed as an asset class on its own. The end-users of the FX market are therefore comprised of a wide variety of financial and non-financial customers around the globe. The trading activity of these agents and their interaction with market intermediaries drives the process of exchange rate determination, which has an impact on virtually all international economic activity. As a result, the FX market is the largest financial market in the world. FX trading volumes are, for example, much larger than global equity market activity (King & Rime, 2010). Currency trading takes place around the world and around the clock, with a weekly cycle beginning early Monday morning in the Asia/Pacific region and ending Friday afternoon in the Americas. Trading activity often peaks when London and New York daytime trading hours overlap and is relatively thin during the so-called “witching hour” period, the late afternoon in New York and early morning in Asia. More than 50 currencies are regularly traded, but the US dollar (USD) has for a long time commanded the dominant status of a vehicle currency. The USD is on one side of almost 90% of all global FX transactions, with the euro (EUR) and Japanese yen (JPY) in distant second and third places (BIS, 2022). Similarly, while the number of market participants around the globe is very high, much of the liquidity provision in FX is concentrated among a relatively small number of global banks and non-bank liquidity providers. Thus, despite its global and dispersed nature, certain aspects of the FX market exhibit a high degree of concentration. The market structure that supports this activity is constantly evolving. Broadly speaking, the FX market is an over-the-counter (OTC) market in which electronic trading has grown rapidly since the early 2000s. It used to be characterized by two very distinct segments, interdealer

1 We are grateful to Refet Gürkaynak and Jonathan Wright, our editors, and to Angelo Ranaldo, our discussant, for providing valuable comments. We have also benefited from conversations and email exchanges with Mark Bruce, Alexei Jiltsov, Colin Lambert, Alexis Laming, Thomas Maag, Eugene Markman, Josh Matthews, James O’Connor, Roel Oomen, Dan Reichgott, Morten Salvesen, Merhrdad Samadi, James Sinclair and Clara Vega. We thank EBS, Refinitiv and the BIS for kindly providing data. The views expressed in this chapter are solely those of the authors and should not be interpreted to represent the views of the Federal Reserve Board or of the Bank for International Settlements. 253

254  Research handbook of financial markets

and dealer to customer. This distinction has become blurred over time, with a proliferation of trading venues, a growing variety of execution methods and some non-bank actors emerging as liquidity providers alongside bank dealers. This chapter discusses how the spot FX market is organized and how it functions, including the main participants, the structure of the market and the role of the official sector. Knowing how the FX market functions is critical to understanding how it converges on equilibrium exchange rates, but this chapter does not focus on exchange rates or exchange rate determination per se, which is the subject of a vast literature.2 We note that the foreign exchange market also includes a large amount of trading in derivatives, including FX swaps, currency swaps, forwards, futures and options. FX swaps and currency swaps, important instruments primarily used for funding and hedging purposes, are covered in Chapter 20.

12.2 MARKET SIZE AND MAIN PLAYERS Measuring global trading activity presents a challenge as the global FX market is obviously not under a single jurisdiction. However, a comprehensive and authoritative source of information, albeit infrequent, is the Triennial Central Bank Survey of Foreign Exchange and Derivatives Market Activity (the Triennial). The Triennial provides a snapshot of daily FX trading activity every third year in the month of April.3 The data for the Triennial are collected by central banks from bank dealers in their jurisdictions and then aggregated, analyzed and published by the Bank for International Settlements (BIS). More frequent estimates can be obtained by using data from surveys conducted twice a year by foreign exchange committees (FXCs), industry groups sponsored by central banks in various countries. 12.2.1 Daily Trading Volumes and the Geography of Trading The average daily trading volume in the global FX spot market has generally been on an upward trend (Figure 12.1).4 According to the latest Triennial, daily trading volume in the global spot FX market averaged $2.11 trillion in April 2022.5 Illustrating the global role of the US dollar as the main vehicle currency, the top three most traded currency pairs were EURUSD, with 23% of the trading volume, followed by USDJPY with 14% and GBPUSD with 10%.6

2 For an overview of the theoretical and empirical literature on exchange rates and exchange rate determination, see, e.g., Maggiori (2022) and the chapters in James et al. (2012). 3 As of the writing of this chapter, the last Triennial survey was conducted in April 2022. 4 The notable spike in trading volume in 2014 was related to monetary easing by the Bank of Japan, while the contraction in 2015 owed importantly to the de-risking that followed the Swiss National Bank’s (SNB) surprise removal of the floor of the Swiss franc against the euro (Moore et al., 2016). 5 Average daily trading volume across all OTC FX instruments was $7.5 trillion in April 2022. Exchange-traded FX futures and options added less than $0.2 trillion to that total. 6 Each currency is assigned a three-letter code (the “ISO 4217” code). Exchange rates are then represented by a currency pair, with the base currency listed first. For instance, GBPUSD, “sterlingdollar,” is quoted in dollars per British pound sterling, while USDJPY, “dollar-yen,” is quoted in yens per dollar.

The foreign exchange market  255 2.25 1.80 1.35 0.90 0.45 0.00 1998

2001

2004

Benchmarked series London New York

2007

2010

Triennial survey Singapore Hong Kong SAR

2013 Tokyo Australia

2016

2019

2022

Canada China

Note:  The semiannual FXC data used to create the benchmarked series are collected in April and October. Turnover from the China Foreign Exchange Trade System (CFETS) and from the Hong Kong Treasury Markets Association survey was added to the stacked bars beginning in April 2015 and April 2017, respectively. Sources:   BIS Triennial Central Bank Survey; Foreign Exchange Committee Surveys; CFETS.

Figure 12.1  Daily FX spot trading volume, in $ trillions Figure 12.1 shows that trading in spot FX is concentrated in a few large financial centers. London alone accounted for 38% in 2022, while the combined share of the top four trading centers, which also included New York, Singapore and Hong Kong, amounted to 74% of global spot FX turnover (BIS, 2022).7 The major FX trading hubs are not necessarily in countries that dominate global international trade, as FX trading for financial motives, such as investments in foreign-denominated securities, far exceeds the transaction volume related to international trade. In addition, the main data centers that host the matching engines that power electronic FX trading venues form a critical part of the global FX market infrastructure. This has incentivized some key players to co-locate their trading activity near these hubs.8 12.2.2 Main Types of Counterparties Figure 12.2 shows the share of trading volume of FX bank dealers (the “reporting” dealers) with three broad counterparty categories covered by the central bank and BIS surveys: other bank dealers, other financial institutions and non-financial customers. Non-financial customers, primarily corporations, use the FX market to support their core business activities, especially international trade. The broad category of “other financial institutions” has traditionally represented the financial customers of the reporting FX dealers. This includes institutional investors, asset managers, hedge funds, commodity trading advisors (CTAs), smaller banks 7 The latter two Asian financial centers have gradually overtaken Tokyo as major trading hubs. 8 Among major data centers used for FX trading are London (LD4), New York (NY4), Tokyo (TY3) and, more recently, Singapore (SG1).

256  Research handbook of financial markets 60 50 40 30 20 10 2001

2004

Reporting dealers

2007

2010

2013

Other financial institutions

2016

2019

2022

0

Non-financial customers

Source:   BIS Triennial Central Bank Survey.

Figure 12.2  FX spot trading: percent shares by type of counterparty that are not FX dealers and central banks.9 More recently, principal trading firms (PTFs), often referred to as high-frequency traders (HFTs), have become important players within the category of “other financial institutions.” Over the past 20 years, the share of global FX trading conducted with non-financial customers has declined from about 20% to less than 10%. In contrast, the share of financial customers, which used to be close to that of non-financial customers, has grown to more than 50%. This again reflects the fact that FX trading has become increasingly driven by the needs of financial customers as opposed to the needs arising directly from international trade.10 For a long time, bank dealers constituted the sole sector that provided liquidity and warehoused risk for the rest of the market. The dealers were the archetypal liquidity providers (LPs) to financial and non-financial customers, which were the archetypal liquidity consumers (LCs). The dealers then traded among themselves to hedge their positions and rebalance their inventories, which generated much of the trading volume in the FX market. Figure 12.2 shows that in the late 1990s interdealer trading accounted for about two-thirds of all FX spot trading. The share of interdealer trading declined substantially up to 2010, partly because dealers came to rely less on interdealer trading for inventory control and partly because the line between traditional LPs and LCs has become blurred. In particular, as PTFs have increased in importance in the FX market, some have taken on a dual role as both LPs and LCs, displacing some of the dealer-to-customer activity. For instance, according to a widely

  9 Note that many of the smaller banks trade FX directly with their own customers. Unlike the reporting dealers, however, they do not act as intermediaries in the global FX market. 10 Some retail investors also participate in the FX market, often trading with high leverage via retail margin brokers. But they generally represent a very small fraction of the trading activity of the “wholesale” market participants discussed earlier. Japan may be an exception, as FX margin trading by retail investors has become more substantial there (Mukoyama et al., 2018).

The foreign exchange market  257

followed yearly survey of the FX market by Euromoney, by 2022 PTFs accounted for almost a third of electronic FX trading with customers.11

12.3 THE TRADING ENVIRONMENT The organization of the FX market may best be understood in light of three key economic frictions: credit risk management, inventory risk management and asymmetric information. Along with advances in technology, these frictions have shaped the evolution of the FX market structure over time. FX spot trades can involve very large sums and are settled two business days after the trade by the transfer of bank balances in the corresponding currencies (“T + 2” settlement, discussed in Section 12.6.3). Hence, an FX trade has historically constituted a bilateral extension of credit, naturally leading to a market structure with banks at its core (Lyons, 2002). The introduction of FX prime brokerage (PB) in the late 1990s was a major advance in the management of credit. A service first offered to hedge funds by large FX dealers, PB enabled non-banks to conduct trades directly with a broad set of counterparties under the umbrella of its PB provider. This allowed a hedge fund to gain access to pricing and liquidity from a large number of bank dealers solely on the basis of its credit relationship with its PB. Over time, prime brokerage has become widespread in the FX market. It allows various types of non-banks, including PTFs, to trade on a large number of venues, including some which were previously viewed as purely “interdealer.” Prime brokerage has therefore contributed to the broadening of liquidity provision beyond bank dealers and likely also to the proliferation of trading venues. The management of inventory imbalances arising from trading with customers used to be a key driver of interdealer trading. The repeated passing of inventory imbalances between dealers in the market, dubbed “hot potato trading” (Lyons, 1996), contributed to the large interdealer share of total volume seen in the early years of Figure 12.2. However, the way bank dealers managed their inventories began to change in the early to mid-2000s when the largest banks began to “internalize” some of their trades, waiting for an offsetting customer trade instead of immediately hedging the position in the interdealer market (Butz & Oomen, 2018). This was likely facilitated by technological advances as well as by the fairly high concentration of trading volume among the largest dealers at the time. The share of internalization in major currency pairs is now estimated to have grown to 80% or more for the largest dealers (Moore et al., 2016; Schrimpf & Sushko, 2019).12 Access to information shapes trading in all assets, and the structure of each market influences the speed of information aggregation. In the FX market, relevant information is dispersed 11 The results of these yearly surveys of the FX market, which Euromoney magazine has published for more than 40 years, are a very useful source of information on the global FX market, including about the relative importance of various liquidity providers. However, as many firms actively seek to be recognised in the survey, the results are best seen as indicative. 12 Butz and Oomen (2018) study how dealers adjust bid and ask quotes to influence the direction and arrival rate of customer orders, and they estimate that an average holding period before an offsetting trade (the “internalization horizon”) may be as short as one minute for a liquid pair like EURUSD. Barzykin et al. (2021) derive rules for dealer choices between internalization and externalization.

258  Research handbook of financial markets

among market participants, and the key intermediaries need to aggregate this information in order to set prices. Examples of relevant dispersed information are international portfolio allocation decisions by institutional investors, observations of the state of the economy in real time by firms (e.g., based on their knowledge of certain flows of imports and exports) or changes in risk preferences. Dealers then learn bits of dispersed information by observing the order flow of their various customers. Customers are not equally informed, and dealers profile them in order to better learn from the information content of their trades. Large banks that have a broad base of financial customers are better informed about FX developments than other banks (Bjønnes et al., 2021; Menkhoff et al., 2016; Ranaldo & Somogyi, 2021). It is also recognized that dealers bring their own independent information to the market (Moore & Payne, 2011). The desire of market participants to manage the information revelation process in the market has likely been one of the factors driving the recent increase in the number of trading venues and available execution protocols. As we will discuss, these trading venues can come with important differences in the degree of disclosure of counterparty identities, the ability to assess a counterparty’s price impact or the option to tailor one’s potential counterparties to a particular subset of market participants. 12.3.1 The Evolution of the FX Trading Environment over Time Figure 12.3 attempts to capture the evolution of the FX market structure over the last three decades. The upper panel, Figure 12.3a, depicts a stylized view of the market structure of the early to mid-1990s, Figure 12.3b the early to mid-2000s and Figure 12.3c the most recent decade. The shaded area denotes the interdealer market, and the surrounding area denotes the dealer-to-customer market segment. Dashed arrows indicate voice execution, and solid arrows indicate electronic execution. FX trading in the earlier period was often described as having a simple two-tier structure (Sager & Taylor, 2006). Customers traded directly with dealer banks, often by telephone or telex, in the “outer tier.” Dealers were compensated in the form of bid-ask spreads and gained private information from the trades of their own customers (Lyons, 1996). Trading among dealers, either to manage inventory risk or for speculation, then constituted the “inner tier.” Interdealer trading could take either a direct (bilateral) form or an indirect (brokered) form, initially only via voice brokers (VB).13 The advent of two electronic brokers, by Reuters (now Refinitiv) and Electronic Broking Services (EBS), in the interdealer market in the early 1990s made the process of inventory risk management and price discovery far more centralized and efficient. The two electronic brokers, both organized as central limit order books (CLOBs), quickly became the main sources of price discovery and reference prices for the entire FX spot market, and therefore began

13 Besides telephone and telex, an electronic messaging platform called Reuters Direct Dealing System was also widely used for direct dealing. According to the 2022 Triennial, 23% of FX spot trading took place via a direct voice method, such as Bloomberg chat, and another 3% used a voice broker. Voice execution could be particularly advantageous to a customer seeking to execute a large trade.

The foreign exchange market  259

Notes:   The figure builds on King et al. (2012); EB = electronic broker; MDP = multi-dealer platform; ECN = electronic communication network; PB = prime broker; PTF = principal trading firm; SDP = single-dealer platform; VB = voice broker.

Figure 12.3  Evolution of the spot FX market

260  Research handbook of financial markets

to be referred to, jointly, as the “primary market” (we henceforth refer to them as “primary CLOBs”).14 The early 2000s offered new electronic trading opportunities in the dealer-to-customer segment, as shown in the middle panel (Figure 12.3b). Multi-dealer platforms (MDPs), likely the most important innovation at that time, enabled customers to submit a request for quote (RFQ) to multiple dealers simultaneously.15 MDPs significantly broadened choices for customers and introduced an important element of competition between dealers. In response, a number of individual FX dealer banks invested in proprietary single-dealer platforms (SDPs) for direct electronic trading. Over time, these SDPs came to replace telephone or telex in much of the bilateral trading.16 The lower panel, Figure 12.3c, provides a stylized view of the current market structure, which has continued to increase in complexity. The remains of the old two-tier structure can be seen, but the structure has been perturbed by two important innovations. First, the dealerto-customer segment saw a proliferation of venues offering multiple types of execution protocols. This allowed an LP to provide liquidity to customers by, for example, streaming prices or placing limit orders in an order book, the latter resembling the trading environment of the interdealer market. We will refer to these venues collectively as electronic communications networks (ECNs), or “secondary ECNs.”17 Second, PTFs have challenged the banks within the inner tier, both as LPs and LCs on the primary CLOBs. They have also, more recently, become active in the outer tier, where they stream prices to customers via ECNs.18 PTFs have thus established a firm foothold in the parts of the FX market that used to be exclusive domains of dealer banks. At the same time, PTF trading in the FX market is enabled via prime brokerage arrangements with dealer banks, including some of the very same banks they now compete with in the market. 12.3.2 Market Fragmentation and the Declining Role of the Primary Market As shown in Figure 12.4, trading volume on the primary electronic brokers has declined substantially since the Global Financial Crisis, even as the overall volume in the FX spot 14 EBS Market became the main venue for exchange rates such as EURUSD (and its predecessors), USDJPY, EURCHF and, more recently, USDCNH, while Reuters Matching became the main venue for the “commonwealth” exchange rates, such as GBPUSD, USDCAD and AUDUSD, the Scandinavian currencies and most EME currencies. 15 An LC can also request a continuous stream of quotes for a given size and a certain amount of time (e.g., a trading day), known as a request for stream (RFS), effectively a continuous form of RFQ. 16 Although the number (and names) of platforms has changed over time, early examples of MDPs such as Currenex, Hostpot FX and FXall continue to be important, and they have often diversified their offerings of execution methods. Banks have also responded to successful third-party MDPs by establishing FXSpotStream, which operates as a bank-owned consortium and provides multibank FX streaming and RFS service to customers. As for the SDPs, as of the writing of this chapter, examples include J.P. Morgan eXecute, Deutsche Bank AutobahnFX, UBS Neo and Citi Velocity. 17 As of the writing of this chapter, examples of secondary ECNs include CboeFX, EuronextFX, LMAX and 360T, among others. For a comprehensive list of various trading venues, we refer the reader to https://www​.marketfactory​.com​/venues/. For tractability, we will continue to refer to platforms with RFQ as the main execution protocol as MDPs. 18 According to the Euromoney FX survey, examples of customer-facing PTF liquidity providers are XTX Markets, Jump Trading, HC Technologies and Citadel Securities.

The foreign exchange market  261

400 300 200 100

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Primary CLOBs

0

Futures

Notes:   Primary CLOBs refer to the wholesale electronic brokers EBS Market and Refinitiv Matching. In the latter part of the sample, the total volume may include other EBS platforms, such as EBS Direct. Exchange-traded futures volume mainly reflects that of the Chicago Mercantile Exchange, plus two smaller exchanges. Source:   EBS; Refinitiv; BIS Exchange Traded Derivatives Statistics. January 2006 to December 2022.

Figure 12.4  Primary CLOBs and currency futures exchanges, daily volumes in $ billions market has increased. The emergence of a large number of alternative trading platforms and the aforementioned decline in interdealer trading of inventory imbalances have very likely contributed to this decline. Furthermore, some have cited the opening of the primary CLOBs to PTFs (via PB arrangements) as another factor. Some participants may indeed prefer to trade on venues with less PTF participation, perhaps out of concern that their order flow may be quickly detected by the PTFs and that they may be subject to more adverse selection. Despite their decline in trading volume, the primary CLOBs are still viewed by most market participants as a key locus of price discovery, and the market data generated by the EBS Market and Refinitiv Matching platforms remain important to the market as a whole. Furthermore, when volatility spikes or market liquidity deteriorates on other venues, FX trading volume often tends to return to the EBS and Refinitiv CLOBs (Moore et al., 2016). This was the case in March 2020, for instance, when financial markets experienced turmoil at the beginning of the COVID pandemic. Some LPs increasingly look to exchange-traded FX futures for reference prices and for hedging their spot activity. This appears to be the case particularly for PTFs, many of which entered the FX market with an already established business model in futures markets. But a growing number of market participants of all types now seems to consider FX futures traded on the Chicago Mercantile Exchange (CME) as at least a close cousin of the primary CLOBs. In contrast to the primary CLOBs, trading volume in currency futures has not declined. It now often exceeds OTC FX spot trading volume in the primary market (Figure 12.4). Figure 12.5 shows the total FX spot electronic trading volume over the past four Triennial surveys. Light grey denotes multilateral platforms (CLOBs, ECNs, MDPs), while medium grey denotes direct e-trading via SDPs. The declining trading volume of the primary CLOBs (which are part of the CLOBs/ECNs/MDPs category) is also shown. Recent growth in the

262  Research handbook of financial markets 1,400 1,200 1,000 800 600 400 200 2013/04 CLOBs/ECNs/MDPs

2016/04

2019/04

SDPs

Other electronic

2022/04

0

EBS & Refinitiv Matching

Notes:   Primary CLOBs refer to the wholesale electronic brokers EBS Market and Refinitiv Matching (in the latter part of the sample, the total may include other EBS platforms, such as EBS Direct). CLOB = central limit order book; ECN = electronic communication network; MDP = multi-dealer platform; SDP = single-dealer platform. Source:   EBS; Refinitiv; BIS Triennial Central Bank Survey.

Figure 12.5  Total electronic spot trading volume, daily average in $ billions dealer-to-customer segment has come from multilateral platforms, mostly at the expense of SDPs. This again reflects mainly the growth of ECNs offering a variety of trading protocols, as market participants can “shop around” in search of the best execution. Faced with an increasing number of trading platforms and a greater variety of execution methods (such as RFQ, streaming and CLOBs), some LCs are also turning to liquidity aggregators. Liquidity aggregators are technology firms that assist their clients in accessing and comparing the various ECNs, MDPs and SDPs in real time, with the goals to obtain the best prices and maximize execution quality. 12.3.3  Mapping the Complex Execution Space Figure 12.6 presents a stylized taxonomy of the variety of electronic trading venues currently used in the spot FX market. The classification of these venues and their associated trading mechanisms is based on two important dimensions: the level of pre-trade anonymity of the counterparties and the level of “firmness” of the liquidity offered on the trading venues. This taxonomy oversimplifies what has become an increasingly complex market structure and likely misses some important exceptions, but it presents a useful way to consider the trading landscape. As shown by the horizontal arrow, the level of pre-trade anonymity can range from fully disclosed to fully anonymous. Trading on an SDP, almost by definition, is fully disclosed, as the single LP knows the identity of all the potential LCs active on its platform. Similarly, on an MDP where an LC submits an RFQ to several LPs, the identities of the potential counterparties are known before an actual trade takes place. At the other end of the spectrum, the primary CLOBs are pre-trade fully anonymous, in the sense that displayed orders do not show a source and all participants can be matched based on

The foreign exchange market  263

Notes:   A stylized taxonomy of electronic trading mechanisms in the dimensions of pre-trade anonymity and the ability to use last look. Note that the ability to use last look may or may not be exercised by an LP. Primary CLOBs refer to the wholesale electronic brokers EBS Market and Refinitiv Matching. CLOB = central limit order book; ECN = electronic communication network; MDP = multi-dealer platform; SDP = single-dealer platform; RFQ = request for quote; RFS = request for stream.

Figure 12.6  A stylized taxonomy of main trading platforms the typical price and time priority rules of a CLOB.19 Note that the futures exchange, appearing to the side, offers the highest possible level of pre-trade (and post-trade) anonymity, basically by definition, as the exchange is the counterparty to each trade. Secondary ECNs, in the middle, give their participants neither full pre-trade anonymity nor full pre-trade disclosure. This can be done, for instance, by assigning an alpha-numeric “tag” to each counterparty on the platform, a common practice on some ECNs. This may then allow a participant, having observed the execution quality of a trade against a counterparty with a particular tag, to decide whether to trade again with that unnamed counterparty. Furthermore, some other ECNs have created several separate CLOBs, each limited to a particular pool or tier of participants with a common trading style or set of characteristics. In that setting, while traders on each CLOB do not know ex-ante the identity of their counterparties, nor can follow them over time, they are assured to be matched only with a particular type of counterparty. The second dimension in Figure 12.6 is the “firmness” of the liquidity offered on an electronic platform, i.e., the level of certainty of execution that follows a request to trade on a posted quote. This depends, importantly, on whether a practice known as “last look” is used by LPs or not. Last look is a process in which an LP, or more precisely the LP’s computer, having received a request to trade on one of its posted quotes, takes a “last look” at the trade

19 Still, reflecting the OTC nature of the FX spot market, participants on the primary CLOBs can trade with each other only if they have established bilateral credit relationships, either directly or via their PBs. In practice most participants on the primary CLOBs can trade with each other.

264  Research handbook of financial markets

request before deciding whether to proceed with the transaction. The last look process occurs in a time period measured in milliseconds (Oomen, 2016; Cartea et al., 2019). Last look was developed to address a number of potential issues arising from latencies within communication networks, the multiplicity of electronic platforms, and counterparties with different levels of technological sophistication. It has become broadly acceptable to use last look to conduct quasi-instantaneous “validity and price” checks, including verifying that sufficient bilateral credit is available and that the quoted price has not become stale. Other uses of last look, including the imposition of “additional hold time” before accepting or rejecting a trade, are much more controversial (see discussion in Section 12.5.1). As shown in Figure 12.6, last look is not allowed on the primary CLOBs, and liquidity there is therefore considered to be “firm.” In contrast, last look is widely used on the secondary ECNs, with a few exceptions, and it is universally used on MDPs and SDPs, at least in the sense that LPs always reserve the option to use it. Liquidity on these platforms is therefore viewed as not completely firm: there is less certainty that a “hit” or request to trade on a quote will result in a trade. Bringing the two dimensions of Figure 12.6 together, it is evident that a higher level of pretrade anonymity is associated with firmer liquidity. Anonymity and certainty of execution in an auction structure make the pre-trade environment on primary CLOBs closest to a full exchange. This is likely one of the reasons why, despite their decline in trading volume, the prices shown on the primary CLOBs still tend to be viewed as reference prices for the entire FX market.

12.4 ALGORITHMIC TRADING IN FX The proliferation of electronic execution in the spot FX market, now near 75% even for dealerto-customer trading (BIS, 2022), has been accompanied by the growth of algorithmic (computer-driven) trading. A significant fraction of both LPs and LCs now use algorithms to drive at least some of their trading activity, with limited “human” intervention. The expansion of algorithmic trading in the FX market occurred first on the two primary CLOBs, but algorithmic trading has now expanded well beyond the primary market.20 Banks and non-banks deploy a variety of algorithms in the FX market, for example execution algorithms that attempt to minimize the price impact of large trades, market-making algorithms which automatically post executable quotes or opportunistic algorithms which monitor the market in search of profit or hedging opportunities. An important distinction among these algorithms is whether they automate only the execution of the trade or whether the decision to trade (the investment decision) is also left to the algorithm. Figure 12.7 shows how the share of “manual trading” and “API trading” on the EBS Market platform has evolved since 2004, illustrating the growth of algorithmic trading on one of the primary CLOBs. At the beginning of the sample, only banks were allowed on the platform and all trading was done manually by entering trading instructions on the specialized keyboard of an EBS terminal. After EBS allowed the direct interface of computers with its platform, banks began to trade algorithmically in 2004 (bank API). Non-banks began in 2005 20 Because of the role that the primary CLOBs play in the market, algorithmic trading has had an important impact on price discovery in FX (Chaboud et al., 2014).

The foreign exchange market  265

80 60 40 20

04

05

06

Manual

07

08

09

10

Bank API

11

12

13

14

15

16

17

18

19

20

21

22

0

Non-bank API

Notes:   The figure shows the proportion of trading volume on EBS Market executed manually versus algorithmically by banks and non-banks. Source:  EBS.

Figure 12.7  Manual versus algorithmic execution on EBS Market (non-bank API) when they were first allowed on the EBS platform through PB arrangements. Note that, broadly speaking, the vast majority of the non-bank API trading shown currently represents the activity of PTFs, mostly using algorithms that automate both the decision to trade and the execution of the trade. In contrast, an important fraction of the bank API activity likely reflects trades where the execution is automated but the initial decision to trade is taken by humans.21 As shown in Figure 12.7, by the end the sample, in 2022, algorithmic trading dominates. Bank API and non-bank API activity each account for a bit more than 40% of total trading volume, shares which have remained fairly even since 2015. The share of manual trading, in contrast, has continued to decline and now accounts for only about 15% of trading volume.22 Numerous types of algorithms are routinely used by banks and non-banks to provide and consume liquidity, as discussed earlier. Among those, “latency arbitrage” algorithms are a somewhat notorious sub-type of opportunistic algorithms which exploit a firm’s speed advantage to generate profit opportunities.23 Latency arbitrage algorithms are believed to be used almost exclusively by a subset of the PTFs active in the market. In an attempt to protect their 21 For example, a bank may decide to cover some of the order flow on its own SDP by trading on a primary CLOB using an execution algorithm. 22 The evolution of these shares over time has some notable features. For instance, manual trading drops in early 2020 as the COVID pandemic begins, probably reflecting the increase in workingfrom-home arrangements, bank algorithmic activity rises beginning in late 2014, likely due to the switch to algorithmic execution around fixings by a number of banks after the fixing scandal (see Section 12.6.2), and non-bank algorithmic activity peaks in 2011 after EBS reduces its tick size (Chaboud et al., 2021). 23 For instance, a trader with faster technology than a certain LP may take advantage of the small time lag between when a market-moving trade takes place and when the LP updates its posted quotes in response, quickly trading on one of these now outdated quotes.

266  Research handbook of financial markets

clients from such strategies, particularly on platforms where manual traders and algorithmic traders coexist, several trading venues in the FX market have introduced various types of “speed bumps.” ParFX, a platform launched in April 2013 by a group of large dealer banks, added randomized pauses (ranging from 20 to 80 milliseconds at the onset) to all trading instructions before processing them. EBS Market later in 2013 introduced a “latency floor” of a few milliseconds, short periods during which all trading messages are batched and randomized before being released to the order book. In 2014, Reuters/Refinitiv Matching, the other primary CLOB, introduced its own version of a speed bump.24 Finally, while execution algorithms have long been used by banks to optimize the execution of large trades, such algorithms are now also being used directly by some of the more sophisticated customers in the FX market. As described in a recent report on the use of execution algorithms in FX (Bank for International Settlements, 2020), sophisticated customers increasingly rely on smart order routing and execution algorithms to spread large orders over time and across multiple electronic venues. Customers in FX also increasingly use Transaction Cost Analysis (TCA) to monitor the execution quality of their trades.

12.5 THE ROLE OF THE OFFICIAL SECTOR 12.5.1 Regulation, Good Practice, the FX Global Code The level of regulatory oversight of the foreign exchange market varies greatly from country to country, depending importantly on the country’s exchange rate regime and the restrictions on cross-border capital movements.25 The regulatory authority can reside with the central bank, the ministry of finance or another government agency. In the case of major floating exchange rates (often called the G-10 exchange rates), regulatory constraints are generally light compared to equity and bond markets. This is likely due in great part to the fact that an FX transaction does not entail the exchange of a security and that, by nature, the currencies of two different countries are involved. Even when they do not have formal regulatory authority over the FX market, many central banks sponsor Foreign Exchange Committees, where a variety of market participants meet regularly to discuss market functioning and encourage “good practice.” In addition, the central banks of countries with major trading centers meet regularly under the auspices of the BIS to discuss the global foreign exchange market. It is under this framework that a broad group of central banks and market participants cooperated to develop the FX Global Code, first published in 2017 and updated in 2021. The FX Global Code, which is maintained by the Global Foreign Exchange Committee (GFXC) lists and discusses principles of good practice for the global FX market.26 Specifically, the Code contains 55 principles of good practice in the areas of ethics, governance, execution, 24 The speed bump introduced by Refinitiv was viewed as an evolution of the EBS speed bump. In particular, the Refinitiv mechanism processes order cancellation messages before other types of messages when releasing each batch to the order book, providing additional protection against aggressive latency arbitrage strategies. That feature was later also adopted by EBS. 25 The IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions is a comprehensive source of information on that topic. 26 The GFXC was created in 2017, and its members are the various FXCs from around the world. The Code is available on its website: http://www​.globalfxc​.org.

The foreign exchange market  267

information sharing, risk and compliance and confirmation and settlement. Market participants are encouraged to sign a “statement of commitment” to the Code to publicly signal their intention to adhere to its principles. The Code has gained broad international acceptance, particularly among large bank dealers, and it is increasingly viewed as having an important influence on the behavior of FX market participants. For instance, the 2021 update of the Code was accompanied by guidance on the proper use of last look. The guidance reaffirmed that last look should only be used for price and validity checks, and recommended that these checks should be applied “without delay.” Market participants viewed that recommendation as advocating against a widely used, but controversial, practice where some liquidity providers wait to observe additional price movements before accepting or rejecting a trade. While LPs argue that this “additional hold time” protects them against possible adverse selection, opponents point out that it allows LPs to only accept the most profitable trades. In the event, the guidance by the GFXC had a rapid and substantial impact on the behavior of market participants: numerous large bank dealers announced an end to additional hold time, and some trading venues announced a reduction in the maximum length of their “last look window.” In the end, while still a fairly new experiment, the effect of the FX Global Code on the FX market is seen by some as evidence that a “soft” approach to regulation, which encourages good practice instead of imposing hard rules, can have a substantial impact on an important financial market.27 12.5.2 FX Intervention and Exchange Rate Management The official sector, mainly central banks and finance ministries, often has a critical influence on floating exchange rates.28 Participants in FX markets are closely attuned to official macroeconomic data releases and central bank communications, and major exchange rates routinely react within milliseconds to the surprise component of these news. In addition, a more direct and willful way in which the official sector can affect exchange rates is via foreign exchange intervention, the purchase or sale of its currency in the FX market with the specific intent to influence exchange rates. Many major industrialized countries with floating exchange rates, including the United States, used FX intervention fairly frequently until the mid-1990s.29 Since then, with a few notable exceptions, including Japan in the early 2000s and Switzerland in the years since the Global Financial Crisis, FX intervention by major industrialized countries has become rare30 FX

27 Skeptics may argue that the “soft” approach to regulation is only effective because deviating from accepted good practice could open a financial firm to civil legal liability. 28 The official sector obviously also plays a critical, but different, role in cases where countries choose to peg or closely manage the exchange value of their currency. 29 During that era, the intervention episodes that followed the 1985 Plaza Accord, an agreement among G5 countries to weaken the dollar, and the 1987 Louvre Accord, a G7 agreement to stabilize dollar exchange rates, are particularly well known. Several European countries also intervened heavily in the 1992 as the European Exchange Rate Mechanism (ERM) came under pressure, especially just before Great Britain withdrew from the ERM. 30 Japan has intervened in FX markets on a few occasions since the 2000s, including in 2011 following the earthquake and nuclear accident and, most recently, in late 2022.

268  Research handbook of financial markets

intervention by developing and newly industrialized countries has, in contrast, remained quite frequent.31 A distinction is often made between “sterilized” and “unsterilized” FX intervention, where sterilizing an intervention operation means offsetting the associated effects on the domestic monetary base, typically by buying or selling bonds. Historically, most FX intervention operations have been sterilized. By design, unsterilized FX intervention is as much monetary policy as it is foreign exchange policy.32 A substantial academic literature has discussed the effects of FX intervention. Earlier work focused mainly on interventions by major industrialized countries (Dominguez & Frankel, 1993), but recent work has expanded to a study of the impact of FX intervention among a broader range of countries (Fratzscher et  al., 2019), including countries which implement a “managed float” for their currencies (Frankel, 2019; Cavallino, 2019). Broadly speaking, research has traditionally discussed three channels through which FX intervention may affect exchange rates: a portfolio balance channel, when investors view domestic and foreign assets as imperfect substitutes; a signaling channel, where FX intervention is interpreted by the market as conveying information about future central bank policy; and a coordination channel, where intervention activity affects the trading behavior of FX traders in a way that reinforces the central bank’s activity. There is broad agreement that FX intervention, particularly if it is large, can have an immediate effect on exchange rates, but there is little consensus about the medium and long-term impact of FX intervention on exchange rates. In general, research has found that announcing an FX intervention, intervening in a direction consistent with monetary policy and coordinating intervention among central banks can enhance the short-term impact of FX intervention. In addition, the impact of FX intervention in currencies outside of the few major currencies often seems to be larger and longer lasting. This suggests that FX interventions can have a larger impact in the presence of market imperfections, such as limited intermediation capacity (Gabaix & Maggiori, 2015) and a number of other financial frictions that render domestic and foreign assets imperfect substitutes (Popper, 2023).

12.6 SPECIAL TOPICS 12.6.1 Flash Events and Other Extreme Events Similar to several other financial markets with a large share of electronic trading, the FX market has in recent years experienced a number of flash events, with very sharp but often quickly retraced exchange rate movements accompanied by a temporary but almost complete disappearance of market liquidity. The analysis of these events can provide insights into the behavior of market participants and into the impact of specific aspects of the market structure.

31 An increasing number of central banks in emerging markets also conduct FX interventions using derivatives, allowing them to economize on FX reserves. 32 The recent intervention purchases of foreign currency by the SNB are an important example. In that case, the SNB specifically decided not to sterilize its intervention operations, resulting in a large increase in the size of its balance sheet, in essence conducting quantitative easing via the purchase of foreign currency.

The foreign exchange market  269

Among prominent recent events, on 7 October 2016, the GBP quickly depreciated almost 9% against the USD before recovering much of its losses within minutes (Bank for International Settlements, 2017); then on 3 January 2019, the JPY appreciated sharply against the USD and a number of other currencies before quickly retracing much of the move. The two events share some interesting characteristics. They both occurred in the late afternoon in the United States (early morning in Asia), a time of day when the FX market typically exhibits its lowest trading volume and liquidity. In both cases, the linkages between the spot and the futures market seemed to be at play. Liquidity in the GBP fell to its lowest level precisely when the futures contracts at the CME experienced a trading halt in response to a large, but still fairly orderly, price movement. The JPY event occurred during the one hour of the day when there is no futures trading on the CME.33 The two flash events also have some individual peculiarities. In particular, the JPY event involved the unwinding by retail FX investors of positions in higher-yielding currencies such as the Turkish lira and the Australian dollar.34 Finally, what is viewed as the most extreme event in the FX market in recent years was not a flash event per se, as it had a clearly understood trigger. On 15 January 2015, the SNB removed the floor that it had imposed on the EURCHF exchange rate since 2011, a floor that had effectively prevented the appreciation of the CHF (Breedon et al., 2022). The SNB had maintained that floor by routinely intervening in the FX market, purchasing euros in an amount sufficient to keep the exchange rate at or above 1.20 Swiss francs per euro. The removal of the floor had not been expected by market participants (Mirkov et al., 2016), and the Swiss franc quickly appreciated up to 40% against the euro, forcing the SNB to resume intervention purchases of euros after about 20 minutes. The Swiss franc ended the day about 15% stronger against the euro. The impact of the associated volatility on leveraged FX positions led to a widespread review of risk-management practices in the FX industry. 12.6.2 Foreign Exchange Benchmark Rates and the “Fixing Scandal” Benchmark exchange rates, also commonly called “fixes,” are calculated and published numerous times a day. They are widely used for a number of purposes, including valuing portfolios that contain foreign-currency-denominated securities, constructing multi-country indices, such as the MSCI stock indices, and even settling some derivatives transactions. The best-known and most widely used of these benchmark rates is the one calculated at 4 p.m. in London by WMR.35 Because some important FX market participants, such as asset managers, routinely request transactions executed at the 4 p.m. WMR rates, the FX market experiences a substantial daily spike in trading volume around that time of the day. Especially large volumes occur at the end of each month, associated with the rebalancing of multi-country portfolios. There are other prominent daily fixes in the FX market, including the ones calculated by the ECB at 2:15 p.m. CET and the Tokyo morning fixes calculated by various Japanese banks at 9:55 a.m. (Financial Stability Board, 2014; Ito & Yamada, 2017). 33 FX futures at the CME trade 23 hours a day five days a week. Trading stops for one hour between 5 p.m. and 6 p.m. ET. 34 The retail FX sector in Japan is larger than in other developed countries, with a substantial number of retail FX margin traders. When the yen appreciates suddenly, currency moves can be amplified by triggering stop-loss orders (Tomohiro Niimi, 2016). 35 WM/Refinitiv (previously known as WM/Reuters) calculates benchmark rates every hour for a large number of spot, forward and NDF exchange rates.

270  Research handbook of financial markets

The role of these fixes, with very few exceptions (Melvin & Prins, 2015), attracted very little attention outside of the industry before 2013. This changed after accusations emerged in 2013 that some traders at large banks may be manipulating benchmark exchange rates (Vaughan et al., 2013). In the end, what became known as the FX fixing scandal resulted in substantial fines being imposed on dealing banks in a number of countries, and reforms of the fixing process were implemented in 2015 after a study by the Financial Stability Board (Financial Stability Board, 2014, 2015). Among other reforms, the window over which the 4 p.m. fix is calculated was lengthened from one minute to five minutes with the goal to reduce the potential for manipulation.36 Despite the 2015 reforms, the issue of benchmark exchange rates continues to be a topic of concern. In particular, traders have pointed to ongoing unusual volatility patterns around the 4 p.m. fix, especially at month ends, leading to some calls for further reforms. The issue of FX benchmarks has also generated some academic research, with analysis focusing on the initial issue (Evans, 2018; Ito & Yamada, 2017; Osler & Turnbull, 2017) and on the impact of the fixing reform (Evans et al., 2018). 12.6.3 Settlement Risk Transactions in the “spot” FX market, basically by definition, settle by the exchange of balances in one currency against balances in another. For the vast majority of currency pairs, settlement occurs two business days after the trading date, that is at “T + 2.”37 FX settlement risk is the risk of loss incurred by a counterparty when it pays out the currency it sold but does not receive the currency it bought. The bankruptcy of Bankhaus Herstatt in 1974 demonstrated how FX settlement risk can undermine financial stability. Herstatt was a medium-sized German bank active in FX markets. On 26 June 1974, the German authorities closed the bank down. While Herstatt had already received Deutsche marks from its counterparties, it had not yet made the corresponding US dollar payments in New York. Herstatt’s failure to pay led a number of banks to stop outgoing payments until their own payments due had been received. The consequences were systemic: the international payment system was severely affected for a period of time, and the erosion of trust caused lending rates to spike and credit to be curtailed. The G-10 central banks launched in 1996 a comprehensive strategy to reduce FX settlement risk, resulting in the launch of Continuous Linked Settlement (CLS) in 2002. CLS is a specialized institution that settles FX transactions on a payment-versus-payment (PvP) basis, addressing settlement risk by ensuring that a payment in a currency occurs if and only if the payment in the other currency takes place.38 The establishment of CLS and other actions led to a substantial reduction in FX settlement risk. Kos and Levich (2016) estimate that the share of FX turnover settled by means of traditional correspondent bank arrangements declined from 85% to 13% between 1997 and 2013. 36 In late 2015, in an effort to discourage trading activity related to its daily FX reference rates, and thereby reduce the potential for manipulation, the ECB delayed their publication from 2:30 p.m. to 4:00 p.m. CET. 37 Transactions involving a handful of currency pairs settle at T + 1, most prominently USDCAD. By convention, the “value date” for almost all FX trades around the world changes from one day to the next at 5 p.m. ET. Trades involving the New Zealand dollar are an exception. 38 Several regional PvP systems, such as Hong Kong CHATS, have also been introduced.

The foreign exchange market  271

The share of global FX transactions settled through CLS reached about 50% in 2013, but more recently the pace of adoption of PvP settlement appears to have slowed (Bech & Holden, 2019). Among likely causes are the growth in trading volume in emerging market currencies that are not CLS-eligible and the internalization of customer flow by LPs, in which case the dealer is likely to settle trades across its own books. There has been a renewed effort in recent years to reduce the remaining FX settlement risk in the system, including efforts to bring PvP settlement to additional currencies. In addition, the 2021 update of the FX Global Code now calls for market participants to use PvP settlement where available.

12.7 TOPICS FOR FUTURE RESEARCH The foreign exchange market has gone through major changes since the early 2000s, and this chapter concludes with a few open questions for future research grouped under three broad topics: liquidity and market fragmentation; new market intermediaries and information; and the future of the FX market. 12.7.1 Liquidity and Market Fragmentation The decline in activity of the interdealer electronic brokers and the rapid growth in the number of electronic trading venues and protocols, particularly in the dealer-to-customer segment, have resulted in an increasingly fragmented trading environment. How has this affected liquidity in the market overall? Moreover, in a market where trading activity is widely dispersed, how does one properly measure market liquidity?39 In such a fragmented market, market depth aggregated across all venues likely overstates the true depth of the market, resulting in a “liquidity mirage.” The existence of last look further exacerbates this issue, because it allows LPs to more easily post the same liquidity simultaneously across many venues, knowing that they can reject some requests to trade. In such an environment, measures of liquidity besides bid-ask spreads and depth, such as estimated price impacts, may therefore be more useful. Last look also allows LPs to post tighter bid-ask spreads, as these spreads no longer have to compensate them for the risk of an adverse price movement before a quote is hit. Importantly, this means that the FX market is now populated by a mixture of firm and non-firm liquidity, and the liquidity subject to last look is likely being offered at a better price. Customers trying to access that cheaper liquidity may however not be able to complete the trade. What is the “true” bid-ask spread in such a market? Looking at the bigger picture, could the non-firm liquidity eventually drive out the firm liquidity, or is the coexistence of firm and non-firm liquidity a stable equilibrium? The answer

39 Research on FX market liquidity has frequently been limited by data availability. Using Reuters data, Evans and Lyons (2002) find that FX liquidity is time-varying and is inversely related to the amount of public information transmitted to the market. Mancini et al. (2013) use EBS data to document that FX market liquidity co-moves with other financial markets, with common effects across currency pairs. More recently, Hasbrouck and Levich (2019) combine CLS data with commercial FX quotes data to design more comprehensive test of liquidity commonality.

272  Research handbook of financial markets

to this question must depend importantly on the amount of information available to LCs, allowing them to evaluate the trade-off between the certainty of execution and the price. Finally, if liquidity is a priced risk factor, reflecting the risk of illiquidity when execution is most needed, could last look also impact the level of exchange rates? 12.7.2 New Intermediaries and Price Discovery Changes in the composition of market intermediaries in the FX spot market also raise interesting research questions. As in several other important financial markets, the bank dealers now compete in the role of main liquidity providers with PTFs. The PTFs are lightly capitalized non-banks that rely more heavily on speed and correlations with other asset classes to hedge their positions. How has the risk-bearing capacity of FX market intermediaries, in total, changed as a result? Some observers argue that the market’s risk-bearing capacity must have declined, even as average trading costs have likely declined as well. If so, is there evidence that this has affected market volatility or the frequency of flash events? The evolution of market intermediaries also raises questions about information and price discovery in this market. The academic literature has long documented the explanatory power of certain customer order flows for exchange rate movements, information that only bank dealers typically have had access to. Has the growing importance of PTFs reduced the explanatory power of customer order flow? Furthermore, has the presence of PTFs in this market lowered the incentives of bank dealers to seek and study fundamental information that may affect the evolution of exchange rates, as in Weller (2018)? There are also important questions about where price discovery occurs in this increasingly fragmented market. Do the primary CLOBs, despite their decreased trading volume, still constitute the main locus of price discovery in the FX market? This chapter has discussed factors that argue in favor of that hypothesis. Alternatively, has price discovery, at least in part, moved to other sections of the market? The smaller size of the primary CLOBs, together with the increased role of PTFs who often rely on correlations with futures to price their spot liquidity, has led some market participants to argue that FX futures do, in some cases, lead price discovery in FX spot.40 Can one envision in the future a highly fragmented FX spot market with price discovery occurring mainly in FX futures?41 12.7.3 The Future of the FX Market The structure of the global FX market has been in a constant state of flux for many years. An interesting aspect of the FX market is that it is generally under very light regulatory oversight, particularly as far as how trading venues operate. As a result, the structure and operation of the FX market have mostly developed endogenously, driven by commercial interests and the market participants’ needs. For instance, platforms can easily set and change parameters such 40 Chaboud et al. (2021) show some evidence of this occurring as early as 2011. Comparing price discovery in FX spot and futures markets used to be an active area of research (e.g., Tse et al. [2006], Rosenberg and Traub [2009] and Chen and Gau [2010]. 41 Note that, to a large extent, this occursin the US equity market (with E-mini S&P 500 futures) and the German sovereign bond market (with Bund futures).

The foreign exchange market  273

as the minimum trade size, minimum price increment, whether last look is allowed, the level of trader anonymity, the content and speed of data updates, etc., as they compete to attract trading volume. The FX market should therefore provide a rich environment where both theoreticians and empirical economists can address questions such as what factors drive the evolution of a market and which platform characteristics are important to lead price discovery. Finally, although much is speculative, the effect of new financial technologies (blockchain, cryptocurrencies, decentralized finance) on the FX market will be an important topic of research in the coming years. This will particularly be the case if central bank digital currencies (CBDC) are issued by major jurisdictions around the world. This could affect not just the settlement of FX transactions,42 but also impact price discovery and the role of the current intermediaries. That, in turn, could eventually lead to fundamental changes in the structure of the foreign exchange market.

REFERENCES Bank for International Settlements. (2017). The sterling “flash event” of 7 October 2016. Markets Committee Papers, 9. Bank for International Settlements. (2022). Triennial Central Bank survey of foreign exchange and over-the-counter (OTC) derivatives markets in 2022, October. Bank for International Settlements. (2020). FX execution algorithms and market functioning. Markets Committee Papers, 13. Banque de France, Bank for International Settlements and Swiss National Bank. (2021). Cross-border settlement using wholesale CBDC. Project Jura, BIS Innovation Hub. Barzykin, A., Bergault, P., & Guéant, O. (2021). Market making by an FX dealer: Tiers, pricing ladders and hedging rates for optimal risk control. Preprint arXiv:2112.02269, arXiv. Bech, M. L., & Holden, H. (2019). FX settlement risk remains significant. BIS Quarterly Review, 48–49, December. Bjønnes, G. H., Osler, C. L., & Rime, D. (2021). Price discovery in two-tier markets. International Journal of Finance and Economics, 26(2), 3109–3133. Breedon, F., Chen, L., Ranaldo, A., & Vause, N. (2022). Judgement day: Algorithmic trading around the Swiss franc cap removal. Journal of International Economics, 140. Butz, M., & Oomen, R. (2018). Internalisation by electronic FX spot dealers. Quantitative Finance, 19(1), 35–56. Cartea, Á., Jaimungal, S., & Walton, J. (2019). Foreign exchange markets with last look. Mathematics and Financial Economics, 13(1), 1–30. Cavallino, P. (2019). Capital flows and foreign exchange intervention. American Economic Journal: Macroeconomics, 11(2), 127–170. Chaboud, A. P., Chiquoine, B., Hjalmarsson, E., & Vega, C. (2014). Rise of the machines: Algorithmic trading in the foreign exchange market. Journal of Finance, 69(5), 2045–2084. Chaboud, A. P., Dao, A., & Vega, C. (2021). What makes HFT tick? Tick size changes and information advantage in a market with fast and slow traders. Available on SSRN. Chen, Y.-L., & Gau, Y.-F. (2010). News announcements and price discovery in foreign exchange spot and futures markets. Journal of Banking and Finance, 34(7), 1628–1636. Dominguez, K. M., & Frankel, J. A. (1993). Does foreign exchange intervention work? New York: Columbia University Press.

42 For instance, the BIS Innovation Hub recently conducted a cross-border settlement of EUR versus CHF FX transaction between a French and a Swiss commercial bank using wholesale CBDC issued by the Banque de France and the Swiss National Bank (Banque de France, Bank for International Settlements and Swiss National Bank, 2021).

274  Research handbook of financial markets

Evans, M. D. (2018). Forex trading and the WMR fix. Journal of Banking and Finance, 87, 233–247. Evans, M. D. D., & Lyons, R. K. (2002). Time-varying liquidity in foreign exchange. Journal of Monetary Economics, 49(5), 1025–1051. Evans, M. D., O’Neill, P., Rime, D., & Saakvitne, J. (2018). Fixing the fix? assessing the effectiveness of the 4pm fix benchmark. Occasional Paper 46, FCA. Financial Stability Board. (2014). Final report on foreign exchange benchmarks, September. Financial Stability Board. (2015). Foreign exchange benchmarks: Report on progress in implementing the September 2014 recommendations, October. Frankel, J. (2019). Systematic managed floating. Open Economies Review, 30(2), 255–295. Fratzscher, M., Gloede, O., Menkhoff, L., Sarno, L., & Stöhr, T. (2019). When is foreign exchange intervention effective? Evidence from 33 countries. American Economic Journal: Macroeconomics, 11(1), 132–156. Gabaix, X., & Maggiori, M. (2015). International liquidity and exchange rate dynamics. Quarterly Journal of Economics, 130(3), 1369–1420. Hasbrouck, J., & Levich, R. M. (2019). FX liquidity and market metrics: New results using CLS Bank settlement data. Typescript, Stern NYU. Ito, T., & Yamada, M. (2017). Puzzles in the Tokyo fixing in the forex market: Order imbalances and bank pricing. Journal of International Economics, 109, 214–234. James, J., Marsh, I. W., & Sarno, L. (Eds.). (2012). The handbook of exchange rates. Chichester: Wiley. King, M. R., Osler, C. L., & Rime, D. (2012). Foreign exchange market structure, players and evolution. In J. James, I. W. Marsh & L. Sarno (Eds.), The handbook of exchange rates. Chichester: Wiley. King, M. R., & Rime, D. (2010). The $4 trillion question: What explains FX growth since the 2007 survey? BIS Quarterly Review, 4, 27–42. Kos, D., & Levich, R. M. (2016). Settlement risk in the global FX market: How much remains? Typescript, Stern NYU. Lyons, R. K. (1996). Foreign exchange volume: Sound and fury signifying nothing? In J. A. Frankel, G. Galli & A. Giovannini (Eds.), The microstructure of foreign exchange markets. Chicago, IL: University of Chicago Press. Lyons, R. K. (2002). The future of the foreign exchange market. In R. E. Litan & R. J. Herring (Eds.), Brookings-Wharton papers on financial services. Brookings Institution. Maggiori, M. (2022). International macroeconomics with imperfect financial markets. In G. Gopinath, E. Helpman, & K. Rogoff (Eds.), Handbook of International Economics (Vol. 6). Elsevier. Mancini, L., Ranaldo, A., & Wrampelmeyer, J. (2013). Liquidity in the foreign exchange market: Measurement, commonality, and risk premiums. Journal of Finance, 68(5), 1805–1841. Melvin, M., & Prins, J. (2015). Equity hedging and exchange rates at the London 4 p.m. fix. Journal of Financial Markets, 22, 50–72. Menkhoff, L., Sarno, L., Schmeling, M., & Schrimpf, A. (2016). Information flows in foreign exchange markets: Dissecting customer currency trades. Journal of Finance, 71(2), 601–634. Mirkov, N., Pozdeev, I., & Söderlind, P. (2016). Toward removal of Swiss franc cap: Market expectations and verbal interventions. Working paper 2016-10, Swiss National Bank. Moore, M. J., & Payne, R. (2011). On the sources of private information in FX markets. Journal of Banking and Finance, 35(5), 1250–1262. Moore, M., Schrimpf, A., & Sushko, V. (2016). Downsized FX markets: Causes and implications. BIS Quarterly Review, 35–51. December. Mukoyama, Y., Kikuta, N., & Washimi, K. (2018). Investment patterns of Japanese retail investors in foreign exchange margin trading. Bank of Japan review Series 18-E-3. Niimi, T. (2016). Recent trends in foreign exchange (FX) margin trading in Japan. Bank of Japan Review 2016-E-5. Oomen, R. (2016). Last look. Quantitative Finance, 17(7), 1057–1070. Osler, C. L., & Turnbull, A. (2017). Dealer trading at the fix. Working Paper 101R, Brandeis University. Popper, H. (2023). Foreign exchange intervention. In Oxford Research Encyclopedia of Economics and Finance. Ranaldo, A., & Somogyi, F. (2021). Asymmetric information risk in FX markets. Journal of Financial Economics, 140(2), 391–411.

The foreign exchange market  275

Rosenberg, J. V., & Traub, L. G. (2009). Price discovery in the foreign currency futures and spot market. Journal of Derivatives, 17(2), 7–25. Sager, M. J., & Taylor, M. P. (2006). Under the microscope: The structure of the foreign exchange market. International Journal of Finance and Economics, 11(1), 81–95. Schrimpf, A., & Sushko, V. (2019). FX trade execution: Complex and highly fragmented. BIS Quarterly Review, 39–51, December. Tse, Y., Xiang, J., & Fung, J. K. W. (2006). Price discovery in the foreign exchange futures market. Journal of Futures Markets, 26(11), 1131–1143. Vaughan, L., Finch, G., & Choudhury, A. (2013, June 12). Traders said to rig currency rates to profit off clients. Bloomberg News. Weller, B. M. (2018). Does algorithmic trading reduce information acquisition? Review of Financial Studies, 31(6), 2184–2226.

PART IV CAPITAL MARKETS

13. The Treasury and when-issued markets J. Benson Durham and Roberto Perli

13.1 A SUMMARY DESCRIPTION OF THE TREASURY MARKET Treasury securities, or “Treasuries,” are securities issued by the U.S. government to finance expenditures not covered by revenues. Because they are backed by the “full faith and credit” of the U.S. government, Treasuries are widely considered credit risk-free assets.1 As U.S. government spending has increased significantly faster than revenues, the amount of debt that the Treasury has needed to fund with Treasury issuance has surged correspondingly. As Figure 13.1 shows, Treasury debt as of the second quarter of 2021 totaled $28.5 trillion.2 13.1.1 Marketable versus Nonmarketable Securities Not all the Treasury debt is financed via publicly tradable securities—a portion is held by government trust funds, revolving funds, and other special funds. These securities are referred to as “intragovernmental holdings” and are “nonmarketable,” because they are not negotiable, transferable, or tradable in the secondary market. As illustrated in Figure 13.1, intragovernmental holdings of Treasuries were about $6.2 trillion at the end of 2021Q2. Marketable securities, also referred to as the “debt held by the public,” were instead about $22.3 trillion.3 Notably, debt held by the Federal Reserve in its System Open Market Account (SOMA) is deemed marketable and is part of publicly held debt. When market participants refer to the “Treasury market,” they generally have in mind the amount of marketable securities, not the size of the U.S. debt.

1 The amount of debt the U.S. Treasury can issue is subject to the “debt ceiling” established by the Congress. When the debt reaches this level, the Congress must raise the ceiling to allow additional Treasury borrowing. At times, the political process of raising the ceiling has been contentious enough to raise the possibility of partial default (the fact that the U.S. Congress can pass revenue and spending bills and yet simultaneously impose a cap on the implied gap between the two is a peculiar complication of the Treasury market). In 1979, the Treasury did in fact default on a small portion of its debt (see Zivney and Marcus, 1989). Even when default happened or is at risk of happening because of debt ceiling tensions, the market widely deems such defaults as “technical” dislocations that will be resolved by subsequent debt ceiling increases. Political shenanigans aside, convention treats Treasuries as credit risk-free securities. 2 The most recent statistics are in the current issue of the Bureau of the Fiscal Service’s Treasury Bulletin. 3 Technically, the debt held by the public also includes a small amount of debt issued by the Federal Financing Bank (FFB), a government entity established to help reduce federal borrowing costs. The FFB issues debt to the public to finance its operations. 277

278  Research handbook of financial markets

Source:   Bloomberg and authors’ calculations.

Figure 13.1  Size and composition of Treasury debt since 1970 13.1.2 Types of Marketable Securities Treasury securities are divided into three broad categories—bills, notes, and bonds. Treasury bills are short-term securities with maturities that range between a few days (referred to as “cash-management” bills) and one year. Bills are “discount securities” issued at less than face value, that pay face value at maturity, and that do not offer a periodic coupon (i.e., they are “zero-coupon” securities). Treasury notes have initial maturities between two and ten years (currently two, three, five, seven, and ten years). Like for bills, the principal is paid back entirely at maturity but unlike bills, notes pay a semiannual interest coupon. Coupons can be “fixed” (constant throughout the life of the security) or “floating” (variable depending on market fluctuations). Most notes (95.4% of the total as of 2021Q2) are fixed-rate. Treasury bonds have initial maturities greater than ten years (currently 20 and 30 years). Similar to notes, the principal is paid entirely at maturity, and interest coupons are paid on a semiannual basis. All bonds issued so far have been fixed rate. Notes and bonds can be either “nominal” and pay the investor a constant amount at maturity regardless of the evolution of inflation over the life of the security, or “real.” Real Treasury securities, more properly referred to as Treasury Inflation-Protected Securities, or TIPS, are characterized by a principal amount that adjusts with inflation, as measured by the headline Consumer Price Index (CPI). At maturity, TIPS pays back the greater of the original principal amount or the principal amount adjusted for the total inflation over the life of the security, albeit with a three-month lag. TIPS, therefore, ensure that inflation will not erode the original principal amount. Also, before maturity, TIPS pay a semiannual fixed coupon applied to the principal adjusted for inflation; consequently, interest payments, like the final principal payment, vary with inflation.4

4 For discussions of pricing and issuance in the TIPS market, see Fleckenstein et  al. (2014) and Dudley et al. (2009) among others.

The Treasury and when-issued markets  279

Source: Bloomberg and authors’ calculations.

Figure 13.2   Average maturity of marketable securities and the slope of the yield curve  13.1.3 Determinants of Treasury Issuance Patterns One interesting question is how and why the Treasury selects what securities to issue. The answer is multifaceted, but in a nutshell the Treasury’s main objective is to finance government borrowing at the lowest cost (see Driessen, 2016).5 In periods of very low interest rates, the Treasury, in principle, could greatly increase bond issuance to secure low cost of funds for a long time. But the demand side of the market needs to be considered too. If bond supply considerably exceeded demand, interest rates would rise and offset the benefit of increasing issuance. And maintaining a balance between demand and supply promotes a stable and efficient market, which in turn reduces financing costs over time.6 In the end, the average maturity of Treasury securities has increased over the decades, and interest rate dynamics (especially the slope of the yield curve, as Figure 13.2 shows) have played a role. But the increase has not been dramatic: At the end of 2021Q2, the average maturity was just 5.7 years (versus 3.7 years in 1981) despite the 30-year Treasury yield hovering around a very low 2% at that time (versus 12% in 1981). Research by members of the Treasury Borrowing Advisory Committee (TBAC, see Belton et  al., 2018) shows that issuing at intermediate maturities can helpfully smooth variations in interest rates and budget deficits. Indeed, Treasury notes constitute the bulk of securities outstanding—almost 56% as of 2021Q2. Bills are a distant second, and TIPS constitute the smallest portion of issuance. So, in practice, the Treasury considers more information than just sheer, near-term cost to choose the maturity structure of its debt.

5 Also see Garbade (2007) for a discussion of the evolution of Treasury debt management strategy. 6 The Treasury has often studied extending the maturity range of bonds but has always encountered resistance from market participants and other experts. For example, in response to Treasury’s stated intention to explore issuing a fifty-year bond in 2017, the Treasury Borrowing Advisory Committee concluded that it does not “currently see evidence of notably strong or sustainable demand for ultra-longs in the US market.” See the presentation on the subject at the end of Office of the Debt Management (2017).

280  Research handbook of financial markets 80% 70%

One-Year Bill

60%

Thirty-Year Bond 57.8% 52.4%

62.3%

41.6%

40%

30.5%

30%

22.7% 18.6%

20%

0%

Ten-Year Note

52.8%

50%

10%

Two-Year Note

19.4% 19.1%

4.9%

0.7% 0.4% 0.1% 0.0% Noncompetitive

16.6%

Primary Dealers

Direct Bidders

Indirect Bidders

Source:   U.S. Treasury and authors’ calculations.

Figure 13.3  Average allocation by bidder at 2021 H1 Treasury auctions 13.1.4 Primary versus Secondary Market The Treasury issues new securities at auction, as discussed in the next section. The auction process is also defined as the “primary market.” All Treasury auctions are open to the public; bidders are divided into “competitive” and “noncompetitive” categories. Competitive bidders are typically institutions that directly participate in the auctions and determine their results. Noncompetitive bidders do not participate in the auction and instead accept the price determined by direct bidders.7 Competitive bidders are in turn subdivided into three subcategories: Primary dealers, direct bidders, and indirect bidders. Primary dealers are selected large financial institutions that are expected to bid at all Treasury auctions at “reasonably competitive” prices; they are also trading counterparties of the Federal Reserve Bank of New York and are expected to maintain an orderly secondary market.8 Institutions are interested in becoming primary dealers because of informational and reputational advantages that could boost business volumes not just in Treasuries but in other credit markets as well.9 Direct bidders are institutions other than primary dealers that submit bidders for their own accounts. Indirect bidders submit bids via a direct bidder; they include foreign and international monetary authorities bidding through the Federal Reserve Bank of New York. Figure 13.3 shows average allotments to the various types of bidders for auctions of oneyear bills, two- and ten-year notes, and 30-year bonds in the first half of 2021; these are representative of typical allotments. Noncompetitive bidders constitute a minuscule fraction of total auction award regardless of type of security. Primary dealers tend to constitute a larger

7 Notably, when the Federal Reserve rolls over (that is, reinvests) maturing Treasury securities, it submits noncompetitive bids that are treated as add-ons to the announced auction sizes. 8 A list of current primary dealers as well as expectations, requirements, and other information can be found on the New York Fed website. 9 For a formal analysis of primary dealers’ informational advantage, Hortaçsu and Kastl (2012).

The Treasury and when-issued markets  281

Source:   Bloomberg and authors’ calculations.

Figure 13.4   Frequency of large price moves in the seven- to ten-year sector and the size of the Treasury market portion for shorter maturity securities, while the opposite is true for indirect bidders.10 Direct bidders account for a relatively steady proportion of allotments across notes and bonds and tend to take down a small fraction of bills. After a security is issued, it can be traded in the open market; this trading of existing securities is referred to as the “secondary market.” Such market is an “over-the-counter” market, meaning that market participants trade on a bilateral basis rather than on an organized exchange. Bids and offers can be submitted by any investor, but primary dealers again have an important role as they are expected to make a market for all Treasury securities. The secondary market is generally very orderly, in the sense that outsized moves are rare outside of periods when unexpected news hits. Demand and supply are typically well matched, and primary dealers meet the expectations that come with their role. As Figure 13.4 shows, the frequency of outsized daily price moves (defined as moves larger than three standard deviations of daily price changes) in the seven- to ten-year sector of the Treasury market has not increased over time even though the market has grown exponentially. Still, two points are worth noting. First, during times of crisis large moves in Treasury prices still occur. For example, during the financial crisis year of 2008, seven- to ten-year Treasury price changes moved (in one direction or the other) by more than three times their pre-crisis standard deviations on a total of 43 trading days, or 18% of the time. That is not a small number. Of course, markets should respond fully to unexpected information, but the data indicate that primary dealers have trouble maintaining orderly markets when shocks hit. Second, the relative low frequency of large moves in recent years masks the increased involvement of the Federal Reserve in the secondary market. Since 2009, Treasuries held on the Fed’s balance sheet have increased a lot, as is well known. At times, the amount of Fed purchases has been astounding. In little more than a month between mid-March and late-April 10 See Duffie (2010), Lou et al. (2013), Boyarchenko et al. (2021), as well as Hortaçsu et al. (2018), for more detailed analysis of pricing dynamics before and after Treasury auctions, including the role of primary dealers. Also, see Adrian et al. (2014) for an asset-pricing model that incorporates the effects of primary dealer leverage. The changing role of primary dealers is discussed in further detail below.

282  Research handbook of financial markets

Source:   SIFMA and authors’ calculations.

Figure 13.5  2020 average daily trading volume, by market 2020, at the onset of the Covid pandemic, the Fed bought $1.5 trillion of Treasuries to “smooth market functioning.”11 In addition, the Fed has intervened aggressively in the repo market in recent years, and indeed in July 2021 the Fed instituted a standing repo facility, thus removing any doubt that it will intervene in the repo market in the future. As discussed elsewhere in this volume, the repo market is crucial to the functioning of the Treasury market. By being willing to finance primary dealers’ inventory of Treasury and other securities, the Fed effectively supports the functioning of the secondary market. In recent years, such support has helped maintain an orderly Treasury market. 13.1.5 Measures of Market Liquidity The Treasury market has a reputation for being one of, if not the, most liquid markets in the world. Although liquidity is a slippery concept that means different things to different people, we define liquidity in terms of market depth and ease of trading even large amounts. There are several measures of Treasury market liquidity, but we will limit our discussion to three in this section and analyze a fourth in more detail below. The first is the daily volume of traded securities. Figure 13.5 shows that, as of 2020, the Treasury market had by far the largest average daily volume of any other major fixed-income U.S. market, even greater than the equity market. Figure 13.6 shows disaggregated data by type of security. Bills are the most frequently traded as a percentage of outstanding issues, followed by notes and bonds. TIPS and floating rate notes see very little trading by comparison. Second, perhaps a better liquidity measure is the bid-ask spread, or the gap between the lowest price at which a security is offered for sale and the highest price that buyers are willing to pay. Essentially, the more buyers and sellers there are in the market, the more liquid it is, and 11 For analysis of concerns about the safe haven status of U.S. Treasuries during the pandemic episode, as well as the role of primary dealer balance sheets, see He et al. (2022).

The Treasury and when-issued markets  283

Source:   SIFMA and authors’ calculations.

Figure 13.6   Daily trading volume as a percentage of outstanding

Source:   Bloomberg and authors’ calculations.

Figure 13.7   Bid/ask spreads for ten-year nominal and TIPS securities the lower we should expect the bid-ask spread to be. Figure 13.7 illustrates the bid-ask spread in recent years in the interdealer market for the most-recently-issued ten-year Treasury note (often considered to be the most liquid security) and the ten-year TIPS note. The spread for the nominal security is typically very small—about a quarter of a basis point or less, whereas the spread for the TIPS security is significantly larger, indicating lower liquidity. Both spreads spiked at the onset of Covid, when the market was temporarily gummed up, as is typical in crises. But the nominal market recovered much more quickly—again, a sign of better overall liquidity. A third measure of liquidity is the spread between off- and on-the-run securities. The Treasury issues new securities on a regular schedule (quarterly for notes and bonds, more frequently for bills), and the most recently issued security of any given maturity is referred to as “on-the-run.” When a new security is issued, the previous on-the-run becomes the first “off-the-run” issue (and older securities are similarly downgraded).

284  Research handbook of financial markets 0.1

NOMINAL 10-YEAR ON-THE-RUN PREMIUM*

%

0.05

0

-0.05

-0.1

Apr-2017 Nov-2017 May-2018 Dec-2018 Jun-2019

Jan-2020

Jul-2020

Feb-2021 Sep-2021

*(03-Jan-2017-21-Oct-2021). Source:   CRSP and authors’ calculations.

Figure 13.8   Nominal ten-year on-the-run premium Typically, on-the-run securities are the most sought after and frequently traded. Activity tends to decrease the farther off-the-run a security is. As a result, on-the-run securities normally trade at a premium to off-the-run securities, and their yield is typically lower. The spread between off- and on-the-run securities is another measure of liquidity of the overall market. Figure 13.8 shows that this spread is typically positive, as expected, and that it spiked at the onset of Covid, just like the bid-ask spread. This indicates a strong preference for the most liquid securities at times of stress, and corresponding illiquidity among off-the-run issues.

13.2 THE AUCTION PROCESS Treasury securities are sold to the public via auctions. This was not always the case: In the early days, the Treasury sold its securities via underwriters who would purchase the totality of offered securities and resell them to the public. The first Treasury auction took place in 1929, for the sale of a three-month bill. Notes and bonds were not sold at auction until the 1970s. 13.2.1 The Mechanics of Treasury Auctions Among the various types of auctions, the Treasury sells its securities via a single-price, descending-bid auction. In a single-price auction, all winning bidders are awarded securities at the same price, even if it differs from their bid. In a descending-bid auction, all bids are ranked from highest to lowest in terms of price and accepted in that order until the entire amount of the offered security is sold. Because Treasury auctions (depending on the type of security being auctioned) are usually quoted in terms of yield or rate or discount margin,

The Treasury and when-issued markets  285

which are inversely related to price, bids are ranked from the lowest to the highest yield, rate, or discount margin. From here on, we will follow convention and talk in terms of yield.12 The auction process begins with an announcement of the auction date.13 The announcement specifies the term and type of security offered (e.g., ten-year note), the amount offered, the auction date and time, the issuance date (which will be after the auction date), the maturity date, the terms and conditions of the offer, and other relevant information. Auctions can be for securities issued for the first time or for additional amounts of previously issued securities; in the latter case, the auction is referred to as a “reopening.” The Treasury does not specify the coupon for originally-issued notes and bonds on the announcement. Rather, the coupon is determined as part of the auction process. Specifically, the coupon is set at the highest level (in increments of one-eighth of a percentage point) that does not result in a price for the security greater than 100% of principal.14 In practice, this ensures that securities are issued at close to their par value on issuance day. Par issuance, in turn, helps predictability and maintains a balance between Treasury’s need for cash and investors’ appetite for lending. Other than for cash-management bills, which cover hard-to-predict, short-term cash needs, the Treasury conducts auction on a predictable schedule. Bills with maturities up to 26 weeks are auctioned every week, while bills with a maturity of one year are offered every four weeks. Treasury notes with two, three, five, and seven years maturities are auctioned monthly. Tenyear notes and all bonds are auctioned quarterly in February, May, August, and November, with reopenings in all other months. TIPS are auctioned either semiannually (in April and October for five-year securities and January and July for ten-year securities) or annually (in February, for 30-year securities); reopenings also follow a regular schedule. Floating rate notes are auctioned quarterly in January, April, July, and October, with reopenings in all other months.15 As said in Section 13.1, bids at auctions can be competitive or noncompetitive. Individual noncompetitive bidders (usually small investors) are limited to a maximum of $5 million bids and are guaranteed to be allocated the securities they bid for. Each individual competitive bidder (primary dealers and other large financial institutions) is limited to a maximum of 35% of securities offered. This ensures competitiveness in the secondary market, given that competitive bidders typically resell the securities they are awarded at auction in the secondary market. Because noncompetitive bidders always receive the amount they bid for, the amount of noncompetitive bids is subtracted from the offered amount. What remains is awarded to competitive bidders according to the following process. First, the competitive bids are ordered from lowest to highest in terms of yield, rate, or discount margin. The lowest bid is accepted, then the second lowest, and so on until the remaining offered amount has been exhausted. If the last viable bid exceeds the remaining offered amount, that bidder is awarded the remaining amount, which will be smaller than the bid.16 The highest rate of all accepted bids is referred 12 Until the 1990s, the Treasury used multiple-price auctions to sell the securities that were auctioned. In multiple-price auctions, winners are awarded securities at the price they bid rather than the price equivalent to highest winning yield. See Malvey et al. (1995). 13 More details and examples of how auctions work can be found at: www​.newyorkfed​.org​/aboutthefed​/fedpoint​/fed41​.html. 14 More details can be found in Grabade and Ingber (2005). 15 For a full description of the timing of Treasury auctions see www​.treasurydirect​.gov​/instit​/auctfund​/work​/auctime​/auctime​.htm. 16 If multiple bidders submit bids at the same yield, rate, or discount margin, they are awarded the remaining amount in proportion to their original bids.

286  Research handbook of financial markets

to as the “stop out rate,” and all winning bidders are awarded securities at that rate, even if they submitted bids at a lower rate. Reopening auctions follow the same process. The securities being sold at reopening auctions have the same maturity and coupon as the securities in the original auction, but a different issue date. Of course, because they are sold at a separate auction, their price and yield will usually be different from the price and yield at which the original securities were sold. 13.2.2 The Diminishing Role of Primary Dealers As said in the previous section, primary dealers are expected to bid at auctions at “reasonably competitive” prices to facilitate an orderly secondary market. As illustrated in the top left panel of Figure 13.9, the amount of primary dealer bids submitted at auction has declined over time but still accounts for most bids submitted across securities. The amount accepted for notes and bonds, however, has clearly declined since the financial crisis of 2008; the amount of accepted bills dropped a few years later. The percentage of securities awarded to primary dealers at ten-year note auctions, for example, went from an average of almost 80% between 2005 and 2007 to just around 20% in the first half of 2021. The decline in securities awarded to primary dealers has been offset by an increase in the take down from other direct bidders and indirect bidders. Before the financial crisis, other direct bidders used to take home about 1.6% of ten-year auctions, on average; by the first half of 2021, that amount had increased tenfold, to more than 16%. Over the same period, indirect bidder allotments increased from 21% to 62% of ten-year securities offered, on average, even though the amount of bids submitted by both other direct bidders and indirect bidders has increased much less. There are several potential explanations for these staggering changes. One could be that primary dealers’ bids have become less competitive, perhaps because of increased risk aversion after the financial crisis, balance sheet pressures induced by regulatory changes, the significant growth of the Treasury market, or other factors. Alternatively, following a change in Treasury rules in 2009, certain dealer purchases on behalf of customers were no longer attributed to the dealer and instead were counted as customers (i.e., indirect) awards.17 A third possibility could be that other direct and indirect bidders have exogenously increased their participation in auctions rather than relying on dealer intermediation. Fleming and Myers (2013) shed some light on the issue by looking at auction prices relative to secondary market prices for comparable securities and whether new auction supply pushes auction prices lower. They show that securities sold at auctions tend to have slightly higher prices (lower yields) than comparable securities sold in the secondary market after 2008. They also find that the price of most recently issued securities tend to react less strongly to the issuance of new securities after the financial crisis. They infer that diminished primary dealer awards at auction are the result of other bidders’ preferences for participating in auctions, rather than reflecting a decline in the quality of primary dealers’ bids that trace to risk aversion or other reasons. They also conclude that the 2009 rule change may have played an ancillary role. We showed in the previous section that outsized secondary market moves have not increased over the years. We cannot therefore say that the diminished primary dealer inventory of securities 17 The change of such rules is outlined in www​.govinfo​.gov​/content​/pkg​/ FR​-2009​- 06​- 01​/ html​/ E9​12787​.htm.

The Treasury and when-issued markets  287

Source:   Bloomberg and authors’ calculations.

Figure 13.9   Treasury auction statistics

288  Research handbook of financial markets

that results from lower auction awards compromises the functioning of the secondary Treasury market. But whether the same could be said in the absence of the greater involvement of the Federal Reserve in the market (also discussed in the previous section) remains an open question. 13.2.3 Measures of Auction Success An auction is considered successful if all the amount offered is sold without unduly moving the market, i.e., without the resulting yield exceeding the rate that prevailed in the secondary or the when-issued market (see the next section) prior to the auction. Yet, success is not always black-and-white: Some auctions can be less successful than others, either because demand is scarce or because the stop-out yield exceeds pre-auction quotes. One readily available measure of auction demand is the “bid-to-cover ratio.” This is simply the amount of bids divided by the offered amount; the larger the bid-to-cover ratio, the larger the demand for the securities, and generally the more successful the auction. In the first half of 2021, bid-to-cover ratios hovered around 2.5 for two-, ten-, and 30-year maturities, which implies that demand was about 2.5 times supply. The problem, however, is that bid-to-cover ratios say nothing about the price or yield associated with the demand. In other words, auctions with high bid-to-cover ratios can still result in stop-out rates above then-prevailing market rates. A better measure of auction success is the difference between the stop-out rate and the price of the same security immediately before issuance. To understand this, note that Treasury securities start trading immediately after the announcement of an auction, even before they are issued. This happens in the “when-issued” market, which is discussed in detail in the next section. If the difference between the auction stop-out yield and the when-issued yield immediately before bidding ends is positive, the auctions is said to have “tailed,” whereas if the difference is negative, the auction is said to have “traded through” when-issued pricing. Significant tailing is an indication of a poor auction, whereas substantial trading through is a sign that the auction was well received.

13.3 THE WHEN-ISSUED MARKET It is not uncommon for auctions with high bid-to-cover ratios to tail significantly; when that happens, the auction is poor despite the high bid-to-cover ratio. In this case, the appetite for the security at current market prices is modest, and most bids likely exceed market yields significantly. This case illustrates well why tail/trade-through metrics better capture auction success. As an example, the August 2021 five-year, fixed-rate note auction was announced on August 19, 2021 and concluded on August 25, 2021 at 1pm ET. The security was issued on August 31, 2021. Between August 19 and August 31 (i.e., between the announcement and issuance) the security traded in the so-called when-issued (WI) market. Just before the end of the auction on August 25, the security had a yield in the when-issued market of 0.822%. The auction settled with a stop-out rate of 0.831%. Therefore, that auction tailed by almost a basis point—not a glaring failure, but not a resounding success either. The when-issued market refers to trading of the security between the announcement date of the security and its issuance. As Fleming, Shachar, and Van Tassel (2020) (FSV) note, WI trading promotes price discovery into the auction and lowers government borrowing costs, on average. In practice, the WI yield comprises a benchmark for comparison with the yield at auction, which market observers use to gauge demand. Also, FSV examine Trade Reporting

The Treasury and when-issued markets  289

and Compliance Engine (TRACE) daily data from FINRA from July 1, 2019, through June 30, 2020, and find that most WI trading occurs in nominal coupons and bills, as opposed to TIPS and nominal FRNs. They also report that trading in nominal coupons, the focus of this chapter, increases after the auction date and nearing issuance. To supplement this research on volumes after the auction date, we examine the pricing (i.e., yields) of WI securities in absolute terms as well as compared to other outstanding issues during the period between auction and issuance going back to the beginning of 1987. To motivate our analysis, academics and practitioners have long observed that more recent Treasury issues, at most times of the auction cycle, tend to trade at higher prices and lower yields relative to issues with close to the same maturity. This premium likely owes to financing advantages (for repurchase agreement transactions and other uses) that reflect the greater liquidity of recent issues, which are yet to be tucked away into buy-and-hold portfolios.18 As Fontaine and Garcia (2012) detail, an increasing amount of the par amount outstanding of a given nominal securities becomes less available for trading through its life. Again, the size of the so-called onthe-run premium for the most recently issued securities is measured relative to a fitted yield curve—or strictly by the pricing-error of the individual security, namely the different between the observed yield and its discount-function-fitted value—on any given day. Our exploratory analysis focuses on these pricing errors for issues specifically during the later phase of the WI trading cycle, again after the auction but before issuance. We calculate these errors or premiums in the following steps. We first fit the yield curve on each day of the sample using all outstanding issued non-callable coupon notes and bonds, using the so-called Nelson-Siegel-Svensson (NSS) methodology.19 Then, given the estimated discount function, we then price any WI securities, given their specified coupon and principal payments. The underlying pricing data come from the daily Center for Research in Security Prices (CRSP) Treasury file, spanning January 3, 1987, through December 30, 2020, which implies 2195 trading days with WI quotes. In turn, we identify WI securities in the CRSP data as those notes and bonds with quoted bid and asked prices at times prior to their reported issue date,20 which also corresponds to unavailability of amounts outstanding in the files. Formally, the (unconditional) WI premium follows,

( )

WI WI WI pi(,t ) = yi(,t ) - yi( ) bt (13.1)

18 This phenomenon is more pronounced in TIPS, as D’Amico, Kim, and Wei (2018) note, for example. But the impact is hardly negligible for nominal coupon securities. In fact, for the purposes of fitting a representative yield curve for monetary policy analysis, GSW omit the on-the-run and first- and second-off-the-run securities before estimating the discount function across all remaining securities. 19 The Svensson (1994) model follows, using the notation in HPW,



æ mö æ mö æ mö m m f ( m, b ) = b0 + b1 exp ç - ÷ + b2 exp ç - ÷ + b3 exp ç - ÷ t2 t1 è t2 ø è t1 ø è t1 ø

where m is the time to maturity and b = ( b0 , b1, b2 , b3 , t1, t2 ) are estimated model parameters. The securities we use to fit the curve using this methodology differ from Gürkaynak, Sack, and Wright 2006 (GSW), who omit the on- as well as the first- and second-off-the-run issues for each tenor. 20 As such, and to be clear, we do not consider the earlier subperiods within the WI cycle that FSV address, namely the period from the announcement to auction. However, in the following analysis of WI premiums we do estimate the effects of the number of days to issue.

290  Research handbook of financial markets

( )

WI WI where yi(,t ) is the quoted yield on the ith WI security, and yi( ) bˆt is the corresponding fitted or predicted value based on the estimated curve at the end of day t given the estimated ˆ discount factor, b. Before presenting these estimates, as an important aside, the literature has increasingly used such pricing errors across all securities, not just on recently issued notes and bonds, as measures of overall Treasury market liquidity, in addition to volumes, bid-asked spreads, and on-the-run premiums discussed in the previous section. For example, as Hu, Pan, and Wang (2013) (HPW) originally argue, large average misses might capture times when the limits to arbitrage bind, insofar as investors cannot secure funding to exploit mis-pricings. Largely folU lowing HPW, the unconditional so-called “noise” measure, Noiset( ) , is the root mean-squared pricing error (RMSE) across all securities from a fitted term structure, as in

U Noiset( ) =



1 Nt

Nt

å éêë y

i ,t

i =1

( )

2

- yi bˆt ùú (13.2) û

where yi,t is the observed yield at time t on the ith of N UST securities, and again as in (13.1), yi bˆt is the corresponding fitted or predicted value based on the estimated curve at the end of ˆ 21 In short, a plausible prior we explore is whether day t given the estimated discount factor, b. U the WI premium is greater, i.e., increasingly negative following (13.1), the greater Noiset( ) , all else equal. Otherwise, the WI premium may contain additional information about Treasury (and by extension broader) market conditions, which related to more idiosyncratic preferences among market participants, perhaps with so-called preferred habitats.22

( )

13.3.1 Empirical Results The charts in Figure 13.10 show estimates following this procedure for selected dates during the 34-year sample. The solid lines show the zero-coupon curve on each day, and the open and shaded small circles show fitted and actual yields—i.e., yi,t and yi ( bt ) , respectively— for each CUSIP used to fit the term structure. The open and shaded large circles indicate WI WI the fitted and actual yields on WI securities traded on those days, i.e., yi(,t ) and yi( ) ( bt ) , respectively. For example, as the top left panel indicates, on May 8, 1987, the ten-year and the 30-year WI securities traded below their fitted values, consistent with a premium, whereas the observed and fitted yields on the two-year WI note were nearly equivalent. Naturally, visual inspection of these randomly chosen dates implies that premiums change over time. For example, the right panels indicate that on August 10, 1990, as well as on February 14, 2020, each of the three WI securities commanded a premium. However, on February 13, 2009, shown in the lower left panel, clearly the ten-year note WI traded well below the fitted term structure, while the WI note and bond at the ends of the curve traded cheap relative to the estimated discount function.​

21 Studies that use curve fitting errors to capture TIPS illiquidity premiums, for example, including Abraham et al. (2016), Grishchenko and Huang (2013), and Durham (2021), who also additionally considers pricing error correction speeds as a gauge of market liquidity. 22 See Vayanos and Villa (2021) on a preferred habitat model of the term structure that directly accounts for heterogeneous idiosyncratic preferences and so-called localized investor demand.

The Treasury and when-issued markets  291 U.S. Nominal Yield Curve: (May-08-1987)

9

Yield (%)

8

7 Zero-Coupon Curve WI Yield WI Fitted Yield

6

5

0

5

10 15 20 25 Years to Maturity. Number of bonds=158.

30

35

U.S. Nominal Yield Curve: (Aug-10-1990)

9

Yield (%)

8.5

8

7.5

Zero-Coupon Curve WI Yield WI Fitted Yield 0

5

10

15

20

25

30

35

Years to Maturity. Number of bonds=172.

Source:   CRSP and authors’ calculations.

Figure 13.10  Fitted yield curves and when-issued yields. The solid lines in these charts show the nominal zero-coupon U.S. Treasury yield curve, based on the NSS fitting methodology. The small black dots and the open small circles indicate the observed and fitted yields, respectively, for those securities used to fit the term structure. The large solid circles are the observed yields on the WI securities, and the large open circles show the corresponding fitted values, based on the estimated discount function To aggregate this information more efficiently for broader inferences, we examine the time series of estimates. The black line in the chart in the top left of Figure 13.11 shows the crosssectional average premium across all WI securities and tenors each trading, which we denote WI U as pt( ) following (13.1).23 For comparison, the solid line shows our estimate of Noiset( ) WI across all CUSIPs. A couple preliminary observations are noteworthy. First, although pt( ) 23 For example, pWI ,t on any given date for the curves shown in Figure 13.10 is the arithmetic average pricing error across all three WI quotes.

292  Research handbook of financial markets U.S. Nominal Yield Curve: (Feb-13-2009)

5

Yield (%)

4 3 2 Zero-Coupon Curve WI Yield WI Fitted Yield

1 0

0

5

10 15 20 25 Years to Maturity. Number of bonds=160.

30

35

U.S. Nominal Yield Curve: (Feb-14-2020)

2.2

Yield (%)

2 1.8 1.6 Zero-Coupon Curve WI Yield WI Fitted Yield

1.4 1.2

0

5

10

15

20

25

30

35

Years to Maturity. Number of bonds=298.

Figure 13.10  (Continued) appears volatile by visual inspection, nonetheless the unconditional full-sample average for WI securities across all maturities is negative, at around 2.44 basis points. Also, contrary to the notion that any WI premium connotes broader market stress, the simple correlation WI U coefficient between pt( ) and Noiset( ) is small yet ultimately positive, which strictly speaking implies a lower WI premium when wider market liquidity conditions worsen, according to this metric of arbitrage constraints.24​ As an alternative approach, we also consider a conditional as opposed to an unconditional measure of the WI premium. For each trading day of the sample and given estimated pricing errors, we run the following cross-sectional regression to ascertain an arguably more precise gauge that nets out common proxies for liquidity and collateral value across other securities, following,

U WI ei(,t ) = jˆ 0 + jˆ t ’X t + bˆ (t ) IWI + ei,t (13.3)

24 Still, both pWI ,t and Noiset(U ) of course could capture different aspects of Treasury market liquidity.

The Treasury and when-issued markets  293 30

Daily Average When-Issue and

Aggregate Curve

Fitting Errors

(NSS)

20

Basis Points

10 0 -10 -20 WI Average Error Noise RMSE All Securities

-30 -40 Sep-1987

Mar-1993

Sep-1998

Feb-2004

Aug-2009

Feb-2015

Jul-2020

Sample average (of averages) =-2.44 basis points. WI sample averages correlation with noise (RMSE)=0.118 Fitting errors based on (NSS) estimation (02-Jan-1987-30-Dec-2020). 50

When-Issue Coefficients: Conditional CUSIP Fitting Error Regressions

(NSS)

40 30

Basis Points

20 10 0 -10 -20 -30 -40 Sep-1987

Mar-1993

Sep-1998

Feb-2004

Aug-2009

Feb-2015

Jul-2020

Sample average =-1.89 (basis points). Regressions include coupon rates and bid-ask spreads as well as dummy variable for WI securities. Fitting errors based on (NSS) estimation. Correlation with noise/RMSE =0.364. (02-Jan-1987-30-Dec-2020). Source:   CRSP and authors’ calculations.

Figure 13.11  Estimated when-issued premiums. The dark line in the top left chart shows the sample time-series of the cross-sectional unconditional average WI WI U premium, pt( ) , and the light line shows the RMSE of Noiset( ) across all nominal U.S. Treasuries. The top right chart shows the time series of WI the conditional WI cross-section premium, bˆ (t ), based on a regression of (U ) i ,t = yi ,t - yi ( bt )on control variables and a dummy variable for whether the individual issue is a WI security, and the lower left chart shows the distriWI bution of bˆ (t ). The lower right chart shows the time-series of unconditional WI WI premium estimates, pi(,t ) , denoting the two- and ten-year WI securities

294  Research handbook of financial markets Nominal U.S. Treasury When-Issue* Coefficients: Conditional CUSIP Fitting Error (NSS) Regressions 350 300 250 200 150 100 50 0

-30

-20

-10

0

10

20

30

40

50

Basis points. Mean =-1.9 basis points. Standard deviation =5.5. Skewness=2.1. Bowley Skew=-0.14. Nominal U.S. Treasury When-Issue Curve Fitting Errors

30

(NSS)

20 10

Basis Points

0 -10 -20 -30 -40 -50

WI All Maturities WI 2-Year Maturity WI 10-Year Maturity

-60 Sep-1987

Mar-1993

Sep-1998

Feb-2004

Aug-2009

Feb-2015

Jul-2020

Sample CUSIP-level average (10-year) =-2.3 (-8.4) basis points. (02-Jan-1987-30-Dec-2020)

Figure 13.11  (Continued)

( )

U where ei(,t ) = yi,t - yi bˆt is the unconditional fitting error of security i based on a given fitting

methodology (in this case NSS); X is a vector of proxies for relevant factors—including the bid-asked spread and coupon size—two variables that may help explain why some specific CUSIPs trade off the fitted yield curve; jare the factor loadings on these control variables; and ei,t is the error term, or the conditional pricing error of each CUSIP.25 But for this analysis,

(C ) 25 Durham (2023) uses ei,t as an alternative conditional measure of “noise”, i.e., Noiset , among other measures to capture the TIPS liquidity premium.

The Treasury and when-issued markets  295

the variable of interest is IWI , a simple dummy variable for whether the security is WI, whatWI ever its tenor, and bˆ (t ) captures the so-called conditional WI premium. WI The chart in the top right of Figure 13.11 shows the path of bˆ (t ) over the sample. True, the pattern resembles that of the unconditional measure. But after netting out other factors that might help account for pricing errors, the average measure over the sample declines to about 1.89 basis points. Also, as the histogram in the lower left of Figure 13.11 suggests, the distribution of conditional time-series estimates is wide, with a standard deviation of about 5.5 basis points, which indeed may increase doubts about a consistent general premium across all WI securities. Measures of skew are somewhat mixed. Standard skew based on the third moment is positive, at 2.1, consistent with positive extreme outliers, whereas Bowley skew relative to the interquartile range by contrast is about –0.14, skewed toward the left or higher premiums WI in the middle section of the distribution. Finally, as with the unconditional measure, pt( ) , we (WI ) (U ) ˆ report an arguably perverse relation between bt and Noiset , insofar as the correlation is positive, at about 0.364, which is greater than the unconditional measure (again, 0.118). WI WI Although we cannot link WI premiums, either pt( ) or bˆ (t ), to a related aggregate measure of market risks, we also examine whether some underlying security characteristics affect WI premiums, whatever market frictions they might capture. For instance, conceivably some WI tenors command greater premiums than others, depending on market segmentation. To wit, the chart in the lower right of Figure 13.11 shows each unconditional WI premium estimate, WI pi(,t ) , for all observations. By visual inspection, we glean the same general common pattern as in the top left. However, the light grey and dark grey circles isolate two- and ten-year WI securities, respectively. Notably, the unconditional averages are –2.3 and –8.4, for the two tenors, respectively, which of course at first unconditional blush implies comparatively more special demand in the ten-year sector after auction and prior to issuance. Even so, other factors might also affect premiums during this latter phase of the WI cycle, and we examine whether there are robust pattens between auction and issuance. As such, we WI run the following simple panel regressions for all pi(,t ) , following

I WI DTI pi(,t ) = jˆ 0 + jˆ t ’X t + bˆ ( ) DTI + bˆ ( n- year ) ’I n - year + ei,t (13.4)

where X includes control variables including the bid-asked spread, the coupon rate, and U Noiset( ) (which in effect may capture relevant time trends); DTI is the number of days until issuance for the WI security; andI n - year is a T ´ ( n - 1)matrix of dummy variables where n is the length of the vector of tenors, éë2 3 5 7 10 20 30 ùû ’, with the indicator variables defined with the 30-year maturity as the omitted condition. Table 13.1 lists the results. Considering the overall fit, the R-squared of the model is about 0.23, and the intuition among the regressors is somewhat mixed. Among the control variables, at the CUSIP-level (and again among only WI notes and bonds in this exercise) we recapture WI U the positive relation between pt( ) and Noiset( ) , with about a 0.29 pass-through increase from a unit increase in the contemporaneous RMSE across all CUSIPs, which is safely statistically significant. Crudely speaking, this result implies that the WI premium falls when wider limits to arbitrage increase. Also, although the estimate for the bid-asked spreads is perhaps surprisingly not robust, the regression implies somewhat higher WI premiums for issues with greater coupon rates. Turning to other security characteristics, DTI does not seem to have a significant effect on the premium over the sample—the miniscule coefficient that implies some decrease toward

296  Research handbook of financial markets

Table 13.1  When-issue pricing error (NSS) regressions CUSIP-level (January 2, 1987– December 30, 2020) Observations = 3472 Adjusted-R˄ 2=0.231 Coefficient

p-value

–0.0072

0.0644

0.2944

0.0000

Coupon

–0.0063

0.0000

Bid-Asked Spread

–0.1778

0.2602

Days to Issue (DTI)

0.0002

0.5221

Two-Year Dummy

0.0105

0.0147

Intercept Unconditional Noise

Three-Year Dummy

0.0007

0.8695

–0.0007

0.8531

0.0141

0.0018

Ten-Year Dummy

–0.0547

0.0000

20-Year Dummy

–0.0066

0.5485

Five-Year Dummy Seven-Year Dummy

Notes:   The table reports regression coefficients and p values. The dependent variable is the WI premium, pi(,t ) , and the regressors include underlying characteristics of WI securities. The sample includes 3,472 CUSIP-days from January 2, 1987, through December 30, 2020. WI

issuance dates is safely insignificant. However, consistent with the chart in the lower right panel of Figure 13.11, the results suggest that some differences across tenors over this lengthy sample are safely robust. WI premiums on two- and seven-year securities are somewhat lower, with pricing errors about 1 to 1.5 basis points greater compared to the 30-year. By contrast, on average and all else equal the WI premiums for ten-year notes are around 5.5 basis points greater, but no other tenor (three, five, and 20 years) produces statistically significant results.26

13.4 SECONDARY MARKET PRICING AND INTERPRETATION OF THE YIELD CURVE After the auction process, again Treasury securities trade in the secondary market. Although as previous sections suggest, there is ample evidence of time-varying nominal liquidity premiums, this market is one of the deepest and most liquid. Also, given the credit-default-risk free status of U.S. Treasury obligations—arguably reinforced by the reserve currency status of the U.S. dollar—both investors and central banks glean critical information from secondary market pricing about underlying borrowing costs in the U.S. economy, the largest in the

26 Please see an online appendix for alternative estimates of the results reported in Figure 11 and Table 1 using an alternative curve-fitting methodology, namely a variance roughness penalty instead of a NSS specification.

The Treasury and when-issued markets  297

world, for horizons up to the longest-term issued obligation to date, thirty years for nominal securities and TIPS alike. This schedule of borrowing costs by maturity is of course known as the yield curve or term structure of interest rates. On average over time, the shape of the yield curve has hardly been flat, as instead borrowing costs tend to differ meaningfully by maturity. The average slope of the nominal term structure based on average values across maturity points from three months to ten years, has been safely positive from the start of 1987 through October 20, 2021. The simplest model of the term structure—namely, to account for why the yield curve takes varying shapes—is the so-called expectations hypothesis (EH), which posits that long-term yields comprise the sum of expected future short-term rates over the tenor of the given security. The EH in its purest form implies that the yield curve slopes up (down) when investors expect the central bank to tighten (ease) monetary policy over a given horizon.27 A clear implication of this view is that realized and expected excess returns on nominal Treasury bonds is zero. But empirical data imply that this framework is very incomplete, and tests of the EH comprise a large academic literature, much of which includes regression analysis of realized excess returns to assess ex post whether required excess Treasury returns are zero.28 But if the EH fails, what instead accounts for observed average positive slope in the both the nominal and TIPS curves over recent decades? Instead of policy expectations, the slope of the curve could also reflect in part non-zero, time-varying required excess returns on Treasury securities, otherwise known as term premiums, defined as the spread between observed yields and the average anticipated short rate, again over the horizon of a given note or bond. Setting aside the ultimate source of non-zero premiums, decomposing yields along the curve is critical for both central bankers as well as investors. For the former, insofar as the slope, or an increase in the spread between longer- and near-dated rates, connotes expectations of some combination of higher inflation or stronger economic growth, the central bank would be disposed to tighten monetary policy in response, all else equal, to prevent an overheating economy. But if instead the same slope traces to higher term premiums on longer-dated notes or bonds, which reflect an increase in investors’ risk aversion amid an unchanged economic outlook, central banks might instead be inclined to ease, to keep the economy from undershooting. The distinction is also crucial for investors, insofar as term premiums, aka expected required excess returns, are naturally a critical input for modern portfolio optimization, which require estimates of anticipated returns over cash as well as variances and covariances among all assets, default risk-free Treasuries among them. Naturally, given the importance of the distinction, the literature on parsing term premiums from observed yields is also vast. A judicious review is beyond the scope of this chapter. But at least two general approaches are noteworthy. First, some researchers use survey data to net term premiums from overall observed yields. Under this metric, the term premium on an n-year bond is simply the difference between the observed yield and the central tendency of survey-based expectations for short rates (e.g., three-month Treasury bills, say) over the corresponding n-year horizon. This formulation is indeed true to concept but nonetheless faces meaningful drawbacks. For example, unlike financial market quotes, survey data are 27 The “weak-form” of the EH suggests that the risk premium is non-zero, and may be maturityspecific, but it is constant. Even so, weaker forms of the EH suggest that any changes in yields reflect changes in expected short rates, only. 28 See Gürkaynak and Wright (2012) for a thorough discussion of early empirical tests of the EH.

298  Research handbook of financial markets

available only at sparse as well as somewhat imprecise intervals. Yet despite some shortcomings, researchers routinely compare their estimates from other methods with survey data to gauge accuracy and intuition. A second approach to decompositions refers to so-called arbitrage-free, affine models of the term structure, which appeared soon after publication of the Black-Scholes-Merton framework for option pricing (e.g., Vasicek, 1977). Other references comprise more fulsome overviews (e.g., Gürkaynak and Wright, 2012; Piazzesi, 2010). But this tack relies on statistical methods rather than surveys of forecasting anticipated short rates. Briefly, the anticipated-rate component under these models ultimately trace to some form of vector autoregression (VAR) forecasts, typically with Gaussian shocks, notably within a framework that precludes the possibility of arbitrage across the full cross-section of yields at any point along the term structure. The short rate in more recent models—as well as all other yields—are affine functions of the underlying factors that drive the term structure,29 which in the literature range from latent variables (e.g., Kim & Orphanides, 2012) to observable factors such as yield curve principal components (e.g., Adrian et al., 2013) or macroeconomic time series (e.g., Ang and Piazzesi, 2003). Given forecasts based on the estimated underlying dynamics of the model factors, as well as the estimated loadings of the short rate on those factors, this approach affords expected short-rate paths over any horizon. The gap between model-implied yields or forwards, which again satisfy the no-arbitrage condition, and these estimates of anticipated short rates comprise model-implied term premiums, for either zero-coupon yields or forward rates.30

13.5 CONCLUSIONS AND FURTHER TOPICS To sum up, naturally this short introductory chapter cannot capture the breadth of open questions related to the functioning and pricing of the world’s foremost credit-risk-free market. In both respects, the persistent secular downward trend in estimate of equilibrium interest rates toward the nominal lower bound comprises a formidable complication. So-called shadow-rate approaches (Krippner, 2012) or quadratic models are an increasingly fruitful line of research that address this problem. Meanwhile, a sensible reminder for practitioners is to provision for as much model uncertainty around term premium estimates as other important yet empirically slippery concepts, such as the equity risk premiums or potential GDP. Of course, the secular decline in neutral rates corresponds with unconventional policy measures that increase the influence of the Federal Reserve’s portfolio on Treasury market functioning and pricing. Whether balance sheet effects transmit through supply or portfolio rebalance channels on the one hand, or signaling mechanisms on the other, remains an open question despite a burgeoning literature. Lessons drawn from the initial foray into so-called quantitative easing (e.g., Gagnon et al., 2011) may not be consistent with subsequent shocks, such as the policy response to the ongoing pandemic and unwind thereof at the time of writing.

29 By contrast, measures of the short rate in early models such as Vasicek (1977) is the single underlying factor that determines the yield curve. 30 Although most models almost ubiquitously adopt the arbitrage-free, affine set-up, including the common assumption of normally distributed disturbances, the literature varies widely across parameter estimation methods, from Kalman-filter-based methods, multi-step maximum likelihood, to linear regression approaches.

The Treasury and when-issued markets  299

REFERENCES Abraham, M., Adrian, T., Crump, R. K., Moench, E., & Rui, Y. (2016). Decomposing real and nominal yield curves. Journal of Monetary Economics, 84, 182–200. Adrian, T., Crump, R. K., & Moench, E. (2013). Pricing the term structure with linear regressions. Journal of Financial Economics, 110(2), 110–138. Adrian, T., Etula, E., & Muir, T. (2014). Financial intermediaries and the cross-section of asset returns. Journal of Finance, 69(6), 2557–2596. Anderson, N., & Sleath, J. (1999). New estimates of the UK nominal and real curves. Bank of England Quarterly Bulletin, Q4, 384–392. Ang, A., & Piazzesi, M. (2003). A no-arbitrage vector autoregression of term structure dynamics with macroeconomic and latent variables. Journal of Monetary Economics, 50(4), 745–787. Belton, T., Li, H., Dawsey, K., Ramaswamy, S., Greenlaw, D., & Sack, B. (2018). Optimizing the maturity structure of U.S. Treasury debt: A model-based framework. Hutchins Center Working Paper 46, Brookings Institution. Birchandani, S., Edsparr, P. L., & Huang, C. F. (2000). The treasury bill auction and the when-issued market: Some evidence. UCLA: Decisions, Operations, and Technology Management Working Paper. Boyarchenko, N., Lucca, D., & Veldkamp, L. (2021). Taking orders and taking notes: Dealer information sharing in treasury auctions. Journal of Political Economy, 129(2), 607–645. Coutinho, P. B. (2014). When-issued market and treasury auctions. Working Paper, UCLA. D’Amico, S., Kim, D. H., & Wei, M. (2018). Tips from TIPS: The informational content of treasury inflation-protected security prices.  Journal of Financial and Quantitative Analysis, 53(1), 395–436. Driessen, G. A. (2016). How treasury issues debt. Congressional Research Service Report R40767. Dudley, W., Roush, J., & Steinberg Ezer, M. (2009). The case for TIPS: An examination of the costs and benefits. Economic Policy Review, Federal Reserve Bank of New York, 15(1), 1–17. Duffie, D. (2010). Presidential address: Asset price dynamics with slow-moving capital. Journal of Finance, 65(4), 1237–1267. Durham, J. B. (2023, forthcoming). What do TIPS say about real interest rates and required returns? Financial Analysts Journal. Fleckenstein, M., Longstaff, F. A., & Lustig, H. (2014). The TIPS-treasury bond puzzle. Journal of Finance, 69(5), 2151–2197. Fleming, M. J., & Myers, S. (2013). Preimary dealers’ waning role in treasury auctions. Liberty Street Economics, February 20, Federal Reserve Bank of New York. Fleming, M., Shachar, O., & Van Tassell, P. (2020). Treasury market when-issued trading activity. Liberty Street Economics, November 30, Federal Reserve Bank of New York. Fontaine, J.-S., & Garcia, R. (2012). Bond liquidity premia. Review of Financial Studies, 25(4), 1207–1254. Gagnon, J. E., Raskin, M., Remache, J., & Sack, B. P. (2011). Large-scale asset purchases by the Federal Reserve: Did they work?,” Economic Policy Review, Federal Reserve Bank of New York, 17, 41–59. Garbade, K. D. (2007). The emergence of ‘regular and predictable’ as a treasury debt management strategy. Economic Policy Review, Federal Reserve Bank of New York, 13(1), 53–71. Garbade, K. D., & Ingber, J. F. (2005). The treasury auction process: Objectives, structure, and recent adaptations. Current Issues in Economics and Finance, 11(2). Federal Reserve Bank of New York. Grishchenko, O. V., & Huang, J. Z. (2013). The inflation risk premium: Evidence from the TIPS market. The Journal of Fixed Income, 22(4), 5–30. Gürkaynak, R. S., Sack, B. P., & Wright, J. (2007). The U.S. Treasury yield curve: 1961 to the present. Journal of Monetary Economics, 54(8), 2291–2304. Gürkaynak, R. S., & Wright, J. (2012). Macroeconomics and the term structure. Journal of Economic Literature, 50(June), 331–367. He, Z., Nagel, S., & Song, Z. (2022). Treasury inconvenience yields during the COVID-19 crisis. Journal of Financial Economics, 173(1), 57–79. Hortaçsu, A., & Kastl, J. (2012). Valuing dealers’ informational advantage: A study of Canadian treasury auctions. Econometrica, 80(6), 2511–2542.

300  Research handbook of financial markets

Hortaçsu, A., Kastl, J., & Zhang, A. (2018). Bid shading and bidder surplus in the US treasury auction system. American Economic Review, 108(1), 147–169. Hu, G. X., Pan, J., & Wang, J. (2013). Noise as information for illiquidity. Journal of Finance, 68(6), 2341–2382. Kim, D. H., & Orphanides, A. (2012). Term structure estimation with survey data on interest rate forecasts. Journal of Financial and Quantitative Analysis, 47(1), 241–272. Krippner, L. (2012). A model for interest rates near the zero lower bound: An overview and discussion. Analytical Note, Reserve Bank of New Zealand AN 2012/05. Lou, D., Yan, H., & Zhang, J. (2013). Anticipated and repeated shocks in liquid markets. The Review of Financial Studies, 26(8), 1891–1912. Malvey, P. F., Archibald, C. M., & Flynn, S. (1995). Uniform-price auctions: Evaluation of the treasury experience. Department of the Treasury, working paper. Nyborg, K. G., & Sundaresan, S. (1996). Discriminatory versus uniform treasury auctions: Evidence from when-issued transactions. Journal of Financial Economics, 42(1), 63–104. Office of Debt Management. (2017). Treasury presentation to TBAC. Retrieved from https://home​ .treasury​.gov​/system ​/files​/276​/Q22​017C​ombi​nedC​harg​esfo​rArchives​.pdf. Piazzesi, M. (2010). Affine Term Structure Models. Handbook of Financial Econometrics. NorthHolland, pp. 691–766. Svensson, L. (1994). Estimating and interpreting forward interest rates: Sweden 1992–1994. Working paper, National Bureau of Economic Research. Tindall, M. L., & Perez, M. A. (2021). Treasury auction during the pandemic: Stresses but few surprises. Federal Reserve Bank of Dallas economics Note, August 31. Vasicek, O. (1977). An equilibrium characterization of the term structure. Journal of Financial Economics, 5(2), 177–188. Vayanos, D. and J.-L. Vila, 2021. A preferred-habitat model of the term structure of interest rate. Econometrica, 89(1), 77–112. Waggoner, D. F. (1997). Spline methods for extracting interest rate curves from coupon bond prices. Working paper, 97–10, Federal Reserve Bank of Atlanta. Zivney, T. L., & Marcus, R. (1989). The day the United States defaulted on treasury bills. Financial Review, 24(3), 475–489.

An online appendix for this chapter is available at: https://www.e-elgar.com/textbooks/gurkaynak.

14. The municipal bond market Daniel Bergstresser1

The goal of this chapter is to describe for academic researchers the important aspects of the municipal bond market and some of the main themes and findings from the existing body of academic research on that market. The municipal bond market is distinct in terms of purposes, participants, regulation, and market structure, and this chapter will emphasize the unique elements of the market’s institutional environment.1 The municipal bond market is made up of instruments issued by states, political subdivisions of states, and by state-chartered entities and authorities like port authorities, universities, and nonprofit hospitals. This market is large, with approximately 50,000 distinct municipal issuers and about $4 trillion in debt outstanding, and it is extremely diverse in terms of issuers and issuance purposes.2 A large share of municipal debt is exempt from federal income tax, so it is common for people to refer to the municipal market as the “tax-exempt” market. This is true even though bonds that are “tax-exempt” can face taxation based on capital gains, even though “taxexempt” issuers will sometimes issue bonds that pay interest that is taxable at the federal level, and even though some issuance that is exempt from federal income taxation turns out, in fact, to be issued by or on behalf of for-profit corporations. This tax exemption means that a relatively simple equation, reflecting the decision problem for an investor who is on the margin between municipal debt and an alternative taxable investment choice, can organize analysis of the municipal market and also of some of the academic research on that market:

E[(1 − τmuni) * Rmuni,i] = E[(1 – τcorp) * Rcorp,i] (14.1)

For the investor who is on the margin between investing in municipal bonds and investing in an alternative asset, for example corporate bonds, the after-tax expected return (in the appropriate probability measure) on municipal debt equals the after-tax expected return on the alternative investment. If the municipal debt is tax-exempt and both assets are free of risk, then the relationship simplifies further: Rmuni,i = (1 − τcorp)*Rcorp,I (14.2) Behind this simple equation lies significant general equilibrium complexity. Assessing the incidence of (i.e. who benefits from) the municipal tax exemption requires taking a view about the supply and demand for capital in both the municipal sector and in competing sectors, 1 I am grateful for helpful comments from the editors Jonathan Wright and Refet Gurkeynak, from the discussant Francisco Roch, and from participants in the Research Handbook on Financial Markets conference in November 2021 2 ​www​.sec​​.gov /news /public -statement /statement -clayton -olsen -2020 -05 -04 301

302  Research handbook of financial markets

as well as the general equilibrium effects of investment in the capital that municipal bonds finance. This leads to the second, related idea behind this chapter’s organization: that when considering the municipal market, one should remember the purpose of this market. Municipal bonds in the United States are, for the most part, used to finance the construction of physical public capital. This would include school buildings, roads, water treatment facilities, hospitals, and other types of public capital. The real impact of tax and regulatory policy that affects the municipal market will depend on the supply of, the demand for, the productivity of, and possible substitutes and complements for the capital that gets financed in this market. Academic research on those questions tends not to be confined to journals that target audiences whose interests are purely focused on financial markets. This chapter proceeds in six sections. The first section describes some basic features of municipal bonds. The second section describes some elements of the municipal bond life cycle, from conception to issuance to maturity and relevant points in between. The third section describes issuers, the fourth section describes the investor base, and the fifth section adds some details on the regulatory environment. This chapter then shifts to focus on the existing research on the municipal bond market and phenomena that are adjacent to and interact with that bond market. This discussion draws on the description of the institutional environment from the earlier sections of the chapter. In line with the organizing frameworks described above, this section attempts to be both broad and also deep specifically with respect to research on the real effects of the taxation and regulation of municipal bonds, and the economic incidence of municipal tax exemptions. A brief final section concludes and suggests potential directions for new research on the municipal market.

14.1 BASIC FEATURES OF MUNICIPAL BONDS This section develops a taxonomy of municipal debt based on the various characteristics of that diverse universe of financial instruments. The most common dividing line involves dividing the debt into two types: debt that is supported by a pledge based in some way on tax revenues versus debt that is backed by a pledge based on a revenue stream from some municipal project. The most well-known type of tax-supported debt is “General Obligation” or “GO” debt, which is backed by a pledge by the issuer to levy taxes to make required debt payments. General Obligation debt is not homogeneous, and within that category are different types of tax pledges that issuers have made to bondholders. An “Unlimited Tax General Obligation” pledge means that an issuer has a statutory requirement to raise taxes to pay debt, with no statutory limit on the level to which tax rates backing that debt can go. A “Limited Tax General Obligation” pledge means that the issuer is required to raise taxes to pay debt, but that a statutory limit on tax rates supersedes commitments to bondholders. General Obligation bonds can also differ with respect to the existence of specific tax pledges for specific bonds, they can differ with respect to specific liens that bondholders have against municipal revenue streams in the event of default, and they can differ with respect to the seniority that competing claims (for example, pension obligations) will have in the event of a default.3 3 ​www​.ncsl​.org​/ Portals​/1​/ Documents​/fiscal​/ LAM​etho​dolo​gyfi​​scal2016​.pdf

The municipal bond market  303

Other types of debt are supported by taxes but do not have a GO pledge. The structuring of securities and pledges in the municipal market is often driven by a desire to issue financial instruments that will have much of the economic substance of “debt” while not being legally defined as debt in a particular state, frequently because many states limit the issuance of financial instruments that are legally defined as debt. One example is lease arrangements where municipal issuers sell to investors very bond-like financial instruments, instruments that are issued by a special-purpose entity established by the underlying municipality. Proceeds from the issuance are used to finance an element of infrastructure, and this infrastructure is leased to the municipality. If all goes according to plan, then the lease payments appropriated by the municipality will eventually arrive in the hands of the investors that have purchased these very bond-like but not-quite-officially-a-bond financial instruments that the special-purpose entity has issued. Financial instruments like this can be characterized as “tax-backed”, but they earn that categorization based on something other than an unlimited pledge by an issuer to raise taxes to pay bondholders. In other cases, debt (of the non-lease variety) is backed by a pledge in some circumstances to allocate funds via some future legislative approval, approval that is presumed to be very likely but that is not promised. The presumption that approval is likely, plausible because of issuers’ desire to maintain access to credit markets, leads investors and others to call these bonds “moral-obligation bonds” or “appropriation-backed bonds”. A further distinction can be made between moral obligation bonds and annual appropriation bonds: with moral obligation bonds, the appropriation comes from a higher level of government than the issuer of the bonds. A typical example of a moral obligation bond would be a situation where a state housing authority issues bonds backed by housing-related revenue, with a “moral obligation” from the state to appropriate funds to pay the bonds if necessary. Credit rating agencies evaluating municipal bonds typically give some credit for non-binding “moral” obligations of this sort, but these non-binding commitments do not get the full amount of credit that a binding legal commitment would receive. The key point of this discussion is that although these very different types of pledges could all stand behind debt could be called “tax-backed” debt, the nature of the pledges can be very different. An unlimited tax general obligation pledge gives a much higher degree of assurance of payment for bondholders than a pledge based on a moral obligation by a state to appropriate funds. Turning to revenue bonds, these bonds are municipal bonds that are backed by specific revenue streams, generally from the project financed by the debt. A very incomplete list of the project types financed by revenue bonds would include water and sewer systems, toll roads, hospitals, and higher-education facilities. From the perspective of an investor, credit analysis of a revenue bond starts with an assessment of the reliability of the revenue streams that secure the bonds. An issuer of tollroad debt would typically commission a traffic study to establish the robustness of the toll revenue streams that securing the bonds. Revenue bonds can also be backed by payment streams that are labeled “taxes”. For example, airport revenue bonds can be backed by dedicated taxes on activities connected to the airport financed by that debt. Bonds can have both a pledge based on a specific revenue stream as well as a general obligation pledge; a bond with both pledges will sometimes be referred to as “Double Barreled” bonds. The hierarchy of credit quality for revenue bonds will reflect the degree to which a piece of infrastructure is essential to users that have the capacity and willingness to pay for its

304  Research handbook of financial markets

services. Bonds backed by revenues from an essential water and sewer system in a developed area will tend to have higher credit quality than bonds backed by a hospital that has close substitutes available nearby. And bonds backed by revenues from a tollroad system in an established area will have a higher credit quality than bonds backed by revenues from a toll road system in an area where future development is expected but not assured. Until the 2013 Detroit bankruptcy, it was conventional to think of revenue bonds as being, in general, riskier than general obligation bonds. But the arguments by Detroit during that bankruptcy process that the city’s general obligation debt should be viewed analogous to unsecured debt (relative to the “secured” water and sewer bonds that were issued by the city’s sewer authority) have made this presumption somewhat less automatic than it had been in before that bankruptcy. The negotiated settlement following the Detroit bankruptcy means that the case delivered fewer precedent-establishing rulings than one might have expected, given the importance of that bankruptcy. Understanding municipal bankruptcy law is critically important, but it is an area where each significant situation (Detroit, Puerto Rico, Jefferson County) has been sui generis in a way that can at times complicate analysis. In addition to this municipal borrowing from bond markets, a significant part of municipal “borrowing” comes in the form of obligations to make payments in the future to retirees. These obligations have been an important part of the Detroit bankruptcy and other significant recent municipal bankruptcies. The value of these future obligations is very large in aggregate, and their value – both in absolute terms and relative to the economic health of the plan sponsor – varies significantly from place to place. The rules that govern restructuring these obligations also vary from place to place, and in some places these obligations are senior by law to any bond-market obligations that a municipal issuer might have. One can also partition the municipal market into long-maturity versus short-maturity instruments. States and localities use short-term debt to cover timing mismatches between issuers’ receipts and their expenditures. A common example of this type of instrument is the “Tax Anticipation Note”, often called a “TAN”, used to bridge timing mismatches between outlays and tax collections. Other types of short-horizon mismatches will be bridged by notes such as “Revenue Anticipation Notes” (RANs) and “Grant Anticipation Notes” (GANs). Short-term instruments issued in advance of to-be-issued-later long-dated instruments are called “Bond Anticipation Notes”, or “BANs”. Longer-maturity municipal debt has maturities stretching out to 30 years; a typical bond issue will include multiple bonds with different maturities and possibly different call and redemption structures. For example, one issue might have 20 unique bonds with maturities going out as far as 30 years, often with the bonds maturing in ten or more years being callable starting at year ten. The design of maturity, coupon, and call structure can be quite complex. As with corporate debt, callability can be a matter of fixed schedules at pre-specified prices, and it can also include “make-whole” call provisions, that allow an issuer to call debt but at a price that varies with measured market interest rate levels. Other taxonomies would involve the taxability of municipal debt. Exemption from taxation provides an implicit subsidy, and the Internal Revenue Service only grants this subsidy for particular purposes. In a situation where more than ten percent of the proceeds of a debt issue are used for trade or business activities, that debt will be taxable at the federal level. Pension Obligation Bonds (sometimes called “POBs”) are bonds that are issued to fund unfunded parts of public pension obligations; these bonds are also not tax exempt. Prior to

The municipal bond market  305

the 2017 changes to federal tax law, an issuer with outstanding tax-exempt debt that was not yet callable had one opportunity to issue tax-exempt debt to “advance refund” the outstanding tax-exempt debt prior to call. The proceeds from advance refunding would be invested in an escrow account with special-issue Treasury securities (called SLGS, for “State and Local Government Securities”), with that escrow dedicated to paying off the original debt to its original call dates. The 2017 TCJA eliminated issuers’ ability to advance-refunded tax-exempt debt with tax-exempt debt. The tax law change has not eliminated advance refunding, but since that change in the law issuers that want to advance-refund use taxable debt to advance refund their tax-exempt debt. The 2009 American Recovery and Reinvestment Act created “Build America Bonds” (often called “BABs”), which are taxable from the perspective of the holder but where the issuer receives a federal payment, originally equal to 35 percent of the interest payment. Instead of an implicit subsidy from tax exemption these bonds involve explicit payment from the United States Treasury to issuers. The appeal of this structure is that it can potentially broaden the set of investors for whom municipal debt might be an attractive investment habitat beyond the high-tax-rate investors that dominate the current investor base. Over $181 billion in BABs were issued in 2009 and 2010, but the failure of Congress to meet deficit targets triggered the significant reduction of this subsidy to issuers of this outstanding debt and has eroded enthusiasm for this type of bond.4 Interest income from municipal bonds is often exempt from state-level income taxes for investors who are in the same state as the issuer. There are some exceptions to this rule: there are states that tax municipal bond income from within-state bonds. Other states do not have income taxes at all. Utah is an example of reciprocity in taxation; an investor in Utah is not taxed on interest income earned on municipal bonds issued from a state when income on bonds issued from Utah would not be taxed. Interest income earned on bonds issued by US territories such as Puerto Rico and Guam is not taxable in any state, which (before the default) made Puerto Rican bonds attractive to investors both in Puerto Rico and also on the mainland. Another way in which municipal debt can vary is in the presence of third-party guarantees of the bonds’ principal and interest payments. Third-party guarantees can come from forprofit bond insurers, from public agencies, or from other entities. The Texas Permanent School Fund’s Bond Guarantee Program is an example of a state-chartered public entity that insures bonds issued by localities within that state. Beyond these public programs, the private market for municipal bond insurance was extremely large until the financial crisis of 2008, and it has shrunk significantly since then. A private guarantor for municipal debt bears (up to the limit of their capital) the risk of loss based on the underlying debt they insure; bond insurance is a way of transferring municipal credit risk from the nominal holders of that debt into the insurance sector. Another dimension along with bonds can differ is the nature of their coupon payments, fixed versus floating. Financial engineers have devised different mechanisms for municipal bonds to pay variable rates, and some of these mechanisms are complicated. Auction Rate Securities are a type of securities that pay interest that is reset based on periodic auctions, typically every week or month. Auction Rate Securities were typically marketed to investors as an asset that was relatively liquid and relatively free of credit risk. The high credit quality was often based on third-party insurance, and the perception of liquidity came from 4 ​www​.brookings​.edu​/ blog​/up​-front ​/2020​/12​/21​/why​-the​-surge​-in​-taxable​-​municipal​-bonds/

306  Research handbook of financial markets

the periodic auctions matching holders, buyers, and sellers of the instruments. During the 2007–2009 financial crisis many third-party insurers ran into financial difficulties. Auctions for ARS were designed so that in the absence of an auction-delivered market yield that was lower than a pre-arranged “maximum” rate, the current holders of the notes would continue to hold them, earning some pre-arranged yield that was generally lower than the latent yield that would have prevailed in the absence of the pre-arranged maximum. Variable Rate Demand Notes (VRDNs or VRDOs) pay interest at a rate that is periodically determined by a remarketing agent to clear the market for those securities. The remarketing agent is obligated to repurchase the securities from holders at these periodic reset dates. A key to the VRDO structure is the presence of a Letter of Credit (LOC) or Standby Purchase Agreement under which VRDOs that cannot be sold back to investors will be put to the liquidity provider. Floating-Rate Notes (FRNs) pay a floating rate that is determined based on a published index, for example the LIBOR or SIFMA index. The terms of ARS, VRDOs, and FRNs can be very important and can vary from bond to bond, making the risk profile of bonds very different, even when they seem similar at first glance. The floating rate instruments in the municipal market are generally issued as part of a “synthetic fixed” structure, where an issuer will issue a floating-rate instrument and attempt to hedge the floating-rate risk with an interest-rate swap contract. These synthetic-fixed structures often involve basis risk; if the swap contract is tied to LIBOR, then that contract’s reference rate may not perfectly capture all the risks in the payments on the bond side of the pair of transactions. While financial innovation can serve the needs of issuers and investors, there is some evidence that financial innovation in the municipal market can amplify principal-agent problems. For issuers, a synthetic-fixed structure can allow officials to borrow at what appear to be lower rates than traditional callable-fixed structures; the apparent lower cost comes from the basis risk taken on between the variable-rate bonds and the swap contract and from the typically non-callable nature of synthetic-fixed issuance.5 The final way in which municipal debt can differ is in legal form – bond versus loan. Other structures, including conduit structures, occur. Many states have charged public entities that serve as issuing conduits for municipal bonds that are issued by and guaranteed from the cash flows from issuers inside of (and outside of) their states. “Brandeis University” bonds are issued by the Massachusetts Development Finance Agency but are secured only by resources available to Brandeis University. Conduit issuance can cross state lines; a recent issue on for the Albert Einstein College of Medicine was issued by the Wisconsin Public Finance Authority.

14.2 THE LIFE CYCLE OF A MUNICIPAL BOND This section describes the life cycle of a municipal bond, from the moment of conception to the moment the debt is repaid or otherwise extinguished. A municipal bond can be issued for one of three reasons. The debt can be issued to raise new funds for a project; these bonds

5 See Perignon and Vallee (2017) for evidence on principal-agent problems and municipal financial innovation in France.

The municipal bond market  307

will be called “new money” bonds. Bonds can be issued to refund existing debt specifically to generate interest savings; this type of transaction will be called a “refunding” transaction. “Restructuring bonds” may be issued, often in times of financial distress or constraint, to provide debt service or covenant relief. Before issuing a municipal bond, an issuer contracts with the third-party agents involved in the issuance process. Most municipal issuers hire financial advisors to help design and structure bond issues. The regulation of the municipal financial advisor market has changed recently and the fiduciary obligation of these advisors to their clients has been made explicit. Issuers engage bond counsel to give an opinion on the legality of the bond issue and on the tax exemption of the bonds’ interest payments. Bond issuance may require approval in the form of a bond election or bond referendum. Rules for these elections vary, but in general a successful bond election is more likely to be required when the proposed bonds would have a GO pledge. This bond referendum requirement may complement or substitute for a requirement that debt not exceed statutory debt limitations. Focusing now on the sales process, issuers occasionally use private placement transactions, selling new bonds directly to investors. Private placement financing will typically place bonds with one investor, often a bank or insurance company. The private placement process can allow an issuer to raise capital more quickly than going to public bond markets. If the issuer decides to issue bonds into public markets, then the issuer must choose whether to sell the bonds via a competitive process or via a negotiated process.6 In a competitive sale the issuer advertises the new issue and the basic structure of the bonds being offered via a Preliminary Official Statement (POS). This overall structure advertised to potential bidders may include the face-value amounts of each maturity to be offered and features of call options each bond in the issue would have. The issuer must also indicate how the winning bid will be chosen. The most common method of selecting a winning bid is to choose the bid that offers the lowest “True Interest Cost” (TIC) for the entire bond issue being offered. The TIC is the net yieldto-maturity of the cash flows in the entire bond offering from the perspective of the issuer. On the day of the bond offering the bids are evaluated, the winning bidding syndicate is awarded the bonds, and these bonds are then sold to investors. Use of the TIC measure is an improvement on the “Net Interest Cost” (NIC) method that prevailed into the 1980s and is still occasionally used. The NIC method does not consider the time value of money; it was used in the pre-computer era because of ease of calculation. Even the TIC method does not consider the impact of call options, an important weakness given the ubiquity of these options. In a negotiated bond offering the issuer, generally working with a financial advisor, first selects an underwriting syndicate to help place the bonds with investors. The selection of an underwriting syndicate can follow a Request for Proposals process. The underwriter and advisor may work with the issuer to structure the bonds being offered, although it is important to distinguish the advisor role (a fiduciary of the issuer) with the underwriter’s role (an intermediary without a fiduciary obligation to the issuer). The underwriter works with the issuer to construct a POS, which is then used to market the bond offering to potential buyers and to establish the details of the issue’s structure. The question of which – between competitive

6 A number of states require issuers to use a competitive sales process, in those states there is not a choice.

308  Research handbook of financial markets

versus negotiated issuance processes – is “better” for issuers has been a highly contested area of municipal research.7 A refunding transaction is distinct from a new-money transaction in that the proceeds of the bond offering will be used to refund outstanding municipal bonds. Refunding transactions can be current refundings, in which the bonds to be refunded will be refunded using the proceeds of the bond offering within 90 days of that offering, and they can be advance refundings, in which the bonds being refunded are not yet near call or maturity dates. In an advance refunding, the proceeds of the bond issue are placed into escrow, invested in U.S. Treasury securities, dedicated to paying off the outstanding debt. Until the 2017 tax reform, an issuer was able to advance-refund a tax-exempt bond issue with new tax-exempt bonds only one time; post-2017 issuers are not able to advance refund bonds with tax-exempt bonds. The timing and structure of bond refunding is a complex problem because of the various levels of optionality involved in issuance and refunding decisions.8 Municipal bonds then trade in over-the-counter markets centered around a network of dealers. Investors wishing to purchase or sell municipal bonds will work, either directly or through a broker, with a dealer. Seasoned municipal bonds tend to be much less liquid than Treasury bonds. Treasury bonds number in the hundreds and have an aggregate par value of over $20 trillion, while municipal bonds number in the millions while having an aggregate par value of near $4 trillion. Municipal bonds also often have unique credit features that make credit analysis more of a bespoke process than for Treasury bonds. And the tax characteristics of municipal bonds – often tax-exempt at the federal level, and frequently tax-exempt for withinstate holders – can have the effect of reducing the breadth of the market for individual bonds. Municipal distress and default, although unusual and especially unusual among rated issuers, does occur. States in the United States are not allowed to file for bankruptcy protection and there have been no state defaults since Arkansas defaulted on its bonds in 1933. Chapter 9 of the United States Bankruptcy Code governs municipal bankruptcy in the set of states that allow issuers to file for bankruptcy protection. In some states the right to file for bankruptcy protection is automatically granted to municipal issuers, while in other states an approval process is required. Bridgeport, Connecticut is an example of a municipality that had a bankruptcy filing rejected after its filing was opposed in court by the state. Recent years have seen some high-profile bankruptcies, including Detroit, Jefferson County, Alabama, and a handful of large cities in California. The Puerto Rico default, a sui generis default by an American territory, not eligible for Chapter 9, is being handled under a law that was specifically designed and passed to handle that commonwealth’s default. Most municipal bankruptcies are not by cities or counties, but are instead by special purpose districts, and defaults on un-rated bonds issued by special purpose districts for relatively speculative purposes are common relative to other parts of the municipal market.

14.3 MUNICIPAL ISSUERS This section describes relevant features of the population of municipal issuers. The goal is to provide context for the discussion of the academic research on municipal credit markets in the 7 See Cestau et al (2020) for a recent paper, and Simonson and Robbins (1996) for an older one. 8 ​https:/​/mrsc​.org​/ Home​/ Explore​-Topics​/ Finance​/ Debt​/ Debt​-Management​-P​olicies​.aspx

The municipal bond market  309

sections that follow. This section has two main points: municipal issuers are large in number and extremely diverse, and the median municipal issuer issues debt in amounts and at frequencies that are low by the standards of issuers in other debt markets. Table 14.1 shows some data on a selection of individual issuers. The data come from the Mergent FISD database of municipal bonds. The first row shows that that database (as of 2017) listed 47,126 distinct municipal issuers, and these issuers had collectively issued $7.6 trillion in debt since 2001. $3.9 trillion of that total remained outstanding as of 2017. The total number of bonds issued over that period was 2.4 million.​ The largest single issuer was the State of California, which issued $241 billion, or 3.14 percent of the total face value of bonds issued during that period. $84 billion issued by this issuer remained outstanding at the end of the period. The top six issuers are likely to be familiar. The first issuer that may be unfamiliar and deserves some specific attention is the issuer listed as “J.P. Morgan Chase Putters/Drivers”. This issuer has roots in IRS rules that prevent interest payments from being deducted from taxes if the interest is on debt used to buy taxexempt bonds. This rule should make it impossible for an investor to use leverage to purchase a portfolio of municipal debt. But “Tender Option Bond” (or “TOB”) programs, of which the J.P. Morgan Putters/Drivers program is one variety, allow an investor to achieve economic substance of using debt to finance municipal bond holdings while retaining the tax benefits of the interest tax exemption. In a one-bond TOB program, a fixed-rate municipal bond is placed into a trust, and that trust issues two securities. The senior security is a floating rate security (in the J.P. Morgan case called the “Putter”) that is senior, has minimal interest rate risk, and intended for investors who want a safe and highly liquid investment. Money-market mutual funds have frequently been the dominant type of investors in “Putters” and similar TOB securities arranged by other dealers. The residual security (in this case the “Driver” security) will have an inverse floating rate structure, and it will be junior in the TOB capital structure to the floating rate security. This residual security has the economic substance of a levered position in the underlying bond, but retains the tax benefits of a tax-exempt municipal bond. TOB programs have shrunk since the 2007–2009 financial crisis, but they highlight the ways in which leverage can occur even in markets when the conventional understanding might be that (whether for tax reasons or otherwise) investors are operating without leverage. The Commonwealth of Puerto Rico has been another large issuer, and Puerto Rico defaulted on its debt in 2016. This bankruptcy process (now occurring under the “PROMESA” legislation designed specifically for that case) has been complex because of the competing claims of different Commonwealth-chartered issuers, which issued bonds secured by different revenue streams. “The Commonwealth of Puerto Rico”, which issued GO bonds, is just one among many issuers involved in the Puerto Rico default. The top 20 issuers represent 0.04 percent of the total number of issuers by count, and their issuance has amounted to 14.75 percent of the total issuance in the market. The remaining rows show the issuers on either side of the decile cutoffs, by volume, and illustrate more elements of the complexity of this market. For example, the 37th-largest individual issuer is the “Golden State Tobacco Securitization Corporation”. State attorneys general settled claims against tobacco manufacturers in the 1990s. The settlements involved payments from the manufacturers to the states that would be based on tobacco revenues, in perpetuity. Many states securitized these revenue streams, issuing bonds whose cash flows will depend on the payments from tobacco manufacturers based on their sales. These bonds have frequently been highly engineered, with complex cash flow waterfalls that are based on tobacco revenues

310

(4)

(5)

0.02%

0.02%

0.02%

8

9

10

0.01%

5

0.01%

0.01%

4

0.01%

0.01%

3

6

0.00%

2

7

State of California

1

45,923

55,458

58,044

59,225

1,07,844

1,13,774

2,41,372

State of Washington

Metropolitan Transportation Authority of New York 39,922

44,023

New York City Municipal Water Finance 45,259 Authority Revenue Bonds

J.P. Morgan Chase Putters/Drivers

State of Illinois

Commonwealth of Massachusetts

New York City Transitional Finance Authority Revenue Bonds

New York City, NY

State of Texas

Note: all issuers in database

0.00%

47,126

10.54%

10.02%

9.45%

8.86%

8.26%

7.54%

6.79%

6.02%

4.62%

3.14%

76,93,529

(3) Total issued ($, Share of total $ issued by millions) issuers of that rank and above

(2)

Rank / Rank / Issuer name number percentile

(1)

(7)

3,012

1,929

1,474

2,713

1,498

1,786

3,179

6,842

2,920

3,975

24,46,384

25,076

26,183

27,861

7,567

26,120

27,060

34,064

43,279

23,527

84,487

38,66,162

Total bond Total outstanding in count issued 2017 ($, millions)

(6)

Table 14.1  Selection of municipal bond issuers in Mergent FISD database, ranked by total nominal dollars issuance between 2001 and 2017

311

0.04%

0.04%

0.04%

17

18

19

0.04%

New York State Dormitory Authority

0.03%

16

20

New Jersey State Transportation Trust Authority

0.03%

15

State of Connecticut

Commonwealth of Puerto Rico

Los Angeles Unified School District

Port Authority of New York and New Jersy

California Statewide Communities Development Authority Revenue Bonds

State of New Jersey

0.03%

0.03%

New Jersey Economic Development Authority Revenue Bonds

13

0.03%

12

Illinois Revenue Authority Revenue Bonds

14

0.02%

11

28,061

28,482

30,860

31,489

32,594

33,056

33,753

34,713

34,966

36,020

14.75%

14.39%

14.02%

13.61%

13.21%

12.78%

12.35%

11.91%

11.46%

11.01%

774

1,274

1,750

848

1,487

1,741

369

2,172

2,023

2,858

13,553

13,351

19,625

25,756

21,746

15,420

2,546

17,922

15,970

23,474

(Continued)

312

0.76%

1.47%

1.47%

694

695

0.39%

184

0.76%

0.39%

183

356

0.19%

90

357

0.19%

89

Philadelphia Authority for Industrial Development Lease Revenue Bonds

Pennsylvania State Public School Building Authority Lease Revenue Bonds

Rhode Island Housing & Mortgage Finance Corp.

Detroit School District

Miami-Dade Country School Board Certificates of Participation

Houston Independent School District

Dallas-Fort Worth International Airport Revenue Bonds

New Jersey Health Care Facilities Financing Authority Revenue Bonds

Chicago, Illinois

1,639

1,640

3,223

3,245

6,039

6,046

11,386

11,413

19,208

0.08%

60.01%

59.99%

50.01%

49.97%

40.01%

39.93%

30.14%

29.99%

20.11%

19.86%

38

Golden State Tobacco Securitization Corporation (California)

19,529

(5)

0.08%

(4)

37

(3) Total issued ($, Share of total $ issued by millions) issuers of that rank and above

(2)

Rank / Rank / Issuer name number percentile

(1)

Table 14.1  (Continued) (6)

(7)

89

138

1,192

328

596

346

614

1,179

1,429

402

741

316

1,018

1,840

4,241

3,260

6,555

6,446

11,490

12,546

Total bond Total outstanding in count issued 2017 ($, millions)

313

15.12%

15.12%

56.39%

56.39%

7,124

7,125

26,574

26,575

6.16%

2,901

Metro-North (Missouri) Fire Protection District

Iowa Finance Authority Mortgage Revenue Bonds

Dalton (Georgia) Development Authority Revenue Bonds

Salina, New York

South San Antonio (Texas) Independent School District

New Braunfels (Texas) Independent School District

East Brunswick Township, New Jersey

2.90%

6.15%

1,369

Marble Falls (Texas) Independent School District

2.90%

2,900

1,368

10

10

105

105

331

331

802

804

99.00%

98.99%

90.00%

89.99%

80.00%

79.99%

70.00%

69.99%

65

4

17

85

199

257

70

232

5

-

42

10

230

194

53

100

314  Research handbook of financial markets

in the distant future. Further restructuring of these highly complex securities has at times allowed issuers to achieve even more short-term budget relief at the expense of future budgets, in what has amounted to an end-run around state balanced-budget requirements.9 The 356th-largest unique issuer in the sample is the Detroit School District, which issued $3.2 billion worth of bonds, of which $328 million remained outstanding at the end of the sample. Issuing at this scale put that school district in the top 0.76 percent of individual issuers. Half of the total issuance in that market was issued by this top 0.76 percent of issuers. Most of the issuers in the market are extremely small issuers; half of the issuance has been by the 46,770 issuers that issued less than $3.2 billion each. Putting this diversity and complexity into context, the U.S. Treasury has 362 distinct CUSIPs outstanding and these bonds have a total face value of over $22 trillion, or about $60 billion per bond. In the municipal market the average bond size is about $3 million and the average amount issued in total by a municipal issuer over the 2001–2017 period was $164 million. The median dollar of debt was issued by an issuer that had issued $3.2 billion in total, and the median unique issuer in the municipal market has issued a total amount of about $10 million. This situation reflects our federal structure and the extent to which control of local decisions is delegated to local authorities, but it creates some challenges for the market. Most issuers are inexperienced at issuance relative to the United States Treasury or relative to issuers in the corporate bond market. Municipal bonds tend to be much less liquid than other securities, in part because the bond sizes are so small, and most issuers are not familiar to typical investors. This makes credit rating agencies particularly important in this market, and it was part of the reason why the bond insurance market enjoyed such high penetration until the financial crisis. The progressive tax rates of the tax code create incentives for high-income individuals to hold the debt, but the disaggregation of the individual issuers makes the market illiquid and creates challenges for credit analysis. This granularity and the scale dominance of a relatively small number of issuers should also be considered by academic researchers studying this market. The results of by-issuer regression analysis may end up being driven by the very large number of miniscule issuers whose characteristics have little in common with the State of California or New York City. Analysis that focuses on the largest issuers may end up missing the specific challenges of the very large number of smaller municipalities in the United States. Another challenge for academic research on the municipal bond market, and for research on local public finance in general, is the overlapping nature of and claims of different municipal authorities. Some authorities (for example the Port Authority of New York and New Jersey) cross state boundaries. In some states, cities, counties, school districts, and other authorities will have partially but incompletely overlapping boundaries. In other places (for example Massachusetts) schools are operated at the city/town level and counties have relatively limited powers and responsibilities.

14.4 MUNICIPAL INVESTORS Households play a very large role in the American municipal bond market; the market tends to have fewer layers of intermediation than in other bond markets. The market has also been 9 See New Jersey’s 2014 tobacco bond restructuring: www​.propublica​.org​/article​/ behind​-new​-jerseys​tobacco​-bond​-bailout​-a​-hedge​-funds​-100​-million​-payday.

The municipal bond market  315

unique in the degree to which issuers purchased third-party insurance. Before the 2007–2009 credit crisis, almost half of municipal debt was sold with bond insurance, so much of the municipal credit was held not by the nominal owners of the bonds but instead by insurance companies that had insured the debt. Because most municipal debt is tax-exempt, the bonds are disproportionately held by investors for whom that tax advantage is particularly useful: the investors who face high marginal tax rates. Table 14.2 shows aggregate ownership of municipal bonds, by sector, since 1960.10 These data come from the Federal Reserve Board’s Financial Accounts database and require some important caveats. First, a change in the methods used to construct these data in 2004 means that the aggregate size of the municipal bond market appears in that year to jump by about $1 trillion; this apparent jump is an artifact how the Federal Reserve constructs these data. Researchers should be careful using this aggregate as a time series because of that jump; a naïve analyst might falsely believe that the size of the municipal market had jumped in that year. In addition, the household numbers are calculated as a residual, meaning that “the household sector” includes assets held through hedge funds and through separately managed accounts. Separately managed accounts are important in the municipal bond market, but this investment channel remains under-studied relative to mutual funds, hedge funds, and other delegated investment channels. Finally, “other financial business” includes a range of holders, including programs established by the Federal Reserve System during the Covid-19 pandemic. Table 14.2 shows the importance of the household sector in the municipal bond market. Of the $24 trillion in U.S. Treasury bonds, under eight percent were held directly by the household sector at the end of 2020, but this figure is over 44 percent for municipal bonds. Including ownership through mutual funds, which are disproportionately held by households and are particularly transparent intermediaries, brings the total to 71 percent. This has not always been the case: until the tax law changes of the 1980s, depository institutions and, to a lesser extent, insurance companies were the dominant habitats for municipal debt. The Federal Reserve’s Financial Accounts data are useful for looking at the ownership of municipal debt by investor type, but these data cannot tell us how that debt is distributed within the household sector. Bergstresser and Cohen (2015) use the Fed’s Survey of Consumer Finances data to investigate how this debt is held within the household sector. This paper finds that the probability that households own municipal debt has fallen significantly over the period leading up to 2013. Municipal bonds, an asset class that in 1989 were frequently held across the broad upper-middle-class, were by 2013 much more concentrated in the portfolios of the top one percent and 0.5 percent of households. This change has coincided with the increasing tilt toward tax-deferred investment vehicles such as 401(k)s and IRAs as the primary locus of investing for relatively affluent households. Ownership of municipal bonds through mutual funds is, on some level, a bit of a puzzle. The main habitat for tax-exempt municipal bonds is with investors who are concerned about taxes, but an investor who invests through a mutual fund cedes a significant element of control over their taxes to the decisions of the fund management team and, indirectly, to the behavior of fellow investors in the funds that they’ve invested in. Some of the decline in ownership through this channel may reflect mutual funds being supplanted by separately 10 Although this chapter focuses on the municipal bond market, there is some direct bank lending to municipalities. For research on the scale and impact of direct bank lending see Bergstresser and Orr (2014), Dagostino (2019), and Ivanov and Zimmermanm (2021).

316

20,327

-

-

400

-

-

1960

100.0%

43.7%

3.4%

Mutual funds and ETFs

GSEs

Brokers and dealers

Other financial business

Rest of the world

(as share of total)

Total assets

Household sector

Nonfinancial business

11,672

4,406

2,747

18,357

State and local governments

Depository institutions

Insurance sector

2,382

Nonfinancial business

Public retirement funds

70,524

31,007

Household sector

1.5%

32.4%

100.0%

1970

-

-

915

-

-

2,032

2,384

2,175

47,144

145,501

70,971

Total assets

1970

1960

(in current $, millions)

2.3%

32.5%

100.0%

1980

467

-

2,509

-

6,387

4,059

87,234

152,362

7,008

9,362

129,980

399,368

1980

2.1%

55.8%

100.0%

1990

2,200

-

7,894

3,685

203,038

473

149,244

120,443

11,583

24,671

659,885

1,183,116

1990

2.3%

32.3%

100.0%

2000

8,000

-

11,289

29,197

583,179

1,707

208,856

117,296

3,886

34,384

475,606

1,473,399

2000

0.8%

50.4%

100.0%

2010

71,736

-

40,017

23,826

962,971

1,921

482,876

257,140

14,061

29,562

1,915,309

3,799,419

2010

0.7%

47.1%

100.0%

2015

87,019

-

13,951

8,608

949,189

2,522

534,812

514,175

15,032

26,613

1,913,499

4,065,420

2015

0.7%

44.6%

100.0%

2020

108,483

6,283

6,780

2,757

1,163,539

41

532,080

522,777

18,639

31,553

1,923,340

4,316,272

2020

Data from Federal reserve. A series break between 2003:Q4 and 2004:Q1 reflects data construction and not real changes. Household sector, as reported, includes both hedge funds and money invested via separately managed accounts (SMAs).

Table 14.2  Holdings of municipal bonds, by investor type

317

0.0%

Rest of the world

0.6%

0.6%

0.0%

0.0%

GSEs

Brokers and dealers

0.0%

Mutual funds and ETFs

Other financial business

0.0%

6.2%

Public retirement funds

48.5%

0.0%

0.0%

0.0%

1.4%

14.0%

25.9%

16.4%

Depository institutions

1.6%

Insurance sector

3.9%

State and local governments

0.1%

0.0%

0.6%

0.0%

1.6%

1.0%

21.8%

38.2%

1.8%

0.2%

0.0%

0.7%

0.3%

17.2%

0.0%

12.6%

10.2%

1.0%

0.5%

0.0%

0.8%

2.0%

39.6%

0.1%

14.2%

8.0%

0.3%

1.9%

0.0%

1.1%

0.6%

25.3%

0.1%

12.7%

6.8%

0.4%

2.1%

0.0%

0.3%

0.2%

23.3%

0.1%

13.2%

12.6%

0.4%

2.5%

0.1%

0.2%

0.1%

27.0%

0.0%

12.3%

12.1%

0.4%

318  Research handbook of financial markets

managed accounts, a channel that permits individual households to have more control while maintaining many of the benefits of a mutual fund, particularly delegated security selection and monitoring.

14.5 SOME FEATURES OF THE REGULATION OF MUNICIPAL MARKETS Three key elements stand out in the regulation of municipal bond markets. First, tax laws establish the circumstances under which municipal bond interest will be taxable or not taxable. The second unusual feature of this market is the limits on the Securities and Exchange Commission. Unlike corporate bond issuers, federal law places limits on what the SEC can require of municipal issuers. A final relevant element is bankruptcy law, in particular Chapter 9 of the United States bankruptcy code, as well as other state and federal laws that affect municipal issuers that have defaulted. Section 103 and sections 141–150 of the Internal Revenue Code govern the tax exemption of interest. Section 103 establishes the basic exemption: “Except as provided in subsection (b), gross income does not include interest on any State and Local bond”. What follows in the Internal Revenue Code are the exceptions to that exclusion (private activity bonds, for example), and then the exceptions to those exceptions. A key principle in tax law is that “arbitrage” should be prevented. In this context, arbitrage refers to the fact that municipal obligations, which are free from federal taxation, can carry lower yields than treasury securities in which municipalities might invest. Absent arbitrage restrictions, municipalities could issue debt, invest the proceeds of that debt issuance in higher-yielding Treasury bonds, and capture the difference in yield. Demonstrating that compliance with anti-arbitrage tax rules involves demonstrating how the proceeds from a bond issue will be used. In addition to the arbitrage restrictions, tax policy pays particular attention to defining the boundaries of and exceptions to the “private activity bond” exception to the tax-exemption of municipal interest. While private activity bonds are generally not tax-exempt, the tax code defines “Qualified Private Activity Bonds” as private activity bonds whose proceeds finance projects deemed worthy of the tax exemption. Hildreth and Zorn (2005) highlight how the 1986 Tax Reform changed the municipal market by tightening the requirements for this municipal tax exemption, in particular for private activity and industrial development bonds. Turning to the Securities and Exchange Commission, a useful starting point is to note that the SEC is not allowed to require much, if anything, from municipal issuers. The Securities Act Amendments of 1975 established the Municipal Securities Rulemaking Board (MSRB), a selfregulatory organization that has authority for establishing rules that govern municipal underwriters, and that has had since the Dodd-Frank reforms authority for governing municipal advisors. The framework for disclosure in the municipal market is SEC Role 15(c)2-12, passed in 1994, which requires underwriters to obtain and distribute an issuers’ official statement, and requires underwriters to be assured by issuers that the issuers will produce information over the life of the bonds. This legal requirement is on dealers, however, and enforcement has been imperfect (see Baber & Gore, 2008). Default by states is rare but not unprecedented; there was a wave of state debt defaults in the 1840s, and Arkansas defaulted on its bonds in the 1930s. The Eleventh Amendment to the

The municipal bond market  319

constitution prevents foreigners or citizens of other states from suing a state in federal court. The abilities of other potential parties to successfully sue states in federal or state court has not been high.

14.6 EXISTING RESEARCH ON THE MUNICIPAL MARKET This discussion begins with some comments about research on the core objective of the municipal market: financing infrastructure. Aschauer (1989) investigates the economic impact of public infrastructure investment, using time-series macroeconomic data from the United States to estimate a generalized Cobb-Douglas model of the economy with an additional “public capital” factor of production. He finds high marginal returns to public capital investment and also finds that slowing public capital investment may have been a partial cause of the observed productivity slowdown of the 1970s. A review article by Gramlich (1994) comes to a more nuanced conclusion about the economic benefits of infrastructure investment, and focuses on the challenges involved in disaggregating and measuring prices and quantities of infrastructure. Fernald (1999) highlights the importance of considering average versus marginal effects of infrastructure. For example, the interstate highway system caused a large increase in productivity, but building a second highway system right next to the first one would offer lower incremental benefits. More recent work (see, for example Redding and Turner, 2015) has begun to introduce spatial general equilibrium models, often based on models from the international trade literature, to estimate the impact of infrastructure investment. Gupta et al. (2020) highlight the importance of this type of spatial analysis. These authors estimate the size and spatial distribution of the value increase that comes from subway extensions in New York City. Many of the papers discussed here previously focus on transportation infrastructure, but there are other important forms of infrastructure as well. Cutler and Miller (2006) show that the growth in the municipal bond market was a driving force behind the expansion of clean water availability in American cities, a development that Cutler and Miller (2005) argue is responsible for half of the total mortality reduction and two thirds of the child mortality reduction in American cities since the 1850s. They argue that clean water investment in that era returned 23 dollars for every dollar invested. Brooks and Liscow (2020) focus on the price of infrastructure investment, focusing on the per-mile cost of building interstate highways in the United States between the 1960s and the 1980s. They show that the real cost of highway construction tripled over this period. The timing of the increase suggests the rise in construction costs was caused at least in part by a changing political environment that increased citizens’ capacity to block or hinder new highway construction. A key insight from public economics is that the person who bears the economic cost of a tax is not necessarily the person who writes the check that nominally pays that tax. In spite of this, traditional empirical and policy analysis has often assumed that the beneficiaries of the interest tax exemption were current holders of municipal bonds, and analysts have often estimated benefits of the tax assumption by maintaining the assumption that pre-tax prices and allocations of financial assets are not affected by the tax exemption.11 But work by Poterba and 11 See, for example, the work of the Congressional Joint Committee on Taxation.

320  Research handbook of financial markets

Verdugo (2011) has used the Surveys of Consumer Finances to show sensitivity of estimates of the revenue cost of the interest tax exemption to assumptions about the portfolio substitution behavior of households. Their work suggests that the revenue cost of the tax exemption in 2004 could have been under $9 billion per year if the alternative household investment in the absence of the exemption would have been corporate stock, and $14 billion per year if the alternative was taxable bonds. Galper et al. (2013; 2014) allow the pre-tax interest rates to vary with effective tax rates and consider the distributional implications of the capital financed by municipal bonds; their work suggests that the benefits of the income tax exemption are concentrated at upper incomes, but also potentially among lower-income households who disproportionately benefit from incremental capital financed in that market.12 More recent work by Garrett et al. (2017) focuses on imperfect competition in the dealer market for newly auctioned municipal bonds, and their work delivers the surprising conclusion that this market structure leads to greater-than-unity passthrough of the value of the tax exemption to issuers. Adelino et al. (2017) develop evidence on the real impact of the municipal bond market using the one-time recalibration of Moody’s credit rating scale for municipal bond issuers that occurred in 2010. This recalibration led to an upgrade of credit ratings for 18,000 municipal issuers. The authors show that this recalibration expanded access to credit for the municipalities that got larger upgrades, and they show that public employment increased in counties where a larger fraction of the issuers were upgraded. Gillette et al. (2020) show that credit ratings and disclosure appear to be substitutes, in the sense that the upgraded municipalities appear to reduce disclosure of continuing financial information over time. It remains to be seen whether this reduced disclosure over time will offset some of the effect that Adelino and co-authors found in their paper. Early work that is relevant for the economic incidence of the municipal interest exemption came from researchers who estimated the implied marginal tax rate based on municipal and comparable taxable yields. The organizing framework for this research has been Equation 14.2, in the opening section of this chapter, rearranged to solve for the equilibrium marginal tax rate of the investor who is indifferent between purchasing tax-exempt municipal bonds and purchasing an equivalent taxable bond: Rmuni,i = (1−τcorp)*Rcorp,i (14.3)

τimplied = 1 − Rmuni/Rcorp. (14.4)

The early literature on implied marginal tax rates (for example, Kidwell and Koch, 1983) found that implied marginal tax rates at different maturities were different, with implied marginal tax rates at longer maturities being lower than implied marginal tax rates at shorter maturities. Another way of describing this finding is that the municipal yield curve tends to have a steeper slope than otherwise equivalent taxable yield curves.13 Poterba (1989) showed 12 See also an earlier exchange in the National Tax Journal with competing theoretical models Fortune (1998) and Gordon and Metcalf (1991). 13 The tendency of municipal yield curves to have steeper slopes than equivalent taxable yield curves has been viewed by practitioners as part of the incentive for municipal issuers to issue floating-rate instruments paired with swaps (tied to taxable benchmarks) in synthetic-fixed packages.

The municipal bond market  321

that implied MTRs appeared to respond to actual personal tax rates, a result that exists in some degree of tension with earlier papers (for example, the Kidwell and Koch [1983] paper) that suggested that commercial banks were the marginal buyers of municipal debt. Green (1993) and Chalmers (1998) address the relative locations of the municipal and taxable yield curves, and make contributions to this literature, often called the “muni puzzle” literature. The puzzle in this literature has been the high yields in the municipal market relative to the market for United States Treasury securities, given that municipal bonds are mostly tax-exempt and given that default have been relatively rare in the modern era. As previously described, the apparent puzzle is particularly pronounced at the longer maturities. Green argues that the different slopes of the yield curves may stem from arbitrage activities by dealers within the taxable bond marketplace, which in his model has the effect of reducing effective tax rates on long-term taxable bonds and reducing the slope of the taxable yield curve. Chalmers exploits the fact that some municipal bonds are “pre-refunded” to contribute to this literature. In an advance refunding, a tax-exempt bond that is not yet callable is refunded with new bonds, and the proceeds of this new issue are placed into an escrow account and invested in U.S. Treasury securities whose cash flows are arranged to match the payments of the now pre-refunded (or “pre-defeased”) bonds. These pre-refunded bonds have the credit risk of U.S. Treasuries and pay interest that is (unlike Treasuries) tax-exempt, and Chalmers shows that even risk-free tax-exempt yield curves have a distinct upward slope relative to Treasuries, and the implied marginal tax rates at longer maturities appear low enough that the high municipal yields can be considered a puzzle. Chalmers (2006) makes the follow-on point that the systematic risk of municipal bonds does not appear different enough from Treasuries to explain the muni puzzle, and also points out the importance of making adjustments to standard practical bond analytics tools (duration calculations in his example) for tax-exempt bonds.14 Schwert (2017) reconsiders the Chalmers results and uses a variety of approaches to estimate the contributions of liquidity and credit risk to municipal spreads. His conclusion is that credit risk (and not liquidity) explains the bulk of observed credit spreads, a result that stands somewhat in contrast with Chalmers’ findings. One of Schwert’s approaches is to use Longstaff, Mithal, and Neis (2005)-style credit risk analysis using CDS spreads in the municipal market, using the very small set of municipalities that have actively traded CDS contracts. The external validity of results based on the small sample of municipal issuers on which municipal CDS contracts exist is an open question. Longstaff (2011) points out that the progressivity of marginal tax rates means that investors will face higher tax rates in good economic times than they do in bad times. This means that the wedge between pre-tax and after-tax returns on taxable assets will widen in good times, a feature that tax-exempt assets will not have. This induces a relative pro-cyclicality in the after-tax returns of municipal bonds that can explain a large part of the apparent muni puzzle, with some differences in the amount explained at different maturities. This explanation for the municipal yield curve slope complements work by Wang et al. (2008), who construct an asset pricing model that includes a liquidity factor and find that the relative illiquidity of longermaturity bonds helps explain the tendency of municipal yield curves to have a steep slope.15 14 See Kalotay and Buursma (2019) for an example of how fixed-income analytical tools can be adjusted in the context of the tax exemption and other institutional details of the municipal market. 15 He and Milbradt (2014), in a paper that focuses on the corporate bond market, highlight how liquidity risk and default risk in that market can interact and reinforce each other, posing a challenge for empirical identification of a pure “default” or a pure “liquidity” effect.

322  Research handbook of financial markets

An important area of research that is getting increasing attention is the optimality (or departures from optimality) of municipal security design and of municipal issuers’ behavior with respect to the call and advance refunding options on their outstanding debt. Ang et al. (2017) explore issuers’ advance refunding decisions, and they argue that issuers prematurely exercise advance refunding options. Their paper highlights that the horizons of municipalities may differ from the horizons of the officials responsible for their debt and refunding decisions; their results suggest that municipal issuers with more binding short-term financial constraints are more likely to engage in negative-NPV but current-budget-relieving advance refunding activity.16 Chen et  al. (2021) look at optimality of issuer call exercise decisions, Mattia Landoni (2018) investigates the structuring of call options on municipal bonds, and Luby and Singla (2020) look into derivatives use by municipal issuers. The rapid growth and decline of the municipal derivatives market, which was tied mostly to synthetic-fixed structures involving floating-rate bond issuance and interest rate swap contracts, deserves attention. The highest-profile municipal bankruptcy tied to a prominent synthetic-fixed issuer was the Jefferson County bankruptcy in 2008, the immediate catalyst for which was the divergence between yields on the county’s variable rate debt and swap contracts tied to the LIBOR reference rate.17 In the Jefferson County case these swap contracts were, at least before the cost of subsequent fines and penalties, lucrative for employees of the banks that arranged them. Official Statements for municipal bonds reveal data on the spreads paid to underwriters, but a synthetic fixed structure can embed spread in a less transparent way into the terms of the swap contract. In the Jefferson County case both municipal officials and bankers went to prison for bribes paid by underwriters to municipal officials that were connected to that issuance. Most municipal issuance is either fixed-rate or synthetic-fixed rate in form. Brooks (2005) argues that the optimal interest rate exposure of municipal market bond debt should depend on the interest risk profile of municipal revenues and costs, and he argues that for many issuers adding unhedged variable-rate debt would be appropriate from a standpoint of balancing revenue and expense risk profiles. It is likely, however, that accounting for pension obligations, which are a long-dated fixed commitment imperfectly reflected on municipal balance sheets, would push in the other directly, possibly in some cases entirely overwhelming any interest rate sensitivities of non-pension revenues and costs. Returning to corruption, Butler et al. (2009) provide larger-sample empirical evidence on public corruption and municipal finance. Using data from 1990 to 2004, they show that statelevel public corruption convictions and the quality of state-level anti-corruption laws are associated with yields on municipal debt; being in the top quartile of their corruption measure is associated with yields that are higher by about seven basis points. This affect is attenuated by the presence of third-party insurance, and places that score higher on corruption measures appear more likely to purchase third-party insurance when they issue debt. Local media play

16 On the different horizons and performance of elected and appointed municipal issuance officials; Whalley (2013), using a regression-discontinuity approach, finds that the yields of bonds issued by municipalities with appointed treasurers are much lower than the yields of bonds issued by municipalities with elected treasurers, and he argues that this observed difference reflects a premium for professional management. 17 See Bergstresser and Cohen (2012) for a case study on the Jefferson County debt issuance and bankruptcy.

The municipal bond market  323

a role in ensuring public accountability, and recent work by Gao et al. (2020) shows that after local newspaper closures municipal borrowing costs increase by as much as 11 basis points. Like corruption, political polarization can handicap the effectiveness of governments, and Li et al. (2018) show that in the period between 2000 and 2014 states whose legislatures were more politically polarized, as measured by the average ideological distance between median Democratic and Republican legislators, paid higher yields when they borrowed in the municipal bond market. Bergstresser et al. (2013) show that areas in the United States that are more fragmented in terms of ethnic and religious divisions pay more to borrow but do not appear to have higher realized default rates than other otherwise-similar areas. An important area of municipal research focuses on different aspects of municipal pension obligations. Municipalities enjoy more discretion in reporting the value of the pension liabilities than other pension sponsors. The nature of these liabilities as fixed commitments – particularly in places where the commitments enjoy statutory or constitutional protections – means that using discount rates that are associated with risky assets to value these liabilities may understate their true value. Brown and Wilcox (2009), and Novy-Marx and Rauh (2009; 2011) illustrate how the size of these pension obligations is similar in magnitude to the value of explicit bond-market borrowing, and Novy-Marx and Rauh (2012) demonstrate that increases in underfunding can increase state’s borrowing costs in the municipal bond market. Boyer (2020) addresses the uncertainty about and differences across states in the relative positions that pensioners and debtholders might have the context of state-level default. He finds that the impact of increases in unfunded pension liabilities on state borrowing costs is higher in states where pension commitments enjoy higher levels of legal protection. In addition to this research on pension liabilities, research on public pensions has also looked into underperformance of pension assets, and there is evidence that political interference can lead to lower returns. Andonov et al. (2018) show that public pensions with state officials on boards underperform in their private equity investments, and that underperformance appears driven by in-state investments and underperformance is higher when these board officials receive more political contributions from the finance industry. Bradley et al. (2016) document similar home bias and underperformance in the public equity holdings of state pension plans. Andonov et al. (2017) link the liability side with the asset side, showing that U.S. public pension funds, which (amazingly) have the accounting leeway to discount pension liabilities with a discount rate based on the expected returns of plan assets, appear to act on the incentives that accounting leeway creates by tilting asset portfolios toward riskier assets. Default by states has been unusual but is not unprecedented, and English (1996) looks at the experiences of American states that defaulted during the 1840s, after a period during which states’ debt grew rapidly. Debt in northern states in that era was taken on largely in order to fund transportation improvements, and debt in southern states was mostly used to finance state banks. Much of that infrastructure underperformed financially; as Fernald (1999) points out, the return to investors in a second canal or railroad linking two points will frequently be lower than the return to the first one. English shows that many of the states that defaulted on their debt in this era eventually resumed debt payments, and he makes the argument that a desire to maintain access to capital markets was behind the resumption of payments, and that this desire can lead borrowers to resume payments after default even in the absence of mechanisms other than reputation for insuring repayment. While defaults by states have been unusual, defaults by cities and other municipal entities have been much larger in number. Cohen (1989) looks at patterns of default over time and

324  Research handbook of financial markets

identifies several waves of municipal defaults, including a wave related to large-city municipal distress in the 1970s, including a default by New York City, and a wave related to the underperformance of various revenue-backed projects after the 1970s. Gorina et al. (2018) broaden the measure of fiscal distress, looking at measures of distress such as layoffs and service cuts. They demonstrate the predictive power of fiscal reserves and revenue mix for predicting distress with the broader distress measure. The role of rating agencies is to offer opinions regarding default, and there is some work on the performance of these agencies. Prior to the 2008 financial crisis municipal default rates were generally lower for a given alpha-numeric rating than default rates on the corporate bond rating scale, which is the reason for the Moody’s recalibration that has been relevant for some of the research discussed. Cornaggia et al. (2017) uses this recalibration to show that investors do appear to respond to ratings, even to credit-information-free upgrades of the sort that occurred in 2010. Cornaggia et al. (2020) look at the individual identities of analysts at credit ratings agencies, following their ratings over time, and present evidence of “home bias” in ratings, with analysts rating issuers from their home states higher than otherwise equivalent issuers from other states. These home-bias-skewed ratings do appear to affect issuers’ debt capacity also, as in Cornaggia et al. (2017). Gao et al. (2019) show that state policy regarding bankruptcy matters for municipal bond yields. Some states allow unconditional access Chapter 9 bankruptcy protection, and these authors show that municipal bond yields are higher and show more cyclical variation and evidence of spread contagion based on other defaults in that subset. This finding of regulation-contingent spread contagion suggests that the finding of Kidwell and Trczinka (1982) of limited spread contagion after the New York City default in 1975 may not be a result that has global generality. There is a large body of work on municipal financial reporting. The Governmental Accounting Standards Board Statement 34 (GASB 34), issued in 1999 and with implementation required by 2002, affected municipal reporting and established the current rules that govern disclosure. Baber et  al. (2020) show that issuers behaved strategically around the implementation of GASB 34, with issuers that were likely to benefit from enhanced disclosure rules delaying issuance relative to issuers that were likely to struggle under the new standards. Baber and Gore (2008) showed that in the pre-GASB 34 period municipal issuers in states that mandated GAAP accounting for financial reporting borrow at lower yields than issuers from other states. Baber et  al. (2013) show that disclosure quality appears to be higher in municipalities that are better governed along other dimensions. Cuny (2016) shows that after the municipal bond insurer Ambac went bankrupt, issuers of debt that had been insured by Ambac increased their disclosure relative to other issuers. Gore et  al. (2004) show that in the bond-insurance era, disclosure and bond insurance could be substitutes. They compare municipalities in Pennsylvania, which did not require GAAP-compliant financial reports, and Michigan, a state that did, and they show that Pennsylvania issuers were much more likely to sell bonds with bond insurance. Recent work by Nakhmurina (2020) shows that the introduction of state-level policies to monitor local fiscal conditions is associated with better local financial reporting and higher turnover of incumbent politicians. Green et al. (2007b) is the seminal work on price dynamics in the market for new bond issues. These authors look at almost 200,000 trades on 12,000 bonds issued from 2000 to 2003; they find evidence of significant price dispersion in how much customers pay for newly issued municipal bonds, with retail investors appearing to pay a larger markup for purchasing

The municipal bond market  325

new bonds relative to other investors. On January 31, 2005 the MSRB began requiring dealers to report trades within fifteen minutes of their execution and Schultz (2012) extends the Green et al analysis to the period after the RTRS (“Real Time Reporting System”) period. Schultz finds that the dispersion in customer prices paid for newly issued bonds fell dramatically when real-time price dissemination began. Butler (2008) looks at individual underwriters and shows that during the 1997–2001 period, in a sample of taxable municipal bonds, “local” underwriters, defined as underwriters with ongoing local presence, sold bonds at lower underwriting cost and at lower yields than other underwriters. Issuers almost invariably use underwriters to place bonds, but they also hire financial advisors to work with them on bond structure and other advice. The Dodd-Frank legislature clarified advisors’ fiduciary obligations to issuers, but Moldogaziev and Luby (2016) show, perhaps contrary to the Butler “familiarity” result, that in situations where underwriters and advisors work together repeatedly, the quality of the services they provide for issuers appears to deteriorate. Garrett (2021) shows that the Dodd-Frank reform of prohibiting an advisor from simultaneously serving as an underwriter appears to have reduced financing costs for the set of school districts that were plausibly particularly exposed to that pre-reform conflict of interest. Once they are issued, municipal bonds trade now in over-the-counter (OTC) markets, but this has not always been the case. Biais and Green (2019) show that until the late 1920s municipal bond trading was concentrated on the NYSE, as was corporate bond trading into the 1940s. Biais and Green argue that exchange-based and OTC-based trading environments can each be stable equilibria, but the movement of institutional trading from NYSE to OTC markets shifted the trading equilibrium away from exchanges. The authors show that transactions costs for trading municipal bonds in the late 1920s was low relative to the modern era, until municipal liquidity on the NYSE collapsed at the end of the 1920s. Harris and Piwowar (2006) use a year of trade data (May 2000 to July 2001) to estimate transactions costs in the secondary market for municipal bonds. This paper is similar to a paper by Green et al. (2007a), except that Harris and Piwowar estimate an econometric model where every trade is its own observation, and Green et al attempt to combine sales to dealers and purchases from dealers into round-trip transactions. Both papers come up with similar estimates of round-trip transactions costs; nearly two percent per transaction for smaller transactions and much lower (but also highly variable) for larger transactions. Chalmers et al. (2021) show that the introduction of real-time transaction reporting appears to have reduced transactions costs in the secondary market as well – a result that contrasts somewhat with Schultz’ finding that RTRS does not appear to have reduced the dispersion of primary-market transaction costs. Cuny (2018) shows the improved disclosure appears to reduce the relative transaction-cost disadvantage of smaller traders in the municipal market. Comparing the markup across transaction size across the implementation of enhanced MSRB dissemination of issuer information. Li and Schuerhoff (2019) explore the network structure of the municipal bond dealer network. They show that while there are more than 2,000 broker-dealer firms on the periphery of the municipal over-the-counter network, trading is dominated by between ten and 30 central dealers that have an extremely dense network of trading connections with each other. Looking at round-trip transactions (sales to dealers, followed by trading chains ending in a sale back from a dealer to a customer), the authors use evidence on pre-arranged trades (trades where a purchase from a customer occurs at almost exactly the same time as the dealer sells the bond

326  Research handbook of financial markets

back to a different customer, suggesting that the customer has had to wait for the dealer to find a customer willing to purchase the bonds) to show that customers face a choice between speed and cost, with more central dealers providing faster execution at a higher explicit cost. Two emerging areas of municipal research deserve particular attention. One area is climate change and the municipal bond market. Recent work (Painter (2020) and Goldsmith-Pinkham et  al. (2021) has shown that municipal bond yields appear somewhat sensitive to exposure to sea level rise. Research in this area has come to different conclusions on the timing and magnitude of this effect, with Painter finding a larger effect than Goldsmith-Pinkham et al. Auh et al. (2022) look at the response of municipal bond yields to natural disaster events and find an immediate increase in yields, which is somewhat attenuated in the presence of federal support. Bourdeau-Brien and Kryzanowski (2019) finds that the increase in municipal yields after floods is largest for counties that are experiencing disaster for the first time and this yield increase appears to revert overt time. A second area deserving particular attention is research on the impact of the Covid pandemic and the responses to that pandemic on the municipal bond market. Although this paper is not merely about the Covid pandemic, Gordon and Guerron-Quintana (2021) develop a model of regional migration and borrowing that shows that migration and place-specific borrowing create externalities that increase the likelihood of municipal default. This model suggests that low market interest rates have ameliorated what would otherwise been harmful effects of Covid-induced migration on some affected cities’ finances. Gustafson et al. (2022) also find that the Covid-induced migration shocks of 2020 appear to have affected bond yields, with areas more affected by Covid-induced emigration seeing relative increases in their bond yields. Other papers have looked at the impact of the extraordinary interventions in the municipal bond markets during the early part of the Covid pandemic. The Municipal Liquidity Facility (MLF) program provided a support backstop for eligible municipal issuers giving them the option to place new borrowing directly with a new Federal Reserve-sponsored facility. Johnson et al. (2021) describe this program in detail, and Haughwout et al. (2021) use discontinuities induced by the cutoffs in eligibility for this program to estimate its effect on bond yields and state and local employment. These authors find some interesting puzzles: the impact of the MLF program on yields appears to have been greatest among low-rated issuers, but the overall impact of the government interventions during the Covid pandemic on government hiring appears to have been strongest among higher-rated municipalities.

14.7 CONCLUSION The municipal bond market plays an essential role in the financing of infrastructure in the United States. Relative to other bond markets the municipal market is particularly fragmented and complex, with an extremely large number of issuers, many of them quite small. Individual household investors play an unusually large role in this market as well. A growing body of high-quality research has engaged successfully with this rich and complex institutional detail and is developing insights about this market that is useful for both practitioners and policymakers.

The municipal bond market  327

REFERENCES Adelino, M., Cunha, I., & Ferreira, M. A. (2017). The economic effects of public financing: Evidence from municipal bond ratings recalibration. Review of Financial Studies, 30(9), 3223–3268. Andonov, A., Bauer, R. M. M. J., & Cremers, K. J. M. (2017). Pension fund asset allocation and liability discount rates. Review of Financial Studies, 30(8), 2555–2595. Andonov, A., Hochberg, Y. V., & Rauh, J. D. (2018). Political representation and governance: Evidence from the investment decisions of public pension funds. Journal of Finance, 73(5), 2041–2086. Ang, A., Green, R., Longstaff, F., & Xing, Y. (2017). Advance refundings of municipal bonds. Journal of Finance, 72(4), 1645–1682. Aschauer, D. A. (1989). Is public expenditure productive? Journal of Monetary Economics, 23(2), 177–200. Auh, J. K., Choi, Jaewon, Derygina, T., & Park, T. (2022). Natural disasters and municipal bonds, working paper. Baber, W., Beck, A., & Koester, A. (2020). Separation in the municipal debt market following GASB 34 implementation, working paper. Baber, W., & Gore, A. (2008). Consequences of GAAP disclosure regulation: Evidence from municipal debt issues. Accounting Review, 83(3), 565–592. Baber, W., Gore, A., Rich, K., & Zhang, J. (2013). Accounting restatements, governance, and municipal debt financing. Journal of Accounting and Economics, 56(2–3), 212–227. Baker, M., Bergstresser, D., Serafeim, G., & Wurgler, J. (2022). The pricing and ownership of U.S. green bonds. Annual Review of Financial Economics [in press]. Bergstresser, D., & Cohen, R. (2012). Jefferson County (A): An EPA mandate. Harvard Business School Case, 213–056. Bergstresser, D., & Cohen, R. (2015). Changing patterns in household ownership of municipal debt: Evidence from the 1989–2013 Surveys of Consumer Finances, working paper. Bergstresser, D., Cohen, R., & Shenai, S. (2013). Demographic fractionalization and the municipal bond market. Municipal Finance Journal, 34(3), 1–38. Bergstresser, D., & Orr, P. (2014). Direct bank investment in municipal debt. Municipal Finance Journal, 35(1), 1–23. Biais, B., & Green, R. C. (2019). The microstructure of the bond market in the 20th century. Review of Economic Dynamics, 33, 250–271. Bourdeau-Brien, M., & Kryzanowski, L. (2019). Municipal financing costs following disasters. Global Finance Journal, 40, 48–64. Boyer, C. (2020). Public pensions and state borrowing costs, working paper. Bradley, D., Pantzalis, C., & Yuan, X. (2016). The influence of political bias in state pension funds. Journal of Financial Economics, 119(1), 69–91. Brooks, L., & Liscow, Z. (2020). Infrastructure costs, working paper. Brooks, R. (2005). A surplus optimization approach to managing municipal debt. Public Finance Review, 33(2), 236–254. Brown, J., & Wilcox, D. (2009). Discounting state and local pension liabilities. American Economic Review: Papers and Proceedings, 99(2), 538–542. Butler, A. W. (2008). Distance still matters: Evidence from municipal bond underwriting. Review of Financial Studies, 21(2), 763–784. Butler, A. W., Fauver, L., & Mortal, S. (2009). Corruption, political connections, and municipal finance. Review of Financial Studies, 22(2), 2873–2905. Cestau, D., Green, R., Hollifield, B., & Schuerhoff, N. (2020). Should governments prohibit negotiated sales of municipal bonds? working paper. Cestau, D., Hollifield, B., Li, D., & Schuerhoff, N. (2019). Municipal bond markets. Annual Review of Financial Economics, 11(1), 65–84. Chalmers, J. M. R. (1998). Default risk cannot explain the muni puzzle: Evidence from municipal bonds that are secured by U.S. Treasury obligations. Review of Financial Studies, 11(2), 281–308. Chalmers, J. M. R. (2006). Systematic risk and the muni puzzle. National Tax Journal, 59(4), 833–848.

328  Research handbook of financial markets

Chalmers, J. M. R., Liu, Y. S., & Wang, Z. J. (2021). The difference a day makes: Timely disclosure and trading efficiency in the muni market. Journal of Financial Economics, 139(1), 313–335. Chen, H., Cohen, L., & Liu, W. (2021). Calling all issuers: The market for debt monitoring, working paper. Cohen, N. R. (1989). Municipal default patterns: An historical study. Public Budgeting and Finance, 9(4), 55–65. Cornaggia, J., Cornaggia, K., & Israelsen, R. (2018). Credit ratings and the cost of municipal financing. Review of Financial Studies, 31(6), 2038–2079. Cornaggia, J., Cornaggia, K., & Israelsen, R. (2020). Where the heart is: Information production and the home bias. Management Science, 66(12), 5532–5557. Cuny, C. (2016). Voluntary disclosure incentives: Evidence from the municipal bond market. Journal of Accounting and Economics, 62(1), 87–102. Cuny, C. (2018). When knowledge is power: Evidence from the municipal bond market. Journal of Accounting and Economics, 65(1), 109–128. Cutler, D., & Miller, G. (2005). The role of public health improvements in health advances: The twentieth-century United States. Demography, 42(1), 1–22. Cutler, D., & Miller, G. (2006). Water, water everywhere: Municipal finance and water supply in American cities. In E. L. Glaeser & C. Goldin (eds.), Corruption and reform: Lessons from America’s economic history (pp. 153–183). Chicago: University of Chicago Press. Dagostino, R. (2019). The impact of bank financing on municipalities’ bond issuance and the real economy, working paper. English, W. (1996). Understanding the costs of sovereign default: U.S. state debts in the 1840s. American Economic Review, 86(1), 259–275. Fernald, J. G. (1999). Roads to prosperity? Assessing the link between public capital and productivity. American Economic Review, 89(3), 619–638. Fortune, P. (1998). Tax-exempt bonds really do subsidize municipal capital! National Tax Journal, 51(1), 43–54. Galper, H., Rosenberg, J., Rueben, K., & Toder, E. (2013). Who benefits from tax-exempt bonds? An application of the theory of tax incidence, Working paper, Urban-Brookings Tax Policy Center, Washington, DC. Galper, H., Rueben, K., Auxier, R., & Eng, A. (2014). Municipal debt. What does it buy and who benefits? National Tax Journal, 67(4), 901–924. Gao, P., Lee, C., & Murphy, D. (2019). Municipal borrowing costs and state policies for distressed municipalities. Journal of Financial Economics, 132(2), 404–426. Gao, P., Lee, C., & Murphy, D. (2020). Financing dies in darkness? The impact of newspaper closures on public finance. Journal of Financial Economics, 135(2), 445–467. Garrett, D. (2021). Conflicts of interest in municipal bond advising and underwriting, working paper. Garrett, D., Ordin, A., Roberts, J. W., & Suárez, J. C. (2017). Tax advantages and imperfect competition in auctions for municipal bonds. NBER Working paper #23473. Gillette, J., Samuels, D., & Zhou, F. (2020). The effect of credit ratings on disclosure: Evidence from the recalibration of Moody’s municipal ratings. Journal of Accounting Research, 58(3), 693–739. Goldsmith-Pinkham, P., Gustafson, M. T., Lewis, R. C., & Schwert, M. (2021). Sea level rise exposure and municipal bond yields, working paper. Gordon, G., & Guerron-Quintana, P. (2021). Public debt, private pain: Regional borrowing, default, and migration, working paper. Gordon, R. H., & Metcalf, G. E. (1991). Do tax-exempt bonds really subsidize municipal capital? National Tax Journal, 44(4), 71–79. Gore, A. K., Sachs, K., & Trzcinka, C. (2004). Financial disclosure and bond insurance. Journal of Law and Economics, 47(1), 275–306. Gorina, E., Maher, C., & Joffe, M. (2018). Local fiscal distress: Measurement and prediction. Public Budgeting and Finance, 38(1), 72–94. Gramlich, E. M. (1994). Infrastructure investment: A review essay. Journal of Economic Literature, 32(3), 1176–1196. Green, R. C. (1993). A simple model of the taxable and tax-exempt yield curves. Review of Financial Studies, 6(2), 233–264.

The municipal bond market  329

Green, R. C., Hollifield, B., & Schuerhoff, N. (2007a). Financial intermediation and the costs of trading in an opaque market. Review of Financial Studies, 20(2), 275–314. Green, R. C., Hollifield, B., & Schuerhoff, N. (2007b). Dealer intermediation and price behavior in the aftermarket for new issues. Journal of Financial Economics, 86(3), 643–682. Gupta, A., van Nieuwerburgh, S., & Kontokosta, C. (2020). Take the Q train: Value capture of public infrastructure projects. NBER Working paper #26789. Gustafson, M., Haslag, P., Weagley, D., & Zihan, Y. (2022). COVID-induced migration and municipal bonds, working paper. Harris, L. E., & Piwowar, M. S. (2006). Secondary trading costs in the municipal bond market. Journal of Finance, 61(3), 1361–1397. He, Z., & Milbradt, K. (2014). Endogenous liquidity and defaultable bonds. Econometrica, 82(4), 1443–1508. Hildreth, W. B., & Zorn, C. K. (2005). The evolution of the state and local government municipal debt market over the past quarter century. Public Budgeting and Finance, 25(4s), 127–153. Ivanov, I., & Zimmermann, T. (2021). The ‘privatization’ of municipal debt, working paper. Johnson, C. L., Moldogaziev, T. T., Luby, M. J., & Winecoff, R. (2021). The Federal Reserve Municipal Liquidity Facility (MLF): Where the municipal securities market and Fed finally meet. Public Budgeting and Finance, 41(3), 42–73. Kalotay, A., & Buursma, J. (2019). The key rate durations of municipal bonds. Journal of Fixed Income, 29(2), 61–64. Kidwell, D. S., & Koch, T. W. (1983). Market segmentation and the term structure of municipal yields. Journal of Money, Credit, and Banking, 18(1), 482–494. Kidwell, D. S., & Trzinka, C. A. (1982). Municipal bond pricing and the New York City fiscal crisis. Journal of Finance, 37(5), 1239–1246. Landoni, M. (2018). Tax distortions and bond issue pricing. Journal of Financial Economics, 129(2), 382–393. Larcker, D. F., & Watts, E. M. (2020). Where’s the greenium. Journal of Accounting and Economics, 69(2–3). Li, D., & Schuerhoff, N. (2019). Dealer networks. Journal of Finance, 74(1), 91–144. Li, P., Tang, L., & Cloyd, B. (2018). The effect of political polarization on state government bonds, working paper. Longstaff, F. (2011). Municipal debt and marginal tax rates: Is there a tax premium in asset prices? Journal of Finance, 66(3), 721–751. Longstaff, F., Mithal, S., & Neis, E. (2005). Corporate yield spreads: Default risk or liquidity? New evidence from the credit default swap market. Journal of Finance, 60(5), 2213–2253. Luby, M. J., & Singla, A. (2020). Financial engineering by city governments: Factors associated with the use of debt-related derivatives. Urban Affairs Review, 56(3), 857–887. Moldogaziev, T. T., & Luby, M. J. (2016). Too close for comfort: Does the intensity of municipal advisor and underwriter relationship impact borrowing costs? Public Budgeting and Finance, 36(3), 69–93. Nakhmurina, A. (2020). Does fiscal monitoring make better governance? Evidence from US municipalities, working paper. Nanda, V., & Singh, R. (2005). Bond insurance: What is special about munis? Journal of Finance, 59(5), 2253–2280. Novy-Marx, R., & Rauh, J. (2009). The liabilities and risks of state-sponsored pension plans. Journal of Economic Perspectives, 23(4), 191–210. Novy-Marx, R., & Rauh, J. (2011). Public pension promises: How big are they are what are they worth? Journal of Finance, 66(4), 1211–1249. Novy-Marx, R., & Rauh, J. (2012). Fiscal imbalances and borrowing costs: Evidence from state investment losses. American Economic Journal: Economic Policy, 4(2), 182–213. Painter, M. (2020). An inconvenient cost: The effects of climate change on municipal bonds. Journal of Financial Economics, 135(2), 468–482. Perignon, C., & Vallee, B. (2017). The political economy of financial innovation: Evidence from local governments. Review of Financial Studies, 30(6), 1903–1934. Poterba, J. M. (1989). Tax reform and the market for tax-exempt debt. Regional Science and Urban Economics, 19(3), 537–562.

330  Research handbook of financial markets

Poterba, J. M., & Ramírez Verdugo, A. (2011). Portfolio substitution and the revenue cost of the federal income tax exemption for state and local government bonds. National Tax Journal, 64(2), 591–614. Redding, S. J., & Turner, M. A. (2015). Transportation costs and the spatial organization of economic activity. Handbook of Regional and Urban Economics, 5, 1339–1398. Schultz, P. (2012). The market for new issues of municipal bonds: The roles of transparency and limited access to retail investors. Journal of Financial Economics, 106(3), 492–512. Schwert, M. (2017). Municipal bond liquidity and default risk. Journal of Finance, 72(4), 1683–1722. Simonsen, W., & Robbins, M. (1996). Does it make any difference anymore? Competitive versus negotiated municipal bond issuance. Public Administration Review, 56(1), 57–64. Wang, J., Wu, C., & Zhang, F. X. (2008). Liquidity, default, taxes, and yields on municipal bonds. Journal of Banking and Finance, 32(6), 1133–1149. Whalley, A. (2013). Elected versus appointed policy makers: Evidence from city treasurers. Journal of Law and Economics, 56(1), 39–81.

15. Mortgage-backed securities1 Andreas Fuster, David Lucca and James Vickery

15.1 INTRODUCTION1 The mortgage-backed securities (MBS) market emerged as a way to decouple mortgage lending from mortgage investing. Until the 1980s, nearly all US mortgages were held on balance sheet by financial intermediaries, predominately savings and loans. Securitization today allows these mortgages to be held and traded by investors all over the world, and the US MBS market is one of the largest and most liquid global fixedincome markets, with more than $11 trillion of securities outstanding and nearly $300 billion in average daily trading volume.2 MBS and a related instrument, covered bonds, are also used for funding mortgages in many European countries as well as some other parts of the world. This chapter presents an overview of the MBS market, including the institutional environment, security design, MBS risks and asset pricing, as well as the economic effects of mortgage securitization. It also assembles descriptive statistics about the MBS market, including market size, growth, security characteristics, prepayment, trading activity and yield spreads. We particularly focus on the large agency residential MBS market in the United States, but also discuss other types of MBS both in the US and around the world. We also consider the role of the Federal Reserve through its quantitative easing program. Throughout, we highlight insights from the growing literature on MBS and mortgage securitization, a body of research catalyzed by the role of MBS markets in the 2008 financial crisis. We also highlight topics for future research. This chapter can only scratch the surface of such a complex topic, however, and for other surveys and further information, we refer the reader to Hayre (2001), McConnell and Buser (2011), Green (2013), Davidson and Levin (2014), Fabozzi (2016) and Kim et al. (2022).

1 We thank the editors, our discussant Benson Durham and Eric Horton, You Suk Kim, Haoyang Liu, Dean Parker, Shane Sherlund and Andreas Strzodka for helpful comments and suggestions. Natalie Newton provided outstanding research assistance. This research was completed while Lucca was an economist at the Federal Reserve Bank of New York. The views expressed in this paper are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York, the Federal Reserve Bank of Philadelphia or the Federal Reserve System. 2  Source: SIFMA. Includes commercial and residential MBS pools and collateralized mortgage obligations. 331

332  Research handbook of financial markets

15.2 THE MBS UNIVERSE 15.2.1 MBS Market Segments and Their Evolution Over Time Mortgage-backed securities are bonds with cash flows tied to the principal and interest payments on a pool of underlying mortgages. Mortgage securitization has a long history (e.g. see Goetzmann & Newman, 2010; de Jong et al., 2022), but the birth of the modern US MBS market is typically dated to the issuance of the first agency MBS pool by Ginnie Mae in 1970. Figure 15.1 documents the enormous growth in the MBS market over the past half-century. The figure breaks down the market along two key dimensions. ●

●

Agency versus nonagency. Agency MBS carry a government-backed credit guarantee from one of three housing agencies: Fannie Mae, Freddie Mac or Ginnie Mae.3 Nonagency MBS, on the other hand, are issued by private financial institutions and are not guaranteed. Instead, securities are tranched in terms of seniority to cater to investors with different credit risk appetites. Residential versus commercial. The bulk of MBS are backed by mortgages on individual residential properties (RMBS). But there is also an active commercial MBS (CMBS) market secured by a diverse range of commercial real estate (e.g. office, multifamily, industrial, hotel and warehouse properties). Commercial mortgages are larger, more complex and more heterogenous than residential mortgages and these features are reflected in the design of CMBS, as we discuss in Section 15.3.

The top panel of Figure 15.1 plots the evolution of the volume of residential MBS. The market began to expand significantly in the early 1980s, driven at the time by high and volatile interest rates and the need to alleviate the maturity and liquidity mismatch faced by savings and loans. Regulatory and tax incentives also played an important role.4 The RMBS market continued to expand rapidly over the following two decades, with the volume of securities outstanding reaching almost 50% of GDP by the time of the Great Recession. Particularly notable is the rapid growth in nonagency MBS during the 2000s due to the issuance of subprime and “alt-A” MBS backed by mortgages with high levels of credit risk.5 3 Fannie Mae and Freddie Mac purchase and securitize “conforming” mortgages, which are typically prime-quality loans. They are not permitted to purchase large jumbo mortgages above the conforming loan limits or mortgages with a loan-to-value (LTV) ratio exceeding 80% unless the loan carries mortgage insurance. Fannie Mae and Freddie Mac are “government-sponsored enterprises” or GSEs. Although not explicitly government-owned, their debt is perceived to carry an implicit public guarantee, and the two GSEs have been in public conservatorship since 2008 (Frame et al., 2015). Ginnie Mae guarantees MBS assembled from mortgages explicitly insured by federal government agencies, primarily the Federal Housing Administration (FHA) and Veterans Affairs (VA). See Burgess et al. (2023) in this volume for more details on these agencies. 4 See the finance classic Liar’s Poker (Lewis, 2010) for a lively account of the MBS market during this period. 5 This included loans to borrowers with low credit scores, interest-only and negative amortization mortgages and loans with incomplete documentation of borrower income and assets. The growth and collapse of the nonagency mortgage market during the 2000s is of course the subject of a very large literature. See, e.g., Gerardi et  al., 2008; Ashcraft and Schuermann, 2008; Mian and Sufi, 2015; Adelino et al.,2016; Foote et al., 2020; Adelino et al., 2020 and references cited therein.

Mortgage-backed securities  50

333

100 agency RMBS, % nominal GDP (LHS) non-agency RMBS, % nominal GDP (LHS) RMBS, % residential mtg debt (RHS)

80

30

60

20

40

10

20

0

0 1970

1980

1990

2000

2010

% of residential mortgage debt

% nominal GDP

40

2020 30

6 agency CMBS, % nominal GDP (LHS) non-agency CMBS, % nominal GDP (LHS)

4

20

2

10

0

0 1970

1980

1990

2000

2010

% of commercial mortgage debt

% of nominal GDP

CMBS, % commercial mtg debt (RHS)

2020

Note:  Shaded areas represent stock of agency MBS (light grey) and nonagency MBS (dark grey) as a percent of nominal GDP. Solid line plots total MBS scaled by the relevant stock of mortgage debt. See Appendix (Section A) for details of figure construction. Source:   Financial Accounts of the United States, BEA.

Figure 15.1  Mortgage-backed securities outstanding

334  Research handbook of financial markets

The housing crash and wave of mortgage defaults that precipitated the Great Recession also caused a freeze in the issuance of nonagency MBS in mid-2007 (Calem et al., 2013; Vickery & Wright, 2013; Fuster & Vickery, 2015; Kruger, 2018). The market has partially recovered but nonagency MBS issuance remains far below pre-crisis levels even to the present day. The MBS market as a whole remains very active, however—as of 2021, 65% of total home mortgage debt is securitized into MBS, up from 60% a decade ago, nearly all of it in the form of agency MBS. The stock of MBS as a percent of nominal GDP is smaller than before the Great Recession though, reflecting the post-crisis normalization of household leverage. The bottom panel of Figure 15.1 focuses on CMBS. The CMBS market is smaller than the residential market, and did not grow in earnest until the 1990s, fueled by the Resolution Trust Corporation, which issued securities backed by distressed commercial real estate in the wake of the savings and loan crisis (An et al., 2009; Chandan, 2012). Like its residential cousin, the nonagency CMBS market experienced an extraordinary boom during the 2000s—almost tripling in size as a percentage of GDP—before CMBS issuance ground to a halt at the start of the Great Recession. The market has returned to health in the post-crisis period, but normalized by GDP, the volume of nonagency CMBS today is only at the level of the early 2000s. In contrast, the agency CMBS market has grown rapidly since the Great Recession. These securities have an explicit or implicit government credit guarantee, like agency RMBS, and are typically backed by mortgages on apartment buildings or another form of multifamily housing (e.g. senior housing, student housing.6 Competition and substitution between government-backed and private securitization is a central feature of both the commercial and residential markets. For instance, Adelino et al. (2020) show that high-LTV lending migrated almost entirely from the FHA and VA to the private subprime market during the 2000s boom, before shifting back during and after the Great Recession. The market price of credit risk is a key driver of these substitution effects, because credit guarantee fees on agency MBS are set administratively and do not reflect market prices. For example, in recent years the rise in the guarantee fees charged by the GSEs has driven some lending from the agency to the nonagency market and has also induced banks to retain low-risk mortgages on balance sheet rather than securitizing them. Regulatory policies also shape the relative size of the agency and nonagency market and likely contributed to the slow post-crisis recovery of nonagency RMBS, including (i) the introduction of higher conforming loan limits in counties with high home prices, which expanded the footprint of the GSEs (Vickery & Wright, 2013); (ii) Dodd-Frank Act risk retention requirements and constraints on household leverage through the “Ability-to-Repay” (ATR) rule (DeFusco et al., 2019)7 and (iii) bank liquidity regulations that favor agency securities (Roberts et al., 2018).

6 See Credit Suisse (2011) for a primer on the agency CMBS market. 7 The Dodd-Frank ATR rule requires lenders to ensure the borrower can repay the loan before extending credit. Lenders can satisfy the ATR requirement by originating “qualified mortgages” (QMs) that satisfy particular criteria (these criteria have recently changed, but previously a key restriction was that the debt-to-income ratio could not exceed 43%). But a rule known as the “QM patch” designated all mortgages sold to the GSEs as qualified mortgages, regardless of their characteristics. The ATR rule combined with this GSE carve-out likely constrained the volume of risky mortgages securitized through the nonagency market. The QM patch expired in 2021.

Mortgage-backed securities 

335

15.2.2 Investors Who invests in MBS? The Financial Accounts of the United States provide a partial answer by tabulating investors in agency and GSE-issued securities, a category which mainly comprises agency MBS. As of mid-2021, depository institutions are the largest class of investors (32% of the total), followed by the Federal Reserve (23%), international investors (11%), mutual funds (7%) and money market funds (5%). A full tabulation is provided in the Appendix. More details on bank MBS holdings are presented in the Federal Reserve Bank of New York (2021). Banks invest very heavily in agency MBS, which account for about half of banks’ total investment security holdings. Banks are also significant investors in agency collateralized mortgage obligations (CMOs), and to a lesser extent, nonagency MBS. The Federal Reserve is the single largest agency MBS investor through its large-scale asset purchase program, with total holdings of $2.5 trillion as of October 2021. Research has found that Fed MBS purchases reduce MBS yields and have a range of other effects on financial markets and the macroeconomy (see Section 15.4). 15.2.3 Agency RMBS in the Cross-Section Table 15.1 provides a cross-sectional snapshot of the population of agency residential MBS pools based on security-level data from eMBS as of March 2021. We divide the universe by agency—Fannie Mae, Freddie Mac or Ginnie Mae. Within the Ginnie Mae population, we separately break out multi-issuer pools, because they are much larger and more diversified than other Ginnie Mae pools and make up the bulk of new Ginnie issuance (see also Section 15.3). Averages and distributional statistics reported in the table are weighted by outstanding pool balance. The population consists of just over a million individual MBS pools, which together comprise $7.7 trillion of home mortgage debt. Almost all of this debt consists of fixed-rate mortgages (FRMs), mainly in the form of 30-year FRMs ($6.5 trillion of the total). For Fannie, Freddie, and Ginnie multi-issuer pools, around 95% of the pool balances are deliverable in the “to-be-announced” (TBA) market, which is the primary venue for agency MBS trading (see Section 15.5). Strikingly, 44% of the outstanding balance reflects pools with an age of less than a year.8 This is an unusually high percentage, due to a record refinancing wave and home price boom in 2020 that resulted in around $4 trillion in mortgage originations (Fuster et al., 2021). Even so, nearly a quarter of the total unpaid balance (UPB) comprises pools with an age exceeding five years. This diversity of vintages is also evident in the distribution of coupons (the rate of interest paid to investors). About 45% of the universe consists of MBS pools with a coupon of 2.5% or lower—these are the typical coupons into which new mortgages would be securitized, reflecting recent record-low mortgage rates. But there is still a substantial population of much higher coupons, with 18% of the total unpaid balance reflecting coupons of 4% or higher. Borrowers represented in these pools would almost surely benefit substantially from refinancing, but for one reason or another have failed to do so (see, e.g., Keys et al., 2016 for discussion.) 8 The pool age closely corresponds to the age of the underlying loans, since agency mortgages are typically securitized shortly after origination. See Section 15.3 for more detail on the securitization process.

336  Research handbook of financial markets

Table 15.1  The cross-section of agency MBS pools Fannie Mae Freddie Mac

Ginnie Mae Multi

Number of active pools

474,062

274,588

Total

Other

8,547 246,025 1,003,222

Aggregate outstanding face value (UPB, in billions) 30yr FRM

2,590.1

1,902.8

1,575.8

386.3

6,455.1

15yr FRM

450.8

339.8

24.5

3.6

818.7

Other FRM

204.2

127.9

0.0

0.3

332.4

34.4

26.7

10.6

46.6

118.3

3,279.5

2,397.2

1,610.9

436.9

7,724.5

94.1

94.1

96.2

21.3

90.5

Other mortgage types Total TBA eligible (%, weighted by UPB) Distribution of pool UPB 10th pctile

4.3

6.2

754.2

1.0

5.0

50th pctile

165.0

287.8

6,226.6

6.2

353.3

90th pctile

26,654.0

6,956.9

34,823.1

36.8

19,598.8

95th pctile

35,199.7

8,932.9

40,937.4

58.6

34,823.1

99th pctile

41,203.9

12,348.9

43,895.2

144.2

41,203.9

Less than 2%

0.05

0.09

0.01

0.03

0.05

2–2.5

0.24

0.27

0.16

0.07

0.22

Distribution by coupon (weighted by UPB)

2.5–3

0.18

0.17

0.20

0.18

0.18

3–3.5

0.20

0.18

0.25

0.14

0.20

3.5–4

0.15

0.13

0.22

0.19

0.16

4–4.5

0.11

0.09

0.10

0.17

0.10

Greater than 4.5%

0.07

0.07

0.06

0.22

0.08

Distribution of pool age (%, weighted by UPB) Less than 1yr

44.22

49.73

40.91

28.60

44.35

1–5yr

30.39

28.52

38.46

40.56

32.07

5–10yr

20.47

17.51

17.94

18.71

18.93

4.92

4.24

2.69

12.13

4.65

1st pctile

0.05

0.10

7.52

0.00

0.04

5th pctile

1.87

3.60

9.83

0.01

2.64

Greater than 10yr Distribution of prepayment speed (weighted by UPB)

(Continued)

Mortgage-backed securities 

337

Table 15.1  (Continued) Fannie Mae Freddie Mac

Ginnie Mae

Total

Multi

Other

25th pctile

13.90

13.24

27.86

3.84

14.82

50th pctile

24.88

23.91

40.34

23.22

27.80

75th pctile

36.26

35.63

47.23

38.81

40.03

95th pctile

50.87

48.93

55.88

66.92

52.57

99th pctile

66.09

61.90

59.98

87.32

66.10

Note:   Reflects the population of agency residential MBS pools measured as of March 2021. All averages and distributional statistics are weighted by outstanding pool unpaid balance (UPB). Source:   Author calculations based on eMBS security-level data.

Pool size also varies widely. The bottom 10% of the universe consists of pools with an outstanding balance of $5 million or below, while the top 10% has a balance exceeding $20 billion. This dispersion reflects differences in the original issue amount as well as the fact that many older pools have partially or almost completely paid down. Pool size is much larger for Ginnie Mae multi-issuer pools than for the other categories. Prepayment speed—a primary driver of security value—is also very heterogeneous across pools. The median three-month prepayment speed, measured by the conditional prepayment rate (CPR), is 27.8%, but the 5th and 95th percentiles are 2.6% and 52.6% respectively. Section 15.4.2 discusses the drivers of prepayments. To sum up, Table 15.1 shows that there is substantial heterogeneity and fragmentation within the agency MBS universe, which consists of more than a million unique individual pools. Even so, trading arrangements have evolved to facilitate a liquid, well-functioning secondary market, with trading concentrated in a small number of forward contracts, as we discuss in Section 15.5. 15.2.4 International MBS Markets Outside the US, securitization is also used as a form of secondary market mortgage finance around the world, including China, continental Europe, Canada, the United Kingdom and Australia. Some countries share features of the US mortgage finance system. For example, the Danish model is similar in many respects to agency securitization, as discussed by Berg et al. (2018). Mortgages in Denmark are originated by a small number of specialist mortgage banks, which then issue bonds with cash flows matching the borrowers’ payments. The mortgage bank retains the loan on balance sheet, however, and bears the credit risk if the borrower defaults. In this sense, Danish mortgage banks play a role similar to Fannie Mae and Freddie Mac. Another example is the Canadian model, which features a significant role for government guarantees, with the public sector insuring all mortgages with a down payment of less than 20%. There is an active market for securitizing these government-insured loans, with payments to investors guaranteed by the Canada Mortgage and Housing Corporation, a government agency also similar in some ways to Fannie Mae and Freddie Mac (Mordel & Stephens, 2015).

338  Research handbook of financial markets

In most other countries, however, securitizations more closely resemble nonagency MBS, with credit risk being borne by capital market investors.9 Standard and Poor’s (2021) provides an overview of market conditions for credit-sensitive MBS around the globe. Covered bonds are a distinct but related form of capital market mortgage financing, and are popular in many European countries.10 Covered bonds are debt instruments that finance a “cover pool” of ring-fenced assets. The bond investor has exclusive recourse to the asset pool in case of default, with further recourse to the issuer’s other assets if needed (see Berg et al., 2018). Unlike securitization, the cover pool is pledged as collateral for the bonds but remains on the issuer’s balance sheet, and mortgage prepayment and default therefore do not typically affect the payments to investors. The US does not have an active covered bond market, in part because banks have access to funding collateralized by mortgages through the Federal Home Loan Bank system (Bernanke, 2009).

15.3 SECURITY DESIGN Aside from the underlying mortgage type (residential or commercial) and the presence or absence of a government-backed credit guarantee, MBS also differ in terms of how cash flows from the mortgages are allocated to investors. The most straightforward MBS design is provided by so-called “pass-through” securities. All cash flows—including scheduled principal and interest, as well as prepayments—are paid to investors on a pro-rata basis, after first subtracting from the interest payments a fee to the loan servicer, and (in case of loans with a credit guarantee) the guarantee fee (“g-fee”). This is how residential agency MBS pools are structured.11 However, there are economic reasons to depart from the simple pass-through structure to appeal to investors with different needs and risk appetites. In the agency MBS segment, pools are resecuritized into collateralized mortgage obligations (or CMOs) to create tranches with different prepayment risk and duration. In a “sequential pay” structure, principal prepayments are first only disbursed to class A bondholders until these bonds are completely paid off; then paid to class B, and so on. This structure caters to investors with different maturity habitats—for example life insurers favor long-duration assets to match their policy liabilities. Also popular, “stripped” CMOs separate cash flows into interest-only (or IO) and principal-only (or PO) tranches. The universe also includes many other security types; see   9 For example, Standard and Poor’s (2020) provides a primer on the China RMBS market, which finances only a small share of Chinese mortgage debt but has grown rapidly in recent years. 10 For example, see Prokopczuk et al. (2013) for a discussion and analysis of the German “Pfandbrief” covered bond market and Meuli et al. (2021) for an analysis of the Swiss covered bond market, which shares features of the Federal Home Loan Bank system. Berg et al. (2018) report that the five largest covered bond markets are Denmark, Germany, France, Spain and Sweden. As we have discussed above, however, Danish covered bonds have the distinctive feature that market risk and prepayment risk are passed through to investors, making these bonds more similar to agency MBS than to other covered bonds. 11 For GSE pools, the servicer fee is 25 basis points (bp), while the periodic g-fee is around 40–50bp. (Fannie Mae and Freddie Mac only disclose average effective g-fees, which include flow equivalents of required upfront payments, so-called loan-level price adjustments. These average effective g-fees have hovered between 50 and 60bp in recent years; see Urban Institute, 2021, p. 26.) For Ginnie Mae-backed loans, the servicer fee is 44bp, while the guarantee fee is 6bp.

Mortgage-backed securities 

339

Arcidiacono et al. (2013) for an overview. As of 2021 there is $1.3 trillion in agency CMOs outstanding (source: SIFMA). In the nonagency market, structured securities are used to allocate credit risk across investors. Typically, nonagency CMOs follow a “senior-subordinated” structure, where principal payments are directed first to the senior tranches, at least during an initial “lockout” period, while lower-ranked “mezzanine” tranches initially receive only coupon payments.12 The lowest-ranked “equity” tranche is the first to absorb credit losses and therefore has the highest risk. In the CMBS market, the most junior tranches are typically retained by the “special servicer” (or “B-piece buyer”) who is also responsible for negotiating workouts for delinquent loans. B-piece buyers therefore have strong incentives to carefully assess the credit risk of the underlying loans before entering a deal, and are considered the gate-keepers in the CMBS market (Ashcraft et  al., 2019). Wong (2018), however, finds evidence that the dual role of B-piece buyers as both investor and servicer leads to conflicts of interest with senior bondholders during workouts. 15.3.1 Process of Securitization How are MBS actually produced? We here provide a brief overview for residential agency MBS pools, following Fuster et al. (2013). See Bhattacharya et al. (2008) for a broader discussion that also covers nonagency securities. The building blocks for any MBS are individual loans. These are typically newly-originated mortgages, but not always. The origination process begins with the borrower and originator agreeing on the terms (interest rate, contract type, and upfront payments including “points”). The loan originator could be a bank or a nonbank mortgage company, with nonbanks accounting for about two-thirds of originations in recent years (see Buchak et al., 2018, Kim et al., 2018 and Kim et al., 2022 for analysis of the growth of nonbank lending and its implications for systemic risk). The terms are guaranteed by the originator for a “lock period” of typically 30 to 90 days, during which time the borrower’s application is evaluated and processed—in particular, to ensure that the loan will satisfy agency guidelines. Assuming the originator decides to fund the loan through an agency securitization (rather than keeping the loan in portfolio), they may at this point already forward-sell the loan in the TBA market, effectively hedging against changes in the value of the loan during the lock period. There are then a number of different ways for the securitization to be executed. In a “lender swap” transaction, the seller directly pools loans and delivers the pool to the securitizing agency (e.g. Fannie Mae) in exchange for an MBS certificate, which can be subsequently sold to investors in the secondary market (or delivered to them, if it had previously been forwardsold). There are also multi-seller swaps, where loans from different financial institutions are pooled into a common MBS (e.g. Fannie Mae Majors) and sellers receive a proportional share of the resulting pool. This is particularly relevant in the Ginnie Mae segment of the agency market, where the Ginnie Mae II multi-issuer pool program accounts for 85–90% of new MBS 12 During the housing and nonagency MBS boom that ended in 2007, these mezzanine tranches were then pooled and resecuritized as collateralized debt obligations (CDOs). These CDOs suffered very high credit losses during the subsequent bust, including the AAA rated tranches (Cordell et al., 2019).

340  Research handbook of financial markets

issuance (Tozer, 2019). The multi-issuer channel is particularly valuable for smaller lenders that would not otherwise have the economies of scale to originate and securitize government mortgages. Finally, Fannie Mae and Freddie Mac also conduct “whole loan conduit” or “cash window” transactions in which they purchase loans directly from originators (typically smaller ones), pool these loans themselves, and sell the issued MBS in the secondary market.13 There are costs and benefits associated with each of these execution methods. For example, the pricing obtained at the cash window is typically worse, but the originator obtains liquidity immediately and does not face the risk of not being able to assemble enough loans for a pool.14 An et al. (2022) provide a detailed analysis of sellers’ choices between these different agency securitization venues and show for example that (i) small issuers and nonbanks are more likely to use the cash window and (ii) sellers with higher-value loans (with more favorable prepayment characteristics) favor single-seller swaps over multi-seller swaps.

15.4 RISKS TO MBS INVESTING, PREPAYMENT, AND THE OAS MBS yields significantly exceed yields on risk-free securities reflecting the risks associated with investing in MBS. We review these risks and then discuss prepayments, both their measurement and their modeling. Finally, we delve more deeply into the valuation of agency MBS through option-adjusted spreads (OAS). 15.4.1 Risks to MBS Investing Risks to MBS investing can be grouped into four main categories: duration, prepayment, credit and liquidity. Duration risk: MBS duration measures two closely related concepts. It is the weighted average time until cash flows, which include both principal and interest payments, are paid out to investors. But duration is also the sensitivity of the MBS price to a change in the general level of interest rates. Because MBS pay fixed coupons to investors and typically have 30-year maturities, duration is high and prices are very sensitive to interest rates. A key distinguishing feature of MBS is that the duration of the security is not fixed but rather uncertain because borrowers can prepay their loans at any time. Prepayment risk: MBS prepayments are either voluntary or involuntary. Voluntary prepayments are largely determined by borrowers’ refinancing and relocation decisions while involuntary prepayments reflect defaults (see credit risk discussion below). Refinancing is the most important source of prepayment, because for typical mortgages securitized into agency MBS, the borrower can pay off the remaining balance at par with no penalty. Since the mortgage rate is fixed, refinancing to a new loan is attractive when market interest rates are low; therefore prepayments rise and MBS duration falls when interest rates decline. In other words, MBS are callable securities, and price appreciation from lower interest rates is therefore capped—MBS exhibit “negative convexity.” 13 There is no cash window for Ginnie Mae MBS, because these pools are issued by private financial institutions rather than Ginnie Mae itself. 14 According to https://www​.machinesp​.com​/post​/a​-close​-look​-at​-the​-gse​-cash​-window, the share of cash window transactions for Freddie Mac-issued pools has been 50% or more in recent years.

Mortgage-backed securities 

341

Credit risk: At any point in time, borrowers can fail to make payments on the mortgages underlying an MBS. Default occurs when the loan no longer pays principal and interest until liquidation. Default is often caused by a “double trigger” of negative equity and a negative life event such as unemployment, illness or divorce (e.g. Ganong & Noel, 2020; Low, 2021).15 Without such negative events, borrowers are generally reluctant to walk away from their mortgage and home even when equity is deeply negative (e.g. Bhutta et al., 2017). More generally, not all delinquencies turn into defaults as loans can cure without “rolling” into more severe credit buckets. The implications of default for investors depend on whether the MBS is an agency or nonagency security. For agency MBS, the GSEs and Ginnie Mae promise full and timely payment of principal and interest, a guarantee that is either explicitly or implicitly backed by the federal government. When a borrower defaults, the issuer repurchases the loan from the MBS pool at par, resulting in a prepayment event for MBS investors (an involuntary prepayment).16 For nonagency MBS, however, borrower default is much more significant because investors bear any credit losses, starting with the most junior security, the equity tranche. Investors model probabilities of default and loss-given-default using reduced-form models that include loan characteristics (e.g. LTVs, credit scores) as well as macroeconomic variables.17 Although agency MBS are essentially free of credit risk, in recent years the GSEs have issued a new instrument—credit risk transfer (CRT) bonds—with cash flows explicitly tied to credit losses on agency mortgages. CRTs are structured debt securities linked to a “reference pool” of securitized loans. These bonds experience principal write-downs if credit losses on the reference pool exceed particular thresholds. See Finkelstein et al. (2018) for more details. CRTs were introduced to reduce the GSEs’ exposure to mortgage credit risk by shifting some of the risk to the private sector. CRT secondary market prices also provide a useful market signal about expected future mortgage credit losses; for example CRT prices fell sharply at the start of the COVID-19 pandemic before recovering. Trading and funding liquidity risk: MBS trading liquidity—the ease with which securities can be traded—and funding liquidity—how easily MBS collateral is funded—are additional risks to investing in MBS and influence returns (Brunnermeier & Pedersen, 2009; Song & Zhu, 2019). Trading liquidity varies quite significantly by type of MBS—liquidity for privatelabel MBS is quite limited, but for agency MBS the TBA forward market provides a high level of trading liquidity as well as funding liquidity through the execution of dollar rolls (see Section 15.5).

15 This double trigger is not a necessary condition for default, however. While most defaulters during and after the financial crisis were indeed underwater on their mortgages, Low (2021) presents evidence that the share of defaults with positive equity is quite high in “normal times.” 16 For MBS guaranteed by the GSEs, it is the relevant agency, Fannie Mae or Freddie Mac, that repurchases the mortgage. For Ginnie Mae pools, it is the financial institution that issued the security rather than Ginnie Mae itself (Tozer, 2019). This is an important distinction, because the financial characteristics of issuers can affect their incentives to repurchase loans. For example bank issuers have been more likely to repurchase nonperforming mortgages from Ginnie Mae pools during the COVID-19 pandemic because of their stronger liquidity position. 17 See Demyanyk and Van Hemert (2011) for an example of this type of model.

342  Research handbook of financial markets

15.4.2 Measuring and Modeling Prepayments We now turn to a more detailed discussion of prepayment risk, the most salient of these four risks for agency MBS. For a given borrower, prepayment often involves paying down the entire loan balance. Such an event only marginally reduces the overall MBS pool balance though, because each pool is backed by many loans. Prepayment is measured by the single monthly mortality (SMM), which is the fraction of an MBS balance prepaid in a month relative to the remaining scheduled principal balance, and by the conditional prepayment rate (CPR), which is simply the SMM expressed at an annual rate. The solid line in the top panel of Figure 15.2 plots the time series of CPR for the aggregate universe of 30-year fixed-rate agency MBS. Aggregating balances hides pool-specific prepayment variation, which is significant as shown earlier in Table 15.1. But even the aggregate prepayment rate exhibits very wide variation, ranging from a CPR of about 55% during the 2003 refinancing wave to a low of about 10% in 2008. The dashed line is the “moneyness” of the mortgage universe, which is the difference between the average interest rate on the universe of outstanding mortgages and the current market mortgage rate. When the moneyness of the universe increases, refinancing becomes more attractive and prepayments therefore

CPR

Moneyness

1.5 1

40 .5 0

20

-.5

Moneyness, percent

CPR, percent

60

-1

0 2000m1

2005m1

2010m1

2015m1

2020m1

CPR, percent

30 25 20 15 10 -2

0

2

4

6

Moneyness, percent

Note:  The top panel shows the time series of the monthly conditional prepayment rate (CPR) on the universe of 30-year fixed-rate agency MBS weighted by their remaining principal balance, against the moneyness of the mortgage universe. Moneyness is calculated as the weighted average coupon rate (WAC) minus the monthly average 30-year fixed-rate mortgage rate. The bottom panel shows a binned scatter plot (Cattaneo et al., 2019) of the crosssectional variation in CPR as a function of their moneyness. All data is monthly and covers the period 2000–2021. Source:   eMBS; Freddie Mac.

Figure 15.2  CPR for agency MBS

Mortgage-backed securities 

343

rise. Even so, similar levels of moneyness in 2003 and 2008 led to very different prepayment outcomes. A much greater degree of heterogeneity also exists at the level of specific pools. Prepayment modeling attempts to explain this variation and to predict prepayments. 15.4.2.1 Modeling prepayments The academic literature considers structural and rational models of mortgage prepayment (e.g. Stanton, 1995), but practitioners rely on reduced-form statistical prepayment models.18 These models do not assume rational borrower behavior but use information on observable borrower characteristics and macroeconomic factors to explain variation in the SMM. Given the many variables and the complexity of the task, models from different investors often disagree significantly about predicted prepayments (Carlin et al., 2014). Explanatory variables in prepayment models can be logically grouped into those related to turnover, refinancing, defaults and curtailments. The first two channels are quantitatively the most important.19 The turnover channel is associated with property sales, which are affected by economic conditions and the strength of the housing market. Turnover is highly seasonal, peaking in the summer, and is also subject to a “seasoning” effect because sales are less likely during the first two to three years after a home purchase. The refinancing channel exhibits the greatest variation over time and across pools. The key variable used in modeling it is a pool’s moneyness: when moneyness is positive, a borrower can lower their rate and monthly payment by refinancing—in other words, the borrower’s prepayment option is “in-the-money” (ITM). Negative moneyness, instead, means that refinancing (or selling the home and buying another) would increase the rate paid—the borrower’s option is “out-of-the-money.” 15.4.2.2 Prepayment versus moneyness: the “S-curve” The bottom panel of Figure 15.2 plots average prepayment rates as a function of moneyness in a monthly panel of MBS indexed by coupon rate and year of origination. Reflecting the shape of the relationship, this is known as an “S-curve.” While prepayments rise with moneyness, on average, they never come close to reaching 100%.20 This reflects the fact that many borrowers fail to refinance when it is in their monetary interest to do so (Keys et al., 2016). In addition, the S-curve bends down as pools become deeply ITM, reflecting the so-called “burnout effect”—over time, an ITM mortgage pool becomes less responsive to interest rates because the borrowers most sensitive to the refinancing incentive have already exited. The S-curve shifts over time due to changes in non-moneyness drivers of prepayment. For instance, when interest rates hit multi-year lows, refinancing for given levels of moneyness 18 Popular statistical frameworks for modeling mortgage prepayment risk as well as credit risk include the Cox proportional hazard model and the logit. See Deng et al. (2000) for a well-known competing risk model of mortgage prepayment and default. Machine learning methods have also been applied to model mortgage prepayment risk and credit risk in recent years (e.g., Sadhwani et al., 2021). 19 Curtailments are partial—rather than full—principal payments that often occur late in the life of a mortgage. Defaults represent a prepayment event for the investor because of the agency credit guarantee, as we have discussed. Default is however relatively unusual among pools issued by Fannie Mae and Freddie Mac, which typically consist of loans to prime borrowers. 20 Note that in constructing this curve, we bin together a large number of securities. As shown in Table 15.1 the underlying variation at the level of single pools is much larger than what the S-curve would imply.

344  Research handbook of financial markets

typically increases due to the “buzz” surrounding these low rates—this is known as the “media effect” in prepayment modeling. This effect was prominent during the record 2002-03 refinancing wave (Figure 15.2, top panel), and may also have contributed to high prepayments during the COVID-19 pandemic after mortgage rates reached all-time lows. Furthermore, an increase in “cash-out” refinancing to extract equity during a home price boom can shift the S-curve upwards, while conversely, falling home prices and/or a weak economy can make it difficult for borrowers to qualify for a refinancing, thereby reducing prepayment for a given moneyness. Changes in agency underwriting guidelines or government policies (such as the Home Affordable Refinance Program introduced in the wake of the Great Recession) can also substantially shift prepayments. Heterogeneity in refinancing across pools also reflects differences in creditworthiness as measured by credit scores and LTVs, and loan size due to the fixed costs involved in refinancing. State-level policies also matter (e.g. New York’s mortgage recording tax). 15.4.3 The OAS and Risks to Investing in Agency MBS We use the option-adjusted spread (OAS) to delve further into the risks associated with agency MBS. The OAS is the most popular metric to assess agency MBS valuations and risk premia. As shown by Boyarchenko et  al. (2019), the OAS is equal to the average expected excess returns over the lifetime of the security.21 Formally, the OAS is the constant spread to baseline rates that sets the expected discounted value of cash flows equal to the security’s market price after accounting for prepayments: T



PM = 

å k =1

X k (rk )

Õ j=1(1 + OAS + rj ) k

, (15.1)

where PM is the market price of an MBS, rj is the riskless interest rate at time j and Xk is the cash flow from the security. The OAS increases the larger the value of discounted cash flows relative to the market price, meaning that when spreads are positive, MBS trade below the discounted price net of the OAS.22 The calculation of the expectation term in the OAS uses Monte Carlo simulations and both a calibrated interest rate and an estimated prepayment 21 In addition, period-by-period excess returns on an MBS can be expressed as the sum of carry income and capital gains. In terms of the OAS, these are equal to the OAS itself plus the (negative) change in the OAS with a weight equal to the MBS duration. One limitation of the OAS is that it is a model-derived measure and thus subject to various assumptions. 22 Another commonly used metric is known as the zero-volatility spread. The ZVS abstracts from rate uncertainty and it is defined as: T



PM =

å k =1

Õ

k

X k (rk )

(1 + ZVS + rj ) j =1

. (15.2)

The ZVS-OAS difference is known as the “option cost.’’ In computing the ZVS both cash flows and discounts are evaluated along a single expected risk-neutral rate path, thus ignoring the effects of uncertainty about the timing of prepayments on the MBS valuation. This implies that the ZVS will be larger than the OAS.

Basis Points

Mortgage-backed securities  160 140 120 100 80 60 40 20 0 -20 -40

QE1

Jan00

Jan03

Jan06

QE3

Jan09

Jan12

345

QE4

Jan15

Jan18

Jan21

Basis points

80

60

40

20 -.01

0

.01

.02

.03

Moneyness

Note:  The top panel shows the time series of the option-adjusted spread (to Treasuries and averaged each month) on the current-coupon agency MBS. The horizontal solid line is the sample average and shaded areas represent periods in which the Federal Reserve purchased agency MBS in QE programs. The bottom panel shows a binned scatter plot (Cattaneo et al., 2019) of the cross-sectional variation in the OAS across MBS coupons as a function of their moneyness. Moneyness is calculated as the coupon rate plus 50 basis points (to account for servicing and the guarantee fee) minus the monthly average 30-year fixed-rate mortgage rate. The bottom panel includes only coupons with remaining principal balance of at least $100 million. All data is monthly average and covers the period 2000–2021. Source:   JP Morgan; Freddie Mac.

Figure 15.3  Time-series and cross-sectional variation of the agency OAS model. The two are combined to simulate interest rate paths and corresponding prepayment flows to obtain model prices and spreads in Equation 15.1.23 We turn to data on OAS from a major dealer (JP Morgan) to present stylized facts in both the time series and cross-section of agency MBS. The top panel of Figure 15.3 shows the evolution of Treasury OAS for the so-called current-coupon MBS, meaning the synthetic coupon trading at par, issued by Fannie Mae. This OAS averaged about 50 basis points since 2000, but spiked to about 150 basis points in the fall of 2008, and turned negative at times during the periods of the QE3 and QE4 programs, when the Federal Reserve purchased large quantities of agency MBS.

23 The interest rate model is calibrated to the term structure of interest rates and volatility surface implied by prices of interest rate derivatives. Once the paths for the interest rates are determined, the cash-flows are obtained from the prepayment model.

346  Research handbook of financial markets

A negative OAS indicates that, after accounting for the embedded prepayment option, MBS are valued more highly than Treasuries. On face value this seems anomalous because Treasuries are more liquid than MBS, are explicitly issued by the federal government, and are generally treated more favorably for regulatory purposes (e.g. bank capital and liquidity requirements). However, a negative OAS could for example be driven by preferred-habitat effects or indicate that some investors reached for a higher yield because they cannot synthetically replicate the higher MBS zero-volatility yield by investing in Treasuries and interest rate options. Note that in this case, the MBS yield will always exceed that on Treasuries, even when OAS are negative. A final explanation is simply that there may be model error leading to mismeasurement of the OAS. The bottom panel of Figure 15.3 focuses on the cross-section of OAS and reveals substantial variation across MBS with different moneyness levels.24 In the cross-section, a smile-like pattern emerges: spreads are lowest for securities for which the prepayment option is at-themoney, and increase if the option moves out-of-the-money and especially when it is in-themoney. The OAS smile pattern was first shown by Boyarchenko et al. (2019), who also find that OAS predicts realized excess returns. A similar smile-shaped pattern in MBS excess returns is documented by Diep et al. (2021). Large positive average OAS over time and across securities suggest that MBS investors require risk compensation to hold MBS over Treasuries. The embedded prepayment option means that even if payments are guaranteed, the timing of cash flow accruals is uncertain. Equation 15.1 explicitly incorporates the feature that cash flows depend on interest rates through prepayments. However, OAS abstracts from uncertainty related to non-interest rate factors that affect prepayments, such as house prices and lending standards. Boyarchenko et al. (2019) show that MBS investors earn risk compensation for these non-interest-rate prepayment factors and that these factors underlie the cross-sectional smile pattern in the OAS.25 In the time series, Boyarchenko et  al. (2019) further show that risk factors unrelated to prepayment, such as liquidity or changes in the perceived strength of the implicit federal government guarantee on the agencies, are important drivers of the average OAS. For example, the non-prepayment component in the OAS co-moves with spreads on other agency debt and corporate securities, reflecting shared risk factors. 15.4.4 Supply Effects and Fed Quantitative Easing The supply of MBS—which is affected by the net volume of new issuance as well as Fed MBS purchases (via quantitative easing, QE) that reduce the net supply available to private investors—is also positively related to the non-prepayment component of OAS. As an indication 24 Here we define moneyness as the difference between the MBS coupon rate and the current level of the 30-year fixed mortgage rate after adding a constant adjustment of 50 basis points. We make this adjustment because the mortgage note rates are typically around 50 basis points higher than the MBS coupon due to the agency guarantee fee as well as servicing fees. 25 They arrive at this conclusion by using matched pairs of IO and PO strips that split a pool’s cash flows into interest payments and principal payments. If the strips in a pair are fairly valued relative to each other, risk-neutral (or market-implied) prepayment rates can be estimated as multiples of physical (realized) ones. OAS computed under these alternative risk-neutral prepayments do not vary much in the cross-section, meaning that it is the prepayment risk premium that leads to the smile-shaped pattern.

Mortgage-backed securities 

347

of these effects, the OAS turned negative during QE3 and QE4 when the Fed purchased large quantities of agency MBS (see grey bars in Figure 15.3, top panel). Consistent with this fact, event studies using high-frequency data find that announcements of new Fed MBS purchases are associated with significant declines in MBS yields and OAS (Gagnon et al., 2011; Hancock & Passmore, 2011; Krishnamurthy & Vissing-Jorgensen, 2011). Chernov et al. (2016) similarly find evidence that MBS risk premiums are affected by the Federal Reserve’s QE programs using a more model-based approach. Krishnamurthy and Vissing-Jorgensen (2011) and Di Maggio et al. (2020) furthermore show that MBS purchases have larger effects on MBS yields than a comparable volume of Treasury purchases, consistent with the presence of some degree of market segmentation. An et al. (2018) study the effects of Fed MBS purchases on market functioning and liquidity, finding modest negative effects. Other research investigates the broader financial and macroeconomic effects of the Federal Reserve’s MBS purchases. For example, Di Maggio et al. (2020) find that Fed QE significantly boosted refinancing activity and as a result, led to higher aggregate consumption. Beraja et al. (2019) show that the effectiveness of QE and monetary policy more generally depends on the distribution of home equity, because insufficient equity reduces the ability of borrowers to refinance.

15.5 TRADING Most agency RMBS trading occurs through the to-be-announced or “TBA” forward market. The key feature of a TBA trade is that the seller does not specify exactly which pools will be delivered at settlement. Instead, the buyer and seller agree on six trade parameters: the agency, coupon, maturity, price, face value, and settlement month, and any combination of pools satisfying the parameters and SIFMA good delivery guidelines can be delivered at settlement.26 The TBA market effectively concentrates the fragmented universe of around a million individual agency MBS pools into a small number of liquid contracts for trading purposes, thereby improving fungibility and liquidity.27 Investors trade TBAs to express price views or for hedging. TBAs are also used by lenders to hedge their origination pipeline, as discussed in Section 15.3. The TBA market can further be used as a funding vehicle, through the execution of “dollar roll” transactions. In a dollar roll, the roll seller sells TBAs for a coming delivery month (the “front” month) and simultaneously purchases TBAs for a later “back” month. This provides short-term funding to the roll seller by postponing the date when she is due to pay cash to settle her long TBA position. The substance of a dollar roll is similar to a repurchase agreement, but there are some important differences; see Song and Zhu (2019) for further discussion and empirical analysis of the dollar roll market. The Federal Reserve has also used dollar rolls to

26 For details of the good delivery guidelines for TBA settlement, see Chapter 8 of SIFMA (2021). The exact CUSIPs to be delivered are specified two days prior to settlement. Settlement occurs once per month (e.g. for 30-year Uniform MBS, settlement day is typically between the 10th and 14th of the month). The TBA settlement schedule is available at www​.sifma​.org​/resources​/general​ /mbs​-notification​-and​-settlement​-dates/. 27 For example, Gao et al. (2017) find that over a sample period of several years, 12 maturity-coupon combinations accounted for 96% of TBA trades.

348  Research handbook of financial markets

support market functioning, and actively employed this tool during the COVID-19 pandemic (Frame et al., 2021). Since mid-2019, Fannie Mae and Freddie Mac pools have traded through a single set of “Uniform MBS” (UMBS) TBA contracts, in which pools issued by either agency can be delivered at settlement. Previously, the two GSEs traded separately in the TBA market. This change in market structure was implemented because TBA trading had historically been highly concentrated in Fannie Mae contracts, leading to an illiquidity discount for Freddie Mac pools which put them at a competitive disadvantage. Liu et al. (2021) find that UMBS implementation successfully improved Freddie Mac TBA liquidity without any obvious adverse effects on overall market functioning. The agency market also features significant trading of individual pools (known as “spec pool” trading). One reason for spec pool trading is that the TBA market operates on a cheapest-to-deliver basis—sellers will deliver the least valuable eligible pools. This leads to a semiseparating equilibrium in which more valuable “pay-up” MBS pools trade individually while less valuable MBS trade on a pooled basis in the TBA market. See Li and Song (2020) for a theoretical model of this structure. Specified pool trades are often arranged to settle on TBA settlement dates, but can settle at any time of the month.28 Specified pool trading also includes agency MBS pools that for various reasons are not TBA-eligible.29 Other mortgage securities, such as CMBS, agency CMOs, and nonagency RMBS, also trade on an individual basis. 15.5.1 Evidence on Trading Activity and Liquidity Table 15.2 presents trading volume statistics based on TRACE data aggregated by SIFMA. Agency residential MBS trading activity dwarfs the other segments of the market, with $288 billion of daily trading volume, compared to $2.2 billion for CMBS and only $0.5 billion for nonagency RMBS. This reflects $261 billion of TBA trading (about 90% of the agency RMBS total), followed by a smaller but still very significant $25.4 billion of specified pool trades and $1.4 billion of agency CMOs. Estimated trading costs are also significantly lower in the TBA market. Bessembinder et al. (2013) estimate one-way trading costs of only one basis point (bp) for TBAs, compared to 40bp for specified pools, and 39bp for nonagency MBS. Gao et al. (2017) find that TBA liquidity has positive spillover effects on the specified pool market—trading costs are lower for specified pools that are TBA eligible and for spec pool trades close to TBA settlement dates. Huh and Kim (2020) trace out the broader effects of TBA liquidity using a TBA-eligibility cutoff at the national conforming loan limit. TBA eligibility is estimated to reduce mortgage rates by 7–28bp, and to spur refinancing activity. Table 15.2 also compares MBS trading volume to activity in other US fixed income markets. TBA activity is lower than activity in the Treasury market, but trading volume is more than six times higher than that in the corporate bond market, despite the larger stock of corporate 28 Chen et al. (2020) show that during the market turmoil of March 2020, stressed market participants scrambled to raise cash by selling MBS in the specified pool market, temporarily driving spec pool prices below TBA prices. To meet the demand for liquidity and stabilize the market, the Federal Reserve responded by executing unconventional MBS purchases on a T+3 settlement basis. 29 TBA-ineligible pools include MBS with more than 10% of super conforming loans exceeding the national conforming limit, and pools with LTVs exceeding 105% (Vickery and Wright, 2013, Huh and Kim, 2020). TBA eligibility rules are designed to limit heterogeneity within each cohort.

Mortgage-backed securities 

349

Table 15.2  MBS trading volume Avg. daily trading volume ($bn) A. Residential: agency MBS TBA Specified pool

260.95 25.34

CMO

1.37

Total

287.67

B. Residential: non-agency MBS CMO (IO/PO)

0.05

CMO (P&I)

0.43

Total

0.48

C. Commercial MBS Agency CMBS

1.22

Non-agency CMBS (IO/PO)

0.28

Non-agency CMBS (P&I)

0.74

Total

2.23

Memo: other USD fixed income securities US Treasury

603.2

Corporate debt

38.9

Municipal bonds

12.0

Federal agency securities

5.3

Asset backed securities

1.9

Note:   Average daily trading volume is calculated over the period from January to September 2021. IO/POs are stripped passthrough pools that pay either interest only or principal only, whereas P&I CMOs are typically sequential pay bonds with cashflows derived from both principal and interest payments on the underlying mortgages. Source:   SIFMA, aggregated from FINRA TRACE data.

bonds outstanding. Trading activity is even lower for municipal bonds, agency debt and asset backed securities, and Bessembinder et al. (2020) further show that TBA trading costs are much lower than for these other markets.30

15.6 ECONOMIC EFFECTS OF MBS AND MORTGAGE SECURITIZATION What are the broader economic effects of MBS markets and mortgage securitization? A sizeable academic literature has studied different aspects of this question and also highlighted 30 Bessembinder et al. (2020) and Gao et al. (2017) speculate that introducing a TBA-like forward market for corporate bonds could improve liquidity in that market. Corporate bonds are more heterogeneous than agency MBS, however, and the number of issuers is much higher, factors which would likely be hurdles to implementing such an idea.

350  Research handbook of financial markets

potential downsides of securitization, especially in the wake of the Great Recession. In this section, we provide a brief overview of some of the main themes. A key benefit of securitization is that it makes mortgages more liquid, thereby significantly de-coupling loan originators’ ability to produce loans from their own financial condition (e.g. funding, risk exposure). As evidence on this point, Loutskina and Strahan (2009) show that bank liquid assets and deposit costs play much less of a role in the origination of conforming mortgages, which can easily be securitized, compared to less-liquid jumbo mortgages. Securitization is also fundamental to the rise of nonbank lenders (financed through wholesale funding) as the dominant origination channel in the US (Buchak et al., 2020; Gete & Reher, 2020; Kim et al., 2022).31 Since securitization increases liquidity and broadens the set of lenders able to originate mortgages, one would naturally expect that it also leads to an outward shift in credit supply, plausibly increasing credit access for otherwise “marginal” borrowers. However, a sizeable literature has argued that securitization also reduces credit quality through an additional “moral hazard” channel: as originators offload the credit risk, they may have weaker incentives to screen borrowers, leading to lower acquisition of soft information and worse ex-post outcomes (e.g. Keys et al., 2010; Nadauld & Sherlund, 2013; Rajan et al., 2015; Choi & Kim, 2020). But the strength of the evidence, and the question of whether securitization is an important cause of the US mortgage boom and bust of the 2000s, remains debated in the literature (e.g. Bubb & Kaufman, 2014; Foote et al., 2012, 2020; Mian & Sufi, 2021). Related literature has focused on the effects of securitization on loan monitoring after origination—in particular, whether the mortgage servicer has insufficient or misaligned incentives to work out appropriate solutions (such as modifications) for delinquent loans. Again, there is a debate in the literature about how important these incentive effects were in explaining (non-)modifications of securitized mortgages during the Great Recession (e.g. Piskorski et al., 2010; Agarwal et al., 2011; Adelino et al., 2013; Aiello, 2022). More recently, Kim et al. (2021) find that mortgage servicers’ financial condition affects forbearance outcomes for securitized mortgages during the COVID-19 crisis. Related, Wong (2018) finds evidence of misaligned servicer incentives in the CMBS market. Securitization may also affect mortgage contract design. Fuster and Vickery (2015) show that lenders reduce the supply of long-term prepayable fixed-rate mortgages (relative to adjustable-rate mortgages) when securitization markets become illiquid. They argue that this is due to lenders’ limited ability to absorb the interest rate and prepayment risk embedded in FRMs.32 Thus, it appears unlikely that the 30-year prepayable FRM, which is by far the dominant mortgage type in the US, could be offered at similarly competitive rates without liquid securitization markets. In turn, the popularity of prepayable FRMs has broader consequences for financial markets and the transmission of monetary policy. In particular, several studies argue that “convexity hedging” flows lead to important interactions between the MBS market and the Treasury yield

31 Without securitization, nonbank lenders could potentially sell whole loans to banks or other financial institutions, but such a market would be far less liquid. Indeed, in the jumbo market, where securitization has been relatively dormant since the Great Recession, nonbanks play a much smaller role than in the agency market. 32 Recent work by Xiao (2021) shows that this ability varies in the cross-section of banks depending on their funding structure, which in turn affects their propensity to securitize mortgages.

Mortgage-backed securities 

351

curve (Hanson, 2014; Malkhozov et al., 2016; Hanson et al., 2021). Furthermore, the fact that US borrowers need to refinance to benefit from a drop in market interest rates means there is much less direct transmission of monetary policy to household balance sheets than in a system with adjustable-rate mortgages (e.g. Campbell, 2013; Di Maggio et al., 2017). Transmission is further blunted by the limited ability of mortgage originators to increase origination capacity during periods of peak demand; instead, originators tend to earn high markups during such periods (Fuster et al., 2013; 2017; 2021).

15.7 DIRECTIONS FOR FUTURE RESEARCH The MBS market was a relatively neglected research topic prior to the 2008 financial crisis, but the literature has grown rapidly in the years since. Rich loan- and security-level datasets are now available to researchers, and the introduction of TRACE data for structured products in 2011 provides new opportunities to study MBS microstructure and liquidity. We end this chapter by highlighting some topics that we believe present opportunities for future research. (i) Securitization and alternative mortgage designs. Various alternative mortgage designs have been proposed to improve macroeconomic stability, reduce transaction costs or produce other benefits. For instance, Eberly and Krishnamurthy (2014), Guren et al. (2021) and Campbell et  al. (2021) study mortgages that can switch from FRMs to ARMs or interest-only loans during recessions, while Greenwald et al. (2021) study shared appreciation mortgages with payments that adjust with home prices. An open question is how such alternative products would be funded and what role securitization markets would play. Securitization may in fact hinder innovation, in the sense that the existence of a thick, liquid secondary market for a particular contract—30-year FRMs—may present a barrier for alternative designs. (ii) What’s holding back nonagency securitization? Nonagency securitization remains far lower than prior to the financial crisis, despite the much higher credit guarantee fees now charged by the GSEs. Stricter post-crisis regulation is a natural reason why, as discussed in Section 15.2.1. But research has not clearly disentangled the role of regulation from other factors such as changes in expectations. (iii) Investor behavior. There is limited research on the determinants of investor behavior in the MBS market (e.g. the striking fact that MBS now make up half of bank security portfolios) and how investors affect pricing and liquidity.33 (iv) Securitization and climate change. An emerging literature studies the interaction between climate change and mortgage and MBS markets (e.g. Ouazad & Kahn, 2019). This is likely to be a fruitful topic for future work. For instance, securitization prices can provide useful high-frequency information about the market’s assessment of climate and natural disaster risk. Another application: Fannie Mae has developed a “green MBS” program for loans backed by buildings with green building certifications—to our knowledge, its effects have not been rigorously studied.

33 One contribution along those lines is by Erel et al. (2013), who study the drivers of bank investments in nonagency MBS tranches.

352  Research handbook of financial markets

To sum up, mortgage-backed securities lie at the heart of housing finance and the US financial system, and also play a significant role in monetary policy and monetary transmission. The MBS market also presents many opportunities for academic study, and we anticipate that this important market will remain a vibrant topic for research in the years to come.

REFERENCES Adelino, M., Gerardi, K., & Willen, P. S. (2013). Why don’t lenders renegotiate more home mortgages? Redefaults, self-cures and securitization. Journal of Monetary Economics, 60(7), 835–853. Adelino, M., McCartney, W. B., & Schoar, A. (2020). The role of government and private institutions in credit cycles in the U.S. mortgage market. Working Paper 20-40, Federal Reserve Bank of Philadelphia. Adelino, M., Schoar, A., & Severino, F. (2016). Loan originations and defaults in the mortgage crisis: The role of the middle class. Review of Financial Studies, 29(7), 1635–1670. Agarwal, S., Amromin, G., Ben-David, I., Chomsisengphet, S., & Evanoff, D. D. (2011). The role of securitization in mortgage renegotiation. Journal of Financial Economics, 102(3), 559–578. Aiello, D. J. (2022). Financially constrained mortgage servicers. Journal of Financial Economics 144(2), 590–610. An, X., Deng, Y., & Gabriel, S. (2009). Value creation through securitization: Evidence from the CMBS market. Journal of Real Estate Finance and Economics, 38(3), 302–326. An, Y., Li, L., Liu, H., & Song, Z. (2018). The costs of quantitative easing: Liquidity and market functioning effects of Federal Reserve MBS purchases. International Journal of Central Banking, 14, 259–304. An, Y., Li, L., Liu, H., & Song, Z. (2022). Shadow bank and fintech mortgage securitization, Working Paper, Johns Hopkins University. Arcidiacono, N., Cordell, L., Davidson, A., & Levin, A. (2013). Understanding and measuring risks in agency CMOs, Working paper no. 13-8, Federal Reserve Bank of Philadelphia. Ashcraft, A. B., Gooriah, K., & Kermani, A. (2019). Does skin-in-the-game affect security performance? Journal of Financial Economics, 134(2), 333–354. Ashcraft, A., & Schuermann, T. (2008). Understanding the securitization of subprime mortgage credit. Foundations and Trends in Finance, 2(3), 191–309. Beraja, M., Fuster, A., Hurst, E., & Vavra, J. (2019). Regional heterogeneity and the refinancing channel of monetary policy. Quarterly Journal of Economics, 134(1), 109–183. Berg, J., Nielsen, M. B., & Vickery, J. (2018). Peas in a pod? Comparing the U.S. and Danish mortgage finance systems, Staff Reports 848, Federal Reserve Bank of New York. Bernanke, B. (2009). The future of mortgage finance in the United States. B.E. Journal of Economic Analysis and Policy, 9(3), 1–10. Bessembinder, H., Maxwell, W. F., & Venkataraman, K. (2013). Trading activity and transaction costs in structured credit products. Financial Analysts Journal, 69(6), 55–67. Bessembinder, H., Spatt, C., & Venkataraman, K. (2020). A survey of the microstructure of fixedincome markets. Journal of Financial and Quantitative Analysis, 55(1), 1–45. Bhattacharya, A. K., Berliner, W. S., & Fabozzi, F. J. (2008). The interaction of MBS markets and primary mortgage rates. Journal of Structured Finance, 14(3), 16–36. Bhutta, N., Shan, H., & Dokko, J. (2017). Consumer ruthlessness and mortgage default during the 2007 to 2009 housing bust. Journal of Finance, 72(6), 2433–2466. Boyarchenko, N., Fuster, A., & Lucca, D. O. (2019). Understanding mortgage spreads. Review of Financial Studies, 32(10), 3799–3850. Brunnermeier, M. K., & Pedersen, L. H. (2009). Market liquidity and funding liquidity. Review of Financial Studies, 22(6), 2201–2238. Bubb, R., & Kaufman, A. (2014). Securitization and moral hazard: Evidence from credit score cutoff rules. Journal of Monetary Economics, 63, 1–18. Buchak, G., Matvos, G., Piskorski, T., & Seru, A. (2018). Fintech, regulatory arbitrage, and the rise of shadow banks. Journal of Financial Economics, 130(3), 453–483.

Mortgage-backed securities 

353

Buchak, G., Matvos, G., Piskorski, T., & Seru, A. (2020). Beyond the balance sheet model of banking: Implications for bank regulation and monetary policy, Working Paper 25149, National Bureau of Economic Research. Burgess, G., Passmore, W., & Sherlund, S. (2023). Government agencies: Fannie Mae and Freddie Mac. In R. S. Gurkaynak & J. H. Wright (Eds.), Research Handbook of Financial Markets. Edward Edgar Publishing. Calem, P., Covas, F., & Wu, J. (2013). The impact of the 2007 liquidity shock on bank jumbo mortgage lending. Journal of Money, Credit, and Banking, 45(s1), 59–91. Campbell, J. Y. (2013). Mortgage market design. Review of Finance, 17(1), 1–33. Campbell, J. Y., Clara, N., & Cocco, J. F. (2021). Structuring mortgages for macroeconomic stability. Journal of Finance, 76(5), 2525–2576. Carlin, B. I., Longstaff, F. A., & Matoba, K. (2014). Disagreement and asset prices. Journal of Financial Economics, 114(2), 226–238. Cattaneo, M. D., Crump, R. K., Farrell, M. H., & Feng, Y. (2019). On binscatter, arXiv preprint arXiv:1902.09608. Chandan, S. (2012). The past, present, and future of CMBS, Tech. rep. Chen, J., Liu, H., Sarkar, A., & Song, Z. (2020). Dealers and the dealer of last resort: Evidence from MBS markets in the COVID-19 crisis, Staff Report 933, Federal Reserve Bank of New York. Chernov, M., Dunn, B. R., & Longstaff, F. A. (2016). Macroeconomic-driven prepayment risk and the valuation of mortgage-backed securities. Working Paper 22096, National Bureau of Economic Research. Choi, D. B., & Kim, J.-E. (2020). Does securitization weaken screening incentives? Journal of Financial and Quantitative Analysis, 1–43. Cordell, L., Feldberg, G., & Sass, D. (2019). The role of ABS CDOs in the financial crisis. Journal of Structured Finance, 25(2), 10–27. Credit Suisse. (2011). Agency CMBS market primer, Tech. rep. Davidson, A., & Levin, A. (2014). Mortgage valuation models: Embedded options, risk, and uncertainty. Oxford: Oxford University Press. de Jong, A., Kooijmans, T., & Koudijs, P. (2022). Plantation mortgage-backed securities: Evidence from Surinam in the 18th century. Journal of Economic History [in press]. DeFusco, A. A., Johnson, S., & Mondragon, J. (2019). Regulating household leverage. Review of Economic Studies, 87, 914–958. Demyanyk, Y., & Van Hemert, O. (2011). Understanding the subprime mortgage crisis. Review of Financial Studies, 24(6), 1848–1880. Deng, Y., Quigley, J. M., & Van Order, R. (2000). Mortgage terminations, heterogeneity and the exercise of mortgage options. Econometrica, 68(2), 275–307. Di Maggio, M., Kermani, A., Keys, B. J., Piskorski, T., Ramcharan, R., Seru, A., & Yao, V. (2017). Interest rate pass-through: Mortgage rates, household consumption, and voluntary deleveraging. American Economic Review, 107(11), 3550–3588. Di Maggio, M., Kermani, A., & Palmer, C. J. (2020). How quantitative easing works: Evidence on the refinancing channel. Review of Economic Studies, 87(3), 1498–1528. Diep, P., Eisfeldt, A. L., & Richardson, S. (2021). The cross section of MBS returns. Journal of Finance, 76(5), 2093–2151. Eberly, J., & Krishnamurthy, A. (2014). Efficient credit policies in a housing debt crisis. Brookings Papers on Economic Activity, 45(2), 73–136. Erel, I., Nadauld, T., & Stulz, R. M. (2013). Why did holdings of highly rated securitization tranches differ so much across banks? Review of Financial Studies, 27(2), 404–453. Fabozzi, F. J. (Eds.). (2016). The handbook of mortgage-backed securities (7th ed.). Oxford: Oxford University Press. Federal Reserve Bank of New York. (2021). Quarterly trends for consolidated U.S. banking organizations. Second Quarter. Finkelstein, D., Strzodka, A., & Vickery, J. (2018). Credit risk transfer and de facto GSE reform. Economic Policy Review, 88–116. Foote, C. L., Gerardi, K. S., & Willen, P. S. (2012). Why did so many people make so many ex post bad decisions? The causes of the foreclosure crisis. Working Paper 18082, National Bureau of Economic Research.

354  Research handbook of financial markets

Foote, C. L., Loewenstein, L., & Willen, P. S. (2020). Cross-sectional patterns of mortgage debt during the housing boom: Evidence and implications. Review of Economic Studies, 88(1), 229–259. Frame, W. S., Fuster, A., Tracy, J., & Vickery, J. (2015). The rescue of Fannie Mae and Freddie Mac. Journal of Economic Perspectives, 29(2), 25–52. Frame, W. S., Greene, B., Hull, C., & Zorsky, J. (2021). Fed’s mortgage-backed securities purchases sought calm, accommodation during pandemic. Tech. rep., Dallas Fed Economics. Fuster, A., Goodman, L., Lucca, D., Madar, L., Molloy, L., & Willen, P. (2013). The rising gap between primary and secondary mortgage rates. Federal Reserve Bank of New York Economic Policy Review, 19, 17–39. Fuster, A., Hizmo, A., Lambie-Hanson, L., Vickery, J., & Willen, P. S. (2021). How resilient is mortgage credit supply? Evidence from the COVID-19 pandemic. Working Paper 28843, National Bureau of Economic Research. Fuster, A., Lo, S. H., & Willen, P. S. (2017). The time-varying price of financial intermediation in the mortgage market. Staff report no. 805, Federal Reserve Bank of New York. Fuster, A., & Vickery, J. (2015). Securitization and the fixed-rate mortgage. Review of Financial Studies, 28(1), 176–211. Gagnon, J., Raskin, M., Remache, J., & Sack, B. (2011). The financial market effects of the Federal Reserve’s large-scale asset purchases. International Journal of Central Banking, 7, 3–43. Ganong, P., & Noel, P. J. (2020). Why do borrowers default on mortgages? A new method for causal attribution. Working Paper 27585, National Bureau of Economic Research. Gao, P., Schultz, P., & Song, Z. (2017). Liquidity in a market for unique assets: Specified pool and to-beannounced trading in the mortgage-backed securities market. Journal of Finance, 72(3), 1119–1170. Gerardi, K., Lehnert, A., Sherlund, S. M., & Willen, P. (2008). Making sense of the subprime crisis. Brookings Papers on Economic Activity, 69–145. Gete, P., & Reher, M. (2020). Mortgage securitization and shadow bank lending. Review of Financial Studies, 34(5), 2236–2274. Goetzmann, W. N., & Newman, F. (2010). Securitization in the 1920’s. Working Paper 15650, National Bureau of Economic Research. Green, R. K. (2013). Introduction to mortgages and mortgage backed securities, no. 9780124017436 in Elsevier Monographs. Elsevier. Greenwald, D. L., Landvoigt, T., & Nieuwerburgh, S. V. (2021). Financial fragility with SAM? Journal of Finance, 76(2), 651–706. Guren, A. M., Krishnamurthy, A., & McQuade, T. J. (2021). Mortgage design in an equilibrium model of the housing market. Journal of Finance, 76(1), 113–168. Hancock, D., & Passmore, W. (2011). Did the Federal Reserve’s MBS purchase program lower mortgage rates? Journal of Monetary Economics, 58(5), 498–514. Hanson, S. G. (2014). Mortgage convexity. Journal of Financial Economics, 113(2), 270–299. Hanson, S. G., Lucca, D. O., & Wright, J. H. (2021). Rate-amplifying demand and the excess sensitivity of long-term rates. Quarterly Journal of Economics, 136(3), 1719–1781. Hayre, L. (2001). Chap. 1. A concise guide to mortgage-backed securities (MBSs). In L. Hayre (ed.), Salomon Smith Barney guide to mortgage-backed and asset-backed securities (pp. 9–68). New York, NY: Wiley and Sons. Huh, Y., & Kim, Y. S. (2020). The real effects of secondary market trading structure: Evidence from the mortgage market. Working Paper, Federal Reserve Board. Keys, B. J., Mukherjee, T., Seru, A., & Vig, V. (2010). Did securitization lead to lax screening? Evidence from subprime loans. Quarterly Journal of Economics, 125(1), 307–362. Keys, B. J., Pope, D. G., & Pope, J. C. (2016). Failure to refinance. Journal of Financial Economics, 122(3), 482–499. Kim, Y. S., Laufer, S. M., Pence, K., Stanton, R., & Wallace, N. (2018). Liquidity crises in the mortgage market. Brookings Papers on Economic Activity, 49, 347–428. Kim, Y. S., Lee, D., Scharlemann, T., & Vickery, J. (2021). Intermediation frictions in debt relief: Evidence from CARES act forbearance. Working Paper. Kim, Y., Pence, K., Stanton, R., Walden, J., & Wallace, N. (2022). Non-banks and mortgage securitization. Annual Review of Financial Economics, 14, 137–166.

Mortgage-backed securities 

355

Krishnamurthy, A., & Vissing-Jorgensen, A. (2011). The effects of quantitative easing on interest rates: Channels and implications for policy. Brookings Papers on Economic Activity, 215–287. Kruger, S. (2018). The effect of mortgage securitization on foreclosure and modification. Journal of Financial Economics, 129(3), 586–607. Lewis, M. (2010). Liar’s poker. W. W. Norton. Li, W., & Song, Z. (2020). Asset heterogeneity, market fragmentation, and quasi-consolidated trading. Working paper. Liu, H., Song, Z., & Vickery, J. (2021). Defragmenting markets: Evidence from agency MBS. Staff Report 965, Federal Reserve Bank of New York. Loutskina, E., & Strahan, P. E. (2009). Securitization and the declining impact of bank finance on loan supply: Evidence from mortgage originations. Journal of Finance, 64(2), 861–889. Low, D. (2021). What triggers mortgage default? New evidence from linked administrative and survey data. Working Paper, CFPB. Malkhozov, A., Mueller, P., Vedolin, A., & Venter, G. (2016). Mortgage risk and the yield curve. Review of Financial Studies, 29(5), 1220–1253. McConnell, J. J., & Buser, S. A. (2011). The origins and evolution of the market for mortgage-backed securities. Annual Review of Financial Economics, 3(1), 173–192. Meuli, J., Nellen, T., & Nitschka, T. (2021). Covered bonds, loan growth and bank funding: The Swiss experience since 1932. International Finance, 24(1), 77–94. Mian, A., & Sufi, A. (2015). House of debt. University of Chicago Press. Mian, A., & Sufi, A. (2021). Credit supply and housing speculation. Review of Financial Studies, 35, 680–719. Mordel, A., & Stephens, N. (2015). Residential mortgage securitization in Canada: A review. Bank of Canada Financial System Review. Nadauld, T. D., & Sherlund, S. M. (2013). The impact of securitization on the expansion of subprime credit. Journal of Financial Economics, 107(2), 454–476. Ouazad, A., & Kahn, M. E. (2019). Mortgage finance and climate change: Securitization dynamics in the aftermath of natural disasters. Working Paper 26322, National Bureau of Economic Research. Piskorski, T., Seru, A., & Vig, V. (2010). Securitization and distressed loan renegotiation: Evidence from the subprime mortgage crisis. Journal of Financial Economics, 97(3), 369–397. Prokopczuk, M., Siewert, J. B., & Vonhoff, V. (2013). Credit risk in covered bonds. Journal of Empirical Finance, 21, 102–120. Rajan, U., Seru, A., & Vig, V. (2015). The failure of models that predict failure: Distance, incentives, and defaults. Journal of Financial Economics, 115(2), 237–260. Roberts, D., Sarkar, A., & Shachar, O. (2018). The costs and benefits of liquidity regulations. Staff Reports 852, Federal Reserve Bank of New York. Sadhwani, A., Giesecke, K., & Sirignano, J. (2021). Deep learning for mortgage risk. Journal of Financial Econometrics, 19(2), 313–368. SIFMA. (2021). Uniform practices for the clearance and settlement of mortgage-backed securities and other related securities. Securities Industry and Financial Markets Association. Song, Z., & Zhu, H. (2019). Mortgage dollar roll. Review of Financial Studies, 32(8), 2955–2996. Standard and Poor’s. (2020). A primer on China’s residential mortgage backed securities market. Technical Report. Standard and Poor’s. (2021). Global structured finance 2021 outlook. Technical Report. Stanton, R. (1995). Rational prepayment and the valuation of mortgage-backed securities. Review of Financial Studies, 8(3), 677–708. Tozer, T. (2019). A primer and perspective on Ginnie Mae. Milken Institute Discussion Paper. Urban Institute. (2021). Housing finance at a glance. Tech. rep., Housing Finance Policy Center. Vickery, J., & Wright, J. (2013). TBA trading and liquidity in the agency MBS market. Federal Reserve Bank of New York Economic Policy Review, 19, 1–18. Wong, M. (2018). CMBS and conflicts of interest: Evidence from ownership changes for servicers. Journal of Finance, 73(5), 2425–2458. Xiao, Z. (2021). Interest rate risk, prepayment risk and banks’ securitization of mortgages. Working Paper, University of British Columbia.

356  Research handbook of financial markets

APPENDIX 15.1 Section A: Variable definitions for construction of Figure 15.1 All statistics are derived from the Financial Accounts of the United States, except for nominal GDP (source: BEA, via FRED). Commercial mortgages and MBS are inclusive of multifamily loans but exclude farm loans. Total mortgage debt: residential: ●

L. 217, FL893065105: One-to-four-family residential mortgages

Total mortgage debt: commercial: ●

L. 217, FL893065405: Multifamily residential + L. 217, FL893065505: Commercial

Total MBS: residential: ●

L. 126, FL413065105: One-to-four-family residential mortgages (GNM) + L. 125, FL403065195: Mortgages, one-to-four-family residential, consolidated trusts (GSEs)+ L. 127, FL673065105: One-to-four-family residential (ABS)

Total MBS: commercial: ●

L. 126, FL413065505: Commercial mortgages (GNM) + L. 126, FL413065405: Multifamily residential mortgages (GNM) + L. 125, FL403065495: Multifamily residential consolidated trusts (GSEs) + L. 127, FL673065405: Multifamily residential (ABS) + L. 127, FL673065505: Commercial (ABS)

Total agency MBS: residential: ●

L. 126, FL413065105: One-to-four-family residential mortgages (GNM) + L. 125, FL403065195: Mortgages, one-to-four-family residential, consolidated trusts (GSEs)

Total agency MBS: commercial: ●

L. 126, FL413065505: Commercial mortgages (GNM) + L. 126, FL413065405: Multifamily residential mortgages (GNM) + L. 125, FL403065495: Multifamily residential, consolidated trusts (GSEs)

Nominal GDP: ●

Nominal GDP, seasonally adjusted annual rate https://fred​ .stlouisfed​ .org​ /series​ /GDP. Note: Statistics in figure normalized by four-quarter-ended GDP.​

Mortgage-backed securities 

357

Table 15A.1  Investors in agency and GSE-backed securities $bn

% of total

Depository institutions

3,357

32%

Federal Reserve

2,414

23%

Rest of the world

1,145

11%

Mutual funds

713

7%

Money market funds

499

5%

State and local governments

428

4%

Life insurance companies

348

3%

Credit unions

297

3%

Pension funds

260

2%

Households and non-profit organizations

247

2%

Government sponsored enterprises

219

2%

State and local government defined benefit pension funds

201

2%

Mortgage real estate investment trusts

188

2%

Property-casualty insurance companies

137

1%

Foreign banking offices

59

1%

Other

78

1%

Includes issues of federal budget agencies; issues of government sponsored enterprises such as Fannie Note:   Mae and FHLB; and agency- and GSE-backed mortgage pool securities issued by Ginnie Mae, Fannie Mae, Freddie Mac, and the Farmers Home Administration. Source: Financial Accounts of the United States, Table L.211, 2021:Q2.

16. Equity trading Caroline Fohlin1

16.1 INTRODUCTION1 Investors have been trading ownership stakes in business organizations at least since the Roman markets in the 2nd century BC. These earliest known equity markets primarily traded shares and bonds of “tax farming companies” (Smith, 2003). Share companies that traded equity appeared in other locations during the medieval era, notably in Toulouse starting in the 11th century, where LeBris et al. (2015) document the emergence and evolution of two such companies into large, incorporated firms with widely held equity by 1373. They highlight how early joint-stock firms were able to use private contractual arrangements under medieval law to create organizations much like a modern corporation—characterized by tradable shares with limited liability and dividend-like payouts—all without the need for government approval. But joint-stock companies remained uncommon for centuries, even after the foundation of the famous East India Companies, founded in 1600 in London and in 1602 in Amsterdam, and most governments typically did require founders to obtain concessions, which involved idiosyncratic and individualized—and costly—permissions. Following the advent of the joint-stock corporation and related creation of standardized, tradable shares, brokers emerged to facilitate the sale of public stocks and intermediate between equity owners. Initially working out of coffee houses, fairs, and even street corners, brokers began trading equity at preexisting commodity exchanges and soon at purpose-built stock exchanges. By the 16th century, there were markets trading equities and bonds in cities across Europe. Antwerp held a continuous fair and rose to become the financial center of the West by the mid-16th century. Much trading activity began to shift to Amsterdam following the sacking of Antwerp by Spanish forces in 1575, which ushered in the Dutch Gilded Age and the rise of Amsterdam to become the European (and world) financial capital in the 17th century. Rapid industrialization in the late 18th and early 19th centuries stimulated increasing demand for larger-scale investments, requiring risk capital from outside investors. While many more corporations formed, limited liability, joint-stock corporations remained restricted in numbers until the mid-19th century, when many early industrializing countries began to shift from the government concession system to a system of “free incorporation” that lowered the barriers to entry and standardized regulations on issuing publicly traded equity. Regulation on equity issuance and corporate governance facilitated rapid growth of equity trading starting in the second half of the 19th century. Modern corporations and corporate equity trading largely operate along these lines, albeit with an array of different specific rules and regulations, depending on jurisdiction. A few

1 I am grateful to Refet Gürkaynak, Jonathan Wright, and John Kim for helpful comments and discussions and to Noah MacDonald and Andrew Smith for careful research assistance. 358

Equity trading 

359

key provisions unify the definition of equity as I use the term in this chapter, which focuses on ownership stakes in publicly traded corporations and surveys a wide range of topics relating to the trading of such shares in marketplaces available to the public. All equity confers a claim on the underlying assets of the corporation and therefore a claim on the related cash flows. Equity shares, or “stocks,” can be divided into two broad categories: those with and those without control rights. Common stockholders generally possess control rights, meaning that they may vote in elections of the board of directors and in general meetings of shareholders and thereby influence the selection of management and the policies and activities of the company. They also gain access to any new equity issued before the corporation offers such shares to the public. By contrast, preferred stockholders relinquish most of their control rights in order to gain a more predictable cash flow on their stake.

16.2 PARTICIPANTS Equity trading participants fall into two large categories: “buy side” and “sell side.” 16.2.1 Buy Side The buy side of the market comprises those actively trading (buying and selling) equity and securities. The main actors include institutional investors, individuals, and hedge funds. The term “institutional investor” applies to corporations or other organizations trading securities for accounts owned by groups or institutions, such as insurance companies, mutual funds, exchange-traded funds (ETFs), or pension funds. The term “retail investors” refers to individual investors trading securities with personal accounts (e.g., 401k holders, IRA holders, etc.) Institutional traders currently account for the majority of trades in most markets and usually have the greatest impact on the markets in which they transact. Koesrindartoto et  al. (2020), for example, show that in the context of a developing economy, both institutional and retail investors significantly impact market movements, but institutional investors exert more influence primarily due to their dominant market share (90 percent, and above, of equity ownership in 2015). Impact may stem from differences in trading behavior. While institutional investors are more likely to use momentum-based trading strategies, and trade less often with larger sums and longer holding periods, retail investors tend to use more contrarian strategies, and trade more often with smaller sums and shorter holding periods.2 Retail traders have wielded growing influence in recent years, by some accounts making up as much as 20 percent of US trading volume in the early 2020s. Morgan Stanley estimated its share peaking at 15 percent of trading in the Russell 3000 in September 2020, declining to ten percent of that market in mid-2021. This category accounts for far less trading in Europe, where Euronext estimates that retail trade peaked at seven percent in mid-2020. Abudy (2020) notes that increased market participation by retail traders contributes positively 2 A breakdown on the number of firms and firm sizes can be found at www​.finra​.org​/sites​/default​ /files​/2021​- 06​/21​_0078​.1​_Industry​_ Snapshot​_v8​.pdf (accessed October 26, 2021). According to their classification of retail investor, there were 1,271 retail trading firms registered with FINRA, with 1,096 being small firms (less than 50 employees).

360  Research handbook of financial markets

to stock market liquidity and overall market quality. Well-diversified retail investors trade more when liquidity is higher, but retail trading does not generate price noise at the market level. Moreover, retail investor market entry is uncorrelated with market liquidity. The third category of traders, hedge funds, participate in a wide range of equity trading. They take both short and long positions, theoretically to protect against adverse price movements. While other organizations operated in many respects in a similar manner, the first modern-style hedge fund opened in 1949. The segment exploded in the 1990s and gained fame, both with a few spectacular success stories and with such dramatic failures as the LongTerm Capital Management collapse and bailout in the late 1990s. Hedge funds are less regulated than other actively managed (mutual) funds. The Securities and Exchange Commission (SEC) sets limits on number of investors a hedge fund can serve and bans advertising, but hedge fund managers are exempt from filing public reports. Hedge fund activism, or a fund acquiring equity in a company for the purpose of influencing management, is a heavily studied topic currently. DesJardine and Durand (2020), provide a helpful review of the issues and analyze the impact. They find that short-term performance gains and increased profits from activism are accompanied by long-term costs, including reduced investment spending and social performance. 16.2.2 Sell Side The “sell side” of the market encompasses organizations that facilitate trading and provide information to traders. The key actors include the stock exchanges and the related intermediaries, such as market makers, brokers, and dealers. In recent decades, alternative trading mechanisms beyond the principal exchanges have gained market share as well. Equity trading mechanisms can be split into two categories: order driven and quote driven, alternatively auction markets and dealer, or over-the-counter (OTC), markets. Auction markets simply facilitate connections between buyers and sellers in a centralized location, while dealer markets rely on intermediaries—the market makers or dealers who conduct transactions. The New York Stock Exchange (NYSE) is a textbook example of an auction market. In this type of market, trades are conducted whenever there is a match between the price a prospective buyer is willing to pay (bid), and the price a prospective seller is willing to accept (ask/ offer). Auction markets have utilized different methods of matching buyers and sellers over time. The NYSE originally maintained a call auction format, where interested buyers and sellers of an equity gathered to announce their bids and offers aloud, and a price was set that maximized transaction volume. This time-consuming process was largely replaced with a continuous trading system in 1872 when the number of listed stocks grew too large to announce each one-by-one (Fohlin, 2016). Under a continuous trading system specialists collect buy and sell orders in an (electronic) order book on a rolling basis, and transactions occur whenever there is a willing buyer and seller. Nowadays, the NYSE operates with continuous trading from 9:30 a.m. to 4:00 p.m., then collects bids and offers throughout the night and opens with a call auction before returning to continuous trading. In a dealer market, like the NASDAQ, there are numerous dealers with their own inventories posting both bid and offer prices for given securities in a limit order book. Potential buyers and sellers transact directly with the dealer, not with one another. Since dealers are both

Equity trading 

361

selling and purchasing securities, they make their profits from the bid-ask spread, which is discussed in depth in section 16.6. Each system has its advantages over the other. Notably, traders in continuous auction markets tend to face lower transaction costs, while traders in dealer markets are better protected against execution risk, the possibility that an agreed-upon trade ultimately fails to take place (Pagano and Röell, 1992). Glode and Opp (2020), Lee and Wang (2021), and Dugast et al. (2021) all compare OTC/dealer markets and centralized/auction markets. Dugast et al. (2021) study the choice between trading OTC or in centralized markets across banks with different levels of trading capacity. They find that trading in centralized markets is optimal for intermediate-capacity banks, while OTC trading is optimal for low- and high-capacity banks due to their complementary roles. Lee and Wang (2021) find that OTC dealers cream-skim less-informed traders, leading to a concentration of informed traders in centralized markets. Overall, they document the negative welfare effects of OTC trading, noting that trading OTC is privately optimal for uninformed traders, yet socially harmful. Glode and Opp (2020) find the reverse allocation of informed and uninformed traders, fueled by asymmetric access to counterparties in the OTC market. Dark pool trading, private exchanges for trading securities that are not available to the public, has made notable strides. Financial Industry Regulatory Authority (FINRA) data shows that 46.4 percent of trades took place off-exchange, with 8.9 percent via dark pools, in February 2021. Zhu (2014) argues that dark pools concentrate uninformed traders, thereby improving price discovery in conventional markets. Others, including Comerton-Forde and Putniņš (2014), echo Zhu (2014) that dark trades are less informed than lit trades. These studies argue that high levels of dark pool trading generate adverse selection effects in the lit market due to the increased concentration of informed traders and find that low levels of nonblock dark trading may have a positive effect on informational efficiency, but that high levels are harmful. However, block trades in the dark pool seem to have no effect on informational efficiency. Buti, Rindi, and Werner (2017) also find that the introduction of dark pool trading leads to a welfare reduction. When a dark pool is introduced in competition with an illiquid limit-order book, there is trade creation, but it is accompanied by a reduction in market quality and average investor welfare.

16.3 PRICE SETTING AND MARKET EFFICIENCY The primary output of any market is prices, that is, a consensus determination of fundamental value. In a perfect market, prices equal the present discounted value of all expected future cash flows. In the presence of risk or uncertainty, expectations over future cash flows vary among market participants, as do estimates of the appropriate rate at which their valuations should discount such cash flows. Thus, equity markets facilitate price discovery through the aggregation of information, encapsulated in orders, from traders with potentially disparate estimates of an equity’s value. Traders must continuously revalue their estimates of the possible states of the world (predicted cash flows) and their probabilities, as relevant, new information arrives. These valuation updates cause equity values, and therefore market prices, to change over time, creating positive or negative returns to traders in the process. Traders balance the expected returns against the risk they must assume in deciding which assets to hold or trade. Assets’ returns

362  Research handbook of financial markets

relate systematically with other assets’ returns over time and thereby offer opportunities to diversify away a portion of a portfolio’s risk. A key breakthrough in the field of finance came in a succession of papers by Treynor (1962), Sharpe (1964), Lintner (1965), and Mossin (1966) that worked through the problem of associating securities returns with risk in the context of a market to which traders can avail themselves of a wide array and combinations of assets and portfolios. The resulting theories have become known as the Capital Asset Pricing Model or, simply, CAPM. Essentially, CAPM argues that an asset’s (or portfolio’s) returns are determined only by its systematic risk, specifically, how it varies with the full set of available assets, because, in theory, all non-systematic (or idiosyncratic) risks are diversifiable. The CAPM formula expresses this relationship as follows:

Es = rf + b ( E M - rf ) (16.1)

where Es is the expected return on an asset, r f is the return on a riskless asset, EM is the return rs on the market overall, and b = s —the ratio of the covariance of the asset that is correlated sM with the market to the variance of the market as a whole (Perold, 2004).3 In other words, a risky asset earns the risk-free rate plus a market risk premium. An important implication of CAPM is that the return on an asset depends only on its systemic risk and that such risk is captured by the security return’s relationship with that of the market, its beta. Inherent in the model are assumptions about investor rationality (risk neutrality) and market efficiency. For CAPM to hold, certain assumptions about market efficiency must be true.4 In particular, it must be the case that asset prices reflect all available information. The CAPM’s assumptions lead further to the efficient markets hypothesis (EMH), which can be formulated in three ways—weak, semi-strong, and strong—depending on how strong the assumptions one makes about the extent and speed of information revelation. The weak form of the EMH holds that market prices fully reflect all information contained in historical prices; the semi-strong EMH hold that market prices reflect all publicly available information (in addition to historical prices); and the strong version holds that the market prices reflect private information as well. (Fama, 1970). Considered together, CAPM and EMH present the “joint hypothesis problem.” When anomalous returns are observed, it is difficult to say whether the anomaly represents a failure of CAPM, a failure of EMH, or both. 16.3.1 Empirical Tests of Asset Pricing and Efficient Markets The CAPM and EMH theories spurred a boom in empirical testing of the theories’ predictions. The surprising theoretical finding that a security’s returns should relate only to its systematic risk, its beta coefficient with respect to the market, prompted researchers to seek out and test whether other specific risk factors might show economically meaningful, statistically significant relationships with returns. The traditional empirical test of CAPM begins with estimation of each stock’s beta using a time series regression of stock returns on the return of a broad market portfolio, such as the S&P 500, both in excess of the risk-free rate, over a period of years—usually calculated using monthly returns. The second stage estimates the 3 For a version of CAPM incorporating a stochastic discount factor, see, e.g., Campbell (2014). 4 Perold (2004, 15–16) provides a summary.

Equity trading 

363

cross-sectional relationship between the time series sample average excess return of each stock and the stocks’ betas, using the slope coefficients from the first stage regressions. Empirical studies have found that estimated betas often do not explain returns, or explain a small share, and that a panoply of other factors may hold explanatory power. After hundreds of empirical studies, the literature most consistently finds robust significant additional explanatory power from including factors for company size (market capitalization), under-valuation (book-tomarket ratio of equity), and momentum (the tendency for positive returns to persist and vice versa).5 The empirical asset pricing literature also often suffers from the problem of noisy statistical estimates at the individual stock level, and even with the common use of portfolios of stocks, significant errors in variables remain. The literature in recent decades has worked on a range of theoretical refinements and econometric innovations to improve estimation of and to make use of larger and more complex data sets that are becoming available. Combined with debates over relevant “factors,” recent work offers new approaches to evaluating the statistical power of any new factor (Feng, Giulio, and Xiu, 2020, for example) and estimating efficiency (such as Liao and Liu, 2021, who provide a useful review of the literature) At the same time, an interconnected strand of the empirical asset pricing literature identifying the correct rate for estimating discounted current values, which has led to the use of stochastic discount factors (SDF) to capture time variation, or state contingency, of the appropriate discounting rate. Much of this literature works to estimate the SDF with low measurement error. Fama (2011) summarizes the key points of the literature to that point. More recent work (such as Kim and Korajczyk, 2022 and Pukthuanthong, Roll, and Wang, 2021) proposes estimators of the stochastic discount factor (SDF). Notably, Kim and Korajczyk (2022) use unbalanced panels of individual stock returns, thereby reducing survivor-ship biases, and resurrect the market factor as a significant premium. Another new line of research involves the use of machine learning techniques to identify the most significant model parameters among the growing “factor zoo.” Karolyi and Van Nieuwerburgh (2020) introductory essay surveys a number of new advances published in their Review of Financial Studies special issue on new methods. Notably Gu, Kelly, Xiu (2020) find large economic gains to machine learning forecasts. Chen, Pelger, and Zhu (2021) employ deep learning techniques, and Son and Lee (2022) propose a graph-based multi-factor model. Asset pricing theory also rests on assumptions of market efficiency: that markets instantly incorporate all relevant information into prices. The recognition that market efficiency may fail in the face of real-world market frictions motivates a whole other strand of empirical research. In 2013 Eugene Fama, Lars Peter Hansen, and Robert Shiller shared the Nobel Prize in economics for their studies of empirical asset pricing and analyses of market behavior. Campbell’s (2014) celebratory treatise on the laureates’ key contributions explains how the strands of research relate and continue to impact the field. In addition to the SDF and factor literature discussed previously, the literature has spanned tests of asset price changes and their time series properties, market reaction to information announcements or revelations (typically employing event studies), and an array of behavioral anomalies. In recent years, the masses of available data have created a high dimensional prediction problem for investors while opening 5 Fama and MacBeth (1973) was one of the first contributions. See Fama and French (1993) on factors in stock and bond returns together, using factor loadings based on portfolio returns. See Bodie, Kane, and Marcus (2021, Chapter 13) for a basic textbook overview of the method.

364  Research handbook of financial markets

the door to applications of “big data” methods in predicting stock returns. Martin and Nagel (2022) and Dong et al. (2022) use machine learning and other shrinkage techniques to evaluate hundreds of market “anomalies” and demonstrate a significant gap between in-sample and out-of-sample market predictability. They also provide clear reviews of several related studies in this rapidly growing area of research.

16.4 CLEARING AND SETTLEMENT Regardless of the trading mechanism and resulting prices, whenever traders execute a trade, the transaction must clear and then settle. Clearing is the bookkeeping process of matching and confirming the buy and sell orders that constitute a trade. Settlement finalizes the transaction, transferring the buyer’s funds to the seller and the seller’s securities to the buyer. The trade is finalized through a clearinghouse, clearing division, or central clearing counterparty, which acts as an intermediary between buyer and seller, buying the equity from the seller and selling it to the buyer. Clearinghouses often act as guarantors of the transaction and also perform settlement procedures, collecting payments from buyers and ensuring delivery of the securities from the seller to the buyer. As a result, clearinghouses take on default risk on both sides of every trade to reduce market risk overall. By “netting” overlapping buy and sell orders in the same securities, clearinghouses also minimize the number of physical transfers needed to complete multiple transactions between two parties on a given day, providing a more efficient and streamlined trading environment. They also complete required reporting of transactions. Historically, clearing and settlement took up to two weeks. The time varied by market. When Amsterdam still served as the financial capital of the world in the early 18th century, several stocks on the Amsterdam Stock Exchange were co-listed on the London Stock Exchange. To settle transactions between the two cities, a courier would need to transport the physical security across the English Channel in a journey that took just under two weeks (Koudijs, 2020). This 14-day window became the standard equity settlement period, and several exchanges kept this settlement period for hundreds of years. The NYSE was an outlier in this period, utilizing overnight clearing as far back as the late 1800s and speeding that process to intraday clearing with the opening of a daytime clearing branch in 1920 (Richter, 1920). With the advent of computerized trading in the late 1900s, settlement periods gradually shortened to the present standard of two days, commonly referred to as T+2. Clearing occurs within this two-day window between execution and settlement. Following the financial crisis of 2008 many countries imposed new regulations requiring central counterparty clearing of derivatives contracts, leading to a spate of new research on the subject. Notably, the fact that equity trading has employed central clearing for many decades and even centuries spurred new historical studies to provide some “predictions” of what effects the new regulations might yield. See, for example, Bernstein et al. (2019) on the introduction of the NYSE clearinghouse in 1892 and McSherry et al. (2017) on the lower rate of broker failures following its inception. Today, the NYSE and NASDAQ remain the two largest clearinghouses in the US, but others work alongside the major exchanges, including the National Securities Clearing Corporation (NSCC), which facilitates the majority of broker-tobroker equity trades.

Equity trading 

365

Significant risk remains in the potential for equity trades to fail to settle, either because buyers fail to deliver funds or sellers fail to deliver securities. The slow rate of settlement has given rise to demand for new technologies, such as distributed ledgers in clearing and settlement. The final section of this chapter discuses blockchain technology and its application to equity trade clearing.

16.5 TRADING COSTS Equity trading involves a variety of costs to traders, both direct and indirect, and these costs relate directly to the liquidity of equity trading. Broker commissions are the most clearly visible transaction cost. Traditional brokers charge fees for their services in the forms of account maintenance fees for managed investments, trade-facilitation fees, and interest fees from those borrowing on margin. 16.5.1 Bid-Ask Spreads and Tick Sizes Traders often propose to buy and sell equity at specific prices, in the form of bid and ask quotations. Bid-ask spreads are the difference between a buyer’s maximum willingness to pay (bid) for an asset and the seller’s minimum willingness to accept (ask). Market makers keep this difference. This bid-ask spread is often used as a proxy for market liquidity, since a wider spread implies a higher “round trip” cost to traders. Tick size refers to the minimum incremental change in the price of an asset. Several studies have found that a reduction in tick size leads to a decrease in bid-ask spreads. Chordia, Roll, and Subrahmanyam (2008) examine the NYSE from 93–02 and find that stock price predictability falls with the reduction of tick sizes, signaling increased market efficiency. Open-to-close/close-to-open variance ratios also increase, which the authors take as evidence that more private information is reflected in stock prices as tick size decreases. Griffith, Roseman, and Shang (2020) examine the effect of the SEC’s Tick Size Pilot Program, which increased tick sizes from one cent to five cents, on transaction costs. They note that increased (decreased) equity tick size leads to increased (decreased) bid-ask spreads for affected stocks, but that these effects are diminished in the presence of options trading. Chung, Lee, and Rösch (2020) also study the SEC’s Tick Size Pilot Program. They find, in line with Griffith, Roseman, and Shang (2020), that the quoted and effective spreads of affected stocks increased by an average of 15 basis points as tick-size increased from one to five cents. However, they also report the reverse effect for large trades (2,000+ shares), and state that the program led to (1) an improvement in pricing efficiency, (2) an increase in average trade size, and (3) a decrease the total number of trades. Market makers, those who place buy/sell orders that are not carried out immediately, provide liquidity to the market, and as such pay a lower fee for trades than market takers, who buy or sell instantly. This phenomenon causes spreads to widen to account for fee disparity, generating consistent profits for market makers (Parlour and Rajan, 2003). Brolley and Malinova (2020) find that if a significant portion of investors pay flat commissions, maker-taker pricing leads to a reduction in trading volume and welfare, while maker-taker traders experience heightened profits.

366  Research handbook of financial markets

16.5.2 Technological Change and Competition from New Platforms Trading costs also relate to trading technology and its impact on transaction speed. Technological change sped up trading in several notable stages: first with the telegraph and associated ticker that transmitted pricing information beyond the exchanges. Then in 1878, the first NYSE installed its first telephone, which allowed much faster dissemination of news and prices, albeit in a relatively restricted geographical area until the development of crosscountry calls in 1914. The most dramatic change to equity trading since the telegraph came with the computerization of the markets in the 1960s. As with all markets, the rise of computers and automation reshaped the way equity markets operate. The first fully automated trading system was the Toronto Stock Exchange’s CATS (Computer Assisted Trading System) in 1977. Starting around this time, the NYSE began gradually incorporating automation into their processes, culminating in the transition to a fully automated, screen-based system in 1985. Computer automation allowed the rise of algorithmic trading, in which market participants trade baskets of securities based on predetermined decision rules that trigger buy and sell orders. In 1987 the nascent technology facilitated a rapid sell-off in the equity markets that led to the largest percentage drop up to that point, which brought about the first major financial crisis since the Great Depression. This episode led to the enactment of “circuit breakers” that pause or even halt trading when prices decline more than a given percent within a given period. The rise and spread of internet technology in the 1990s, along with further advances in computer technology, again dramatically altered equity trading at the turn of the 21st century. Algorithmic trading has grown to encompass the majority of equity trading in the US in the last two decades. Despite its prominence, views are mixed regarding the effects of algorithmic trading—and in particular high frequency trading—on trading costs and social welfare. For example, Putniņš and Barbara (2020) find a benign effect, while Aquilina et al. (2021) document strictly negative effects. According to Putniņš and Barbara (2020), algorithmic and high frequency traders are heterogenous and can largely be sorted into toxic and beneficial traders. Toxic traders, who tend to trade in the same direction as institutional traders, can almost double institutional trading costs on their own. However, these negative effects are nullified by liquidity-providing beneficial traders, leading to no net effect on trading costs. On the other hand, Aquilina et al. (2021) focus on a particular practice of high frequency traders known as latency arbitrage, and they document strictly negative effects. Latency arbitrage refers to any obvious arbitrage opportunity where success is essentially determined by how fast one can act on it. For example, if the price of a stock changes significantly at one trading venue but not others, then there is a clear latency arbitrage opportunity where the fastest actor in the market will profit. The authors find that the negative effects of latency arbitrage account for roughly a third of price impact and effective spread. Each technological advance empowered more access for traders, which mostly lowered the direct costs of trading. The latest internet trading technologies have culminated in recent years with the ability of traders to trade directly and bypass traditional market makers. Keim and Madhavan (1998) noted several years ago that the rise of alternative trading platforms may incentivize large traders to reduce trading costs. Colliard and Foucault (2012) concur, noting that the rising prominence of alternative platforms has fragmented trading and left platforms dropping fees to capture a higher market share. They note this has two effects:

Equity trading 

367

(1) available gains from trade increase for investors, but (2) the lower liquidity can incentivize investors to choose limit orders with lower execution probability, thereby reducing investors’ chance of actually trading and capitalizing on the mispricing. Still, as noted previously, recent research also shows that dark pool trading reduces overall welfare. In the post-financial crisis era, new trading platforms, such as Robinhood, have used low and no fee trading via smartphone-based apps to drive new users into equity trading. These platforms ostensibly have the noble goal to make equity trading accessible to a wider swath of the public by lowering barriers to entry. However, their business model involves profiting off aggregating and selling order flow in what is commonly called “payment-for-order-flow.” Victoria (2021) argues that Robinhood does not fulfill the principle of increased equity in markets hinted at by a slogan like “democratizing finance,” but hinders it by creating a gamified, casino-like environment. Tan (2021) finds that this simple, gamified structure leads investors to take on greater amounts of risk than they otherwise would. While the platforms are new, the concept is age-old. Historically, small investors used lowcost, storefront brokers known as “bucket shops” to trade small lots of shares. Hochfelder (2006) highlights the role of bucket shops in increasing participation in financial markets among the broader population. At the same time, these organizations were positioned to front run their clients, who were likely far less informed about the stocks they traded or the market mechanisms.

16.6 FUNDING MARKETS AND EQUITY MARKET LIQUIDITY Equity trading relies heavily on short-term financing, since transactions require funding to bridge the gap between a seller transferring the shares to the buyer and the buyer transferring payment to the seller. In addition, a significant volume of equity trading takes place on margin, meaning that only a relatively small percentage of the purchase price is paid in cash. The remainder is financed through margin lending, which provides substantial liquidity to equity markets. Kahraman and Tookes (2017) document (in India) a positive causal relationship between traders’ ability to borrow and stock market liquidity: liquidity increases when stocks become eligible for margin trading, driven by the contrarian trading strategies of margin traders. They also find that the effect reverses during crises. In one of the most widely cited analyses of the link between equity and funding markets, Brunnermeier and Pedersen (2009) find that equity market liquidity and funding liquidity are mutually reinforcing. The availability of funding affects the number of traders in a market, which affects market liquidity. However, the availability of funding depends on market liquidity. This two-way effect creates mutual reinforcement that can explain liquidity spirals and several features of market liquidity found in the empirical literature: (1) liquidity can dry up suddenly, (2) liquidity of different assets tend to move together, (3) liquidity and volatility are correlated, (4) liquidity is subject to “flight to quality,” meaning that risky assets can become disproportionately illiquid during economic downturns, and (5) market liquidity moves with market performance since funding conditions move with market performance Chiu et al. (2012) provide empirical support for Brunnermeier and Pederson using ETFs in the wake of the 2008 recession. They find that the increase in funding illiquidity during the 2008 recession led to increased bid-ask spreads and decreased market depth, signaling an increase in market illiquidity. These effects are stronger for financial ETFs compared to index

368  Research handbook of financial markets

ETFs. Macchiavelli and Zhou (2021) also find that dealers’ low funding liquidity is correlated with smaller market shares and more frequent trades. Notably, Fohlin, and Gehrig (2021) find similar results in the historical context, showing that funding market illiquidity at the peak of the Panic of 1907 led to increased stock market illiquidity, and that this effect was compounded by market opacity. Given the crucial role of short-term financing in equity trading, empirical research in to asset pricing has studied the possible impact of individual and market level liquidity on stock returns, mostly in the form of additional factors in a standard CAPM “factor model” framework. Amihud and Mendelson (1986) found a cross-sectional relationship between stock-level illiquidity and returns, whereas Chordia et al. (2000) examined the commonality in liquidity, and Pastor and Stambaugh (2003) found a market-wide, time-varying liquidity factor.6 That factor also seems to explain a portion of the momentum factor (discussed previously in the section on empirical asset pricing), although more recent studies undermine that finding (Li, Novy-Marx, and Velikof (2019)). Moreover, the liquidity factor turns up negative in studies of the 1930s through 50s (Pontiff and Singla (2019) and Pastor and Stambaugh (2019)). Notably, the liquidity factor becomes most significant in periods of extreme liquidity freezes, such as the 2008 financial crisis. Still, an interesting open question remains for future research whether it is possible to predict liquidity freezes in advance.

16.7 REGULATION The Securities and Exchange Commission (SEC) is the government agency responsible for ensuring transparency from those offering securities and fair and honest treatment of investors by brokers, dealers, and exchanges. Their official mission is threefold: (1) protect investors, (2) maintain fair, orderly, and efficient markets, and (3) facilitate capital formation. The SEC was created when Congress passed the Securities Exchange Act of 1934 during the height of the Great Depression. At the time, the organization was only responsible for enforcement of the Securities Act of 1933, but the body of enforceable legislation has grown substantially over time in response to various financial scandals and economic downturns. The Financial Industry Regulatory Authority (FINRA) is a nonprofit self-regulatory agency authorized by Congress to oversee the licensing and regulation of broker-dealers. FINRA’s mission is four-fold: (1) write and enforce rules for registered dealers and broker-dealers, (2) ensure firm compliance with these rules, (3) foster market transparency, and (4) educate investors. FINRA is overseen by the SEC, which handles the highest level of appeals from brokerdealers that disagree with accusations or charges levied by FINRA. Unlike the SEC, FINRA was not the first organization of its kind but was instead formed through the merger of the now defunct National Association of Securities Dealers and internal regulators of the NYSE. 16.7.1 Key Regulations on Equity Trading A key regulatory provision in US equity trading is FINRA Rule 5310, the “Best Execution” rule. When executing a trade, the “Best Execution” rule requires brokers and other entities to 6 Le and Gregoriou (2020) provide a thorough survey of liquidity measurement and its role in asset pricing.

Equity trading 

369

“use reasonable diligence” to find the best market for the transaction. Specifically, the broker should consider A) the character of the market for the security (e.g., price, volatility, relative liquidity, and pressure on available communications); (B) the size and type of transaction; (C) the number of markets checked; (D) accessibility of the quotation; and (E) the terms and conditions of the order which result in the transaction.

Brokers and other entities must either review their compliance with the Best Execution rule on an order-by-order basis or conduct a “regular and rigorous” review of the quality of their executed orders. Similarly, the SEC’s Regulation National Market System (Reg NMS) requires brokers to buy and sell on behalf of customers such that they receive at least the National Best Bid and Offer (NBBO).7 Interestingly, neither the Best Execution rule nor Reg NMS prohibit brokers from accepting “payment for order flow” (PFOF). Under PFOF, market makers and transactional counterparties pay brokers a commission for directing trades their way. PFOF is only against the law if the trade price is worse than the NBBO or causes brokers to fail to meet their Best Execution obligations. However, by paying brokers for their clients’ transactions, PFOF creates potential conflicts of interest and has therefore generated controversy since its introduction in the 1990s. Free trading platforms whose business models depend on PFOF, like Robinhood, and their role in meme stock trading have attracted renewed attention to PFOF. The SEC and FINRA are clearly wary of PFOF’s potential for conflicts of interest. (FINRA, 2021). However, unlike the UK’s securities regulators, the SEC and FINRA do not appear poised to ban the practice of PFOF. 16.7.2 Regulations on Equity Issuance (IPOs) An initial public offering (IPO) is the first issuance of publicly traded equity by a corporation. IPOs are typically managed and marketed by investment banks, referred to as underwriters. While much of the leg work in going public is performed by underwriters, the corporation going public ultimately decides the starting share price. In order to issue an IPO, a firm must first file a registration statement with the SEC, then maintain compliance with all SEC regulations throughout their public tenure. The average number of IPO filings from 2010–2020 was just under 200 per year. The number of IPO filings is typically correlated with the overall health of the economy—filings rise in expansion periods and fall during recessions. However, the nearly 500 IPOs filed in 2020 during the coronavirus pandemic stands in stark contrast to this trend but fits with the pattern of hot IPO markets during upswings in equity market valuations. Much of the boom in the latter half of 2020 came in the form of Special Purpose Acquisition Companies (SPACs), which provide funding for mergers and acquisitions, reorganization, or other forms of capital exchanges among existing businesses.

7 See 17 C.F.R. § 242.611; see generally 17 C.F.R. § 242.600 et seq.

370  Research handbook of financial markets

16.7.3 Short Selling and Margin Requirements Investors speculating that the price of a given security will fall over time often take a “short position.” This most often takes place in the form of short selling, which involves borrowing shares from an existing owner (oftentimes from a broker in exchange for fees) and selling those shares with the intention of buying them back after a price decrease. This final step in the process, buying back the shares and returning them to the original owner, is known as “covering the short.” The SEC has placed additional regulations on short selling, the most important of which are part of Regulation SHO.8 Regulation SHO sets four key rules for different investment actors: (1) broker-dealers must clearly mark their investments as long, short, or shortexempt, (2) trading centers must adopt procedures to halt the execution of short sales if prices drop more than 10 percent, (3) broker and dealers must verify that the proper shares can be borrowed in a timely manner before accepting short sale orders, and (4) brokers and dealers must close out short sales according to the guidelines set by the SEC. Together, these tenets help to prevent short sales that intentionally drive down share prices in the market, thereby protecting investors. Critically, Regulation SHO bans short sales placed with the intention of manipulating security prices. While covered short selling is usually permitted, many jurisdictions periodically ban the practice in times of severe downturns in equity prices, such as during the 2020 onset of the COVID-19 pandemic and during the 2008 subprime mortgage “meltdown” and associated financial crisis. Several studies examine the effects of such bans on market liquidity and price discovery. Bessler and Vendrasco (2021) study the implementation of short-selling bans in response to the COVID-19 pandemic in six European countries and find that countries implementing bans experienced lower market liquidity and trading volume and wider bid-ask spreads relative to countries that did not implement a ban. They also note the lack of a price effect for bans, as there was no significant difference between price movements in ban and no-ban countries. Fohlin, Liu, and Zhou (2022) look more closely at liquidity quartiles and find that a priori low-liquidity stocks actually improve liquidity and that the middle 50 percent of stocks see no effect of the bans. Only the most liquid stocks suffer temporary liquidity loss from the bans. Beber and Pagano (2013) similarly study the effects of different short-selling bans imposed around the world in the wake of the 2008 financial crisis. They find that the bans decrease market liquidity, particularly among small-cap and unlisted stocks; slow price discovery, particularly in bear markets; and are not associated with improved stock performance, aside from in the US. Relatedly, the Federal Reserve Board and FINRA, as well as many exchanges, regulate margin trading. While brokerage firms may establish additional rules, they must uphold (in the US) Federal Reserve Board and self-regulatory organization (SRO) rules. For example, rules currently dictate that no more than 50 percent of the purchase price can be borrowed and that a minimum deposit is required for margin trading (the minimum of either $2000 or 100 percent of the margin security purchase price). In the US, the SEC provides comprehensive information on equity trading regulation via its web publications (such as Investor​ .g​ov).

8 17 C.F.R. §§ 242.200–242.204.

Equity trading 

371

16.7.4 Regulatory Failures Among other regulatory shortcomings, the 2008 financial crisis highlighted a number of critical failures in SEC oversight of securities trading. Coffee Jr. (2009) argues that the SEC’s 2004 deregulation of the maximum leverage ceiling contributed to the eventual insolvency of Bear Stearns, Lehman Brothers, and Merrill Lynch in 2008. And, of course, the infamous Madoff affair underscored the lack of sufficient staffing and attention to blatant warnings about fraud. Bernard Madoff operated as one of the largest market makers for the NYSE and was chair of NASDAQ. A well-known and highly regarded figure on Wall Street, he was able to convince thousands of small investors to turn over funds to him to manage in his investment funds. He was, in reality, running a Ponzi scheme and not investing their funds. It took many attempts by whistleblowers to get the SEC to investigate Madoff, and even once they did, they overlooked the scheme. Eventually, Madoff confessed to his crimes and went to prison, but the episode—the largest of its type in US history—elucidated the severe problems with SEC oversight and enforcement of securities trading regulations and investor protection.

16.8 RECENT TRENDS: BLOCKCHAIN TECHNOLOGY AND “TOKENIZATION” The most recent development in equity trading technology is decentralization using blockchains and, more recently, tokenization of equity. Tokens are digital assets whose ownership and transactions are recorded on blockchains. Blockchain is a method for recording and storing transactional and ownership data. Although first conceived of in the early 1990s as a method of recording intellectual property ownership, interest in blockchain exploded following the distribution of Nakamoto (2008) in a cryptography mailing list, which led to the creation of Bitcoin and myriad other cryptocurrencies. Essentially, the well-named blockchain works by gathering transaction information into a block and concatenating these blocks into a sequential chain. Once blocks have been appended to a chain, the information they contain cannot be edited without editing all subsequent blocks. An important dimension along which blockchains can differ is in terms of who can edit the chain by verifying and appending blocks. With distributed ledger blockchains—alternatively, “permissionless” or “public” blockchains—the right/duty to append the chain is crowdsourced to members of the public based on the computing power they apply to arbitrary math problems—as in the case of Proof-of-Work (PoW) protocols such as Bitcoin or Ethereum—or based on their ownership of the blockchain’s native tokens—as in Proof-of-Stake (PoS). By contrast, “private” (or “permissioned”) blockchains reserve the right to amend the blockchain to a single, centralized entity. As one can imagine, coordinating the activities of many distributed actors to accurately maintain a blockchain is no trivial task. Nakamoto relies on heuristics to argue that PoW protocols generate an equilibrium called the Longest Chain Rule (LCR) in which there are no persistent forks and only the longest chain is added to. However, Biais et al. (2019) use game theory to show that while the LCR is a Markov perfect equilibrium, it is not unique; PoW protocols support equilibria with persistent forks. And, indeed, the schism between Ethereum and Ethereum Classic is an example of one such persistent fork. By contrast, Saleh (2021) shows that in a PoS-based blockchain, not only is appending the longest chain a subgame

372  Research handbook of financial markets

perfect equilibrium, but with sufficiently small block rewards, persistent forking equilibria are precluded. Blockchain technology is now being applied in the creation and trading of tokens, which have proliferated as a new way to create and trade equity stakes and other types of digital assets via blockchain. Howell, Niessner, and Yermack (2020) helpfully provide a taxonomy of these different tokens. Coins are intended to serve as a store of value and medium of exchange, security tokens are traditional securities whose ownership is recorded on a blockchain, and a utility token “gives its holder consumptive rights to access a product or service.” It is important to note that these are not perfectly distinct classes of assets; many tokens have a mix of utility, security, and coin features. Gryglewicz, Mayer, and Morellec (2021) model an exchange platform’s decision as to which features should be included in a token issued to finance its development. Roth, Schär, and Schöpfer (2021) provide an overview of the benefits of tokenized equity in a general sense and as it relates to crowdfunding. Tokenized equity examples include Neufund, a blockchain company, and Slice, an Indian credit card company, which have both raised funds by issuing equity tokens. Howell, Niessner, and Yermack (2018) study initial coin offerings (ICOs) and underscore the parallels with IPOs of stocks and the positive real effects of entrepreneurs’ ability to access liquidity through token exchanges. Goforth (2019) provides a survey of how tokenized securities, including tokenized equities, are viewed and treated under US law. In the US, equity tokens are taxed as property, thus assessed according to short-term or long-term capital gains tax. Ulyanava (2018) provides an overview of how cryptocurrencies including tokenized equities are taxed around the world, and details how countries like Belarus are providing tax exemptions to promote adoption. Blockchain technology, however, goes beyond tokens and is a general technological advance useful in financial market clearing. Catyas (2016) discusses the potential for blockchain in clearing and settlement, noting the potential for increased security and speed. Priem (2020) shares this sentiment, noting that blockchain would reduce many of the repetitive components in the clearing process and thereby reduce trading costs. What blockchains have to offer equity trading is the potential of a secure, immutable, and transparent way to record transactions very quickly and at low cost. Chiu and Koeppl (2019) address clearing and settlement on a permissionless, PoW-based blockchain by constructing a model that relates block time, block size, settlement time, and transaction fees. While blockchains may be capable of real-time transaction processing, Chiu and Koepple (2019) describe the tradeoff between consensus and transaction speed: to prevent bad-faith forking (i.e., settlement failure), there needs to be sufficient validation activity; however, there will only be enough validation activity if there is enough congestion to generate sufficiently high transaction fees to compensate the validators. In any case, even the most conservative estimates for blockchain transaction throughput are much, much faster than the current T+2 standard for equity trading. Yermack (2017) points out that blockchains would primarily affect equity trading by greatly increasing liquidity and transparency. Faster transactions with fewer intermediaries would not only reduce direct trading costs (e.g., intermediaries’ fees) but also reduce the need to tie up assets as collateral during the period between trade execution and settlement, which would greatly increase market liquidity. Cong and He (2019) point out that decentralized consensus would necessarily entail more distributed information. While this could lead to fewer informational asymmetries and therefore greater market entry, the greater distribution of information would also facilitate more efficient collusion and punishment of defectors.

Equity trading 

373

Yermack (2017) further highlights the many ways blockchain technology could affect equity trading and corporate governance. For example, while the greater transparency might deter institutional and activist shareholders, who generally prefer some degree of anonymity, the increased liquidity would make it easier for them exercise their “exit” option by selling off stock (or at least threaten to). Yermack (2017) further hypothesizes that blockchains would increase price discovery, better allow observers to infer whether sales were information or liquidity driven and provide greater incentives for outsiders to gather (private) information on firms. One possible implication of blockchain for equity trading seems to have received little attention relative to its potential impact. Combined with smart contracts, blockchain-based equity exchanges could obviate entire classes of financial intermediaries, such as clearinghouses and broker-dealers. The impact of potentially eliminating key players in the financial sector merits further scholarly investigation. Just what effects blockchain might have on equity trading depend largely on the actions of regulators. While tokens potentially implicate various regulatory schemes and regulators, the SEC would remain the most important regulator for blockchain as it relates to equity trading in the United States (or anywhere in the world if securities are offered to US citizens). Under the Securities Act of 1933 and the Exchange Act of 1934, the SEC has jurisdiction over the sale of “securities.” The SEC and the US Supreme Court have interpreted the term securities broadly and functionally enough to encompass just about any token with security features. While regulatory frameworks are still under development and debate as of the early 2020s, the SEC appears open to the idea of blockchains but wary of endorsing its most revolutionary features. In a recent statement, SEC Commissioner Crenshaw extols the gatekeeping and consumer protection functions of financial intermediaries and expresses skepticism about decentralized finance (DeFi) (Crenshaw, 2021). Consistent with this position, it seems likely that the SEC will only authorize more limited versions of permissioned blockchains. Indeed, while the SEC has authorized the US’s first blockchain-based stock exchange, BSTX, its features are limited. Trades on BSTX will pass through a registered clearing agency, traders can only choose same-day trading under certain circumstances, and BSTX only releases certain limited information to its users (SEC, 2022). In so doing, BSTX shuns the radical transparency and disintermediation that attract blockchain’s most fervent supporters. By contrast, the Australian Securities Exchange is planning to replace its Clearing House Electronic Subregister System (CHESS) with a distributed-ledger blockchain-based clearing and settlement system. Originally scheduled to be complete in 2021, ASX has delayed implementation of the CHESS replacement until 2023. The CHESS replacement will be an important test of blockchain technology and its potential impact on securities trading. Once online, it may allow the hypotheses posited and theories developed in the economics and finance literature to be tested empirically.

16.9 CONCLUSION Globally, listed equity market capitalization exceeds $125 trillion as of early 2022 (SIFMA), 42 percent of which traded on US markets. In the US, equities account for 44.2 percent of households’ liquid assets, totaling well over $25 trillion (Federal Reserve Board). Given its impact on individuals and the economy, equity trading attracts widespread public attention

374  Research handbook of financial markets

and constitutes one of the most important areas for academic research. This chapter lays out the key areas of research into equity trading, providing both brief historical background and patterns over time in the relevant institutions and on the current state of research. Among the most promising new areas for research into equity trading include the theoretical and empirical analysis of technological advances in equity trading and related clearing and settlement systems (“fintech”) and the application of machine learning and big data to the empirical analysis of equity pricing and market behavior.

REFERENCES Abudy, M. M. (2020). Retail investors’ trading and stock market liquidity. The North American Journal of Economics and Finance, 54. Retrieved from https://www​.sciencedirect​.com​/science​/article​/pii​/ S1062940820301741. Amihud, Y., & Mendelson, H. (1986). Asset pricing and the bid-ask spread.  Journal of Financial Economics, 17(2), 223–249. Aquilina, M., Budish, E. B., & O’Neill, P. (2021). Quantifying the high-frequency trading. Chicago booth research paper (20–16), 2020–86. Beber, A., & Pagano, M. (2013). Short-selling bans around the world: Evidence from the 2007–09 crisis. The Journal of Finance, 68(1), 343–381. Retrieved from https://www​.jstor​.org​/stable​/pdf​/23324399​. pdf​?refreqid​=excelsior​%3Ad​7394​6cc0​5a54​2957​bc56​5291​b0c6651. Bernstein, A., Hughson, E., & Weidenmier, M. (2019). Counterparty risk and the establishment of the New York stock exchange clearinghouse. Journal of Political Economy, 127(2), 689–729. Bessler, W., & Vendrasco, M. (2021). The 2020 European short-selling ban and the effects on market quality. Finance Research Letters, 42. Retrieved from https://www​.sciencedirect​.com​/science​/article​/ pii​/S1544612320317001​?via​%3Dihub. Biais, B., Bisière, C., Bouvard, M., & Casamatta, C. (2019). The blockchain folk theorem. Review of Financial Studies, 32(5), 1662–1715. Bodie, Z., Kane, A., & Marcus, A. (2021). Investments (12th ed.). New York: McGraw-Hill. Boulton, T. J., & Braga-Alves, M. V. (2020). Short selling and dark pool volume. Managerial Finance, 46(10), 1263–1282. Retrieved from https://papers​.ssrn​.com​/sol3​/papers​.cfm​?abstract​_id​=3882880. Brolley, M., & Malinova, K. (2020). Maker-taker fees and liquidity: The role of commission structures. Working paper. Retrieved from https://papers​.ssrn​.com ​/sol3​/papers​.cfm​?abstract​_id​=3726190. Brunnermeier, M. K., & Pederson, L. H. (2009). Market liquidity and funding liquidity. The Review of Financial Studies, 22(6), 2201–2238. Retrieved from https://academic​.oup​.com ​/rfs​/article​/22​/6​/2201​/ 1592184​?login​=true. Buti, S., Rindi, B., & Werner, I. M. (2017). Dark pool trading strategies, market quality and welfare. Journal of Financial Economics, 124(2), 244–265. Retrieved from https://www​.sciencedirect​.com​/ science​/article​/pii​/S0304405X16300022. Campbell, J. Y. (2014). Empirical asset pricing: Eugene Fama, Lars Peter Hansen, and Robert Shiller. The Scandinavian Journal of Economics, 116(3), 593–634. Catyas, J. (2016). Developing blockchain real-time clearing and settlement in the EU, U.S., and globally. Columbia Journal of European Law. Working paper. https://papers​.ssrn​.com ​/sol3​/papers​.cfm​ ?abstract​_id​=2807675. Chen, L., Pelger, M., & Zhu, J. (2021). Deep learning in asset pricing. Available at SSRN 3350138. Chiu, J., Chung, H., Ho, K., & Wang, G. H. K. (2012). Funding liquidity and equity liquidity in the subprime crisis period: Evidence from the ETF market. Journal of Banking and Finance, 36(9), 2660–2671. Retrieved from https://www​.sciencedirect​.com ​/science​/article​/pii​/S0378426612001550/. Chiu, J., & Koeppl, T. (2019). Blockchain-based settlement for asset trading. Review of Financial Studies, 32(5), 1716. Chordia, T., Roll, R., & Subrahmanyam, A. (2000). Commonality in liquidity. Journal of Financial Economics, 56(1), 3–28. Chordia, T., Roll, R., & Subrahmanyam, A. (2001). Market liquidity and trading activity. Journal of Finance, 56(2), 501–530.

Equity trading 

375

Chordia, T., Roll, R., & Subrahmanyam, A. (2008). Liquidity and market efficiency. Journal of Financial Economics, 87(2), 249–268. Retrieved from https://www​.sciencedirect​.com​/science​/ article​/pii​/S0304405X07001833. Chung, K. H., Lee, A. J., & Rösch, D. (2020). Tick size, liquidity for small and large orders, and price informativeness: Evidence from the Tick Size Pilot Program. Journal of Financial Economics, 136(3), 879–899. Retrieved from https://www​.sciencedirect​.com​/science​/article​/pii​/S0304405X19302831. Coffee Jr., J. C. (2009). What went wrong? An initial inquiry into the causes of the 2008 financial crisis. Journal of Corporate Law Studies, 9(1), 1–22. Retrieved from https://www​.tandfonline​.com​/doi​/pdf​ /10​.1080​/14735970​.2009​.11421533​?needAccess​=true/. Colliard, J.-E., & Foucault, T. (2012). Trading fees and efficiency in limit order markets. Review of Financial Studies, 25(11), 3389–3421. Retrieved from https://academic​.oup​.com​/rfs​/article​-abstract​ /25​/11​/3389​/1566107​?redirectedFrom​=fulltext. Comerton-Forde, C., & Putniņš, T. J. (2014). Dark trading and price discovery. Journal of Financial Economics, 118(1), 70–92. Retrieved from https://www​.sciencedirect​.com​/science​/article​/abs​/pii​/ S0304405X15001191. Cong, L. W., & He, Z. (2019). Blockchain disruption and smart contracts. Review of Financial Studies, 32(5), 1754. Crenshaw, C. (2021). DeFi risks, regulations, and opportunities. International Journal of Blockchain Law, 1, 4–11. Retrieved from https://www​.sec​.gov​/news​/statement​/crenshaw​-defi​-20211109. DesJardine, M. R., & Durand, R. (2020). Disentangling the effects of hedge fund activism on firm financial and social performance. Strategic Management Journal, 41(6), 1054–1082. Retrieved from https://onlinelibrary​.wiley​.com​/doi​/full​/10​.1002​/smj​.3126. Dong, X., Li, Y., Rapach, D. E., & Zhou, G. (2022). Anomalies and the expected market return. The Journal of Finance, 77(1), 639–681. Dugast, J., Üslü, S., & Weill, P. (2021). A theory of participation in OTC and centralized markets. NBER Working Paper 25887. Retrieved from https://www​.nber​.org​/papers​/w25887. Fama, E. F. (1970). Efficient Capital Markets: A review of theory and empirical work. Journal of Finance, 25(2), 383–417. Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 3–56. Fama, E. F. (2011). My life in finance. Annual Review of Financial Economics, 3(1), 1–15. Fama, E., & MacBeth, J. (1973). Risk, return, and equilibrium: Empirical tests. Journal of Political Economy, 81(3). Feng, G., Giglio, S., & Xiu, D. (2020). Taming the factor zoo: A test of new factors. The Journal of Finance, 75(3), 1327–1370. Financial Industry Regulatory Authority. (2021). What we do. Retrieved from www​.finra​.org​/about​/ what​-we​-do (accessed 16 October 2021). FINRA. (2021). Regulatory notice 21-23. FINRA reminds member firms of requirements concerning best execution and payment for order flow. Fohlin, C. (2016). Frictions: Lessons from the history of financial market microstructure. In D. Chambers & E. Dimson (Eds.), Financial market history: Reflections on the past for investors today. (pp. 115– 131). New York: CFA Institute Research Foundation. Fohlin, C., & Gehrig, T. (2021). Stock market liquidity in the panic of 1907. Working paper. Fohlin, C., Lu, Z., & Zhou, N. (2022). Do short sale bans impact liquidity? Evidence from the European Covid-Era Bans, Emory University Mimeo. Glode, V., & Opp, C. C. (2020). Over-the-counter versus limit-order markets: The role of traders’ expertise. The Review of Financial Studies, 33(2), 866–915. https://doi​.org​/10​.1093​/rfs​/ hhz061. Goforth, C. (2019). Securities treatment of tokenized securities under U. S. Law. Pepperdine Law Review, 46(3), 405–470. Retrieved from https://heinonline​.org​/ HOL​/ Page​?handle​=hein​.journals​/ pepplr46​&id​= 429​&collection​=journals​&index=. Griffith, T., Roseman, B., & Shang, D. (2020). The effects of an increase in equity tick size on stock and option transaction costs. Journal of Banking and Finance, 114. https://doi​.org​/10​.1016​/j​.jbankfin​ .2020​.105782. Gryglewicz, S., Mayer, S., & Morellec, E. (2021). Optimal financing with tokens. Journal of Financial Economics, 142(3), 1038–1067.

376  Research handbook of financial markets

Gu, S., Kelly, B., & Xiu, D. (2020). Empirical asset pricing via machine learning. Review of Financial Studies, 33(5). Hochfelder, D. (2006). “Where the common people could speculate”: The ticker, bucket shops, and the origins of popular participation in financial markets, 1880–1920. Journal of American History, 93(2), 335–358. Retrieved from https://academic​.oup​.com​/jah​/article​/93​/2​/335​/830224​?login​=true. Howell, S. T., Niessner, M., & Yermack, D. (2020). Initial coin offerings: Financing growth with cryptocurrency token sales. The Review of Financial Studies, 33(9), 3925–3974. Kahraman, B., & Tookes, H. E. (2017). Trader leverage and liquidity. Journal of Finance, 72(4), 1567– 1610. Retrieved from https://onlinelibrary​.wiley​.com​/doi​/full​/10​.1111​/jofi​.12507. Karolyi, G. A., & Van Nieuwerburgh, S. (2020). New methods for the cross-section of returns. Review of Financial Studies, 33(5). Keim, D. B., & Madhavan, A. (1998). The cost of institutional equity trades. Financial Analysts Journal, 54(4), 50–69. Retrieved from https://www​.jstor​.org​/stable​/4480093​?seq​=1​# metadata​_info​ _tab​_contents. Kim, S., & Korajczyk, R. A. (2022). Large sample estimators of the stochastic discount factor. Available at SSRN 3131274. Koesrindartoto, D. P., Aaron, A., Yusgiantoro, I., Dharma, W. A., & Arroisi, A. (2020). Who moves the stock market in an emerging country — Institutional or retail investors? Research in International Business and Finance, 51. Retrieved from https://www​.sciencedirect​.com​/science​/article​/pii​/ S0275531919300327. Koudijs, P. (2020). Those who know most: Insider trading in eighteenth-century Amsterdam. Journal of Political Economy, 123(6), 1356–1409. Retrieved from https://www​.jstor​.org​/stable​/10​.1086​/683839. Le Bris, D., Goetzmann, W. N., & Pouget, S. (2015). The development of corporate governance in Toulouse: 1372–1946. NBER working paper. Le, H., & Gregoriou, A. (2020). How do you capture liquidity? A review of the literature on low‐frequency stock liquidity. Journal of Economic Surveys, 34(5), 1170–1186. Lee, T., & Wang, C. (2021). Why trade over-the-counter? When investors want price discrimination. Working paper. Retrieved from https://www​.google​.com​/url​?q​=https​%3A​%2F​%2Fwww​ .dropbox ​ .com​ %2Fs​ %2F11f3rdjq0fh5gpl​ %2FOTC ​ _ EX ​ .pdf ​ %3Fdl​ %3D0 ​ & sa ​ = D​ & sntz ​ =1​& usg​ =AFQjCNFmCBOAE​-uKg​93B4​cJts​xggp7gt8w. Li, H., Novy-Marx, R., & Velikov, M. (2019). Liquidity risk and asset pricing. Critical Finance Review, 8(1–2), 223–255. Liao, Z., & Liu, Y. (2021). Optimal cross-sectional regression. Available at SSRN 3719299. Lintner, J. (1965). The valuation of risky assets and the selection of risky investments in stock portfolios and capital budgets. Review of Economics and Statistics, 47(1), 13–37. Macchiavelli, M., & Zhou, X. (2021). Funding liquidity and market liquidity: The broker-dealer perspective. Management Science. Retrieved from https://papers​.ssrn​.com​/sol3​/papers​.cfm​?abstract​ _id​=3311786; https://pubsonline​.informs​.org​/doi​/abs​/10​.1287​/mnsc​.2021​.4053. Martin, I. W. R., & Nagel, S. (2021). Market efficiency in the age of big data. Journal of Financial Economics, 145(1), 154–177. McSherry, B., Wilson, B. K., & McAndrews, J. J. (2017). Net settlement and counterparty risk: Evidence from the formation of the New York Stock Exchange Clearing House in 1892. Journal of Money, Credit and Banking, 49(6), 1273–1298. Mossin, J. (1966). Equilibrium in a capital asset market. Econometrica, 34(4), 768–783. Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system [Unpublished manuscript]. Retrieved from https://bitcoin​.org​/ bitcoin​.pdf. Pagano, M., & Röell, A. (1992). Auction and dealership markets: What is the difference? European Economic Review, 36(2–3), 613–623. Parlour, C. A., & Rajan, U. (2003). Payment for order flow. Journal of Financial Economics, 68(3), 379–411. Retrieved from https://www​.sciencedirect​.com​/science​/article​/pii​/S0304405X03000710. Pástor, Ľ., & Stambaugh, R. F. (2003). Liquidity risk and expected stock returns. Journal of Political Economy, 111(3), 642–685. Perold, A. F. (2004). The capital asset pricing model. Journal of Economic Perspectives, 18(3), 3–24. Pontiff, J., & Singla, R. (2019). Liquidity risk? Critical Finance Review, 8(1–2), 257–276.

Equity trading 

377

Priem, R. (2020). Distributed ledger technology for securities clearing and settlement: Benefits, risks, and regulatory implications. Financial Innovation, 6(1). Retrieved from https://jfin​-swufe​ .springeropen​.com ​/articles​/10​.1186​/s40854​- 019​- 0169​-6. Pukthuanthong, K., Roll, R. W., & Wang, J. L. (October 3, 2021). An agnostic and practically useful estimator of the stochastic discount factor. Available at SSRN.  Retrieved from https://ssrn​.com​/ abstract​=3503974 or http://dx​.doi​.org​/10​.2139​/ssrn​.3503974. Putniņš, T. J., & Barbara, J. (2020, December). The good, the bad, and the ugly: How algorithmic traders impact institutional trading costs. NBER Big Data and Securities Markets Conference. Richter, F. E. (1920). The stock clearing corporation. The Quarterly Journal of Economics, 34(3), 538– 544. https://doi​.org​/10​.2307​/1883366. Rohr, J., & Wright, A. (2019). Blockchain-based token sales, initial coin offerings and the democratization of public Capital Markets. Hastings Law Journal, 70(2), 463. Roth, J. et  al. (2021). The tokenization of assets: Using blockchains for equity crowdfunding. In K. Wendt (ed.), Theories of change: Change leadership tools, models and applications for investing in sustainable development (pp. 329–350). Retrieved from https://books​.google​.com​/ books​?hl​=en​ &lr=​&id=​-g8tEAAAQBAJ​&oi​=fnd​&pg​=PA329​&dq​=tokenized​+equity​&ots=​-MJpCWNSqd​&sig​ =A​​UaT1U​​Q1lyW​​yvpxz​​fu-​-C​​CZpxM​​A​# v​=o​​nepag​​e​&q​=tokenized​%20equity​&f​=false. Saleh, F. (2021). Blockchain without waste: Proof-of-stake. The Review of Financial Studies, 34(3), 1156–1190. Schulhofer-Wohl, S. (2021). Externalities in securities clearing and settlement: Should securities CCPs clear trades for everyone?, Policy Discussion Paper Series PDP-2021-02, Federal Reserve Bank of Chicago. SEC. (2022). Notice of filing of amendment Nos. 2 and 3 and order granting accelerated approval of a proposed rule change, as modified by amendment Nos. 2 and 3, to adopt rules governing the trading of equity securities on the exchange through a facility of the exchange known as BSTX LLC. Release No. 34-94092; File No. SR-BOX-2021-06. Sharpe, W. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. Journal of Finance, 19(3), 425–442. Smith, M. B. (2003). A history of the global stock market: From ancient Rome to Silicon Valley. University of Chicago Press. Retrieved from https://books​.google​.com​/ books​?hl​=en​&lr=​&id​ =KJBJzXLEQggC ​ & oi​ =fnd ​ & pg​ = PP6 ​ & dq ​ = stock​+market​ +history​& ots ​ = MFAtkqXQ8T​& sig​ = 9​​ ClGpm​​8be9o​​DBb​-b​​rGCx2​​2​-3Q_​​Y​#v​=o​​nepag​​e​&q​=stock​%20market​%20history​&f​=false. Son, B., & Lee, J. (2022). Graph-based multi-factor asset pricing model. Finance Research Letters, 44, 102032. Tan, G. K. S. (2021). Democratizing finance with RobinHood: Financial infrastructure, interface design and platform capitalism. Environment and Planning A: Economy and Space. Retrieved from https:// journals​.sagepub​.com​/doi​/abs​/10​.1177​/0308518X211042378. Treynor, J. L. (1962). Toward a theory of market value of risky assets. Unpublished manuscript. Final version in R. A. Korajczyk (ed.), Asset pricing and portfolio performance, 1999 (pp. 15–22). London: Risk Books. Ulyanava, K. (2018). Legal regulation of the crypto-currency taxation. Open Journal for Legal Studies, 1(1), 1–8. Victoria, V. (2021). False idols: The perils of ‘democratizing’ financial markets. Unpublished Master’s thesis. Retrieved from https://shareok​.org​/ bitstream​/ handle​/11244​/330102​/2021​_Victoria​_Violet​ _Thesis​.pdf​?sequence​=10​&isAllowed​=y. What we do. U.S. securities and exchange commission. Retrieved from https://www​.sec​.gov​/about​/what​ -we​-do (accessed 16 October 2021). Yermack, D. (2017). Corporate governance and blockchains. Review of Finance, 21(1), 7–31. Zhu, H. (2014). Do dark pools harm price discovery? Review of Financial Studies, 27(3), 747–789. Retrieved from https://www​.mit​.edu/​~zhuh​/Zhu​_darkpool​_RFS​.pdf.

17. Sovereign debt Leonardo Martinez, Francisco Roch, Francisco Roldán and Jeromin Zettelmeyer1

17.1 INTRODUCTION1 Sovereign debt—debt issued by national governments—is an unusual sort of asset. In the context of the domestic financial system, it is often viewed as safe and liquid—at least safer and more liquid than privately issued debt. The deep reason for this is the sovereign’s power to tax future income. Buying and selling sovereign debt help economic agents cope with liquidity shocks, smooth consumption, and realize investment opportunities. But there is also another perspective on sovereign debt. This focuses on the difficulty of enforcing sovereign debt contracts, particularly if the holders of debt are foreign investors. The problem is not necessarily the lack of a legal framework: sovereign debtors can issue debt in foreign jurisdictions whose legal systems are removed from the influence of the debtor. Rather, the problem is that in the event of default, foreign courts rarely have the power to force a sovereign to hand over assets, since most of these assets are located within the sovereign’s borders. This raises the question of why externally held sovereign debt can exist at all. The first perspective dominates in advanced countries, in which large volumes of tradeable debt are held by domestic residents that the government cannot politically afford to default on. The second tends to dominate in emerging markets and developing countries, where political institutions may be weaker and external and domestic debt markets may be segmented. Importantly, however, both perspectives are two sides of the same coin, namely, the extraordinary power of the sovereign. In the first case, what makes sovereign debt special is the sovereign’s power to tax; in the second, its power over its own territory and assets. This chapter aims to help readers understand the assumptions and drivers behind both characterizations of sovereign debt, and what they imply for the economy—that is, the benefits of safe sovereign debt and the costs of living with sovereign risk. It does so by providing an overview of the enormous literature on sovereign debt held by private creditors. This literature dates back to the early 1980s and was originally focused on the question of why externally held sovereign debt can exist at all—or equivalently, on the costs of default. More recently, the literature has tried to measure these default costs and disentangle them from the output dislocation that might be inducing default, to study how the presence of sovereign risk changes the properties of the macroeconomy in emerging markets, and (very recently) to shed light on the valuation and sustainability of sovereign debt in advanced countries.

1 We are grateful to our discussant Daniel Bergstresser, the editors, Refet Gurkaynak and Jonathan Wright, and Tim Willems for helpful comments and suggestions. The view expressed in this article are those of the authors and do not necessarily represent the views of the IMF, its Executive Board, or IMF management. 378

Sovereign debt 

379

This chapter starts by summarizing some of the unique characteristics of sovereign borrowing, and then briefly describes the main institutional features of the sovereign debt market. It then discusses the uses and implications of default-risk-free sovereign debt. This is useful both as a benchmark and because it has been the empirically relevant case for many advanced countries. The chapter then switches gear, and recalls answers to some of the core questions of the literature on externally held sovereign debt: what are the costs of default? What causes defaults? Finally, the chapter describes the implications of sovereign risk—living in the shadow of possible default, even if default never materializes. It ends by describing some strategies for how sovereign risk might be reduced.2

17.2 WHY SOVEREIGN BORROWING IS DIFFERENT Sovereign borrowing and borrowing by private parties (households and corporations) have broadly similar motives. Like private agents, governments borrow to finance long-lived investments. Furthermore, just like households borrow to smooth their consumption through periods of temporary hardship, there are good reasons for governments to borrow in recessions and reduce public debt in good times. Countercyclical borrowing supports private consumption smoothing by maintaining government services that cannot be easily substituted (such as law enforcement, public health, and public education), and by stabilizing the income of households that cannot easily borrow themselves (for example, by keeping public employment stable and paying unemployment benefits in recessions). At the same time, it avoids the distortions that would be created by fluctuating tax rates (Barro, 1979), and the contractionary effect of raising taxes in the middle of a recession.3 But sovereign borrowing also has distinctive features. These imply that the costs of borrowing by sovereigns may not be the same as the borrowing cost of private parties, and that the economics of sovereign defaults may differ from that of personal or corporate bankruptcy. Consider first the case where sovereign debt is default-risk free. This case is empirically relevant for countries with large domestic debt markets and domestic institutions that enforce repayment (e.g., because governments need to respect constitutional principles or because they face prohibitive political penalties from defaulting on their citizens). In such cases, the key difference between privately issued and sovereign debt is that the latter is both safer and more liquid. While private domestic borrowers can post collateral, they often do not have sufficient collateral to fully eliminate borrowing constraints. In contrast, sovereign debt is backed by a “collateral” of sorts: the power to tax in the future. One implication, which we elaborate on in Section 17.4, is that governments that can issue “safe” debt may be able to sustain positive debt without ever having to generate primary surpluses to repay it, as investors are willing to hold this debt because of its unique safety and liquidity features. In emerging and developing economies, as well as some advanced countries in crisis times, however, sovereign debt is not safe. One reason for this is that the repayment of sovereign 2 Some of the material presented in Sections 17.2 and 17.4–17.7 of this chapter draws on surveys by Sturzenegger and Zettelmeyer (2007), Hatchondo et al. (2007), Panizza et al. (2009), and Willems and Zettelmeyer (2022). 3 As we explain in Section 17.8, sovereign risk challenges this view of the optimality of countercyclical fiscal policy.

380  Research handbook of financial markets

debt may be more difficult to enforce, particularly when the debtholders are foreign investors. While domestic courts can enforce corporate and household debt, they are subject to laws that can be changed by the sovereign. Foreign courts are outside the reach of a sovereign, but they cannot dictate repayment. What they can do, and have increasingly done since the 1990s, is to issue an order which gives a private creditor the right to attach sovereign assets. Most of these assets, however, are located inside the sovereign’s national borders, where the sovereign will not allow them to be attached. Hatchondo and Martinez (2011) and Wright (2002) discuss how attempts to attach sovereign assets have had limited success. Finally, an important difference between sovereign and privately issued debt is that the former is affected by politics (Alesina & Tabellini, 2005; Persson & Svensson, 1989). For example, a politician who cares mostly about the period during which she will be in office may not fully internalize the costs of issuing debt. Moreover, some governments could decide to borrow strategically to bind the hands of future governments with different political preferences. To summarize, sovereign debt is very different from privately issued debt, but the differences do not all point in the same direction. They depend on the institutional and political characteristics of the issuer, the identity of the creditor, the liquidity of the domestic debt market, and the integration of domestic and international debt markets. These characteristics can have an important influence on whether sovereign debt is a boon or a source of problems. An important question, which we will return to at the end of this survey, is whether policy can influence these characteristics.

17.3 SOVEREIGN BOND MARKET BASICS The sovereign debt market is the oldest and largest bond market in existence. As of end-2018, the amount of global debt securities issued by general governments exceeded $45 trillion (BIS, 2019). Securities-level data on sovereign borrowing is available from the 1800s, particularly for advanced economies (Flandreau et al., 2010). Hall et al. (2021) construct series for all US debt since independence. Meyer, Reinhart, and Trebesch (2022) compile a database of foreign-currency government bonds traded in London and New York since 1815, covering 91 countries. Government debt plays a crucial role in financing governments worldwide and serves as a benchmark for capital costs and long-term bank loans (Neumeyer & Perri, 2005; Mendoza & Yue, 2012). Government bonds are initially sold in primary markets with the purpose of raising funds. Auctions (e.g., Dutch style or minimum-price offering) are widely used in developed countries as they prove more cost-effective and transparent, while countries in which a smaller number of bidding institutions might create collusion concerns tend to resort to syndication, underwriting, or tap sales (Kimmel, 2019). Governments structure issuances to affect their redemption profile across the yield curve as well as to control their rollover, exchange rate, and interest rate risks. Secondary markets convert government securities which arise from long-term financing needs into the liquid instruments demanded by market participants for portfolio or collateral purposes. Monitoring secondary markets is critical to the government’s debt strategy, as new information is typically reflected in secondary market prices. Primary and secondary markets support each other, as higher liquidity in secondary markets improves participation (and prices) in primary markets as securities become easier to offload (see also Passadore & Xu, 2020) while issuing at key maturities in primary markets

Sovereign debt 

381

can support the growth of secondary markets by creating a benchmark yield curve to anchor prices. Sovereign bonds differ with respect to their governing law (domestic or foreign), financial terms (currency denomination, maturity, amortization profile, floating or fixed rate, nominal or CPI-indexed), and non-financial contractual provisions such as listing requirements; cross-default, acceleration, and negative pledge clauses; and majority restructuring provisions. These provisions allow bonds to be traded, protect investor interests and in some cases determine conditions for changing the payment terms. They typically follow a market standard that tends to change slowly (Gelpern, Gulati, & Zettelmeyer, 2019). Advanced countries almost always issue under domestic law, while emerging markets and developing economies (EMDEs) have tended to issue under foreign law (typically New York or English). However, the share of domestic debt in total debt of EMDEs has been rising from about 30 percent to 46 percent between 2000 and 2020 (IMF, 2021). Currency composition is a major issue influencing the risk of sovereign bonds from the perspective of both the issuer and the borrower. While most advanced economies issue debt in their own currencies, the governments of emerging and developing markets often find this difficult or too expensive. Most foreign-currency debt is denominated in US dollars, with Japanese yen, euros, UK sterling, or Swiss francs playing minor roles (Arsanalp et al., 2019). The causes and consequences of foreign-currency issuance, sometimes referred to as “original sin,” studied by a long literature pioneered by Eichengreen and Hausmann (1999), are still a matter of research and debate to which we return in Section 17.10. More recently, some emerging countries have managed to develop local capital markets and issue bonds denominated in local currency (see, for example EBRD, 2010). Since government debt is usually traded in secondary markets, the holders of bonds are difficult to track. Non-residents may participate in domestic bond markets and vice versa. As Broner, Martin, and Ventura (2010) point out, this effectively removes a sovereign’s ability to impose different terms of repayment to different agents (selective defaults). OECD (2019) compile information on the investor base for sovereign debt of OECD countries and how it has changed over time. They highlight the increasing role of domestic and foreign central banks and institutional investors (insurance companies, investment funds, pension funds) in the 2010s, as well as how domestic central bank demand seems to crowd out domestic private bank demand. Arslanalp and Tsuda (2014, updated) provide similar decompositions for EMs.

17.4 SAFE SOVEREIGN DEBT4 This section focuses on the case of an economy in which investors regard government debt as default-risk free. In such a setting, sovereign debt could be “special” in the sense that it provides liquidity services to investors facing borrowing constraints. This could have important implications for the pricing and sustainability of sovereign debt. To see this, consider the benchmark case in which financial markets are complete and there are no constraints to private borrowing. In this setting, it can be shown that in the presence of rational investors—which rules out “Ponzi games” in which sovereigns roll over their debts forever—the market value of outstanding debt must be equal to the net present value of 4 This section draws on Willems and Zettelmeyer (2021).

382  Research handbook of financial markets

expected future primary surpluses calculated using stochastic discount factors (Bohn, 1995). This relationship is sometimes referred to as the (model-based) intertemporal government budget constraint (IGBC). The use of stochastic (“risky”) discount rates rather than risk-free rates reflects the fact that—although debt is default-risk free—the market value of debt fluctuates, exposing the investor to risks.5 In a setting in which investors/consumers are borrowing constrained (e.g., because of scarce collateral), buying a liquid asset which can be sold in times of need is the only way to smooth consumption. It may also be the only way of ensuring that investors/consumers have sufficient liquidity to realize productive investment opportunities when these arise (Woodford, 1990). As a result, sovereign debt may “crowd in” investment and raise growth. These welfare enhancing effects—consumption smoothing, higher investment—arise independently of any welfare enhancing effects due to government spending financed by debt. In principle, liquidity could also be provided by private financial intermediaries. However, Holmstrom and Tirole (1998) show that when the liquidity needs of investors/consumers are correlated and private borrowing is constrained, aggregate shocks will result in a shortage of private liquidity. In such a setting, the government can achieve a Pareto improvement by issuing sovereign debt. In effect, the power to tax consumers’ future endowments enables the government to commit funds on behalf of private agents. Sovereign debt offers a workaround that enables private agents to borrow against their future income after all. As a result, holding sovereign debt provides a service to investors/consumers that exists independently from its value as a claim on future primary surpluses. This insight has important implications for the question of whether today’s debts in advanced economies can be repaid without some combination of high inflation, financial repression, or even default (contradicting the assumption that advanced country debt is default-risk free). Several authors, such as Olijslagers et al. (2020) and Jiang et al. (2021), have argued that there is an inconsistency between today’s high debt levels, primary fiscal balances that are expected to remain in deficit far into the future, and low (expected) inflation. In other words, modelbased forecasts of primary surpluses and stochastic discount factors suggest that the IGBC, value of debt stock = E{PV(future primary surpluses)} is violated. Jiang et al. (2021) refer to this as the “valuation puzzle.” One interpretation for this is that markets are expecting higher future primary surpluses than seems to be justified based on the past fiscal behavior of countries. Validating these market expectations will require atypically large fiscal adjustments. If governments are not capable of such adjustments, the IGBC will eventually be restored through an unexpected—and possibly disruptive—event that lowers the value of the debt stock. However, as observed in recent papers by Berentsen and Waller (2018), Brunnermeier et al. (2020), and Reis (2021), these conclusions might change if the liquidity services provided by safe government debt in the presence of private borrowing constraints are taken into account. In that case, the intertemporal government budget constraint contains an extra right-hand-side term:

5 In particular, the sovereign will seek to repay debt in booms, when the marginal utility of consumption is low, and issue additional debt in recessions, when the marginal utility of consumption is high. This debt needs to be absorbed by private investors.

Sovereign debt 



{

}

value of debt stock = E PV ( future primary surpluses )

{

}

+ E PV ( future service flow ) .

383

(17.1)

This implies that positive levels of debt could be sustainable even if the government never produces primary surpluses. The fact that government debt provides liquidity services enables an issuing sovereign to mine a (finite) bubble. Importantly, however, the expanded IGBC will continue to set a limit for sustainable primary deficits—it is just that this limit is looser in the presence of a service flow. This implies that “overmining” the bubble—pushing the debt above what seems justified by the right-hand side of the expanded IGBC—could be dangerous. To the extent that the fiscal adjustment that is required to bring the present value of future primary surpluses back in line with debt stock remains plausible, markets may expect such fiscal adjustment, allowing the government to temporarily “overmine” the bubble. But beyond that point, the bubble might burst if investors lose faith in the safe asset status of the debt (Farhi & Maggiori, 2018). The result would be a sharp, disruptive tightening of the IGBC. In light of this risk, governments for whom the market value of the debt significantly exceeds the present value of forecast future primary surpluses should start thinking about managing down their debt levels.

17.5 SOVEREIGN DEBT RESTRUCTURINGS AND DEFAULTS We now turn to the case when debt is not default-risk free. To understand the economic implications, it is worth starting by describing some of the stylized facts and costs associated with sovereign default, before turning to the implications of sovereign risk (even in the absence of default) for economic performance. There are different definitions of a sovereign default. From a legal perspective, a default event is a contractually specified breach of the debt contract—most notably, a failure to pay scheduled debt service beyond the grace period specified in the contract. On this basis, the academic literature and policy institutions such as the IMF distinguish between “preemptive” debt restructurings (renegotiation of debt terms before a payment has been missed) and “post-default” restructurings. According to Asonuma and Trebesch (2016), 38 percent of debt restructurings between 1978 and 2010 were preemptive (i.e., prior to a payment default). Not surprisingly, preemptive restructurings tend be quicker, with an average negotiation time of 12 months, than post-default restructurings (60 months). In contrast, credit-rating agencies tend to define as default an episode in which the sovereign makes a restructuring offer at terms that are less favorable than the original debt.6 Sovereign defaults and restructurings have occurred in cycles, reflecting the boom-bust nature of international capital flows. According to Mitchener and Trebesch (2021), there are four main peaks in emerging market defaults in the last 200 years: in the 1830s, the 1880s, the 1930–40s, and the 1980s. More than half of emerging market countries were in default during these periods. All waves included many Latin American countries, but the 1930–40s wave triggered by the Great Depression was global, and the 1980s defaults included many 6 Peter (2002) discusses the rating agencies’ definitions of default.

384  Research handbook of financial markets

African and some Asian developing countries. Advanced countries defaulted less frequently. At least ten percent of advanced economies were in default during and immediately after the Napoleonic wars in the early 19th century, during a long period from the 1830s to the 1870s, and during the 1930s–40s. In 2012, Greece became the first advanced country since the 1960s to undergo a deep debt restructuring (Zettelmeyer, Trebesch, & Gulati, 2013). A debt restructuring typically leads to new payment promises involving a combination of lower principal, lower interest payments, and longer maturities (according to Dvorkin et al., 2020, by 3.4 years on average). The promised cash flows can be summarized as a present value, which is typically evaluated at the secondary market yield prevailing immediately after a debt restructuring (the “exit yield”). The losses suffered by investors, known as the “haircut,” can be measured as the percentage difference between this present value and the value of the pre-restructuring debt. There are two conventions for computing the haircut. Market practitioners tend to compare the present value of the new debt with the face value of the old debt (HM):

HM = 1 -

Present value of new debt obtained in the restructuring . (17.2) Face value of old debt surrendered in the restructuing

Apart from convenience, a justification for this approach is that in defaults (but not in predefault restructurings) old debt is typically accelerated, i.e., becomes due and payable immediately. At that moment, the face value and the present value of the old debt coincide. An alternative measure, proposed by Sturzenegger and Zettelmeyer (2007; 2008) and used in much of the empirical academic literature, compares the present value of the new and old (originally promised) payment stream, both evaluated at the exit yield (HSZ):

HSZ = 1 -

Present value of new debt obtained in the restructuring . (17.3) Present value of old debt surrendered in the restructuing

The HSZ measures the loss experienced by creditors accepting a restructuring exchange offer compared to a counterfactual in which they would have been repaid with the same probability as the investors accepting the offer. If the accepting creditors’ decision was rational, they were expecting to lose more than HSZ by rejecting the restructuring offer. In that sense, HSZ is a measure of the harshness of the restructuring.7 Figure 17.1 presents HSZs computed by Meyer et al. (2022) for more than 300 sovereign debt crises since 1815.8 They find an average HSZ of 44 percent. Haircuts are typically smaller in preemptive debt restructurings than in post default restructurings (18 versus 48 percent according to Asonuma & Trebesch, 2016) and larger for short-term debt than for long-term debt (Asonuma et al., 2021). 7 Because HSZ discounts the face value (expected amortizations) of the old debt using the (fairly high) exist yield, it tends to deliver lower estimates of investor losses than HM. Hatchondo et al. (2014) question the interpretation of HSZ as a measure of investors’ losses associated with a debt restructuring, showing that while the crisis that preceded a restructuring lowered the market value of sovereign debt, the restructuring often seems to have increased this market value. 8 Note that the figure shows some events with negative HSZ. According to Meyer et al. (2022), these events occurred in early stages of a crisis when to avoid defaulting on debt payments, sovereigns may be willing to postpone payments even at a very high interest rate. They were typically followed by other restructurings with larger haircuts. Cases of total debt repudiation (HSZ = 100 percent) correspond to extreme events such as wars, revolutions, or the break-up of empires.

Sovereign debt 

385

Note:   Size of each circle represents the nominal value of debt restructured, expressed in 2009 US$. Source:   Meyer et al. (2022).

Figure 17.1  Haircuts in sovereign debt restructurings with foreign private creditors since 1815

17.6 THE COSTS OF SOVEREIGN DEFAULTS If there were no costs of defaulting, sovereigns would always want to default. Anticipating this, investors would never lend to sovereigns and there would be no sovereign debt. Hence, for sovereign debt to exist, it must be more costly for a sovereign to default than to pay back its debt in at least some circumstances. Conversely, for sovereign defaults to exist, there must be some circumstances in which it is more costly for a sovereign to pay back its debt than to default. The insight that the existence of sovereign debt requires default costs has motivated an extensive literature, both theoretical and empirical, that tries to pinpoint such costs. These can be grouped into four classes (Panizza et al., 2009): financial penalties in the form of higher borrowing costs and/or capital market exclusion, direct sanctions and trade costs, reputational spillovers, and domestic financial and political costs. 17.6.1 Access to and Costs of External Borrowing The paper that created the modern sovereign debt literature, by Eaton and Gersovitz (1981) showed that in the absence of any other enforcement mechanism, defaults could be deterred by the threat of permanent exclusion of the borrower from international capital markets. As

386  Research handbook of financial markets

the literature quickly pointed out, this radical punishment is implausible on both theoretical and empirical grounds: theoretically, because it benefits both creditors and the debtor to eventually resume borrowing; and empirically, because indefinite capital market exclusion of a defaulting country from international capital markets is extremely rare. This said, temporary exclusion periods can be both theoretically rationalized (Cole et al., 1995; Kletzer & Wright, 2000; Wright, 2002) and are empirically observed (Cole et  al., 1995; Cruces & Trebesch, 2013). Furthermore, sovereign borrowing costs rise after defaults, although the effects are temporary (Özler, 1993; Flandreau & Zumer, 2004). Cruces and Trebesch (2013) find that the extent of the rise is associated with the investor losses in the default, with a one standard deviation increase in investor losses associated with a medium-term increase in the costs of borrowing by 120 basis points. The effect can no longer be detected after about seven years. In the past 20 years or so, exclusion from international capital markets has become more effective as a channel of legal enforcement of sovereign debt contracts issued under foreign law. While court orders rarely lead to successful attachment of assets of the debtor, they have sometimes allowed the holders of defaulted bonds to interfere with cross-border payments to other creditors who had previously agreed to a debt restructuring. Most famously, a 2011 New York court decision ordered Argentina not to pay the holders of restructured bonds unless “holdout creditors” that had refused to accept these bonds were also paid. This order was enforced by threatening to sanction private parties that helped Argentina evade the order (Gelpern, 2013), eventually forcing Argentina to settle with its holdouts. Schumacher et al. (2021) document that creditor lawsuits have become increasingly common in sovereign debt markets and that these lawsuits disrupt government access to international capital markets, as foreign courts are able to impose financial embargos on sovereigns. 17.6.2 Direct Sanctions and Trade Costs Governments have, on occasions, intervened actively in support of their constituents who are holders of defaulted debt issued by other governments. These interventions have taken the form of diplomatic dissuasion, withholding of official credit, threat of trade sanctions, and in exceptional cases, armed interventions. Mitchener and Weidenmier (2005) document about a dozen cases of sanctions of this type during the period 1870–1914. There is also substantial evidence that sovereign defaults disrupt international trade (Rose, 2005; Asonuma et al., 2016; Serfaty, 2020). Why this happens remains unclear, however. 20th and 21st century defaults no longer lead to trade sanctions, and the evidence does seem to back a link via the reduction in trade finance (Borensztein & Panizza, 2009). 17.6.3 Default as a Negative Signal about the Government or the State of the Economy Several studies argue that a sovereign default is costly because of the information it signals. For example, a default decision could be seen as revealing information about policy preferences, such as that the government being willing to ignore property rights. A default could also be viewed as signaling the government’s private information about the weak state of the economy. Besides increasing the cost of future government borrowing, these signals could have negative consequences for the broader economy. Cole and Kehoe (1998) argue that a sovereign

Sovereign debt 

387

default may imply that the government is considered untrustworthy in other areas besides the credit relationship with lenders. One consequence may be capital flight. In Sandleris (2008), the negative information about the state of the economy revealed by the default decision depresses the net worth of firms, with negative implications for investment. Consistent with these arguments, Hébert and Schreger (2017) find that legal rulings that increase the probability of default caused a depreciation of the exchange rate and a decline in the value of Argentine equity, disproportionally hurting foreign-owned firms, exporters, and large firms. 17.6.4 Domestic Financial and Political Costs To the extent that governments default on debt held by domestic residents (who are generally voters), it is not surprising that a sovereign default may have political costs. Broner et al. (2010) present a theory of sovereign debt in which default is deterred by the combination of domestic political default costs (assumed to prevent governments from opportunistically default on their own citizens) and the presence of well-functioning secondary markets (which make it impossible to selectively default on foreigners). Reinhart and Rogoff (2011) document that a significant share of sovereign debt is issued under domestic jurisdiction and held mainly by residents (see also IMF, 2021; Erce et al., 2022). Gennaioli et al. (2018) document the significant exposure of banks to their own sovereigns. When banks hold sovereign bonds, a sovereign default may hurt their balance sheets, causing a decrease in lending, a banking crisis, and a decline in economic activity (Gennaioli et al., 2014). In addition, banks could be hurt indirectly, including through the signaling channels discussed previously. Bocola (2016) finds that sovereign risk is recessionary because it tightens the funding constraints of banks. Asonuma et al. (2021) show that reduction in bank credit to the private sector is an important channel through which debt restructurings hurt the economy. Borensztein and Panizza (2009) and Malone (2011) find that sovereign defaults are associated with an increased probability of job loss by political leaders. These political costs have been associated with inefficient delays of sovereign debt restructurings (Levy Yeyati & Panizza, 2011). 17.6.5 Quantifying the Output Cost of Sovereign Defaults Most default costs described above—rises in borrowing costs, financial embargos, trade reductions, reputational spillovers, losses imposed on domestic financial intermediaries— would be expected to have an impact on output. Measuring these output costs is intrinsically difficult, however. While it is easy to find a negative correlation between default and growth, it is difficult to determine whether this negative correlation is driven by the default or by other factors that explain both the default and low growth. Furthermore, while the default could trigger low growth, low growth (and the expectation of low growth in the future) may trigger the default. Consistent with this “reverse causality,” Levy Yeyati and Panizza (2011) find that output contractions typically precede defaults. Despite these difficulties, a few recent papers have tried to identify the causal incremental impact of default on output. Kuvshinov and Zimmermann (2019) use an inverse propensity score weighting (IPSW) approach (Jordá & Taylor, 2016) which first estimates a government’s propensity to default using a logit model and then uses the predicted default probability to give more surprising (exogenous) default events a greater weight in the estimation (this approach works if

388  Research handbook of financial markets

the decision to default is only influenced by the variables used to estimate the propensity score). Their main finding is that defaults have a large but temporary output cost, which peaks at an average of almost four percent of GDP after five years, followed by a recovery in the long run. Marchesi and Masi (2021) use a synthetic controls approach in which a group of fifteen countries that defaulted on official debts and seven countries that defaulted on private debts are compared with a control group constructed to mimic the pre-default behavior of the defaulting countries. Unlike Kuvshinov and Zimmermann, they find a permanent effect of default on private creditors on debtor output: a default depresses growth in the debtor country by about 1 to 1.8 percentage points per year in which the country is in default, but not in the long run. In contrast a default on the official sector has no impact on growth during the default years and appears to raise growth in the long run. This literature also points to some factors that modify the costs of default. Defaults appear to have higher costs in countries with large banking sectors (Asonuma et al., 2021) and if they are followed by banking crises (Kuvshinov & Zimmermann, 2019). Debt restructurings that preempt (and hence avoid) default appear to have much lower output costs than post-default restructurings (Asonuma et  al., 2016; 2021). “Hard” defaults (defined either using an index that captures the “coerciveness” of the government’s negotiating tactics or based on the losses inflicted on investors) lead to larger output losses than “soft” defaults (Trebesch & Zabel, 2017).

17.7 WHEN DO GOVERNMENTS DEFAULT?9 Closely related to the costs of default are the circumstances in which governments default— namely, settings in which the costs of defaulting are smaller than the costs of servicing the debt. This could be due to the state of the economy, the costs of rolling over debt coming due (or the inability to do so), or political factors. 17.7.1 Affordability of Debt Payments (“Ability to Pay”) When government resources are low relative to scheduled debt service, paying debt obligations may require large adjustments to expenditures or revenues. Such adjustments can be economically or politically costly. Circumstances that may depress the affordability of debt payments include: ●

Economic downturns. Governments tend to default in periods of low growth (Manasse & Roubini, 2009), when fiscal revenues are typically lower, and expenditures are sometimes higher. Tomz and Wright (2007) report that 62 percent of default episodes occurred in years when the output level in the defaulting country was below its trend. Consistent with this, sovereign credit ratings respond to macroeconomic factors, such as the GDP growth rate and per capita income (Cantor & Packer, 1996). The countercyclicality of the interest rate paid by governments in developing countries (Section 17.8) suggests that markets expect more defaults when economic conditions are worse. Wars or civil conflicts can trigger a collapse of economic activity and thus increase default risk (Rivoli & Brewer, 1997).

9 This section draws on Hatchondo et al. (2007).

Sovereign debt  ●

●

●

389

Terms of trade shocks. Many emerging economies strongly rely on commodity exports as a source of tax revenue and foreign exchange (Mendoza, 1995). Many studies show that terms of trade fluctuations are a significant predictor of sovereign default and interest rate spreads in emerging economies.10 For example, in Ecuador, falling commodity prices led to a sovereign default in 1999.11 The sharp declines in oil prices during the second half of the 1990s contributed to the worsening of the macroeconomic and fiscal situation that led to the Russian default of 1998 (Sturzenegger & Zettelmeyer, 2007). A devaluation of the local currency. The level of public debt obligations is an important determinant of affordability. Moreno Badia et  al. (2020) find that the levels of public debt and public debt service are important predictors of fiscal crises. When sovereign debt is denominated in foreign currency (Section 17.10) and revenues rely mainly on nontradable goods production and taxation, a depreciation of the local currency damages the government’s ability to afford debt payments. Currency mismatches in the banking, corporate, and household sectors can magnify the effects of the depreciation, by leading to bankruptcies, a drop in investment, and a fall in government revenue. Manasse and Roubini (2009) find that exchange rate overvaluation and exchange rate volatility predict sovereign debt crisis. Ghulam and Derber (2018), Moreno Badia et al. (2020), and Moreno Badia et al. (2021), find a significant role of exchange rate fluctuations as a predictor of crises. Moreno Badia et al. (2020) document that fiscal crises often overlap with currency crises. Contingent liabilities. The materialization of contingent liabilities, including from the banking sector, may damage the government capacity to pay and thus increase sovereign risk (Moreno Badia et al., 2021). Balteanu and Erce (2018) and Ghulam and Derber (2018) discuss how bank and sovereign distress feed into each other.

17.7.2 High Borrowing Costs (“Ability to Pay”) A high cost of borrowing could trigger a default if it leads debt to grow faster than revenues and the economy, and thus causes a loss of access to the external debt market (at which point the cost of any action to mobilize the resources to repay may be prohibitive, unless the country can access official financing from crisis lenders such as the IMF). Apart from the adverse events discussed in the previous subsection, higher borrowing costs could have two triggers. First, a sharp rise in advanced country interest rates. For example, sharply higher interest rates in the United States in the early 1980s (a result of Federal Reserve Board chairman Paul Volcker’s efforts to bring down inflation in the US) were one of the main triggers of the developing country debt crisis of the 1980s (Cline, 1995). Many studies have documented that the borrowing cost of developing countries are influenced by US interest rates (Lambertini, 2001; Arora & Cerisola, 2001; Uribe & Yue, 2006; Ghulam & Derber, 2018).

10 See, for instance, Catao and Sutton (2002), Catao and Kapur (2004), Min (1998), Caballero (2003), Caballero and Panageas (2003), Calvo, Izquierdo and Mejia (2004), Cuadra and Sapriza (2006), Sturzenegger and Zettelmeyer (2007), and Hilscher and Nosbusch (2010). 11 Oil and bananas together accounted for 59 percent of Ecuadorian exports in 2001. Ecuador was the first country to default on Brady bonds (Brady bonds arose from an effort in the late 1980s to reduce the debt held by less-developed countries that were frequently defaulting on loans).

390  Research handbook of financial markets

Second, a collapse in confidence or increase in risk aversion, also referred to as a “debt run,” or a “sudden stop” (Calvo, 1998). Unlike a rise in borrowing costs triggered by foreign interest rates, this type of crisis can be self-fulfilling, as the expectation of default triggers a jump in borrowing costs which raises the cost of repaying and induces default, validating the original expectation (Sachs, 1984; Calvo, 1988; Cole & Kehoe, 1996, 2000; Lorenzoni & Werning, 2019). Sudden stops of this kind contributed to the Mexican (1995) debt crisis, the Asian (1997) crisis, the international propagation of the global financial crisis after the collapse of Lehman brothers in September of 2008, the sharp but thankfully brief tightening of emerging market borrowing conditions in March 2020, and many other international financial crisis episodes. Longstaff et al. (2011) find that global factors account for 64 percent of the variation in sovereign spreads and that on average, the risk premium represents about a third of sovereign spreads. 17.7.3 Political Factors (“Willingness to Pay”) Manasse and Roubini (2009) and Ghulam and Derber (2018) find that political factors influence sovereign default risk. This occurs because government turnover may trigger significant changes in the sovereign’s willingness to pay (Van Rijckeghem & Weder, 2004; Hatchondo et al., 2010). In a survey of default episodes, Sturzenegger and Zettelmeyer (2007) conclude that “a solvency crisis could be triggered by a shift in the parameters that govern the country’s willingness to make sacrifices in order to repay, due to changes in the domestic political economy (a revolution, a coup, an election etc.).” Goretti (2005) describes how concerns about the elected candidate explain the increase of the sovereign spread in Brazil around the 2002 presidential election. Alfaro and Kanczuk (2005), Cole et  al. (1995), and Hatchondo et al. (2009) present models in which both default and difficulties in market access after a default may be triggered by political turnover.

17.8 THE COSTS OF SOVEREIGN DEFAULT RISK Starting with Neumeyer and Perri (2005), several studies have documented that business cycles in small emerging economies differ from those in small advanced economies, and suggested that these differences relate to the presence of default risk in the former but not the latter.12 Default risk influences interest rates which in turn influence economic activity. Figure 17.2 and Table 17.A1 present the empirical regularities documented by Neumeyer and Perri (2005), updated with available 1983–2018 data. Compared with developed economies, emerging economies feature: ● ●

●

●

Higher volatility of output, real interest rates, and net exports; Higher volatility of consumption relative to income (in emerging economies, consumption is typically more volatile than income, while the opposite is true in advanced economies); Countercyclical real interest rates, with higher interest in recessions (in contrast, interest rates in advanced economies are procyclical); More countercyclical net exports.

12 See Neumeyer and Perri (2005), Uribe and Yue (2006) Aguiar and Gopinath (2007) and GarciaCicco et al. (2010), among others.

Sovereign debt 

391

Emerging Economies

Correlation between GDP(t) and R(t+j)

0.50

0.25 Country Argentina Brazil Korea Mexico Philippines Average

0.00

–0.25

–0.50

–4

–2

0

2

4

j (Quarters) Developed Economies

Correlation between GDP(t) and R(t+j)

0.75

0.50 Country Australia Canada Netherlands New Zealand Sweden Average

0.25

0.00

–0.25

–4

–2

0

2

4

j (Quarters) Source:   Authors’ calculations based on official data.

Figure 17.2  Cross-correlations between GDP and real interest rates at different lags Additional distinctive features of emerging economies include procyclical government expenditure (while government expenditure is acyclical or slightly countercyclical in advanced countries) and a countercyclical inflation tax (while that latter is procyclical in advanced economies).13 13 See Gavin and Perotti (1997), Talvi and Vegh (2005), Kaminsky et al. (2004), and Alesina et al. (2008).

392  Research handbook of financial markets

Macroeconomic models incorporating sovereign risk can account for the business cycle regularities of emerging economies.14 High interest rates paid by developing countries reflect a compensation for the probability of a sovereign default, while the countercyclicality of spreads is consistent with the fact that sovereigns are more likely to default when economic conditions are bad (see Section 17.7). Hence, borrowing conditions are more expensive in bad times, and thus borrowing levels may be lower. This leads to countercyclical net exports (lower imports in bad times) and higher volatility of consumption relative to income compared to advanced economies. Furthermore, if borrowing is more expensive in bad times, it may be optimal to tax more and decrease government expenditures in such times, which would help to explain the procyclicality of public expenditures and the countercyclicality of tax rates in emerging countries (Cuadra, Sanchez & Sapriza, 2010). However, emerging markets needs not be stuck with these adverse business cycle properties forever. Frankel et al. (2013) show that many emerging economies have “graduated” from fiscal procyclicality, shifting from procyclical to countercyclical fiscal policy. They also suggest that improving institutional quality is the key to the ability to conduct a more countercyclical fiscal policy. Similarly, Vegh and Vuletin (2014) show that many Latin American countries have been able to switch from procyclical to countercyclical fiscal policy responses to crises, and that this graduation proved effective in reducing the severity of the crises. Amador and Phelan (2021) investigate the possibility of “graduation” in the context of a model of sovereign default and reputation. A recent branch of the literature inspired by the Eurozone crisis focuses on direct costs that sovereign risk creates for the economy at large. A literature pioneered by Farhi and Tirole (2018) studies the nexus between a sovereign and the banking system. Because of feedback created by a combination of exposure of domestic banks to sovereign bonds and the possibility and anticipation of bailouts, so-called “doom loops” can arise.15 In these events, negative shocks for one of the actors can spell trouble for the other, amplifying the recession caused by the original shock. At the same time, also inspired by the Eurozone crisis, Bocola (2016) provides a model of the correlation between the interest rates paid by sovereigns and those available for investment by the private sector of the same countries, which Neumeyer and Perri (2005) assume. Because of the pass-through of sovereign risk, investment becomes more costly when sovereign default seems likely, which hurts growth prospects for the economy (see also Arellano, Bai, & Bocola, 2017 and Arellano, Bai, & Mihalache, 2018). Balke (2017) pursues a related effect in job postings instead of real investment. Bianchi, Ottonello, and Presno (2021) studies feedback between sovereign risk and domestic fundamentals with nominal rigidities, and Roldán (2020) investigates costs of sovereign risk stemming from an aggregate-demand doom loop.

14 See Aguiar and Gopinath (2006), Arellano (2008), Cuadra and Sapriza (2008), Hatchondo and Martinez (2009), Hatchondo, Martinez and Sapriza (2010), Lizarazo (2010) and Yue (2010) among others. 15 For instance, Burnside et al. (2001) argue that the Asian crisis in 1997 was originated due to the expectation that future deficits associated with implicit bailout guarantees to the banking system would be financed through an inflation tax on outstanding nominal debt.

Sovereign debt 

393

17.9 WHY DO GOVERNMENTS CHOOSE HIGH AND VOLATILE SOVEREIGN RISK? The quantitative-theoretic literature on sovereign default has shown that a rational and benevolent government only chooses to live with high and volatile sovereign risk because of its inability to commit to future repayment and borrowing policies.16 Furthermore, even without commitment to future repayment policies, if governments could commit to future borrowing policies, this would eliminate the majority of sovereign risk (Hatchondo & Martinez, 2009; Chatterjee & Eyigungor, 2012; Hatchondo et al., 2021a).17 Governments’ inabilities to commit to future borrowing policies generates the so-called debt dilution problem. Debt dilution refers to the reduction in the value of existing debt triggered by the issuance of new debt. Issuing new debt reduces the value of existing debt because it increases the probability of default. Three factors generate the sovereign debt dilution problem: (i) governments issue long-term debt, (ii) the current government cannot control debt issuances by future governments, and (iii) bonds are priced by rational investors. Rational investors anticipate that additional borrowing by future governments will increase the risk of default on long-term bonds issued by the current government and, thus, offer a lower price for these bonds. The current government could benefit from constraining future borrowing because this could increase the price of the bonds it issues. However, governments are typically unable to constrain borrowing by future governments. This literature also assumes that the government can only issue noncontingent bonds (i.e., financial markets are incomplete). In the standard model, debt and income are the only determinants of default. With complete markets, the government could make a different payment promise for each level of next-period income, eliminating uncertainty about repayment. Then, lenders would never pay for a payment promise on which they know the government would default and a bond making such a promise is not traded. In contrast, with non-contingent bonds, when the government borrows it promises the same payment regardless of the level of income. In this case, the government tends to default if the income realization is sufficiently low.

17.10 WHAT CAN BE DONE TO MITIGATE SOVEREIGN RISK? 17.10.1 Fiscal Frameworks In order to improve commitment to future government borrowing, an increasing number of countries are adopting fiscal rules—restrictions imposed upon future governments’ ability to conduct fiscal policy—that can help anchor expectations about future policy, reducing excessive future deficits, and hence relaxing current borrowing constraints. Hatchondo et  al. (2022a; 2022b) use a model of sovereign default with long-term debt to show how substantial gains could be achieved by introducing simple fiscal rules that 16 Following Aguiar and Gopinath (2006) and Arellano (2008), this literature has extended the Eaton and Gersovitz (1981) framework for quantitative studies of fiscal policy and sovereign debt crises. 17 Hatchondo el al. (2022a) argue that it would be easier to commit to future borrowing policies than to commit to future repayment policies.

394  Research handbook of financial markets

implement fiscal anchors.18 While debt levels play a predominant role in discussions of fiscal rules, Hatchondo et al. (2022a; 2022b) show that the sovereign spread is better suited to be the fiscal anchor. They find that a common spread limit generates significant welfare gains for sets of economies with different levels of debt intolerance (i.e., with different mappings from sovereign debt levels to sovereign spreads). In contrast, a common debt limit may fail to generate welfare gains for many economies in the set and may even generate welfare losses for some of these economies. Since the sovereign spread incorporates information about the degree of debt intolerance in each economy, the common spread limit forces economies with more debt intolerance to borrow less while allowing economies with less debt intolerance to borrow more. The performance of a common rule limit for sets of economies with different levels of debt intolerance is important for two reasons. First, fiscal rules often impose common limits on several economies. In 2014, 48 of the 85 countries with fiscal rules had supranational rules. Second, robust policy recommendations should acknowledge uncertainty about a single economy’s characteristics. The exercises presented by Hatchondo et al. (2022a; 2022b) can be interpreted as searching for policy recommendations that are robust to this uncertainty. Blanchard et al. (2021) point out that debt-limit fiscal rules are bound to lead to mistakes due to this “Knightian uncertainty,” and instead propose replacing numerical debt targets with fiscal standards (i.e., qualitative prescriptions that leave room for judgment together with a process to decide whether the standards are met). Of course, the possibility of constraining future governments’ borrowing with a fiscal rule depends on those governments’ ability to commit to respecting the fiscal rule in the future. Countries have strengthened compliance with their fiscal rules by introducing independent fiscal councils that provide public assessments of fiscal plans and performance, and evaluation or provision of macroeconomic and budgetary forecasts. In addition, an increasing number of countries has implemented fiscal responsibility laws that set out procedural and transparency requirements. Fiscal rules are also being complemented with automatic sanctioning and enforcement procedures. Schaechter et al. (2012), Debrun and Kinda (2014), and Eyraud et al. (2018) discuss countries’ experiences with fiscal rules, transparency laws, and fiscal councils. Hatchondo et al. (2022a) argue that committing to follow good fiscal rules may not be too costly. Chatterjee and Eyigungor (2015) and Hatchondo et al. (2016) propose modifying sovereign debt contracts to constrain future borrowing. 17.10.2 State Contingent Debt As discussed in Section 17.7, one of the inefficiencies in sovereign debt models stems from the assumption of incomplete markets, which reflects the lack of state-contingent debt instruments in practice. The recent European sovereign debt crises and the increase in public debt levels after the COVID-19 shock have brought proposals for state-contingent debt instruments to the forefront of policy debates as a strategy to avoid costly defaults (United Nations, 2006; Blanchard et al., 2016; IMF, 2017; IMF, 2020).

18 Hatchondo et  al. (2021) compute the borrowing path that a government with a commitment to future borrowing would choose and find that a simple debt brake rule can achieve 60 percent of the welfare gains obtained with the optimal borrowing path.

Sovereign debt 

395

Several studies highlight the possible benefits from tying sovereign debt obligations to domestic GDP (Shiller, 1993; Borensztein & Mauro, 2004; Hatchondo & Martinez, 2012). One benefit from GDP indexation is that issuing debt that promises lower payments when GDP takes low values may facilitate the financing of automatic stabilizers (such as an increase in unemployment benefits during economic downturns) and countercyclical fiscal policy. Another benefit is that GDP indexation could diminish the likelihood of fiscal crises for governments that face a countercyclical borrowing cost (in part because of countercyclical default risk). Kamstra and Shiller (2010) argue that GDP indexation would help investors who want exposure to income growth (for instance, to protect relative standards of living in retirement) and protection against inflation. Bolton and Jeanne (2009) argue that it is somewhat of a puzzle that the overwhelming majority of sovereign debts are not GDP indexed. Despite these well-understood advantages, the use of state-contingent debt instruments is scarce in practice and countries have not been able to issue such financial instruments at a reasonable premium—as in the recent cases of Argentina (2005), Greece (2012), and Ukraine (2015). Surprisingly, while some practical implementation challenges have been discussed among policymakers, there is little theoretical analysis investigating them and the lack of indexation in sovereign debt markets remains a puzzle. IMF (2017) and Benford et al. (2018) point to myopia on the part of issuers, who might be out of office before the gains fully materialize. Krugman (1988) argues that GDP-indexed bonds could create moral hazard problems by disincentivizing the government to conduct growth-friendly policies or leading governments to misreport GDP statistics. However, these arguments do not seem to be empirically relevant.19 Others argue that markets for these instruments tend to be shallow and, thus, these bonds would carry a large liquidity premium. Moretti (2020) investigates this liquidity channel and finds that state-contingent debt is still welfare-improving. Overall, there are no compelling arguments in the literature to outweigh the aforementioned merits of indexation and justify their little use in practice. One exception is the work by Roch and Roldan (2023) who analyze how concerns for model misspecification (à la Hansen & Sargent, 2001) on the part of international lenders affect the desirability of issuing state-contingent debt instruments in a standard sovereign default model. They show that for the commonly used threshold state-contingent bond structure (e.g., in the GDP-linked bond issued by Argentina in 2005), the model with robustness generates ambiguity premia in bond spreads that can explain most of what the literature has labeled as novelty premium. While the government would be better off with this bond when facing rational-expectations lenders, this additional source of premia leads to welfare losses when facing robust lenders. Furthermore, the optimal bond design crucially depends on the degree of the lenders’ preference for robustness. At the calibrated level of robustness, the optimal state-contingent bond provides insurance to the home country but also avoids jumps in the repayments offered to lenders as a function of domestic income. In contrast to the commonly used threshold bond, the optimal design generates substantial welfare gains, although these gains are decreasing in the level of robustness.

19 These arguments should also apply to inflation-linked bonds, but many countries have issued these type of securities.

396  Research handbook of financial markets

17.10.3 Monetary Policy Credibility Eichengreen and Hausmann (1999) were the first to highlight that emerging economies’ sovereign debt was mainly denominated in foreign currency during the 1990s and introduced the term “original sin” to describe the inability of these countries to borrow abroad in local currency. Since then, there has been extensive literature examining both the consequences and causes of original sin. In particular, Hausmann (2003) shows that original sin lowers the creditworthiness of a country because it makes the real exchange rate (which is volatile and tends to depreciate in bad times) a relevant price in determining the capacity to pay. Thus, countries with original sin are charged an additional risk premium when they borrow. A strand of the literature has related the original sin to the lack of a credible and independent monetary policy. Jeanne (2003) develops a simple conceptual model in which borrowers are induced to borrow in foreign currency because unpredictable monetary policy makes them uncertain about the future real value of their local currency liabilities. Rajan and Tokatlidis (2005) also argue that liability dollarization is a response to a lack of monetary policy credibility. More recently, Du et al. (2020) consider a two-period framework with risk-averse lenders and varying degrees of inflation commitment. As governments with lower commitment resort more often to inflation in the second period, debt denominated in local currency is riskier and lenders demand higher compensation for holding it. These risk premia make local-currency debt less desirable and induce the government to borrow in foreign currency. Ottonello and Perez (2019) and Engel and Park (2022) build quantitative models of sovereign default and currency composition to rationalize debt issuing in foreign currency as a commitment to not partially default the debt with future inflation when monetary policy is discretionary. Aguiar et al. (2013) show that while low commitment to inflation renders an economy with domestic currency bonds more vulnerable to a rollover crisis, extreme commitment is not desirable either as it eliminates the possibility of inflating during a crisis. They argue that there is a range of moderate inflation credibility that makes domestic currency bonds strictly preferable for intermediate levels of debt, where the reduction in rollover crisis vulnerability is at work. In the context of a monetary union, Bianchi and Mondragon (2022) use a quantitative model of sovereign default with nominal rigidities and self-fulfilling crises to show that the lack of monetary policy independence increases sovereign default risk by making countries more vulnerable to rollover crises. Without monetary policy autonomy, a self-fulfilling crisis can generate a recession in the presence of nominal rigidities under a fixed exchange rate regime but not with a flexible exchange rate. A final channel through which central bank credibility can make debt safer does not operate through default risk but rather through the cyclical insurance properties of sovereign debt. By lowering interest rates in recessions, the central bank generates a capital gain for domestic holders of sovereign debt at the time when they need it most. This turns government debt into a “negative beta” asset that helps investors hedge (Brunnermeier et al., 2021, Cochrane, 2021). Only central banks that do not have to worry that expansionary monetary policy may lead to a de-anchoring of inflation expectations can do this. Overall, this line of research suggests that a credible and independent monetary policy can function like a fiscal asset (Willems & Zettelmeyer, 2022). It is an essential condition for the development of local-currency debt markets which would reduce the premium associated with the original sin and, at the same time, reduce the premium related to rollover crises. It

Sovereign debt 

397

contributes to market liquidity and boosts the insurance value of holding sovereign debt. Seen through this lens, giving central banks more independence and adopting inflation-targeting frameworks are essential steps in de-risking debt and establishing it as a safe asset.

17.11 CONCLUSION In conclusion, we summarize what we view as the four main takeaways of this chapter. First, sovereign debt is a unique asset class because of the unique powers of the sovereign. From the perspective of a private creditor, however, this can cut both ways. In countries in which political accountability of governments and constitutionally anchored property rights make defaults on domestic creditors extremely unlikely, and from which foreign investors can freely enter and exit through liquid secondary bond markets, the main power that matters is the power to tax in the future. This acts like a form of collateral, making sovereign debt safe. In contrast, in countries with lower political accountability and/or greater ability to discriminate between foreign and domestic creditors, the main trait that sets sovereign debt aside is the limited ability of foreign investors to enforce their claims against the sovereign in the event of default. This makes sovereign debt risky, particularly when economic institutions are weak. Advanced countries are predominantly in the first group, while emerging markets and developing economies predominantly in the second. Second, countries issuing risky sovereign debt can do so because of the associated costs of default. These include temporary exclusion from external capital markets, higher future borrowing costs, negative reputational effects that can hurt the domestic economy, and the financial and political costs that arise from defaulting on one’s own citizens. Defaults have two main causes: economic or political shocks that raise the costs of fiscal and/or external adjustment that would be needed to generate the resources to repay above the costs of default, and sharp rises in borrowing costs. The latter could be triggered by an increase in foreign (world) borrowing rates, but also by a shift in expectations, leading to a sudden change in market appetite for risky debt, a rollover crisis, and a depreciation of the currency. As a result, debt crises can be self-fulfilling. Third, countries issuing risky sovereign debt suffer from a severe disadvantage relative to countries that issue safe debt. Their real interest rates, output, consumption, and external balances are more volatile. While in advanced economies, consumption is smoother than output, the opposite is true in economies with risky debt. In such economies, government fiscal policies tend to be procyclical, exacerbating recessions and booms. The reason is that recessions increase sovereign risk and borrowing costs, forcing governments to borrow less, making recessions worse. This can further increase sovereign risk. Finally, risky-debt economies can morph into safe-debt economies and vice versa. In the last two decades, many emerging market countries have “graduated” from procyclical to countercyclical fiscal policies; leading to less severe crises and reduced sovereign risk. Conversely, during the euro area crisis, the debt of several European countries that was previously thought of as safe became risky, and one country (Greece) had to restructure its debts. The causes for graduation or relegation mostly relate to slow-moving institutional change (and with regard to relegation, to large shocks that overwhelm the capacity of institutions to deal with them). However, they can also be influenced by policies that are to some extent under the control of governments. These include commitment devices against overborrowing

398  Research handbook of financial markets

and debt dilution, such as fiscal rules; adoption of state-contingent debt instruments that reduce the set of economic states that could lead to default, and credible and independent monetary policy. The latter enables the government to issue bonds in local currency, makes bond markets more liquid, and enables central banks to intervene in a panic, reducing the risk of rollover crises.

REFERENCES Aguiar, M., Amador, M., Farhi, E., & Gopinath, G. (2013). Crisis and commitment: Inflation credibility and the vulnerability to sovereign debt crises. NBER Working Papers 19516, National Bureau of Economic Research, Inc. Aguiar, M., & Gopinath, G. (2006). Defaultable debt, interest rates and the current account. Journal of International Economics, 69(1), 64–83. https://doi​.org​/10​.1016​/j​.jinteco​.2005​.05​.005. Alesina, A. F., & Tabellini, G. (2005). Why is fiscal policy often procyclical? SSRN Electronic Journal. https://doi​.org​/10​.2139​/ssrn​.780225. Alesina, A., Campante, F. R., & Tabellini, G. (2008). Why is fiscal policy often procyclical? Journal of the European Economic Association, 6(5), 1006–1036. https://doi​.org​/10​.1162​/JEEA​.2008​.6​.5​.1006. Alfaro, L., & Kanczuk, F. (2005). Sovereign debt as a contingent claim: A quantitative approach. Journal of International Economics, 65(2), 297–314. https://doi​.org​/10​.1016​/j​.jinteco​.2004​.02​.006. Amador, M., & Phelan, C. (2021). Reputation and sovereign default. Econometrica, 89(4), 1979–2010. https://doi​.org​/10​.3982​/ ECTA16685. Andreasen, E., Sandleris, G., & Van der Ghote, A. (2019). The political economy of sovereign defaults. Journal of Monetary Economics, 104, 23–36. https://doi​.org​/10​.1016​/j​.jmoneco​.2018​.09​.003. Arellano, C. (2008). Default risk and income fluctuations in emerging economies. American Economic Review, 98(3), 690–712. https://doi​.org​/10​.1257​/aer​.98​.3​.690. Arellano, C., Bai, Y., Arslanalp, S., & Tsuda, T. (2014). Tracking global demand for emerging market sovereign debt. IMF Working Papers 2014/039, International Monetary Fund. Arellano, C., Bai, Y., & Mihalache, G. (2017). Default risk, sectoral reallocation, and persistent recessions (No. w23835; p. w23835). National Bureau of Economic Research. https://doi​.org​/10​.3386​ /w23835. Arora, V., & Cerisola, M. (2001). How does U.S. Monetary policy influence sovereign spreads in emerging markets? IMF Staff Papers, 48(3), 474–498. Arslanalp, S., Bergthaler, W., Stokoe, P., & Tieman, A. (2019). Concepts, definitions and composition. In S. Abbas, A. Pienkowski & K. Rogoff (Eds.), Sovereign debt: A guide for economists and practitioners. Oxford: Oxford University Press. Arellano, C., Bai, Y., & Bocola, L. (2017). Sovereign default risk and firm heterogeneity (No. w23314; p. w23314). National Bureau of Economic Research. https://doi​.org​/10​.3386​/w23314. Asonuma, T., Niepelt, D., & Rancière, R. (2017). Sovereign bond prices, haircuts and maturity (No. w23864; p. w23864). National Bureau of Economic Research. https://doi​.org​/10​.3386​/w23864. Asonuma, T., & Trebesch, C. (2016). Sovereign debt restructurings: Preemptive or post-default: Sovereign debt restructurings. Journal of the European Economic Association, 14(1), 175–214. https://doi​.org​/10​.1111​/jeea​.12156. Balkan, E. M. (1992). Political instability, country risk and probability of default. Applied Economics, 24(9), 999–1008. https://doi​.org​/10​.1080​/00036849200000077. Balke, N. (2017). The employment cost of sovereign default. Available at: https://www.neelebalke.com/ Balke-employment-cost-of-default.pdf Barro, R. J. (1979). On the determination of the public debt. Journal of Political Economy, 87(5), 940– 971. https://doi​.org​/10​.1086​/260807. Benford, J., Ostry, J. D., Shiller, R., Haldane, A., & Obstfeld, M. (2018). Sovereign GDP-linked bonds: Rationale and design, CEPR Press, London, UK. Berentsen, A., & Waller, C. (2018). Liquidity premiums on government debt and the fiscal theory of the price level. Journal of Economic Dynamics and Control, 89, 173–182. https://doi​.org​/10​.1016​/j​.jedc​ .2018​.01​.038.

Sovereign debt 

399

Bianchi, J., & Mondragon, J. (2022). Monetary independence and rollover crises. Quarterly Journal of Economics, 137(1), 435–491. Bianchi, J., Ottonello, P., & Presno, I. (2019). Fiscal stimulus under sovereign risk (No. w26307; p. w26307). National Bureau of Economic Research. https://doi​.org​/10​.3386​/w26307. Blanchard, O. (2016). The Phillips curve: Back to the ’60s? American Economic Review, 106(5), 31–34. https://doi​.org​/10​.1257​/aer​.p20161003. Blanchard, O. J., Zettelmeyer, J., & Leandro, A. (2021). Redesigning EU fiscal rules: From rules to standards. PIIE Working Paper. https://doi​.org​/10​.2139​/ssrn​.3810052. Bocola, L. (2016). The pass-through of sovereign risk. Journal of Political Economy, 124(4), 879–926. https://doi​.org​/10​.1086​/686734. Bohn, H. (1995). The sustainability of budget deficits in a stochastic economy. Journal of Money, Credit, and Banking, 27(1), 257. https://doi​.org​/10​.2307​/2077862. Borensztein, E., Mauro, P., Ottaviani, M., & Claessens, S. (2004). The case for gdp-indexed bonds. Economic Policy, 19(38), 167–216. Borensztein, E., & Panizza, U. (2009). The costs of sovereign default. IMF Staff Papers, 56(4), 683–741. Broner, F., Martin, A., & Ventura, J. (2010). Sovereign risk and secondary markets. American Economic Review, 100(4), 1523–1555. Brunnermeier, M. K., & Krishnamurthy, A. (2020). Corporate debt overhang and credit policy. Brookings Papers on Economic Activity, 2020(2), 447–502. https://doi​.org​/10​.1353​/eca​.2020​.0014. Brunnermeier, M. K., Merkel, S., & Sannikov, Y. (2021). Debt as safe asset: Mining the bubble [Unpublished manuscript]. Princeton: Princeton University. Burnside, C., Eichenbaum, M., & Rebelo, S. (2001). Prospective deficits and the Asian currency crisis. Journal of Political Economy, 109(6). https://doi​.org​/10​.1086​/323271. Calvo, G. A. (1988). Servicing the public debt: The role of expectations. American Economic Review, 78(4), 647–661. Calvo, G. A. (1998). Capital flows and capital market crises: The simple economics of sudden stops. Journal of Applied Economics, 1(1), 35–54. https://doi​.org​/10​.1080​/15140326​.1998​.12040516. Cantor, R., & Packer, F. (1996). Determinants and impact of sovereign credit ratings. Journal of Fixed Income, 6(3), 76–91. https://doi​.org​/10​.3905​/jfi​.1996​.408185. Chatterjee, S., & Eyigungor, B. (2012). Maturity, indebtedness, and default risk. American Economic Review, 102(6), 2674–2699. https://doi​.org​/10​.1257​/aer​.102​.6​.2674. Chatterjee, S., & Eyigungor, B. (2015). A seniority arrangement for sovereign debt. American Economic Review, 105(12), 3740–3765. https://doi​.org​/10​.1257​/aer​.20130932. Chaumont, G. (2020). Sovereign debt, default risk, and the liquidity of government bonds (SSRN Scholarly Paper ID 3714870). Social Science Research Network. Retrieved from https://papers​.ssrn​ .com​/abstract​=3714870. Citron, J.-T., & Nickelsburg, G. (1987). Country risk and political instability. Journal of Development Economics, 25(2), 385–392. https://doi​.org​/10​.1016​/0304​-3878(87)90092-7. Cochrane John, H. (2021). r < g. Unpublished manuscript, Hoover Institution, Stanford, CA. Cole, H. L., Dow, J., & English, W. B. (1995). Default, settlement, and signalling: Lending resumption in a reputational model of sovereign debt. International Economic Review, 36(2), 365. https://doi​.org​ /10​.2307​/2527201. Cole, H. L., & Kehoe, P. J. (1998). Models of sovereign debt: Partial versus general reputations. International Economic Review, 39(1), 55. https://doi​.org​/10​.2307​/2527230. Cole, H. L., & Kehoe, T. J. (1996). A self-fulfilling model of Mexico’s 1994–1995 debt crisis. Journal of International Economics, 41(3–4), 309–330. Cole, H. L., & Kehoe, T. J. (1998). Models of sovereign debt: Partial versus general reputations. International Economic Review, 39(1), 55–70. Cole, H. L., & Kehoe, T. J. (2000). Self-fulfilling debt crises. Review of Economic Studies, 67(1), 91–116. Cruces, J.J. and Christoph Trebesch, T. (2013): Sovereign Defaults: The Price of Haircuts, American Economic Journal: Macroeconomics, 5, 85–117. Cuadra, G., Sanchez, J. M., & Sapriza, H. (2010). Fiscal policy and default risk in emerging markets. Review of Economic Dynamics, 13(2), 452–469. https://doi​.org​/10​.1016​/j​.red​.2009​.07​.002. Dixit, A., & Lambertini, L. (2001). Monetary–fiscal policy interactions and commitment versus discretion in a monetary union. European Economic Review, 45(4–6), 977–987. https://doi​.org​/10​ .1016​/S0014​-2921(01)00134-9.

400  Research handbook of financial markets

Debrun, X. and Kinda, T. (2014): Strengthening Post-Crisis Fiscal Credibility: Fiscal Councils on the Rise – A New Dataset, IMF Working Paper. Dovis, A., Golosov, M., & Shourideh, A. (2016). Political economy of sovereign debt: A theory of cycles of populism and austerity. Social Science Research Network. Retrieved from https://papers​.ssrn​.com​ /abstract​=2725723. Du, W., Pflueger, C. E., & Schreger, J. (2020). Sovereign debt portfolios, bond risks, and the credibility of monetary policy. Journal of Finance, 75(6), 3097–3138. Dvorkin, M., Sánchez, J. M., Sapriza, H., & Yurdagul, E. (2021). Sovereign debt restructurings. American Economic Journal: Macroeconomics, 13(2), 26–77. https://doi​.org​/10​.1257​/mac​.20190220. Eaton, J., & Gersovitz, M. (1981). Debt with potential repudiation: Theoretical and empirical analysis. Review of Economic Studies, 48(2), 289. https://doi​.org​/10​.2307​/2296886. EBRD. (2010). Developing local currency finance. Chapter 3, Transition Report, 2010 (pp. 46–65). London: European Bank for Reconstruction and Development. Eichengreen, B., & Hausmann, R. (1999). Exchange rates and financial fragility. Proceedings – Economic Policy Symposium – Jackson Hole, Federal Reserve Bank of Kansas City, 329–368. Engel, C., & Park, J.J. (2022). Debauchery and original sin: The currency composition of sovereign debt. Journal of the European Economic Association. Erce, A., Mallucci, E., & Picarelli, M. (2022). A journey in the history of sovereign defaults on domesticlaw public debt. Universidad Pública De Navarra Working Paper 2106. Eyraud, L., X. Debrun, A. Hodge, V. Lledó, & Pattillo, C. 2018. Second-Generation Fiscal Rules: Balancing Simplicity, Flexibility, and Enforceability. IMF Staff Discussion Note 18/04. Farhi, E., & Maggiori, M. (2018). A model of the international monetary system. The Quarterly Journal of Economics, 133(1), 295–355. https://doi​.org​/10​.1093​/qje​/qjx031. Farhi, E., & Tirole, J. (2018). Deadly embrace: Sovereign and financial balance sheets doom loops. The Review of Economic Studies, 85(3), 1781–1823. https://doi​.org​/10​.1093​/restud​/rdx059. Flandreau, M., Flores, J. H., Gaillard, N., & Nieto-Parra, S. (2010). The end of gatekeeping: Underwriters and the quality of sovereign bond markets, 1815–2007. NBER International Seminar on Macroeconomics, University of Chicago Press, 6(1), 53–92. Frankel, J. A., Vegh, C. A., & Vuletin, G. (2013). On graduation from fiscal procyclicality. Journal of Development Economics, 100(1), 32–47. https://doi​.org​/10​.1016​/j​.jdeveco​.2012​.07​.001. García-Cicco, J., Pancrazi, R., & Uribe, M. (2010). Real business cycles in emerging countries? American Economic Review, 100(5), 2510–2531. Gavin, M., & Perotti, R. (1997). Fiscal policy in Latin America. NBER Macroeconomics Annual, 12, 11–61. https://doi​.org​/10​.1086​/654320. Gelpern, A., Gulati, Mitu, & Zettelmeyer, J. (2019). If boilerplate could talk: The work of standard terms in sovereign bond contracts. Law and Social Inquiry, 44(3), 617–646. Gennaioli, N., Martin, A., & Rossi, S. (2014). Sovereign default, domestic banks, and financial institutions: Sovereign default, domestic banks, and financial institutions. The Journal of Finance, 69(2), 819–866. https://doi​.org​/10​.1111​/jofi​.12124. Gennaioli, N., Martin, A., & Rossi, S. (2018). Banks, government bonds, and default: What do the data say? Journal of Monetary Economics, 98, 98–113. https://doi​.org​/10​.1016​/j​.jmoneco​.2018​.04​.011. Ghulam, Y., & Derber, J. (2018). Determinants of sovereign defaults. The Quarterly Review of Economics and Finance, 69, 43–55. Goretti, M. (2004). The Brazilian currency turmoil of 2002: A nonlinear analysis. SSRN Electronic Journal. https://doi​.org​/10​.2139​/ssrn​.736383. Guembel, A., & Sussman, O. (2012). Sovereign debt without default penalties (SSRN Scholarly Paper ID 885434). Social Science Research Network. Retrieved from https://papers​ .ssrn​ .com​ /abstract​ =885434. Hall, G., Payne, J., Sargent, T. J., & Szőke, B. (2021). Costs of financing US federal debt: 1791–1933. Princeton University Working Paper. Hansen, L. P., & Sargent, T. J. (2001). Robust control and model uncertainty. American Economic Review, 91(2), 60–66. https://doi​.org​/10​.1257​/aer​.91​.2​.60. Hatchondo, J. C., & Martinez, L. (2009). Long-duration bonds and sovereign defaults (revised July 2009) (SSRN Scholarly Paper ID 2187910). Social Science Research Network. Retrieved from https://papers​.ssrn​.com ​/abstract​=2187910.

Sovereign debt 

401

Hatchondo, J. C., & Martinez, L. (2012). On the benefits of GDP-indexed government debt: Lessons from a model of sovereign defaults. Economic Quarterly Federal Reserve Bank of Richmond, 139–157. Hatchondo, J. C., Martinez, L., & Sapriza, H. (2007). The economics of sovereign defaults. Federal Reserve Bank of Richmond Economic Quarterly, 93(2), 163–187. Hatchondo, J. C., Martinez, L., & Sapriza, H. (2009). Heterogeneous borrowers in quantitative models of sovereign default. International Economic Review, 50(4), 1129–1151. https://doi​.org​/10​.1111​/j​.1468​ -2354​.2009​.00562​.x. Hatchondo, J. C., Martinez, L., & Sosa-Padilla, C. (2016). Debt dilution and sovereign default risk. Journal of Political Economy, 124(5), 1383–1422. https://doi​.org​/10​.1086​/688081. Hatchondo, J., Martinez, L., & Roch, F. (2021). Constrained efficient borrowing with sovereign default risk. Mimeo. Hatchondo, J., Martinez, L., & Roch, F. (2022a). Fiscal rules and the sovereign default premium. American Economic Journal: Macroeconomics, 14(4): 244–73. Hatchondo, J., Martinez, L., & Roch, F. (2022b). Numerical fiscal rules for economic unions: The role of sovereign spreads. Economics Letters, 210. https://doi.org/10.1016/j.econlet.2021.110168. Hausmann, R. (2003). Good credit ratios, bad credit ratings: The role of debt denomination. Rulesbased fiscal policy in emerging markets: Background, analysis and prospects, G. Kopits (ed.). London: Macmillan. Hébert, B., & Schreger, J. (2017). The costs of sovereign default: Evidence from Argentina. American Economic Review, 107(10), 3119–3145. https://doi​.org​/10​.1257​/aer​.20151667. Hilscher, J., & Nosbusch, Y. (2010). Determinants of sovereign risk: Macroeconomic fundamentals and the pricing of sovereign debt. Review of Finance, 14(2), 235–262. Holmström, B., & Tirole, J. (1998). Private and public supply of liquidity. Journal of Political Economy, 106(1), 1–40. https://doi​.org​/10​.1086​/250001. IMF. (2017). State-contingent debt instruments for sovereigns. IMF policy paper. IMF. (2020). The role of state-contingent debt instruments in sovereign debt restructurings. Staff discussion notes. International Monetary Fund. IMF. (2021). Issues in restructuring of sovereign domestic debt. Policy Paper, November. Washington: International Monetary Fund. Jeanne, O. (2003). Why do emerging economies borrow in foreign currency? Working Paper 03/177. Washington, DC: International Monetary Fund. Jiang, Z., Lustig, H., Van Nieuwerburgh, S., & Xiaolan, M. (2019). The U.S. public debt valuation puzzle (No. w26583; p. w26583). National Bureau of Economic Research. https://doi​.org​/10​.3386​/w26583. Jiang, Z., Lustig, H., Van Nieuwerburgh, S., & Xiaolan, M. (2021). What drives variation in the U. S. Debt/output ratio? The dogs that didn’t bark (No. w29351; p. w29351). National Bureau of Economic Research. https://doi​.org​/10​.3386​/w29351. Kaminsky, G. L., Reinhart, C. M., & Végh, C. A. (2004). When it rains, it pours: Procyclical capital flows and macroeconomic policies. NBER Macroeconomics Annual, 19, 11–53. https://doi​.org​/10​ .1086​/ma​.19​.3585327. Kamstra, M. J., & Shiller, R. J. (2010). Trills instead of t-bills: It’s time to replace part of government debt with shares in gdp. The Economists’ Voice, 7(3). https://doi​.org​/10​.2202​/1553​-3832​.1782. Kimmel, C. (2019). Characteristics of government bond markets: Primary vs. secondary markets. Mimeo, Deutschland Bundesbank. Kohlscheen, E. (2010). Sovereign risk: Constitutions rule. Oxford Economic Papers, 62(1), 62–85. Krugman, P. (1988). Financing vs. forgiving a debt overhang (No. w2486; p. w2486). National Bureau of Economic Research. https://doi​.org​/10​.3386​/w2486. Kuvshinov, D. & Zimmermann, K. (2017): Sovereigns Going Bust: Estimating the cost of default, Working Paper. Kuvshinov, D., & Zimmermann, K. (2019). Sovereigns going bust: Estimating the cost of default, European Economic Review, 119, 1–21. https://doi​.org​/10​.1016​/j​.euroecorev​.2019​.04​.009. Lizarazo, S. V. (2013). Default risk and risk averse international investors. Journal of International Economics, 89(2), 317–330. https://doi​.org​/10​.1016​/j​.jinteco​.2012​.08​.006. Longstaff, F. A., Pan, J., Pedersen, L. H., & Singleton, K. J. (2011). How sovereign is sovereign credit risk? American Economic Journal: Macroeconomics, 3(2), 75–103. https://doi​.org​/10​.1257​/mac​.3​.2​.75.

402  Research handbook of financial markets

Lorenzoni, G., & Werning, I. (2019). Slow moving debt crises. American Economic Review, 109(9), 3229–3263. Malone, S. W. (2011). Sovereign indebtedness, default, and gambling for redemption. Oxford Economic Papers, 63(2), 331–354. https://doi​.org​/10​.1093​/oep​/gpq021. Manasse, P., & Roubini, N. (2009). “Rules of thumb” for sovereign debt crises. Journal of International Economics, 78(2), 192–205. Marchesi, S., & Masi, T. (2021). Life after default: Official vs. private sovereign debt restructurings. Journal of International Money and Finance, 113, 1–23. Marchesi, S., & Masi, T. (2021). Life after default: Private and official deals. Journal of International Money and Finance, 113, 102339. Martínez, J., & Santiso, J. (2003). Financial markets and politics: The confidence game in Latin American emerging economies. International Political Science Review, 24(3), 363–395. https://doi​ .org​/10​.1177​/0192512103024003005. Mendoza, E. G. (1995). The terms of trade, the real exchange rate, and economic fluctuations. International Economic Review, 36(1), 101. https://doi​.org​/10​.2307​/2527429. Mendoza, E. G., & Yue, V. Z. (2012). A general equilibrium model of sovereign default and business cycles. Quarterly Journal of Economics, 127(2), 889–946. Meyer, J., Reinhart, C., & Trebesch, C. (2021). Sovereign bonds since Waterloo. Quarterly Journal of Economics, 137(3), 1615–1680. Mitchener, K. J., & Trebesch, C. (2021). Sovereign debt in the 21st century: Looking backward, looking forward. Journal of Economic Literature [in press] (also published as NBER Working Paper 28598). Mitchener, K. J., & Weidenmier, M. (2005). Empire, public goods, and the Roosevelt corollary. The Journal of Economic History, 65(3). https://doi​.org​/10​.1017​/S0022050705000240. Moreno Badia, M., Medas, P., Gupta, P., & Xiang, Y. (2020). Debt is not free. IMF Working Paper 20/1. Moreno Badia, M., Gamboa Arbelaez, J., & Xiang, Y. (2021). Debt dynamics in emerging and developing economies: Is R-G a red herring?. IMF Working Paper 21/229. Moretti, M. (2020). Financial innovation and liquidity premia in sovereign markets: The case of GDPlinked bonds. Retrieved from https://mmorettifiles​.github​.io​/ Paper​_2020​_GDPBonds​.pdf. Moser-Boehm, P. (2006). The relationship between the central bank and the government. In Central banks and the challenge of development. Bank for International Settlements. Neumeyer, P. A., & Perri, F. (2005). Business cycles in emerging economies: The role of interest rates. Journal of Monetary Economics, 52(2), 345–380. https://doi​.org​/10​.1016​/j​.jmoneco​.2004​.04​.011. OECD. (2019). OECD sovereign borrowing outlook 2019. Paris: OECD Publishing. Ottonello, P., & Perez, D. J. (2019). The currency composition of sovereign debt. American Economic Journal: Macroeconomics, 11(3), 174–208. Panizza, U., Sturzenegger, F., & Zettelmeyer, J. (2009). The economics and law of sovereign debt and default. Journal of Economic Literature, 47(3), 651–698. Passadore, J., & Xu, Y. (2020). Illiquidity in sovereign debt markets (SSRN Scholarly Paper ID 3195968). Social Science Research Network. Retrieved from https://papers​.ssrn​.com​/abstract​=3195968. Persson, T., & Svensson, L. E. O. (1989). Why a stubborn conservative would run a deficit: Policy with time-inconsistent preferences. The Quarterly Journal of Economics, 104(2), 325. https://doi​.org​/10​ .2307​/2937850. Pouzo, D., & Presno, I. (2016). Sovereign default risk and uncertainty premia. American Economic Journal: Macroeconomics, 8(3), 230–266. https://doi​.org​/10​.1257​/mac​.20140337. Rajan, R., & Tokatlidis, I. (2005). Dollar shortages and crises. International Journal of Central Banking, 1(2), 177–220. Reinhart, C. M., & Rogoff, K. S. (2011). From financial crash to debt crisis. American Economic Review, 101(5), 1676–1706. https://doi​.org​/10​.1257​/aer​.101​.5​.1676. Reis, R. (2021). The constraint on public debt when r < g but g < m. BIS Working Papers, No 939. Retrieved from https://www​.bis​.org​/publ​/work939​.htm. Rivoli, P., & Brewer, T. L. (1997). Political instability and country risk. Global Finance Journal, 8(2), 309–321. https://doi​.org​/10​.1016​/S1044​- 0283(97)90022-3. Roch, F., & Roldán, F. (2021). Uncertainty premia, sovereign default risk, and state-contingent debt. IMF working papers, forthcoming at JPE Macroeconomics. https://doi.org/10.1086/723950.

Sovereign debt 

403

Roldán, F. (2020). The aggregate-demand doom loop: Precautionary motives and the welfare costs of sovereign risk (SSRN Scholarly Paper ID 3774928). Social Science Research Network. Retrieved from https://papers​.ssrn​.com ​/abstract​=3774928. Sachs, J. (1984). Theoretical issues in international borrowing. Princeton Studies in International Finance, No 54. Sandleris, G. (2008). Sovereign defaults: Information, investment and credit. Journal of International Economics, 76(2), 267–275. Schaechter, M. A., Kinda, M. T., Budina, M. N., & Weber, A. (2012). Fiscal Rules in Response to the Crisis: Toward the Next-Generation Rules: A New Dataset, IMF Working Paper. Schumacher, J., Trebesch, C., & Enderlein, H. (2021). Sovereign defaults in court. Journal of International Economics, 131, 103388. https://doi​.org​/10​.1016​/j​.jinteco​.2020​.103388. Shiller, R. J. (2007). Macro markets: Creating institutions for managing society’s largest economic risks (Repr). Oxford: Oxford University Press. Sturzenegger, F., & Zettelmeyer, J. (2007). Debt defaults and lessons from a decade of crises. The MIT Press. https://doi​.org​/10​.7551​/mitpress​/2295​.001​.0001. Sturzenegger, F., & Zettelmeyer, J. (2008). Haircuts: Estimating investor losses in sovereign debt restructurings, 1998–2005. Journal of International Money and Finance, 27(5), 780–805. https://doi​ .org​/10​.1016​/j​.jimonfin​.2007​.04​.014. Talvi, E., & Végh, C. A. (2005). Tax base variability and procyclical fiscal policy in developing countries. Journal of Development Economics, 78(1), 156–190. https://doi​.org​/10​.1016​/j​.jdeveco​.2004​.07​.002. Tomz, M., & Wright, M. L. J. (2007). Do countries default in “bad times”? Journal of the European Economic Association, 5(2–3), 352–360. https://doi​.org​/10​.1162​/jeea​.2007​.5​.2​-3​.352. Tourre, F. (2017). A macro-finance approach to sovereign debt spreads and returns. Retrieved from https://ideas​.repec​.org​/p​/red​/sed017​/13​.html. Trebesch, C., & Zabel, M. (2017). The output costs of hard and soft sovereign default. European Economic Review, 92, 416–432. https://doi​.org​/10​.1016​/j​.euroecorev​.2016​.10​.004. Turner, P. (2002). Bond markets in emerging economies: An overview of policy issues. In Bank for International Settlements (Ed.), The development of bond markets in emerging economies (Vol. 11, pp. 1–12). Bank for International Settlements. United Nations. (2006). GDP-indexed bonds: Making it happen. Economic & Social Affairs. Uribe, M., & Yue, V. Z. (2006). Country spreads and emerging countries: Who drives whom? Journal of International Economics, 69(1), 6–36. https://doi​.org​/10​.1016​/j​.jinteco​.2005​.04​.003. Van Rijckeghem, C., & Weder di Mauro, B. (2004). The politics of debt crises (SSRN Scholarly Paper ID 636043). Social Science Research Network. Retrieved from https://papers​ .ssrn​ .com​ /abstract​ =636043. Van Wijnbergen, S., Olijslagers, S., & de Vette, N. (2020). Debt sustainability when r −g < 0: No free lunch after all (SSRN Scholarly Paper ID 3736921). Social Science Research Network. Retrieved from https://papers​.ssrn​.com ​/abstract​=3736921. Vegh, C., & Vuletin, G. (2014). The road to redemption: Policy response to crises in Latin America (No. w20675; p. w20675). National Bureau of Economic Research. https://doi​.org​/10​.3386​/w20675. Willems, T., & Zettelmeyer, J. (2022). Sovereign debt sustainability. Annual Review of Financial Economics [in press] (also available as IMF Working Paper No. 2022/016). Woodford, M. (1990). Public debt as private liquidity. The American Economic Review, 80(2), 382–388. Wright, M. (2002). Reputations and sovereign debt. Retrieved from https://eml​.berkeley​.edu/​~obstfeld​/ e281​_sp03​/wright​.pdf. Wright, M. L. J. (2005). Coordinating creditors. American Economic Review, 95(2), 388–392. https:// doi​.org​/10​.1257​/000282805774670437. Yeyati, E. L., & Panizza, U. (2011). The elusive costs of sovereign defaults. Journal of Development Economics, 94(1), 95–105. https://doi​.org​/10​.1016​/j​.jdeveco​.2009​.12​.005. Zettelmeyer, J., Trebesch, C., & Gulati, M. (2013). The Greek debt restructuring: An autopsy. Economic Policy, 75, 513–563. https://doi​.org​/10​.1111​/1468​- 0327​.12014.

​

404  Research handbook of financial markets

Table 17A.1  Business cycles in emerging and developed economies (a) Standard deviations % Standard deviation

% Standard deviation

% Standard deviation of GDP GDP

R

NX

PC

TC

INV

EMP

HRS

Emerging economies Argentina

3.51

3.71

1.16

3.84

3.54

9.60

—

—

Brazil

1.66

1.93

0.92

2.10

3.37

5.47

1.34

2.50

Korea

3.29

1.30

2.60

5.15

5.60

8.34

2.04

1.31

Mexico

2.22

1.68

1.21

2.22

1.96

7.96

1.21

1.57

Philippines

1.10

1.11

1.51

0.75

1.10

6.41

2.19

—

Average

2.46

3.56

1.55

2.95

3.19

8.33

1.69

1.79

Developed economies Australia

0.92

1.40

0.86

0.88

0.76

4.26

1.31

1.71

Canada

1.23

1.16

0.68

0.82

0.56

3.73

2.04

0.81

Netherlands

1.17

0.80

1.62

0.98

0.80

7.42

1.15

—

New Zealand

1.62

1.74

1.12

1.59

1.26

5.71

1.69

0.39

Sweden

1.54

1.16

0.54

1.21

0.77

4.82

1.77

2.99

Average

1.29

1.25

0.97

1.10

0.83

5.19

1.59

1.48

(b) Correlations with GDP R

NX

PC

TC

INV

EMP

HRS

-0.72

-0.91

0.98

0.98

0.99

—

—

Emerging economies Argentina Brazil

-0.11

-0.70

0.81

0.50

0.90

0.60

0.54

Korea

-0.60

-0.91

0.97

0.93

0.96

0.89

0.37

Mexico

-0.24

-0.62

0.93

0.92

0.91

0.70

-0.28 —

Philippines

-0.10

0.16

0.22

0.23

0.65

0.12

Average

-0.40

-0.59

0.78

0.71

0.88

0.58

0.21

Developed economies Australia

0.40

-0.45

0.46

0.41

0.77

0.49

0.16

Canada

0.60

-0.03

0.74

0.62

0.75

0.31

0.29

Netherlands

0.72

-0.05

0.75

0.67

0.40

0.43

—

New Zealand

0.14

-0.11

0.70

0.67

0.63

0.39

0.31

Sweden

0.37

-0.10

0.65

0.55

0.81

0.52

0.70

Average

0.45

-0.15

0.66

0.59

0.67

0.43

0.37 (Continued)

Sovereign debt 

405

Table 17A.1  (Continued) (c) Correlations with interest rate NX

PC

TC

INV

EMP

HRS

0.84

-0.77

-0.80

-0.70

—

—

Emerging economies Argentina Brazil

0.24

-0.14

-0.12

-0.10

-0.38

-0.25

Korea

0.67

-0.65

-0.69

-0.57

-0.57

-0.27

Mexico

0.38

-0.21

-0.22

-0.43

-0.60

0.18

Philippines

0.30

0.21

0.12

-0.25

-0.42

—

Average

0.49

-0.31

-0.34

-0.41

-0.49

-0.11

-0.47

0.54

0.50

0.39

0.70

0.11

0.00

0.47

0.44

0.41

0.28

-0.23 —

Developed economies Australia Canada Netherlands

0.00

0.53

0.41

0.24

0.49

New Zealand

-0.30

0.14

0.18

0.44

0.66

0.22

Sweden

-0.25

0.00

0.04

0.45

0.45

0.16

Average

-0.20

0.33

0.31

0.39

0.52

0.07

Note:   Net exports (NX) are exports minus imports over GDP. Real interest rates (R) are in percentage points. Total consumption (TC) includes private (PC) and government consumption, changes in inventories, and statistical discrepancy. Investment (INV) is gross fixed capital formation. Employment (EMP) is number of workers, and total hours (HRS) is number of workers times weekly hours of work per worker. All series except net exports and real interest rates are in logs. All series have been Hodrick–Prescott filtered. Sources:   Authors’ calculations based on quarterly official data.

PART V DERIVATIVE MARKETS

18. Interest rate swaps1 Bin Wei and Vivian Z. Yue

18.1 INTRODUCTION AND HISTORY1 Interest rate swaps are important derivative contracts. According to the 2009 ISDA Derivatives Usage Survey, 83% of the world’s 500 largest companies use interest rate derivatives for risk management.2 An interest rate swap is an agreement to exchange streams of interest payments between two counterparties during the tenor (maturity) of the swap. We refer to the streams of payments as legs. Both legs of a swap typically have the same first and last dates when interest begins and ends accruing. These dates are referred to as the swap’s Effective Date and Maturity Date, respectively. The interest payments on both legs are based on the same hypothetical principal, which is referred to as the notional of the swap. Depending on the (fixed or floating) nature of payment streams, interest rate swaps are categorized as either coupon swaps or basis swaps. A coupon swap, also known as a fixedfloating swap, refers to a swap in which one counterparty—called “fixed payer” or “floating receiver”—agrees to pay a fixed rate of interest in exchange for receiving a floating rate of interest and the other counterparty—called “floating payer” or “fixed receiver”—agrees to pay floating and receive fixed. In contrast, in a basis swap both counterparties receive floating rates of interest. Therefore, a fixed-floating swap’s cash flows consist of a fixed leg and a floating leg, whereas a basis swap has both floating legs. For swaps with streams of payments in different currencies, we refer to them as cross-currency swaps. The first cross-currency swap was executed between the World Bank and IBM in 1981. Since then, the interest rate swap market has grown rapidly. Figure 18.1 plots the outstanding notional of interest rate swaps between 1998 and 2020. The total outstanding notional for interest rate swaps (coupon, basis, or cross-currency swaps combined) exceeded $100 trillion in 2003 for the first time and then reached almost $500 trillion in 2008. According to the BIS’s “Triennial Central Bank Survey”, the average daily turnover of interest rate swaps increased from $63 billion in 1995 to above $4 trillion in 2019. The share of USD-denominated interest rate swaps steadily increased from 27% in 1995 to 45% in 2019. At the same time, the dealer share of trading volume decreased substantially from 66% in 1995 to 22% in 2019. Figure 18.2 plots the outstanding notional of fixed-floating swaps between 2014 and 2020 based on data from weekly CFTC Swaps Reports. As shown in the figure, the vast majority of fixed-floating

1 We are very grateful for helpful comments from Refet Gürkaynak, Ricardo Reis, and Jonathan Wright. The views expressed here are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Federal Reserve Bank of Atlanta or the Federal Reserve System. Any remaining errors are the authors’ responsibility. 2 The survey is available from the following website: www​.isda​.org​/2009​/04​/22​/derivative​-usage​survey/.

407

408  Research handbook of financial markets 700 Coupon/Basis Swaps Cross-currency Swaps Other

600

USD Trillions

500 400 300 200 100 0

1998

2000

2002

2004

2006

2008

2010

2012

2014

2016

2018

2020

Note:   The figure shows the size (notional) of the interest rate swap market in trillion dollars between 1998 and 2020. The blue bar depicts the notional value of swaps denominated in the same currency. The red bar depicts the notional for cross-currency swaps. For comparison, the gold bar depicts the notional for interest rate options and forward rate agreements (FRA). Source:   Bank for International Settlements.

Figure 18.1  Size of the interest rate swap market

250

USD Trillions

12-24 24-60 60+

0-3 3-6 6-12

200 150 100 50 0 2014Jan

2015Jan

2016Jan

2017Jan

2018Jan

2019Jan

2020Jan

Note:   The figure shows the size (notional) of fixed-float interest rate swaps by tenor in trillion dollars between 2014 and 2020. Totally six tenor buckets are considered: less than three months (“0–3”), between three and six months (“3–6”), between six and 12 months (“6–12”), between 12 and 24 months (“12–24”), between twenty four and 60 months (“24–60”), more than 60 months (“60+”). The gap in January 2019 is due to the lack of Swaps Report data during this period. Source:   Commodity Futures Trading Commission (CFTC) Swaps Report.

Figure 18.2  Fixed-floating swap gross notional by tenor

Interest rate swaps 

409

swaps have a tenor of two years or longer, which account for about 60% to 80% of the total outstanding notional. Interest rate swaps are typically traded bilaterally over-the-counter (OTC). In the OTC swap market, trades are negotiated directly between counterparties and not through an exchange. In the early days of the market, the interest rate swaps were custom-tailored. If an intermediary had a customer for one side of a swap but could not find another customer to take the other side, the intermediary would need to warehouse the swap to get the deal done. To make warehousing practical and less risky to intermediaries, standardization of swap terms emerged as a solution to make the swaps market more liquid. As a product of standardization, plain vanilla interest rate swaps have become the most widely used among all interest rate swaps in the market. The benchmark floating rate for plain vanilla swaps is the three-month London interbank offered rate (LIBOR). However, the integrity of LIBOR was called into question during the 2008–2009 financial crisis. There is mounting evidence that a number of LIBOR panel banks deliberately under-reported their costs of borrowing at the height of the finance crisis.3 The United Kingdom’s Financial Conduct Authority (FCA) and ICE Benchmark Administration (IBA) announced definitive end dates for LIBOR; for example, no USD LIBOR tenors will be available after June 30, 2023. The Federal Reserve and other regulators also published guidance that supervised institutions should stop produce new LIBOR contracts by the end of 2021. The Secured Overnight Financing Rate (SOFR) is likely the new benchmark rate replacing USD LIBOR. The last decade has seen other new developments in the market. The 2008–2009 financial crisis exposed significant weaknesses in the OTC derivatives markets such as heightened risks and opaqueness in counterparty exposures. In response, the G20 Leaders agreed in 2009 to reforms in the OTC derivatives markets. The reforms require trading of standardized OTC derivatives on exchanges or electronic trading platforms, central clearing through central counterparties, reporting of all transactions to trade repositories, and higher capital and margin requirements for non-centrally cleared transactions.

18.2 TYPES OF INTEREST RATE SWAPS There are three types of interest rate swaps: coupon or fixed-floating swaps, basis swaps, and cross-currency swaps. The streams of payments are denominated in the same currency in coupon or basis swaps, but in different currencies in cross-currency swaps. In this section, we provide more details about each type.

3 See, e.g., a Wall Street Journal article on May 29, 2008, titled “Study Casts Doubt on Key Rate: WSJ Analysis Suggests Banks May Have Reported Flawed Interest Data for Libor”. In some cases, LIBOR manipulation existed well before the financial crisis (see, e.g., Statement of Facts, U.S. Dept. of Justice and Barclays Bank PLC (June 26, 2012), available at www​.justice​.gov /iso/​opa/r​esour​ces/9​​ 31201​​27101​​73426​​36594​​​1​.pdf​).

410  Research handbook of financial markets

Swap rate (fixed rate payment)

Fixed Rate Receiver

Fixed Rate Payer LIBOR (floating rate payment)

Note:   The figure shows cash flows for an example of a plain vanilla fixed-floating interest rate swap. We consider a five-year plain vanilla swap in the example.

Figure 18.3  Plain vanilla fixed-floating swap 18.2.1 Coupon or Fixed-Floating Swaps To be concrete, we focus on two most popular types of fixed-floating swaps: plain vanilla swaps and Overnight Index Swaps (OIS). In the USD swap market, the floating leg in an OIS is tied to the Federal Funds effective rate, whereas the floating leg in a plain vanilla swap is tied to a longer-term USD LIBOR rate (e.g., three-month).4 18.2.1.1 Plain vanilla swaps The point of reference in a plain vanilla swap is the fixed rate: if one is receiving (respectively, paying) the fixed rate, one is said to be receiving in a swap, or simply receiving (respectively, paying in a swap, or simply paying). We can think of receiving in a swap as a long position in a bond since one is receiving coupon payments and paying periodic financing at the floating rate. Similarly, paying in a swap can be thought of as a short position in a bond. The structure of a plain vanilla swap is shown in Figure 18.3. Under standard market conventions, one counterparty (“Fixed Rate Receiver”) pays a floating rate (e.g., three-month LIBOR), typically quarterly on an Act/360 basis, and the other counterparty (“Fixed Rate Payer”) pays a fixed rate, typically semiannually on a 30/360 basis. The fixed rate is known as the swap rate and set when the swap is first executed such that its net present value is zero to either counterparty. Both counterparties have an International Swaps and Derivatives Association (ISDA) Master Agreement, which provides standardized documentation for most swaps nowadays. A spot starting interest rate swap in USD has an effective date, which is two New York and London business days after the trade date (i.e., “T + 2”). The effective date is the date from which interest payments start to accrue. For example, for a USD swap transacted on August 16, 2017, its effective date is August 18, 2017. On the fixed leg of the swap, the fixed rate is set on the trade date such that the swap has a net present value of zero, which is called the swap rate or the par swap rate. As a result, the fixed payment is determined as follows

Fixed Payment = Fixed Rate ´ Notional ´ (Bond Days/360), (18.1)

4 The Federal Funds effective rate is the rate that banks lend to one another overnight in the Federal Funds market. The reference floating rates for other major currencies are “Euro Overnight Index Average” (EONIA) for the euro, “Sterling Overnight Index Average” (SONIA) for British pound, and “Tokyo Overnight Average Rate” (TONAR) for Japanese yen.

Interest rate swaps 

411

where “Bond Days” denotes the number of days in the semiannual interest accrual period on a 30/360 basis. Oftentimes, it is the swap spread that is quoted instead of the swap rate. The swap spread for a swap of a given maturity is defined as the difference in basis points between the swap rate and the yield of the on-the-run Treasury of comparable maturity. On the floating leg of the swap, we assume that the three-month LIBOR rate is paid quarterly on an Act/360 basis for concreteness. The LIBOR rate is set on each reset date, which is actually the LIBOR setting determined by the BBA two London business days prior. Once the LIBOR rate is reset, that rate will accrue for the quarter and payment will be made three months later on a Modified Following basis. A Modified Following basis means that if the payment date falls on a “bad day” (i.e., a weekend or holiday), then the payment is made on the first “good” day after the bad day if that good day is not in the next month, or, otherwise, made on the first good day before the bad day.5 As a result, a floating payment is determined as follows

Floating Payment = 3 - month LIBOR ´ Notional ´ (Act/360), (18.2)

where “Act” denotes the number of days in the quarterly interest accrual period on a Act/360 basis. Example 1: Suppose two counterparties transact a spot starting five-year interest rate swap in USD on August 16, 2017. The swap has $100 million notional value. It thus has an effective date of August 18, 2017, and a maturity date of August 18, 2022. The floating rate payer is a swap dealer, and the fixed rate payer is its client. The ten fixed rate payment dates fall on the 18th of each February and August between the effective and maturity dates inclusive. The 20 floating rate payment dates fall on the 18th of each February, May, August, and November between the effective and maturity dates inclusive. In the swap market, the bid side swap rate is the highest rate at which a dealer is willing to pay fixed, while the offer side swap rate is the lowest rate at which a dealer is willing to receive fixed. In this example, as shown in Table 18.1, the offer side (five-year) swap rate was 1.858% on August 16, 2017. Therefore, each fixed rate payment is given by

1.858% ´ $100 million ´

180 = $929,000. 360

The three-month LIBOR rate was 1.317% on August 16, 2017. There are 92 actual days in the first interest accrual period, implying the first floating rate payment below:

1.317% ´ $100 million ´

92 = $336,567. 360

5 Note that the Modified Following basis is a hybrid of the other two bases. “Preceding” (payments are made on the first good day before the bad day) and “Following” (payments are made on the first good day after the bad day) if scheduled payment dates fall on a bad day.

412  Research handbook of financial markets

Table 18.1  Quotes of Treasury securities and Interest rate swaps (August 16, 2017) Term 1-year 2-year 3-year 4-year 5-year 6-year 7-year

Treasury

Swap

Price

Yield

Mid-Yield

Rate

Spread

—

—

1.198

1.466–1.470

27.0

1.330–1.326

1.328

1.586–1.591

26.1

1.487–1.484

1.485

1.686–1.691

20.3

—

1.633

1.772–1.777

14.2

1.782–1.781

1.781

1.853–1.858

7.4

—

1.909

1.932–1.936

2.5

2.038–2.035

2.037

2.002–2.007

–3.2

3 100-02 –100-03 4 1 1 100-01 –100-01 4 2 — 1 100-14–100-14 4 — 1 100-18–100-18 2

8-year

—

—

2.099

2.066 –2.070

–3.1

9-year

—

—

2.161

2.123–2.127

–3.6

10-year

1 100-07 –100-08 2

2.224–2.222

2.223

2.173–2.178

–4.7

12-year

—

—

2.282

2.259–2.263

–2.0

15-year

—

—

2.369

2.345–2.349

–2.3

20-year

—

—

2.516

2.425–2.430

–8.8

25-year

—

—

2.662

2.459–2.463

–20.0

30-year

1 98-26 –98-27 2

2.808–2.807

2.808

2.472–2.476

–33.4

Note:   This table reports pricing information for Treasuries and swaps of different tenors. Source:  Bloomberg.

However, floating rate payments will not be known until each LIBOR setting date, except for the very first floating rate payment.6 18.2.1.2 OIS swaps In an OIS swap, one counterparty pays the fixed rate known as the OIS rate, and the other pays an average of the Federal Funds effective rate over the life of the swap. The OIS rate for a given maturity is thus considered as the expected Federal Funds rate over the life of the swap. The floating payment in an OIS swap is typically the daily compounded interest over the calculation period, while the fixed payment is the fixed rate accrued for the same calculation

6 For a maturity between adjacent ones of on-the-run Treasury securities we use linear interpolation to compute a mid-market yield for that maturity. For example, using the six-month yield of 1.133% and the two-year yield of 1.328%, we find the interpolated one-year yield as 1-1 / 2 ´ (1.328% - 1.133%) = 1.198%. 1.133% + 2 -1 / 2

Interest rate swaps 

413

400 350

Basic points

300 250 200 150 100 50 0 Jan06 Jan07 Jan08 Jan09 Jan10 Jan11 Jan12 Jan13 Jan14 Jan15 Jan16 Jan17 Jan18 Jan19 Jan20 Jan21

Note:   The figure shows the historical three-month LIBOR-OIS spread between January 2006 and September 2021. Source:   Bloomberg and authors’ calculations.

Figure 18.4  Historical LIBOR-OIS spread period. OIS swaps of maturity less than one year have one calculation period, while those of longer maturity are broken into annual calculation periods.7 The LIBOR-OIS spread, defined as the difference between a LIBOR rate and an OIS rate of the same maturity, is considered by many an important gauge for the health of the financial system. It is viewed as the risk premium demanded by banks to lend to one another for a longer period (e.g., three-month), compared to lending overnight in the Fed Funds market. Figure 18.4 shows the historical LIBOR-OIS spread since 2006. As shown in the figure, the LIBOR-OIS spread is typically very stable and low during normal times, but widened out significantly in times of stress. For example, it peaked in excess of 350 basis points when Lehman failed, and spiked to nearly 150 basis points at the outbreak of recent COVID-19 shock. 18.2.2 Basis Swaps In a basis swap, both counterparties receive floating rate payments. A spread is added to one of the floating legs to make the NPV of the swap equal to zero. Consider the basis swap in Figure 18.5 as an example. In the example, a counterparty (“Client”) pays one-month LIBOR plus a spread α on the monthly Act/360 basis, and the other counterparty (“Dealer”) pays three-month LIBOR on the quarterly Act/360 basis. The tenor of the swap is two years and the notional is $100 million. This swap is often called a 1s3s (pronounced “ones threes”) basis swap. The spread α is typically nonzero and is determined by various factors. The most important determinant is supply and demand: the spread increases with demand for paying one-month LIBOR, and vice-versa. Another important determinant is the slope of the LIBOR curve: the spread would be positive if market participants expect that one-month LIBOR will be lower 7  More precisely, the floating payment over the calculation period is given by (1 + a1 ´ r1 ) ´ (1 + a1 ´ r1 ) ´  ´ (1 + a N ´ rN ) - 1, where N denotes the number of compounding days in the calculation period, αi is the length of each compounding day (e.g., ai = 1 / 360 or 3 / 360 for Fridays in USD), and ri denotes the overnight reference rate.

414  Research handbook of financial markets 1-month LIBOR +

Dealer

Client 3-month LIBOR

Note:   The figure shows cash flows for an example of a basis swap. In the example, we consider a two-year basis swap.

Figure 18.5   Basis swap than three-month LIBOR during the life of the swap. Lastly, the frequency of floating-rate payments matters as well: the monthly payments on the one-month LIBOR side mean more frequent and sooner payments, which helps lower the spread or turn it into negative. The swap spreads of fixed-floating and basis swaps are intrinsically intertwined. Consider the two-year basis swap as an example, from Table 18.1; the dealer is willing to pay fixed 1.586% (semi-annual 30/360) versus three-month LIBOR (quarterly Act/360). If the dealer in the basis swap example simultaneously enters into the above fixed-floating swap with another client, the two three-month LIBOR legs in both swaps cancel out each other and the dealer ends up paying fixed 1.586% versus receiving one-month LIBOR plus α. The immediate implication is that the dealer would be willing to pay fixed (1.586% - a) versus receiving one-month LIBOR, or (1.586% - a) would be the bid side of the swap rate for a fixed-floating swap versus one-month LIBOR.8 18.2.3 Cross-Currency Swaps A cross-currency swap is one in which the streams of payment are in different currencies during the life of the swap. The streams of payment can both be fixed, both be floating, or consist of one fixed stream and one floating stream. At the beginning and end of the transaction, the two counterparties typically exchange specific amounts of the two different currencies, which are determined by the spot exchange rate between the two currencies at the start of the transaction. Note that regardless of how the exchange rate fluctuates at maturity, the final exchange of the currencies is based solely on the same spot exchange rate at the beginning of the transaction. Consider the cross-currency swap between U.S. dollars and Japanese yen in Figure 18.6 as an example. Suppose the effective date is August 16, 2017. On the effective date, the spot yen/$ exchange rate was 110.19 (i.e., $1 can be exchanged for 110.19 Japanese yen) and the spread for a one-year cross-currency swap between USD and JPY was –49.125 basis points (i.e., a = -49.125). Initially, the dealer receives USD $100 million and pays Japanese yen 11.019 billion. Then the dealer pays USD three-month LIBOR on the quarterly Act/360 basis, and the client pays JPY three-month LIBOR minus 49.125 basis points on the quarterly Act/360 basis.

8 Note that the spread α in the 1s3s basis swap is quoted on the monthly Act/360 basis, whereas the fixed leg of the fixed-floating swap is semi-annual 30/360. For simplicity, we ignore the difference in bases (the resulting error is small in magnitude) and simply subtract α from 1.586% to derive the swap rate.

Interest rate swaps 

415

JPY LIBOR +

Dealer

Client USD LIBOR

Note:   The figure shows the cash flows for an example of a one-year cross-currency swap.

Figure 18.6   Cross-currency swap The tenor of the swap is one year. At maturity, the dealer pays Japanese yen 11.019 billion and receives USD $100 million. One way to interpret the above cross-currency swap is as follows. The dealer receives a “dollar loan” of $100 million on the effective date and pays USD LIBOR as the financing cost. Similarly, the client receives a “yen loan” of 11.019 billion and pays JPY LIBOR as the financing cost. At maturity, both counterparities pay off the loans to each other.

18.3 RISKS An investor of a fixed-floating swap faces three primary sources of risk: interest rate risk, spread risk, and counterparty risk. There is an additional source of risk in a cross-currency swap: currency risk. As the other side of the same coin, swaps are used as an effective hedge against these risks, which is an important use of swaps. Movements in interest rates cause changes in the value of a swap. For example, when the swap curve shifts downward (upward) in a market rally (sell-off), the fixed rate receiver gains (loses) and the floating rate receiver loses (gains). So investors face interest rate risk as the value of their swap positions is sensitive to interest rate changes. Investors can engage in a spread trade to hedge the interest rate risk. If one receives fixed and sells Treasuries to hedge, one is being short spreads or having sold spreads, or receiving in spreads. Conversely, if one pays fixed and buys Treasuries to hedge, one is being long spreads or having bought spreads, or paying in spreads. If the swap rate changes one for one with the Treasury yield of the same maturity (i.e., the swap spread is unchanged), the value of the swap in a spread trade is hedged against movements in the Treasury yield for investors. However, investors still face spread risk: when spreads go up (down), investors who are long (short) spreads will gain (lose). Therefore, another important use of interest rate swaps is speculation, whereby investors express a view on movements in swap spreads. We examine the determinants of swap spreads in Section 18.6. Counterparty risk is another source of risk in the swap market. It refers to the risk that leads to a loss for one counterparty when the other counterparty defaults and cannot honor its obligations. To reduce the counterparty risk, a Credit Support Annex (CSA) is included in addition to an ISDA Master Agreement. Under a CSA, the counterparty who is out of the money will have to post collateral to the other counterparty. The terms of a CSA will, among other things, specify the types of acceptable collateral, thresholds below which no collateral is posted, and minimum transfer amounts of collateral. As we will discuss shortly in Section 18.5, the reforms post the financial crisis require central clearing in the swap market, which further mitigates the counterparty risk.

416  Research handbook of financial markets

Currency risk is an additional risk in a cross-currency swap. As explained in Section 18.2.3, exchanges of the notional value of principal in different currencies are made at the effective and maturity date based on the initial spot exchange rate. If the exchange rate had moved dramatically over the life of the swap, one counterparty would suffer a potentially large loss (and may choose not to honor its obligations) if the exchange rate movements were not in its favor. As such, a cross-currency swap is subject to high counterparty risk even toward the end of its life, which is different from coupon or basis swaps that have no exchange of principal at maturity and thus have little counterparty risk near maturity.

18.4 SWAP USAGE AND TYPES OF USERS Primary uses of interest rate swaps are for balance sheet management, hedging, and speculation. Major participants in the swap market include corporations, banks, insurance companies, Government Sponsored Enterprises (GSEs), asset managers, hedge funds, besides the intermediaries such as broker-dealers and central counterparties. 18.4.1 Balance Sheet Management Interest rate swaps are one of most popular tools for balance sheet management. For example, banks tend to issue fixed coupon corporate bonds and enter into “receive fixed/pay floating” swaps. In this way, they can transform their liabilities into floating-rate ones to better match with their assets that consist of floating rate loans. Similarly, corporations with floating rate liabilities (e.g., loans linked to LIBOR) can enter into swaps where they pay fixed and receive floating. In the longer-tenor swap market (i.e., 10–30 years) the key players are insurance companies and pension funds, both of which have long-duration liabilities. To actively match the duration of their assets to that of liabilities, they have a strong demand for receiver swaps where they receive fixed and pay floating. Specifically, pension funds are required to minimize underfunding under the Pension Protection Act of 2006 and hence have the incentive to avoid duration mismatch that can produce future shortfalls (Klingler & Sundaresan, 2016). According to the surveys by the Chief Investment Officer magazine in 2013 and 2014, around 80% of more than 100 pension fund managers who were surveyed stated that they used interest rate swaps among other derivatives. GSEs and other large mortgage-backed securities (MBS) holders use interest rate swaps to manage their asset/liability profiles. Due to the prepayment option, MBS have negative convexity, meaning that their duration increases (decreases) as rates rise (fall). This is not desirable for GSEs or MBS portfolio managers, because the duration of their portfolios would increase exactly at times when rates increase and no one wants to hold a longer-duration MBS portfolio. To manage the undesirable increase in the duration of their portfolios, GSEs and MBS portfolio managers will turn to the swap market by paying fixed. 18.4.2 Hedging Corporations also use interest rate swaps for “rate-lock hedging” against uncertainty in the cost of their corporate bond issuance. Most corporate bond offerings are underwritten on

Interest rate swaps 

417

a “best-efforts” basis, whereby the underwriter promises to do its best to achieve the lowest possible yield for the issuer. The final yield remains unknown until the bond issuance is completed. As a result, corporate bond issuers who face uncertainty in the actual cost of their bond offerings due to interest rate changes employ interest rate swaps for “rate-lock hedging”. Specifically, issuers enter into forward-starting contracts to pay fixed and receive float. In this way, they are protected from rising interest rates: as rates rise, the increase in the value of their swap positions helps offset the losses from issuing bonds at higher market rates. On the other hand, if interest rates fall, the losses from their swap positions are compensated by benefits from issuing bonds at a lower yield. Essentially, entering into forward-starting swaps help the issuers lock in the cost of bonds that are to be issued in the near future. 18.4.3 Speculation Because interest rate swaps require little capital up front, investors can use them to speculate on movements in interest rates. For example, to speculate that five-year interest rates will fall, an investor could simply receive fixed in a five-year swap transaction, instead of using cash or borrowed capital to buy a five-year Treasury note. Another important use of interest rate swaps is to speculate on the direction of swap spreads. To express the view that swap spreads will increase (decrease) dramatically in the future, investors can speculate by entering into a long (short) spread position. In Section 18.6, we provide more details about spread trades. There is also evidence for corporations’ use of interest rate swaps for timing the market (Faulkender, 2005): when the yield curve is steep, firms use swaps to pay floating, whereas, when the yield curve is flat or inverted, they use swaps to lock in the relatively attractive longer-term fixed rate. The dependence of a firm’s swap positions on the slope of the yield curve suggests myopia on the part of the firm, or a failure of the expectations hypothesis, or some specific hedging motive.

18.5 REGULATION AND REFORMS Title VII of the Dodd-Frank Wall Street Reform and Consumer Protection Act (Dodd-Frank Act), signed into law in July 2010, provides a comprehensive framework for the regulation of the OTC swaps markets. It contains three broad mandates: the clearing mandate, the trading mandate, and the registration/reporting mandate. The first clearing mandate requires that certain derivative transactions must be centrally cleared through clearing houses. The U.S. Commodity Futures Trading Commission (CFTC) is responsible for determining which interest rate swaps are to be cleared.9 Non-centrally cleared transactions have higher capital and margin requirements, which raise the costs of OTC trading and thereby encourage a shift of trading activities to exchanges. The move to 9 The CFTC is also responsible for regulating other swaps, namely, FX swaps, credit default swaps on broad-based indices, and physical commodity swaps. The SEC is responsible for regulating security-based swaps, namely, single name credit default swaps, total return swaps on single name equity securities, loans, and narrow-based security indices. Both the CFTC and SEC jointly regulate mixed swaps.

418  Research handbook of financial markets

central clearing implies that each counterparty in a swap trade must face a central clearing house. In this way, counterparty risk, a prominent risk in the OTC derivative markets, is mitigated by the existence of central counterparties who provide guarantees on the contract. At the same time, because a single trade between two parities now become two trades of each party with the central counterparty, outstanding notionals tend to be inflated. Trade compression has gained traction partly because banks move toward more efficient balance sheet usage due to regulatory changes. The second trading mandate requires any swap that is required to be centrally cleared must be traded on either an exchange or a “Swap Execution Facility” (SEF). A SEF is defined in Section 721 of the Act as a “trading system or platform in which multiple participants have the ability to execute or trade swaps by accepting bids and offers made by multiple participants in the facility or system”.10 Trading on a SEF has the ability to maintain an OTC marketplace so that counterparties can continue to transact bilaterally but on a regulated trading system. The SEFs have served to move a large share of OTC swap trading to electronic platforms. The third registration/reporting mandate requires that swap dealers and major swap participants in the interest rate swap market must register with the CFTC, and requires reporting of all transactions to trade repositories.

18.6 DYNAMICS OF SWAP SPREADS In this section, we take a closer look at the dynamics and determinants of swap spreads. Figure 18.7 depicts the spreads of plain vanilla interest rate swaps for the tenors of two years (solid blue line), ten years (dashed red line), and 30 years (dotted black line) between January 2001 and September 2021. As we can see from the figure, the swap spreads moved closely and were similar in magnitude prior to the financial crisis, but started to diverge afterwards. In particular, the spreads for longer-tenor swaps turned negative and continued to be negative (e.g., 30-year swap spread). Interest rate swaps can trade either as rates or spreads. When one is trading rates, one is having a single (long or short) position in an interest rate swap and thus is taking duration risk. For example, receiving in a swap has a similar risk to buying the same notional amount of Treasuries of the same maturity. On the other hand, one can directly enter into a spread trade by entering into a swap and holding simultaneous offsetting positions in Treasuries, and possibly repurchase agreements known as “repo”. A repo is the sale of a security with the simultaneous agreement to repurchase the security at a predetermined price at some time later in the future. Consider “buying a spread” as an example. Its mechanics is illustrated in Figure 18.8. Suppose a swap dealer agrees to pay fixed in a $100 million two-year swap at a bid side swap rate of 1.586% (see Table 18.1), and at the same time buys $100 million two-year Treasuries at

10 The list of SEFs can be found at the CFTC’s website: https://sirt​.cftc​.gov​/SIRT​/SIRT​.aspx​?Topic​= Swa​pExe​cuti​onFa​cilities. Reforms in Europe resulted in multilateral trading facilities (“MTFs”) and organized trading facilities (“OTFs”). As the key difference, execution of transactions on MTFs is non-discretionary and rule-based, whereas an OTF is operated with a degree of discretion over how a transaction will be executed (e.g., discretion over who are accepted to trade and how trades are executed).

Interest rate swaps 

419

200

Basic points

150

2-year 10-year 30-year

100 50 0 -50 -100 Jan01

Jan03

Jan05

Jan07

Jan09

Jan11

Jan13

Jan15

Jan17

Jan19

Jan21

Note:   This figure depicts the spreads of plain vanilla interest rate swaps for the tenors of two years (solid blue line), ten years (dashed red line), and 30 years (dotted black line) between January 2001 and September 2021. Source:   Bloomberg and authors’ calculations.

Figure 18.7  Historical swap spreads the offer price 100-03 to hedge his swap position.11 Suppose the dealer is able to enter into a repo to fund the purchase of the Treasury securities for two years.12 When the repo matures, the dealer will take delivery of the Treasury securities and return the amount borrowed plus interest accrued using the repo rate. As a result, the net interest (i.e., carry) of this transaction is given by:

Treasury Yield - Swap Rate + LIBOR - Repo Rate

= (LIBOR - Repo Rate) - Swap Spread.

If the LIBOR-repo spread is larger than the swap spread, the dealer will be able to earn the positive difference. Conversely, if the LIBOR-repo spread is smaller than the swap spread, the dealer can still earn the opposite of the difference by shorting the spread (i.e., receive in the swap, short the Treasury securities, and invest the proceeds in a reverse repo). Note that the trade of longing or shorting the spread requires no capital upfront. Therefore, in theory, the difference should be zero in a frictionless market. That is, changes in swap spreads should be closely related to changes in the LIBOR-repo spread. In Figure 18.9, we plot the two-year swap spread as well as the spread between three-month LIBOR and three-month repo rates (used as a proxy for the LIBOR-repo spread). Even though the three-month LIBOR-repo spread is not a perfect proxy for the LIBOR-repo spread, it comoves closely with the two-year swap spread, especially after the financial crisis, as shown in 11 The purchase of the Treasury securities on August 16, 2017 would settle the next day on August 17, 2017. The total invoice price paid would include accrued interest. 12 A repo typically lasts for a relatively short period of time. For simplicity, we assume that a two-year repo exists. In practice, one needs to roll into another repo from time to time to get then-prevailing repo rates.

420  Research handbook of financial markets Treasury Yield

Treasury Market

Swap Rate

Investor LIBOR

Cash

Swap Market

Repo Rate

Cash

Repo Market

Note:   This figure depicts the mechanics behind “buying a spread” or “paying in a spread” in which an investors takes a long position in a swap spread.

Figure 18.8  Long swap spread position 400

Basic points

300

2-year swap spread LIBOR-repo spread

200 100 0 -100 Jan01

Jan03

Jan05

Jan07

Jan09

Jan11

Jan13

Jan15

Jan17

Jan19

Jan21

Note:   This figure depicts the two-year swap spread (solid blue line) and the three-month LIBOR-repo spread (dashed red line) between January 2001 and September 2021. Source:   Bloomberg and authors’ calculations.

Figure 18.9  The LIBOR-repo spread versus swap spread the figure. Note that the explanatory power of the three-month LIBOR-repo spread is much limited for longer-tenor swap spreads; for example, it has difficulty explaining the negative 30-year swap spread. Nevertheless, it plays a more important role in determining swap rates with shorter tenors. In practice, there are numerous other factors that drive swap spreads. Next, we discuss in more detail those determinants of swap spreads. First, one of the biggest factors that determines the short-run movement in swap spreads is demand and supply. For example, an increase (decrease) in the aggregate demand for paying fixed to dealers would widen (narrow) swap spreads. Demand and supply factors that affect Treasury yields can also impact swap spreads. Treasury yields, on the other hand, reflect mostly the risk of the U.S. government. In distressful times, investors may flock to the

Interest rate swaps 

421

Treasury market and safe-haven flows drive down Treasury yields and widen swap spreads. Similarly, changes in demand for Treasuries from large investors (e.g., foreign central banks) or changes in Treasury supply can also have a significant impact on swap spreads. Second, swap spreads tend to widen during periods of heightened financial market stress. During such periods, investors flock to the Treasury market and “flight-to-quality” flows drive down Treasury yields. In addition, LIBOR rates, which measure average borrowing costs for the LIBOR panel banks, tend to rise when there is a perceived deterioration in bank credit/ liquidity conditions. As a result, both lower Treasury yield and higher LIBOR rates work to widen swap spreads. Third, the slope of the yield curve is another factor. As the yield curve steepens, the difference between long-term and short-term rates increases. As a result, investors have more incentive to receive fixed in a long-term swap (e.g., ten-year) and pay a three-month LIBOR. Their strong demand for “receive fixed/pay float” swaps tend to cause swap spreads to tighten. Fourth, the level of rates impacts swap spreads. As rates decrease, mortgage refinances pick up and GSEs or MBS portfolio managers tend to receive fixed, tightening swap spreads. Conversely, a dramatic increase in rates can prompt them to pay fixed-in swaps and widen swap spreads. So we would expect swap spreads to tighten (widen) when rates decrease (increase) dramatically. The selloff in the Treasury market in the summer of 2003 is a case in point. Investors who had purchased Treasuries in anticipation of bond purchases by the Fed were quickly selling them when the Fed announced no intention of purchasing Treasuries. Both Treasury yields and swap spreads increased significantly within a few weeks. Fifth, the impact of the above factors is more pronounced in the short and intermediate tenors. In the longer-tenor swap market (i.e., 10–30 years) the key players are insurance companies and pension funds, which both have long-duration liabilities. To actively match the duration of their assets to that of liabilities, they have a strong demand for receiver swaps, which tend to make swap spreads narrower. The sustained negative level of the 30-year swap spread is attributable to a stronger demand from these institutions for payer swaps as a result of the higher volatility in the market during and after the financial crisis. Lastly, regulation changes may make spread trades costly and thus impact swap spreads. As mentioned above, swap spread arbitrage strategies involve a swap, a (long or short) position in a Treasury, and a repo (or reverse repo). Regulatory changes can lead to transaction costs of swap spread arbitrage. For example, the rules of Basel III Leverage Rate and U.S. Supplementary Leverage Ratio (SLR) aim to restrict leverage in the banking sector. Arguably, these rules led banks to reduce the size of repo businesses, which in turn could cheapen Treasuries and narrow swap spreads. Therefore, limits to arbitrage arising from regulatory changes may impact swap spreads across the spectrum of tenors.

18.7 PRICING To value an interest rate swap, one needs to compute the net present value (PV) of the fixed and floating legs. The value of the swap for the fixed rate receiver is thus

V (Swap) = PV (Fixed Leg) - PV (Floating Leg). (18.3)

422  Research handbook of financial markets

Pricing a fixed-floating interest rate swap when it is first executed amounts to determining the fixed coupon that makes the net present value of the trade zero. For swaps that have been in existence for some time, pricing these swaps is determining their mark-to-market value. It is straightforward to compute PV (Fixed Leg) since the fixed leg involves a stream of known cash flows. Denote the (annualized) fixed rate by s and the swap pays the fixed rate n 1 2 times a year at times T1 = < T2 = <  < TNn = N . Let Tii=0 , n n Nn



PV (Fixed Leg) =

åa ´ D(T ), (18.4) i

i

i =1

where D(Ti) denote the present value of a dollar to be received in time Ti and αi denotes the length of each calculation period in years calculated using the appropriate basis. Pricing the float leg seems to be more complicated because its future cash flows are unknown at the time of pricing. However, it turns out to be much simpler, in that on the effective date when a swap is executed, the present value of the floating leg, including the notional at maturity, is par, which also holds true on any reset date on which the floating rate is reset. In general, for a N-year swap that pays a floating rate m times a year, we now determine the 1 value of its floating leg at the effective date T0 = 0. In Example 1, N = 5 and m = 4. Let T1 = , ..., m TNm = N denote the dates on the floating rate payments are made. We first divide the N-year dura-1  1 < T2 = 2 < … < TNm = N . tion into Nm calculation periods, [Ti , Ti +1 ]iNm =0 , where 0 = T0 < T1 = m m At the beginning of the calculation period [Ti , Ti+1 ], the (annualized) floating rate r (Ti ) is deter1 mined and accrued for the next -year. At the end of the period Ti+1 , the amount r (Ti ) / m is m paid per unit notional. The present value of the cash flow from this period is given by

r (Ti ) / m ´ D(Ti +1 ) = D(Ti ) - D(Ti +1 ), (18.5)

D(Ti ) . 1 + r (Ti ) / m Applying the result in equation (5) to all cash flows of the floating leg per unit notional, which are r (T0 ) / m at time T1 , r (T1 ) / m at time T2 , ..., r (TNm -1 ) / m at time TNm , we obtain the value of the foloating leg per unit notional as:

where in deriving the above expression we have used the result D(Ti +1 ) =

Nm -1



PV (Float Leg) =

å(D(T ) - D(T i

i +1

) = D(T0 ) - D(TNm ). (18.6)

i =0

Intuitively, the cash flows of the floating leg can be replicated by ensuring that one owns $1 at the effective date T0, but owes $1 at the maturity date TNm . In this case, one can invest $1 at the prevailing floating rate r (Ti ) at the beginning of the calculation period [Ti , Ti+1 ] and then use the proceeds, 1 + r (Ti ) / m , to pay r (Ti ) / m to settle the floating obligation and keep $1 to invest in the next calculation period. At the maturity date TNm , after paying the interest r (TNm -1 ) / m one can use the $1 left to pay off the loan. The initial value of the replicating portfolio is also equal to 1 - D(TNm ).

Interest rate swaps 

423

A par-swap is a swap whose value today is zero. The fixed rate of this swap is called parswap rate. As shown above, for a swap with effective date T0 and maturity date T, the par-swap rate, S (T0 , T ), is given by

S (T0 , T ) =

D(T0 ) - D(T )

å

Nn i =1

ai ´ D(Ti )

. (18.7)

Note that the receiver in a swap at the effective date has effectively a long position in a fixedcoupon bond at par. For this reason, a swap with zero value is called a “par swap”. Rearranging equation (3) yields Nn



1=

åS(T , T ) ´ a ´ D(T ) + D(T ). (18.8) 0

i

i

i =1

The right-hand side of the above equation represents the value of a T-maturity coupon bond with coupon rate S (0, T ), which is shown to be equal to unity, implying that the bond is worth par. Therefore, when a swap’s fixed rate is the same as the market par swap rate, its value is zero, which is analogous to the fact that a bond is worth par if its coupon rate is the same as market yield. Put differently, par-swap rates are analogous to bond yield to maturities. A swap is spot-starting, if T0 = 0; or, otherwise, forward-starting or a forward swap. The par-swap rate of a forward swap is called the forward (par) swap rate. For spot-starting swaps, a graph of S (0, T ) versus maturity T is called the par-swap curve.

18.8 ACADEMIC LITERATURE ON INTEREST RATE SWAPS Several strands of the literature study interest rate swaps. The first strand of the literature contains theoretical studies on why firms use interest rate swaps. The second strand calibrates multi-factor term-structure models to understand the dynamics of swap spreads. 18.8.1 Why Do Firms Use Interest Rate Swaps? Several theories have been proposed for why firms use interest rate swaps. Titman (1992) develops an asymmetric information model and shows that lower-rated firms with favorable private information that its borrowing costs will be lower in the future have incentives to borrow short-term and swap floating for fixed rates in order to reduce interest rate expenses. Intuitively, such firms prefer to borrow short-term since they expect their borrowing costs to decrease in the future. However, borrowing short-term subjects them to interest rate risks and hence to higher costs of financial distress. Entering into a floating-for-fixed swap simultaneously can reduce interest rate risks and lower financial distress costs. Wall (1989) develops an agency cost theory for the use of swaps. Issuing long-term bonds entails agency costs due to underinvestment because of the incentive to bypass positive net present value projects (Myers, 1977), and due to asset substitution because of the incentive to shift to high-risk projects (Jensen & Meckling, 1976). Wall proposes that to avoid the agency

424  Research handbook of financial markets

costs associated with long-term debt, firms can instead issue short-term bonds and enter a swap as a fixed-rate payer. Bicksler and Chen (1986) propose a theory based on comparative advantages. The idea is that some firms (e.g., higher-rated firms) have a comparative advantage borrowing long-term and others (e.g., lower-rated firms) a comparative advantage borrowing short-term. The authors show that swap transactions allow higher-rated (or lower-rated) firms to borrow at more attractive rates than they would otherwise obtain from borrowing short-term (or long-term).13 Wall and Pringle (1989) find empirical evidence supporting the agency theory based on information contained in the footnotes of Annual Reports for the year of 1986. The Statement of Financial Accounting Standard (SFAS) numbers 105 and 107, published by the Financial Accounting Standards Board, require disclosure of notional and fair values of interest rate swap usage in financial statements, effective after 1992. The availability of firm-level swap usage makes it possible to empirically test the above theories. Saunders (1999) finds strong evidence for the information asymmetry theory of swap usage in Titman (1992) and some evidence for the agency cost theory in Wall (1989). The firm-level swap usage data are also used by some authors to study the usage of swaps by non-financial companies. Empirically, Li and Mao (2003) show that U.S. non-financial firms that use swaps are mostly fixed payers and that fixed-rate swap payers generally have lower credit ratings, higher leverage ratios, higher percentages of long-term floating-rate loans, and are more likely to use bank loans than floating-rate swap payers. Faulkender (2005) constructs the final interest rate exposure on the firm level by combining the initial exposure of newly issued debt with the use of swaps, and provides evidence that non-financial firms engage in timing the market or speculation: firms increas their floating-rate exposure as the yield curve steepens. Chernenko and Faulkender (2011) further use the cross-section and time-series variations in the final interest rate exposure of the firm’s debt to distinguish between a firm’s hedging and speculative activities, respectively. They find that hedging is concentrated among high-investment firms. Lastly, Jermann and Yue (2018) and Bretscher et al. (2018) develop quantitative dynamic models of investment, financing, and risk management and use the models to study firms’ swap choices. The model in Jermann and Yue (2018) reproduces the stylized facts that firms tend to be fixed-rate payers and time the market. Bretscher et al. (2018) emphasize the use of interest rate swaps to hedge interest rate uncertainty which is shown to have adverse effects on future economic activity. They show that risk management using swaps is effectively risky and thus interest rate uncertainty depresses financially constrained firms’ investments. 18.8.2 What Drives Swap Spreads? Earlier studies focus on default risk and liquidity components in swap spreads, and use different multi-factor term structure models to estimate these two components. As discussed earlier in Section 18.6, default risk arises from the possibility of default in the LIBOR market. On the other hand, swap spreads are considered as compensation for a liquidity-based convenience yield associated with Treasury securities (Grinblatt, 2001).

13 The first cross-currency swap between the World Bank and IBM in 1981 is another example of comparative advantages.

Interest rate swaps 

425

Duffie and Singleton (1997) and Liu et al. (2006) empirically estimate both default risk and liquidity components in swap spreads and find that both components are important sources of variation in swap spreads. Through a joint pricing model for Treasury securities, corporate bonds, and swap rates using six latent factors, Feldhütter and Lando (2008) further decompose the term structure of swap spreads into a convenience yield for holding Treasury securities, a credit spread arising from the LIBOR market, and a residual component. They find that the convenience yield is the largest contributing factor to the swap spread. Gupta and Subrahmanyam (2000) study the swap-futures differential, or the difference between swap prices and the prices of interest rate futures. They show that the convexity bias—caused by the non-linearity of payoffs—has only been taken into account since 1996 and explains the empirically observed swap-futures differential. Some recent studies examine other factors that drive swap rates or spreads. Collin-Dufresne and Solnik (2001) focus on the impact of the LIBOR panel selection for swap pricing. More recently, Hanson (2014) documents the relationship between MBS duration and swap spreads. Eom et  al. (2002) study the links between USD and JPY interest rate swaps and find that changes in the dollar interest rate swap spreads “Granger-cause” changes in the spreads of yen interest rate swaps for the long (ten-year) maturities. Johannes and Sundaresan (2007) develop a swap valuation theory under marking-to-market and costly collateral. They show, theoretically and empirically, that collateralization generates an increase in swap rates. 18.8.3 Negative Swap Spreads As shown in Figure 18.7, the 30-year swap spread turned negative shortly after the bankruptcy of Lehman Brothers at the height of the financial crisis. Since then, it continues to be negative, and fluctuates between –20 and –35 basis points in 2021. Swap spreads for other tenors (e.g., ten-year) also become negative from time to time. Negative swap spreads are a pricing anomaly because swap spreads usually widen during times of stress as noted earlier. The negative spreads are driven by a confluence of factors: namely, heightened demand to receive fixed-in swaps from pension funds, insurance companies, and corporate bond issuers (see Section 18.4.1), or increased Treasury supply. In addition, recent regulatory changes have reportedly resulted in decreased swap market liquidity, exacerbating the effects of those factors. For example, the significant decline in longer-dated Treasury yields in 2014 and early 2015 increased the value of liabilities of pension funds and insurance companies that typically resemble long-term and fixed-rate debt obligations. As a result, these institutions had a higher demand to receive fixed-in swaps to increase the duration of their assets to hedge. As another example, increased Treasury supply or the sale of Treasuries by foreign reserve managers would exert upward pressure on Treasury yields and narrow swap spreads. Jermann (2020) presents a swap pricing model to explain negative swap spreads. In the model, frictions for holding bonds limit arbitrage, implying that negative swap spreads cannot be fully arbitraged. Intuitively, dealers with smaller bond positions due to frictions are less exposed to long-term interest rate risk and thus require less compensation for the exposure to the fixed swap rate as well, which can lower the swap rate to the extent that swap spreads become negative. Boyarchenko et al. (2018) argue that regulatory changes such as leverage requirements reduced incentives for supervised institutions to enter into trades against negative swap spreads. Klingler and Sundaresan (2016) document a relation between pension funds duration hedging and negative swap spreads.

426  Research handbook of financial markets

18.8.4 Expectations of Monetary Policy The pricing of OIS swaps reflects market expectations regarding the likely path of a future policy rate.14 Similar to work in the literature on measuring U.S. monetary policy surprises based on fed feds futures and Eurodollar futures (see, e.g., Kuttner, 2001; Gürkaynak et al., 2005), a number of studies in the literature use changes in the OIS rates as a proxy for revisions to the expected path of future interest rate rates for the U.S., Canada, and Sweden (Woodford, 2012), for the U.K. (Joyce et al., 2011), and for the Euro area (Abbassi & Linzert, 2012; Altavilla et al., 2019). Specifically, Woodford (2012) provides empirical evidence for the effectiveness of forward guidance policy in the U.S., Canada, and Sweden. Joyce et al. (2011) investigate the impact of the Bank of England’s quantitative easing policy on U.K. asset prices. Abbassi and Linzert (2012) study the effectiveness of the ECB’s monetary policy by analyzing changes in Euribor rates as a response to changes in the OIS rates (which proxy for changes in the expected path of future interest rates). They find that the interest rate channel of monetary policy transition remained effective during the financial crisis, but less so compared to the pre-crisis period. Altavilla et al. (2019) construct an event study database (called “Euro Area Monetary Policy Event-Study Database” or “EA-MPD”)—which features price changes for a broad class of assets and various maturities, including OIS, sovereign yields, stock prices, and exchange rates—to measure and assess the ECB monetary policy.

18.9 CONCLUSIONS Interest rate swaps are among the most popular derivative contracts. Since the 1980s, both financial institutions and non-financial firms are using interest rate swaps for managing interest rate risk. Depending on the types of interest rate swaps, investors and traders engage in interest rate swap markets by exchanging streams of interest payments. The usage and pricing of interest rate swaps are an important topic of research that also touches the literature on corporate finance, risk management, term structure, and monetary policy research. Empirical and theoretical studies have examined why firms use swaps, and how firm characteristics explain the use of swaps. Still, the impact of interest rate swaps on macroeconomic activity including corporate default, investment and production decisions, and borrowing decisions has not seen in-depth study until recently. Researchers have examined whether the size of swap positions of non-financial firms and the negative co-movements between swap usage and the term spread can be accounted for by a quantitative model. The economic gains from swap usage are found to be small in such research work. Yet such models abstract from other considerations, such as more liquidity of swaps relative to corporate bonds which is also subject to agency problems for long-term debt. Empirical results based on panel data on corporate swap usage show that firms use swaps both to hedge to reduce exposure to interest rate uncertainty and to speculate to time the market. The negative spreads for longer-tenor swaps remain a pricing anomaly. The regulatory landscape for interest rate swaps and OTC derivatives has changed dramatically. The recent regulation and reforms after the 2008–2009 financial crisis aimed to mitigate 14 Note that OIS rates are broadly interpreted as the market’s expectation for the policy rate at given intervals in the future. The interpretation implicitly assumes zero term premium in OIS rates.

Interest rate swaps 

427

some of the risks in the OTC derivative markets. How regulatory changes have affected the dynamics of swap spreads opens a new avenue for future research.

REFERENCES Abbassi, P., & Linzert, T. (2012). The effectiveness of monetary policy in steering money market rates during the financial crisis. Journal of Macroeconomics, 34(4), 945–954. Altavilla, C., Brugnolini, L., Gürkaynak, R. S., Motto, R., & Ragusa, G. (2019). Measuring euro area monetary policy. Journal of Monetary Economics, 108, 162–179. Bicksler, J., & Chen, A. H. (1986). An economic analysis of interest rate swaps. Journal of Finance, 41(3), 645–655. Boyarchenko, N., Gupta, P., Steele, N., & Yen, J. (2018). Negative swap spreads. Federal Reserve Bank of New York Economic Policy Review, 24(2). Bretscher, L., Schmid, L., & Vedolin, A. (2018). Interest rate risk management in uncertain times. Review of Financial Studies, 31(8), 3019–3060. Chernenko, S., & Faulkender, M. (2011). The two sides of derivatives usage: Hedging and speculating with interest rate swaps. Journal of Financial and Quantitative Analysis, 46(6), 1727–1754. Collin-Dufresne, P., & Solnik, B. (2001). On the term structure of default premia in the swap and LIBOR markets. Journal of Finance, 56(3), 1095–1115. Duffie, D., & Singleton, K. J. (1997). An econometric model of the term structure of interest-rate swap yields. Journal of Finance, 52(4), 1287–1321. Eom, Y. H., Subrahmanyam, M. G., & Uno, J. (2002). Transmission of swap spreads and volatilities in the Japanese swap market. Journal of Fixed Income, 12(1), 6–28. Faulkender, M. (2005). Hedging or market timing? Selecting the interest rate exposure of corporate debt. Journal of Finance, 60(2), 931–962. Feldhütter, P., & Lando, D. (2008). Decomposing swap spreads. Journal of Financial Economics, 88(2), 375–405. Grinblatt, M. (2001). An analytic solution for interest-rate swap spreads. International Review of Finance, 2(3), 113–149. Gupta, A., & Subrahmanyam, M. G. (2000). An empirical examination of the convexity bias in the pricing of interest rate swaps. Journal of Financial Economics, 55(2), 239–279. Gürkaynak, R. S., Sack, B. P., & Swanson, E. T. (2005). Do actions speak louder than words? The response of asset prices to monetary policy actions and statements. International Journal of Central Banking, 1, 55–93. Hanson, S. G. (2014). Mortgage convexity. Journal of Financial Economics, 113(2), 270–299. Jermann, U. J. (2020). Negative swap spreads and limited arbitrage. Review of Financial Studies, 33(1), 212–238. Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305–360. Jermann, U. J. and Yue, V. Z. (2018). Interest rate swaps and corporate default. Journal of Economic Dynamics and Control, 88, 104–120. Johannes, M., & Sundaresan, S. (2007). The impact of collateralization on swap rates. Journal of Finance, 62(1), 383–410. Joyce, M. A. S., Lasaosa, A., Stevens, I., & Tong, M. (2011). The financial market impact of quantitative easing in the United Kingdom. International Journal of Central Banking, 7, 113–161. Klingler, S., & Sundaresan, S. (2016). An explanation of negative swap spreads: Demand for duration from underfunded pension plans [Working paper]. Kuttner, K. N. (2001). Monetary policy surprises and interest rates: Evidence from the fed funds futures market. Journal of Monetary Economics, 47(3), 523–544. Li, H., & Mao, C. X. (2003). Corporate use of interest rate swaps: Theory and evidence. Journal of Banking and Finance, 27(8), 1511–1538. Liu, J., Longstaff, F. A., & Mandell, R. E. (2006). The market price of risk in interest rate swaps: The roles of default and liquidity risks. Journal of Business, 79(5), 2337–2359.

428  Research handbook of financial markets

Myers, S. C. (1977). Determinants of corporate borrowing. Journal of Financial Economics, 5(2), 147–175. Saunders, K. T. (1999). The interest rate swap: Theory and evidence. Journal of Corporate Finance, 5(1), 55–78. Titman, S. (1992). Interest rate swaps and corporate financing choices. Journal of Finance, 47(4), 1503–1516. Wall, L. D. (1989). Interest rate swaps in an agency theoretic model with uncertain interest rates. Journal of Banking and Finance, 13(2), 261–270. Wall, L. D., & Pringle, J. J. (1989). Alternative explanations of interest rate swaps: A theoretical and empirical analysis. Financial Management, 18(2), 59–73. Woodford, M. (2012). Methods of policy accommodation at the interest-rate lower bound. In Paper Presented at the 2012 Jackson Hole Conference, Federal Reserve Bank of Kansas City.

19. Credit default swaps1 Antulio N. Bomfim

19.1 INTRODUCTION AND HISTORY1 Credit default swaps (CDS) are, by far, the most common type of credit derivative. They are financial instruments that allow the transfer of credit risk among market participants, potentially facilitating greater efficiency in the pricing and distribution of credit risk. In its most basic form, a CDS is a contract where a “protection buyer” agrees to make periodic payments (the CDS “spread” or premium) over a predetermined number of years (the maturity or term of the CDS) to a “protection seller” in exchange for a payment from the protection seller in the event of default by a “reference entity.” CDS premiums tend to be paid quarterly and are set as a percentage of the total amount of protection bought (the “notional amount” of the contract). CDS maturities generally range from one to ten years, with the five-year maturity being particularly common. Major dealers regularly disseminate quotes for credit default swaps. Along with risk spreads in the corporate bond market, CDS quotes are now commonly relied upon as indicators of investors’ perceptions of credit risk regarding individual borrowers and investors’ willingness to bear this risk. In addition, quotes from the CDS market are used as inputs in the pricing of other traditional credit products such as bank loans and corporate bonds, potentially helping promote greater integration of the various segments of the credit market. Broadly speaking, there are two types of CDS: Single-name CDS are contracts that name a single reference entity, such as a corporation or a sovereign borrower, whereas multi-name CDS reference more than one borrower. For instance, a multi-name contract could be written to cover defaults in a reference portfolio (such as in the case of a basket credit default swap) or, as has been increasingly common over the past couple of decades, be based on an index of commonly negotiated single-name CDS. The latter are generally called CDS indexes. Historically, credit default swaps have been mostly negotiated in a decentralized over-thecounter market. As a result, unlike exchanged-based markets, there are no readily available historical aggregate volume or notional amount statistics that one can draw upon. Instead, most discussions of the evolution of the market, its size, and activity tend to center on results of surveys of market participants and on anecdotal accounts by key market players. For instance, the Bank for International Settlements (BIS) conducts a semiannual survey of dealers’ derivatives

1 The analysis and conclusions set forth are those of the author and do not indicate concurrence by Northern Trust Asset Management or other members of its Global Fixed Income Team. This chapter is partly based on Bomfim (2016). I am grateful for insightful comments provided by Refet Gurkaynak and Jonathan Wright.

429

430  Research handbook of financial markets 100% 90% 80% 70%

Index product, 54%

Banks & securities firms, 22% Hedge funds, 15% Insurers, 6% Spec purp, 3%

60% 50% 40%

Other MN, 4%

30% 20%

Single name, 42% Nonfinancial firms, 16%

10% 0%

Other financial firms, 38%

CDS Type

End Users at Dealers

Sovereign, 14%

Non-rated, 21%

Financial firm, 20%

Non-invest. grade, 21%

Nonfinancial firm, 28% Sec prod, 4%

Investment grade, 58%

Multisector, 34% Ref Entity Type

Ref Entity Rating

Note:   “End Users” refers to non-dealer end users. Source:   Bank for International Settlements (2021) and author’s calculations. List of abbreviations used in Figure 19.1: Other MN: Other multi-name CDS (non-index product) Spec purp: Special-purpose vehicle, special-purpose corporation, or special-purpose entity Sec prod: Securitized products Ref Entity: Reference entity

Figure 19.1  Global CDS market at a glance, 2020:H2 activities around the world, and that survey has included information on credit default swaps since 2004. Other surveys, some of which have been discontinued, go a little further back in the past, but the credit default swap market is relatively young; it was virtually non-existent in the early 1990s.2 Based on data collected by the BIS (2021), notional amounts outstanding in the global credit default swap market totaled close to $8.5 trillion at the end of 2020, with a little more than half corresponding to index products—first bar in Figure 19.1. That said, the credit default swap market is still small relative to the overall global derivatives market: Notional amounts outstanding in credit default swaps accounted for approximately 1.5 percent of notional amounts in the global derivatives market in late 2020. The CDS market would likely be larger today, were it not for the 2008 global financial crisis (GFC). Underlying factors contributing to the decline in notional amounts outstanding since the crisis include, initially, heightened uncertainty about the regulatory environment, risk aversion, and increased use of netting/trade

2 See, for instance, the surveys run by the British Bankers Association (2006) and the International Swaps and Derivatives Association (2010).

Credit default swaps 

431

compression arrangements (Fitch Ratings, 2011, and Wooldridge, 2019) and, especially over the past several years, greater reliance on central clearing (Aldasoro & Ehlers, 2018).3

19.2 MAIN MARKET PARTICIPANTS Approximately 62 percent of the notional amounts outstanding of credit default swaps at the major derivatives dealers surveyed by the BIS in 2020 corresponded to contracts with central counterparties, which are exchange-like entities that act as a protection buyer to every participating seller and a protection seller to every participating buyer. It is quite remarkable that central counterparties are now such important participants in the global credit default swap market. These entities essentially came about as a response to widespread counterparty credit risk concerns that emerged during and after the GFC. Interdealer contracts are also common and include CDS entered as part of a dealer’s market-making activities as well as contracts where a dealer is an end user of credit derivatives. For instance, a dealer that is also a commercial bank might enter into a contract as a way to hedge part of the credit risk in its loan book. Excluding interdealer contracts and contracts between dealers and central counterparties, provides a glimpse into the main (non-dealer) end users of credit default swaps (or at least those non-dealer end users who entered contracts with dealers, as opposed to central counterparties). As shown in the second bar in Figure 19.1, the main non-dealer end users include smaller (though still large) banks or securities firms that do not participate in the BIS surveys, hedge funds, insurance firms, and other financial customers (a category that includes mutual funds) as well as non-financial customers. Hedge funds, in particular, have become increasingly important participants in the global credit default swap market over the past several years, both in relative and absolute terms. As a group, non-dealer end users of credit default swaps have tended to be net sellers of default protection in their transactions with dealers. The main net sellers have been banks and securities firms, insurers, and mutual funds. Special-purpose vehicles were important sellers before the GFC, but they have been less active in the CDS market since (see Figure 19.1). Historically, many of these have tended to view selling default protection as a way to increase the yield on their portfolios, though, especially for some insurance companies, this view backfired very badly during the GFC. For instance, risk exposures in protection-selling positions in credit default swaps were widely seen as having contributed to the near-collapse of American insurer AIG at the height of the 2008 crisis. The AIG experience was a stark reminder of the different risks associated with protection-selling positions in the CDS market. Reportedly, AIG was a significant seller of protection in AAA-rated tranches of securities backed by subprime residential mortgages and other assets. But the bulk of AIG’s CDS-related woes did not stem from actual defaults in those tranches. Instead, the company came under intense financial pressure in part as a result of recurring calls by its CDS counterparties for additional

3 Trade compression entails closing out ‘‘redundant contracts,’’ i.e., contracts where a market participant has offsetting exposures to a given reference entity (or entities). For instance, suppose that market participant A has entered into offsetting CDS contracts with counterparties B and C. In that case, provided B and C agree, A could close out its contracts with B and C, replacing these two contracts with a single contract between B and C directly. This would show up as a decline in notional amounts outstanding.

432  Research handbook of financial markets

collateral as the credit quality of the referenced tranches was deteriorating rapidly amid the broader financial crisis (Stulz, 2010).

19.3 TYPES OF REFERENCE ENTITIES Credit default swaps are written on both sovereign and non-sovereign entities. In practice, however, data from the BIS (2021) suggest that most contracts reference non-sovereign entities (approximately 86 percent of the notional amount outstanding in 2020, as shown in the third bar in Figure 19.1). Considering only CDS written on non-sovereigns, approximately one-third of the notional amounts outstanding at year-end 2020 corresponded to contracts that referenced only nonfinancial firms. Contracts that referenced only financial firms accounted for roughly 23 percent of the notional amounts outstanding in contracts referencing non-sovereign entities, with multi-name contracts that reference both financial and nonfinancial firms accounting for most of the remainder. The share of CDS written on sovereigns rose in the years after the GFC, from less than 4 percent of the notional amounts outstanding in the global CDS market in 2007 to around 14 percent in 2020. For single-name credit default swaps, the share of the sovereign sector went from around six percent in 2007 to close to one-third in 2020. Indeed, despite the GFC’s adverse impact on the market, notional amounts outstanding in the sovereign segment of the single-name CDS market managed to nearly double between 2007 and 2013. In contrast, notional amounts outstanding in the overall single-name CDS market shrunk by almost half over the same period (BIS, 2013). CDS contracts written on emerging-market government debt remain a dominant segment of the sovereign CDS market. Indeed, contracts that reference the sovereign debt of key emerging-market countries—such as Brazil, Mexico, and Turkey—are often cited as being among the most frequently negotiated in the sovereign sector of the global CDS market. Still, rising concern about debt-to-GDP ratios and fiscal deficits in the aftermath of the GFC—related in part to the economic downturn that followed the crisis and the high budgetary costs of repairing national financial systems—likely contributed to increased investor interest in CDS written on the sovereign debt of a much broader set of countries, including some advanced economies (Mahadevan, Naraparju, & Musfeldt, 2011). The sovereign debt crisis in Europe also played an important factor in a notable expansion of the sovereign CDS market to contracts that covered more than just the usual set of key emerging-market countries. Italy, for instance, was reported as one of the most frequently negotiated contracts during the height of the European crisis (Fitch Ratings, 2011). In fact, the CDS curve for sovereign Italian debt widened dramatically during the European sovereign debt crisis and—as shown in Figure 19.2—became inverted in late 2011, with the one-year CDS spread significantly wider than the five- and ten-year spreads. An inverted CDS curve typically signals heightened concern by market participants about the near-term prospects of the reference entity. Abstracting from market frictions that might have otherwise distorted the shape of the Italian CDS curve in late 2011, an inverted curve suggests that market participants believed at the time that the Italian government was facing formidable near-term challenges but that, if those immediate hurdles could be overcome, Italy would fare better down the road. In the end, Italy’s most pressing challenges were indeed overcome. Its sovereign CDS spreads have narrowed considerably since, and its sovereign CDS curve has returned to its normal upward-sloping pattern (dashed curve in Figure 19.2).

Credit default swaps  12/31/2021 (right axis)

640

Basis points

620 600 580 560 540 520

1

2

3

4 Maturity (years)

5

7

10

90 80 70 60 50 40 30 20 10 0

Basis points

11/15/2011 (left axis)

433

Source:   IHS Markit.

Figure 19.2  Sovereign CDS curve, Italy While the increased use of sovereign CDS written on the debt of some advanced economies in the aftermath of the 2008 and European sovereign debt crises is notable, the market for CDS contracts that reference the sovereign debt of countries with the highest credit ratings has mostly remained small and relatively illiquid. In general, for these countries, investors’ most pressing concerns are less about the risk of a “fundamental” default—where the government is simply unable to service or repay its debt—and more about the risk of a “technical” default triggered by a presumably rare and temporary disruption that could result in a delay of payments due (Boyarchenko & Shachar, 2020). This observation suggests that movements in the sovereign CDS spreads for the most creditworthy countries should primarily reflect shifting market views on technical default risk. A case in point is the United States, where technical default risks have emerged occasionally because of sometimes-contentious negotiations around the U.S. government’s statutory ceiling on its own federal debt. For instance, in 2011, amid intense debt-ceiling discussions in Washington, DC and a related downgrading of U.S. government debt by the credit-rating agency Standard and Poor’s, one-year CDS spreads on U.S. sovereign debt did widen notably, before narrowing back following the passing of the Budget Control Act of 2011, which included provisions to raise the debt limit and reduce future federal deficits (Austin & Miller, 2011). While U.S. debt ceiling negotiations appear to have left some imprint on U.S. sovereign CDS spreads in the few years that immediately followed the 2011 episode, a study by Boyarchenko and Sachar (2020) points to a less tight connection in recent years. That study concluded that investor interest in U.S. sovereign CDS has diminished over the past few years, with the market becoming less liquid and U.S. sovereign CDS spreads becoming a less reliable indicator of investors’ views on U.S. technical default risk. 19.3.1 Credit Quality of Reference Entities Credit default swaps are written on investment-grade (rated BBB or higher), speculative-grade (rated BB or below), as well as unrated debt instruments, but contracts written on investment-grade instruments correspond to the majority of the notional amounts outstanding in

434  Research handbook of financial markets

the global credit default swap market. For instance, as shown in the last bar in Figure 19.1, notional amounts outstanding in credit default swaps written on investment-grade entities accounted for roughly 58 percent of the notional amounts in global CDS market in 2020, with the remainder being about evenly split between contracts written on speculative-grade instruments and instruments that either had an unknown credit rating or no credit rating (BIS, 2021). The larger share of contracts written on investment-grade entities may seem counterintuitive. After all, one might have expected that protection buyers would be more interested in protecting themselves from their riskier debtors, rather than from highly-rated borrowers. In part, the preponderance of contracts written on investment-grade instruments traces its roots to the early days of the credit derivatives market, when bank capital requirements set by regulators tended not to differentiate between lending to investment- and speculative-grade borrowers. For instance, the terms of the 1988 Basel Accord called on financial regulators to require banks to hold the same amount of capital in reserve for monies lent to, say, an investment-grade, A-rated borrower as they would for a speculative-grade borrower. Nonetheless, lending to the former yielded a lower expected return, giving banks an incentive to free up the regulatory capital associated with loans to investment-grade borrowers by buying protection in the CDS market. But the treatment of bank regulatory capital and banks’ use of credit default swaps has changed dramatically since the 1988 Basel Accord, especially after the GFC. Indeed, even before the crisis, the terms of the Basel II Accord provided for greater discrimination among differently rated borrowers for the purposes of setting regulatory capital requirements. This and other regulatory changes partly explain why, as large as the market share of contracts written on investment-grade instruments is today, it is much smaller than it was in earlier times. For instance, a survey run by FitchRatings (2003) reported that contracts written on investment-grade instruments accounted for 92 percent of the credit derivatives market in 2003 (compared to about 58 percent in 2020). Issues unrelated to regulatory capital also help explain the rising share of contracts written on speculative-grade entities. Indeed, respondents to a survey run by the British Bankers Association (2002) reported that they expected credit derivative uses directly related to regulatory capital management to eventually play a less prominent role in the evolution of the market. In part, that view reflected the expectation that market participants would become more focused on using credit derivatives as tools for overall portfolio management. In addition, protection buyers’ attention was expected to continue to shift from regulatory to economic capital. As a result, some market participants expected (correctly, it turned out) that the market share of credit default swaps written on speculative-grade entities would increase.

19.4 BASIC MECHANICS OF A CREDIT DEFAULT SWAP As an illustration, consider a credit default swap where the two parties agree on a notional amount of $100 million, a reference entity, a term (the period covered by the contract), and a CDS spread of 40 basis points. In this case, the protection buyer will pay $100,000 every quarter to the protection seller. If no default by the reference entity occurs during the life of the CDS, the protection seller simply pockets the premium payments. Should a default event occur, however, the protection seller becomes liable for the difference between the face value of the debt obligations issued by the reference entity and their recovery value. For instance,

Credit default swaps 

435

assuming that the reference entities’ obligations are worth 20 cents on the dollar after default, the protection seller's liability to the protection buyer in the event of default would be $80 million. The contract is terminated once the seller fulfills its default-related obligation. In the event of default by the reference entity, a CDS can be settled physically or in cash, with the settlement choice determined upfront when entering the contract. In a physically settled swap, the protection buyer has the right to sell (deliver) a range of defaulted assets to the protection seller, receiving as payment the full face value of the assets. The types of deliverable assets are also prespecified in the contract. For instance, the contract may determine that any form of senior unsecured debt issued by the reference entity is a deliverable asset, and thus any bank loan or bond that meets this criterion is a deliverable asset. In a cash settled swap, the counterparties may agree to poll market participants to determine the recovery value of the defaulted assets, and the protection seller is liable for the difference between face and recovery values. The asset or types of assets that will be used in the poll are prespecified in the contract. In the earlier days of the CDS market, cash settlement was more common in Europe than in the United States, where, by far, the majority of CDS were physically settled. Since 2009, however, settlements in both the U.S. and Europe have increasingly happened through auctions involving the relevant defaulted instruments, where the auctions generally determine a common recovery rate for cash settling the contract. Auctions have become the standard settlement method in the global CDS market. The legal terms of credit default swap agreements are highly standardized and have been so for a long time. Indeed, the adoption of standardized documentation for CDS agreements has played an important role in the development and increasing liquidity of the CDS market. The use of master agreements sponsored by the International Swaps and Derivatives Association (ISDA) is a common market practice, significantly reducing setup and negotiation costs. The standard contract specifies all the obligations and rights of the parties as well as key definitions, such as which situations constitute a “credit event”—a default by the reference entity— and how a default can be verified. Regarding the former, CDS contracts generally allow for the following types of default events: ● ● ● ● ● ●

Bankruptcy Failure to pay Debt moratorium Debt repudiation Restructuring of debt Acceleration or default

Some of these events are more common in contracts involving certain types of reference names. For instance, moratorium and repudiation are typically covered in contracts referencing sovereign borrowers. In addition, CDS contracts have historically been negotiated both with and without restructuring included in the list of credit events. The maturity of a credit default swap does not have to match that of any particular debt instrument issued by the reference entity. The five-year maturity is especially common, although it is possible, and increasingly easier, to terminate or unwind a credit default swap before its maturity (and this is commonly done) in order to extract or monetize the market value of the position. Typically, unwinding a CDS position requires both parties in the contract to agree on the market value of the position. The party for whom the position has negative

436  Research handbook of financial markets

market value then compensates the other accordingly. Alternatively, a party may be able to close out its position by assigning it to a third party, but this generally requires the approval of both new counterparties.

19.5 MAIN USES As with any other derivative instrument, credit default swaps can be used to either avoid or take on risk, in this case credit risk. Protection buyers are credit risk avoiders; protection sellers are credit risk takers, and, of course, the market would not exist without either of them. At the most basic level, protection buyers use credit default swaps to buy default insurance, and protection sellers use CDS contracts as a source of income. In practice, however, market participants’ uses of credit default swaps go well beyond this simple insurance analogy. 19.5.1 Protection Buyers Credit default swaps allow banks and other holders of credit instruments to hedge anonymously their exposure to the credit risk associated with particular debtors. Thus, while the credit instruments may remain in the protection buyer’s balance sheet—which may be important particularly to banks for client-relationship reasons—the associated credit risk is transferred to the protection seller under the CDS contract. Some market participants, however, may want to buy protection through credit default swaps even if they have no exposure to the reference entity in question. Buying protection is akin to shorting the reference entity's debt—in that the market value of the protection buyer’s position would increase in the event of a subsequent deterioration in the credit quality of the reference entity. 19.5.2 Protection Sellers For protection sellers, the credit default swap market offers an opportunity to increase the yield on their portfolios or diversify their credit risk exposure. Here again there is a straightforward analogy to selling traditional insurance policies. For as long as the events covered in the contract do not occur, protection sellers receive a steady stream of payments that essentially amount to default insurance premiums. Of course, prospective protection sellers could, in principle, simply buy debt instruments issued by the desired reference entities directly in the marketplace in order to earn a potentially higher yield or to benefit from greater portfolio diversification. Furthermore, buying credit risk through outright long positions in, say, corporate bonds and loans, has the advantage of not exposing one to counterparty credit risk in the CDS contract. This begs the question of what motivates someone to sell protection in the CDS market. The largely unfunded nature of credit default swaps distinguishes them importantly from cash market instruments such as bonds and bank loans. For instance, credit default swaps allow an investor to obtain, say, exposure to $10 million worth of debt issued by the reference entity with essentially no initial capital outlay or at an upfront cost that is much smaller than the notional amount of protection provided by the contract. In contrast, that same exposure

Credit default swaps 

437

would have required a sizable initial cash outlay by the investor if the exposure were obtained in the form of a direct purchase of the reference entity’s bonds or loans. Of course, if the investor could finance the cost of the outright purchase in the repo market for bonds or loans issued by the reference entity, the transaction could also be characterized as one requiring little or no initial net cash outlay. In particular, the investor would buy the bond or loan for its market price of, say, Y and immediately repo it out, essentially using the bond or loan as collateral for a loan in the amount of Y. This repo-financed transaction may not work well in practice, however, because the repo market for corporate debt is typically not that liquid. Credit default swaps can also be used to create synthetic long positions in corporate debt: Instead of holding the credit risk assets outright, one can simply sell protection in a CDS contract. This use of credit default swaps highlights the fact that, in addition to their largely unfunded nature, credit default swaps might be particularly attractive to investors when outright positions in debt instruments issued by the reference entity are difficult to establish. Consider, for instance, a firm whose debt is closely held by a small number of investors. For an investor who wants to obtain credit risk exposure to that firm, but who cannot buy its debt instruments directly in the cash market, selling protection via a CDS contract becomes a potentially appealing alternative. Our financial intuition should tell us that, at least in theory, the income that the investor will receive under the CDS contract will be closely linked to the cash flow that it would have received if buying the reference entity’s debt directly. 19.5.3 CDS Spreads as Market Indicators We have thus far focused on the main uses of credit default swaps strictly from the standpoint of those who participate in the credit derivatives market. Not so obvious, but potentially very important, is the growing use by market participants and non-participants alike of pricing information from credit default swaps as indicators of market sentiment regarding specific reference entities and credit risk in general. Credit default swap indexes, in particular, have become common barometers of conditions in the overall credit market—please see Section 19.8. Indeed, some market observers have even suggested that prices in the credit default swap market have a tendency to incorporate information more quickly than prices in the corporate bond market given that, at times, it may be easier to enter into swap positions than to buy or sell certain corporate bonds and loans. Whether information truly is reflected first in the credit derivatives or cash markets remains a point of empirical debate, but the fact is that both investors and regulators have started to pay closer attention to signals sent out by the credit default swap.

19.6 VALUATION CONSIDERATIONS Valuing a CDS can be thought of as arriving at the value of the CDS spread that would result in no money changing hands at the inception of the CDS contract. This spread is commonly called the “par CDS spread.” We can use a simple, static replication approach to arrive at the par CDS spread. That approach tells us that if we can devise a portfolio made up of simple

438  Research handbook of financial markets

securities that replicates the cash flows and risk characteristics of the contract we want to price, the price of that contract is, in the absence of arbitrage opportunities, simply the price of setting up the replicating portfolio.4 Following Bomfim (2002), we will consider two stylized examples that provide some basic insights regarding the valuation of credit default swaps. Example 1: Let Rf denote the current coupon rate for a floating-rate note that is not subject to default risk, and Rf+S denote the coupon rate for an otherwise identical floating-rate note that is subject to default risk (S is the credit spread). Consider an investor who is offered the choice of one of two portfolios: ● ●

a long position in this risky floater combined with a short position in the riskless floater; a protection-selling position in a CDS written on the issuer of the risky floater.

We assume that both floaters have the same maturity, coupon dates, and face values ($1 each), and that they sell at par immediately after their coupon payment dates. To keep things even simpler, assume further that the recovery rate on the risky floater is zero and that default can only occur immediately after the coupon payment dates. In addition, we assume that the CDS position involves no upfront payments. What are the cash flows associated with each portfolio? For as long as the issuer of the risky floater does not default, the first portfolio yields S every period. As for the second portfolio, the CDS has a cash flow of Scds every period, where Scds is the CDS premium. In the event of default, the holder of the portfolio of floaters ends up with a short position in the risk-free floater, which translates into a liability of $1, given that the floater is valued at par on its coupon payment dates. The protection seller in the CDS is liable for the CDS payoff, which is also worth $1. Thus, when there is a default, both portfolios have the same payoff. Let’s pause here to make two key points: ●

●

With time-varying interest rates, the static replication argument outlined would generally fail if, instead of using floating-rate notes to replicate the CDS cash flows, we had used fixed-rate notes. This is so because a fixed-rate note is not generally valued at par after it is issued and thus the liability of the short seller in the event of default could well be different from $1. Neither portfolio required a cash outlay when they were set up: The proceeds of the short sale of the riskless floater were used to finance the purchase of the risky floater, and it costed nothing to enter into the CDS.

Given the same initial cost, the same payoffs in the event of default, and the same risk exposure of the CDS transaction and the portfolio of floaters, it must be the case that the CDS and the floater portfolio have the same cash flow in the absence of a default by the reference entity. 4 There are other technical conditions that the replicating portfolio must satisfy, such as the requirement that it must constitute a self-financing investment strategy, but we will just assume that all these conditions are satisfied here. Baxter and Rennie (2001) provide an intuitive discussion of this topic. For a more rigorous, but still accessible exposition of replicating strategies, see, e.g., Bjork (2009).

Credit default swaps 

439

This requires that Scds=S. Thus, under the conditions set out previously, the premium that should be specified in a CDS written on a given reference entity is the same as the risk spread associated with a par floater issued by that entity. Let’s now try to make our example a little more realistic. We have thus far ignored the fact that the first portfolio ultimately has to be funded on the balance sheet, whereas the CDS does not. This is where Example 2 comes in. Example 2: Using the same notation and assumptions of Example 1, consider the following two scenarios: ●

●

The investor finances the purchase of the risky floater, by repoing it out, paying the repo rate Rf+F. (Alternatively, we can think of Rf+F as the rate at which the investor can obtain financing for the portfolio.) Assuming no default by the issuer of the risky floater, the investor receives Rf+S every period and pays out Rf+F to its repo counterparty. In the event of default, the risky floater becomes worthless, and the investor ends up owing $1 to its repo counterparty. To sum up, the investor's cash flows are: S-F (no default) and -$1 (default). The investor sells $1 worth of protection in a CDS written on the issuer of the same risky floater considered in the previous scenario. The cash flows associated with such a CDS position are: Scds (no default) and -$1 (default).

Again, notice that neither strategy required an initial cash outlay and both have the same payoff in the event of default. Thus, in the absence of arbitrage opportunities and market frictions, it must be the case that they have the same payoff in the absence of default, i.e., the CDS premium, Scds, must be the equal to the difference between the risky floater spread, S, and the investor's funding cost, F:

Scds = S – F (19.1)

where Equation 19.1 differs from the result obtained from Example 1 because we are now explicitly taking into account the fact that the first portfolio has to be funded on the balance sheet of a leveraged investor whereas the CDS is an unfunded instrument. To bring the discussion of the given examples even closer to the real world, we should note the following: Although this approach for pricing a CDS relied on rates on par floaters issued by the reference entity, most corporate debt issued in the United States are fixed-rate liabilities. In practice, however, one can circumvent this problem by resorting to the asset swap market.5 In particular, the previous examples can be made more realistic as illustrations of how to obtain an (approximate) value for the CDS premium if we substitute the par floater spread,

5 Asset swaps are a common form of derivative contract written on fixed-rate debt instruments. The end result of an asset swap is to separate the credit and interest rate risks embedded in the fixed-rate instrument. Effectively, one of the parties in an asset swap transfers the interest rate risk in a fixedrate note or loan to the other party, retaining only the credit risk component. As such, asset swaps are mainly used to create positions that closely mimic the cash flows and risk exposure of floatingrate notes (Bomfim, 2016).

440  Research handbook of financial markets

S, with the par asset swap spread associated with the reference entity. This would lead to the following version of Equation 19.1: Scds = Sa (19.2)



where Sa is the asset swap spread associated with the reference entity. 19.6.1 CDS versus Cash Spreads in Practice In practice, CDS spreads can vary substantially from the corresponding reference entity’s asset swap or floater spread. For instance, when Scds > Sa, market participants say that there is a positive “CDS-cash basis” for that reference entity.

( CDS-cash basis ) = ( CDS spread ) - ( par asset swap spread ) (19.3)

CDS and asset swap spreads can diverge for a number of reasons, including market segmentation and idiosyncratic supply and demand factors. For example, in the earlier days of the CDS market, instances of substantial positive bias were sometimes associated with strong demand by convertible bond investors for buying default protection against the issuers of those bonds. Those investors were focusing primarily on the cheapness of convertible bonds’ embedded equity call options while using the CDS market to shed the credit risk associated with the bonds. Another factor that contributes to positive bias is the fact that participation in the CDS market is limited either by some investors’ lack of familiarity with credit derivatives or by regulatory restrictions and internal investment policies of certain institutional investors. In addition, for some reference entities, a CDS liquidity premium, reflecting the poorer liquidity of the CDS market relative to the cash (corporate bond) market for those entities, may also be a factor leading to positive bias. Other factors that could contribute to a nonzero CDS-cash basis include ● ● ● ● ● ●

Cheapest-to-deliver feature of CDS contracts (for physically settled contracts) Default-contingent exposure in asset swaps Accrued premiums in CDS contracts Funding risk in asset swaps Counterparty credit risk Liquidity risk differentials

19.6.2 Counterparty Credit Risk In principle, a significant consideration for purchasers of protection in the credit default swaps market is the credit quality of the protection seller. The protection seller may itself go bankrupt either before or at the same time as the reference entity. This is, of course, what is meant by counterparty credit risk. As will be noted, market participants commonly use credit-enhancement mechanisms—such as the posting of collateral—to mitigate the effects of counterparty credit risk in the pricing of CDS contracts. In the absence of these mechanisms,

Credit default swaps 

441

however, and other things being equal, the higher the credit quality of a given protection seller relative to other protection sellers, the more it can charge for the protection it provides. Regarding its credit default swap counterparty, the protection buyer is subject to two types of risk: Should the protection seller become insolvent before the reference entity, the protection buyer is exposed to “replacement risk” or the risk that the price of default insurance on the reference entity might have risen since the original default swap was negotiated. The protection buyer’s greatest loss, however, would occur when both the protection seller and the reference entity default at the same time, and hence the importance of assessing the extent of default correlation between the reference entity and the protection seller. 19.6.3 Collateralization and Netting As the credit default swap market has grown, so have market participants’ exposures to one another. An important step that market participants took early on to reduce counterparty credit risk is to require the posting of collateral against the net exposures resulting from CDS positions. Collateral, primarily cash, is commonly adjusted daily to reflect the net markedto-market value of the CDS positions between two counterparties (see Section 19.6.5 for a discussion of a simple marking-to-market approach). Typically, when both counterparties are dealers—or even for a non-dealer counterparty that is active in many sectors of the derivatives market—collateral arrangements tend to be more complex, in that they will cover net exposures on all over-the-counter derivative positions, not just CDS. Note that the focus is on “net” exposures between counterparties, and, indeed, this netting of counterparty credit risk exposures is an important feature of market functioning. Consider a simple example. AZZ Bank and XYZ Bank have a large number of credit default swaps between the two of them. AZZ’s total exposure to XYZ amounts to $100 billion, whereas XYZ’s exposure to AZZ is $90 billion. Netting means that the exposures of the two banks to one another are offset before any collateral is posted so that what matters in the end is the $10 billion net exposure of AZZ to XYZ. This would be the only amount against which any collateral would be calculated, and this would be the claim that AZZ would have on XYZ in the event of a default by XYZ. Taken together, early adoption of collateralization and netting practices, along with standardized documentation, have had the effect of helping overcome some of the “growing pains” of the credit default swap market. While standardized documentation has helped reduce legal risk and transaction costs, collateralization and netting have eased concerns about counterparty credit risk, especially as potential risk exposures through credit derivatives have grown. 19.6.4 CDS-Implied Credit Curves Thus far, our main input for determining the fair value of a CDS spread has been pricing information from other credit instruments, such as par asset swap spreads and par floater spreads. Certain reference entities, however, may not have marketable debt outstanding, or the market for their debt may be very illiquid and available quotes may be uninformative. An alternative approach to valuing credit default swaps that is especially useful when reliable market prices on related instruments are not available is the one based on credit risk

442  Research handbook of financial markets

models.6 Suppose we have a model that gives us the default probabilities associated with a given reference entity. Consider now the simple case of a one-year CDS with a $1 notional amount and a single premium payment, Scds, due at the end of the contract. Assume further that a default by the reference entity, if any, occurs only at the maturity date of the contract. (To keep things even simpler, assume no counterparty credit risk and no market frictions such as illiquidity or market segmentation.) Let’s start with a credit default swap that has zero market value at its inception, which means that the CDS spread is such that the value of the “protection leg”—defined as the present value of the expected payment made by the protection seller in the event of default by the reference entity—is equal to the value of the “premium leg”—defined as the present value of the premium payments made by the protection buyer. The current value of the protection leg is simply the present value of the premium:

PV ( premiums ) = PV ( Scds ) (19.4)

where PV(.) denotes the present value of the variable in parenthesis. To arrive at an expression for the protection leg, define was the probability that the reference entity will default in one year’s time. The protection seller will have to pay (1-X) with probability w and 0 otherwise, where X is the recovery rate associated with the defaulted instrument. Thus, we can write the present value of the protection leg as:

(

)

PV ( protection ) = PV w éë1 - X ùû (19.5)

If the CDS is to have zero market value at its inception, the present values in Equations 19.4 and 19.5 must be equal, and that will happen when

Scds = w éë1 - X ùû (19.6)

and we get the result that the cost of protection, Scds, is increasing in the probability of default and decreasing in the recovery rate associated with the reference entity. In particular, in the limiting case of no recovery, the CDS premium is equal to the probability of default. Thus, if we have a theoretical model that gives us the default probabilities associated with the reference entity, we can price a CDS written on that entity accordingly. The above results can be easily generalized, with a few modifications, for more realistic cases, such as multi-period credit default swaps. In particular, let Q(t,Tj) denote the risk-neutral probability, as of time t, that the reference entity will not default by a future time Tj, where Q(t,Tj) is conditional on all information available at time t and, naturally, on the issuer not having defaulted by time t. Q(t,Tj) is commonly called a risk-neutral survival probability of the reference entity. More generally, for any given issuer or reference entity, one can imagine an entire term structure of survival probabilities, which is one way of thinking of that entity’s credit curve. Consider now a CDS with a notional amount of $1, written at time t on this reference entity, and with premium payment dates at [T1, T2 …, Tn]. Let Sn be the corresponding annualized CDS premium. For simplicity, assume that, in the event of default, the protection seller will 6 See, for instance, Bomfim (2016) for a basic introduction to credit risk modeling.

Credit default swaps 

443

pay (1-X) at the premium payment date immediately following the default, where X, restricted to be between 0 and 1, is the reference entity’s recovery rate. As noted, we can think of a CDS as having two “legs:” The premium leg is made up of the periodic payments made by the protection buyer; the protection leg is the default-contingent payment made by the protection seller. Assuming that the occurrence of defaults is independent of the risk-free interest rate embedded in the prices of riskless bonds, the present discounted value of the premium leg can be written as:

PV ( premiums ) =

n

åZ ( t, T ) Q ( t, T ) d S (19.7) j

j

j n

j =1

where Z(t,Tj) is the time-t price of a riskless zero-coupon bond that matures at Tj with a face value of $1, and δj is the accrual factor for the jth premium payment (the number of days between the (j-1)th and jth premium payment dates divided by the number of days in the year, based on the appropriate day-count convention). Equation 19.7 shows that there are two elements to discounting future premiums. When computing the present value of a future payment, first, there is the time-value of money, captured by Z(t,Tj), and, second, one must take account of the fact that a future premium due, say, at Tj will only be received if the reference entity has not defaulted by then, and, conditional on all information available at time t, the risk-neutral probability of that happening is Q(t,Tj).7 The present value of the protection leg can be written in a similar way:

PV ( protection ) =

n

åZ ( t, T ) Prob éëT j

j -1

< t £ T j ùû éë1 - X ùû (19.8)

j =1

where τ is the time of default, and Prob[Tj-1 < τ ≤ Tj] denotes the probability, conditional on information available at time t, that the reference entity will default between Tj-1 and Tj. The intuition behind (19.8) is straightforward: One does not know whether and when a default will occur, but there is some probability Prob[Tj-1 < τ ≤ Tj] that the reference entity will default during the interval [Tj-1, Tj], in which case the protection seller would have to pay (1-X) at Tj, which is worth Z(t, Tj) (1-X) in today’s dollars. As a result, the present value of the protection leg of the CDS is the probability-weighted sum of all possible default scenarios. We can then write:

PV ( protection ) =

n

åZ ( t, T ) éëQ ( t, T j

j =1

j -1

) - Q ( t, T j )ùû

éë1 - X ùû (19.9)

7 Equation 19.7 can be modified to account for the premium accrued between the last premium payment date and the date of default—see, for instance, Bomfim (2016).

444  Research handbook of financial markets

Equations 19.7 and 19.9 allow us to write an expression for the par CDS spread:

ìï Sn = í ïî

üï Z ( t , T j ) éëQ ( t , T j -1 ) - Q ( t , T j ) ùû éë1 - X ùû ý ïþ j =1 n

å

ìï í ïî

üï Z ( t , T j ) Q ( t , T j ) d j ý (19.10) ïþ j =1 n

å

Our starting point for arriving at Equation 19.10 was model-based estimates of the default probabilities of the reference entity. But we can turn this process the other way around. In particular, suppose we see quotes for CDS spreads for a given reference entity for contracts with maturities ranging from T1 and Tn. In this case, and assuming we know the risk-free discount factors Z(t,Tj), for j = 1 to n, we can use (19.10) to “bootstrap” the CDS-implied risk-neutral survival probabilities of the reference entity. For, instance, starting with j=1, Q(t,T1) is easily computed as (1-X)/(1-X+ δ1 S1), which can then be used to compute Q(t,T2), and so on. 19.6.5 Marking to Market a CDS Position The market value of a CDS position varies over time, mostly reflecting changes in how the market assesses the creditworthiness of the reference entity. Marking a CDS position to market is the act of determining today’s value of a CDS agreement that was entered into at some time in the past. Consider, for instance, a credit default swap that was written one year ago, with an original maturity of five years. Suppose that the reference entity is a corporation that was perceived to have a substantially more favorable profit outlook one year ago (and hence higher survival probabilities) than today. As a result, we assume that the CDS premium that was written into the year-old contract is substantially lower than the one that would be written into a brand-new contract today. Under these circumstances, the protection seller in the year-old contract is collecting a premium that is well below the going market rate. This contract then has a negative market value to the seller of protection. The protection buyer, on the other hand, is holding a contract with positive market value as she is paying well below the premium she would pay if setting up a brand-new contract today. How can we value the year-old contract? One simple approach is to think of the problem of valuing the year-old contract as that of computing its “replacement cost.” To put it simply,

replacement cost = value of new contract - value of old contract (19.11)

Assuming that the newly minted four-year contract was negotiated such that it has no market value at its inception—that is, the new contract calls for a CDS spread, S4, that is equal to the reference entity’s par CDS spread. Equation 19.11 shows that the replacement cost of the old contract is simply the negative of its current market value. After some simple manipulations, and assuming (for simplicity) that the premiums are paid annually, we can write the value (to a protection buyer) of the year-old contract (which had an original maturity of five years) as: 4



Time-t value of year-old contract =

åZ ( t, T ) Q ( t, T ) d éëS - S j

j =1

j

j

4

5,old

ùû (19.12)

Credit default swaps 

445

where S5,old is the CDS spread written into the year-old contract, which now has a remaining maturity of four years.

19.7 EFFECTS OF THE GLOBAL FINANCIAL CRISIS ON CDS MARKET CONVENTIONS In many ways, the GFC left a lasting imprint on the CDS market. We have already mentioned the growing importance of central counterparties in the global CDS market in the post-crisis environment. We discuss next two additional post-crisis innovations: The widespread use of standardized coupons and upfront premiums (which also had the effect of facilitating the rise of central counterparties) and the growing popularity of CDS indexes. 19.7.1 Upfront Payments and Standardized Coupons The simple examples addressed thus far assumed that the CDS contract had no market value at its inception. As discussed, the CDS spread for such a contract is such that, at the time of the contract’s inception, the present value of the protection leg of the swap is equal to the present value of the premiums leg. Contracts with zero market value at inception were the norm in the years before the 2008 crisis. Back then, it was generally the case that only contracts written on reference entities that were viewed as potentially headed for trouble in the near term required upfront payment of a portion of the protection premiums. The upfront payments helped attract protection sellers to a market that could otherwise be severely one-sided, especially if the entities referenced in those contracts were perceived as being subject to “jump-to-default risk.”8 Since the 2008 crisis, however, contracts involving upfront payments have become standard, and not just because of jump-to-default considerations. Indeed, a major driver of the shift to upfront payments was a push for further standardization of contract terms and practices in the global CDS market. An important part of that push was the adoption of a fixed CDS spread and predetermined payment dates for all reference entities in a given class. For instance, in North America, standard CDS contracts written on high-grade corporates generally specify a fixed spread of 100 basis points, regardless of the characteristics of the reference entity. Contracts written on corporates in the high-yield sector generally specify a spread of 500 basis points. To see how upfront payments relate to coupon standardization, consider the case of a highyield corporate that is perceived to be sufficiently risky that its par CDS spread is significantly higher than the standard 500 basis points specified in the contract. In this case, a CDS contract with the standard premium would require an upfront payment from the protection buyer to the protection seller in order to compensate the latter for a contract-specified spread that is essentially too low. (The upfront payment would flow in the opposite direction if the par CDS spread were perceived as being lower than the standardized spread specified in the contract.) 8 The rationale for requiring upfront payment for contracts where jump-to-default risk was perceived as being high was (and remains) simple: Without an upfront payment, the protection seller might be called upon to cover a default event before it has had the opportunity to earn much of the protection premiums.

446  Research handbook of financial markets

19.7.2 Valuing Upfront Payments At its most basic level, valuing an upfront payment comes down to marking to market a CDS contract whose premium is set to a predetermined (standardized) value, instead of the reference entity’s par CDS spread. To make the point a little more concrete, let Sstd denote the standardized spread written into the contract. Using the same logic followed here to mark to market an old CDS position, we can write the market value (to the protection buyer) of this brand-new CDS contract with a standardized spread of Sstd and a par spread of Spar as n



Market Value to Protection Buyer =

åZ ( t, T ) Q ( t, T ) d éëSpar - Sstd ùû (19.13) j

j

j

j =1

Equation 19.13 says that the upfront payment from the protection buyer to the protection buyer is equal to the present value of the difference between two premium payment streams, one associated with the reference entity’s par CDS spread, Spar, the other with the standardized coupon written into the contract, Sstd. Again, if the standardized coupon happened to be above the par CDS spread, in which case the market value to the buyer would be negative, the upfront payment would flow from the protection seller to the protection buyer. 19.7.3 Standardized Coupons and Central Clearing Standardized coupons have facilitated the central clearing of CDS trades and made it easier (less costly) for market participants to close or unwind existing CDS positions. As such, they have enhanced the overall liquidity of the CDS market. For instance, from a dealer’s perspective, the unwinding of a CDS position by an end user generally results in transferring (assigning) that position to the dealer, with the dealer seeking to hedge the risks in its newly acquired position by entering into at least one offsetting position on the same reference entity with another counterparty. With standardized spreads, the dealer’s cost of hedging the position acquired from the end user is generally lower than with non-standardized spreads. In particular, there is no cash flow mismatch for the dealer between the old and new contracts, which helps limit the dealer’s exposure to risks associated with the timing of an eventual default by the reference entity. Ultimately, the lower cost of hedging for the dealer translates into a lower cost of unwinding the position for the end user, with the end result being enhanced market liquidity. Without standardized spreads, the hedging position sought by the dealer would typically involve a very different spread from the one in the contract acquired from the end user. For instance, if the position acquired from the end user involves a coupon that is substantially lower than the one received in the newly entered hedging position, the dealer would be exposed to the risk of a sudden default by the reference entity, which would cut short the stream of higher CDS payments embedded in the hedging position.

19.8 CDS INDEXES Credit default swap indexes are a relatively recent addition to the credit derivatives market. The earliest versions of CDS indexes were introduced in 2001. Since then, especially after the

Credit default swaps 

447

2008 financial crisis, index products have become an increasingly important segment of the credit derivatives market, reaching close to 42 percent of the global notional amount outstanding by mid-2013—Bank for International Settlements (2013). In its simplest form, a CDS index can be thought of as a portfolio of the most actively transacted single-name CDS contracts in each segment of the market. Indeed, for an equalweighted index, the index par spread should be sufficiently close to the average of the individual par spreads associated with the index’s constituents. In practice, however, liquidity and demand and supply factors can drive a wedge between the valuation of the index and that of its components. 19.8.1 The Mechanics of the CDX.NA.IG Index We will use the Markit North American CDX Investment Grade Index—CDX.NA.IG—to illustrate how CDS indexes generally work.9 CDX.NA.IG is one of the most actively traded CDS indexes. It is an equal-weighted index that tracks the 125 most liquid single-name CDS contracts transacted in North America, provided the credit rating of the constituent reference entities is at least “BBB-” or “Baa3.” Once the constituents of a given vintage (“series”) of the index are chosen, the composition of that series remains fixed, as long as there are no defaults among the constituents. Buying and selling the CDX.NA.IG index is akin to selling and buying a portfolio of debt instruments issued by the reference entities included in the index. Coupon conventions for an index generally are the same as for its constituent singlename contracts. The composition of CDX.NA.IG is refreshed every six months, at which point a new index “series’’ is generated. For instance, in late 2014, series 23 of CDX.NA.IG—or CDX. NA.IG.23—was the on-the-run index. Older series—CDX.NA.IG.22, CDX.NA.IG.21, etc.— continue to trade until their maturity date, but liquidity tends to migrate to the on-the-run series. Credit events related to a given index generally follow the definitions and procedures specified in the single-name CDS contracts that make up the index. For instance, for CDX.NA.IG, the relevant credit events are bankruptcy and failure to pay. We will use a simple numerical example to illustrate how a default by an index constituent is handled. Let’s start with a position in CDX.NA.IG with a total notional amount of $10 million. As noted, CDX.NA.IG is an equal-weight index, which means that each constituent corresponds to 1/125 of the position, or a notional amount of $80,000 per constituent in this case. A default by one of the index components would trigger the following changes: ● ● ●

The weight of the defaulted entity in the index falls to zero. The total notional amount of the position in the index falls by 1/125, to $9.92 million. The index weight of each remaining constituent increases to 1/124, reflecting the fact that each non-defaulted constituent still corresponds to a notional amount of $80,000.

9 Detailed information on the CDX.NA.IG index, as well as several other CDS indexes, can be obtained at the markit​.c​om website. Markit owns, manages, and compiles the CDX family of indexes, as well as other indexes. Markit (2014) provides a useful overview of its indexes.

448  Research handbook of financial markets

A default by an index constituent would also trigger a payment from the protection seller in the index position to the protection buyer, just as it would be the case in a single-name CDS. Suppose the recovery rate associated with the defaulted reference entity is 45 percent. Assuming that the position is cash settled, the protection seller would have to pay $44,000 to the protection buyer, or (1-.45) x 80,000. The newly re-weighted index continues to trade until its maturity date, though it would still be subject to further adjustments in the event of additional defaults among its constituents. 19.8.2 CDS Indexes as Market Indicators CDS indexes make it easier for investors and market observers to obtain exposure to, or simply track, a specific sector of the credit market. Focusing on the North American investment-grade sector, Figure 19.3 shows the evolution of the CDX.NA.IG index since 2004, highlighting a few key episodes of financial and macroeconomic stress. Not surprisingly, the GFC had a profound effect on CDS market pricing in the United States, resulting in a broad, steep, and persistent widening in spreads that was first particularly noticeable in early 2008, around the time of the collapse of the investment bank Bear Stearns and its immediate acquisition by J.P. Morgan Chase. It would not be until mid-2014 when the CDX.NA.IG index narrowed to levels close to its pre-GFC range, although the European sovereign debt crisis did play a significant role in slowing the U.S. investment-grade sector’s recovery from the GFC. In particular, by the spring of 2010, investors were becoming increasingly worried that problems associated with a worsening fiscal picture in Europe—mostly focused on Greece and a few other so-called peripheral EU countries at the time—could eventually hurt global growth prospects, with negative spillover effects for the U.S. corporate sector. The fiscal outlook for Europe darkened further in 2011, spreading beyond the periphery—as illustrated by the inversion of the Italian 300

Basis Points

250 200

May'10: concerns about Europe

Mar'08: Bear Stearns

Mar'20: Covid pandemic

Nov'11: Italian CDS curve inverts

150

Jan'16: Chinese market turmoil

100 50

Source:   IHS Markit.

Figure 19.3   CDX.NA.IG index

8-Jul-21

8-Jul-20

8-Jul-19

8-Jul-18

8-Jul-17

8-Jul-16

8-Jul-15

8-Jul-14

8-Jul-13

8-Jul-12

8-Jul-11

8-Jul-10

8-Jul-09

8-Jul-08

8-Jul-07

8-Jul-06

8-Jul-05

8-Jul-04

0

Credit default swaps 

449

sovereign CDS curve later that year (Figure 19.2)—and the cost of buying default protection in the U.S. corporate sector rose markedly in late 2011. Following the resolution of the European sovereign debt crisis, the U.S. corporate debt market enjoyed a period of relatively calm that was only briefly interrupted by a bout of financial market turmoil in China around early 2016 and rising global trade frictions in late 2018. It took the onset of the Covid-19 pandemic in early 2020 for CDS spreads in the U.S. investment-grade sector to really spike again, but that spike was short lived. One interesting aspect of the CDS market reaction to the economic effects of the pandemic was the different behavior of the costs of buying default protection in contracts that referenced consumer-facing businesses, such as airlines, relative to those that referenced producers of consumer goods. For instance, while CDS spreads in both the consumer-goods and consumer-services sectors widened markedly in the earliest stage of the pandemic, CDS spreads for consumer-facing reference entities took longer to recover, consistent with changes in consumer spending patterns at the time that resulted in a shift towards goods consumption and away from services consumption.

19.9 CONCLUDING REMARKS AND THE USE OF CDS IN THE LITERATURE The market for credit default swaps is still relatively young but has already become an integral part of the global credit market. Market participants actively use CDS to adjust their risk exposures as well as to infer evolving market views on credit risk. In addition, there is a growing academic and practitioner literature on credit default swaps. That literature has examined a range of CDS-specific issues, including the role of counterparty credit risk in the determination of par CDS spreads (Hull & White, 2001); how quickly new information is reflected in the CDS pricing, relative to other markets (Tolikas & Topaloglou, 2017), and the degree of integration between the cash and CDS markets in developing economies (Chan-Lau & Kim, 2008). These are just a few of the topics that are ripe for further investigation in the professional literature. Other paths for further research include the use of pricing data from the CDS market to examine broader questions, such as the cost of sovereign defaults (Hebert & Schreger, 2017) and the systemic and country-specific nature of credit risk in the U.S. and Europe (Ang & Longstaff, 2013).

REFERENCES Aldasoro, I., & Ehlers, T. (2018). The credit default swap market: What a difference a decade makes. BIS Quarterly Review, June 1, 14. Ang, A., & Longstaff, F. A. (2013). Systemic sovereign credit risk: Lessons from the U.S. and Europe. Journal of Monetary Economics, 60(5), 493–510. Austin, D., & Miller, R. (2011, August 15). Treasury securities and the U. S. Sovereign Credit default swap market. Congressional Research Service Report or Congress. Bank for International Settlements. (2013). OTC derivatives statistics at end-June 2013. Bank for International Settlements. (2021). OTC derivatives statistics at end-December 2020. Baxter, M., & Rennie, A. (2001). Financial calculus: An introduction to derivative pricing. Cambridge: Cambridge University Press. Bjork, T. (2009). Arbitrage theory in continuous time (3rd ed.). Oxford: Oxford University Press.

450  Research handbook of financial markets

Bomfim, A. N. (2002). Credit derivatives and their potential to synthesize riskless assets. Journal of Fixed Income, 12(3), 6–16. Bomfim, A. N. (2016). Understanding credit derivates and related instruments (2nd ed.). Cambridge, MA: Academic Press. Boyarchenko, N., & Shachar, O. (2020, January 6). The evolving market for U. S. Sovereign credit risk. Liberty street economics. Federal Reserve Bank of New York. British Bankers Association. (2006). Credit derivatives report 2006. Chan-Lau, J. A., & Kim, Y. S. (2008). Equity prices, bond spreads, and credit default swaps in emerging markets. ICFAI Journal of Derivatives Markets, 2, 7–26. Hebert, B., & Schreger, J. (2017). The costs of sovereign default: Evidence from Argentina. American Economic Review, 107(10), 3119–3145. Hull, J. C., & White, A. D. (2001). Valuing credit default swaps II: Modeling default correlations. Journal of Derivatives, 8(3), 12–21. International Swaps and Derivatives Association. (2010). ISDA market survey summaries 2010–1995. International Swaps and Derivatives Association. Mahadevan, S., Naraparaju, P., & Musfeldt, A. (2011). Sovereign CDS markets, a corporate perspective. Credit Derivatives Insights. Morgan Stanley. Markit. (2014). Markit credit indices: A primer. Markit. Ratings, F. (2011). Credit derivatives survey: Focus on sovereigns and regulatory issues [Special report]. FitchRatings, September. Stulz, R. M. (2010). Credit default swaps and the credit crisis. Journal of Economic Perspectives, 24(1), 73–92. Tolikas, K., & Topaloglu, N. (2017). Is default risk priced equally fast in the credit default swap and the stock markets? An empirical investigation. Journal of International Financial Markets, Institutions and Money, 51, 39–57. Wooldridge, P. (2019). FX and OTC derivatives markets through the lens of the triennial survey. BIS Quarterly Review, December 15, 19.

20. Foreign exchange swaps and cross-currency swaps Angelo Ranaldo1

20.1 INTRODUCTION1 In the first part of this chapter, we explain the main characteristics of foreign exchange (FX) swap and cross-currency swap contracts. We emphasize the importance of the valuation adjustment (XVA) approach and then map the FX swap market in terms of currencies, parties, maturities, and size. The second part is devoted to the institutional framework of the FX swap market, specifically its over-the-counter (OTC) nature, recent technological trends, and policy actions such as the regulatory reform and process to oust the Libor rate. We conclude this chapter by pointing to future research directions to better understand the pricing and market functioning of FX swaps. 20.1.1 Definition and Usages of Foreign Exchange Swaps An FX swap is an agreement for two reciprocal transfers of funds in two different currencies such that the transfer at maturity cancels out the initial exchange, which is usually conducted at spot.2 One party borrows one currency and simultaneously lends another currency to the same counterparty. The notional amounts in each currency are exchanged at the beginning and end of the life of the swap. The exchanged notional amounts at the beginning act as collateral. The difference of the repayment obligation is fixed on the day of writing the contract at the FX forward rate. At the time the contract is agreed, all transfers of funds are known. An FX swap can be seen as a low-risk, collateralized borrowing or lending facility for a foreign currency. It also can be viewed as combining a spot and forward FX transaction into one instrument. Let’s take the example illustrated in Figure 20.1: a European bank “A” has “x” euros in its books and needs USD for one year. We assume x equals ten million euros. On the spot market (S0), bank A can buy 1.18979 USD for one euro. At the same time, the ask one-year forward rate (F1) is 1.2032 USD for one euro.The US-based bank B agrees to be the counterparty. Today, at t0, bank A sends x to bank B, corresponding to 10,000,000 euros in the 1 Angelo Ranaldo: I am very thankful to Vincent Wolff for his excellent research assistance and the following colleagues for their comments: Benjamin Anderegg, Roman Baumann, Gino Cenedese, Pasquale Della Corte, Oliver Gloede, Peteris Kloks, Harald Hau, Edouard Mattille, Omar Misuraca, Femi Opeodu, Vlad Sushko, Olav Syrstad, Yannick Timmer, Vladimir Visipkov, Jonathan Wright, and Rafael Wyss. I acknowledge the financial support from the Swiss National Science Foundation (SNSF grant 182303). 2 “Spot” is an FX naming convention referring to the fact that whereas the transaction terms (and economic substance) of a spot trade are instantaneous, the delivery of the currency occurs two days later, which is a time frame referred to as spot. 451

452  Research handbook of financial markets

Figure 20.1  Payments of a EUR/USD FX swap example. Bank B sends x * S0 USD to bank A, corresponding to 11,897,900 USD. One year later at maturity, bank B sends x euros to bank A, 10,000,000 euros, and bank A sends F1 * x , 12,032,000 USD to bank B. A cross-currency swap resembles an FX swap but with two main differences. First, both parties of the cross-currency swap periodically exchange interest payments throughout the life of the contract. Second, the final rate at which the last payment is exchanged is the same FX spot rate as at the start of the contract. Hence, a cross-currency swap is an agreement for two reciprocal transfers of funds at initiation and maturity, along with the recurring exchanges of floating rates during the life of the contract. The notional amounts in each currency are usually exchanged at the beginning and end of the life of the swap. The repayment obligations of both parties and margins act as collateral. The rate of exchange of the floating payments during the contract term is specified when writing the contract. The specified rate is usually based on the Libor.3 For example, the USD Libor could be exchanged against the EUR Libor for floating payments throughout the life of the contract. At the moment of writing this chapter, the market is in the process of replacing the Libor with some risk-free rates (RFRs). Below, we will briefly discuss the possible consequences for cross-currency swaps. Now, we will take a look at example of a cross-currency swap, as illustrated in Figure 20.2: the Swiss bank C has x Swiss francs in its books and needs euros for three years. We assume x equals ten million CHF. On the spot market, bank C can buy 1.1068 CHF for 1 EUR (S0). The German bank D agrees to be the counterparty. Banks C and D also exchange SARON4 + 0.3% against the EONIA on a yearly basis. At the moment the swap goes live (t0), bank C sends x = 10 million CHF to bank D. Bank D sends x / S0 = 11,068,000 EUR to bank C. After one year (t1), bank C sends (SARON + 0.3%)*10 million CHF to bank D and D sends EONIA* x / S1 to C. The same payment is repeated in year two. After three years, the contract has matured, and the banks exchange the final interest payments due, as well as the notional amounts at the same FX spot rate as at the start of the contract (S0). That is, bank C sends (SARON + 0.3%)*10 million. CHF to D, and D sends EONIA* x / S3 to C. Bank C also sends x / S0 EUR to bank D, and bank D sends 10 million CHF to Bank C.

3 The Libor (London Inter-bank Offered Rate) is an average of the estimates provided by the leading banks for which interest rate they would be willing to borrow and lend from and to other banks. It is a nontraded interest rate computed for different maturities and currencies. 4 The SARON (Swiss Average Rate Overnight) represents the overnight interest rate of the secured funding market for the Swiss franc (CHF).

Foreign exchange swaps and cross-currency swaps  453

Figure 20.2  Payments of an EUR/USD cross-currency swap FX swap and cross-currency contracts are different in two other respects. First, the former is a pure FX instrument referencing only the spot and forward rate and is quoted in forward points (or swap points). By no-arbitrage conditions, the forward premium (discount) remains closely tied to the differential of the key interest rates in the two currencies, establishing an indirect link to money market rates. By contrast, cross-currency swaps pricing is directly determined by the interest rates referenced in the contract. This renders the cross-currency swap similar to an interest rate swap but with one collateral leg in a different currency, hence actually requiring an exchange of collateral at the prevailing spot rate. Second, these two contracts are used differently. For instance, FX swaps are predominantly used by banks for managing (short-term) funding liquidity across currencies or for hedging FX risk on a rolling basis. By contrast, by design, cross-currency swaps are term instruments suitable for longterm hedges such as hedging corporate bond issuances.5 In addition to making the market more complete, the utility of FX swaps is multifaceted and depends on the initial situation of the two parties. Here, we highlight three issues: first, comparative advantages emerge when two parties have different borrowing costs and/or creditworthiness. For example, a US company would like to borrow in euros, while a European bank needs USD. However, both benefit from better terms and cheaper rates if they borrow in their home currencies. Thus, a swap contract facilitates the exploitation of these benefits and leads to a Pareto improvement. Additional sources of comparative advantages come from different creditworthiness, tax treatments of company earnings related to the place of taxation, or regulatory requirements of FX risk exposures. Second, the balance sheet of international firms involves multicurrency assets and liabilities that can be modified by FX and cross-currency swaps, as shown in Figures 20.1 and 20.2. Imagine a multinational US company holds a USD-denominated bond with an interest rate of 2.5%. With a cross-currency swap, all incoming payments of the bond can be transformed into, for example, the corresponding payments of a euro bond with an interest of 3%. Depending on the regulatory framework, changing the nature of a balance sheet position from a foreign into a home currency or vice versa can free up or retain capital. Third, FX swap contracts allow agents to hedge currency mismatches in

5 For example, European corporate issuers of reverse Yankee bonds would typically also ask the underwriting bank to provide the hedge in the form of a cross-currency swap matching the maturity of their newly issued euro-denominated bond.

454  Research handbook of financial markets

cash flows. For instance, firms in emerging markets often receive cash flows that are denominated in a currency other than the entity’s functional currency. This represents a source of risk that can be hedged through swaps. 20.1.2 Further Considerations on Swap Pricing In introductory textbooks, FX swap pricing begins as either the difference of two bonds or as a portfolio of forward contracts.6 In an idealized setting, it is assumed that no “frictions,” such as credit risk and funding spreads, exist and that the forward exchange rates are realized. In such a frictionless environment, the covered interest rate parity (CIP) condition is satisfied; that is, converting the amount borrowed in one currency to lend it in another currency over a given period while hedging the exchange rate risk makes no profit. The Global Financial Crisis of 2008 has proven that these assumptions, as well as the CIP condition, can be violated, as will be discussed later. Nowadays, swap pricing takes into account market frictions and other practical considerations such as funding costs and regulatory issues related to margins, credit risk, and capital costs. Banks actually apply valuation adjustments to address theses issues, a practice termed XVA (“X-value adjustments”). XVA can be understood as applying some adjustments to a base value, which is the theoretical market price of a perfectly collateralized FX swap in such an idealized setting. XVA adds or subtracts from this base value margins; this depends on several factors, such as which side of the contract the counterparty is on, the underlying risks, and the shadow costs of regulation. The main types of XVA are the credit (CVA), debit (DVA), funding (FVA), margin (MVA), and capital (KVA) valuation adjustments. Because the XVA quantification is computationally intensive and depends on what kind of model and inputs are used, the XVA estimation for the same trade can significantly differ across financial firms.7 Although an in-depth discussion of XVA swap pricing is outside the scope of this chapter,8 we illustrate the main idea behind the XVA adjustment for counterparty risk, that is, CVA and DVA, which are two sides of the same coin. Derivative contracts are always a zero-sum game, so whether a party deals with CVA or DVA depends on which side of an eventual default it is on. If the counterparty default occurs and the swap has a positive (negative) value to the company and a negative (positive) value to the counterparty, the company will be an unsecured creditor (debtor) in the outstanding amount. The adjustment is then called a CVA (DVA). A pricing adjustment that occurs because of counterparty default is the base value of a swap minus the probability-weighted value if the counterparty defaults, plus the probabilityweighted value if the adjusting company defaults. For this, it is assumed that a base value of the swap (VND ) exists. The expected cost if the counterparty defaults (CVA) depends on the intervals the contract is active N, the present value of the expected loss given default ELGD in period i, and the probability of default p in period i. N



CVA =

åp (ELGD ) (20.1) i

i

i =1

6 See Hull (2017) for a detailed discussion of classical pricing methods. 7 Although large dealers have specific desks computing XVA exposures, it is more difficult to compute XVA for small banks that might ignore them or rely on optimization vendors. 8 See Green (2015) and Gregory (2015) for a comprehensive discussion of XVA.

Foreign exchange swaps and cross-currency swaps  455

However, the value adjusting company itself, not just the counterparty, may (also) default. This may lead to a loss for the counterparty but a gain to the company itself. Similar to the CVA, the DVA is as follows: N



DVA =

åp

H i

( ELGDiH ) (20.2)

i =1

where piH reflects the probability of default in period i of the company itself and ELGDiH the expected gain given this default. Taking both value adjustments into account, the value of the portfolio (PF) of the swap contracts becomes the following:

PF = VND - CVA + DVA (20.3)

Depending on the ELGD, CVA and DVA can be positive or negative. This implies that the XVA adjustment can lead to a price greater or smaller than the base value. It is important to emphasize that XVAs are usually calculated for the entire portfolio and that only the net exposure between two parties matters. In general, the XVA approach tends to generate heterogeneity in (FX) derivative pricing, especially given the fact that the price resulting from its application varies according to the characteristics of the contract, counterparties, and regulatory framework (as will be discussed later). Therefore, a CVA (FVA) can play a more pronounced role for long- (short-)term contracts such as FX cross-currency swaps (FX swaps), while more tightly regulated financial firms such as global systemically important banks (G-SIBs) are more receptive to regulatory issues such as a KVA.​ 20.1.3  Mapping the FX Swap Market The FX market is the largest financial market in the world. However, it is quite opaque and decentralized; this is one of the main reasons why the market is not easy to accurately map. The BIS central bank survey is the most comprehensive source of global FX spot and OTC FX derivatives trading activity. Starting from 1992, every three years, the BIS has taken a snapshot of the activity taking place only in the month of April.9 Figure 20.4 shows the estimated average daily trading volume of the FX swap and spot market, which was around three and two trillion US dollars (USD) in 2019, respectively. Additionally, the continuous linked settlement system (CLS, which will be discussed later) provides aggregate high-frequency data of FX prices and volumes. Although CLS data are only available from 2012 onwards, CLS data are accessible at very high frequencies, even at intraday intervals. Figure 20.4 shows that CLS settles about one third of the total FX activity, suggesting that it is representative of global

9 April is chosen because, historically, it tends to be one of the calmer months in the financial markets because it does not fall on a fiscal quarter or year end and is not known for historical periods of global financial turbulence. FX dealer trading volume for the month are then converted into a daily average in the published results.

456  Research handbook of financial markets

Note:  Figure 20.3 shows the average daily FX swap and spot volume in USD. The solid line represents the volume reported by the Triennial Central Bank Survey of Foreign Exchange and Over-the-counter Derivatives Markets in 2019 (BIS, 2019). The dashed line represents the volume settled by CLS. Sources:   BIS and CLS Group.

Figure 20.3  FX swaps: huge and growing market FX trading. It also shows that although the spot market has flattened out since 2013, the swap market has continuously grown. To gain more insights into the key features of the FX swap market, the pie charts in Figure 20.4 show the average daily FX swap volume by currency, as provided by BIS and CLS. Three considerations stand out: first, for the major currencies, both BIS and CLS data sets exhibit the same order and same size of shares, providing further support to the representativeness of CLS data. Second, the FX market is dominated by the USD, followed by the euro. Almost half of all FX trades (by volume) include the USD, and almost 20% include the euro. Third, the major difference in the data sets is that the Russian ruble (RUB), the Chinese renminbi (CNY), and some other emerging market currencies are missing from the CLS figures because they are not part of the CLS system.10 The time series in Figure 20.5 shows the smoothed daily FX swap volume settled by CLS and then converted into USD. The total daily volume is gradually increased without remarkable fluctuations, even in times of stress, from a level of around 800 billion USD in 2016 to levels of around 1,000 billion USD in 2020, implying that the yearly volume corresponds to almost three times the 2019 global GDP. The most traded currency pair is EUR–USD, with around 40% of volume, followed by USD–JPY, with around 18%.

10 The CLS settlement settles 18 currencies, and to carry out these operations, CLS has accounts with each of those 18 central banks.

Foreign exchange swaps and cross-currency swaps  457 BIS

CLS 2% 4% 4%

8%

8%

47%

10% 17%

USD EUR JPY GBP AUD CHF CAD Other

14% 3% 2% 3% 45%

7% 8% 18%

USD EUR JPY GBP AUD CHF CAD Other

Figure 20.4 shows the relative weights of different currencies. The left pie chart shows the average daily Note:   2019 FX swap volume as settled by CLS. The right pie chart shows the average daily FX swap volume reported by the Triennial Central Bank Survey of Foreign Exchange and Over-the-counter Derivatives Markets in 2019 (BIS, 2019). Sources:   CLS Group and BIS.

Figure 20.4  Daily FX swap volume provided by CLS and BIS The l.h.s. pie chart in Figure 20.6 shows the total FX swap volume by the two trading parties involved in a swap contract. CLS data allow us to differentiate between banks, corporates, funds, and nonbank financial institutions. The main component (93%) is driven by bank-to-bank transactions. This includes clients’ order placement funneled through the bank, as well as banks’ order placements. Another 6% are transactions between funds and banks. The remaining 1% are transactions between banks and nonbank financial institutions or corporations (a bank must virtually always be one of the counterparties in the CLS network). The r.h.s. pie chart in Figure 20.6 shows the total FX swap volume by its maturity. Although less short-term oriented than the repurchase agreements (repo) market, it is clear that the FX swaps market concentrates on short- and medium-term maturities because overnight swaps (labeled “0 days”) account for almost half of total volume and around 80% of the volume is within a one-month maturity.11 CLS data allow us to uncover two temporal patterns of FX swaps: how it evolves during the day and through the quarter. First, the intraday time series in Figure 20.7 documents the hourly average intraday FX swap volume for EUR–USD and USD–JPY. The area shaded with upward-sloping stripes reflects the so-called London trading hours (from 7 a.m. to 5 p.m. local time) and the downward-sloping stripes area the Tokyo hours, which are eight hours ahead of London. The average USD–JPY swap volume is above the EUR–USD swap volume for the Tokyo hours (0–6 a.m.), but both peak in the overlapping opening hours in London (7–9 a.m.), which is part of the CLS settlement window. These patterns are consistent with the common perception that FX markets are active 24 hours a day but also that dealers significantly reduce their inventory exposures outside of the most liquid trading hours.

11 The tenor labeled “0 days” mainly refers to swaps with the near leg of tomorrow and a far leg of after tomorrow (Tomorrow-Next). Swaps with “one–three-day” maturities are contracts having a near leg at the spot date and settlement date one–three days after that.

458  Research handbook of financial markets

Note:  Figure 20.5 is constructed as the one-month moving average of the CLS FX swap volume converted into USD. The exchange rates are volume-weighted, as settled by CLS. Source:   CLS Group.

Figure 20.5  Daily FX swap volume settled by CLS Second, the time series in Figure 20.8 is centered around the International Monetary Market (IMM) dates, which are the third Wednesday of March, June, September, and December and, therefore, are located closely to the quarter end.12 The IMM dates are of particular importance for FX markets because many FX swaps and other FX derivative contracts expire. In the days leading up to the maturity date, FX derivatives are frequently rolled to maintain the position. Figure 20.8 shows the change of CLS FX swap volumes in billion USD a month before and after the IMM date. The dark-grey shaded bar represents the IMM day and the light-grey shaded bar the quarter-end days. FX swap volumes show strong seasonalities in that they sharply decline on the IMM date. End-of-quarter effects are more nuanced; Kloks, Mattille, and Ranaldo (2023) show that once year-ends are controlled for, swap volumes actually increase at the quarter-end, in particular for short-term tenors. We will return to this point when we discuss recent research analyzing quarter-end periods.13

20.2 INSTITUTIONAL FRAMEWORK We now discuss three aspects of the FX swaps market and its institutional framework: the OTC setting, recent technological changes, and the policy context. 12 Because of weekends, holidays, and alternating month lengths, the days between IMM date and quarter end vary. 13  Figure 20.8 highlights the seasonality for the EUR–USD and USD–JPY, but the pattern holds for all prominent currency pairs.

Foreign exchange swaps and cross-currency swaps  459 Maturities

Party to counterparty 1%

6%

5% 13%

45%

21% 93%

Bank to Bank Other

6%

Bank to Fund

0 days 8 - 35 days

10%

1 - 3 days 36 - 95 days

4 - 7 days >= 96 days

Note:  Figure 20.6 shows on the l.h.s. left pie chart the total FX swap volume from 2016–2021 grouped by the respective trading parties. The r.h.s. pie chart shows the volume by its maturity. Only around 1% of the volume has a maturity longer than one year. Source:   CLS Group.

Figure 20.6  Weights of FX swap volume grouped by parties and maturities

CLS FX swap volume in billion USD

0.5 0.4 0.3 0.2 0.1 0.0 0

2

4

EURUSD

6

8

10

JPYUSD

12

14

16

London hours

18

20

22

24

Tokyo hours

Note:  Figure 20.7 shows the hourly average intraday FX swap volume for the euro and the Japanese yen against the US dollar from May 2019 until April 2020. The x-axis denotes London time, and the striped areas represent the London and Tokyo trading hours (see legend). Source:   CLS Group.

Figure 20.7  Intraday FX swap volume

460  Research handbook of financial markets

Change of CLS FX swap volume in billion USD

250

EURUSD JPYUSD IMM day

200 150

Quarter-end

100 50 0 –50 –100 –150 –200

–30

–20

–10

0

10

20

30

Note:  Figure 20.8 shows the FX CLS swap volume for EUR–USD and USD–JPY from 2016 to 2021. The x-axis represents the days before and after the IMM date, that is, the third Wednesdays of March, June, September, and December when many FX swaps and other FX derivative contracts expire. It is calibrated in a way that day zero is the IMM day. The IMM day and the quarter-end days are represented by the dark grey and light grey bars respectively. Source:   CLS Group.

Figure 20.8  FX swap volume around the IMM date 20.2.1 OTC Market The FX swap market is an over-the-counter (OTC) market in which each transaction is executed between two parties away from regulated exchanges. The ISDA Master Agreement is the standard document used to govern FX swap transactions. However, the party and counterparty bilaterally bargain, eventually agreeing on a specific contract that can be customized in various aspects, including price, notional amounts in either currency, and date of the near and far legs. Given this OTC setting, the FX swap market is fragmented. Apart from the limit order book platform Refinitiv FX Trading, there is no centralized exchange facilitating a uniform price formation process. In this way, FX spot and swap markets differ because the interdealer segment of the former relies on few central electronic limit order book platforms, such as the Electronic Broking Services (EBS), Refinitiv FX Matching, or Cboe FX ECN. The network of FX swap trading is a two-tier market that encompasses different types of market participants, among which dealers have a central role. In the outer tier, dealers act as market makers and liquidity providers to their customers, who are very different from each other: banks, large multinational corporations, hedge funds, pension funds, insurance companies, mutual funds, other institutional investors, retail clients, and central banks. It should be stressed that this dealership is concentrated on about 50 financial institutions, of which a

Foreign exchange swaps and cross-currency swaps  461

dozen of the largest global banks represents the lion’s share of the FX swap market.14 In the inner tier, dealers trade among themselves. In addition to supporting the price formation process, the interdealer segment facilitates the adjustment of inventory imbalances and hedging positions. A clear separating line between the interdealer and the dealer–customer market has blurred somewhat for at least two reasons: first, the emergence of electronic trading venues and, second, the attempt to penetrate brokerage and market-making services by some financial institutions that are not traditional dealers. Nevertheless, the FX swap market remains a global OTC, two-tier, dealer-centric network dealing with a highly diverse client base. 20.2.2 Technological Changes Since the first swap contracts were written in the early 1980s, many technological and institutional improvements have occurred. Here, we discuss just two that have made an impact in the past decade: electronification (i.e., the advent of electronic and automated trading) and settlement issues. Electronic facilities in the FX swap market have become more popular in recent years. Although it cannot be compared with the degree of electronification achieved in the spot market, screen-based trading is spreading at least for what concerns the entry-level e-trading.15 In the interdealer segment, the main trend over the last two decades has been to move from a pure voice trading system in which a dealer bilaterally meets the request for quote (RFQ) of its customers to a broker-based market. More specifically, brokerage firms collect multiple dealers’ interests and orders for several currency pairs and tenors to then be published on their electronic pages, such as the Tradition, ICAP, or GFI. In the past, there have been attempts by brokers to strengthen a more centralized interdealer segment, such as by coordinating regular auctions where dealers contribute to the liquidity provision for at least some relevant tenors, which is similar to a “dark pool” concept. However, these attempts never took off.16 Recent evidence indicates that the interdealer segment is approximately evenly split into voice and electronic trading, with the latter being fragmented across many different trading venues (Schrimpf and Sushko, 2019). In the dealer-to-customer segment, electronification has been more pronounced. In recent years, electronic trading has caught up with the swaps and forwards market before moving on to non-deliverable forwards.17 Trading has become increasingly electronic on the

14 The biggest players include (custodian) banks such as JP Morgan, Deutsche Bank, UBS, XTX Markets, Citi, HSBC, Jump Trading, Goldman Sachs, State Street, and Bank of America (Euromoney, 2020). 15 A recent survey indicates that between 60% and 95% of FX swap trading performed by the surveyed bank flows now happen as electronic trading (Risk​.ne​t, 2021b). 16 Despite the pre-trade anonymity, one of the deterrents has apparently been the post-trade transparency, that is, the need to disclose the identity of traders, revealing sensitive information about one’s interests and trading positions. 17 A survey ran in 2021 by fx​-markets​.​com indicates that electronic trading in OTC FX options is between 66% and 90% but only between 10% and 40% of the total traded volume.

462  Research handbook of financial markets

following types of venues: electronic communication network (ECN), bank based, or platforms.18 Regarding the trading platforms, a single-dealer platform (SDP) is a gateway by which traders are granted access to quotes from one dealer, whereas a multiple-dealer platform (MDP) grants access to those posted by various traders. Many trading platforms feature tools and services based on the latest technologies, such as artificial intelligence (AI) and algorithmic trading.19 SDPs are predominant in the FX swap market, even if MDPs enable clients to solicit competitive quotes from multiple dealers simultaneously by indicating the desired currency pair, tenor, amount, and trade direction (if possible). Of course, this creates competitive pressures on dealers, who can respond either with a static quote or with a quote stream that updates in real-time as market conditions evolve. These trading facilities are not anonymous because dealers can observe the client’s identity and possibly tailor their quotes accordingly. Furthermore, MDPs facilitate the best execution and regulatory requirements such as MiFID II.20 The natural question that arises is as follows: Why has the FX swap market not taken advantage of new technologies supporting electronification such as AI and data science, which have improved significantly in recent years? There are at least three structural impediments that keep the FX swap market it is, which is predominantly arranged bilaterally via RFQ, discouraging the substitution of SDPs with MDPs. First, despite the progress and attempts mentioned, there is no central pool of interdealer liquidity and no widely accepted market “mid” (or reference price) leading to an efficient price discovery process and facilitating execution algorithms. Second, there are multiple pricing factors that structurally make FX swap contracts much less prone to standardization than FX spot contracts, thereby hindering automatization. These factors often require customized services and drive price dispersion, as well as XVA adjustments, as discussed above. Even if de facto an FX swap is a collateralized instrument, one important factor is the counterparty risk embedded in forwards and swaps that increases with maturity, which is generally longer for FX swaps than the day or two of settlement risk in a spot trade (Figure 20.6). Third, the FX swap market hinges on the dominant role of dealers, who obviously have an interest in maintaining their network centrality, bargaining power, and sources of revenue. Rather than facing more competition between them and from other (nondealer) firms as with what occurred in the FX spot market, traditional FX dealers have an incentive to maintain the status quo, including RFQ and their SDPs. The natural reaction of the dealer’s customers is to maintain relationships with more than one dealer to obtain the best execution and competitive price.21 On the other hand, dealers undertake valuable tasks 18 For FX swaps, ECN includes 24 Exchange, 360T, BGC, Bloomberg, CME/NEX Group, Currenex, FXSpotStream, ICAP, Integral, MOEX, and Refinitiv. Bank venues are addressed to active clients looking for faster(algorithm) and include BAML, Barclays, BNP, Citi, Commerzbank, Credit Suisse, Deutsche Bank, HSBC, JP Morgan, Lloyds, Morgan Stanley, Nordea, RBC, SEB, Société Générale, Standard Chartered, State Street, and UBS. FlexTrade is a trading platform. Additional information is available on www​.marketfactory​.com​/venues/ 19 An important advantage has been to customize orders and make them contingent on evolving market conditions, for example, by changing the order size and tenor depending on the movement of the underlying FX spot and/or the movements in the money market rates. 20 Although challenging, platforms such as 360T are trying to provide a centralized and transparent price and automated forms of limit checking. 21 This is especially important for banks that cannot obtain a sufficiently large credit line from a single dealer, which demonstrates to their clients that they are protecting their interests by ensuring the best execution.

Foreign exchange swaps and cross-currency swaps  463

that are difficult to replace. For instance, allocating (managing) credit lines to (for) clients and enabling hard-and-fast prices to be quoted can be trickier in an MDP setting. Settlement is another area of the FX infrastructure that has witnessed considerable progress. Settlement or “Herstatt” risk, that is, the danger that the buyer (seller) may not receive delivery (payment) of the bought (sold) currency by the settlement date, has been significantly reduced over the last two decades. In this respect, CLS plays a crucial role.22 It operates the world’s largest multicurrency cash settlement system, handling 40% of global spot, swap, and forward FX transaction volumes—or even more, depending on estimates.23 After agreeing on a transaction, CLS members send the payment instructions to CLS to be matched, confirmed, and stored until value date.24 At the beginning and end of every business day, each settlement member’s multicurrency account has a zero balance, and the funding and pay-out of multilateral net positions is conducted using a predefined schedule.25 Along with settlement risk reduction, the multilateral netting approach has improved operational efficiency and reduced funding requirements. Efficiency in terms of collateral allocation and use has been particularly important in recent years, especially given the scarcity of (high-quality and liquid) collateral assets induced by various factors such as regulation and large asset purchase (quantitative easing) programs conducted by central banks. 20.2.3 Policy Actions In this section, we focus on the post–financial crisis regulation that has affected FX swaps. After that, we briefly discuss the dismissal of the Libor rate.26 Before the Global Financial Crisis erupted in 2007–2008, the regulation of derivative markets was quite lax and uneven across jurisdictions. In 2009, the Group of Twenty (G20) leaders agreed on strengthening and harmonizing derivative market regulation. This policy agenda has given rise to various regulatory initiatives whose exhaustive treatment is beyond the scope of this chapter. Here, we focus on a few important aspects of the Basel III legislation, which are arguably having significant and heterogeneous consequences for FX swap contracts and counterparties.27 Specifically, we discuss the regulation of derivative contracts and banking, which are also incorporated in part into important legislative frameworks such as the Dodd– Frank Act and the European Financial Market Infrastructure Regulation (EMIR). These new 22 In reaction to the failure of Bankhaus Herstatt in 1974, the Committee on Payment and Settlement Systems (CPSS) called on banks to develop multicurrency settlement and netting arrangements to reduce settlement risk with the support of central banks. The first step was the creation of CLS, which became operational in 2002. 23 Cespa, Gargano, Riddiough, and Sarno (2021) estimate that CLS handles around over 50% of the global FX transaction volume, while Bech and Holden (2019) estimate that 40% of FX trading was settled using a PvP method. 24 CLS is member-owned comprising over 70 of the world’s largest financial institutions. Over 25,000 third parties, primarily buy-side institutions, access CLS Settlement via CLS’s settlement members. 25 In this framework, settlement members pay and receive funds “irrevocably" through CLS’s central bank account in each currency via their own accounts or nostro bank accounts. 26 Another important policy measure is represented by the central bank swap lines, which are discussed in another chapter of this handbook. 27 Basel III refers to a new regulatory framework on bank capitalization, stress testing, and liquidity risk announced by the Basel Committee on Banking Supervision in July 2010.

464  Research handbook of financial markets

rules have been joined by other guidelines, such as the FX Global Code published in May 2017 and updated in July 2021.28 One of the central parts of the political agenda affecting FX swaps has been reform of over-the-counter (OTC) derivatives. The regulatory requirements vary across jurisdictions and depend on whether the counterparties are regulated entities (a bank, securities dealer, insurance company, fund management company, or asset manager) and the size of the outstanding notional of OTC derivatives. In general, trading in FX derivatives might be subject to obligations that aim to improve post-trading transparency (reporting, trade confirmation, and reconciliation/dispute resolution), decrease ex-ante uncertainty (risk mitigation, valuation, and initial and variation margins), and portfolio compression intended to net out OTC transactions. However, FX swaps are generally less regulated than other derivatives, such as interest rate swaps (IRSs). For instance, contrary to IRSs, it is not mandatory that an FX swap be cleared by a central counterparty (CCP), and it is not subjected to platform trading obligations such as swap execution facilities (“SEF”). Importantly, FX swaps are physically settled derivatives and generally not subject to margin requirements in the form of initial or variation margin (ISDA, 2021). Although a variation margin obligation is generally applied, the initial margin requirement applies only to entities if the aggregated month-end average gross position of OTC derivatives not cleared through a central counterparty exceeds a given size. What has certainly increased is the data collection promoted by market regulators, whose analysis should improve the ability to conduct financial stability policy, but also internal risk assessment and peer monitoring. Even if less regulated than other derivatives, the new legislative apparatus has created direct and indirect (“shadow”) costs that heterogeneously affect FX swaps, depending on which categories market participants formally belong to, their characteristics (e.g., balance sheet composition and size, business models, etc.), and how they act in the markets. Here, we focus on two measures: the Basel III leverage ratio and the special treatment of Globally Systematically Important Banks (G-SIB). The common wisdom is that the leverage ratio rule impacts FX swaps, but whether and how it does so is not at all clear. The leverage ratio is a nonrisk weighted capital requirement according to which a bank has to hold a minimum level of high-quality loss-absorbing (Tier 1) capital in proportion to “on-balance sheet” instruments, such as loans, securities, or repurchase agreements. Conversely, FX swaps fall in the category of the so-called “off-balance sheet” instruments that marginally contribute to the leverage ratio computation.29 On the other hand, by weighing all exposures equally, the leverage ratio rule triggers an increase in intermediation costs of those assets that are characterized by a low margin and high volume, such as FX swaps. As discussed in the section on XVA, these regulatory costs become even more relevant when banks have limited capacity to net out derivatives exposure that offset

28 The Global Code provides a set of global principles of good practice in the foreign exchange market. 29 Being an off-balance sheet instrument, FX swaps’ contribution to exposure under the leverage ratio goes through what is known as an “add-on factor” for potential future exposure (PFE). For FX and gold derivatives of maturities less than or equal to one year, the PFE factor is 1% (BIS, 2014). There may be indirect effects, though. For instance, an (arbitrage) trade might need additional funding that might increase the LR exposure by 100%, here depending on the balance sheet structure of the institution performing it.

Foreign exchange swaps and cross-currency swaps  465

each other across different counterparties and when counterparties entail higher credit risk exposure. Because FX dealers are typically large global banks, the G-SIB rules involve them. For a bank classified as a G-SIB, FX swap exposures add to the overall G-SIB score, creating an additional capital surcharge. Depending on the bank’s business model, FX swap exposures can contribute to several indicator categories all at once.30 However, assessing the exact mechanism of how the G-SIB framework ultimately affects the incentives to trade FX swaps is again a complex task and is likely to depend on each bank individually. The same applies to other (risk-weighted) capital requirements31 and liquidity requirements such as the liquidity coverage ratio (LCR) and the net stable funding ratio (NSFR), which requires banks to hold high-quality liquidity assets (HQLA) against potential net cash outflows during a short and longer stress period, respectively. Furthermore, some nations apply additional or countercyclical requirements. Let us now move to the discontinuation of the Libor rate, that is, the fact that the vast majority of Libor tenors will not be published after the end of 2021.32 Three considerations are noteworthy. First, it impacts more cross-currency swaps than FX swaps because a crosscurrency swap must explicitly name an interest rate in the contract to determine the rates to be exchanged, which has always been the Libor in the past. Second, it is happening now and although the way forward is not entirely clear at the moment, regulators, central banks, and industry actors33 are working on best practice guidelines and calling for market participants to switch to alternative risk-free rates (RFRs) such as SOFR (secured overnight financing rate for the the USD), Sonia (Sterling overnight index average), TONA (Tokyo overnight average), and Saron (Swiss average Rrate overnight).34 Although at the end of 2021, cross-currency swaps involving the Libor were still predominant, the interbank market has been gradually moving over to RFR, at least for four currency settings: the US dollar, sterling, Swiss franc, and Japanese yen (Risk​.ne​t, 2021a). Third, we should expect that the adoption of different RFRs based on different calculation methods will fuel price dispersion in derivative contracts. Recent studies show that the dispersion of money market rates in individual currencies has substantially increased since the Global Financial Crisis (Ballensiefen & Ranaldo, 2022), and 30 FX swaps contribute the most to the G-SIB score through the so-called complexity component because it considers the total notional amount of OTC derivatives (BIS, 2013). Other components can also be affected, depending on the exact nature of FX swap exposures. 31 For example, for US G-SIBs, the Tier 1 capital ratio increased from 4% precrisis to the 9.5% to 13% range under Basel III, and the total capital ratio increased from 8% to the 11.5% to 15% range. 32 A few Libor tenors linked to the USD, however, will continue until the end of June 2023 to allow most “legacy” or outstanding contracts to mature. 33 For instance, institutions such as the US Commodity Futures Trading Commission’s Market Risk Advisory Committee, the UK Financial Conduct Authority, and the RFR working groups are promoting the transition into the new regime for the four most affected currencies. There was some uncertainty about the fate of the reference rate for the euro, but this is fading following the ESMA’s recommendation to switch EONIA to €STR. 34 A survey conducted by Duff and Phelps (2021) shows that half of the firms surveyed did not have a transition plan in place in early 2021. However, concrete changes are taking place. For instance, interdealer trading conventions for cross-currency swaps between USD, JPY, GBP, and CHF LIBOR have moved to each currency’s risk-free rate (i.e., “RFR first”) as of September 21, 2021, and the share of transaction references in RFR increased sharply starting in the summer of 2021. Anecdotal evidence also suggests there is also more reliance on overnight index swap (OIS) rates, at least for short and medium maturities.

466  Research handbook of financial markets

this is reflected in benchmark rates referenced in derivative contracts (Klingler & Syrstad, 2021). In some cases such as Sonia and TONA, the Libor is replaced by another unsecured overnight rate and, therefore, will incorporate risk aspects such as counterparty credit risk. In other cases, such as SOFR and Saron, the Libor is substituted by a secure (collateralized) rate, making the benchmark rate less exposed to risk but more subject to collateral assets, which, in turn, depend on all those factors that are influencing the supply and demand for (safe) assets such as fiscal and monetary policy (Ballensiefen, Ranaldo, & Winterberg, 2022).

20.3 RESEARCH The OTC nature of the FX swap market has prevented us from thoroughly studying many aspects of its functioning and pricing. A major obstacle has been the lack of comprehensive and granular data. However, some new data sources give hope that we will soon improve our knowledge of this market. Among them, it is worth noting two sources: first, repository and supervisory data offering information at transaction and identity (ID) levels and, second, aggregate high-frequency data from CLS representative of global trading volume. There are at least three areas of future research to highlight, first of which is the market design of FX swaps. Some frictions of OTC markets, such as search costs and bargaining, have been studied theoretically (e.g., Duffie, Garleanu, & Pedersen, 2005, 2007; Colliard, Foucault, & Hoffmann, 2018). However, the discussion above has shown that the FX swap market features many more important characteristics, including the two-tier, dealer-centric network structure that involves a highly diverse client base.35 It would be helpful to have a theoretical framework that more closely matches with these characteristics to better understand the effects of market frictions and policy-relevant instruments involving FX swaps, such as central bank FX swap lines,36 (unconventional) monetary policies,37 or the regulatory issues discussed above.38 The second research area deserving more focus is asset pricing. Much research has been done in the past to shed light on spot FX pricing, but less attention has been paid to FX swaps. It is true that the deviations from the CIP condition, which is the cornerstone of currency forward and swap pricing, have recently been the subject of advanced research.39 By studying FX rates at quarter ends, recent research suggests that the limited intermediaries’ balance sheet capacity induced by the post-crisis regulation causes CIP deviations.40 Yet as previously 35 Hasbrouck and Levich (forthcoming) show that traders who hold a more central position in the FX spot market network benefit from a centrality premium and bargaining power. 36 See, e.g., Bahaj and Reis (2021). 37 e.g., Brazil is an interesting case because its central bank has used FX swaps to conduct (sterilized) FX interventions (Chamon, Garcia, & Souza, 2017). 38 Comparing IRSs that are centrally and bilaterally cleared, Cenedese, Ranaldo, and Vasios (2020) show that the latter bear an OTC premium consistent with a higher regulatory cost. 39 Akram, Rime, and Sarno (2008) provide evidence that the CIP held far in the past, while Baba, Packer, and Nagano (2008) and Mancini-Griffoli and Ranaldo (2011) relate CIP deviations to credit risk and (dollar) funding liquidity constraints during the Global Financial Crisis. 40 Du, Tepper, and Verdelhan (2018) analyze the CIP basis at quarter ends, pointing to a causal effect of the leverage ratio requirement on FX rates. Conducting a dealer-level analysis, Cenedese, Della Corte, and Wang (forthcoming) exploit the exogenous variation introduced by the UK leverage ratio framework. Abbassi and Bräuning (2021) provide evidence that banks with large dollar funding needs have a strong incentive to sell dollars forward before the quarter ends.

Foreign exchange swaps and cross-currency swaps  467

discussed, it is quite challenging to pinpoint whether and how the new regulation impacts FX swap pricing and arbitrage conditions. The FX swap transaction volume has remained relatively high at quarter ends, as shown in Figure 20.8, suggesting that dealers have fulfilled their role as intermediaries to meet the demand of their clients. What emerges most clearly from Figure 20.8 is the dramatic reduction in FX swap transaction volume during IMM days, a phenomenon that has not yet been studied. In addition to the supply effects (represented by FX dealers) on FX forward and swap prices, future research should shed light on the demand side, that is, dealers’ customers and their primary reasons for demanding swaps: hedging and funding. Some of the overarching questions that arise are as follows: How and why do dealer customers use FX swaps as hedging instruments?41 To what extent, and how, are FX swaps used as a funding instrument? How do alternative funding sources such as money market instruments affect FX swaps pricing and why?42 In addition to studying the forces driving the supply and demand for FX swaps, we should better understand which systematic pricing factors explain the cross-sectional and time-series variation of FX swap rates. One promising way to hone in on FX swap asset pricing is to adopt the same approach that is used in practice: XVA.43 XVA components are natural candidates to represent FX swap pricing factors. The third research area calling for more examination is market microstructure. It is important to better understand how price formation occurs and which factors determine it. In many markets, including the FX spot market, it has been shown that order flow determines the price formation process. The initial research work carried out so far on FX swap order flow is promising, but much remains to be understood.44 Future research should highlight how the various “frictions” discussed above impact the price formation process and information content of order flows in various market segments, such as in the interdealer and dealer-customer segments.45 Another under-researched microstructure issue is market liquidity, which may not be uniform and abundant for every FX swap, for every maturity, and at every point in time.46

41 Borio, Iqbal, McCauley, McGuire, and Sushko (2016) analyze how FX hedging demand and costly balance sheets affect the CIP basis. Liao and Zhang (2021) theoretically and empirically study how the currency hedging channel impacts FX spot and forward rates in countries with external imbalances. Alfaro, Calani, and Varela (2021) analyze firms’ currency risk exposure and their hedging strategies. Casas, Meleshchuk, and Timmer (2021) study whether FX derivatives protect the import and export of firms enduring a currency revaluation. 42 For instance, segmentation in money markets (Ranaldo, Schaffner, & Tsatsaronis, 2019) might affect the market quality of FX swaps. Moreover, large (international) banks raise significant funding in US dollars, while smaller banks tend to fund themselves in their domestic currencies. 43 For instance, dealers intermediating IRS appear to be charging regulatory costs and risk premiums to their customers, which is consistent with XVA (Cenedese, Ranaldo, and Vasios, 2020), including funding valuation adjustments (Andersen, Duffie, & Song, 2019). 44 Using data on interdealer transactions, Syrstad and Viswanath-Natraj (2020) provide evidence that order flow determines the FX swap rate, thus increasing the cost of dollar funding, especially at quarter-end periods. 45 Using FX spot flow data, Ranaldo and Somogyi (2021) shows that the asymmetric information risk is priced. Using FX volume data, Cespa, Gargano, Riddiough, and Sarno (2021) show that currencies with abnormally low volumes display strong return reversals. Regarding FX derivatives, a first step in this direction has been taken in Hau, Hoffmann, Langfield, and Timmer (forthcoming), showing that FX dealers exert discriminatory pricing in FX forward and swap rates. 46 Krohn and Sushko (2021) show strong commonality in liquidity between FX spot and swap markets that can be impaired by funding constraints and when the largest dealers pull back.

468  Research handbook of financial markets

REFERENCES Abbassi, P., & Bräuning, F. (2021). Demand effects in the FX forward market: Micro evidence from banks’ dollar hedging. Review of Financial Studies, 34(9), 4177–4215. Akram, F. Q., Rime, D., & Sarno, L. (2008). Arbitrage in the foreign exchange market: Turning on the microscope. Journal of International Economics, 76(2), 237–253. Alfaro, L., Calani, M., & Varela, L. (2021). Central Bank swap lines: Evidence on the effects of the lender of last resort. NBER Working Paper NBR. 28910. Andersen, L. B. G., Duffie, D., & Song, Y. (2019). Funding value adjustments. Journal of Finance, 74(1), 145–192. Baba, N., Packer, F., & Nagano, T. (2008). FX and OTC derivatives markets through the lens of the triennial survey. BIS Quarterly Review, March. Bahaj, S., & Reis, R. (2021). Currency hedging: Managing cash flow exposure [Working paper]. Ballensiefen, B., & Ranaldo, A. (2022). Safe asset carry trade. Review of Asset Pricing Studies. Ballensiefen, B., Ranaldo, A., & Winterberg, H. (2020). Money market disconnect, Review of Financial Studies, forthcoming. Bech, M. L., & Holden, H. (2019). FX settlement risk remains significant. BIS Quarterly Review, December. BIS. (2013). Global systemically important banks: Updated assessment methodology and the higher loss absorbency requirement [Technical report]. Basel Committee on Banking Supervision. BIS. (2014). Basel III leverage ratio framework and disclosure requirements [Technical report]. Basel Committee on Banking Supervision. BIS. (2019). Triennial central bank survey, foreign exchange turnover in April 2019. Monetary and Economic Department. Borio, C., Iqbal, M., McCauley, R. N., McGuire, P., & Sushko, V. (2016). The failure of covered interest parity: Fx hedging demand and costly balance sheets. BIS Working Paper No. 590. Revised November 2018. Casas, C., Meleshchuk, S., & Timmer, Y. (2021). The dominant currency financing channel of external adjustment [Working paper]. Cenedese, G., Della Corte, P., & Wang, T. (2021). Currency mispricing and dealer balance sheets. Journal of Finance, 76(6), 2763–2803. Cenedese, G., Ranaldo, A., & Vasios, M. (2020). OTC premia. Journal of Financial Economics, 136(1), 86–105. Cespa, G., Gargano, A., Riddiough, S. J., & Sarno, L. (2021). Foreign exchange volume. Review of Financial Studies, 00, 1–42. Chamon, M., Garcia, M., & Souza, L. (2017). FX interventions in Brazil: A synthetic control approach. Journal of International Economics, 108, 157–168. Colliard, J. E., Foucault, T., & Hoffmann, P. (2021). Inventory management, dealers’ connections, and prices in otc markets. Journal of Finance, 76(5), 2199–2247. Du, W., Tepper, A., & Verdelhan, A. (2018). Deviations from covered interest rate parity. Journal of Finance, 73(3), 915–957. Duff and Phelps. (2021). Nearly half of organizations still do not have a firm libor transition plan in place. Retrieved from https://www​.duffandphelps​.com​/services​/ libor​-transition​-advisory. Duffie, D., Garleanu, N., & Pedersen, L. H. (2005). Over-the-counter markets. Econometrica, 73(6), 1815–1847. Duffie, D., Garleanu, N., & Pedersen, L. H. (2007). Valuation in over-the-counter markets. Review of Financial Studies, 20(6), 1865–1900. Euromoney. (2020). Euromoney FX survey 2020. Euromoney. Green, A. (2015). XVA: Credit, funding and capital valuation adjustments. Hoboken, NJ: John Wiley & Sons. Gregory, J. (2015). The XVA challenge: Counterparty credit risk, funding, collateral and capital. Hoboken, NJ: John Wiley & Sons. Hasbrouck, J., & Levich, R. M. (2021). Network structure and pricing in the FX market. Journal of Financial Economics, 141(2), 705–729.

Foreign exchange swaps and cross-currency swaps  469

Hau, H., Hoffmann, P., Langfield, S., & Timmer, Y. (2021). Discriminatory pricing of over-the-counter derivatives. Management Science, 67(11). Hull, J. (2017). Options, futures and other derivatives management (10th ed.). Pearson. ISDA. (2021). Derivatives subject to non-cleared margin rules. Retrieved from https://www​.isda​.org​/a​/ 9ZDgE​/ ISDA​-In​-Scope​-Products​-Chart​_UnclearedMargin​_In​-process–6​.10​.21​.​pdf. Klingler, S., & Syrstad, O. (2021). Life after LIBOR. Journal of Financial Economics, 141(2), 783–801. Kloks, P., Mattille, E., & Ronaldo, A. (2023). Foreign Exchange Swap Liquidity, Swiss Finance Institute Research Paper, 23–22. Krohn, I., & Sushko, V. (2022). Fx spot and swap market liquidity spillovers. Journal of International Money and Finance, Elsevier, vol. 120(C). Liao, G. Y., & Zhang, T. (2021). The hedging channel of exchange rate determination [Working paper]. Mancini-Griffoli, T., & Ranaldo, A. (2011). Limits to arbitrage during the crisis: Funding liquidity constraints and covered interest parity. Working paper Swiss National Bank. Ranaldo, A., Schaffner, P., & Tsatsaronis, K. (2019). Euro repo market functioning: Collateral is king. BIS Quarterly Review, 95–108. Ranaldo, A., & Somogyi, F. (2021). Asymmetric information risk in FX markets. Journal of Financial Economics, 140(2), 391–411. Risk​.ne​t. (2021a). Cross-currency swaps set to ditch libor in ‘RFR first’ drive. Retrieved from https:// www​.risk​.net ​/derivatives​/7860941​/cross​-currency​-swaps​-set​-to​-ditch​-libor​-in​-rfr​-first​-drive. Risk​.ne​t. (2021b). E-trading takes hold for FX swaps – Sort of. Retrieved from https://www​.risk​.net​/ infrastructure​/7485516​/e​-trading​-takes​-hold​-for​-fx​-swaps​-sort​-of. Schrimpf, A., & Sushko, V. (2019). FX trade execution: Complex and highly fragmented. BIS Quarterly Review, December. Syrstad, O., & Viswanath-Natraj, G. (2022). Price-setting in the foreign exchange swap market: Evidence from order flow. Journal of Financial Economics, 146(1), 119–142.

21. Inflation hedging products Stefania D’Amico and Thomas B. King1

21.1 INTRODUCTION1 This chapter reviews the efficacy and cost of some simple strategies for hedging inflation risk in financial markets, focusing on evidence from the United States. Although the question of which assets provide good inflation hedges may seem straightforward, we emphasize that there is no one-size-fits-all portfolio prescription as different investors may have different types of inflation-hedging objectives. Economists’ knee-jerk instinct is to deflate every nominal series by some broad index like the CPI or PCE deflator. But for individual investors this may be a very misleading way of measuring the relevant “real” cash flows. For example, suppose a consumer’s only income is her monthly wage, her only asset is her house, and her only expense is her fixed-rate mortgage payment. This consumer is exposed to the risk of wage inflation (or, more precisely, wage disinflation). She also may be exposed to house-price changes, but the nature of that exposure will depend on her housing plans—whether to move to a substantially different housing market or switch to renting, and at what point in the future this might occur. In any case, a calculation of her real debt burden that deflated her mortgage by the CPI would make no sense, and a hedging strategy that involved a CPI-linked product would leave her exposed to the basis between consumer prices and her wage, and possibly to the basis between consumer prices and house prices. Of course, most consumers don’t just pay a mortgage. But even a household that consumes exactly the CPI basket every month will also have assets, liabilities, and an income stream that will make the CPI an imperfect proxy for its net exposure to price changes. Similarly, even well-diversified firms have costs and revenues that are likely to diverge dramatically from consumer and producer price indices and that may diverge by different amounts over different horizons. In general terms, any investor i (household, business, financial institution, pension fund, insurance company) wants to protect the real value of her own expected future net worth over an horizon τ:

ét + t ù NW = Et ê xiA,s - xiL,s M sN ds ú (21.1) ê ú ët û i t ,t

ò(

)

1 For helpful comments, discussions, and suggestions we thank Francois Gourio, Shane Sherlund, and the editors. Corey Feldman provided superb research assistance, and Santiago Sordo Palacios was a huge help in updating estimates of the inflation risk premium. The views expressed are not official positions of the Federal Reserve Bank of Chicago or the Federal Reserve System. 470

Inflation hedging products 

471

where xiA,t and xiL,t are nominal cash flows that the investor receives and pays, respectively, at time t, and M tN is the nominal pricing kernel. Equivalently, the investor discounts real quantities by the real pricing kernel MiR,s :



ù ét + t æ x A x L ö NW = Et ê ç sA - sL ÷ MiR,s ds ú (21.2) Ps ø ú ê è Ps û ët i t ,t

ò

where Pt A and Pt L are the prices associated with the respective cash flows (i.e., the asset and liability prices). The real pricing kernel is indexed by i because it is defined in relation to a particular price index, which, in turn, is unique to each specific agent’s situation:

Pi,t = MiR,t / M tN (21.3)

Pi,t is a weighted difference of Pt A and Pt L . For investor i, hedging inflation risk means minimizing the variance of this index. This simple framework allows us to think about different aspects of hedging inflation risk. If the asset and liability prices do not change at the same rate, then the inflation risk originates from the gap between the two inflation rates. Furthermore, asset inflation and liability inflation can vary across investor types. For instance, most households will be concerned about the difference between wage inflation and consumer-price inflation, while businesses will be concerned about input inflation (wage and capital-investment inflation) versus output inflation (producer prices). Long-term institutional investors, such as insurance companies and public pension funds, typically have cost-of-living-adjusted liabilities and therefore need assets that adjust accordingly. Even within a certain investor class, the need for protection against inflation may change with specific investor characteristics. For instance, within the household category, an average retiree may be more exposed to medical expenses than an average urban consumer; and an average student may be more exposed to education expenses (Parikh et al., 2019). Finally, the timing of the cash flows matters. The differences in the asset and liability inflation rates can be magnified if there is a duration gap between assets and liabilities. Even when the price index for assets and liabilities is the same, the investor may still face inflation risk due to the duration gap. A special case that illustrates the importance of this point is an investor who needs to hedge a cash flow at a particular horizon. In the absence of real-yield fluctuations, a ten-year inflation-linked bond should be a perfect hedge for its underlying price index over a period of exactly ten years. But, as we show, it can be quite a poor hedge at shorter horizons. These considerations are relevant because different types of inflation are imperfectly correlated with each other. Consequently, a given asset’s hedging abilities can be good for one type of inflation but not for another type. For example, commonly used “real assets” such as stocks, commodity futures, and real estate are generally good hedges for energy inflation but not for core inflation (Fang, Liu, & Roussanov, 2021; henceforth FLR). Treasury Inflation Protected Securities (TIPS) on the other hand, are highly exposed to core CPI inflation but do not protect from energy inflation at short to medium horizons. We review the literature on the inflation-hedging ability of different assets and add some new evidence of our own. Over the last twenty years, we find that commodities are generally successful in hedging headline consumer inflation, but this mostly seems to reflect their

472  Research handbook of financial markets

significantly positive relation with energy prices. Related assets, such as the stocks of oil and gas and metals-mining firms and some emerging-market (EM) currencies, share this property. Hedging core inflation is harder. At horizons of less than a year, there is little protection available, except for TIPS. At longer horizons, short-term nominal bonds and real estate provide a decent hedge. Certain stock-market strategies can also work, but here one has to be careful because performance varies substantially across different types of stocks. There are also some significant differences in how stock and bond strategies perform against core CPI versus core PCE. For instance, post-1999, two-year nominal bonds are a good hedge for core CPI at longer horizons, while certain stock-market sectors such as oil and gas are a good hedge for core PCE. We find that core producer prices (PPI) and wages are the most difficult types of inflation to hedge, although real estate and short-term nominal bonds provide some protection. Most consumer-price inflation stems from three sources: real estate costs, the passthrough of materials and energy prices to consumer goods, and the passthrough of labor costs to goods and services. Materials and energy prices are closely tied to commodities, which can be hedged very effectively through futures markets or closely related assets like sectoral equities. Real estate can now be effectively hedged through real estate investment trusts (REITs) and other instruments that provide broad exposure to this sector. While the literature does not typically focus on hedging labor costs, our empirical exercise identifies short-term nominal bonds and real estate as a reasonably good hedge in this dimension. And, since wages are a particularly large component of the cost of non-housing services, the same instruments do a decent job hedging service inflation. Another consideration is that the nature of inflation risk that needs to be hedged depends on the composite growth-inflation regime (e.g., high inflation, high growth–demand shocks dominate; high inflation, low growth–supply shocks dominate; low inflation, low growth–deflation bias dominates; and transition among those regimes which creates inflation uncertainty as shown in David & Veronesi, 2013). Apparently, core and energy prices are also driven by different demand shocks, hence the simple distinction between regimes in which supply shocks dominate and regimes in which demand shocks dominate is not sufficient (FLR). We also document that the properties of inflation hedging products have changed over time, especially because, in the U.S., after the global financial crisis (2007–2009), disinflation/deflation risk has dominated inflation risk. Finally, we turn to the cost of inflation risk by focusing on measures of inflation risk premia (IRP). These premia are a strong indicator of the attractiveness of an asset as an inflation hedge. They have typically been measured from no-arbitrage dynamic pricing models of the nominal and real-term structure, but methods have recently been developed to extract them from other asset classes as well. The changing nature of inflation risks over time has strongly affected the IRP embedded in various assets. The size and sign of the IRP are determined by the covariance between inflation and the real side of the economy, and as this covariance has transitioned from mostly negative to mostly positive over the last 20 years, nominal bonds have typically commanded a negative IRP and stocks a positive IRP.

21.2 LITERATURE REVIEW In this section, we review the evidence from previous studies on how different assets have performed in hedging inflation. It is useful to distinguish between “real assets” and “nominal

Inflation hedging products 

473

assets.” By “real assets” we mean those whose value is directly tied to physical assets. The primary assets in this category are equities, real estate, and commodities. Real assets contrast with currencies and bonds. Currencies generally lose value in response to inflation in the home country, which means that in some cases foreign currencies can be a good inflation hedge. Bond yields compensate investors for expected inflation over the life of the bond, but if inflation is higher than expected the real purchasing power of the cash flows will be eroded. To the extent that unexpected inflation leads to revisions of future expected inflation, this loss of real purchasing power can be significant. An exception to this is inflation-indexed bonds (which might also be considered “real assets,” though in a slightly different sense). 21.2.1 Stocks: Broad Indices and Individual Stocks Stocks represent claims against real assets, such as factories, equipment, and inventories. However, as pointed out in Gorton and Rouwenhorst (2006), firms also have contracts with suppliers of inputs, labor, and capital, that are fixed in nominal terms and hence act very much like nominal bonds. Furthermore, if inflation is negatively correlated with real economic performance, firms may suffer losses in high-inflation regimes: their equity prices may fall even if they are adjusting perfectly to inflation. These observations imply that whether stocks provide a good hedge against inflation is an empirical question. Numerous studies have documented the poor inflation-hedging ability of the aggregate stock market in the U.S. and in other countries. Some work shows that it is possible to construct good inflation-hedging portfolios using individual equities or equity sectors, but the performance is time-varying and depends on the measure of inflation under consideration. Ang et  al. (2012) show that there is considerable heterogeneity in how individual stock returns covary with CPI inflation, as different companies have different pricing power.2 They find that since the 1990s, the stock portfolios with the highest inflation exposure overweight oil/gas, which benefits from rising commodity prices, and technology, which benefits from technological innovation. However, they also show that trying to forecast inflation exposure at the individual stock level is difficult, as co-movements with inflation exhibit pronounced time variation, including a change in sign post-2008. This makes it hard to construct portfolios of stocks that are good out-of-sample inflation hedges. However, Parikh et al. (2019) show that some equities also have a good out-of-sample performance against CPI inflation over the period between January 1990 and January 2014.3 They find that the best-performing stocks are concentrated in the energy and technology sectors, in line with Ang et al. (2012). While the previous two studies evaluate the short-run performance of individual stocks, Bampinas and Panagiotidis (2016) focus on the long-run (LR) perspective. That is, the stocks are ranked based on their cointegrating relation with CPI. Their in-sample estimates over the period 1993–2012 indicate that the LR relation between the aggregate stock market and CPI

2 They construct portfolios based on individual S&P 500 stocks’ realized inflation betas (i.e., the slope coefficient in rolling regressions of stock returns on inflation, which can be measured either by realized, expected, or unexpected inflation rates) and analyze their performance in- and out-of-sample. 3 Unlike Ang et al. (2012), instead of using current-month CPI, Parihk et al. (2019) use the subsequent one-month change in CPI as the regressor, and they do not use five-year rolling regressions but the full sample available at each point in time.

474  Research handbook of financial markets

inflation is insignificant, while individual stocks in the healthcare, industrials, and energy sectors perform quite well. Out-of-sample, the energy sector has the highest covariation with inflation, followed by the materials and consumer staples sectors. However, there is considerable time variation in those LR relations. The results about a statistically insignificant LR relation between the aggregate stock market and CPI inflation contrast with the results of Boudoukh and Richardson (1993) over the sample period 1802–1990. They find that five-year stock returns and five-year inflation are positively and significantly related. Further, these findings are robust with respect to subperiods as well as in the case of the U.S. and U.K. markets. This suggests that the inflation-hedging ability of the aggregate stock market deteriorated post-1990. Importantly, after decomposing headline CPI inflation into core and energy, FLR finds that stocks’ exposures to the two components are very different.4 Stock portfolios over the 1963:Q2–2019:Q4 sample period have significantly negative core-inflation exposure but positive energy-inflation exposure, which implies that the insignificant relation between stocks and headline CPI may be due to the two opposite exposures offsetting each other. Even across five industry stock portfolios (consumer, manufacturing, high tech, health, and others), all the core exposures are significant and negative, while the energy exposures are positive (though only the one for the manufacturing sector is statistically significant). Overall, these results suggest that an asset’s inflation-hedging properties may depend on the measure of inflation. This motivates our analysis of multiple inflation indices in Section 21.3. 21.2.2 House Prices and REITs A REIT is essentially an investment company that owns real estate–related assets. Shares in REITs trade on organized exchanges or in the over-the-counter market and their ownership allows investors to participate in large real-estate investments selected and managed by professionals. There are three major types of REITs. Mortgage trusts primarily hold long-term mortgages. Equity trusts are more focused on ownership of commercial property such as shopping centers, office headquarters, and so on. And, finally, hybrid REITs hold a mix of mortgages and property. Since equity REITs are backed directly by physical real-estate assets, one might expect a priori that they would be superior to mortgage REITs in hedging against inflation. However, Park et al. (1990) find that none of the three types of REIT investments display reliable CPI- or PPI-inflation hedging properties, in an analysis of their returns over monthly, quarterly, semiannual, and annual horizons over the period 1972 to 1986. FLR obtain similar results with respect to headline CPI over the sample period 1980–2019, but they find that all REITs have highly significantly negative core-inflation exposure, similar to stocks. They also find that equity REITs are positively exposed to energy inflation over this period. 21.2.3 Commodities: Broad Indices and Individual Futures A commodity futures contract is an agreement to buy (or sell) a specified quantity of a commodity at a future date, at a predetermined price specified in the contract—the futures price. 4 In FLR, inflation hedging is defined as the correlation between monthly excess returns and unexpected inflation, which is different from the definition we use in Section 21.3.

Inflation hedging products 

475

Commodities, and hence commodity futures, come in many forms. Some commodities are storable and some are not; some are input goods, while some are intermediate goods. Many commodity futures are traded on U.S. exchanges—with the exception of some metals that are traded in London. Physical delivery occurs at a location within the contiguous 48 states, and settlement is in U.S. dollars. The U.S. markets for some commodity futures (gold, crude oil) should be integrated with global markets, but prices of others are likely to be influenced by local conditions (natural gas, live hogs). Gorton and Rouwenhorst (2006) construct monthly time series of equally-weighted indices of both commodity spot and commodity futures prices to assess the performance of this asset class as a whole. They show that average returns of monthly and annually re-balanced futures and spot indices have outpaced CPI inflation over the period between 1959 and 2004. But the average buy-and-hold spot return of 3.47% per annum is lower than the average inflation of 4.15% over the same sample period, suggesting that over the long term commodities’ inflationhedging performance deteriorates. They also show that commodity futures are a better inflation hedge than stocks or bonds. One obvious reason is that commodity futures represent a bet on commodity prices, which are directly linked to the components of inflation. Further, all the results carry through when the authors use unexpected rather than actual CPI inflation. Erb and Harvey (2006) point out that, over the 1982–2003 period, not all individual commodity futures were good inflation hedges. In particular, the precious metal sector, gold, and silver had statistically significant negative covariation with headline CPI. Finally, Kat and Oomen (2006) use daily settlement prices on 142 commodity futures contracts covering the period from 1973 to early 2005. Importantly, they also analyze three different types of inflation: CPI, PPI, and the employment cost index (ECI, from 1982). And, in the case of CPI and PPI, they consider both headline and core. They show that overall, at an annual frequency, commodity futures returns and inflation are positively correlated. The average correlation with CPI is 25.1%, with PPI 23.3%, and with ECI 22.8%. Energy, metals, cattle, and sugar display the highest correlation with inflation. However, those correlations are substantially lower with ECI, core CPI, and PPI. These results are broadly consistent with those we will report in Section 21.3. 21.2.4 Treasury Inflation-Protected Securities TIPS are fixed-income securities whose semiannual coupons and principal payments are indexed to the non-seasonally adjusted CPI for all urban consumers. The Treasury began issuing TIPS in 1997, and as of 2021 they constituted about 8% of outstanding marketable Treasury debt. Because they adjust for inflation, TIPS yields are sometimes known as “real yields,” and, to a first approximation, they are equal to the risk-neutral expectation of the average real shortterm rate over the life of the bond. The difference between a nominal Treasury yield and the TIPS yield at the same maturity is a proxy for the risk-neutral expected rate of inflation over that horizon and is known as “inflation compensation” or the “break-even inflation" (BEI) rate. Two technical features complicate TIPS’ ability to hedge the CPI perfectly, even when held to maturity. First, when TIPS mature, the investor is paid the adjusted principal amount or the original principal, whichever is greater. Hence, there is an embedded floor for the principal value that cannot decrease in case of deflation.5 Second, the adjustment of the TIPS principal 5 See Grishchenko et al. (2016) and Christensen et al. (2012).

476  Research handbook of financial markets

is done by scaling it up by the “index ratio,” which is obtained by dividing the reference CPI by the CPI at the time of issuance. The reference CPI has an indexation lag of about 2.5 months, which means that investors cannot precisely lock in a real return. The price at time t of a zero-coupon (ZC) TIPS with maturity τ and indexation lag l is determined as follows:

é M tN+ t CPI t + t -l ù é M tN+ t CPI t + t -l ù CPI t E Pt ZC = ; (21.4) , t = Et ê t ú ê N ú N CPI t û CPI t -l ë M t CPI t -l û ë Mt

where M tN is the nominal discount factor (also known as pricing kernel). The last term illustrates that the price of TIPS can be decomposed in the price of a ZC bond that perfectly hedges CPI inflation until maturity τ − l times the CPI inflation realized over the previous 2.5 months (l). Hence, TIPS holders receive compensation for the inflation that occurred 2.5 months before the purchase date, but are still exposed to inflation during the 2.5 months just preceding the maturity or sale of the bond. As a consequence, the price of TIPS will reflect the expected divergence between these two rates of inflation plus an indexation lag premium for the inflation risk associated to this expected divergence. D’Amico et al. (2018) show that the indexation lag premium is generally very small, but at times it can rise to 30 or 70 basis points at the ten- and five-year maturity respectively, as happened in December 2008. In general, the shorter the horizon of the inflation hedging strategy, the more relevant the indexation lag premium tends to be. Abstracting from these technicalities, if the real yield did not vary much, then, when held to maturity, TIPS could provide an almost perfect hedge against headline CPI. But, if the real yield fluctuates, TIPS can perform quite poorly if the investment horizon is different than the bond maturity. This is because, under the Taylor principle, real yields generally move in the same direction as inflation, causing TIPS prices to fall with a strong economy. Thus, capital losses on TIPS typically offset some of their built-in inflation protection. Further, these losses can be worse for longer-duration TIPS, which are more sensitive to yield volatility. Based on returns from 2001 to 2019, FLR report that TIPS were not a perfect hedge against headline CPI inflation. However, their exposure to core CPI inflation is quite large, as the TIPS index return increases by 4.54% in response to 1% increase in core inflation rate. In contrast, the TIPS index does not hedge against energy inflation. TIPS are the only real asset found to be useful against core inflation in the FLR study, and this is consistent with our results below that it is quite difficult to find instruments to hedge core measures of inflation. It is particularly interesting that TIPS are a good core hedge, given that they are indexed to headline CPI, not core. From a portfolio management perspective, TIPS provide a stream of known “real” payments at horizons up to 30 years and are therefore highly attractive to long-term investors, such as retirement savings accounts. Alternatively, investors can hold TIPS mutual funds and ETFs, which generally hold a portfolio of bonds across the term structure and may not hold them to maturity. In 2021, as inflation concerns were rising, flows into TIPS mutual funds and ETFs were exceptionally strong. As of July 2021, those flows accounted for roughly 25% of total assets under management. Since 2018, the cumulative fund flows into TIPS ETFs has been of about $250 billion.6 6 Based on Haver and EPFR Global, which tracks fund flows, and Bloomberg data.

Inflation hedging products 

477

21.2.5 Inflation Derivatives We mainly focus on the most-traded U.S. inflation derivatives and mention some of the less popular ones briefly, just for completeness. An inflation swap (IS) is a derivative transaction in which one party agrees to swap fixed payments for floating payments tied to an inflation index for a given notional amount and period of time. In the U.S., the standard contract references the non-seasonally-adjusted CPI for urban consumers (Fleming and Sporn, 2013). U.S. ISs were introduced when the Treasury began issuing TIPS in 1997. Indeed, the inflation index used in a standard IS matches precisely the inflation index for TIPS, including the same indexation lag. Hence, in theory, the IS rate and the TIPS-implied BEI rate should be equal in the absence of market frictions. In practice, IS rates are almost always higher (with the spread exceeding 100 basis points during the global financial crisis). This is mostly because the market is characterized by one-way flows with a lack of natural sellers of inflation protection. The only parties that sell protection are dealers and real money investors whose positions are ultimately hedged with TIPS cash flows and, therefore, they pass on the hedging costs to the receivers of inflation, charging a fee. ZC ISs are the most commonly used. They have only one cash flow (CF) at expiration. For instance, in the case of a ten-year ZC IS with rate equal to 200 basis points at inception, the cash flow exchange at maturity will be:

CFt + t = (1 + 0.0200)10 - I t + t ; (21.5)

where I t +t is the inflation index. ISs trade in a dealer-based over-the-counter market. The predominant market makers are the G14 dealers, which trade with one another and with their customers.7 Differently from TIPS, ISs can be tailored to more precisely meet investor needs because the IS maturity, notional amount, and other terms are agreed upon at the time of the trade. Hence, some entities might prefer ISs to TIPS. Further, ISs are often favored by pension funds and insurance companies because they allow to hedge inflation risk without funding a bond. Fleming and Sporn (2013) analyze all electronically matched ZC IS U.S. trades involving a G14 dealer from June 1 to August 31, 2010. They find that just over two trades per day occurred. Daily notional trading volume was estimated to average $65 million. In the TIPS market, in comparison, an estimated $5 billion per day traded over the same period, on average. However, their analysis reveals that, despite the over-the-counter nature and low level of trading activity, the market is reasonably liquid and transparent. Transaction prices are typically near widely available end-of-day quoted prices and realized bid-ask spreads are modest. In recent years, following regulatory initiatives, ISs have increasingly moved to central clearing, which may enhance their liquidity further. It is worth mentioning that, in February 2004, the Chicago Mercantile Exchange (CME) started trading futures on the U.S. CPI inflation index. The main advantage of CPI futures over ZC ISs was to mitigate counterparty risk. However, as reported in Kerkhof (2005), likely 7 As reported in Fleming and Sporn (2013), the G14 dealers are the largest derivatives dealers and, during the period covered by their study, include Bank of America, Barclays, BNP Paribas, Citigroup, Credit Suisse, Deutsche Bank, Goldman Sachs, HSBC, JP Morgan Chase, Morgan Stanley, Royal Bank of Scotland, Société Générale, UBS, and Wells Fargo.

478  Research handbook of financial markets

due to the ill-design of the CPI future (the contract traded annualized quarterly inflation), the market never really took off. Finally, in the U.S., the inflation options market started in 2002 with the introduction of caps and floors on the realized inflation rate. An inflation-indexed caplet is a call option on the forward inflation rate implied by the CPI index. Analogously, an inflation-indexed floorlet is a put option on the same inflation rate. Inflation caps and floors are bundles of inflation caplets and floorlets (respectively) over a given horizon. Trading in inflation caps and floors gained momentum following the global financial crisis, but has dried up recently. Using quotes from inflation caps and floors during the years in which the market was active, Kitsul and Wright (2013) analyze the economic properties of the probability density of inflation. Interestingly, by comparing the option-implied densities to those derived from time-series models, they show that the empirical pricing kernel is U-shaped, with investors having high marginal utility in states of the world characterized by either deflation or high inflation. This would explain why assets that pay off in those states are particularly attractive, as we will discuss in Section 21.4. 21.2.6 Currencies The rationale for currencies to hedge inflation risk is purchasing power parity (PPP). When the U.S. experiences higher inflation, all else equal, the purchasing power of the dollar declines and the foreign currency appreciates. The literature has considered different types of currency portfolios. Menkhoff et al. (2017) built value portfolios that are sorted based on the deviation from PPP. For instance, their first value portfolio would contain currencies that are most undervalued relative to their real exchange rates five years before. Undervalued currencies are expected to appreciate as they should revert back to their fundamental values. Verdelhan (2018) has constructed currency portfolios sorted on dollar betas, interacted with the sign of average forward discount. Lustig et al. (2011) use interest-rate-sorted currency carry portfolios, and Lustig et al. (2014) consider the dollar carry portfolio. FLR analyze the performance of all of these portfolios relative to headline CPI, as well as core and energy inflation, finding that they mostly hedge only against energy inflation. Another strategy to hedge inflation risk using currencies is to focus on those of emerging markets (EM) that are commodities exporters, such as the Russian ruble, South African rand, and Brazilian real. Those currencies typically gain when there is a rise in the prices of major commodities exported by the country. For instance, in July 2021, the South African rand was the best performing EM currency, supported by rising gold, base metal, and iron ore prices. The ruble also gained as oil prices in July 2021 reached their highest level since 2018.

21.3 EMPIRICAL PROPERTIES OF INFLATION HEDGES In this section, we present evidence on the inflation-hedging properties of various types of financial assets. As emphasized above, there is no single metric for this—it depends on what type of inflation an investor needs to hedge and over what horizon. With that in mind, we consider several different price indices over investment horizons ranging from one month to 30 years. The inflation measures we consider are listed in Table 21.1, while all the financial assets are listed in Table 21.2. (See Table B1 in the Appendix of D’Amico and King [2023] for data sources.) Note that we include commodity prices in both categories, since, depending on

Inflation hedging products 

479

Table 21.1  Measures of inflation CPI: headline; core; energy; services; durables; nondurables PCE: headline and core PPI: finished goods and finished goods core Wage inflation: hourly earnings Broad commodity indices: BCOM and GSCI

Table 21.2  Asset price indices Equities indices: Wilshire 5000; S&P 500 Equity portfolios: S&P industry sub-indices; Fama-French 5 factors Bonds: GI EM-bond index; Treasuries and TIPS at various maturities Commodity indices: BCOM index; GS index and subindices Commodities: gold, silver, wheat, soybeans, hogs; WTI and Brent spot and futures Real estate: Case-Shiller; Wilshire REIT ETF Currencies: dollar versus yen, euro, rand, ruble, real

the investor, a particular commodity may either represent a cost that needs to be hedged or a potential hedging instrument. We also include the CPI among the asset returns, even though it is not itself a traded instrument, as a proxy for how a standard inflation swap would have performed in hedging various types of inflation over time. We begin our baseline sample in 1999. The focus on this relatively short period allows us to bring in TIPS and many other inflation and asset-return series that cannot be considered in longer samples. It is also valuable to look mainly at recent behavior, since there is evidence that correlations have shifted over time. We also consider a second sample beginning in 1972 using a subset of the series that are available over that period. Both samples end in August 2021. Our measure of a given asset’s inflation-hedging ability is its nominal price’s simple (Pearson) correlation with each type of inflation (abstracting from transactions costs). Computing these correlations at short horizons would be straightforward. However, at investment horizons of more than a few years, the raw data would not be sufficient to estimate correlations precisely. For ten-year correlations, for example, even our longer sample contains only five non-overlapping observations. To overcome this problem, we estimate time-series models of each price index/asset price pair, and we use these models to project the correlations at different horizons. Specifically, for each pair of variables for which we have monthly data, we search across possible VARIMA(p,1, q) models, where p Î[0,12] and q Î[0,3], in log-levels. For the few quarterly series, we take p Î[0, 4] and q Î[0,1]. In each case, we select the best model using the AIC, simulate one million observations, and compute the correlations between changes in log-levels at horizons of one month up to 30 years. 21.3.1 Buying and Holding Bonds The treatment of equities and commodities in this exercise is relatively straightforward, but some difficulty arises when considering hedging using bonds in this context. A strategy of

480  Research handbook of financial markets

holding a bond for its entire life is very different from a strategy of continuously rolling over to maintain a bond portfolio of constant maturity. For instance, recall the discussion of TIPS performance over short and long investment horizons in Section 21.2.4. Consider a ZC nominal bond of maturity τ and an investment horizon h £ t.8 Denote the initial (time-t) yield on the nominal bond by ytN,t and real yield on the TIPS by ytR,t . Define each bond’s return as the change in its log price, and recall that the price of a nominal bond is just exp[ -tytN,t ] . Then, the total nominal return on the nominal bond over the investment horizon is:

(

)

rtN,t +( th) = t ytN,t - ytN+ h,t - h + hytN+ h,t - h . (21.6)

When it comes to TIPS, the nominal return is:

(

)

rtR,t (+th) = t ytR,t - ytR+ h,t - h + hytR+ h,t - h + pt ,t + h (21.7)

CPI t + h is the log change in the CPI between periods t and t + h. Thus, to calCPI t culate the return on a holding strategy for both types of bonds, one needs to know the τ-period yields at the beginning of the investment and the (τ − h)-period yields at the end. In the case of TIPS, one also needs to know the intervening rate of inflation.9 With this in mind, we modify our strategy by extending the considered VARIMA models to include three variables rather than two: a price index, an τ-maturity bond yield, and an (τ − h)-maturity bond yield. Then, in our simulations, we calculate the returns on the bonds using Equations 21.6 and 21.7 and compute the correlations with the simulated inflation series over the same period. where pt ,t + h = log

21.3.2 Results Here we summarize our most important results, starting with the post-1999 sample (Tables 1A through 3A in the Appendix of D’Amico and King [2023]) and then turning briefly to the post1972 sample (Tables 4A and A5 in the Appendix of D’Amico and King [2023]). 21.3.2.1 Headline inflation Hedging food and energy inflation is relatively easy, since these prices are closely linked to commodities, with oil and natural gas spot and future prices displaying some of the highest correlations. Because variation in food and energy prices makes up most of the variation in headline inflation indices, this also means that commodities are generally a good hedge for headline inflation. The broad commodity indices and oil futures have correlations as high as 70% at the six-month horizon and beyond with headline CPI, PCE, and PPI. Broad stock indices can also provide a good hedge for headline inflation, although much of the correlation

8 If h ˃ τ some rollover will be necessary. We sidestep this more-complicated case in the interest of space. 9 The calculation for TIPS abstracts from the deflation floor and the indexation lag.

Inflation hedging products 

481

is driven by energy-related stocks.10 For similar reasons, EM currencies tied to energy prices also provide some protection. Contrary to conventional wisdom, we do not find any ability of gold to hedge headline inflation over the post-1999 sample. Finally, real estate, as proxied by the Case-Shiller index, does a surprisingly good job of hedging consumer prices at longer horizons, while the Wilshire REIT index effectively hedges the PPI.11 Generally speaking, returns on longer-term nominal bonds are negatively correlated with headline inflation, so that their inclusion in an investor’s portfolio increases the exposure to inflation risk, rather than hedging it. This is true whether the bonds are held to maturity, continually rolled over to constant maturities, or held for an intermediate period.12 On the other hand, three-month Treasury bills (T-bills) provide relatively good protection against headline inflation because these rates rise when monetary policy tightens. Further, as we discuss in Section 21.4, the inflation-hedging ability of nominal bonds has improved over the last decade, as disinflation/deflation risk has dominated inflation risk until very recently. Finally, short- and medium-term TIPS have performed quite well at protecting against headline CPI over short investment horizons (from one month to one year). At those same horizons, also the ten-year TIPS has provided some protection against headline PCE and PPI, with correlations higher than those for the two- and five-year TIPS at horizons between six months and two years. 21.3.2.2 Consumer and producer inflation components While there are multiple attractive strategies for hedging non-core inflation, the prospects are somewhat dimmer when it comes to core. At horizons of less than a year, few of the assets we consider provide good protection. (This is arguably not a very serious problem, however, because core inflation displays very little volatility at short horizons.) One exception is TIPS. The ten-year TIPS returns from the three-month to the one-year horizon have correlations of about 0.3–0.4 with core CPI, while the two-year TIPS has a similar size correlation at the onemonth horizon. Correlations with core PCE are also positive but a bit smaller. The correlations with core PPI are mostly negative, however. At longer horizons, there are substantial differences across the different core indices. Core PCE behaves somewhat similarly to headline PCE. It is correlated with the broad commodity indices and oil futures, certain stock-market sectors including oil and gas, and real estate. However, in all of these cases the correlations are at most around 60%, which is significantly lower than the best-performing assets for headline inflation. For core CPI, on the other hand, the only assets that provide some hedging value are the two- and five-year nominal bonds rolling returns, some of the Fama-French factors, and the Case-Schiller price index.13 Unlike 10 The strong correspondence between energy sector stocks and headline CPI is consistent with Ang et al. (2012) and Parikh et al. (2019). However, those studies also find that technology stocks are important. In contrast, we find only weak correlations between headline inflation and the semiconductor and telecommunications sectors and significant negative correlations with the software sector at longer horizons. 11 We caution that the results involving the Case-Shiller index in this sample are strongly influenced by the run-up to the 2008 housing crisis and its aftermath. 12 In principle, a strategy of shorting nominal bonds provides protection against headline inflation, although such strategies can be costly to implement. 13 It is perhaps unsurprising that house prices perform well with core CPI as housing services constitute a large percentage of the core basket.

482  Research handbook of financial markets

with headline inflation, broad commodity indices and oil futures do not perform well with core CPI. Gold and most stock-market indices are significantly negatively correlated with core CPI.14 To hedge core PPI is even harder. Only the 30-year nominal bond rolling returns and the Case-Shiller price index offer some protection. Decomposed somewhat differently, the nondurable (ND) components of the CPI are dominated by food and energy, so their results are similar to the headline CPI correlations discussed above. Broad commodity indices and oil futures provide an almost perfect hedge against ND CPI at the one-year horizon and beyond. Moreover, energy prices have high passthrough to the cost of durable goods, so CPI durables are also highly correlated with broad commodity indices and oil futures. Stock market sectors such as metals-mining, financials, and insurance also perform well with durables. In contrast, very few assets provide a decent hedge for CPI Services. Only the two-year nominal bond rolling returns, T-bills, and the Case-Shiller price index display positive correlation larger than 30%. (At very long horizons, the S&P Oil and Gas Exploration and Production sector returns and two-year future on WTI also have correlations around 30%.) Thus, the weak correlations noted for core CPI inflation stems from the lack of a good hedge for prices in the service sector. Further, also TIPS can provide a decent protection against CPI services. In particular, the two-year TIPS works well at the three-month horizon, the five-year TIPS at the six-month horizon, and the ten-year TIPS for horizons longer than six months. Not surprisingly, strategies of holding nominal bonds for long periods, including to maturity, almost always generate additional exposure to inflation, rather than providing a hedge. Across bond maturities and holding periods, the correlations with various components of consumer and producer inflation range from slightly positive to –40%. The exception is CPI services, where we find that ten-year bonds held to maturity have a positive 50% correlation. 21.3.2.3 Wages and house prices One reason that hedging service prices is difficult seems to be that a large fraction of those prices reflect labor costs. We find few good hedges for wage inflation. Indeed, most of the asset returns we consider display a negative correlation with average hourly earnings. However, these are also generally small in magnitude, so that even potential short positions would not be successful in hedging wages. The main exception is rolling returns in shorter-term nominal Treasuries. T-bills and two-year bonds have correlations with average hourly earnings of more than 50% at horizons beyond one year, and ten-year bonds have a 34% correlation if held to maturity. Further, also TIPS provide some protection, with the two-year TIPS hedging relatively well at the three-month horizon, the five-year TIPS at the six-month horizon, and the ten-year TIPS at the ten-year horizon.15

14 The opposite signs on the correlation of the stock market with headline and core inflation are roughly consistent with the findings in FLR. Interestingly, although the overall stock market is negatively exposed to core inflation, the “robust-minus-weak” and “conservative-minus-aggressive” Fama-French factors seem to provide good protection at longer horizons, perhaps suggesting that profitable and conservative firms are more resilient to inflation. However, this result is not robust to the longer sample discussed below. 15 The Case-Shiller index also appears to provide a good hedge for wages at long horizons, but this result may be spurious since it does not hold for our other measure of real estate prices (the Wilshire REIT) and there is no obvious economic reason that it should be true.

Inflation hedging products 

483

Finally, although we have discussed real estate as a potential hedging instrument (and, indeed, we have shown that it performs well as a hedge in many cases), one might also want to hedge real-estate costs themselves. Our results using the Case-Shiller index show that there are a variety of ways of doing this successfully at horizons of one year and longer. Most components of the stock market, including the SMB and HML factors, as well as most commodities and some currencies are strongly correlated with house prices at these horizons.16 Rolling returns on longer-term bonds, on the other hand, display strong negative correlations. 21.3.2.4 Results for the longer sample To examine the stability of the above results over time, we re-estimate the models using data since 1972, where possible. One interesting feature of the longer sample is that measures of headline and core inflation are considerably more highly correlated than in the 1999 sample.17 This means that they are more likely to be hedged well by the same set of instruments. Most notably, commodities—in particular oil—do a better job of hedging core inflation over the longer sample period. This is consistent with oil’s large role in driving business-cycle fluctuations throughout the 1970s and is quite different from what is emphasized in FLR. Gold and silver perform quite well against most inflation measures in the post-1972 sample, suggesting that those commodities’ inflation-hedging abilities must have been particularly good in the pre-1999 period. This behavior may have contributed to the common perception that precious metals are robust inflation hedges. However, as we showed above, that property seems to have disappeared over the last 20 years. In contrast, the two currencies that we can track do worse (against both headline or core) over the longer sample period. We continue to find relatively few possibilities for hedging wage inflation in the longer sample. Unlike in the post-1999 sample, average hourly earnings are positively correlated with commodity prices, but, except at very long horizons, those correlations are still quite modest. 21.3.3 Summing Up Summarizing the implications of the above results for investor portfolios, assets such as shortterm nominal bonds and the Case-Shiller price index, which we find display sizable correlations with CPI headline and hourly earnings, would be a good hedge for household balance sheets. Those same assets would also provide a good inflation hedge for firms in the services sector, as they hedge wage and CPI services inflation quite similarly. Meanwhile, very robust inflation-hedging products, such as broad commodity indices and oil futures, would protect against input and output inflation in most non-services firms’ balance sheet. Finally, investors mostly focused on hedging core inflation might want to use TIPS over shorter horizons, but the two- and five-year nominal bonds rolling returns, some of the Fama-French factors, and the Case-Shiller price index for longer horizons investments.

16 Of course, all of these results apply to national house prices at an aggregate level; since local housing markets are very idiosyncratic, any hedging strategy based on financial instruments is unlikely to offer much protection for an individual homeowner. 17 For example, at the five-year horizon, core and headline CPI have a correlation of 0.97 in the post1972 sample, compared to just 0.55 in the post-1999 sample.

484  Research handbook of financial markets

21.4 COST OF INFLATION HEDGES: THE INFLATION RISK PREMIUM In this section, we consider the cost of hedging inflation, that is, the returns investors are willing to forgo to hold inflation-hedging products, also known as inflation risk premium (IRP). For this purpose, we have to consider the price of inflation risk embedded in various asset classes. While it is natural to start from the IRP implied by the nominal term structure, as noarbitrage dynamic term-structure models (DTSMs) have become increasingly more sophisticated and deliver estimates of the IRP that can adjust to different inflation regimes, the literature has provided new insights in the price of inflation risk embedded in the stock market (Boons et al., 2020) and in other real assets (FLR). Recent findings indicate that, similarly to the IRP implied by DTSMs, the inflation risk priced in stock returns is strongly time-varying and sometimes even changes sign. Just as there is no one-size-fits-all strategy for hedging inflation, there is no unique premium for inflation risk, even within a given asset class. Each type of inflation commands its own IRP. In particular, across the various real assets examined in Sections 21.2 and 21.3, the price of core inflation risk is negative, while the price of energy inflation risk is positive but statistically insignificant. This implies that it is only hedging against core inflation that is costly. The magnitude of the IRP for other types of inflation—including producer prices, housing, and wages—remains an open empirical question. 21.4.1 Cost of Inflation Risk in the Nominal Term Structure We start by focusing on estimates of the IRP implied by state-of-the-art DTSMs, especially those augmented with data on inflation and inflation hedges.18 In these models, the IRP at any maturity τ is obtained by subtracting from the nominal yield the real yield and expected inflation:



é æ M R CPI t ö ù Covt ç t +Rt , ê ÷ ú 1 è M t CPI t + t ø ú + J . (21.8) IRPt ,t = ytN,t - ytR,t - pt ,t = - log ê1 + t ,t ê æ M tR+ t ö æ CPI t ö ú t ê Et ç R ÷ Et ç ÷ú êë è M t ø è CPI t + t ø úû

The last two terms indicate that in DTSMs the IRP is given by the negative covariance between the real pricing kernel and inflation and a Jensen’s inequality term, which is fairly small and therefore ignored from now on. Under the simplifying assumption that the real pricing kernel can be approximated by the real yield (Campbell et al., 2017), the IRP can be rewritten as follows:

(

)

( )

1 IRPt ,t » - log é1 + Covt ytR,t , pt ,t / Et ytR,t Et ( pt ,t ) ù , (21.9) ë û t

18 See for example, Chernov and Mueller, 2012; Haubrich et al., 2012; Fleckenstein et al., 2017; D’Amico et al. 2018; Ajello et al., 2020; Breach et al., 2020.

Inflation hedging products 

485

which implies that in DTSMs the sign and size of the IRP is determined by the sign and size of the conditional covariance between the real economy and inflation, also known as the nominal-real covariance in the literature.19 Empirical evidence suggests that this covariance has been changing over time (e.g., Campbell et  al., 2017; Gourio & Ngo, 2020), reflecting the changing hedging characteristics of nominal bonds, especially at the ZLB. Specifically, nominal bonds become more attractive when this covariance is positive, as they pay off in inflationary or deflationary scenarios. They therefore command a very low or even negative IRP (Kitsul & Wright, 2013). Based on the evidence for the nominal-real covariance, it should not be surprising that the IRP estimated in very flexible DTSMs, such as those of D’Amico et al. (2018) and Breach et al. (2020), displays a time-varying size and at times flips sign. DTSMs can encompass the diverse dynamics of the IRP over extreme episodes like the early 1980s, characterized by high actual and expected inflation as well as high inflation uncertainty, and the post-2008 period, characterized by low inflation (and mild deflation) as well as very low expected inflation and inflation uncertainty. The early 1980s and post-2008 years are periods during which investors’ perceptions of inflation risk evolved from high inflation scares (Goodfriend, 1993) to deflation fear (Kitsul & Wright, 2013). Figure 21.1 plots the time series of the five- and ten-year IRP estimates implied by the Breach et  al. (2020) model. The estimates are larger and positive in the earlier part of the sample, with peaks of about 1 and 1.4 percentage points, respectively, in the early 1980s; in contrast, they are quite small and often negative after 2008. This is the “deflationary bias” period, during which inflation kept falling rather than stabilizing around the Federal Reserve’s desired target of 2% (Bianchi et al., 2021). These negative estimates of the IRP have important implications for the inflation-hedging properties of nominal bonds. As shown in Fleckenstein et al. (2017) the average probability of deflation perceived by investors in the 2009–2015 sample period was about 19%, 14%, 6%, and 1.4%, at the one-, two-, five-, and ten-year horizon, respectively. But, at the shorter horizons, this probability sometimes went above 30%. These are all episodes in which nominal bonds are a very good hedge as they offer high real returns when investors need them the most. These are also times in which stocks are very poor hedges. For example, Fleckenstein et al. (2017) find a strong negative relation between deflation risk and the stock market for all horizons. The evidence reported in the next section expands on this finding. 21.4.2 Cost of Inflation Risk in Other Asset Classes In the consumption-based equilibrium model of Boons et al. (2020), each shock drives a separate component of the real pricing kernel (mt +1 ); hence, in considering the risk premium of asset i, it is possible to zoom in on the IRP while ignoring the other components (e.g., consumption shock, η):

-Covt ( mt +1, ri,t +1 ) = l u,t b P ,i,t + l h,t bC ,i,t +  (21.10)

19 A similar result holds in Gourio and Ngo, 2020, where the real pricing kernel is approximated by stock-market returns.

486  Research handbook of financial markets 1.5 10yr 5yr 1

0.5

0

–0.5

1987

1993

1998

2004 Date

2009

2015

2020

Source:   Breach, D’Amico, and Orphanides (2020).

Figure 21.1  Estimates of the five- and ten-year inflation risk premium where ri,t +1 is the asset return, l u,t is the price of inflation risk, and bP,i,t is the quantity of inflation risk, given by the asset’s inflation beta. In their empirical analysis, the inflation beta is defined as the slope coefficient in a rolling regression of returns on inflation shocks (up,t +1):

b P ,i , t =

Covt ( uP ,t +1, ri,t +1 ) ; (21.11) var ( uP ,t +1 )

and the price of inflation risk is obtained running monthly cross-sectional regressions of individual stock returns on lagged inflation betas, controlling for capitalization, book-to-market, and momentum:

ri,t +1 = l 0 + l u,tbP ,i,t + controls (21.12)

Since bP,i,t is just a different approximation of the nominal-real covariance shown in Equation 21.9 (i.e., the real pricing kernel is approximated by stock-market returns rather than real yields), the variation in stocks’ IRP is driven by the same conceptual object that drives the IRP embedded in nominal bonds. Boons et al. (2020) find that over the 1962–2014 period, for a one-standard-deviation increase in this covariance, the IRP increases by about 3% to 4%, with larger effects as the holding period grows from one to 12 months. In particular, from the 1960s to the early 2000s, the IRP is strongly statistically significant and almost monotonically decreasing in inflation betas, indicating that investors are willing to lose about a 7% return to hold the stock portfolio with the highest inflation beta. Post-2002, the IRP becomes positive,

Inflation hedging products 

487

increasing rather than decreasing in inflation betas, indicating that investors require extra compensation to hold stock portfolios highly exposed to inflation. This is consistent with the reversal in the nominal-real covariance from negative to positive since the early 2000s. In other words, during periods characterized by disinflation or deflation risk, stocks are not attractive as a hedge anymore, while nominal bonds become extremely attractive, as demonstrated by their very low or negative IRP estimates. Turning to the IRP estimates for other real assets, FLR also runs Fama-McBeth crosssectional regressions of average returns onto asset inflation betas. Across all 38 portfolios that include stock industry portfolios, Treasury maturity-sorted portfolios, corporate bond maturity-sorted portfolios, currency carry portfolios, commodities, REITs, and international stocks, the price of headline CPI inflation risk is not statistically significant, the price of energy inflation risk is positive but also insignificant, and the price of core inflation risk is negative and significant (for nearly all assets). Their estimates imply that if an asset increases one unit of exposure to core inflation, investors require 1.07% of excess returns per annum.

21.5 CONCLUSION This chapter has discussed the possibilities for hedging inflation in U.S. financial markets, reviewing the literature and providing some new evidence. We emphasize that there is no one-size-fits-all approach to inflation hedging; the optimal hedge depends on the particular types of prices that an investor is exposed to and at which horizons. Over the last 20 years, we find that commodities and commodity-related stocks and currencies are generally successful in hedging headline consumer inflation, but this mostly seems to reflect their significantly positive relation with energy prices. Hedging core inflation is harder. At horizons of less than a year, there is little protection available, except for TIPS. At longer horizons, short-term nominal bonds and real estate provide a decent hedge, and there is some evidence supporting certain stock-market strategies. Core producer prices and wages are the most difficult types of inflation to hedge, although real estate and short-term nominal bonds provide some protection. House prices can be effectively hedged through REITs and other instruments that provide broad exposure to this sector. Our empirical exercise identifies short-term nominal bonds and real estate as reasonably good hedges for employment costs. And, since wages are a particularly large component of the cost of non-housing services, the same instruments do a decent job hedging service inflation. The relative attractiveness of different inflation hedging instruments evolves over time, depending on the growth-inflation regime. This is reflected in the IRP, which measures the cost of inflation hedging. IRPs have typically been measured from no-arbitrage dynamic pricing models of the nominal and real term structure, but methods have recently been developed to extract them from other asset classes as well. The size and sign of the IRP are determined by the covariance between inflation and the real side of the economy, and as this covariance has transitioned from mostly negative to mostly positive over the last 20 years, nominal bonds have commanded a negative IRP and stocks a positive IRP. After the global financial crisis, inflation risk evolved into disinflation risk, dictating a negative IRP for nominal bonds and a positive IRP for equities. It remains to be seen whether these trends will continue to hold following the volatile growth and inflation experience of the COVID period.

488  Research handbook of financial markets

REFERENCES Ajello, A., Benzoni, L., & Chyruk, O. (2020). Core and ’crust’: Consumer prices and the term structure of interest rates. Review of Financial Studies, 33(8), 3719–3765. Ang, A., Brière, M., & Signori, O. (2012). Inflation and individual equities. Financial Analysts Journal, 68(4), 36–55. Bakshi, G., Gao, X., & Rossi, A. G. (2019). Understanding the sources of risk underlying the cross section of commodity returns. Management Science, 65(2), 619–641. Bampinas, G., & Panagiotidis, T. (2016). Hedging inflation with individual stocks: A long-run portfolio analysis. North American Journal of Economics and Finance, 37, 374–392. Bianchi, F., Melosi, L., & Rottner, M. (2021). Hitting the elusive inflation target. Journal of Monetary Economics, 124, 107–122. Boons, M., Duarte, F., de Roon, F., & Szymanowska, M. (2020). Time-varying inflation risk and stock returns. Journal of Financial Economics, 136(2), 444–470. Boudoukh, J., & Richardson, M. (1993). Stock returns and inflation: A long-horizon perspective. American Economic Review, 83(5), 1346–1355. Breach, T., D’Amico, S., & Orphanides, A. (2020). The term structure and inflation uncertainty. Journal of Financial Economics, 138(2), 388–414. Campbell, J. Y., Sunderam, A., & Viceira, L. M. (2017). Inflation bets or deflation hedges? The changing risks of nominal bonds. Critical Finance Review, 6(2), 263–301. Chernov, M., & Mueller, P. (2012). The term structure of inflation expectations. Journal of Financial Economics, 106(2), 367–394. Christensen, J. H. E., Lopez, J. A., & Rudebusch, G. D. (2012). Extracting deflation probability forecasts from treasury yields. International Journal of Central Banking, 8(4), 21–60. D’Amico, S., Kim, D., & Wei, M. (2018). Tips from TIPS: The information content of Treasury inflationprotected security prices. Journal of Financial and Quantitative Analysis, 53(1), 395–436. D’Amico, S., & King, T. B. (2023). One asset does not fit all: Inflation hedging by index and horizon. Federal Reserve Bank of Chicago Working Paper 2023-08. David, A., & Veronesi, P. (2013). What ties return volatilities to price valuations and fundamentals? Journal of Political Economy, 121(4), 682–746. Downing, C. T., Longstaff, F. A., & Rierson, M. A. (2012). Inflation tracking portfolios. National Bureau of Economic Research, w18135. Erb, C., & Harvey, C. R. (2006). The strategical and tactical value of commodity futures. Financial Analysts Journal, 62(2), 69–97. Fang, X., Liu, Y., & Roussanov, N. (2021). Getting to the core: Inflation risks within and across asset classes [Working paper], May. Fleckenstein, M., Longstaff, F. A., & Lustig, H. (2017). Deflation risk. Review of Financial Studies, 30(8), 2719–2760. Fleming, M. J., & Sporn, J. R. (2013). Trading activity and price transparency in the inflation swap market. FRBNY: Economic Policy Review, May. Goodfriend, M. (1993). Interest rate policy and the inflation scare problem: 1979–1992. Reserve Bank of Richmond Economic Quarterly, 79(1). Gorton, G., & Rouwenhorst, K. G. (2006). Facts and fantasies about commodity futures. Financial Analysts Journal, 62(2), 47–68. Gourio, F., & Ngo, P. (2020). Risk premia at the ZLB: A macroeconomic interpretation Federal Reserve Bank of Chicago Working Paper 2020-01. Grishchenko, O. V., Vanden, J., & Zhang, J. (2016). The information content of the embedded deflation option in TIPS. Journal of Banking and Finance, 65, 1–26. Gürkaynak, R. S., Sack, B., & Wright, J. H. (2007). The U.S. Treasury yield curve: 1961 to the present. Journal of Monetary Economics, 54(8), 2291–2304. Haubrich, J., Pennacchi, G., & Ritchken, P. (2012). Inflation expectations, real rates, and risk premia: Evidence from inflation swaps. Review of Finacial Studies, 25(5), 1588–1629. Kat, H. M., & Oomen, R. C. A. (2006). What every investor should know about commodities (Part II). Alternative Investment Research Center Working Paper, No.0033.

Inflation hedging products 

489

Katz, M., Lustig, H., & Nielsen, L. (2017). Are stocks real assets? Sticky discount rates in stock markets. Review of Financial Studies, 30(2), 539–587. Kerkhof, J. (2005). Inflation derivatives explained. Fixed income quantitative research. Lehman Brothers (July), 1–80. Kitsul, Y., & Wright, J. H. (2013). The economics of options-implied inflation probability density functions. Journal of Financial Economics, 110(3), 696–711. Leombroni, M., Piazzesi, M., Schneider, M., & Rogers, C. (2020). Inflation and the price of real assets [Technical report]. National Bureau of Economic Research. Lustig, H., Roussanov, N., & Verdelhan, A. (2011). Common risk factors in currency markets. Review of Financial Studies, 24(11), 3731–3777. Lustig, H., Roussanov, N., & Verdelhan, A. (2014). Countercyclical currency risk premia. Journal of Financial Economics, 111(3), 527–553. Menkhoff, L., Sarno, L., Schmeling, M., & Schrimpf, A. (2017). Currency value. Review of Financial Studies, 30(2), 416–441. Parikh, H., Malladi, R. K., & Fabozzi, F. J. (2019). Preparing for higher inflation: Portfolio solutions using U.S. Equities. Review of Financial Economics, 542–554. Park, J., Mullineaux, D. J., & Chew, I. K. (1990). Are REITs inflation hedges? Journal of Real Estate Finance and Economics, 3(1), 91–103. Verdelhan, A. (2018). The share of systematic variation in bilateral exchange rates. Journal of Finance, 73(1), 375–418.

22. Futures and options1 Refet S. Gürkaynak and Jonathan H. Wright

22.1 INTRODUCTION AND HISTORY1 Futures and options are important derivative contracts that serve as means of transferring risk. A futures contract is an agreement to buy or sell an asset at a future date, but for a price that is agreed today. The money, however, does not change hands until the future delivery date. Futures contracts are traded on exchanges and are standardized in terms of the quantity, time of delivery, and other terms. This makes trading very liquid. The party who is agreeing to buy the asset is said to be “long,” and the party that agrees to sell it is “short.” The total number of contracts outstanding is called the open interest. Similar contracts that trade over the counter and without standardization are known as forward contracts. Depending on the contract, futures may be either physically settled, meaning that payment is made for the asset, or cash settled, meaning that the difference between purchase and settlement price will be credited to, or debited from, the investor’s account. An investor can buy an asset for immediate delivery, referred to as the spot price, or in the futures market. The situation in which futures prices are below spot prices is known as backwardation; the opposite and more common case is known as contango. These terms were coined in 19th century England. An options contract gives the holder the right, but not the obligation, to buy or sell an asset at a future date at a price that is agreed today. The holder of the options contract can, however, let it expire unused. The holder of the contract has to pay the writer of the contract a premium, and that is collected at the time the contract is written. An option to let the holder buy the asset at a fixed price is a call option; one to let the holder sell the asset at a fixed price is a put option. Some options can be exercised only on the expiration date; these are called European options. Other options can be exercised at any time up to and including the expiration date; these are called American options. Options are likewise important tools for risk management and transfer. Futures and options have the feature that it is easy for an investor to bet on the price of the asset falling; without derivatives, this requires shorting the asset—that is borrowing the asset and selling it—which is often possible, but difficult. Futures and options are also useful to economists because they give information that may allow us to reverse engineer agents’ beliefs.

1 The authors thank Bin Wei for very helpful comments on an earlier draft and to Mahmut Sefa Ipek and Senem Turan for research assistance.

490

Futures and options 

491

The fundamental idea of futures and options have been around for a very long time. The Babylonian King Hammurabi set up a legal code that essentially included both futures and options contracts around 1750 BC. The Code of Hammurabi states that: If anyone owe a debt for a loan, and a storm prostrates the grain, or the harvest fail, or the grain does not grow for lack of water; in that year he need not give his creditor any grain, he washes his debttablet in water and pays no rent for the year.

This amounts to what we think of today as a “put option” (Whaley, 2006). The first organized futures exchange was a rice futures market in Osaka, Japan circa 1730. The Chicago Board of Trade (CBOT) was formed in 1848, and in 1865 introduced standardized futures contracts including central clearing and margin requirements, that are central to modern derivatives markets. For most of the history of futures markets, trading was in physical commodities—agricultural commodities and metals, but in the 1970s trading in financial futures began, and financial futures are now the dominant part of futures trading.2 The last few decades have seen two more developments: concentration and switching to electronic trading. The Chicago Mercantile Exchange launched its Globex electronic trading platform in 1992. Although floor trading continues, all the major exchanges now have electronic trading systems as well. The Chicago Board of Trade, Chicago Mercantile Exchange and New York Mercantile Exchange merged to form the Chicago Mercantile Exchange Group which is the largest derivatives exchange in the world. While futures exchanges go back centuries, options were traded over the counter until more recently. The first exchange to trade options was the Chicago Board Options Exchange (CBOE) that opened in 1973 (Miller, 1986), but there are now large options exchanges around the world, and options are also traded on some stock exchanges. Exchange traded futures and options do not always last, some contracts are discontinued if sufficient open interest cannot be sustained, and some are even brought back to life later. Hegde (2004) provides the illustrative example of broiler (poultry, not kitchen appliance) futures. Economic derivatives on outcomes of macroeconomic data releases similarly briefly traded on the CME before the market for those contracts shuttered. Contracts for inflation that were introduced in 1980s were discontinued (Shiller, 1993).3 But over-the-counter trading in swaps and options written on inflation (Kitsul & Wright, 2013) have been more successful recently.4 Till (2015) provides a detailed discussion of examples of futures contracts that succeeded and those that failed.

2 In the late 19th century, so-called “bucket shops” began offering their own futures contracts with settlement prices based on the prices of CBOT contracts. CBOT eliminated the competition by successfully arguing that contracts where only cash settlement is possible, with no option for physical delivery, are equivalent to gambling and should be outlawed. That argument hindered the introduction of financial index futures, such as futures on S&P 500, and many other contracts such as futures on inflation, where the delivery of the underlying is either impossible or entails prohibitive transaction costs, until the 1980s. Millo (2007) provides details. 3 There was another short-lived but unsuccessful attempt in 2004. 4 In any derivatives market, there naturally have to be market participants on both sides and typically few are interested in betting on low inflation. The existence of Treasury Inflation Protected Securities has helped by creating a natural provider of inflation protection.

492  Research handbook of financial markets

22.2 TYPES OF FUTURES AND OPTIONS TRADERS A basic distinction is often made between participants in futures and options markets who are using the market to protect some other business interest, known as hedgers or commercial traders, and participants who are just trying to make profits, known as speculators or non-commercial traders. For example, a producer of agricultural commodities who uses futures markets to lock in a price would be considered a hedger, whereas a hedge fund would be a considered a speculator. Speculators serve an important role in futures and options markets, because the hedgers will often tend to be one side of the market. Speculators often take the other side of these bets and provide liquidity in normal times. Unlike hedgers, who are insuring existing risk and on net do not have uncovered positions, speculators often take on risk and exacerbate market swings by short covering (buying the underlying when prices go up while in a short position) or reversing their positions to mitigate their exposure. The Commodity and Futures Trading Commission (CFTC) requires large participants in futures markets to report their position and whether their predominant motive is hedging or speculating and the aggregated positions are published weekly. In recent years, the CFTC has produced two other classifications of large traders. One is called the supplemental report, and this splits out index funds (a portfolio that is constructed to track a financial index, such as the S&P 500) as a separate category. Index funds might be either hedgers or speculators. The other is called the disaggregated report and includes four categories: producers, money funds, swap dealers, and others. The producers are hedgers, the money funds are speculators and the other two groups could be either hedgers or speculators. 5 The distinctions between these categories are not always clear, but a large literature has evolved studying the relationship between positions of these different trader types and price changes (see, for example, Houthakker, 1957; Chang, 1985 and Kang et al., 2020).

22.3 OPEN OUTCRY AND ELECTRONIC TRADING The traditional form of trading in futures and options is open outcry, where there is a trading pit in which traders for a particular type of futures execute trades either verbally or by hand signals. Some exchanges, like the Intercontinental Exchange have eliminated this floor trading entirely, while others still continue it, but electronic trading is now far more prevalent. Electronic trading is faster, cheaper in terms of spreads and other costs, less error prone, and it facilitates regulatory compliance. Nonetheless, large trades are also often executed away from electronic platforms. There are concerns that electronic trading may occasionally be destabilizing, such as in the 2010 “flash crash" which involved interactions between trading in cash and futures markets.

5 Note the often confusing distinction between hedgers, who hedge and hedge funds, who are usually speculators.

Futures and options 

493

22.4 CLEARING Participants in the futures market have little reason to be concerned with the creditworthiness of their counterparties as the exchange stands in the middle as the counterparty to both the buyer and the seller, and that is of course essential to the liquidity and functioning of futures markets. This function of the exchange is called “being the clearinghouse.” This means that contracts are guaranteed by the exchange/clearinghouse. Once a contract is made between a long party and a short party, this is immediately recast or novated as two separate contracts: one between the long party and the clearinghouse, and the other between the short party and the clearinghouse. This is of course risky for the clearinghouse, but they manage this risk carefully. Investors are required to post an initial margin with the clearinghouse when they enter into a trade, and if the price moves against them they must post additional maintenance margins, or else the position will be liquidated. In addition, all trades must go through a clearing member who is responsible for any default by the end customer. Clearing members are large financial institutions that have been carefully vetted and have to post their own margin. The clearing members are, moreover, required to keep client funds in segregated accounts. These clearing safeguards are very robust and have multiple layers of safety. An extreme example is that of MF Global, which was a clearing member of the CME that filed for bankruptcy in 2011, and turned out to have breached the rules about keeping client funds segregated, but even this did not cause any broader systemic problems. Indeed we are not aware of any futures exchange ever failing. Futures trading rules allow for netting, meaning that if a trader has both long and short positions, these can be cancelled out. Hence, if a trader has a long position and wants to exit this, they can simply enter into an offsetting short position. Clearing services in the options market are provided by the Options Clearing Corporation (OCC) for options traded on CBOE and other exchanges.

22.5 REGULATION In the US, futures and options are regulated both by the exchanges themselves, and by the state and federal governments. Since 1974, the Commodity and Futures Trading Commission (CFTC) has been the regulatory agency charged with regulation of futures and options on commodities. CFTC and the Securities and Exchange Commission (SEC) share regulatory responsibilities for derivatives written on underlying financial securities and indices. Both the exchanges and the CFTC seek to prevent market manipulation. For example, a potential form of market manipulation is a short squeeze, also known as cornering. Where there is a futures contract requiring physical delivery, a market participant might take a very large long position and then simply hold the contract to expiration and require delivery, while also holding a large amount of the commodity. The counterparties who have taken short positions will then be forced to buy large amounts of the commodity, driving its price up and generating profits for the market participant with the large long position. There was a famous example of such a short squeeze in the 1970 when the Hunt brothers took large long positions in silver futures while also accumulating physical silver. The price of silver was driven up from $5 an ounce in 1978 to nearly $50 an ounce in 1980. The exchanges imposed regulations

494  Research handbook of financial markets

to try to curb the short squeeze, but it only came to an end when the CFTC banned all purchases of long silver futures contracts, only allowing liquidation.6 The aftermath of the 2008 financial crisis brought new regulations into place. In the US, this was mainly through the Dodd-Frank Act. The Financial Stability Oversight Council can designate clearinghouses as systemically important financial market utilities (SIFMUs), that are subject to enhanced regulation. The CME, the Intercontinental Exchange, and the OCC have all received this designation. These SIFMUs also have access to central bank credit in extreme circumstances, with the approval of the Treasury secretary.

22.6 PRICING AND USAGE The pricing of a futures contract depends heavily on whether the underlying asset is storable or not. Many futures contracts are written based on assets that can be stored at a trivial cost, such as financial assets and precious metals. For these assets, there is a tight arbitrage relationship between the futures price and the price for immediate delivery, or the spot price, known as spot-futures parity. This states that if F is the futures price for delivery T periods hence, S is the spot price today, r is the continuously compounded interest rate and d is the dividend yield of the asset (0 if the asset pays no dividends), then:

F = Se(r - d )T (22.1)

The logic is simple. There are two ways of buying the asset: either in the spot market today, or in the futures market for delivery in T periods. Both are equivalent, except that buying it in the spot market requires immediate payment whereas the futures contract does not require payment until delivery. And, buying the asset in the spot market entitles the holder to any dividends (as in the case of a stock) immediately, whereas buying in the futures market will lead the investor to miss out on these dividends. Spot futures parity makes adjustments for these two effects. In the case of a non-dividend paying asset with positive interest rates, the futures price should be above the spot price. The basis is the name given to the difference between the futures and spot price, and as time approaches the maturity, the basis should converge to zero. This arbitrage relationship applies to stock index futures relative to the underlying index. Violations of the relationship do exist, but program trading quickly identifies and exploits them. As a result of index arbitrage, mispricings are small and do not last long. Some futures contracts are written on things that are not storable, such as the weather or electricity, in which case there is no arbitrage relationship between spot and futures prices. Other futures contracts are written on assets that can be stored, but are quite costly to store,

6 A similar short squeeze in the 1950s led to passing of a 1958 law in the US that bans futures contracts on onions. The law was later amended to also ban futures on motion picture box office receipts in 2010. The authors of this chapter do not know what connects these two markets. The ban on onion futures has served as a natural experiment to study whether futures trading dampens (Working, 1960; Gray, 1963) or exacerbates (Johnson, 1973) price fluctuations in the underlying. The question remains unsettled.

Futures and options 

495

such as crude oil. In this case, the absence of arbitrage instead implies bounds on the relationship between spot and futures prices. Arbitrage relationships also apply to options prices. A put and a call option can be combined to a form a portfolio that is the equivalent of the underlying asset and a risk-free bond. Using this reasoning gives the relationship:

P + S = C + Ke -rT (22.2)

where P and C are the prices of put and call options, respectively, at the same strike price, K, and maturity, T, where S is the price of the underlying asset and r is the continuously compounded risk-free interest rate. Stoll (1969) is credited with the first modern treatment of put-call parity, although Knoll (2008) gives examples of the idea going much further back. Options pricing is based on the insight that under the condition that investors know the volatility of the underyling asset ex ante, the options are redundant securities, and can be perfectly replicated by a continuously rebalanced portfolio of the underlying asset and a risk free bond. This insight underlies the famous pricing equation of Black and Scholes (1973) who assume that innovations to the asset price are lognormally distributed and show that a European call option expiring in T periods with a strike price of K on an asset with a spot price of S and a per period standard deviation of σ is:



æ æSö æ æ æSö æ s2 ö ö s2 ö ç ln ç ÷ + ç r - ÷ T ç ln ç ÷ + ç r + ÷T ÷ 2 ø 2 ø ÷ èKø è èKø è C = SF ç - Ke -rT F ç 1/2 1/2 ç ç ÷ sT sT ç ç ÷ ç ç ÷ è è ø

ö ÷ ÷ (22.3) ÷ ÷ ÷ ø

where r is the risk-free rate and Φ is the standard normal cumulative distribution function. This equation does not make any assumption about the preferences of investors, other than that they do not leave arbitrage opportunities on the table. But the assumptions of BlackScholes rule out many real-world features of the underyling asset including nonnormality, stochastic volatility, and jumps. Subsequent work has considered pricing incorporating these features. For example, Heston (1993) provides the extension to the case of stochastic volatility and Duffie et al. (2000) price options in the case of a jump diffusion. Carr and Wu (2004) consider a very general framework for option pricing including jumps, stochastic volatility, and negative correlation between returns and volatility. The Black-Scholes implied volatility of an option is the value of σ that would set the BlackScholes price equal to the actual observed price. Under the Black-Scholes assumptions, this should be the same no matter what the strike price of the option is. In practice, however, options with low strike prices have higher implied volatilities than those with higher strike prices, a feature known as the volatility smirk. A key result with options is that the second derivative of the price of a European call option with respect to the strike price is equal to the probability density function of the underyling asset price at the expiration date, if agents are risk neutral (Breeden & Litzenberger, 1978),

496  Research handbook of financial markets 6

3 month forward 5

4

3

2

1

0 1.2

1.3

1.4

1.5

1.6

1.7

Note:   Probability density function for the pound to US dollar exchange rate in three months (pounds per dollar), as of the day of the Brexit referendum, June 23, 2016 (before voting had concluded). This is computed by converting call option prices to implied volatilities, fitting a smoothed curve to these volatilities, converting back to call options prices and taking the second derivative of these numerically. The vertical line shows the three-month forward exchange rate. Source:  Bloomberg.

Figure 22.1  Probability density function for pound–dollar exchange rate in three months as of June 23, 2016 which we call the risk-neutral probability density function.7 Given a set of options at different strike prices, we can fit a smoothed curve to get the fitted options price at all strike prices, and then take the second derivative numerically to work out the risk-neutral probability density function.8 These can yield interesting insights that are not evident from implied volatilities alone. For example, the probability density for the pound–dollar exchange rate right before the June 2016 Brexit referendum, shown in Figure 22.1 was notably skewed in the direction of depreciation of the British pound. To measure risk aversion, the true subjective probabilities, the P-measure, and options-implied Q-measure probability densities can be compared. Under risk neutrality, they will be the same. 7 The risk-neutral probability density functions are derived from prices assuming that agents are risk neutral. Even if agents are not risk neutral, the probabilities derived in this way will be nonnegative and integrate to one and so can be interpreted as probabilities, although they are not necessarily the actual probabilities perceived by the agents. The potentially distorted probabilities assuming risk-neutrality are referred to as Q-measure probabilities, while the actual physical probabilities are P-measure probabilities. 8 More technically, it is standard to convert the options prices to implied volatilities, smooth these, and then convert back to options prices (Shimko, 1993).

Futures and options 

497

If we think of a broad stock market index like the S&P 500 as a proxy for aggregate consumption, then the comparison of the P- and Q-measure densities give us a way of measuring a representative agent’s risk aversion (Aït-Sahalia & Lo, 2000; Bliss & Panigirtzoglou, 2004).

22.7 SPECIFIC FUTURES AND OPTIONS CONTRACTS AND THEIR USAGE IN THE LITERATURE Futures and options have three distinct uses in the literature. The first, for short maturity futures, is as a placeholder for the underlying. This is often because the standardized and exchange traded futures contracts have more accessible intraday prices, as in the case of oil prices where the nearest maturity futures price is often used as the oil price (Rosa, 2014; Chatrath et al., 2012), or because the trading hours of the futures contract are different from that of the underlying and measuring intraday reactions outside the trading hours of the underlying require resorting to the nearest maturity futures price. This latter reason makes stock futures usually employed in intraday studies of effects of macroeconomic data releases—such as the employment report—many of which are released at 8.30am Eastern Time, before the stock markets are open, while trading in stock futures are taking place (Faust & Wright, 2018; Andersen et al., 2007; Miao et al., 2014; Lahaye et al., 2011).9 The other two uses of futures and options in the academic literature are for the related tasks of pricing and information extraction. Pricing, briefly described earlier, relates futures and options prices to current and expected states of the world. This literature has to do with what the relevant variables are, and which function maps these variables to the derivatives’ prices. Information extraction takes this relationship and, essentially, inverts the function to go from the observed futures and options prices to unobserved expectations and uncertainty. This is where risk aversion and risk pricing becomes important, as risk pricing gives rise to term premia and prices are not only functions of (statistical) expected outcomes but the overall distribution of possible outcomes and their pricing. In a CAPM-type framework risk pricing will depend on marginal utility in different states of the world. The history of academic research into futures and options pricing goes back to mid-20th century. For futures, some of the early treatments that follow what we would now call asset pricing methods are by Working (1948; 1953) who discusses storage costs, basis, hedging demand, and backwardation and contango in futures prices. His focus was on grain futures, the dominant futures market at the time. Interest rate futures gained in importance in terms of research attention soon after, both because the market itself deepened and partly because absence of storage costs for Treasuries simplified the research question. Cox et al. (1981) summarized the state of the literature in interest rate futures pricing up to then and clarified the role of what is now called term premia, arising from interest rate uncertainty. Research (which goes back to Cootner, 1960) then began to focus on separating term premia from expected future spot prices (as well as tax effects and shorting costs) and that effort is ongoing. Fama and French (1987) is an early paper that articulates and tests the implications of storage-based futures pricing versus term premia-based ones. 9 Some of these papers also use bond futures for bond prices although the bond market is open, as the bond market is over the counter and individual bonds are less liquid than bond futures which makes the futures prices more reliable.

498  Research handbook of financial markets

22.7.1 Expectations and Risk Premia and Drawing Inferences from Futures and Options Prices Under risk neutrality, the futures price should be the expected future spot price, but with risk aversion the futures price can also incorporate a risk premium. Keynes (1930) considered risk premia in agricultural futures contracts and argued that it would generally be negative so that the futures price would be below the expected future spot price. His reasoning was that producers used the futures market to hedge against price declines, whereas consumers were too small to participate in the futures market and so the hedging demand was concentrated on one side of the market. But every short position needs a long counterparty, and often speculators are needed to take these long positions. Risk averse speculators will do so, but only if they get positive expected returns. Thus the hedging pressure view of Keynes (1930) would imply that backwardation is the usual configuration of futures prices. Hirshleifer (1990) reexamines this hedging pressure idea in a general equilibrium framework where producers face both price and quantity risk, and shows that futures prices can be either above or below the expected future spot price. It is very common for policymakers to use futures to attempt to infer the expected future path of an asset price. This is correct under an assumption of risk neutrality or if investors’ marginal utility is uncorrelated with the value of the asset, but not otherwise. A common way to assess whether futures prices are indeed expectations of future spot prices is to run a forecast efficiency regression of the form:

st + h = a + bft ,h + et + h (22.4)

where ft ,h is the futures price at time t for maturity h periods hence and st + h is the spot price at time t + h. Clearly for the futures price to be the expected future spot price, it is necessary that α = 0 and β = 1. Other variables that are in the information set at time t can be added to the right-hand side; if the futures price is the rational expectation of the futures spot price, then their coefficients should be zero. Chernenko et al. (2004) apply the test based on Equation 22.4 to a range of futures prices and in most cases reject the hypothesis that the futures price is equal to the expected future spot price, implying that there are risk premia. Moreover, β is significantly different from 1, implying that these risk premia are time-varying. But there are some exceptions to this. One is that for very near-term interest rate futures contracts, the hypothesis that α = 0 and β = 1 is not rejected. Another is that it is generally not rejected for energy prices, a conclusion also reached by Chinn and Coibion (2014) and Alquist and Kilian (2010). This lends some support to the idea of using futures prices to approximate market expectations of interest rates in the very near-term (but not further ahead), and also using it to approximate expectations of future oil prices. A caveat is that for α and β to be zero and one, respectively, is necessary but not sufficient for futures prices to be expected future spot prices. Indeed, Alquist and Kilian (2010) show that a simple random walk no change forecast gives better predictions of future crude oil prices than futures markets which implies that there must be time-varying risk premia in this market. And of course another caveat is that these tests have limited power and are dependent on the sample period. Gürkaynak et al. (2007) find a very small but statistically significant upward bias in even the shortest term interest rate futures contracts, viewed as expectations of future interest rates.

Futures and options 

499

A number of authors have studied the properties of futures returns. For example, Gorton and Rouwenhorst (2006) compute the average returns on an equally weighted index of commodity futures returns, taking careful account of collateral requirements, so that this in effect represents the excess returns. They show that it has similar risk and return properties to US equities, but is negatively correlated with equity returns which makes this asset very attractive from the point of view of the CAPM.10 Again, this is evidence that risk premia are generally important in the pricing of futures. 22.7.2 Some Specific Contracts 22.7.2.1  Interest rate futures Interest rate futures include Treasury futures which are physically settled in a rather complicated way. Futures traded on the CME include two-year, five-year, and ten-year Treasury notes and thirty-year bonds. Let us take the ten-year as an example. If the contract goes to delivery, the short side has to deliver any Treasury notes chosen by the short side with a face value of $100,000 and a remaining time to maturity of between 6.5 years and ten years. The price that they receive for this is the futures price multiplied by a conversion factor, plus accrued interest. The conversion factor is the price at which a particular coupon security with a face value of $1 would trade if its yield to maturity were 6 percent. Since yields are far below 6 percent and have been for a long time, it will generally be the shorter maturity securities that are cheapest to deliver (Livingston, 1984). The open interest in CME Treasury futures is enormous. For example, in recent years, the open interest in the ten-year Treasury future has averaged 3–4 million contracts with a face value of $300–400 billion. Physical delivery of that volume of Treasuries would be very difficult. Indeed, CME rules actually prevent traders from holding large long positions into delivery. Contracts are listed expiring in March, June, September, and December of each year, and most of the trading is in the front contract, except for the last month of the contract when most traders roll their positions into the next contract. Barth and Kahn (2021) document that in recent years Treasury futures have been somewhat overvalued relative to cash Treasuries, presenting an opportunity for a basis trade and documented that it was widely used by hedge funds from 2016 until the turmoil in Treasury markets in March 2020. Barth and Kahn (2021) and Schrimpf et al. (2020) both find that this basis trade might have contributed to the financial market disruptions in that month. There are also short-term interest rate futures, of which the most important are Eurodollar futures. It is possible to enter into a forward rate agreement in which one counterparty lends to another for three months starting at a future date, but at an interest rate that is agreed today. This is however a risky proposition as the borrower may turn out to be on the brink of bankruptcy at the start of the loan. Eurodollar futures are designed as a bet on the level of three-month interest rates that strips out any credit risk. The long side of the contract is agreeing to pay the three-month London Interbank Offered Rate (LIBOR) at the maturity date as interest rate on a notional deposit of $1 million in exchange for a fixed interest rate on that same notional deposit.

10 However Büyükşahin and Robe (2014) find that since 2008, the correlation of equity and commodity returns has increased substantially.

500  Research handbook of financial markets

The price of the instrument is quoted as 100 minus the fixed interest rate in percentage points; hence, as in fixed income space, lower interest rates correspond to higher prices. The contract is cash settled with only the net changing hands. If the futures price is F, then the corresponding interest rate is iF = 100 - F and at the expiration of the futures contract, if the actual interest rate is i in percentage points, the long party makes a payment to the short party of $2,500(i - iF ) .11 This may be negative, with the short party making the payment. Eurodollar futures trade at maturities up to ten years, allowing their complete term structure to be modeled. Researchers have compared the Eurodollar futures and forward rates, explaining their differences in terms of the mark-to-market features of the futures contract, taxation, and default risk (see, e.g. Grinblatt and Jegadeesh, 1996). At the time of writing, there is considerable uncertainty about the future of the Eurodollar futures contract. Futures are listed for March, June, September, and December12 going out ten years, and yet the LIBOR interest rate was supposed to cease being produced in December 2021. The Federal Reserve Board, the Office of the Comptroller of the Currency and the Federal Deposit Insurance Commission have issued joint supervisory guidance discouraging banks from entering into new derivatives contracts that reference LIBOR after December 2021, however Eurodollar futures continue to settle on LIBOR.13 Federal funds futures are bets on the average level of the effective federal funds rate in a given month and, in conjunction with Eurodollar futures, these are widely used to measure expectations for the future path of monetary policy (Gürkaynak et al., 2007). Kuttner (2001) proposed using high-frequency changes in federal funds futures to measure surprises in the target federal funds rate at Federal Open Market Committee (FOMC) meetings, and these measures of unexpected changes in the target funds rate at FOMC meetings are very widely used as monetary policy surprises (see, for example, Bernanke and Kuttner (2005)). But since the FOMC has long sought to influence longer-term interest rates by communicating its intentions about policy well ahead of time, the effective stance of monetary policy may be better measured by a somewhat longer maturity interest rate. Accordingly, some authors have measured monetary policy surprises at FOMC meetings from changes in short-term interest rate futures maturing a few quarters out (Gertler & Karadi, 2015; Miranda-Agrippino & Rey, 2020). As an alternative, Gürkaynak et  al. (2005) showed how to combine Federal Funds and Eurodollar futures of various maturities to measure two separate monetary policy surprises— surprises in the funds rate target and in the path of future policy. Moving on to the era of unconventional policy, Swanson (2021) combines high-frequency changes in futures and other asset prices to measure three dimensions to FOMC monetary policy surprises—the target surprises, path surprises and asset purchase surprises. The surprises obtained in these ways can be used directly, or as “external instruments” in a structural VAR (e.g. Gertler & Karadi, 2015; Miranda-Agrippino & Rey, 2020; Eberly et al., 2020).

11 $2,500 is the change in the value of the interest on a one million dollar loan per percentage point change in the interest rate, per quarter. 12 There are also contracts for the first four months that are not in the quarterly cycle. 13 The secured overnight funding rate (SOFR) is an overnight repo rate calculated based on actual transactions by the Federal Reserve Bank of New York. It is seen as the likely successor to LIBOR as a benchmark short-term interest rate. The CME now trades futures on the SOFR as well.

Futures and options 

501

Most of the literature that uses interest rate futures to measure monetary policy surprises treat these as temporary deviations from the policy rule of the central bank. But some authors have raised the possibility that they also reveal to the public information that the central bank may have, or be perceived to have, about the underlying state of the economy (Campbell et al., 2012). Jarocinski and Karadi (2020) and Cieslak and Schrimpf (2019) use the behavior of stock returns around monetary policy announcements to parse monetary policy surprises into different components. Cieslak and Schrimpf (2019) find that there is an “information” component but that it is small. One argument for an information effect, made by Campbell et al. (2012) and Nakamura and Steinsson (2018) is that surprise monetary policy easings are followed by Blue Chip survey respondents subsequently lowering their growth projections or raising their unemployment projections. Bauer and Swanson (2021) make a strong argument against this interpretation. 22.7.2.2 Commodity futures Commodity futures are the oldest type of futures contracts and are traded for commodities like wheat, oil, or precious metals. They involve physical delivery with very precise specifications. For example, the West Texas Intermediate crude oil futures contract requires delivery of 1,000 barrels of oil with specific chemical properties in Cushing, Oklahoma. Most commodity futures contracts do not actually go to physical settlement and are instead reversed before delivery. Oil futures switch between being in contango and backwardation. Two notable episodes of contango, shown in Figure 22.2, were in the winter of 2008 and in the spring of 2020. On both occasions, spot and short-dated futures oil prices were low because of economic weakness (the Great Recession and the COVID recession), but the economy was expected to recover in future years and so longer-dated futures prices were much higher. Indeed in April 2020, some short-dated oil futures prices went negative because of a glut of oil. Under these circumstances, buying oil in the spot market and storing it for delivery in the futures market was appealing, but was limited by storage capacity. While contango is the usual configuration of futures curves due to the presence of positive term premia pushing longer-term prices up, backwardation is not all that rare. The two examples shown in Figure 22.2 are October 1990, when the Iraqi invasion of Kuwait led to an increase in oil prices but this was expected to be short lived, and September 2008, when the onset of the Global Financial Crisis and ensuing US policies led to a decline in the value of the dollar, leading investors to flock to precious metals and oil. As the date is towards the end of the month, the expected transitory nature of the run up in oil prices manifests itself in the strong relative decline in the two-month, rather than the one-month forward contract. Känzig (2021) uses high-frequency data on oil futures around OPEC announcements to measure oil supply news shocks, rather analogous to the futures based monetary policy shocks discussed above. He proposes using these as external instruments in a structural VAR. 22.7.2.3 Index futures Another important class of cash-settled futures are stock index futures. The CME trades S&P futures where the long party receives from the short $250 for every point that the index exceeds the futures price at maturity. There are also E-mini S&P futures that are identical except that the payment is $50 per point. Contracts are listed expiring in March, June, September and December. Researchers have studied the relationship between cash stock market indices and

502  Research handbook of financial markets

Note:   Price of West Texas Intermediate Oil Future contracts against maturity. Source:  Bloomberg.

Figure 22.2  Oil futures prices by maturity on selected dates index futures. Hasbrouck (2003) finds that price discovery for the S&P 500 index occurs mainly in the E-mini futures market. Martin (2017) shows that equity index options can be used to construct a time-varying lower bound on the equity risk premium. This lower bound averages around five percentage points, but spikes to higher levels at times of financial market turmoil. The fact that these lower bounds are quite high suggests that they are at least close to being sharp bounds on the equity risk premium. 22.7.2.4 Currency futures Currency futures have been around since the collapse of the Bretton-Woods system. They involve physical delivery of a foreign currency, but they are not in fact the dominant way in which exchange rate movements are hedged. Hedging of exchange rate movements is done more by forward contracts and swaps that are traded over-the-counter and can be customized to the client’s specific needs (Papaioannou, 2006). Géczy et al. (1997) document the ways in which different types of firms use foreign exchange derivatives. 22.7.2.5 Currency options A risk-reversal is an options strategy consisting of buying out-of-the money calls and selling comparably out-of-the money puts, or the other way around. This strategy is essentially a bet on the skewness of the underlying asset. Risk-reversals are particularly common for currency options. Some currencies, such as the Australian and New Zealand dollar, have riskreversal prices that imply that large depreciations vis-à-vis the US dollar are more likely than

Futures and options 

503

large appreciations (Brunnermeier et al., 2008). Other currencies, such as the Japanese Yen and Swiss Franc have risk-reversal prices that imply that large appreciations are more likely. (Brunnermeier et al., 2008) and Farhi and Gabaix (2016) show that the currencies which have downward/upward crash risk tend to be the investing/funding currencies in the famous carry trade. This therefore provides an explanation for the profitability of the carry trade and the failure of uncovered interest parity—there are expected excess returns on borrowing in a low interest currency and investing in the high interest currency, but they are compensation for the risk of rare but large losses. The availability of options prices was key to researchers reaching this insight. 22.7.2.6 Options on individual stocks Options on individual stocks trade on exchanges, such as the CBOE. Options on individual stocks are typically American options, that can be exercised either at expiry or sooner. Most of these options have an expiration of one year or less, but there are longer maturity options known as long-term equity anticipation securities (LEAPs). Driessen et al. (2009) compare the prices of individual options with those on index options, and use this to measure options-implied correlations between stock returns. Kelly et al. (2016) apply the same idea to comparing prices of options on individual bank shares with option prices on a financial sector index, with findings pointing that especially since the 2008–2009 financial crisis, options prices imply that there is an implicit government backstop to the financial sector as a whole. 22.7.2.7 Options on futures Investors can buy options to give them the right, but not the obligation, to enter into either long or short futures contracts. Some CME futures options are European, while others, like Eurodollar and Treasury futures options, are American. Closely related to these, swaptions give the holder the right but not the obligation to enter into an interest rate swap contract. All of these options are useful for quantifying interest rate uncertainty as they allow calculation of implied volatility (e.g. Swanson, 2006; Wright, 2020). Bikbov and Chernov (2011) fit a term structure model to Eurodollar futures and options prices, studying the relationship between options prices and the term structure of interest rates. 22.7.2.8 VIX The VIX index is the annualized options-implied volatility from 30-day options on the S&P 500 index. The VIX has been calculated by the CBOE since 1993. In its early years it was based on the S&P 100 and used the Black-Scholes formula to get the implied volatility. Since 2003, the VIX has been based on options on the broader index, and the VXO index, based on options on the S&P 100 index, has been constructed as a separate index. Furthermore, the volatility is now computed directly from options prices without relying on the Black-Scholes or any other specific model as:

VIX =

2 exp(r t) æ ç t è

ò

F

0

P( K ) dK + K2

ò

¥

F

ö C(K ) dK ÷ (22.5) 2 K ø

where P(K) and C(K) denote put and call option prices at strikes of K, respectively, r is the risk-free rate and τ is the time to maturity (Carr & Madan, 1998, Britten-Jones & Neuberger,

504  Research handbook of financial markets

2000). Equation 22.5 is known as the “model free implied volatility.” The VIX and VXO receive a lot of attention from investors, the press and academia, and the VIX is often referred to as the “fear index.” For example, Bloom (2009) proposed a model in which uncertainty shocks are important drivers of the business cycle, and used the VXO to measure uncertainty. While the VIX is an index rather than an asset that can be directly traded, there are now cash settled futures on the VIX. These have a payout that is based on the difference between the futures price and the VIX on the settlement day. VIX futures are available for a range of maturity dates and the term structure of VIX futures generally slopes up, so that a strategy of going short in longer-maturity VIX futures will on average be profitable, although it is subject to the risk of an abrupt crash. There are also options on VIX futures—options on futures on implied volatility that are in turn extracted from the price of another option contract. The difference between the options-implied volatility and physical volatility can be shown to be a function of investor risk aversion (Bollerslev et al., 2011). Consequently, the gap between the VIX and an econometrician’s forecast of volatility is known as the variance risk premium, and is used as a time-varying measure of risk aversion. It can be used, for example, to predict short-term excess stock returns (Bekaert & Hoerova, 2014; Bollerslev et al., 2009, 2011). And we can in this way decompose the VIX into a “quantity” and “price” of risk.

22.8 CONCLUSION In a world of complete markets, futures and options are redundant securities. But markets are not complete and these securities perform roles that cannot be trivially replicated. For market participants, buying and selling futures and options allows for hedging risks and taking risky positions—making bets—based on their information and beliefs. They are useful for risk sharing, but at the same time can on occasion contribute to financial instability, putting them firmly on regulators’ radar screens. Since futures and options are unique in their usage, they are also unique in their information content. Hence, research has progressed in both pricing and reverse engineering prices to extract information. Doing this right requires paying attention to the specifics of contracts, the regulatory environment, and other risks facing the users. A change in prices because a regulatory change gave a tax advantage to a contract should not be confused with one due to changes in investors’ beliefs or risk preferences. The literature has been making strides in separating risk premia from expected future outcomes but there is much to be done and this remains an active area of research. Risk premia are not nuisance components—depending on the application the risk premium and its decomposition into price and quantity of risk may be as important as backing out the expected future prices and probabilities. Another avenue of research is following the creation of new markets and securities, and understanding which beliefs go into pricing these. As the spectrum of futures and options grows, research utilizing the prices of these securities grows in parallel.

REFERENCES Aït-Sahalia, Y., & Lo, A. W. (2000). Nonparametric risk management and implied risk aversion. Journal of Econometrics, 94(1–2), 9–51.

Futures and options 

505

Alquist, R., & Kilian, L. (2010). What do we learn from the price of crude oil futures? Journal of Applied Econometrics, 25(4), 539–573. Andersen, T., Bollerslev, T., Diebold, F., & Vega, C. (2007). Real-time price discovery in global stock, bond and foreign exchange markets. Journal of International Economics, 73(2), 251–277. Barth, D., & Kahn, R. J. (2021). Hedge funds and the Treasury cash-futures disconnect. Office of Financial Research working paper 21-01. Bauer, M. D., & Swanson, E. T. (2021). An alternative explanation for the “Fed information effect”. San Francisco Fed working paper 2020-06. Bekaert, G., & Hoerova, M. (2014). The VIX, the variance premium and stock market volatility. Journal of Econometrics, 183(2), 181–192. Bernanke, B. S., & Kuttner, K. N. (2005). What explains the stock market’s reaction to Federal Reserve policy? Journal of Finance, 60(3), 1221–1257. Bikbov, R., & Chernov, M. (2011). Yield curve and volatility: Lessons from Eurodollar futures and options. Journal of Financial Econometrics, 9(1), 66–105. Black, F., & Scholes, M. (1973). The pricing of options and corporate liabilities. Journal of Political Economy, 81(3), 637–654. Bliss, R. R., & Panigirtzoglou, N. (2004). Option-implied risk aversion estimates. Journal of Finance, 59(1), 407–446. Bloom, N. (2009). The impact of uncertainty shocks. Econometrica, 77(3), 623–685. Bollerslev, T., Gibson, M., & Zhou, H. (2011). Dynamic estimation of volatility risk premia and investor risk aversion from option-implied and realized volatilities. Journal of Econometrics, 160(1), 235–245. Bollerslev, T., Tauchen, G., & Zhou, H. (2009). Expected stock returns and variance risk premia. Review of Financial Studies, 11, 4463–4492. Breeden, D. T., & Litzenberger, R. H. (1978). Prices of state contingent claims implicit in options prices. Journal of Business, 51(4), 621–651. Britten-Jones, M., & Neuberger, A. (2000). Option prices, implied price processes, and stochastic volatility. Journal of Finance, 55(2), 839–866. Brunnermeier, M. K., Nagel, S., & Pedersen, L. H. (2008). Carry trades and currency crashes. NBER Macroeconomics Annual, 23(1), 313–348. Büyükşahin, B., & Robe, M. A. (2014). Speculators, commodities and cross-market linkages. Journal of International Money and Finance, 42, 38–70. Campbell, J. R., Evans, C. L., Fisher, J. D. M., & Justiniano, A. (2012). Macroeconomic effects of Federal Reserve forward guidance. Brookings Papers on Economic Activity, 1, 2899–2934. Carr, P., & Madan, D. (1998). Towards a theory of volatility trading. In R. Jarrow (Ed.), Volatility: New estimation techniques for pricing derivatives. London: Risk Books. Carr, P., & Wu, L. (2004). Time-changed lévy processes and option pricing. Journal of Financial Economics, 71(1), 113–141. Chang, E. C. (1985). Returns to speculators and the theory of normal backwardation. Journal of Finance, 40(1), 193–208. Chatrath, A., Miao, H., & Ramchander, S. (2012). Does the price of crude oil respond to macroeconomic news? Journal of Futures Markets, 32(6), 536–559. Chernenko, S., Schwarz, K. B., & Wright, J. H. (2004). The information content of forward and futures prices: Market expectations and the price of risk. International Finance Discussion Paper, 808. Chinn, M. D., & Coibion, O. (2014). The predictive content of commodity futures. Journal of Futures Markets, 34(7), 607–636. Cieslak, A., & Schrimpf, A. (2019). Non-monetary news in central bank communication. Journal of International Economics, 118, 293–315. Cootner, P. H. (1960). Returns to speculators: Telser vs. Keynes. Journal of Political Economy, 68(4), 396–404. Cox, J. C., Ingersoll, J. E., & Ross, S. A. (1981). The relation between forward prices and futures prices. Journal of Financial Economics, 9(4), 321–346. Driessen, J., Maenhout, P. J., & Vilkov, G. (2009). The price of correlation risk: Evidence from equity options. Journal of Finance, 64(3), 1377–1406. Duffie, D., Pan, J., & Singleton, K. (2000). Transform analysis and asset pricing for affine jumpdiffusions. Econometrica, 68(6), 1343–1376.

506  Research handbook of financial markets

Eberly, J. C., Stock, J. H., & Wright, J. H. (2020). The Federal Reserve’s current framework for monetary policy: A review and assessment. International Journal of Central Banking, 16, 5–71. Fama, E. F., & French, K. R. (1987). Commodity futures prices: Some evidence on forecast power, premiums, and the theory of storage. Journal of Business, 60(1), 55–73. Farhi, E., & Gabaix, X. (2016). Rare disasters and exchange rates. Quarterly Journal of Economics, 131(1), 1–52. Faust, J., & Wright, J. H. (2018). Risk premia in the 8:30 economy. Quarterly Journal of Quantitative Finance, 8. Géczy, C., Minton, B. A., & Schrand, C. (1997). Why firms use currency derivatives. Journal of Finance, 52(4), 1323–1354. Gertler, M., & Karadi, P. (2015). Monetary policy surprises, credit costs, and economic activity. American Economic Journal: Macroeconomics, 7(1), 44–76. Gorton, G., & Rouwenhorst, K. G. (2006). Facts and fantasies about commodity futures. Financial Analysts Journal, 62(2), 47–68. Gray, R. W. (1963). Onions revisited. Journal of Farm Economics, 45(2), 273–276. Grinblatt, M., & Jegadeesh, N. (1996). Relative pricing of Eurodollar futures and forward contracts. Journal of Finance, 51(4), 1499–1522. Gürkaynak, R. S., Sack, B., & Swanson, E. T. (2005). Do actions speak louder than words: The response of asset prices to monetary policy actions and statements. International Journal of Central Banking, 1, 55–93. Gürkaynak, R. S., Sack, B., & Swanson, E. T. (2007). Market-based measures of monetary policy expectations. Journal of Business and Economic Statistics, 25(2), 201–212. Hasbrouck, J. (2003). Intraday price formation in us equity index markets. Journal of Finance, 58(6), 2375–2400. Hegde, A. (2004). An economic history of the failure of broiler futures. mimeo, North Carolina State University. Heston, S. L. (1993). A closed-form solution for options with stochastic volatility with applications to bond and currency options. Review of Financial Studies, 6(2), 327–343. Hirshleifer, D. (1990). Hedging pressure and future price movements in a general equilibrium model. Econometrica, 58(2), 411–428. Houthakker, H. S. (1957). Can speculators forecast prices? Review of Economics and Statistics, 39(2), 143–151. Jarocinski, M., & Karadi, P. (2020). Deconstructing monetary policy surprises—The role of information shocks. American Economic Journal: Macroeconomics, 12(2), 1–43. Johnson, A. C. (1973). Effects of futures trading on price performance in the cash onion market (pp. 1930–1968). US Department of Agriculture working paper 1470. Kang, W., Rouwenhorst, K. G., & Tang, K. (2020). A tale of two premiums: The role of hedgers and speculators in commodity futures markets. Journal of Finance, 75(1), 377–417. Känzig, D. R. (2021). The macroeconomic effects of oil supply news: Evidence from OPEC announcements. American Economic Review, 111(4), 1092–1125. Kelly, B., Lustig, H., & Van Nieuwerburgh, S. (2016). Too-systemic-to-fail: What option markets imply about sector-wide government guarantees. American Economic Review, 106(6), 1278–1319. Keynes, J. M. (1930). Treatise on money. London: Macmillan. Kitsul, Y., & Wright, J. H. (2013). The economics of options-implied inflation probability density functions. Journal of Financial Economics, 110(3), 696–711. Knoll, M. S. (2008). The ancient roots of modern financial innovation: The early history of regulatory arbitrage. Oregon Law Review, 87, 93–116. Kuttner, K. N. (2001). Monetary policy surprises and interest rates: Evidence from the fed funds futures market. Journal of Monetary Economics, 47(3), 523–544. Lahaye, J., Laurent, S., & Neely, C. J. (2011). Jumps, cojumps and macro announcements. Journal of Applied Econometrics, 26(6), 893–921. Livingston, M. (1984). The cheapest deliverable bond for the CBT Treasury bond futures contract. Journal of Futures Markets, 4(2), 161–172. Martin, I. (2017). What is the expected return on the market? Quarterly Journal of Economics, 132(1), 367–433.

Futures and options 

507

Miao, H., Ramchander, S., & Zumwalt, J. K. (2014). S&P 500 index-futures price jumps and macroeconomic news. Journal of Futures Markets, 34(10), 980–1001. Miller, M. H. (1986). Financial innovation: The last twenty years and the next. Journal of Financial and Quantitative Analysis, 21(4), 459–471. Millo, Y. (2007). From green fields to green felt tables and back: The origins of index-based derivatives. Centre for Analysis of Risk and Regulation Discussion paper, 44. Miranda-Agrippino, S., & Rey, H. (2020). U.S. monetary policy and the global financial cycle. Review of Economic Studies, 87(6), 2754–2776. Nakamura, E., & Steinsson, J. (2018). High-frequency identification of monetary non-neutrality: The information effect. Quarterly Journal of Economics, 133(3), 1283–1330. Papaioannou, M. G. (2006). Exchange rate risk measurement and management: Issues and approaches for firms. IMF Working paper 06/255. Rosa, C. (2014). The high-frequency response of energy prices to U.S. monetary policy: Understanding the empirical evidence. Energy Economics, 45, 295–303. Schrimpf, A., Shin, H. S., & Sushko, V. (2020). Leverage and margin spirals in fixed income markets during the Covid-19 crisis. BIS Bulletin 2. Shiller, R. J. (1993). Macro markets: Creating institutions for managing society’s largest economic risks. Oxford: Clarendon Press. Shimko, D. (1993). Bounds of probability. Risk, 6, 33–37. Stoll, H. R. (1969). The relationship between put and call option prices. Journal of Finance, 24(5), 801–824. Swanson, E. T. (2006). Have increases in Federal Reserve transparency improved private sector interest rate forecasts? Journal of Money, Credit, and Banking, 38(3), 791–819. Swanson, E. T. (2021). Measuring the effects of Federal Reserve forward guidance and asset purchases on financial markets. Journal of Monetary Economics, 118, 32–53. Till, H. (2015). Case studies on the success or failure of futures contracts. Journal of Governance and Regulation, 4(3). Whaley, R. E. (2006). Derivatives: Markets, valuation, and risk management. Chichester: John Wiley and Sons. Working, H. (1948). Theory of the inverse carrying charge in futures markets. Journal of Farm Economics, 30(1), 1–28. Working, H. (1953). Futures trading and hedging. American Economic Review, 43(3), 314–343. Working, H. (1960). Price effects of futures trading. Food Research Institute Studies, 1, 3–31. Wright, J. H. (2020). Event-day options. NBER working paper 28306.

Index

Asian financial crisis 105 Asian Infrastructure Investment Bank 120 Asonuma, T. 383, 387 Asset-backed commercial paper (ABCP) 89, 208 Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility (AMLF) 89 Asset-backed securities (ABS) 25, 89, 137 asset holdings 7–9, 19, 21, 27 asset inflation 471, 487 asset managers 255 asset pricing 1, 331, 362–4, 466, 497 asset purchase programmes 45, 61, 62 asset swaps 439–41 asymmetric information 257 auction markets 360 auction process auction success, measures of 288 mechanics of treasury auctions 284–6 primary dealers 286–8 secondary market pricing 296–8 when-issued market 288–96 yield curve interpretation 296–8 auction rate securities 305 Auh, J. K. 326 Authorised Participants (APs) 157 Avdjiev, S. 115

30-year fixed-rate mortgages 175–6 2009 American Recovery and Reinvestment Act 305 2013 Detroit bankruptcy 304 Abbassi, P. 466n39 ability to pay 20, 389–90 “Ability-to-Repay” (ATR) rule 334 abnormal returns 66 Abraham, M. 290n21 Abudy, M. M. 359 Adachi, K. 63, 69 Adelino, M. 320, 334 Adrian, T. 90, 165n18 advance refund 305, 308, 322 agency debt 8, 12, 17, 19, 349 agency MBS 17 agency vs. nonagency 332 Aguiar, M. 396 Aitken, J. 241 Akram, F. Q. 466n39 Albertazzi, U. 49, 135 Alfaro, L. 390 algorithmic trading 264, 265, 366, 462 Allahrakha, M. 249 Alquist, R. 498 Altavilla, C. 47 Alves, N. 46 Amador, M. 392 Ambrose, B. W. 180 American International Group (AIG) 89 American municipal bond market 314 Amihud, Y. 368 Amsterdam Stock Exchange 364 An, X. 347 Anbil, S. 82, 91, 244 Anderson, A. G. 244 Andonov, A. 323 Andrade, P. 46, 47 Ang, A. 473 Aoki, K. 2 appropriation-backed bonds 303 Arai, N. 73 Aramonte, S. 2 Arce, Ó. 74 Arslanalp, S. 381 artificial intelligence (AI) 462 Aschauer, D. A. 319 Ashcraft, A. 90

Baber, W. 324 Bagehot, W. 80, 109, 128 Bahaj, S. 2, 115, 117, 119, 122 balance-sheet identity 98 Balke, N. 392 Balteanu, I. 389 Bampinas, G. 473 bank-dependent borrowers 139 bank deposits bank deposit-rate ceilings 195 and investment funds 195 Bank for International Settlements (BIS) 103, 254, 429 banking system 126 Bank of America 26n30, 461, 477 Bank of Canada 104 Bank of England 79, 103, 104, 113, 426 Bank of Japan’s balance sheet balance sheet expansion and expectations management 66–8 bank lending 70–71 effects of 75 509

510  Research handbook of financial markets

ETF purchases 68–70 Japanese economy and 57–65 JGB purchases 70 macroeconomy 71 model-based approaches 73–5 time-series approach 71–3 Bank of Korea 111, 117 bank-run equilibrium 130 banks commercial banking 126 goldsmiths in London and Amsterdam 126 intermediation 129–30, 137–40 liquidity providers 130–32 Medici bank 126 producers of information 136–7 relationships and banking 132–6 bank-sponsored short-term investment funds (STIFs) 214 Barbara, J. 366 Barbon, A. 68 Barth, D. 499 Basel Accord 127, 434 Basel Committee on Banking Supervision (BCBS) 127 Basel III bank capital rules 163 basic balances 64 basis swaps 409, 413–14 Bauer, M. D. 501 Baxter, M. 438 Bayesian techniques 74 Beber, A. 370 Belton, T. M. 243n7 Benford, J. 395 Berentsen, A. 382 Berger, A. N. 132, 134, 135 Bergstresser, D. 3, 315, 323 Berlin, M. 134 Bernanke, B. S. 18, 138, 139, 188, 234 Berndt, A. 137 Bessembinder, H. 348, 349 Bessler, W. 370 Biais, B. 325, 371 Bianchi, J. 392 Bicksler, J. 424 bid side swap rate 411 Bikbov, R. 503 bilateral repo market 247 Binance USD 140 Bindseil, U. 165 Black, F. 495 Blackrock 198n14 Blanchard, O. J. 394 Blinder, A. S. 139 blockchain technology 371–3 implementations 140

Bloom, N. 504 Bocola, L. 44n18, 47, 387, 392 Bofondi, M. 135 Bollerslev, T. 504 Bolton, P. 395 Bomfim, A. N. 3 bond issuance 307 bond market 302 Boons, M. 485 Bord, V. M. 137 Borensztein, E. 387 borrowed reserves 84, 85, 223 borrowers 57 Bottero, M. 135 Boudoukh, J. 474 Bourdeau-Brien, M. 326 Bouveret, A. 2, 205 Bouwman, C. H. 132 Bowman, D. 71 Boyarchenko, N. 165n18, 247, 344, 346, 425 Brainard, L. 28 Bräuning, F. 466n39 Breach, T. 485 break-even inflation (BEI) rate 475 Bretscher, L. 424 Bretton Woods era 103 broker-dealers 21, 150, 151 balance sheets 150, 152 Brolley, M. 365 Broner, F. 381 Brooks, L. 319 Brooks, R. 322 Brown, J. 323 Brunnermeier, M. K. 367, 382 Budget Control Act 433 Buffet, W. 3 Bundesbank 80 Burgess, G. 3 Burgess, W. R. 92 Burghardt, G. D. 243n7 Busch, M. O. 115 Buser, S. A. 331 Buti, S. 361 Butler, A. W. 322, 325 Buttimer, R. 180 Butz, M. 257 buy side 359–60 Cahn, C. 47 Calomiris, C. W. 130 Campbell, J. Y. 351 Campbell, S. 90 capital asset pricing model 362 capital constraint channel 66, 68, 69 capital efficiency 154

Index 

capital investment 69 capital markets 2, 3 Carlson, M. 82, 90, 96, 223n6 Carpenter, S. 75 Carpinelli, L. 135 Carr, P. 495 cash holdings 161 Casolaro, L. 136 Catyas, J. 372 Cboe FX ECN 460 CDX.NA.IG index 447–8 Cecchetti, S. 142, 143 Cenedese, G. 466n38 central bank digital currencies (CBDC) 28, 140–43, 273 central bank lending Federal Reserve’s regular discount window programs 86–8 history and rationale 79–82 moral hazard 95–8 repo lending 91 role in monetary policy 82–5 stigma and lender of last resort 91–5 unusual and exigent circumstances 88–91 central clearing 446 central counterparties (CCPs) 147 central limit order books (CLOBs) 258 certificates of deposit (CD) 156, 197 Cespa, G. 463 Cetina, J. 249 Cetorelli, N. 136 Chaboud, A. P. 2, 272n40 Chalmers, J. M. R. 321 Charoenwong, B. 69 Chatterjee, S. 394 cheap cash bonds 163 Chen, A. H. 424 Chen, H. 74, 322 Chen, J. 348 Chen, L. 363 Chernov, M. 503 Chiang-Mai initiative 103, 107n3 Chicago Board of Trade (CBOT) 491 Chicago Board Options Exchange (CBOE) 491 Chicago Mercantile Exchange (CME) 261, 477, 491 China 210 Chinn, M. D. 498 Chiu, J. 367, 372 Chodorow-Reich, G. 139 Chordia, T. 365 Choudhry, M. 243n7 Chung, K. H. 365 Cieslak, A. 501 Cingano, F. 139

511

Citigroup 26n30 Clarida, R. H. 90 clearing 493 and settlement 364–5 Clearing House Electronic Subregister System (CHESS) 373 Cohen, N. R. 323 Cohen, R. 315 Coibion, O. 498 Cole, H. L. 386 collateral 109 collateralization and netting practices 441 Colliard, J.-E. 366 Collin-Dufresne, P. 425 commercial banks 9n9, 86 commercial paper (CP) 62, 131, 197 Commercial Paper Funding Facility (CPFF) 89 Committee on Payment and Settlement Systems (CPSS) 463 commodities 474–5 Commodity and Futures Trading Commission (CFTC) 492, 493 commodity futures 501 conditional prepayment rate (CPR) for agency MBS 342f conduit issuance 306 Cong, L. W. 372 “constant” NAVs (CNAVs) 197 constructive ambiguity 96 consumer and producer inflation components 481–2 consumer price index (CPI) 62, 278 consumer-price inflation 472 consumer prices 470 Continental Illinois National Bank 85, 132 contingent liabilities 389 Continuous Linked Settlement (CLS) 270 Copeland, A. 246–8, 250 core deposits 134n7 Cornaggia, J. 324 corporate bonds 62, 301 mutual funds 161 corporate debt 73 Corradin, S. 248 Correia, I. 44n18 corridor system 16n15 countercyclical borrowing 379 counterparty credit risk 440–41 coupon/fixed-floating swaps 409, 410 covered interest rate parity (CIP) 454 Covid-19 pandemic 56, 64, 80, 81, 87, 90, 104, 161, 195, 315, 344, 348, 369 CCP margins and fund flows 162 Cox, J. C. 497 credit default swaps (CDS) 434–6

512  Research handbook of financial markets

global financial crisis, market conventions 445–6 indexes 446–9 market indicators 437 market participants 431–2 protection buyers 436 protection sellers 436–7 reference entities 432–4 single-name CDS 429 valuation considerations 437–45 credit enhancement 174 credit rating agencies (CRAs) 198, 303 credit risk management 257 credit risk transfers (CRTs) 174 credit support annex (CSA) 415 credit unions 9n9 Crosignani, M. 46 cross-border payments 141 cross-currency swaps 409, 414–15 Cruces, J. J. 386 Cuny, C. 325 Cúrdia, V. 74 currencies 478 currency futures 502 currency options 502–3 Cutler, D. 319 D’Amico, S. 4, 289, 485 Danielli, L. 205 Danish National Bank (DNB) 113, 118, 233 Davidson, A. 331 Davis, I. 247 Dawsey, K. 2 dealer-to-customer segment 260, 460, 461 Debrun, X. 394 debt dilution 393 debt payments 388–9 decentralised finance 158 Dedola, L. 49 default, liquidity line 110 De Fiore, F. 129 deflationary pressure 59 deflation risk 4 Degryse, H. 134, 135 De Haas, R. 135 Dell’Ariccia, G. 45 De Michelis, A. 67 depository institutions (DIs) 86 Depository Trust & Clearing Corporation (DTCC) 245 Derber, J. 389, 390 derivatives markets 3 DesJardine, M. R. 360 Diamond, D. W. 81, 129–31, 201 Diamond, J. 59

Diamond-Dybvig model 81 digital euro 51 Di Maggio, M. 347 discount rate 82, 222 discount window 82, 222 loans 88 disinflation/deflation risk 472 Dodd–Frank Act 463, 494 Dodd-Frank ATR rule 334n7 domestic financial crisis 56 Du, W. 396, 466n39 Duffie, D. 165n17, 244, 250, 425, 495 Dugast, J. 361 Durand, R. 360 Durham, J. B. 3, 290n21, 294n25 Duygan-Bump, B. 90 Dybvig, P. H. 81, 129–31, 201 dynamic stochastic general equilibrium (DSGE) model 74 Eaton, J. 385 Eberly, J. 351 economic and financial developments 34 Economic and Monetary Union (EMU) 33 Eggertsson, G. B. 44n18, 70 Eguren-Martin, F. 115 Eichengreen, B. 381, 396 Electronic Broking Services (EBS) 258, 460 electronic communication network (ECN) 260, 462 electronic markets 153 electronification 460 Emergency Economic Stabilization Act 19 emergency lending programs 25 emerging markets (EM) 472, 478 emerging markets and developing economies (EMDEs) 381 English, W. B. 2, 139 Ennis, H. M. 249 Eom, Y. H. 425 equilibrium, in federal funds market actual equilibrium after 2008 228–30 Federal Reserve’s reverse repurchase facility 230–31 interest on reserves 227–8 short-term U. S. interest rate 231–3 equilibrium exchange rates 254 equilibrium federal funds rate 223n6 equity 3 market liquidity 367–8 shares 359 equity trading blockchain technology and “tokenization” 371–3 buy side 359–60

Index 

clearing and settlement 364–5 funding markets and equity market liquidity 367–8 price setting and market efficiency 361–4 regulation 368–71 sell side 360–61 trading costs 365–7 equivalent cash bond 154 Erb, C. 475 Erce, A. 389 Erel, I. 351n33 Eren, E. 115 Eser, F. 49 EURCHF exchange rate 269 Euro area banks 117 Eurocurrency borrowings 134n7 Eurodollar 134n7 Euromoney FX survey 260n18 Euro Overnight Index Average (EONIA) 410n4 European banks 121 European Central Bank (ECB) 33, 56, 79 European crisis 1, 104, 105, 432 European Deposit Insurance Scheme 131 European Exchange Rate Mechanism (ERM) 267 European Financial Market Infrastructure Regulation (EMIR) 463 European mutual funds 198 European Union 208, 212–13 Eurosystem’s balance sheet between 2007 and 2014 37 ABS purchase programme (ABSPP) and corporate sector purchase programme (CSPP) 40 covered bond purchase programmes (CBPPs) and securities markets programme (SMP) 38–9 before Financial Crisis 35–6 liabilities 39, 41 longer-term refinancing operations (LTROs) 37–8 monetary policy 44–50 multi-country nature of EMU 42–3 pandemic emergency longer-term refinancing operations (PELTROs) 41 pandemic emergency purchase programme (PEPP) 40–41 public sector purchase programme (PSPP) 40 since 2014 40 targeted longer-term refinancing operations (TLTROs) 41 Eurosystem repo facility for central banks (EUREP) 105 Eurozone 61

513

Evans, M. D. D. 271 exchange rates 75 exchange-traded funds (ETFs) 62, 157, 359 expectations hypothesis (EH) 297, 417 externalities 81, 149 market participants 165 Eyigungor, B. 394 Fabo, B. 45 Fabozzi, F. J. 331 Fama, E. F. 136, 137, 363n5 Fannie Mae 2, 173–4, 177, 178, 179, 182, 184, 186, 189 Farhi, E. 392, 503 Faulkender, M. 424 federal agency obligations 83 Federal Deposit Insurance Corporation (FDIC) 82, 127 Federal Deposit Insurance Corporation Improvement Act 87 federal funds 134n7 defined 220 rate 221 federal funds market 2 brief history of 220–21 equilibrium in 225–33 features of 221–5 future of 233–4 Federal Home Loan Bank Board (FHLBB) 182 Federal Home Loan Banks 24, 221 Federal Home Loan Mortgage Corporation (FHLMC) 171, 221 Federal Housing Enterprises Financial Safety and Soundness Act 182 Federal Housing Finance Agency 174 Federal National Mortgage Association (FNMA) see Fannie Mae Federal Open Market Committee (FOMC) 7, 83, 83n7 Federal Reserve 2 Federal Reserve Act (FRA) 8, 9, 19, 25, 29, 79n3, 86n9, 87, 88 Federal Reserve asset purchase programs 19 Federal Reserve balance sheet Covid-19 pandemic 6 Global Financial Crisis (GFC) 6 historical perspective on 11–14 leading up to the GFC 14–16 lending and private asset programs 24–7 longer-term interest rates 6 mechanics of 7–10 mortgage-backed securities (MBS) 6 outlook for 27–9 post-crisis policy implementation framework 21–4

514  Research handbook of financial markets

short-term interest rates 6 sources of information on 32 Federal Reserve Bank (FRB) 56, 87, 88, 92 Federal Reserve Bank of New York 8, 82n6, 221 Federal Reserve Board 19 Federal Reserve Board’s Financial Accounts database 315 Federal Reserve District 88 Federal Reserve liquidity lines 112 Federal Reserve System 82, 126, 127, 315 Fernald, J. G. 319 Fernández-Villaverde, J. 142 Ferrero, A. 74 FHLMC see Federal Home Loan Mortgage Corporation (FHLMC) Fieldhouse, A. J. 173 Financial Conduct Authority (FCA) 409 financial crisis 1, 364 financial customers 258 financial deregulation 57, 59 financial flows 110 financial inclusion 28 Financial Industry Regulatory Authority (FINRA) 361, 368 financial innovation 306 financial instability 1 financial institutions 45, 255 Financial Institutions Reform, Recovery, and Enforcement Act (FIRREA) 182 financial instruments 303 financial intermediaries 129 financial intermediation 1 financial market 1, 3, 98 infrastructures 154 participants 121–2 Financial Services Regulatory Relief Act 19 financial stability 102 Financial Stability Board 270 Financial Stability Oversight Council 214, 494 financial turmoil 155 Finkelstein, D. 174 fiscal policy 29 Fischer, S. 120 fixed-floating swap 408 fixed income markets 153 fixed payer 407 Flandreau, M. 103 Fleckenstein, M. 485 Fleming, M. J. 241 Floating-Rate Notes (FRNs) 306 floating receiver 407 Focarelli, D. 136 Fohlin, C. 3, 368, 370 Foreign and International Monetary Authorities (FIMA) 105

foreign exchange (FX) 102 algorithmic trading in 264–6 Committees 266 counterparties 255–7 daily trading volumes and geography of trading 254–5 end-users of 253 flash events and extreme events 268–9 foreign exchange benchmark rates and “fixing scandal” 269–70 future of 272–3 liquidity and market fragmentation 271–2 new intermediaries and price discovery 272 official sector 266–8 over-the-counter (OTC) market 253 settlement risk 270–71 trading environment 257–64 foreign exchange swaps definition and usages of 451–4 further considerations on 454–5 institutional framework 458–66 market 455–8 over-the-counter (OTC) market 460–61 policy actions 463–6 research 466–7 technological changes 461–3 foreign official institutions 24 forward swap 423 Foucault, T. 366 Frankel, J. A. 392 Freddie Mac 2, 173–5, 182–6, 188–9 free of default risk 3 Friedman rule 51 Frost, J. 134 Fujiki, H. 75 Fujiwara, I. 62, 67 funding markets 367–8 Fuster, A. 3, 173, 350 futures and options clearing 493 commodity futures 501 currency futures 502 currency options 502–3 expectations and risk premia 498–9 history 490–91 index futures 501–2 on individual stocks 503 interest rate futures 499–501 open outcry and electronic trading 491 options contracts 497–504 pricing and usage 494–7 regulation 493–4 types of futures and options traders 491 VIX index 503–4 FX prime brokerage (PB) 257

Index 

Gabaix, X. 503 Galper, H. 320 Gambacorta, L. 73 Gao, P. 323, 324, 348 Garbade, K. D. 238, 241, 279n6 Garca-Posada, M. 46 Gargano, A. 463 Géczy, C. 502 Gehrig, T. 368 General Obligation debt 302–3 Gennaioli, N. 387 Georg, C. 205 German sovereign bond market 272n41 Gersovitz, M. 385 Gertler, M. 44, 138 GFC lending programs 26 Ghent, A. C. 180 Ghulam, Y. 389, 390 Gianinazzi, V. 68 Gilchrist, S. 138 Gillette, J. 320 Glass-Steagall Act 127, 131 Global Financial Crisis (GFC) 1, 3, 4, 61, 70, 80, 105, 127, 195, 260, 267, 430, 463, 465, 472, 501 Global Foreign Exchange Committee (GFXC) 266 Globally Systematically Important Banks (G-SIB) 464 Glode, V. 361 Goforth, C. 372 Goldberg, L. S. 136 Goldsmith-Pinkham, P. 326 Goodman, L. 181 Gordon, G. 326 Goretti, M. 390 Gorina, E. 324 Gorton, G. B. 248, 473, 475, 499 government agencies 6 government bonds 66 government guarantees 187 Government National Mortgage Association (Ginnie Mae) 172 governments default 388–90 government securities 98 government-sponsored enterprises (GSE) 221 advantages of no reform 188 conservatorship and preferred stock purchase agreements 183–4 GSE charter 172–3 guarantee of mortgage tail risk 187–8 history and oversight of 181–3 implicit guarantee 185–6 limitations 188–9 and money supply 186

regulated utility/government ownership 189 special privileges 184–5 true privatization 188 U. S. mortgage markets 173–81 Gramlich, E. M. 319 Great Depression 67, 103, 126, 138, 383 Great Financial Crisis (GFC) 150 Great Inflation 4 Great Recession 334 Green, R. C. 321, 324, 325 Green, R. K. 331 green sovereign securities 29n37 Greenwald, D. L. 351 Griffith, T. 365 Grishchenko, O. V. 290n21 Group of Seven (G-7) intervention 9n6 Gu, S. 363 guarantee fees 177–9 Guerron-Quintana, P. 326 Gupta, A. 137, 319, 425 Guren, A. M. 351 Gürkaynak, R. S. 297n28, 500 Gustafson, M. 326 Hall, G. 380 Hall, R. E. 75 Hamilton, A. 128 Han, S. 244 Hanisch, M. 72 Hanson, S. G. 425 Harada, K. 68 Harris, L. E. 325 Harvey, C. R. 475 Hasbrouck, J. 271, 502 Hatchondo, J. C. 380, 393, 394 Hattori, T. 67, 69 Hausman, J. K. 71 Hausmann, R. 381, 396 Hayashi, F. 72 Hayre, L. 331 He, Z. 372 headline inflation 480–81 Hébert, B. 387 hedge funds 153, 157–8, 255, 360, 431 liquidity provision 163 hedging 416–17 inflation risk 471 see also inflation hedging products Hegde, A. 491 Hernandez-Murillo, R. 180 Herstatt Bank 127 Heston, S. L. 495 high-frequency traders (HFTs) 256 high-quality liquidity assets (HQLA) 465 Hildreth, W. B. 318

515

516  Research handbook of financial markets

Hiraki, K. 63, 69 Hirshleifer, D. 498 Hochfelder 367 Hofmann, B. 73 Holmstrom, B. 382 Honda, Y. 72 Hong Kong Monetary Authority (HKMA) 111 Hoshi, T. 132 Hosono, K. 71 Housing and Economic Recovery Act (HERA) 182 Housing and Home Finance Agency (HHFA) 182 Howell, S. T. 372 Hrung, W. B. 90 Hu, G. X. 290 Huang, J. Z. 290n21 Huh, Y. 249, 348 Hungarian National Bank 104 Iacoviello, M. 67 ICE Benchmark Administration (IBA) 409 illiquid assets 132 IMF 102, 119, 378, 383 implicit guarantee 185–6 index futures 501–2 industrialization 358 Infante, S. 249 inflation 34 caps and floors 478 compensation 475 expectations 63 inflation risk premia (IRP) 472 inflation swap (IS) 477 options market 478 risk 4 inflation hedging products broad indices and individual futures 474–5 broad indices and individual stocks 473–4 buying and holding bonds 479–80 consumer and producer inflation components 481–2 currencies 478 empirical properties of 478–83 headline inflation 480–81 house prices and REITs 474 inflation derivatives 477–8 inflation risk premium 484–7 treasury inflation-protected securities 475–6 wages and house prices 482–3 information aggregation 257 initial margins (IMs) 160 initial public offering (IPO) 369 institutional investors 202–4, 207, 209, 255 insurance companies 66, 359, 471

Intercontinental Exchange 492 interdealer contracts 431 interdealer market 257, 258 interest on excess reserves (IOER) 250 interest on reserve balances (IORB) 19–20 interest on reserves 223 interest rates 64, 109 derivatives 154 futures 499–501 interest rate swaps (IRSs) 409, 423–6, 464 dynamics of swap spreads 418–21 and history 407–9 pricing 421–3 regulation and reforms 417–18 risks 415–16 swap usage and types of users 416–17 types of 409–15 Internal Revenue Code 318 Internal Revenue Service 304 international financial system 105 international financial transactions 141n12 International Monetary Market (IMM) 457, 458 International Swaps and Derivatives Association (ISDA) 435 international trade 105 in-the-money (ITM) 343 intraday FX swap volume 459 intraday time series 457 inventory risk management 257 inverse propensity score weighting (IPSW) approach 387 investment-grade entities 434 investor discounts real quantities 471 investors 335 Ippolito, F. 131 ISDA Master Agreement 460 Ivashina, V. 115 James, C. 136 Japan 199, 207–8 inflation 57 Japanese government bonds (JGBs) 62 Japan real estate investment trusts (J-REITs) 62 Jarocinski, M. 501 Jasova, M. 46 Jeanne, O. 396 Jermann, U. J. 424, 425 Jeske, K. 181 Jiang, Z. 382 Jiménez, G. 139 Johannes, M. 425 Johnson, C. L. 326 JP Morgan Chase 82

Index 

Kahn, C. M. 130 Kahn, R. J. 499 Kahraman, B. 367 Kalman-filter-based methods 298 Kamstra, M. J. 395 Kan, K. 62, 74 Kanczuk, F. 390 Känzig, D. R. 501 Karadi, P. 44, 501 Kashyap, A. K. 131, 138 Kawamoto, T. 62, 63 Keane, F. M. 241 Kehoe, P. J. 386 Keim, D. B. 366 Keister, T. 142 Kelly, B. 363, 503 Keynes, J. M. 498 Keys, B. J. 137 Kidwell, D. S. 324 Kilian, L. 498 Kim, D. H. 289 Kim, Y. 363 Kim, Y. S. 331, 348, 350 Kimura, T. 72 Kimura, Y. 70 King, M. 98 King, T. B. 4 Kirby, A. S. 241 Kishaba, Y. 62, 74 Kitamura, T. 63, 69 Kiyotaki, N. 44 Klingler, S. 425 Knoll, M. S. 495 Koeda, J. 70, 72 Koeppl, T. 372 Koesrindartoto, D. P. 359 Koijen, R. S. J. 49 Korajczyk, R. A. 363 Kos, D. 270 Kotidis, A. 249 Krishnamurthy, A. 46, 246, 248, 347, 351 Krohn, I. 467n46 Krueger, D. 181 Krugman, P. 395 Kryzanowski, L. 326 Kubota, H. 73 Kuroki, Y. 72 Kuvshinov, D. 387 land and stock prices 59 Landoni, M. 322 large hedge funds 153 latency arbitrage algorithms 265 Latin American debt problems 85 Layton, D. 188

517

Lee, A. J. 365 Lee, J. 363 Lee, T. 361 Lehman Brothers 16, 17, 19, 133 Lehnert, A. 173 Leland, H. E. 129 Lemke, W. 49 “lender of last resort” (LOLR) 42, 91, 105, 114, 117, 119, 121 lending programmes 45 LeSueur, E. 247 Letter of Credit (LOC) 306 Levich, R. M. 270, 271 Levin, A. 331 Levine, R. 128 Li, D. 325 Li, H. 424 Li, L. 91 Li, W. 181, 348 liability inflation 471 Liang, N. 165n17 Lintner, J. 362 liquidity aggregators 262 crisis 96, 165 demand pressures 150 liquidity-based demand 222 liquidity-constrained banks 71 liquidity coverage ratio (LCR) 465 liquidity providers (LPs) 256 and market fragmentation 271–2 mirage 271 regulations 96 transformation 194, 201–2 liquidity imbalances 147, 160–63 liquidity line, central banks agreement 107–9 collateral 109 default 110 different uses 116 economic consequences 114–19 evolution of 103–6 financial flows 110 for FX Interventions 118–19 interest rates 109 international trade and currency usage 117–18 limits 109–10 maturity 109 operation of 107–14 policy response 102 recipient central bank 111–14 reciprocity 109 setting up 114 terms and conditions of 108 liquid liabilities 132

518  Research handbook of financial markets

Liscow, Z. 319 Liu, H. 348 Liu, J. 425 loan limits 176–7 localized investor demand 290n22 Logan, L. 241 London Interbank Offered Rate (LIBOR) 409, 452, 499 longer-maturity municipal debt 304 Longer-term securities 27 Longest Chain Rule (LCR) 371 long-short hedge funds 160 Longstaff, F. 321 long-term equity anticipation securities (LEAPs) 503 long-term government bonds 66 long-term institutional investors 471 long-term interest rates 75 Long-Term Refinancing Operation (LTRO) 238 Lopez, J. A. 91 Loutskina, E. 350 Lu, Z. 370 Luby, M. J. 322, 325 Lucca, D. 3, 173 Lummer, S. L. 136 Lustig, H. 478 Lyons, R. K. 271 MacBeth, J. 363n5 Macchiavelli, M. 90, 249, 368 macro add-on balances 64 macroeconomy 1, 64, 71 and distributional effects 49–50 effects 46–7 fluctuations 143 variables 75 macroprudential approach 51, 164 Maddaloni, A. 248 Madhavan, A. 366 Madigan, B. 2, 94 Main Street Lending Program 90 Malinova, K. 365 Malone, S. W. 387 Manasse, P. 390 Mancini, L. 271 Mao, C. X. 424 Marchesi, S. 388 Marchetti, M. 46 margin requirements 165 market dysfunction 166 fragmentation 260–62 indicators 437, 448–9 liquidity 282–4 market-determined rate 82

market-making algorithms 264 microstructure 467 participants 166, 221, 258, 267 stress 24, 26 structure 1 marketable securities 278–9 Martin, A. 2, 246–9, 381 Martin, I. 502 Martinez, L. 3, 380 Masaru, H. 61 Masi, T. 388 Massachusetts Development Finance Agency 306 maturity extension program (MEP) 17 McCabe, P. 2 McCauley, R. N. 107 McConnell, J. J. 136, 331 McCormick, M. J. 246 McGuire, P. 115 McSherry, B. 364 Mendelson, H. 368 Menkhoff, L. 478 Mergent FISD database 310–13 Mertens, K. 173 Mester, L. J. 134 Metrick, A. 90, 248 Meyer, J. 380, 384 Mihov, I. 234 Mikami, T. 63 Miller, G. 319 Minsky, H. 1 Mitchener, K. J. 383, 386 Mitman, K. 181 Miyakawa, D. 71 Miyao, R. 72 MMF liquidation 206 model-based approaches 73–5 modeling prepayments 343 Modigliani, F. 70 Modigliani–Miller theorem 137 Moldogaziev, T. T. 325 Monetary Authority of Singapore (MAS) 109, 111 Monetary Control Act of 1980 (MCA) 86n10 monetary liabilities 80 monetary policy 33 asset prices 46, 47–9 bank lending 46 credibility 396–7 formulation and implementation 85 implementation 84 large-scale asset purchases 47–50 long-term liquidity provision 46–7 macroeconomic effects 46–7 portfolio balance channel 44 statement 64

Index 

toolkit 33 money management funds 199 money market funds (MMFs) 2, 80 assets under management (AUM) 194 China, 2003 200 contagion risks 205 crises 207–13 France, 1981 197–8 institutional investors 202–4 Japan, 1992 199 liquidity transformation 194, 201–2 Luxembourg and Ireland, Late 1980s–Early 1990s 198–9 policy implications and 213–14 private money-like assets 202 South Africa, 1995 200 threshold effects 205–6 United States, 1972 195–7 vulnerabilities 202 vulnerabilities threaten financial stability 206 Money Market Mutual Fund Liquidity Facility (MMLF) 90 money markets 2 bidding behavior in 94 mutual funds 90 money reserve fund (MRF) 199 money supply 186 Monte Carlo simulations 344 Mookherjee, D. 129 moral-obligation bonds 303 Morck, R. 69 mortgage-backed securities (MBS) 3, 137, 171, 186, 416 agency RMBS in cross-section 335–7 credit risk 341 duration risk 340 economic effects of 349–51 future research 351–2 international MBS markets 337–8 investors 335 market segments and evolution 332–4 measuring and modeling prepayments 342–3 mortgage securitization 349–51 OAS and 344–6 outstanding 333 prepayment risk 340 S-curve 343–4 security design 338–40 supply effects and Fed quantitative easing 346–7 trading see trading trading and funding liquidity risk 341 trading volume 349 mortgages outstanding 173

519

Mossin, J. 362 multi-dealer platforms (MDPs) 260 multiple-equilibrium game-theoretic model 81 municipal bonds 3, 301, 302, 314 basic features of 302–6 holdings of 316–17 life cycle of 306–8 municipal investors 314–18 municipal issuers 308–14 municipal debt 301 municipal distress and default 308 municipal issuers 308–14 Municipal Liquidity Facility (MLF) 27, 90, 326 municipal markets existing research on 319–26 features of regulation 318–19 Municipal Securities Rulemaking Board (MSRB) 318 municipal tax exemption 301 Munyan, B. 2, 249 mutual funds 156, 359 Naber, J. 223n7 Nagel, S. 246, 248 Nakahama, M. 62 Nakajima, J. 63, 72 Nakamoto, S. 371 Nakamura, E. 501 Nakazono, Y. 67 Nakhmurina, A. 324 National Best Bid and Offer (NBBO) 369 National Central Banks (NCBs) 33 National Housing Act 181n30 National Securities Clearing Corporation (NSCC) 364 NAV rounding 205–6 Nelson, W. R. 2, 94 Net Interest Cost (NIC) method 307 net stable funding ratio (NSFR) 465 Neumeyer, P. A. 390, 392 New York Stock Exchange (NYSE) 360 Niessner, M. 372 Nikolaou, K. 244 Nishino, K. 62, 67 non-bank financial intermediaries (NBFIs) characteristics 148 decentralised finance 158 exchange-traded funds 157 for financial stability 147 hedge funds 157–8 and liquidity demand 155–8 liquidity imbalances 147, 160–63 and market-based intermediation 147, 150–55 money market funds 156–7 mutual funds 156

520  Research handbook of financial markets

policy considerations 163–7 propagation of systemic risks 158–60 systemic risk in 155–8 nonbanks firms 9 intermediation 2 in unusual and exigent circumstances 88–91 nonborrowed reserves 83–5, 223 nondurable (ND) components 482 non-financial customers 255 non-firm liquidity 271 non-fungible tokens (NFTs) 140 non-mortgage US corporate debt 150 no-questions-asked (NQA) property 202 Novy-Marx, R. 323 offer side swap rate 411 OIS swaps 412–13 Okimoto, T. 68, 72 Olijslagers, S. 382 Ongena, S. 135 Ono, A. 71 Oomen, R. 257 open market 16, 98, 107 banking system 82 government securities in 80 Open Market Desk 8 open outcry and electronic trading 491 Opp, C. C. 361 option-adjusted spread (OAS) 340, 344–6 Options Clearing Corporation (OCC) 493 options contracts 497–504 options pricing 495 Orlov, D. 246, 248 Ottonello, P. 392 Overnight Bank Funding Rate (OBFR) 24 Overnight Index Swaps (OIS) 410 overnight reverse repurchase (ON RRP) offer rate 224 over-the-counter (OTC) 253, 409, 464 Owyang, M. T. 180 Paddrik, M. E. 246 Pagano, M. 370 Painter, G. 181 Paludkiewicz, K. 49 Pan, J. 290 Panagiotidis, T. 473 Panizza, U. 387 parameter estimation methods 298 Paravisini, D. 133 Parikh, H. 473 Park, J. 474 Parkinson, P. 165n17 par-swap curve 423

par-swap rate 423 Passmore, W. 2, 173 Pastor, L. 368 Pattipeilohy, C. 67 Paulson, H. 183 Paycheck Protection Program Liquidity Facility 90 payment for order flow (PFOF) 369 Pederson, L. H. 367 Peek, J. 135 Peersman, G. 73 Pelger, M. 363 pension funds 66, 359 Pension Obligation Bonds 304 Pension Protection Act 416 People’s Bank of China (PBoC) 105 Perignon, C. 306n5 Perli, R. 3 Perri, F. 390, 392 Petersen, M. A. 134, 135 Pettit, L. 249 Phelan, C. 392 Piwowar, M. S. 325 plain vanilla fixed-floating swap 410 plain vanilla swaps 410 Plaza Accord 57 Plona, C. 243n7 Png, I. 129 policy discussions 33 implementation 22, 23 interventions 34 rate balances 64 and research 166–7 Polish National Bank 104 political factors 390 Popov, A. 135 portfolio balance channel 18, 268 potential future exposure (PFE) 464 Powell, C. 184 Pozzolo, A. F. 136 Preferred Stock Purchase Agreement (PSPA) 183 Preliminary Official Statement (POS) 307 premium leg 442, 443 Presno, I. 392 price formation 1 pricing 421–3 errors 292 price setting and market efficiency 361–4 and usage 494–7 Primary and Secondary Market Corporate Credit Facilities 26, 90 primary credit program 86, 87 Primary Dealer Credit Facility (PDCF) 89 primary dealers 8, 222, 286–8

Index 

primary vs. secondary market 280–82 prime brokerage 153, 257 principal-agent problem 131 principal trading firms (PTFs) 149, 151, 153, 256 Pringle, J. J. 424 private currencies 141 liquidity funds 214 money-like assets 202 mortgage insurers 174 placement financing 307 placement transactions 307 private-debt instruments 194 productive equilibrium 81 profit-and-loss (PnL) 159 proof-of-stake (PoS) 371 proof-of-work (PoW) protocols 371 protection buyers 436 protection leg 442 protection sellers 436–7 prudential regulation 166 public debate 33 Puonti, P. 73 Putniņš, T. J. 366 Pyle, D. H. 129 Q-JEM 74 qualified mortgages (QMs) 334n7 Quantitative and Qualitative Easing (QQE) 56, 62, 65 Quantitative easing (QE) 17, 17n17, 33, 56, 59 QUICK survey system 67 Quint, D. 47 quote driven 360 Rajan, R. G. 130, 134, 135 Ramírez, C. A. 246 Ranaldo, A. 3, 248, 249, 466n38, 467n45 Rauh, J. 323 Ravn, M. O. 173 real assets 471, 473 real estate 85, 471, 472, 487 real estate investment trust (REIT) 474 recipient central bank collateral 113 historical parallel 113 purpose 111 settlement, frequency, tenor and limits 113 reciprocity 109 Redfearn, C. 181 redundant contracts 431n3 Refinitiv FX Matching 460 refunding transaction 308 regulatory interventions 132 Reinhardt, D. 115

521

Reinhart, C. M. 380, 387 Reis, R. 2, 75, 115, 117, 119, 122, 382 relationship-based lending 134 Rennie, A. 438 repo contracts substitutes for 245–6 variations on 245 repo financing 200 repo lending 91 repo market 238–9 acceptable collateral 240 classic repo transaction 240 default vs. fail-to-deliver 241 growth of 238–9 haircuts 240 invention of 237–8 maturity 239 rate of interest 239–40 regulation and central bank interactions with 249–51 rehypothecation 241 repo vs. reverse repo 241 settlement date 239 structure of 239–41 and systemic risk 248–9 trading repo 246–8 repo pricing 242–4 implied repo rates 243 market segmentation 244 SOFR 244 specialness 244 repurchase agreements (RPs) 8, 134n7 request for quote (RFQ) 260, 460 resale agreements 238 reserve balances 64 Reserve Primary Fund 89 reserve sharing agreements 107n3 residential vs. commercial 332 retail traders 359 Reuters 258 revenue bonds 303 reverse repurchase agreements (RRPs) 10 reverse repurchase facility 224, 238, 251 Richardson, G. 114 Richardson, M. 474 Riddiough, S. J. 463 Rime, D. 3, 466n39 Rindi, B. 361 risk-free counterparty 81 risk-free rates (RFRs) 452, 465 risk-reversals 502–3 Roch, F. 3, 395 Rogers, J. H. 73 Rogoff, K. S. 387 Roldán, F. 3, 395

522  Research handbook of financial markets

Roll, R. 365 Romer, C. D. 67 Roosevelt, F. D. 67 Rösch, D. 365 Rose, J. 223n6 Roseman, B. 365 Rosengren, E. S. 135 Ross, C. P. 248 Rossi, B. 45 Roubini, N. 390 Rouwenhorst, K. G. 473, 499 Sack, B. 2 Safety and Soundness Act 180 Saleh, F. 371 Sambasivam, R. 223n7 Sanches, D. R. 142 Santos, J. A. 137 Sarno, L. 463, 466n39 Saunders, K. T. 424 Schaechter, A. 394 Schaffner, P. 248, 249 Scharfstein, D. S. 115 Schenk, C. R. 107 Schenkelberg, H. 72 Schoenholtz, K. 142, 143 Scholes, M. 495 Schreger, J. 387 Schrimpf, A. 2, 499, 501 Schuerhoff, N. 325 Schwert, M. 321 S-curve 343–4 seasonal credit 88 interest rates 86 seasoned municipal bonds 308 secondary credit interest rates 86 secondary market 296–8 Secondary Market Corporate Credit Facility (SMCCF) 26 Secured Overnight Financing Rate (SOFR) 24, 244, 250, 409 Securities and Exchange Commission (SEC) 360, 493 securities and lending operations 33 Securities Exchange Act 368 securitization 339–40, 350 Sedunov, J. 132 self-regulatory organization (SRO) rules 370 Seligman, J. S. 90 sell side 360–61 Senyuz, Z. 244 Sette, E. 135 settlement risk 270–71 Shang, D. 365 Sharpe, W. 362

Sherlund, S. M. 2, 173 Shiller, R. J. 395 Shin, H. S. 2, 158 Shintani, M. 67, 73 Shioji, E. 63, 71 Shiratsuka, S. 66 Shirota, T. 69 short-term interest 33 Singh, M. 241 Singla, A. 322 single-dealer platforms (SDPs) 260 single monthly mortality (SMM) 342 Singleton, K. J. 425 Skeie, D. 249 Slovin, M. B. 132 small and medium-sized enterprises (SMEs) 134 smooth-transition vector auto-regression model (STVAR) 72 Society for Worldwide Interbank Financial Telecommunication (SWIFT) 141n12 Solnik, B. 425 Soma, N. 67 Somogyi, F. 467n45 Son, B. 363 Song, Z. 347, 348 South Africa 200, 210 sovereign bonds 381 sovereign debt 382 costs of sovereign default risk 388–92 costs of sovereign defaults 385–8 crisis 46, 49, 432 fiscal frameworks 393–4 governments default 388–90 high and volatile sovereign risk 393 markets 3 monetary policy credibility 396–7 restructurings and defaults 383–5 safe sovereign debt 381–3 sovereign bond market basics 380–81 sovereign borrowing 379–80 state contingent debt 394–5 special privileges 184–5 Special Purpose Acquisition Companies (SPACs) 369 speculation 417 Spiegel, M. M. 91 stablecoins 28, 140–43 Stambaugh, R. F. 368 standardized coupons 445, 446 Standing Repo Facility (SRF) 91 state contingent debt 394–5 Stein, J. C. 96, 115, 132 Steinsson, J. 501 Sterling Overnight Index Average (SONIA) 410n4

Index 

Stiglitz, J. E. 96 stochastic discount factors (SDF) 363 stock-market strategies 472 stock portfolios 474 stock prices 75 Stoll, H. R. 495 Strahan, P. E. 350 Sturzenegger, F. 390 Styczynski, M.-F. 223n7 Subrahmanyam, A. 365 Subrahmanyam, M. G. 425 Sudo, N. 74 Sundaresan, S. 425 Supplementary Financing Program 17n16 Sushko, V. 3, 467n46 Sutch, R. 70 Swanson, E. T. 2, 47, 500, 501 swap execution facilities 464 swap pricing 454–5 swap spread 411 Swiss Average Rate Overnight 452 Swiss National Bank 103, 104 syndicated loans 136 Syrstad, O. 467n44 systemic risk decentralised finance 158 defined 155 exchange-traded funds 157 hedge funds 157–8 money market funds 156–7 mutual funds 156 System Open Market Account (SOMA) 8, 277 Szczerbowicz, U. 46 Tachibana, M. 72 Tan, G. K. S. 367 Tanaka, M. 74 tax-exempt market 301 Te Kaat, D. M. 49 Tender Option Bond programs 309 Tepper, A. 466n39 Term Asset-Backed Securities Loan Facility (TALF) 89 Term Auction Facility (TAF) 25, 94, 95 Term Securities Lending Facility (TSLF) 25, 26, 89 Tether 140 Texas Permanent School Fund’s Bond Guarantee Program 305 Thibodeau, T. 180 three-tier system 64 thrift institutions 9n9 thrifts and credit unions 86 Till, H. 491 time-series approach 71–3

523

Tirole, J. 382, 392 Titman, S. 424 to-be-announced (TBA) market 335 Tobin, J. 141 tokenization 371–3 Tokyo Overnight Average Rate (TONAR) 410n4 Tomura, H. 75 Tomz, M. 388 too big to fail (TBTF) 82 Tookes, H. E. 367 total electronic spot trading volume 262 Townsend, R. M. 129 trade shocks 389 trading costs 365–7 equity trading see equity trading trading repo European repo markets 247–8 U. S. repo markets 246–7 transaction lending 132 Treasury auctions 284–6 Treasury bills 17 Treasury Borrowing Advisory Committee 278 Treasury General Account (TGA) 6, 10 Treasury inflation-protected securities (TIPS) 278, 471, 475–6 Treasury market 3 auction process see auction process illiquidity 161 marketable securities 278–9 marketable vs. nonmarketable securities 277–8 market liquidity 282–4 primary vs. secondary market 280–82 Treasury issuance patterns 279–80 Treasury securities 9 and interest rate swaps 412 Treasury’s Housing Reform Plan 189 Trebesch, C. 380, 383, 386 Treynor, J. L. 362 Tristani, O. 2, 47 Troost, W. 114 True Interest Cost (TIC) 307 Trzinka, C. A. 324 Tsatsaronis, K. 248, 249 Tsuruga, T. 62, 74 Tucker, P. 97 Udell, G. F. 134 Ueda, K. 66, 67 Ugai, H. 66, 70 Uhlig, H. 129 Ulyanava, K. 372 unconventional policies 33, 34 underwriting 179 United States 195–7, 208–13

524  Research handbook of financial markets

commercial banking system 127 Commodity Futures Trading Commission (CFTC) 417 debt ceiling negotiations 432 domestic market 2 financial developments 27 government 3 government securities 83 monetary policy 136 mortgage-backed securities 3 prime funds 194 repo market 2 small and medium-sized firms 139 Treasury Department 6 Treasury markets 153, 163, 164f Treasury obligations 296 Treasury securities 8, 308 U. S. Treasury 79 U. S. Treasury bills 131 U. S. Treasury debt 237 USD Coin 140 USD credit markets 103 universal banks 127 University of Chicago Initiative on Global Markets 128 unsecured debt 161 upfront payments 445, 446 Vallee, B. 306n5 Value-at-Risk (VaR) 159 Van Cayseele, P. 134 Van den End, J. W. 67 Van Horen, N. 135, 249 Vardoulakis, A. P. 249 variable net asset values 197 Variable Rate Demand Notes (VRDNs) 306 variation margins (VMs) 160 VARIMA models 480 Vasios, M. 249, 466n38 Vayanos, D. 45, 70 vector autoregression (VAR) 67, 298 Vegh, C. 392 Vendrasco, M. 370 Ventura, J. 381 Verdelhan, A. 466n39, 478 Vickery, J. 3, 173, 350 Victoria, V. 367 Vila, J.-L. 45, 70 Vissing-Jorgensen, A. 347 Viswanath-Natraj, G. 467n44 VIX index 503–4 von Thadden, E.-L. 249 Vuletin, G. 392

wages and house prices 482–3 Walker, M. 246, 248 Wall, L. D. 423, 424 Wallace, N. 44n18 Waller, C. J. 28, 382 Wang, C. 361 Wang, J. 290, 321 Watanabe, K. 59 Watanabe, T. 59 Watzka, S. 72 weekly liquid asset (WLA) thresholds 205 Wei, B. 3 Wei, M. 289 Weidenmier, M. 386 Weiss, A. 96 Weller, B. M. 272 Werner, I. M. 361 Werner, T. 49 Wieland, J. F. 71 Wiggins, R. 90 Wilcox, D. 323 willingness to pay 390 Wiwattanakantang, Y. 69 Wolgemuth, J. 241 Wong, M. 350 Woodford, M. 44n18, 70 World War I 103 Wright, J. H. 47, 297n28 Wright, M. L. J. 380, 388 Wu, L. 495 Xiao, Z. 350 Xiu, D. 363 X-value adjustments 454 Yang, Y. 250 Yermack, D. 372 Yeyati, E. L. 387 Yield Curve Control (YCC) 63, 65 Yoshida, J. 69 Yue, V. Z. 3, 424 Yun, Y. 115 Zakrajšek, E. 2 zero interest rate policy 56, 59, 61, 66 Zervos, S. 128 Zettelmeyer, J. 3, 390 Zhou, N. 370 Zhou, X. 249, 368 Zhu, H. 347, 361 Zhu, J. 363 Zimmermann, K. 387 Zorn, C. K. 318