Devaluing to Prosperity : Misaligned Currencies and Their Growth Consequences [1 ed.] 9780881326512, 9780881326239

Citation preview

D E VA L U I N G TO PROSPERITY Misaligned Currencies and Their Growth Consequences

SURJIT S. BHALLA P E T E R S O N I N S T I T U T E F O R I N T E R N AT I O N A L E C O N O M I C S

D E VA L U I N G TO PROSPERITY M is a lign ed C ur r encies and T he ir Grow th C onsequences

SURJIT S. BHALLA

D E VA L U I N G TO PROSPERITY M is a lign ed C ur r encies and T he ir Grow th C onsequences

SURJIT S. BHALLA

P E T E R S O N I N S T I T U T E F O R I N T E R N AT I O N A L E C O N O M I C S WASHINGTON, DC | AUGUST 2012

Surjit S.Bhalla is chairman of Oxus Investments, a New Delhi–based “hedge fund” and emerging markets advisory firm. He was a research economist at Rand Corporation, the Brookings Institution, the World Bank, and the Policy Group and an economist/strategist at the World Bank, Goldman Sachs, Deutsche Bank, and Oxus Investments. He has served on several committees of the government of India (the capital account convertibility committees of 1997 and 2006 and the advisory committee on secondary markets for the Securities and Exchange Board of India, among others). He was an appointed member of the National Statistical Commission of India (2007–10). He is on the governing board of the National Council of Applied Economic Research, and since 2002, has been a regular invitee to the Aspen Institute Program on the World Economy. Bhalla is author of several academic articles as well as books on globalization and its effects on the world economy: Imagine There’s No Country: Poverty, Inequality, and Growth in the Era of Globalization  (2002),  Second Among Equals: The Middle Class Kingdoms of China and India  (2007, unpublished),  and  The Old Order Is Fading—Or How the Middle Class Is Reshaping the World (forthcoming, 2013). He received his PhD in economics from Princeton University. PETER G. PETERSON INSTITUTE FOR INTERNATIONAL ECONOMICS 1750 Massachusetts Avenue, NW Washington, DC 20036-1903 (202) 328-9000 FAX: (202) 659-3225 www.piie.com C. Fred Bergsten, Director Edward A. Tureen, Director of Publications, Marketing, and Web Development

Printing by Victor Graphics, Inc. Cover design by Sese-Paul Designs Cover photo: © iStockphoto Copyright © 2012 by the Peter G. Peterson Institute for International Economics. All rights reserved. No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, or by information storage or retrieval system, without permission from the Institute. For reprints/permission to photocopy please contact the APS customer service department at Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; or email requests to: [email protected] Printed in the United States of America 14 13 12 5 4 3 2 1 Library of Congress Cataloging-inPublication Data Bhalla, Surjit S. Devaluing to prosperity : misaligned currencies and their growth consequence / Surjit S. Bhalla. p. cm. Includes bibliographical references and index. ISBN 978-0-88132-623-9 1. Devaluation of currency. 2. Economic development—Developing countries. 3. Developing countries—Commerce. 4. International trade. 5. Foreign exchange rates. 6. Monetary policy. I. Title. HG3852.B495 2012 332.4’142--dc23 2012025158

The views expressed in this publication are those of the author. This publication is part of the overall program of the Institute, as endorsed by its Board of Directors, but does not necessarily reflect the views of individual members of the Board or the Advisory Committee.

To All ye faithful members of the Book Club

Contents

Preface

xiii

Acknowledgments

xvii

1 Introduction What Determines Growth: Geography? Technology? Policy? The Primacy of Exchange Rate Policy A Guide to the Book

2 Determinants of Economic Growth The Historical Context Some Explanations of Growth Growth since 1950 Popular Theories Growth Policies: The Washington Consensus

3 Currency Valuation, Savings, and the Current Account Currency Valuation and Savings Current Account Balance and Growth Currency Valuation and the Current Account

1 2 4 5

11 12 12 13 15 29

33 34 37 39

4 Measuring Currency Valuation

41

Equilibrium Exchange Rates Real Exchange Rates Currency Misalignments Real Exchange Rates and Income

41 42 48 52

VII

Different Measures of Currency Valuation Income, Currency Valuation, and Growth: A Review

5 The Yin and Yang of Investment

58 64

67

Currency Valuation and Investment Investment Impact of Currency Devaluation Investment as the Channel of Influence

67 69 72

6 Is the Real Exchange Rate Endogenous?

77

An Endogenous Real Exchange Rate: The Theory The Impossible Trinity Passive Devaluation Changes in Currency Valuation, 1980–2011 Is China a Currency Manipulator In Summary

78 78 85 86 90 91

7 Rashomon Rules: US Dollar, Euro Dollar…

93

US Current Account Deficit Valuations of the Dollar Currency Valuations and the US Current Account Deficit Whither the Dollar?

8 Currency Valuation and Growth Overview Econometric Models of Growth Tests of Simple Growth Models Conclusion

9 Policy Failures and Growth Miracles

94 95 98 108

115 116 117 119 134

135

Growth Successes and Failures Explanations for Failure Structural Breaks in Growth Explaining Growth Acceleration

136 137 139 141

10 Mercantilism and Miracles

149

Defining Mercantilism Mercantilism and Growth Miracle Economics

149 153 153

11 Institutions versus Exchange Rate Policy

165

The Conventional Wisdom New Evidence Institutional Measures Institutions and Growth: Revisiting the Evidence

166 168 168 171

VIII

12 Currency Undervaluation: A Time-Tested Policy for Growth Nineteenth-Century Exchange Rates Tariffs and Growth The Yen Exchange Rate in 1950

13 Economics of the Yen and the Renminbi Japan in the 1980s and China in the 2010s Déjà Vu? China Is Different Revisiting Paul Samuelson

14 Changing Times, Changing Views Currency Wars Why Currency Undervaluation? Evidence of Malfunctions and Imbalances The Seductive Appeal of Currency Undervaluation One Country’s Ceiling Is Another Country’s Floor—Evidence on Stolen Growth Breaking the Cycle

15 Conclusion Currency Undervaluation Affects Investment and Generates Growth The Dual of Currency Undervaluation is Currency Overvaluation The Real Exchange Rate Can Be Influenced by the Nominal Exchange Rate The Real Exchange Rate Can Be Influenced by Standing Still Mercantilism is Alive and Well There Are Parallels between 1870 and 1950 Institutions Don’t Rule A Postcrisis Realignment

179 180 183 186

189 189 191 196 205

209 211 211 212 216 219 221

225 225 226 227 227 228 228 228 228

Appendix A Data and Methods

233

Appendix B Bhalla (2007a) Dataset Extended to 2011

239

References

247

Index

258

Tables 2.1 2.2 2.3

Global population and income, 1950–2011 Average (log) growth in per capita income by region, 1951–2011 Running out of good luck? Growth persistence, 1960–2011

14 16 28

IX

2.4 3.1 3.2 3.3 4.1 4.2 4.3 4.4 4.5 5.1 6.1 6.2 7.1 7.2 7.3 7.4 7.5 7.6 7.7 8.1 8.2 8.3 8.4 8.5 8.6 8.7 9.1 9.2 9.3 10.1 10.2 10.3 10.4 X

Policies of the Washington Consensus Effect of currency valuation on economic variables, 1950–2011 Current account balance, savings, and currency valuation for selected countries, 1970–2011 Effect of currency valuation on the current account balance, 1965–2011 Estimates of currency misalignments, 1960 Estimating the real exchange rate: Methods and findings Currency valuation and growth: Methods and findings Currency valuations for selected countries, 1970–2011 Currency valuation estimates, 2011 Effects of currency valuation on investment Direct and indirect (standing still) estimates of real currency devaluations, 1980–2011 Different estimates of (log) changes in currency valuation, 1980–2011 Dollar valuation by different measures, 1960–2011 Correlation between different currency valuation measures for the US dollar, 1993–2011 Explaining the US current account balance: Log (exports/ imports), 1978–2006 Explaining the US current account balance: Percent of GDP, 1978–2006 The euro area: Current account balances and related indicators, 1991–2008 Currency valuation and current account balance for the euro area, 1980–2011 Toward an adjustment of the US dollar, 2011 Comparison of currency valuation estimates for selected countries, 1970–2011 Estimating the impact of currency valuation on the growth of income per capita, 1950–2011 Two-stage least squares regression result Currency valuation and growth: Different econometric methods Determinants of growth, fixed effects model, 1950–2010 Do outliers matter? Regional growth and regional currency valuation, 1980–2011 Currency valuation and growth accelerations and decelerations, 1950–2011 Explaining growth accelerations, different definitions, 1950–2011 Is there a middle income trap? Mercantilism rankings for selected countries, 1990–2011 Mercantilism and growth, 1950–2011 Growth in GDP and factors of production, 1960–2011 Miracle growth, 1951–2011

30 36 38 40 47 50 51 60 61 75 87 89 97 98 99 99 105 107 110 121 122 124 126 127 129 133 142 144 145 151 154 155 157

10.5 11.1 11.2 11.3 11.4 11.5 11.6 12.1 12.2 12.3 12.4 12.5 13.1 13.2 13.3 14.1 14.2 14.3 14.4 14.5 A.1 B.1

Miracle growth, 1960–2011: Results (and associated data) Institutions and growth, 2010 and 1980–2010 Importance of institutions in explaining growth Importance of instruments in explaining growth Currency valuation versus institutions Average values and significance of currency valuation in the presence of institutions Role of institutions in growth Real exchange rate and currency valuation, 1870–1950 Tariffs and currency valuation, 1870–1913 Tariffs, currency valuation, and growth, 1870–1913 Tariffs, currency valuation, and growth, 1913–38 How the rate of 360 yen to $1 was selected in 1950 Japan and China: No déjà vu The big debate: The Chinese renminbi is not undervalued The big debate: The Chinese renminbi is undervalued China and India: Education, income, and wages, 1985–2010 How costly is foreign reserve accumulation? World crisis foretold through world currency valuations Effect of negative currency valuation on world current account balances, 1980–2011 Is there any evidence of stolen growth? Country composition in Bhalla (2007a), gross country panel dataset, 1950–2011 Country data, 2011

162 172 175 175 176 177 178 182 184 185 186 187 193 198 201 213 215 218 219 222 237 240

Figures 2.1 4.1 4.2 4.3 4.4 4.5 5.1 5.2 5.3 6.1 7.1 7.2 7.3

Growth in India according to the labor reallocation theory, 20 1960–2010 Real exchange rate and income per capita, 2011: An S-shaped 55 relationship Elasticity of real exchange rate to income per capita, by income 57 group, 1950–2010 How accurate are the 1980 currency valuation measures? 62 How accurate are the 2011 currency valuation estimates? 63 Real exchange rate currency valuations for selected countries, 2011 64 Investment and currency valuation: A strong relationship, 73 1960–2011 Investment growth compared with initial currency valuation 75 Investment growth compared with average change in currency 76 valuation Evolution of actual and predicted real exchange rates, 1960–2011 82 Three estimates of the real value of the US dollar, 1973–2011 101 Explaining the US current account deficit, 1990–2011 102 Explaining US current account imbalances, 1978–2011 103 XI

7.4 8.1 8.2 9.1 9.2 10.1 10.2 13.1 13.2 14.1 14.2

XII

Does currency valuation matter for current account balance? Core and peripheral economies of the euro area, 1993–2011 Distribution of currency valuation, 1950–2011 How does currency valuation impact growth? Added variable plot: Growth acceleration and initial level of currency valuation Added variable plot: Growth acceleration and average change in currency valuation Doubling per capita income, helped by currency undervaluation Added variable plot: Number of years taken to double per capita income and average value of currency valuation Japan: Evolution of per capita income and currency valuation, 1970–85 China: Evolution of per capita income and currency valuation, 1991–2006 World real interest rate, 1950–2011 World current account balances affected by currency valuations, 1980–2011

107 130 132 146 147 159 160 194 195 217 220

Preface

The last three decades have witnessed high economic growth in developing countries, a widening of global imbalances, and a sharp increase in reserve accumulation, especially among high-growth Asian economies. This book argues that these events are strongly linked via a consistent policy of currency undervaluation in Asian economies. World or regional currency crises have occurred with increasing frequency in the last few decades. The book also argues that the Mexican peso crisis of 1994, the East Asian currency crisis of 1997, the US financial crisis of 2006–08, and the euro area crisis of 2010–12 all relate to currency misalignment, the investigation of which is the major task this book has set for itself. Simultaneously, the world has witnessed a radical transformation of China. From being one of the poorest countries just 30 years back, China is today arguably the largest country in the world in terms of purchasing power parity (PPP) output. Its role in the global economy has been linked, at least in part, to the manner in which it has set the exchange value of its currency. The demand for significant revaluation of the Chinese currency has been near universal in recent years. In Europe, a common currency, the euro, brought large benefits to the member economies. However, this was accompanied by a significant cost in terms of loss of competitiveness for some euro area countries, a loss that precipitated large current account imbalances and hence the ongoing euro crisis. Traditionally, the view has been that overvalued currencies hurt economic growth, but undervalued currencies cannot help in growth acceleration. There has also been a parallel belief: The real exchange rate—that is, a country’s competitive ranking—cannot be affected by merely changing the nominal xiii

exchange rate. This view is grounded in the thesis that inflation follows any devaluation of currency and hence the real exchange rate remains unaffected. Against this background, this book examines the veracity of various propositions relating to currency misalignments and their effect on various items of policy interest. It exposes methodological limitations caused by measurement problems in estimates of currency undervaluation in previous analyses of exchange rate management and growth. It does so by subjecting more than a century of global exchange rate management and growth outcomes to rigorous empirical analysis. The main conclusions are that the real exchange rate can be affected by nominal devaluations and currency undervaluation does help growth but the global economic system becomes unstable if too many countries, and especially large countries, pursue this beggar-thy-neighbor approach. The Peter G. Peterson Institute for International Economics is a private, nonprofit institution for the study and discussion of international economic policy. Its purpose is to analyze important issues in that area and to develop and communicate practical new approaches for dealing with them. The Institute is completely nonpartisan. The Institute is funded by a highly diversified group of philanthropic foundations, private corporations, and interested individuals. About 35 percent of the Institute’s resources in our latest fiscal year were provided by contributors outside the United States. The Institute’s Board of Directors bears overall responsibilities for the Institute and gives general guidance and approval to its research program, including the identification of topics that are likely to become important over the medium run (one to three years) and that should be addressed by the Institute. The director, working closely with the staff and outside Advisory Committee, is responsible for the development of particular projects and makes the final decision to publish an individual study. The Institute hopes that its studies and other activities will contribute to building a stronger foundation for international economic policy around the world. We invite readers of these publications to let us know how they think we can best accomplish this objective. C. FRED BERGSTEN Director July 2012

xiv

PETER G. PETERSON INSTITUTE FOR INTERNATIONAL ECONOMICS 1750 Massachusetts Avenue, NW, Washington, DC 20036-1903 (202) 328-9000 Fax: (202) 659-3225 C. Fred Bergsten, Director BOARD OF DIRECTORS * Peter G. Peterson, Chairman * George David, Vice Chairman * James W. Owens, Chairman, Executive Committee Leszek Balcerowicz Ronnie C. Chan Chen Yuan * Andreas C. Dracopoulos * Jessica Einhorn Stanley Fischer Arminio Fraga Jacob A. Frenkel Maurice R. Greenberg Herbjorn Hansson Tony Hayward * Carla A. Hills Yoshimi Inaba Karen Katen W. M. Keck II Michael Klein * Caio Koch-Weser Andrew N. Liveris Sergio Marchionne Donald F. McHenry Indra K. Nooyi Paul O’Neill David J. O’Reilly Hutham Olayan Peter R. Orszag Samuel J. Palmisano Michael A. Peterson Victor Pinchuk Lynn Forester de Rothschild * Richard E. Salomon Sheikh Hamad Saud Al-Sayari Edward W. Scott, Jr. *Lawrence H. Summers Jean-Claude Trichet Laura D’Andrea Tyson Paul A. Volcker Peter Voser Jacob Wallenberg Marina v.N. Whitman Ronald A. Williams Ernesto Zedillo

ADVISORY COMMITTEE Barry Eichengreen, Chairman Richard Baldwin, Vice Chairman Kristin Forbes, Vice Chairwoman Isher Judge Ahluwalia Steve Beckman Olivier Blanchard Barry P. Bosworth Menzie Chinn Susan M. Collins Wendy Dobson Jeffrey A. Frankel Daniel Gros Sergei Guriev Stephan Haggard Gordon H. Hanson Takatoshi Ito John Jackson Peter B. Kenen Anne O. Krueger Paul R. Krugman Justin Yifu Lin Jessica T. Mathews Rachel McCulloch Thierry de Montbrial Sylvia Ostry Jean Pisani-Ferry Eswar S. Prasad Raghuram Rajan Changyong Rhee Kenneth S. Rogoff Andrew K. Rose Fabrizio Saccomanni Jeffrey D. Sachs Nicholas H. Stern Joseph E. Stiglitz William White Alan Wm. Wolff Daniel Yergin Richard N. Cooper, Chairman Emeritus

Ex officio * C. Fred Bergsten Nancy Birdsall Richard N. Cooper Barry Eichengreen Honorary Directors Alan Greenspan Lee Kuan Yew Frank E. Loy David Rockefeller George P. Shultz

* Member of the Executive Committee

XV

XVI

Acknowledgments

My obsession with the workings of exchange rates goes way back. As a student in the United States from a developing country, I was sensitive to, and all too painfully aware of, official and purchasing power parity exchange rates. I realized that even more when, coming back to India in the early 1980s, my salary took a fifteen-fold decline. But my productivity did not change, or so I felt. I was doing exactly the same work as I did in the United States. So how could my productivity change just by transplantation into India? It could not and it did not; but my real wage declined big time. Therein lies the genesis of one of the key factors behind the development of this book’s thesis. On September 18, 1992, just two days after “Black Wednesday,” and working under pressurized conditions, I concluded that the British devaluation could have been predicted on the basis of good old-fashioned “productivity-adjusted purchasing power parity economics”—the same reasoning about the difference between productivity and cost noted above. This model also worked ex ante in the Mexican (1994) and Thai (1997) devaluations. While it might seem that I am following the dictum of my professor Ray Fair too faithfully—he said that as a forecaster one should forecast often and always remind people when one is right—the point about the story is that exchange rates, both as an economist and as a trader, have been my major preoccupation for the last 25 years. My first-ever trade was not the stock market but trading currencies, dollars into yen or Deutsche mark, etc. The gains and losses from currency trading, I would like to think, have helped fine-tune my understanding of exchange rates and their role in generating or inhibiting economic growth. No intellectual effort is possible without a lot of help—and I am fortunate to have friends, the usual suspects, who have been ever generous with their time and intellect in contributing to the rejection, and sometimes acceptance, xvii

of back-of-the-envelope findings and ideas. These include Shankar Acharya, Montek Ahluwalia, Suman Bery, Ravinder Kaur, Farrukh Iqbal, Robert Lawrence, and Arvind Virmani. Thanks also to seminar participants at various places where the findings of this book were presented, from an ICRIER seminar entitled “Chinese Mercantilism: Currency Wars and How the East was Lost” in 1999 to an IMF seminar entitled “There Are No Growth Miracles” in 2007 to a World Bank-New York University 2010 presentation “The Euro and the Yuan—Different Peas in the Same Pod.” These are part of a more than a decadelong quest to understand the workings of exchange rates and the dynamics of economic growth. Seminars at the Peterson Institute for International Economics in 2005, 2006, and 2007 were also useful in generating some ideas and cleaning up others. Richard Cooper, Don Hanna, Josh Felman, Simon Johnson, Vijay Joshi, Lant Pritchett, and an anonymous referee provided useful comments. Also encouraging were discussions with Abhijit Banerjee in early 2007 on the role of my definition of the middle class and the controversial result that institutions did not seem to affect growth. Paul Wachtel and Olivier Jeanne provided extremely useful and constructive comments—a role beyond the call of duty of any editor/referee. Thank you. The penultimate chapter on how the currency undervaluation movie ends was suggested by Stan Fischer and I am extremely grateful for this intervention. Finally, my family—Ravinder, Simran, and Sahil— have in no small way contributed to my sanity through this seven-year effort. The team at Oxus Investments was ever willing to work on this data- and time-intensive book; no religious holiday was sacrosanct, no time was over time. Thanks to all, and especially Tirthantomoy Das, for giving a special meaning to “research assistance.” Finally, thanks to C. Fred Bergsten for comments on various drafts and for being the policy wonk that all of us aspire to be. SURJIT S. BHALLA July 2, 2012

xviii

1 Introduction

Armaments, universal debt and planned obsolescence— those are the three pillars of Western prosperity. —Aldous Huxley, Island Times have changed. This crisis is different. It is impossible to exaggerate the deep funk into which the world economy sank with the financial crisis of 2008. Or to exaggerate how different things have become in its wake. The transformation started with the financial crisis of 2008, metamorphosed into a currency war during 2010, and shape-shifted into a European identity crisis during 2011. These crises have taken a toll on all the world’s economies, developed and developing, and have brought the world to a crossroads. The future is more uncertain than at any time since the Great Depression and World War II. As during the Great Depression, Western nations have lost the most ground economically, and there is palpable fear that they may lose much more. In contrast, emerging-market and developing economies, led by China, contain large pockets of prosperity, although there is uncertainty about the future in these quarters too, particularly among members of their large new middle classes. It seems that globalization, which increased growth for all economies during the past two decades, especially for emerging-market economies, may have carried a hidden cost for the developed world. To determine whether this is true, and how best to advance balanced global growth in the future, it is imperative that policymakers correctly understand the determinants of the ongoing crises. The story is relatively simple. It begins with growth, although it also involves inequality and the pain endured by people whose incomes have stagnated and whose jobs have been lost or are in jeopardy, as underscored by the Occupy protests on Wall Street and elsewhere during 2011. But growth is at the heart of the story. Without growth, there is equality only for poverty, not income. Before the Great Recession of 2008, the world enjoyed a several decades– 1

long period of unparalleled prosperity. The number one economy, the United States, grew at more than 2 percent per capita per year for more than 60 years. Developing economies in Africa and Latin America experienced a long period of stagnation from 1980 until just after 2000 but had begun to improve and were growing at high rates leading into the crisis. Eastern Europe boomed after the fall of the Soviet Union in 1989. The emerging-market economies of Asia performed consistently well during the past 30 years, despite some domestic and international crises. Among these, China performed superbly since the start of economic reform in 1978. There were signs of danger during the years before the crisis. For example, China strictly followed, and nearly perfected, the strategy of devaluing its currency in order to improve its competitiveness and achieve previously unimagined prosperity. In retrospect, the sheer size of China’s economy, its outsize growth, and the growing global imbalances that resulted should have predicted the world crisis, particularly because the Asian financial crisis a decade earlier had its roots in similar circumstances. More attention should have been paid to these early warning signs, but boom times rarely engender wise insights. The precipitating event for the global crisis was the subprime mortgage crisis in the United States. There are two competing perspectives on the causes of the US mortgage crisis. One contends that overly energetic US consumers took on excessive debt to pay for cars, homes, and other trappings of an illusory prosperity. The other contends that an oversupply of goods and capital from the developing and emerging-market economies, primarily China, artificially lowered prices in the United States and eventually caused consumers to make irrational economic decisions. Almost the same explanation is offered for the euro area sovereign debt crisis. One anecdotal story is about a Greek businessman who always drove a Volkswagen and changed his car every three years. In 2001, he went to his dealer to get the latest model, and was told that, with the same monthly payments, he could now afford a Mercedes. With the introduction of the euro, interest rates were the same for a saver living in the fast-growing German economy and a spender in the slower-growing Greek economy. The Greek businessman drove away with a Mercedes because he could borrow at 3 percent instead of the 7 percent rate he faced before the euro. As this book explores, the German-Greek story is not really different from the Chinese-American story. At the root of both stories lie disparate rates of economic growth.

What Determines Growth: Geography? Technology? Policy? Countries’ economic growth experiences vary widely, as do interpretations of the reasons behind them. Attempting to understand the fundamental determinants of economic growth has preoccupied the minds of economists and policymakers for close to a millennium. The Industrial Revolution from the 1790s to the 1860s brought the first significant departure from the static economic 2

DEVALUING TO PROSPERITY

dullness of the previous several hundred years, launching the first period of strong and persistent economic growth. It was fueled by technological progress—faster and cheaper methods of extracting the same from less. It is easy today to take for granted the recent breadth and speed of technological progress, but technological discoveries were hard to come by just a few hundred years ago. The technological progress during the Industrial Revolution and the resulting productivity increases provided the first firm basis for studying economic growth. Many contend that generating growth is more a matter of art than science, but a rich and abundant literature examines the many determinants of growth, some exogenous and some a matter of policy. The pioneering work of Angus Maddison (2001, 2003) allows us to probe the determinants of growth across space, time, and technology. His extensive collation of data has allowed researchers to examine the role of geography, institutions, and policy. Is there a magic formula? Of course not. In fact, it is risky and inaccurate to attach too much weight to any single determinant. But Maddison’s data do suggest some broad generalizations that are especially relevant today. For example, institutions are vastly overrated as an explanation for differences in income levels. The role of institutions in promoting growth has been a recent focus of economists, who tell us that countries are rich and poor depending on whether they have institutions that protect property rights or institutions that promote democracy. But it could be the case that institutions are a luxury service—that is, wealth and good institutions go together because a rich country demands good institutions—and so whether a country is rich or poor today may have relatively little to do with whether it had good institutions yesterday. Geography is the oldest known factor in economic growth and therefore the oldest explanation of why some countries are rich and others poor. Before the Industrial Revolution, we were all equal—first as hunters and gatherers, then as farmers and pastoralists, crop yields or herd sizes depended exclusively on soil quality, rainfall, and agricultural fertility. One country may have been richer than another, but the gap between the two remained the same over time. According to Maddison’s data (2003), in 1500 the rich countries in the world (those with incomes in the 95th percentile and above) had only three times the income level of the poor countries (those in the 5th percentile and below). This remained true as late as 1820. But by 2011 the gap had widened significantly: Rich countries had more than 30 times the income level of poor countries. In short, the world was a lot more equal in 1500, but it was a lot poorer as well. After the Industrial Revolution, technological change played a greater role in economic growth. Then came two World Wars, with the Great Depression in between, and the focus turned to the role of government in economic growth. The market had not quite worked—business cycles were tolerable, but cycles that led to deep economywide failures were not. There had to be another way, and John Maynard Keynes provided an alternative in the form of “government intervention,” also generally referred to as “policy.” John Williamson (1990) INTRODUCTION 3

outlined the standard set of ten policies that were judged to enhance longterm economic outcomes and were generally prescribed for countries facing economic crisis, particularly in the developing world, which he memorably dubbed the Washington Consensus. For several years, most economists and policymakers swore by that consensus, though questions were later raised about the efficacy of these policies. Indeed, some argue that the Washington Consensus never worked in the first place (Rodrik 2006). However, this book argues that the Washington Consensus remains alive and well and a very effective—if not politically correct—guide to fostering economic growth.

The Primacy of Exchange Rate Policy Since the Keynesian argument for government involvement in economic affairs was broadly accepted 60 years ago, several broad policies and many more specific policy inventions have been pursued. There has been enough variety in policy to suit every intervention palate, including Keynesianism and monetarism, import substitution and export promotion, nationalization and openness, and infrastructure investment and enhanced social spending. Overall, however, regardless of the policies pursued at a given time or place, most if not all economists agreed that there were good and bad policies and that economic growth was a function of the policy mix in place. Trade policy has been a favorite. How open should an economy be to trade? Should exports be subsidized? How should importers be taxed? Most every imaginable policy has been followed, often to satisfy the priorities of a particular interest group including both exporters and importers. Quite often in the past policies were tailored to the needs of import substituters, who did not want foreign competition. This book makes a case for the reexamination of the opposite policy, export-led growth, a politically correct version of exchange rate “management.” It does so by exploring the critical role in economic growth of currency undervaluation, the real variable behind policies ostensibly directed at export-led growth. This book outlines how this key variable affects exports, trade, and economic growth by giving authorities the ability to improve their economy’s competitiveness by reducing the production costs of their exports and making them cheaper in foreign markets. This book is based on an analysis of more than 180 countries over some 500 years, but especially the last 150 years. The analysis shows that real exchange rates play a critical role in economic growth and that the strongest asset for growth is a weak currency—a currency weaker than the “unified and competitive exchange rate” prescribed by the Washington Consensus. In short, real currency devaluations can help a country grow toward prosperity. It is widely accepted that currency overvaluation hurts growth, but there is very little theoretical or empirical acceptance of the mirror-image effect, namely, that currency undervaluation helps growth. This book examines the reasons behind this anomaly. One reason is that, while the effects of currency overvaluation are well known and easily observed, the same is not true for the 4

DEVALUING TO PROSPERITY

effects of undervaluation. For example, the East Asian success story during the 1980s and early 1990s was largely attributed (including by the World Bank) to either intelligent state intervention or effective export promotion, but little attention was paid to what may have been the dominant cause: real currency depreciation. Verifying the potential growth role of currency depreciation is made difficult by both theoretical and empirical considerations. Theoretically, it is not possible for changes in a nominal variable like the exchange rate to affect a real variable like growth. Empirically, there are two major difficulties. First, there are large methodological errors in measuring currency under- or overvaluation, and these result in only weak, if any, noticeable effects on growth. Second, as shown recently by Michael Woodford (2009), errors (“construction bias”) in models used to study the relationship between currency valuations and growth may preclude the type of analysis that would show clear evidence of a significant relationship between the two.

A Guide to the Book Chapter 2 examines the evolution of theories on economic growth over time. In particular, it reviews the seminal contributions of a pioneer in the field, Sir Arthur Lewis (1955), who established four key principles of growth that continue to underpin the literature in the field, including the importance of a key factor in determining growth during the early stages of development—the reallocation of labor from low-productivity agriculture to high-productivity industry. Chapter 3 explores the role of savings and the current account balance. The Asian crisis of 1997–98 was precipitated by a current account deficit in Thailand that exceeded 8 percent of GDP. In 2006, two years before the Great Recession that started with the failure of Lehman Brothers, US policymakers started to worry about a deficit that was approaching 6 percent of GDP—and the mirror image of this deficit, a household saving rate near zero. In early 2012, there were large debts and current account deficits in several euro area economies, including Greece, Ireland, Italy, Portugal, and Spain. The real exchange rate measures the relative price of traded goods, and therefore should be an important element of such current account imbalances. In fact, if currency valuations affect growth, they should exert an even greater effect on the current account and saving rates. Chapter 3 tests and confirms this hypothesis. The chapter also discusses the link between current account surpluses and growth. It shows that current account surpluses do not necessarily lead to higher growth, as some argue. Instead, the causal link is from currency undervaluation to higher growth: If policy interventions are used to maintain the currency at an undervalued level, then current account surpluses are a necessary consequence. The contention that current account surpluses “cause” high growth is proven false. INTRODUCTION 5

Chapter 4 reexamines the question of how best to measure currency valuation. Given the widespread rejection of the hypothesis that undervaluation affects growth, and the fact that this study finds not only positive effects but also large and significant effects, it is important to compare and contrast the empirical foundation of both the old and new approaches to measuring currency valuation. A primary reason the analysis in this book shows strong effects is the correction of the “measurement errors” that plague many existing currency valuation measures. The chapter documents how existing methods for measuring equilibrium exchange rates (EERs) produce very large errors—in fact, the explanatory power of the traditional method is less than 50 percent. And the chapter introduces a new, nonlinear functional relationship between income and real exchange rates (RERs) that suggests an S-shaped relationship between the two and explains more than 86 percent of the variation in RERs between 1996 and 2011—a near perfect fit.1 Greater precision in measuring equilibrium exchange rates means greater precision in measuring any deviation from such equilibrium (currency undervaluation or overvaluation), which in turn leads to a better understanding of the role of currency undervaluation in generating growth. Chapter 5 explains the mechanisms by which currency valuation affects economic growth. The primary channel is investment: Currency undervaluation raises investment profits by helping to reduce labor costs. The analysis here follows the so-called reduced-form link between undervaluation and growth, but the emphasis is that the link occurs because currency undervaluation affects production costs and thereby profits and investment. In addition, currency undervaluation can be expected to affect the efficiency of investment, because currency undervaluation often spurs foreign investment, a process that is theoretically modeled and empirically verified in the chapter. Chapter 6 examines the arguments of the critics. Use of the RER in models of growth has been criticized on theoretical and empirical grounds. The contention is that the RER is endogenous or set within a broader economic system characterized by the mutual interdependence of growth, inflation, exchange rate change, fiscal deficits, and more. This argument derives support from the Impossible Trinity argument, which holds that a country cannot simultaneously have a free flow of capital, an independent monetary policy, and a freely floating exchange rate. Why? Because one affects and neutralizes the other. For example, large capital flows emanating from an open capital account will appreciate the RER and prevent any real devaluation. Hence, one cannot really devalue the currency and benefit from higher growth. The Impossible Trinity argument holds in theory but not in reality, certainly not in emerging-market and developing economies, which are the main focus of those who argue that the RER can be affected by changing the nominal exchange rate. Chapter 5 presents data for both developed and developing and emerging-market economies showing that the Impossible Trinity 1. I first offered this relationship in Bhalla (2007a).

6

DEVALUING TO PROSPERITY

rarely occurs. The occasions when it holds in practice can be considered “black swan” events (Taleb 2010), that is, they are highly improbable, unexpected, but have a major impact. Chapter 7 is devoted to “smell tests,” which are meant to differentiate between those measures and models that yield “sensible” results about currency valuation for different countries and those that do not. The first and most important smell test is an evaluation of how each measure explains the value of the US dollar—the numeraire for all calculations of purchasing power parity and hence currency valuation. A particular measure passes the smell test if it explains the performance of the US current account, one of the most important and controversial components of current global imbalances. My currency valuation measurement for the United States (Bhalla 2007a) explains the movements in the US current account over the last 30 years much more closely than the official trade-weighted value of the dollar (the Federal Reserve’s Broad Index of currencies of the 37 major US trading partners). No other popular measure of the interplay between currency valuation, RERs, and growth uses this type of dollar valuation exercise as a test. If such a measure cannot adequately explain movements in the US current account, can it be trusted to explain growth in the developing world? The chapter also looks at the likely future direction of the dollar exchange rate, which is central to the world financial system, and predicts depreciation against Asia (especially China) and steadiness and appreciation versus the euro and yen. In fact, this forecast will need to become reality for global imbalances to be corrected. Chapter 8 examines the hypothesis that policies that promote cheap and cheaper currency can lead to high and higher growth. It finds that this single policy can change a country’s destiny, more so than the 10 policies constituting the Washington Consensus or the 20 policies in its augmented version. This single policy worked more than 200 years ago and appears to work as reliably today: produce a product cheaply and conquer world markets. Some consider this trade-oriented use of currency undervaluation to be a pure expression of market economics; others see it, when pushed to its extremes, to be synonymous with mercantilism. The chapter explores several popular determinants of growth (including demography, institutions, and openness) using conventional models for several time periods, most exhaustively for the period 1950–2011. The individual effect of each popular determinant is examined in isolation and in combination with two currency valuation variables: the initial level currency valuation, at the beginning of any growth period, and the average percentage change in valuation during the period in question—two variables that recur throughout this book. Conventional methods use only one currency valuation variable—the mean level over the time period under analysis—and the chapter also examines these results. Chapters 9 and 10 present additional evidence to substantiate the simple conclusion that currency valuation can explain a large proportion of growth accelerations (and decelerations) discussed in the literature, as well as a large INTRODUCTION 7

proportion of unexplained “miracle” or black swan growth. Even though there are relatively few such growth miracles, one surprising example may well be the United States during the long post–World War II period. When practiced in the extreme, currency undervaluation is mercantilism, roughly defined as a system of political economy based on policies to maintain a favorable trade balance.2 Chapter 10 presents an index of mercantilism that attempts to differentiate between current account surpluses resulting from high savings generated by cultural and historical factors and those resulting more from currency undervaluation. If a country scores high on both factors, then it can be considered mercantilist as defined by this index, and crosscountry tests confirm the importance and significance of this new variable. Chapter 11 conducts several tests of the hypothesis that good institutions equal good growth. Some tests in the chapter are identical to those already presented in the literature, while others are new. Since the ascendancy of institutional explanations of growth, there has been little emphasis not only on geography, culture, and religion as growth determinants but also on traditional policy explanations. The Great Recession may have caused a resurgent belief in the efficacy of policy, but it is not clear how long this will last. Anticipating the results, when a horse race is run between institutions and a policy of currency undervaluation, there is no contest. Policy wins overwhelmingly. Chapter 12 examines whether development patterns since the mid-19th century reveal any common threads. They do. It may well have been a weak currency during the 19th century that set the stage for rapid growth in the United States and Europe during the Industrial Revolution and through the end of World War II. In other words, a weak currency can lead to faster growth. The chapter also provides an explanation for MacArthur’s choice of 360 yen/ dollar as Japan’s fixed exchange rate after the War. Chapter 13 addresses the evolution of thought on the question of whether the Chinese renminbi is, or has been, substantially undervalued. The last section of the chapter draws an analogy to Paul Samuelson’s (1964) seminal article on the RER and the presumed overvaluation of the dollar in the early 1960s. If currency undervaluation is so beneficial, why isn’t everybody practicing it? Chapter 14 addresses this important question. The answer is that virtually every country has done so. The global net currency valuation level moved from an average overvaluation of about 10 percent for 1970–2000 to an average undervaluation level of 11 percent for 2010–11. This chapter also examines the rising global imbalances, especially whether currency changes in the dollar, euro, and renminbi affect the growth of other countries—that is, is currency undervaluation a zero-sum game? The chapter also examines, at a micro level, wages of Chinese and Indian workers. Somewhat surprisingly, for most of the last three decades, Chinese wages have been below Indian wages although theory predicts they should have been substantially higher, which 2. One possible title for this book was The Discreet Charm of Mercantilism.

8

DEVALUING TO PROSPERITY

offers further evidence for the fact that the Chinese exchange rate has been substantially undervalued for most of the last 30 years. Hence, the popular puzzle about why China has grown so much faster than India has an answer: a significantly more undervalued exchange rate. The chapter also assesses the effects of recent global crises on the evolving world financial order, notably the increase of global cooperation toward the goal of greater growth for all. And why the future is likely to have considerably less imbalances in currency values and therefore less imbalances in global trade. Chapter 15 summarizes the major conclusions of the book. Institutions are overrated for their effect on economic prosperity. The real exchange rate is not endogenous but can be influenced by changes in the nominal exchange rate. Currency undervaluation can and does affect growth. Taken to an extreme, currency undervaluation is mercantilism. When a large set of countries, or large countries practice mercantilism, a financial and/or a currency crisis is the inevitable outcome. Optimistically, however, having faced three major crises over the last 15 years—the East Asian crisis, the Great Recession, and the euro area crisis—the world is poised for more collaborative currency arrangements and a better globalized future.

INTRODUCTION 9

2 Determinants of Economic Growth We shall not cease from exploration, And the end of all our exploring Will be to arrive where we started And know the place for the first time. —T. S. Eliot, Little Gidding, Four Quartets The level of income in an economy at any point in time represents the accumulated growth in incomes over time, so investigating what produces higher incomes is really investigating the determinants of economic growth. But that investigation is complicated by the fact that country experiences with growth are enormously varied and often confusing. Numerous potential growth determinants have been identified over the years, but mapping reliable channels of growth has been a major problem for analysis. Economic outcomes are often confounded by many causes, and more explanations have been offered for per capita income as an outcome than there are economists, sociologists, and political scientists, not to mention politicians and policymakers. What determines your relative welfare today? Religion? The weather? Your culture? Does democracy promote economic, social, and political progress? Or is democracy the ultimate luxury good, desired by those who can afford it? What further confounds the search for channels of growth is that what matters for growth differs over time. In the 16th century, geography likely played a critical role. For example, territories with access to the sea fared better than landlocked areas, as did those in a Mediterranean, agriculture-friendly climate. In the 19th century, the strongest determinant became the ability to capitalize on technological change—that is, the capacity to partake in and profit from the Industrial Revolution. After World War II, reconstruction and American aid likely provided the primary impetus for growth. And since 1980, the approximate start of the era of globalization, catching up with the technological frontier is likely the major reason developing economies have grown at rates of more than 5 percent. But that is all conjecture. In the end, the question remains the same: Is there one basic set of causes, levers that can be pulled, to enhance growth? 11

The Historical Context Perhaps exploring the history of the economies that are developing rapidly today can provide clues to the mystery of growth? In 1500, the average incomes of India and China were equal to the world average. By 1960, these societies had been considered doomed to extreme poverty. And in 1980, their average incomes stood at one-fifth the global average. But by 2020, they are expected to again match the global mean.1 The ground these economies lost over 450 years, they seem poised to gain back in 45. That is both the beauty and the beast of economic growth. When the mid-18th century heralded the arrival of the Industrial Revolution, why was the West able to grasp it first, and why did the two great civilizations of China and India go into a tailspin and stay there long thereafter? Some argue, with facts and conviction, that colonialism intervened to stymie global progress and development. But, although there are several reasons colonialism was a contributing cause to the lack of progress in a number of economies, this explanation is particularly inapplicable to India and China: India was colonized by the British, the least harmful colonial power (Bhalla 1997), and China was never colonized. There are other confounding details. Latin American economies were rich when they gained independence from their colonial powers in the early 19th century. Sub-Saharan Africa was considerably richer than Asia (excluding Japan) in 1960. Yet between 1980 and 2000, neither Latin America nor sub-Saharan Africa showed any gains in per capita income. In contrast, poor Asia, given up for lost, grew at one of the most rapid annual rates in world history—more than 4.5 percent per capita. And the West grew at an average annual rate of only 1 percent from 1820 to 1913, with the US annual growth rate at 1.7 percent per capita during the same years. This is one example of how varied the global experience of growth has been over the past several hundred years. And this varied experience has been accompanied by an equal variety of investigations and interpretations—in fact, the oldest type of investigation in economics is into the determinants of economic growth.

Some Explanations of Growth In the modern era, the earliest answer to the question of what determined growth was from the master himself, Adam Smith, in The Wealth of Nations, published in 1776. Nobel Laureate Sir Arthur Lewis laid down the basics in The Theory of Economic Growth, which remains as rich and relevant now as at its publication in 1955. In the 1960s, two notable scholars focused on related aspects of the age-old question: Alexander Gerschenkron (1962) voiced concern about the relative pace of growth in some European countries, and Gunnar Myrdal 1. See Bhalla (2007a) and Bhalla (forthcoming) for details.

12

DEVALUING TO PROSPERITY

(1968) offered a rather gloomy outlook on poverty (his book was subtitled An Inquiry into the Poverty of Nations). Robert Solow (1970), another Nobel laureate, and Trevor Swan (1956) provided the mathematical underpinnings of growth with a theoretical framework that still serves as the foundation for discussions of growth. The field has become crowded and the approaches more creative. John Hause (1971) asked, “If You’re So Smart, Why Aren’t You Rich?” Richard Easterlin (1981) wondered, “Why Isn’t the Whole World Developed?” William Easterly (2002) characterized the “quest for growth” as “elusive.” Elhanan Helpman (2005) described it as a “mystery,” and Benjamin Friedman (2005) considered it a moral question. I examined the issue through the prism of the Chinese and Indian middle classes (Bhalla 2007a).2 Such a historical review offers the helpful reminder that there may be nothing really new emerging about the oldest economic investigation. For example, the economics of institutions was fully anticipated by Lewis (1955)— The Theory of Economic Growth had a section entitled “Economic Institutions.” Another popular modern theory, on the importance of economic freedom, was articulated and emphasized by both Friedrich von Hayek (1944) and Milton Friedman (1962), and was given prominent mention by Lewis (1955). In addition to emphasizing the importance of institutions, Lewis (1955) outlined four principles of growth: factor accumulation, human capital, institutions, and policy. He also offered three proximate causes of growth: “effort to economize (efficiency); increase of knowledge and its application; and increasing the amount of capital or other resources per head” (Lewis 1955, 11). In short, growth is about increasing the application of factors, particularly human and physical factors, and obtaining an extra yield (productivity) from these factors, a concept today known as total factor productivity growth. Each generation of economists has built on this framework, but the foundation principles have remained the same.

Growth since 1950 The historical pattern of development can be summarized by looking at growth over time in two groups: developed economies (the so-called Western world, plus Japan, Eastern Europe, and the former Soviet Union) and the developing economies (the rest of the world). Data for these two groups as a whole and for selected economies in each group are reported here. In 1950, there were 51 developed economies,3 with a population of 0.83 billion or about one-third of the world’s population of 2.55 billion (table 2.1). Purchasing power parity (PPP) 1996 base current income in these economies 2. This study was titled Second Among Equals: The Middle Class Kingdoms of India and China. A revised version is in preparation (Bhalla forthcoming). 3. See appendix table A.1 for details on the country composition of the world and other details pertaining to countries, for example, number of small countries and oil-exporting countries.

DETERMINANTS OF ECONOMIC GROWTH 13

Table 2.1

Year

Global population and income, 1950–2011

Developed economies

Developing economies

Global total

Developed economies’ share of total (percent)

Population (billions) 1950

0.83

1.72

2.55

32.5

1960

0.95

2.10

3.05

31.1

1980

1.14

3.30

4.44

25.7

2011

1.31

5.60

6.91

19.0

Income (trillions) In current US dollars 1950

—

—

—

—

1960

1.0

0.3

1.3

76.9

1980

8.0

2.4

10.4

76.9

2011

41.7

22.8

64.5

64.7

4.6

1.7

6.3

73.0

1960

7.1

2.7

9.8

72.4

1980

15.9

7.4

23.4

67.9

2011

30.5

38.1

68.7

44.4

In constant PPP dollars 1950

Per capita income In current US dollars 1950

—

—

—

1960

1,972

283

498

1980

9,897

1,265

2,669

2011

39,010

9,496

10,283

In constant PPP dollars 1950

5,542

988

2,471

1960

7,474

1,286

3,213

1980

13,947

2,242

5,270

2011

23,282

6,804

9,942

PPP = purchasing power parity Notes: The sample comprises 204 economies. Current US dollar results are based on a reduced sample of 125 countries. Constant PPP dollars have a base year of 1996. Sources: Penn World Table 6.1 (Heston, Summers, and Aten 2002); Maddison (2001); IMF, World Economic Outlook database; and World Bank, World Development Indicators. Data for 2011 are estimates obtained from IMF (2011). See appendix A for details. Hereafter, the constructed dataset is referred to in text and tables as Bhalla (2007a) dataset extended to 2011.

14

DEVALUING TO PROSPERITY

totaled about $800 billion, or close to 73 percent of global income of $1.1 trillion. Sixty years later, in 2011, the developed world had only 19 percent of the global population and about 44 percent of global income. In 1950, an average person in a developing economy earned in constant PPP terms $988 a year, considerably lower than the $5,542 earned by the average person in a developed economy.4 Table 2.2 reports average real growth in income during 1951–2011. By 2011, there had been considerable improvement in both the absolute and relative income of the average developingeconomy inhabitant, whose per capita income had risen to $6,804, reflecting an average annual growth in income of 3.4 percent during this 71-year period. In the developed economies, per capita incomes increased to $23,282 a year, reflecting an average annual growth rate of 2.3 percent.

Popular Theories The remainder of this chapter examines some of the theories offered to explain this differential pattern of income growth in the developed and developing economies. To the extent possible, the theories are presented in historical order, the evolution of which can broadly be described as moving from an emphasis on inheritance (what societies are endowed with), to the importance of hard work and enterprise, and finally to a focus on policy-induced changes. The importance of particular growth determinants is quantified using an “impact coefficient.” Sometimes this coefficient is culled from the available literature on stylized facts or from conventional wisdom. When possible, it is calculated as an additional variable in the basic growth model pioneered by Robert Barro and Xavier Sala-i-Martín (1992). This model has two variables: The dependent variable is per capita income growth, and the independent variable is the per capita at the start of the time period. The most common form of this basic model uses a cross-country panel dataset with five-year time periods.5

Geography Is Not Destiny The despair over lack of growth in Africa in the postcolonial period (after about 1960) led to the hypothesis that “geography is destiny” (Landes 1998). This thesis contends that countries close to the equator are at a natural disadvantage compared with countries further away, which are destined to be richer because they have a more temperate climate and better soil, which leads to higher productivity and thereby to higher growth. The tropical climate is more favorable to disease and less conducive to work, and therefore slows the potential pace of economic development. In other words, these economies inherit a 4. All income figures are in real dollars, adjusted for inflation using a purchasing power parity (PPP) index with a 1996 base. 5. Unless otherwise specified, the basic model includes time and country dummies in addition to the log of initial per capita income.

DETERMINANTS OF ECONOMIC GROWTH 15

16 DEVALUING TO PROSPERITY

Table 2.2

Average (log) growth in per capita income by region, 1951–2011 (percent)

Country

1951–59

1960–72

1973–82

1983–95

1996–2007

2008–11

1960–79

1980–11

1951–2011

World

2.7

2.7

2.2

2.6

4.3

4.0

2.5

3.5

3.2

Developed economies

3.4

4.0

1.6

0.7

2.8

0.2

3.4

1.4

2.3

Germany

6.8

3.6

1.9

2.2

1.4

0.6

3.3

1.5

2.9

Japan

6.6

8.8

2.5

2.7

1.3

–1.1

6.7

1.7

4.0

United Kingdom

2.3

2.4

1.3

2.4

2.2

–1.2

2.3

1.6

1.9

United States

1.4

2.9

1.6

2.4

2.1

–0.7

2.9

1.6

2.0

Developing economies

2.4

2.2

2.4

3.2

4.7

4.9

2.2

4.0

3.4

Brazil

3.7

4.8

3.2

1.0

1.6

2.8

4.8

1.3

2.4

China

3.9

1.7

3.6

6.5

8.7

8.7

2.1

7.5

5.2

India

1.5

2.0

1.6

3.7

5.4

6.0

1.6

4.6

3.2

Korea

4.0

5.2

5.8

7.2

3.8

2.8

6.0

4.9

5.1

Notes: Based on data for 204 countries. See appendix A for details and appendix B for full data. Developed economies include members of the Organization for Economic Cooperation and Development (OECD) and countries in Eastern Europe and the former Soviet Union. Sources: Bhalla (2007a) dataset extended to 2011. See table 2.1 and appendix A for details.

deck that is stacked against them, which is an additional important reason for slow historical growth in sub-Saharan Africa. Variants on this theory use various proxies for geography, including latitude, number of days with tropical weather, number of days in frost, minimum temperature, minimum monthly rainfall, and maximum temperature. Such geographic variables tend to be the most “moody” empirically—meaning that they are sometimes significant but are not often meaningful. The most common variable is latitude, and there are enough examples of high-growth economies or faster-growing economies near the equator to throw the entire theory into question. Singapore, which is virtually on the equator, has had one of the world’s fastest growth rates. Kerala, the southernmost state in India (10q north latitude), is its most developed state and has social indicators (such as infant mortality) comparable to those of economies in the West. In Africa itself, Ghana and Uganda performed much better than Lesotho or Mali but are closer to the equator. The list goes on, and the results corroborate that latitude is not at all important in explaining growth differences, over the long term (1960–2011) or the shorter term (1980–2011). None of several other geographical variables—including temperature and rainfall—proved significant in explaining growth performance.

Culture and Religion A variant on theories that emphasize the role of institutions is the assertion that cultural or religious background can influence the pace of development. For example, this theory would posit that East Asian countries have been successful because they subscribe to Confucianism, or that some Western economies are rich because of the Protestant work ethic. In his detailed studies, Xavier Salai-Martín (1997a, 1997b) finds that Confucianism was consistently conducive to higher growth, second in importance to the openness of the economy. I document (Bhalla 1997) how confusing this hypothesis can be: Confucianism was also associated with authoritarianism and with economic freedom. In the context of this book, Confucianism also correlates with a deeply undervalued exchange rate. Therefore, while the cultural and religious theory may be appealing, it has rarely, if ever, been empirically strong: If the world is separated according to religion into four groups—Catholic, Protestant, Muslim, and others—both Catholic- and Muslim-dominated societies have a significantly lower economic growth rate (–0.9 percent below the world average). Protestantdominated societies do better (–0.5 percent below the gobal average), but the coefficient is not statistically significant. All the religion variables, however, become insignificant when more variables are introduced into the analysis.

Inequality Nobel Laureate Simon Kuznets (1955) repopularized the study of the effect of growth on economic equality. He asserted that with development (growth), inequality first worsened and then improved, an effect that became known as DETERMINANTS OF ECONOMIC GROWTH 17

the Kuznets Curve. Testing this theory became a major focus of growth economists, particularly during the 1970s and 1980s. The conventional conclusion is that inequality is empirically intractable but does have a dampening effect on growth. For 42 developing economies for which there are at least two years of inequality data since the mid-1970s, there is support for the conclusion that high initial inequality leads to lower subsequent growth.6 These preliminary results should be treated with caution, however, because inequality data are not available on a consistent basis for the same country, let alone on a consistent cross-country basis. Furthermore, for some countries there are only consumption inequality data, while for others there are only income distribution data, a difference that can and does bias the results.

Factor Accumulation: Capital, Labor, and Human Capital The simple idea behind the factor accumulation theory is that higher inputs can mean higher outputs. Capital is the oldest known determinant of economic growth: accumulate more capital and grow faster. But during the 1960s and 1970s, before the opening of the global economy, several developing economies, especially India and China, along with Russia and the Latin American countries, demonstrated that investment without openness, or investment without competition, would produce some immediate growth but would be disastrous for growth in the longer term. In the mid-1990s, the conventional wisdom was that East Asia’s success was mostly about factor accumulation (Young 1995, Krugman 1994). But recent research indicates that this conclusion was incorrect: There was more capital investment in these economies to be sure, but more important was the fact that their openness to trade and investment and their more-than-competitive exchange rates enabled higher productivity growth. Another factor of production (and Lewis’s second principle of growth) is human capital—knowledge or education. The theory behind the important role of education was not developed until the late 1950s; empirical estimates started trickling out in the mid-1960s. Today, it is a stylized fact and conventional wisdom that private returns to education are high, often very high. No one has yet argued that education is unnecessary for sustained economic growth, development, and catch-up. In the mid-1980s, there was increased attention to the indirect and increasing social returns to education, inspired in part by Paul Romer (1986). However, even though various production function models find that education generates positive returns on an economywide

6. Initial inequality is strongly significant and has an estimated magnitude of –0.15, that is, each 10 percentage point increase in the Gini inequality measure lowers subsequent growth by 1.5 percent a year. However, once the currency valuation measures are introduced into the analysis, the impact coefficient is reduced to less than half its original level (–0.07), although the significance is retained. This issue is examined in Bhalla (forthcoming).

18

DEVALUING TO PROSPERITY

basis, it has proven difficult to find a significant externality effect for education on a smaller scale.7 These caveats notwithstanding—there are times when too much capital may not yield extra growth and education may not have significant externalities—a key aspect of growth empirics remains to establish the determinants of human and capital factor accumulation. What causes investment to increase and countries to become richer as a result? I return to this later but first address an important and nearly forgotten determinant of initial step-jump growth: the reallocation of labor from traditional agriculture to modern industry.

Reallocation of Labor Agriculture is always the starting point for economic growth, whether for Western economies on the eve of the Industrial Revolution or for developing economies at their independence in the mid-20th century. Lewis (1955) famously and accurately described the transition of an economy from agriculture to nonagriculture as the sine qua non of economic transformation and economic growth. Cheap and unlimited supplies of labor are available in most economies before the onset of rapid economic growth. At this point, the ecosystem is in equilibrium; there is little productivity growth in agriculture; and most people are, and remain, on the farm. Technological growth, either within or outside agriculture, releases labor. If this growth is outside the economy, then outward migration causes the supply of domestic agricultural labor to decline. Alternatively, technological growth in the domestic industry induces labor to leave the farm. In both cases, growth is accompanied by a decline in the share of agriculture in the economy’s GDP. Some of the development literature presumes an unlimited supply of labor à la Lewis, and some follow the structural change school of Hollis Chenery and Moises Syrquin (1975). But all the literature recognizes that, in the early stages of development, any extra growth in an economy is due to the reallocation of labor from the low-productivity agricultural sector to the higher-productivity nonagricultural (industrial or services) sectors. Only in later stages of development do factor accumulation and technological change begin to contribute to higher growth. This factor reallocation has been estimated by Sherman Robinson (1976) to average about 16 to 18 percent during the early growth stages in developing economies. The reallocation theory has important implications for the evolution of economic growth and therefore standards of living. This reallocation of labor generates an income growth stream (and income stream) that is S-shaped. At first, there is a gradual release of labor from a low-productivity enterprise

7. Some analysis rejects an externality effect of education—a prime ingredient in endogenous growth models—including Bhalla (1994) and Pritchett (2001). For a contrasting view, see Vandenbussche, Aghion, and Meghir (2004).

DETERMINANTS OF ECONOMIC GROWTH 19

Figure 2.1

Growth in India according to the labor reallocation theory, 1960–2010

share of agriculture in GDP

average 5-year growth (percent)

0.50

10

0.45

9

Share of agriculture in GDP

8

0.40 7 0.35 0.30

6 5 percent GDP growth

5 4

0.25

3 0.20

Average 5-year growth (percent)

2

0.15

1

0.10 1960

0 1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

Notes: The share of agriculture in GDP is simulated by the reallocation growth model; see text. It is very close to the level actually observed. Source: World Bank, World Development Indicators, 2011.

(agriculture, which has a typical annual growth rate of 3 percent) to a highproductivity enterprise (industry and services, which have a typical annual growth rate of 6 percent). Agriculture initially accounts for a large share (60 percent) of output. Over time, higher-productivity growth and the movement of factors means that the share of agriculture eventually declines to 20 percent and below. At these levels, the reallocation of labor offers little extra growth, and while the descent to this trough is steep, at the base, there is a flattening out. This is the S-shaped pattern of the evolution of income levels, a property that is used to estimate equilibrium real exchange rates in chapter 4. Under the reallocation theory, growth can be modeled as the weighted average of the growth rates in agriculture and industry, with the relative weights being the shares of each sector in the economy.8 Figure 2.1 presents some simulated data for India. In the first year, 1950, it is assumed that 55 percent of real

8. “Industry” is used to refer to both industry and services, given that the two sectors move broadly in tandem and have similar labor productivities due to the operation of wage arbitrage.

20

DEVALUING TO PROSPERITY

output originates in agriculture, and the annual growth rates of agriculture and nonagriculture are assumed to be constant at 2.75 and 6 percent, respectively. (All these assumptions are very close to what has actually been observed in the Indian economy during this period.) These differential growth rates predict a path for agriculture and for overall GDP growth. By 2011, this path brings the share of agriculture down to 17 percent and GDP growth to 5.5 percent. The actual percentages are 16 percent and over 8 percent, respectively. This pure application of the reallocation model shows that India would reach a GDP growth rate of 5 percent, sometimes considered a threshold to more accelerated growth, in 1983, independent of the presence or absence of economic reforms.9

Trade Policy As noted, trade has long been considered an important, if not the most important, determinant of why some countries are rich and others poor: Comparative advantage rules, and maximization of incomes is best achieved through the enhancement of trade. Many developing economies adopted an autarkic, closed, Soviet-style model after independence, in an effort to develop fast. This most often failed, and these economies opened up to external trade as a means of recovering from their self-induced ills. Because many of these economies then grew faster than ever before, “trade causes growth” became a stylized fact. A contrary line of reasoning, embodied in such works as Rigobon and Rodrik (2004), asserts that, while openness could facilitate trade and trade could facilitate growth, the causality could easily be the reverse. As economies grow, demand increases for a variety of products, leading to expanded trade (more imports and, in turn, more exports to finance the imports). Thus, econometric models that purport to show an acceleration of growth caused by increased trade could actually be showing the reverse. This problem can be addressed econometrically using techniques to identify the direction of causation, specifically through “instrument” variables— that is, variables that are correlated to one of the independent variables (trade) but not to the other (growth). The operative word is “econometric,” which means an estimate of reality. Estimates are subject to error, and the mere presence of a potential error, regardless of its size, allows both sides to claim victory. The protagonists say they have identified the problem away; the opponents say the instruments are weak. The debate goes on. Trade policy can be assessed either in terms of outcomes—that is, the share of trade in GDP and changes therein—or in terms of the instruments that affect trade. The latter are captured by indicators representing tariff policy, and there is a lack of strong statistical support for such indicators. In many 9. See Bhalla (2010) for a detailed discussion of the role of policy in India’s development from 1950 to 2010.

DETERMINANTS OF ECONOMIC GROWTH 21

instances, tariffs have been reduced and import protection has declined, yet growth has failed to accelerate. In others, growth has persisted despite high tariffs. However, no study has actually found that high tariffs have led to faster growth during the postwar era. There has been more success in measuring trade policy using the share of trade in GDP. If initial trade shares (in 1980) are introduced into the basic growth model, the impact coefficient of trade shares is significant at the 1 percent level of confidence, and this significance remains even when other related determinants of growth (e.g., currency valuation) are introduced into the analysis. Each 10 percentage points of extra trade share adds 0.4 percentage points to subsequent per capita growth.

Economic Openness The more that economies look outward, the richer they become—a result that holds no matter how far back we go in history. However, openness as an empirical concept was not formally and econometrically investigated until the World Bank’s 1991 World Development Report. But there has been a constant outpouring of articles and indices since (see Harrison 1996). Openness had its fair share of critics. A number of scholars and politicians feared that decreasing import tariffs would allow foreign firms, with modern methods and lower costs, to swamp all competition and prevent the development of domestic industry and expertise. As evidence of the potential costs, they cited the experience of developing economies during the colonial era. At that time, free trade was prevalent, and yet the developing economies fell far below the income levels of their colonial masters. Further, the argument went, there was evidence that high tariffs actually helped developed economies grow faster.10 Jeffrey Sachs and Andrew Warner (1995) developed a popular measure of openness, which was extended by Romain Wacziarg and Karen Horn Welch (2003). Others have offered different measures, including Robert Hall and Charles Jones (1999) and, more recently, Menzie Chinn and Hiro Ito (2008). Somewhat surprisingly, none of these measures of openness are statistically significant in most growth models. The significance of the openness variable improves only after currency valuation measures are introduced. This is consistent with theory and expectations: Openness does not help much when the exchange rate is overvalued—just consider Japan. But if the economy is open and the country has a competitive currency, then growth is likely on the way.

10. See O’Rourke (2000). See also chapter 9, where the evidence presented suggests that the argument that tariffs helped growth in the 19th century is fragile and very weak.

22

DEVALUING TO PROSPERITY

Currency Undervaluation One of the earliest and most ardent advocates of currency undervaluation11 as a policy determinant of growth was Béla Balassa (1964). Both as an advisor to the World Bank and as an academic, Balassa emphasized the advantages of export-led growth, which came about via competitive exchange rates, a politically correct term for an undervalued exchange rate. There are large differences between an export-led growth strategy and a currency undervaluation strategy. The former is a classic industrial policy, under which certain firms or sectors are chosen by the government to become highly competitive in the international marketplace (that is, to be successful exporters). The effects of a currency undervaluation strategy are not confined to particular firms or to the export sector. It deliberately has economywide positive ramifications. Market discipline is tapped through trade and the exposure of domestic firms to international markets, thus nullifying the potential political and economic distortions that could result from state favoritism and state intervention. Theoretically, it is not easy to follow a policy of deliberate currency devaluation, because it involves changing the terms of trade, and not just for the short term. It is virtually impossible for developed economies to follow such a policy, given the large free market in exchange rates, but a number of developing economies may be able to influence the exchange rate by being active players in the currency market. They can use “intervention”—the government either buys or sells dollars depending on whether it wants its currency to go down (depreciate) or go up (appreciate). Using dollars to buy its own currency signals that the authorities believe that the exchange rate is too low (e.g., a rate of 50 rupees per dollar), and selling its own currency to buy dollars signals that they believe the nominal exchange rate is too high (e.g., a rate of 40 rupees per dollar). There are several problems with the concept of domestic authorities influencing a trillion-dollar daily exchange market by buying and selling a small sum of dollars. Even if they are able to influence the nominal exchange rate, they are unlikely to influence the exchange rate that really matters—the real exchange rate (RER),which can move in counter directions through exchangerate-induced inflation or deflation. For example, the government injects money into the economy when it buys dollars and sells the domestic currency. This extra money can generate inflation, which can negate any competitive advantage the economy would gain from devaluing its currency. Whether this actually happens and, if so, to what degree is explored in subsequent chapters. There are other problems with the theory that a competitive exchange rate

11. Currency undervaluation involves greater emphasis on negative values of currency valuation, that is, when a currency is cheap (positive values are when a currency is expensive). A −10 percent valuation means that a currency is undervalued or cheaper by 10 percent; a +10 percent valuation means that a currency is overvalued or more expensive by 10 percent.

DETERMINANTS OF ECONOMIC GROWTH 23

equals higher economic growth. There is considerable debate about the methodology to use for measuring the equilibrium value of the exchange rate, a first step in evaluating whether a given exchange rate is competitive. Furthermore, the equilibrium exchange rate is a moving target—it is not constant over time or during different stages of development. All these issues are examined in the rest of this book.

Government Intervention: Fiscal Policy It is important to analyze the contribution to growth of policy changes. There is general agreement that bad policy outcomes, such as high inflation, are a major handicap for higher growth. Another favorite policy recommendation for both developed and developing economies—one that is almost synonymous with the Washington Consensus—is to reduce the fiscal deficit. The promised benefits are manifold, including greater efficiency in production, fewer losses in government undertakings, and less crowding out of private investment. Government deficits matter, and their reduction is necessary for macroeconomic stability and sustained growth. Indeed, the European Union was founded in part on the notion that government deficits matter a lot. A related notion is that interest rates matter. High fiscal deficits, financed by higher government borrowing, translate into higher real interest rates, which may crowd out private investors.12 Fiscal policy is an important explanatory variable in growth models that assess short time periods, such as the five-year periods used extensively in this book. The coefficient for fiscal policy is almost always significant and robustly so, and it has the correct sign. However, this variable is not significant for models that cover longer periods, say, 20 years or more. A simple explanation for the variable’s significance in short-term models is that fiscal deficits increase by definition when growth falls and hence there is a robust negative association between growth and fiscal deficits. William Easterly (2005) finds that only outliers cause the fiscal deficit coefficient to be significant. However, no matter what the empirical specification, or which country is under study, the best empirical effect of fiscal deficits in the literature is about 0.1—that is, for each 1 percentage point reduction in the fiscal deficit, growth is 0.1 percentage point higher. That’s it. As a country moves from a fiscal deficit of 4 percent of GDP to a fiscal deficit of zero, it adds only about 0.4 percent to annual GDP growth. This “extra growth” may seem reasonable for developed economies whose potential GDP growth is about 2 to 3 percent, but the total effect is insignificantly small for developing economies, whose average annual GDP growth in the last several years has been above 5 percent. There is, however, a measurement problem associated with fiscal deficits as traditionally evaluated. The most popular source for the fiscal deficit vari12. For India, the causation was likely the other way around, with high managed real interest rates leading to larger interest payments and larger fiscal deficits (Bhalla 2000).

24

DEVALUING TO PROSPERITY

able has been the World Bank’s World Development Indicators, but this source reports the data for deficits at only the central level, ignoring deficits at the state and local levels. The International Monetary Fund (IMF) now publishes fiscal deficit data on an aggregate basis, including central, state, and local levels. Tests for nearly 100 countries fail to reveal any effect of an initial fiscal deficit (in the initial year of a five-year period) on subsequent growth. This result is obtained with both the central and aggregate deficits. However, if the fiscal deficit variable is entered as an average for each five-year period, then the coefficient is robustly significant, with a magnitude of about 0.16. The reason behind this transformation in significance and magnitude is most likely that high growth is associated with better fiscal outcomes. The most obvious conclusion is that there is very little historical evidence to suggest that fiscal deficits matter for economic growth.

Education It is almost a tautology that education brings about higher incomes for individuals and societies. Education helps make investment more productive and leads to higher growth. Micro data across the world yield very robust results for the positive effects of education: Each extra year of education adds about 12 percent to a person’s lifetime income, although in macro data (five-year periods) there is no significant coefficient for education. Over the long term, 1980–2011 or 1960–2011, there is a robustly significant effect of education: Each extra year of mean education in the labor force adds about 0.4 to 0.6 percentage points to annual GDP growth.13 This effect is magnified if the country is open, or more accurately, has an undervalued exchange rate, because this allows for importation of technology that workers with a higher level of education can use to improve their productivity and that allows the economy to catch-up with the frontier.14 This was a lesson learned by India and China between 1950 and 1980, as well as by several other developing economies (and the former Soviet Union and its empire in Eastern Europe).

Catch-Up Some additional phenomena are often considered as factors of growth, some of which are rephrasing or extensions of Lewis’s (1955) four proximate causes. One, in particular, deserves mention. In a pioneering contribution, Barro and Sala-i-Martín (1992) introduced the now ubiquitous term “catch-up.” Catch up means the ability of poorer economies to grow faster than wealthier economies because of two factors. First, income growth for poorer economies can be 13. See Barro and Lee (2010) for details on construction of a variable representing the mean years of educational attainment for a large sample of countries; data were extended until 2011 by trend extrapolation. 14. The interaction between education and openness was first explored in World Bank (1991).

DETERMINANTS OF ECONOMIC GROWTH 25

enhanced by simple technology transfer rather than the research and production of new technology, which is an inherently slow process. Second, poorer economies have wages that are lower in international productivity-adjusted terms. The basic growth model is defined with the catch-up term on the righthand side; the Barro–Sala-i-Martín framework allows us to put a value on this term. Poor economies can grow at about 0.5 to 1.5 percent per annum faster than rich economies, other things equal. This is an extremely robust result. In other words, in seeking to identify patterns of growth, it is important to recognize that poor economies have a natural advantage: They grow faster.

Institutions The theory that institutions play an important role in growth rests on two arguments. First, economic freedom (property rights) reduces uncertainty and enhances entrepreneurship, among other things, and this leads to greater efficiency and higher growth. Second, political freedom (political liberties and democracy) allows for more sensible decisions because of greater checks and balances. However, the evidence for the second argument is decidedly mixed, indeed the opposite is often argued, that authoritarianism helps growth, as shown by the strong growth example of East Asia. On the other hand, for every East Asian dictator whose economy produced high growth, there are 10 African and Latin American dictators whose did not. The importance of institutions in the determination of economic growth is explored in chapter 10.

Demography Demographics is a recent addition to the determinants literature. Population growth is a drag on development, but growth in the worker population (an increased supply of workers) is not. The argument is straightforward. In a period of high population growth, the future supply of workers increases sharply. In the transition period, when birth rates decline, the dependency burden on the existing workforce decreases (from an elevated level due to the high fertility earlier). The decline in births results in a higher proportion of the female population being employed, higher saving and investment, and therefore faster growth. There are two key assumptions underpinning this thesis. First and foremost, the pace of job growth increases to accommodate the expanded number of workers. This may not always be the case. Second, the labor force participation rate (LFPR) of women can differ sharply across countries. In India, the urban LFPR of women is as low as 25 percent, although it is sharply above the 15 percent level prevailing just a decade ago. An increase in the LFPR is likely to be the biggest positive shock for most developing economies—subject, of course, to pursuit of macroeconomic policies conducive to economic growth and employment. To summarize, the demographic dividend is a prolonged period of higher saving and growth that springs from the extra availability 26

DEVALUING TO PROSPERITY

of workers and a decrease in the dependency ratio, coupled with the life-cycle hypothesis of consumption. Empirical tests confirm the importance of the demographic dividend. Demographic effects are captured by the initial worker-to-dependent ratio (the ratio of the population ages 15–64 in the total population in 1980). This variable is significant and has an elasticity of 0.19 when it is added to the basic growth model—that is, each 1 percentage point increase in the ratio leads to an increase in the per capita annual growth rate of 0.19 percent.

The Middle Class The importance of a vibrant middle class was discussed by Aristotle and later by John Stuart Mills, Thomas Malthus, Karl Marx, and Barrington Moore (although the latter two had a somewhat different definition of the middle class than the first three).15 Briefly, the middle class can be expected to positively affect growth because of its commitment to economic reforms and to a level playing field. The middle class holds these purely out of self-interest: The surest way for it to benefit is if merit is rewarded, and the sine qua non of the middle-class mentality is dedication to education and hard work. Who constitutes the middle class? Following Bhalla (2002a), the line that defines the middle class, much like the poverty line, is absolute and is the same for all peoples of the world. It is based on the population-weighted average of the highest poverty lines or the poverty lines in the wealthy economies of the West (and Japan). In these economies, by definition, the poverty line divides the poor and the nonpoor. Likewise, the lines that define the middle class are those that define the nonpoor on one end and the rich on the other, with the rich defined as those with an income level 10 times higher than the middleclass line (i.e., the beginning of the middle class). In 1996, based on PPP prices, the per capita daily poverty line for the developed world was $8.19 (the US poverty line was somewhat higher at $10.80). In 2011 prices, with the US GDP deflator having increased by 36.6 percent since 1996, the middle-class line for the developed world is $11.20 per capita per day; for a family of four, this is $16,350 per year. The rich are those with incomes starting at 10 times the starting level of the middle class, or $163,500 for a family of four. This means that the middle class in the United States in 2011 is those with incomes between $16,350 and $163,500.16 Does the initial size of the middle class matter in affecting growth? Yes. Even after controlling for currency valuations, the percent of the population

15. I examine the similarities and differences and the theoretical bases for their theories in Bhalla (2007a). 16. Given this definition of the middle class, and given quintile estimates of the distribution of income, I outline a method in Bhalla (2002a) for estimating the percentage of the population below any given income level, such as a dollar a day poverty line and middle-class line.

DETERMINANTS OF ECONOMIC GROWTH 27

Table 2.3

Running out of good luck? Growth persistence, 1960–2011 Persistence coefficient

R2

Correlation coefficient

1960s

0.09

0.13

0.23

1970s

0.35***

0.14

0.37

1980s

0.46***

0.23

0.47

1990s

0.34***

0.15

0.38

2000s

0.29

0.12

0.31

1970s versus 1950s

–0.27**

0.11

–0.10

1980s versus 1960s

0.29**

0.12

0.34

1990s versus 1970s

0.36***

0.14

0.37

0.45***

0.23

0.48

Period Versus previous decade

Versus two decades earlier

Versus two-plus decades earlier 1980–2011 versus 1960–80

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. The model of persistence is . . . as follows: y t = α + β1*y0 + β1* y t – 1, where y is the growth of per capita income in period t and t – 1, and y0 is the log of initial per capita income. All countries are in the sample included in the persistence model, except those with populations of less than 1 million, oil-exporting economies, and countries of Eastern Europe and the former Soviet Union. Source: Bhalla (2007a) dataset extended to 2011.

that is middle class in each country in 1980 has an important statistical effect: Each 10 percentage point increase in the size of the middle class in 1980 results in 0.3 percent additional annual growth between 1980 and 2011. Five-year panel data reveal almost an identical effect.

Good Fortune In an oft-cited paper, “Good Policy or Good Luck?” William Easterly, Michael Kremer, Lant Pritchett, and Lawrence Summers (1993) offer a novel hypothesis for differences in levels of development and growth rates. It is all luck. They hold out little hope for countries to control their economic destinies. On the basis of a wide-ranging cross-country analysis, the authors rather convincingly argue: Growth rates are highly unstable over time, while country characteristics are highly persistent. The correlation across decades of countries’ growth rates of income per capita is around 0.1 to 0.3, while most country characteristics display cross-decade correlations of 0.6 to 0.9. Correlations of growth across periods as long as two decades—period lengths comparable to those used in the cross-section empirical literature—are similarly low. (Easterly et al. 1993, 460)

Table 2.3 retests the “good luck” hypothesis. Contrary to Easterly et al., persistence is remarkably stable across decades, with the impact coefficient 28

DEVALUING TO PROSPERITY

averaging about one-third—each 1 percent growth in the previous time period persists (translates) into a 0.3 percent growth during subsequent decades.17 There is an even larger persistence if the comparison is made between 1980– 2011 and 1960–80; the impact coefficient is 0.45. Luck matters, but persistence matters much more!

Growth Policies: The Washington Consensus Returning to where this chapter started, even if the causes of growth are identified, there remain important prior questions. For example, what causes savings and investment to increase? What differentiates a successful catch-up developing economy from an unsuccessful one? Since the end of World War II, there has been a new “exogenous” factor in the growth process—government policy. This is distinct from the longestablished role of government as a regulator of economic activity—even Adam Smith emphasized the importance of government (as a nonmarket participant) in setting and policing the rules. For some 150 years, regulation was the extent of government involvement, until John Maynard Keynes set off the intellectual transformation of the government’s role, from passive rule setter to active player. This radical development in economic thought, no surprise, remains controversial even today—perhaps even more so in the wake of the Great Recession, when the old rules of the market seem to have broken down. Keynesian theory as actually practiced during the past 60 years has produced a wide variety of policies, from the good, to the bad, to the downright ugly. The most popular and comprehensive catalog of good policies is John Williamson’s (1990) Washington Consensus, so called because Washington is home to the two central Keynesian Bretton Woods institutions: the International Monetary Fund and the World Bank (formally, the International Bank for Reconstruction and Development). And informed advice for policy interventions to enhance economic growth has come from economists working with those institutions. Of late, the Washington Consensus has become a term of disdain, because the many recent policy crashes and economic crises have left both economists and the general public without much confidence in the ability of government to act positively through policy. The five years leading up to the crisis in 2008— ironically, the period of the fastest growth in the developing world—were also the time when the consensus among academics diverged significantly from that among practitioners, who implemented policy, especially exchange rate policy, differently than it was prescribed by the experts.18 Since 2008, policy

17. This is for a sample of 97 developed and developing economies; excluded are those with less than 1 million population in 2008, oil-exporting economies, and countries in Eastern Europe and the former Soviet Union. The difference in the results with the Easterly et al. analysis may result from the use of different sets of Penn data (1996 base rather than a 1985 base) and the use of data for 20 additional years (1990–2010). 18. The 2008 crisis and beyond has made amost everyone look foolish, from those who favor a

DETERMINANTS OF ECONOMIC GROWTH 29

Table 2.4

Policies of the Washington Consensus Old

Washington Consensus 1. Fiscal discipline

New “Augmented” Washington Consensus 11. Corporate governance

2. Reorientation of public expenditures

12. Anticorruption

3. Tax reform

13. Flexible labor markets

4. Financial liberalization

14. World Trade Organization agreements

5. Unified and competitive exchange rates

15. Financial codes and standards

6. Trade liberalization

16. Prudent capital-account opening

7. Openness to direct foreign investment

17. Nonintermediate exchange rate regimes

8. Privatization

18. Independent central banks/inflation targeting

9. Deregulation

19. Social safety nets

10. Secure property rights

20. Targeted poverty reduction

Sources: Williamson (1990) for the old Washington Consensus; Rodrik (2006) for the “augmented” Washington Consensus.

has made a comeback of sorts, at least temporarily, as countries resorted to policy and government bailouts to correct market failures of all sorts. Over the longer term, however, the majority conclusion—and one that has significant support among academics—is that policy does not matter. However, a slight change in phrasing would lead the proposition to have near-universal support among economists: “Policy matters only when all the appropriate conditions are in place, and when the government is well intentioned.” Table 2.4 outlines the broad mean of policies available. The left column lists the policies in the original Washington Consensus; the right column contains an enhanced list compiled by Dani Rodrik (2006). There is an argument to be made that the 10 policies in the original Washington Consensus are really only subcomponents of two main policies (openness and fiscal discipline) and one institutional arrangement (secure property rights). In the left column of the table, rows 1 to 3 relate directly or indirectly to fiscal policy: Tax reform enhances tax revenue, and reorientation of public expenditures also helps fiscal discipline. Rows 4 to 7 pertain to openness, including rules for foreign direct investment or competitive exchange rates. Rounding out the Washington Consensus are policies that a decade after their identification were identified as measures of “institutions”: Privatization, deregulation, and secure property rights (rows 8 to 10) all help growth via their effects on incentives to the private sector—that is, these are the institutions needed to enhance growth. The Washington Consensus list is not in priority order, but it is inter-

“markets only” approach to those who favor active institutions, and including some who correctly forecast the crash.

30

DEVALUING TO PROSPERITY

esting that “secure property rights” is last, given the fact that for the advocates of the institutional approach, property rights come first, almost to the exclusion of everything else. With measures covering fiscal matters and institutions, and a total of 20 policies including Rodrik’s (2006) augmentation, there is no shortage of policy instruments available for promoting economic growth. The track record of countries that have heeded this policy advice, however, is inconsistent. Often, especially in Latin America, the policies seem to have not worked. This raises the all-important question: How robust is the empirical basis of the advocacy? Stated differently, if the Washington Consensus policies are good and desirable, then why are there mixed results? The rest of the book is concerned with answering these important questions.

DETERMINANTS OF ECONOMIC GROWTH 31

3 Currency Valuation, Savings, and the Current Account

I’m not young enough to know everything. —J. M. Barrie, The Admirable Crichton, Act I (1903) The thesis that growth can be explained in terms of real currency valuations is controversial, contentious, and, some may contend, convoluted. The purpose of this book therefore is to explore why the conclusion that overvaluation hurts growth is well accepted, but the corollary, that undervaluation helps growth, is not. What we know is that, broadly, growth is a function of investment, and investment is helped by savings. A growing young population helps (demographic dividend), as do education and openness. And because of catch-up, poorer economies can, and do, grow at a faster rate than richer economies. These are well-established stylized facts. Now add currency valuation to the mix.1 It can be argued that currency undervaluation makes an economy more competitive by reducing costs and increasing profitability and that this increases investment, which increases growth. This channel is clear-cut and theoretically sound but remains to be tested empirically. Interestingly, most of the literature tests not the direct relationship between currency valuation and investment, but rather the indirect (reduced form) relationship between currency valuation and economic growth. This book does both, testing the indirect relations for the recent and distant past and also developing a model to test the direct relationship between currency valuation and investment. There are five variables of interest: growth, currency valuation, investment, savings, and current account balance. Most of the book examines the relationship between the first two variables, chapter 5 examines investment, and this

1. Currency valuation is a neutral term; undervaluation is currency valuation with a negative sign. A positive relationship between currency undervaluation and savings means that, as a currency becomes more undervalued, the saving rate increases.

33

chapter examines the role of the final two variables, savings and the related topic of current account balance (which is also covered in chapter 7). This chapter discusses the controversies and stylized assumptions surrounding the effect of currency misalignments, deliberate or otherwise, on savings and the current account balance. There is broad acceptance that currency valuation affects the demand for imports and exports. There is the expectation that the cheaper a currency, the better the current account balance. The next steps are murkier. Assume for a moment that (measurement difficulties notwithstanding) we could actually know when a currency was undervalued by a significant amount. Would that country have a surplus on the current account? Given that the current account balance is the difference between savings and investment, does it follow that such currency undervaluation would positively affect savings?

Currency Valuation and Savings Some countries have persistently large current account surpluses. One explanation is that these countries have a cheap exchange rate. Another is that these countries have a higher proclivity to save. There are a number of reasons for a higher saving rate, including demographics (a higher dependency ratio) and a desire to insure against future shocks by accumulating foreign reserves. In any event, higher savings cause a current account surplus when savings are unmatched by domestic investment. The relationship between savings and currency valuation is not straightforward. John Williamson (2003, 5) sees the saving rate to be positively affected by currency overvaluation: “Savings increase as the current account deficit increases because of a less competitive exchange rate.” There is an opposite view. Michael Dooley, David Folkerts-Landau, and Peter Garber (2005) conclude that export-led growth requires a depreciated exchange rate, which leads to an improvement in the current account balance, which means higher savings. In short, undervaluation causes higher savings. There is a third view, which is developed in the next chapter. It is similar to that of Dooley, Folkerts-Landau, and Garber (2005), except there is no intermediation through the current account balance. Instead, a depreciated real exchange rate makes investment more profitable, which causes investment to increase, which causes higher growth, which leads to higher savings.2 This suggests that both investment and saving rates are affected by currency valuation, with the effect on savings possibly greater. The investment model offered in chapter 4 suggests that an undervalued exchange rate, and/or movement toward less overvaluation, leads to more investment, part of which may be financed by foreign capital inflows. The increase in investment leads to higher growth. This growth is first viewed as transitory and a large fraction of the 2. Bhalla (2007a) and Levy-Yeyati and Sturzenegger (2007). See Montiel and Servén (2008) for a useful survey of the issues.

34

DEVALUING TO PROSPERITY

“extra” income is therefore saved.3 As the process continues, this growth leads to a higher income level and higher saving and investment rates. Finally, there is the view that a currency becomes undervalued because of a need for higher savings. The East Asian financial crisis underscored the potential need for economies to insure themselves against external shocks by accumulating foreign reserves, which can be accomplished by intervening in the foreign exchange market and preventing the exchange rate from appreciating. This demand for self-insurance causes savings to increase, the external balance to improve, the real exchange rate to depreciate, and domestic demand (consumption) to be lower than it would otherwise be. In a simple form, this is Ben Bernanke’s (2005) savings glut thesis. This explanation fits Chinese and East Asian high savings and weak currencies. It also fits the explanation that these countries had higher investment demand, and therefore higher savings, both brought about by a cheaper exchange rate. Formal tests of the relationship between currency valuation and savings (and investment) are reported in table 3.1. All models have (log) initial per capita income, mean level of currency valuation, and time dummies in the regression. The dependent variable is the share of savings and investment in GDP. The results for both ordinary least squares (OLS) and fixed effects models, are reported. The latter specification has country-specific effects (country dummies) in addition to the above variables. This fixed effects model is preferred; the OLS results are reported for comparative purposes. Given the diversity of possible outcomes, it is encouraging to note that currency valuation—as measured in Bhalla (2007a) and described in detail in the next few chapters—has a very significant effect on savings and investment: Each 10 percent initial level of currency undervaluation adds about 0.4 percentage points to the saving rate. The second row in the table reports on a test of the savings glut hypothesis. A primary explanation for excess savings in developing economies after the Asian crisis of 1997–98 was their desire to build foreign reserves in the event of another crisis. The raw data do indicate a large jump in savings; for non-oil-exporting developing countries, the saving rate increased from 15.8 percent before 1995 to 18.1 percent after 1995. One test of the savings glut hypothesis is to use a dummy variable in a savings model. This dummy variable is zero for the pre-1995 period and 1 for 1995–2011. The coefficient is expected to be positive and significant. The dummy variable for years after 1994 is statistically insignificant and negative—that is, savings were lower after the Asian crisis, other things equal. There is little empirical evidence to support the hypothesis that countries are saving more because of their desire to selfinsure against crises.

3. Both the permanent-income and life-cycle hypotheses of savings contend that the saving rate is independent of the level of income. While that is true for developed economies, it is likely that the saving rate increases with the level of income for developing economies. This was documented in Bhalla (1980).

CURRENCY VALUATION, SAVINGS, AND THE CURRENT ACCOUNT 35

36 DEVALUING TO PROSPERITY

Table 3.1

Effect of currency valuation on economic variables, 1950–2011 OLS model

Dependent variable

Currency valuation

Savings as percent of GDP

–0.008

Investment as percent of GDP

–0.031***

Fixed effects R2 0.29

Currency valuation

Adjusted R2 0.65

979

–1.47

0.65

979

0.49

993

0.89

0.49

993

–0.038*** –0.038***

0.16

Time dummy, 1995

–0.051*** –0.051***

Number of observations

OLS = ordinary least squares Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Data are compiled in five-year intervals with the last five-year observation comprising years 2005 to 2011; see appendix A for details. All models include initial log per capita income and time dummies. The fixed effects model includes country dummies and reports the coefficients for the time dummy = 1995 for the savings and investment models. Source: Author’s calculations.

Current Account Balance and Growth Table 3.2 presents data on the three variables of interest—current account, savings, and currency valuation—for six countries: Germany, Japan, and the United States among the developed economies; and China, India, and Korea among the developing economies. Only India and the United States have a persistent current account deficit; the rest are surplus economies, despite the fact that this group includes both poor and rich economies and those with both undervalued and overvalued currencies. The pattern for these selected economies is reflective of the world at large and should prevent any hasty conclusions about the relationship between surpluses, levels of income, growth rates, and currency valuations. For example, both Germany and Japan have been thought to have undervalued exchange rates because they consistently have current account surpluses. However, in the case of Japan, the current account balance (CAB) has increased with an appreciation in the exchange rate. Between 1970 and 1985, considered the peak in Japanese trade surpluses, the CAB averaged 1 percent of GDP. Since 1985, the CAB has averaged 2.9 percent. Almost the same result holds for Germany—a CAB of 0.6 percent during 1970–85 and a CAB of 2.6 percent during 1986– 2011. This pattern is consistent, and not only for a few outlier years. China has simultaneously had high savings and a current account surplus since at least the 1970s. In the mid-1980s, China’s saving rate was high in the 30 percent range, and by the early 1990s it exceeded 40 percent of GDP. China’s currency at that time was overvalued by about 30 percent.4 By the late 1990s China’s currency was undervalued by 20 percent. However, the saving rate did not budge from the level of a decade earlier, about 40 percent. The current account surplus also stayed close to the level prevailing in the late 1980s, about 2 percent. After 2000 the undervaluation of the renminbi began to really deepen. In 2006, China’s currency valuation reached its low of −45 percent. (This means that the equilibrium level of the renminbi was close to half the prevailing exchange rate of 8 yuan to the dollar.) In 2006 and 2007, China’s current account balance averaged 9.5 percent of GDP. It is not likely a coincidence that China’s saving rate should rise above 50 percent precisely when its currency nosedived in real terms and its surplus widened to more than 10 percent of GDP. The per capita growth rate also accelerated above 11 percent just before the crisis in 2008, and reserves (made possible through active intervention in the currency markets that prevented the renminbi from appreciating) rose above $1.8 trillion from below $150 million in 2000. In 2011, the comparable indicators were: currency valuation, −44 percent; saving rate, 54 percent; current account surplus, above 5 percent; foreign exchange reserves, comfortably above $3 trillion; and per capita growth, close to 9 percent. The important observation is the absence of a direct relationship between CAB and currency valuation. Germany consistently had a current account 4. See Bhalla (2007a and forthcoming) for an examination of saving behavior in China and India.

CURRENCY VALUATION, SAVINGS, AND THE CURRENT ACCOUNT 37

38 DEVALUING TO PROSPERITY

Table 3.2

Current account balance, savings, and currency valuation for selected countries, 1970–2011

Period

Current account (percent of GDP)

Savings (percent of GDP)

Currency valuation (percent)

Current account (percent of GDP)

Germany

Savings (percent of GDP)

Currency valuation (percent)

Current account (percent of GDP)

Japan

Savings (percent of GDP)

Currency valuation (percent)

United States

1970–85

0.6

22.7

15.8

1.0

33.6

0

–0.4

19.0

–9.6

1986–95

1.4

23.3

34.8

2.7

32.8

55.2

–1.8

16.0

–12.6

1996–04

0.4

20.2

21.8

2.7

27.9

41.4

–3.5

16.7

0.3

2005–11

5.9

24.1

33.9

3.5

25.8

23.7

–4.4

13.8

–2.2

1970–85

0.7

34.4

383.9

–0.6

18.1

222.6

–4.3

24.9

1986–95

0.3

39.2

41.1

–1.6

21.0

69.5

1.6

37.0

0.1

1996–04

2.3

40.8

–18.5

–0.3

24.4

–9.2

2.7

33.4

–20.9

2005–11

7.1

52.2

–43.4

–1.8

34.4

–26.6

2.0

31.1

–27.5

China

India

Notes: A minus sign for currency valuation means that the currency is undervalued. Sources: Bhalla (2007a) dataset extended to 2011.

Korea 28.1

surplus, and indeed the surplus has been much higher since the introduction of the euro than under the much weaker Deutsche mark in the early 1980s. The same is true for Japan—the surplus is higher now than when the yen was weaker during the 1980s. For China and Korea, there is a close correspondence between CAB and currency, but for India, the currency has been undervalued for the last 15 years and the current account has worsened. Admittedly, this is a small sample of only six countries (regression analysis for a large set of countries is reported below), but the data for these selected countries are presented to underscore the fact that there is no obvious or straightforward relationship between current account imbalances and currency valuation. Trevor Swan (1956) pointed out that there needn’t be a predictable relationship between the two when he noted that internal and external balance could not be achieved with only one instrument, the real exchange rate.

Currency Valuation and the Current Account There is clearly a need for different hypotheses, empirical linkages, and tests. Eswar Prasad, Raghuram Rajan, and Arvind Subramanian (2007) contend that current account surpluses are a determinant of higher growth. Their reasoning is indirect and oriented more toward the ineffectiveness of foreign capital in aiding growth: Poor countries are less able to absorb foreign savings, and capital inflows therefore often lead to currency overvaluation. So their story is really about currency valuation, rather than the current account, and one could expect that, after accounting for the effects of the exchange rate, the significant positive relationship between the CAB and growth would disappear. However, according to their detailed analysis, the ineffectiveness of foreign savings to promote growth persists even after controlling for currency overvaluation. In particular, as measured by the method introduced by Simon Johnson, Jonathan Ostry, and Arvind Subramanian (2007), each 1 percentage point increase in the current account surplus raises per capita growth by 0.1 percent. I examine their results (Bhalla 2007b) and find that the relationship is even stronger when oil surplus countries are excluded.5 This strong relationship is also fragile, however; with just the exclusion of one outlier country, Singapore, the coefficient on the current account surplus (the dependent variable is per capita growth) is reduced to significance at only the 15 percent level. These results confirm that the Prasad-Rajan-Subramanian relationship between the CAB and economic growth is most likely unbalanced and weak. What is the effect of currency valuation on a current account surplus? It is expected that the larger the undervaluation (negative currency valuation), the larger the surplus. This is confirmed by regressions relating surplus to valuation for different time periods and specifications. The real price of foreign exchange strongly affects exports and imports, as reported in table 3.3 for nonoil-exporting countries for several five-year periods since 1965. The magnitude 5. I thank Arvind Subramanian for providing their data.

CURRENCY VALUATION, SAVINGS, AND THE CURRENT ACCOUNT 39

Table 3.3

Period 1965–79

Effect of currency valuation on the current account balance, 1965–2011 Coefficient

R2

Number of countries

Number of observations

–0.02***

0.082

81

162

1980–94

–0.026***

0.094

93

274

1980–2011

–0.037***

0.157

94

555

1995–2011

–0.048***

0.219

94

281

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Current account balance (percent of GDP) = α + β*currency valuation. Five-year panel data, 1965–2011, are as follows: 1965–69, 1970–74, 1975–79, and so on. Source: Author’s calculations.

of the coefficient on currency valuation has a strong tendency to rise over time. In the preglobalization years (before 1980), the coefficient was −.02—that is, each 10 percent increase in undervaluation raised the CAB by 0.2 percentage point. For 1995–2011, this coefficient is more than doubled, at −.048. More tests of the relationship between currency valuation and economic phenomena are reported throughout the book. These preliminary results are meant to suggest the likely impact of currency valuation on matters of considerable policy interest. That currency valuation affects the CAB is reassuring, but expected. It is the relationship with other economic variables—specifically, investment and per capita growth—that is controversial and needs rigorous theoretical and empirical analysis. This is undertaken in subsequent chapters.

40

DEVALUING TO PROSPERITY

4 Measuring Currency Valuation

In all affairs, it’s a healthy thing now and then to hang a question mark on the things you have long taken for granted. —Bertrand Russell This chapter deals with the critical issues of defining and measuring the real exchange rate and gauging the extent to which it deviates from its equilibrium.1

Equilibrium Exchange Rates One of the earliest attempts to define an equilibrium exchange rate was in terms of purchasing power parity (PPP). The phrase itself implies that when there is parity, prices are in equilibrium. The concept of PPP originated with Gustav Cassel (1921), who offered it in terms of the “law of one price”—that is, prices of traded goods, adjusted for tariffs and transportation costs, should be the same in all countries. The reasoning was that when this is not the case, arbitrage would lead to a convergence in prices. According to the law of one price, goods in country or region A should be priced the same as goods in country B. The transformation of country A’s price into country B’s price is, of course, via the nominal exchange rate. A weaker version of the law of one price is the law of relative prices, which posits that any differences in prices across countries at a point in time (called the base year or base period) can be expected to be equalized over time through changes in these relative prices. An example helps illustrate. Assume a computer costs $1,000 in the United States. Given that there are hardly any transportation costs (a computer is lightweight) or transaction costs (zero tariffs), the computer should cost 6,850 yuan 1. Again, when the valuation of a currency is below its equilibrium—when it is cheaper than it should be—the currency is deemed to be undervalued, and the valuation has a negative sign; when it is above its equilibrium value—more expensive than it should be—it is deemed to be overvalued, and the valuation has a positive sign.

41

in China if the exchange rate is 6.85 yuan to $1. If the US price were to increase by 10 percent, to $1,100, then importers would bring additional supply into the United States from other countries that can produce the computers at the same price, and the price in the United States would fall back to $1,000.2 This arbitrage argument, however, does not apply to the prices of services such as a haircut, a schoolteacher’s salary, local transportation, maid services, or the like.3 As noted, the transformation of the price in one country into the price in another country occurs through the nominal exchange rate. The law of one price assumes that prices will converge to equilibrium independently of whether the nominal exchange rates are fixed, pegged, dirty float, floating, or something in between. In fact, it is precisely because the nominal exchange rate does not reflect underlying real forces in the economy that the nature of the currency regime is irrelevant for calculating the equilibrium exchange rate. The equilibrium exchange rate is therefore the one that ensures the same purchasing power for one unit of currency (say, PPP dollars) in all countries, including the country issuing the comparator currency (in this case, the United States). Measuring changes in purchasing power, or deriving the PPP exchange rate for different years, therefore only requires information on relative inflation rates. Using the above example, let us substitute a composite good for the computer and international financial markets for the computer sales company. In this new world, domestic price changes take place via inflation (changes in the consumer price index) and arbitrage takes place through changes in the nominal exchange rate. To determine the PPP dollar exchange rate for a particular country, the price of the composite good in the local (home) currency is divided by the nominal exchange rate with respect to the US dollar.

Real Exchange Rates The concept of a real exchange rate (RER) extends the concept of the PPP exchange rate beyond the relative prices of a single good. The RER represents the ratio of prices for an identical set of goods in terms of a single reference currency, such as the dollar, at a given point in time. Hence, the phrase purchasing power parity. By defining an identical goods space at the same point in time, US$1 becomes PPP$1. RERh = (Ph/Pus),

(4.1)

where Ph is the domestic price level of country h, and Pus is the price level in the United States. Since Ph is measured in a local (home) currency, it is actu-

2. Only if there were full employment in every country that could produce additional computers at the same “fixed” price would it be impossible to supply the additional computers. 3. Although in one sense the market never ceases to work, as witnessed by the number of illegal immigrants working as maids in the United States. It is said that other than haircuts, there is nothing nontradable anymore.

42

DEVALUING TO PROSPERITY

ally obtained by dividing the domestic price of goods in local currency by the prevailing exchange rate with respect to the US dollar. This division translates a domestic price into a US dollar price. The price could be for a particular good or a set of goods. Equation 4.2 restates the definition: RER = (Eh × Ph )/Pf ,

(4.2a)

where Eh is the home currency exchange rate ($/Re, or how many dollars one unit of home currency can purchase), Ph is the price level in the home country, and Pf is the price level in the foreign country (the United States). Or with a dot above signifying log changes, 5(‫ ۽‬ƠK‫ۻ‬Kí‫ۻ‬I.

(4.2b)

Or, 5(‫ ۽‬ƠG‫ۻ‬Kí‫ۻ‬I,

(4.2c)

where ƠGis the change in the value of the domestic currency (with depreciation having a negative sign). Several aspects of the RER (a ratio) follow very clearly from these deceptively simple equations. First, (log) changes in the RER are identically equal to currency change ( ƠG), and the difference between domestic and foreign inflation ( ‫ۻ‬Kí‫ۻ‬I). This holds true for short periods of time but may not hold true over longer periods, during which relative prices within the domestic economy may change and may diverge from the changes occurring during the same period in the comparator country (the United States). This is the reason price surveys are taken in all countries every 5 to 10 years, with each new survey year becoming a new reference or base year. Second, a first (and most likely incorrect) approximation of currency misalignment is given by the RER. If it is less than 1—that is, goods cost less in the foreign country than in the United States—then the currency can be deemed to be undervalued. If the RER is greater than unity, then goods are more expensive than they should be, and the currency is overvalued. An example helps illustrate the concept. In 1970, India took part in the first United Nations International Comparison Program (ICP) along with 11 other countries. According to that price survey, an identical basket of goods costing $1 in the United States was priced at 2.53 rupees in India. The prevailing official exchange rate was 7.5 rupees to $1. Thus, the RER for India in 1970 was 0.34, the ratio of the prices in dollars in the two economies (2.53 divided by 7.5). In the same survey, the RER for Germany was 0.87. Given that both these RERs were less than 1, both currencies could be deemed to be undervalued— India by about 70 percent and Germany by 13 percent.4

4. As described in the next section, the Balassa-Samuelson effect (Balassa 1964, Samuelson 1964),

MEASURING CURRENCY VALUATION 43

If a price survey could be conducted every year, it would be straightforward to develop a time series of RERs for each country, but instead, doing so is difficult, time-consuming, and expensive. The ICP has conducted its survey three times since 1970—in 1985, 1993, and 2005.5 Hence the shortcut of using changes in domestic (and US) price levels to approximate the RER for each year. There can be discussion about which price measure should be used to signify inflation, the consumer price index, wholesale price index, GDP deflator, or another. Since the basket of goods being compared across economies comprises all goods—consumption, investment, and government goods— the GDP deflator seems most appropriate. There are also choices about which basket of currencies should be used, a bilateral exchange rate (typically against the dollar) or a multicurrency basket. The bilateral rate seems most appropriate because, although including additional currencies may improve the measure, the extra benefit is small.

The Balassa-Samuelson Effect The most original, most popular, and likely most accurate method of measuring currency misalignments resulted from independent papers by Béla Balassa (1964) and Paul Samuelson (1964). These grew out of a controversy over whether the dollar was overvalued in 1960. The conventional wisdom was that the dollar was indeed overvalued by as much as 20 to 30 percent. At exchange rates prevailing at the time, compared against the United States, the cost of the average composite good was about two-thirds in Japan, Italy, and the Netherlands and about three-fourths in Denmark, France, and West Germany. Given that there was very little variation in costs due to tariffs or transportation costs, these large differences in average prices were taken as a clear indication that the dollar was overvalued with respect to the yen and a variety of European currencies. Hendrik Houthakker, a distinguished US economist, summarized the conventional wisdom: Recent figures indicate that an average basket of commodities bought for $1 in the US would cost only 3.11 marks in Germany, while the official exchange rate is 4 marks to the dollar. We may say, therefore, that the dollar was overvalued against the mark by 22 percent. (Quoted in Samuelson 1964, 147)

In order for these relative prices to move toward equilibrium, the exchange rates would need to adjust; specifically, the dollar would need to depreciate which accounts for differences between countries in the relative productivity of the tradable and nontradable sectors, can modify the degree and even the direction of currency undervaluation. 5. The 2005 ICP survey results have sparked considerable controversy (see Bhalla 2008b, Deaton and Heston 2010), with the effect that a new round of base-year price measurements is ongoing for 2011. The implications of this new PPP exchange rate for GDP levels, growth, and currency valuation are discussed in chapters 8 and 14.

44

DEVALUING TO PROSPERITY

by about 20 percent. Looking at the same data, however, Balassa developed empirical estimates that suggested the dollar was not overvalued at all, and Samuelson formulated theoretical arguments that supported Balassa’s estimates. Their simple explanation is that, relative to Europe and Japan, the United States was more productive and wealthier and therefore had a higher overall price level, which meant a higher RER. This higher RER—as implied by equation 4.1, which defines the RER as the ratio of price levels—meant that the dollar was less overvalued. In other words, a strict comparison of price levels per se, or unadjusted RERs, would lead to the erroneous conclusion that the dollar was overvalued and that the Deutsche mark and Japanese yen were undervalued. Even if the link between the price level and the RER is clear, what is less obvious is why the price level is higher in wealthier countries. Balassa (1964) and Samuelson (1964) both point out that in wealthy or wealthier countries, productivity growth in the manufacturing/industrial sector is higher than in the services sector. With the labor market equilibrating wages across sectors, the prices of services would be higher in wealthier countries, even as goods arbitrage would ensure that the prices of goods were approximately equal across countries. Thus, the overall price level (the price of a composite item composed of both manufactures and services) would be higher in wealthier economies, and there would be no pure purchasing power parity. Goods prices would be the same in all countries, but overall price levels would be higher in the wealthier economies. As a result, differences in pure or unadjusted RERs are not a valid proxy for measuring real currency undervaluation. However, if the RER is adjusted to account for the differences in average productivity, then it is possible to make an appropriate comparison of currency valuation. Both authors contend that the dollar would be seen to be less overvalued after making the appropriate productivity adjustments. How much less overvalued the dollar was is a matter of estimating the ratio between the prices of tradables and nontradables.

RER as a Price Ratio Given the spotlight shined by Balassa and Samuelson on the relationship between the RER and the share of nontradables in the economy, research thereafter proceeded toward refining the definition of the real exchange rate. The Balassa-Samuelson explanation of differential price levels centered on the expectation that prices for tradable goods would be equal across countries; if not, then the changes in such prices would be very close to equal. Either way, that left changes in the prices of nontradables as the central determinant of movements in the RER. The RER within an economy therefore was redefined as the ratio of the price of nontradables (PNT) to tradables (PT): RERh = (PNT /PT)

(4.3)

MEASURING CURRENCY VALUATION 45

The RER between economies would then be the ratio of the individual country ratios PNT/PT ; and with the assumption of equality in PT, the RER would just reflect differences in the prices of nontradables: RER = (Eh × PhNT)/PfNT

(4.4)

The expectation is that this ratio would move in the same direction as per capita income, according to the hypotheses of Balassa (1964) and Samuelson (1964) and later research by Alan Heston, Daniel Nuxoll, and Robert Summers (1994). While theoretically elegant, equation 4.3 is problematic in empirical application. For starters, it is difficult to test for the relationship between PNT and PT because data are not readily available to decompose prices into nontradables and tradables. However, one method for doing this is to use the consumer price index as a proxy for nontradables and the wholesale price index as a proxy for tradables. This method may have worked in the past, when industry generally comprised only tradable goods and services included sectors such as government, banking, and construction—all nontradables. However, given that there are no longer many services that are nontradable, this approximation fails the plausibility test. Today, telecommunications, banking, and new services such as software are all tradables, and their prices reflect this. Thus, the ratio of PNT to PT has become considerably less relevant today than it was in the 1960s. The most damning evidence against the tradable-nontradable decomposition is that tradable goods prices do not move in tandem across countries. In a comprehensive evaluation of these prices in developed economies, Charles Engel (1999) finds that more than 80 percent of the variation in the RER was due to variations in tradable goods prices, a result that completely contradicts the prediction that variations in RER are due to variations in nontradables prices. His conclusion: The important impression from the evidence presented here is that the nontraded goods component has contributed to little of the movement in the real exchange rates for the United States at any horizon…. Knowledge of the behavior of the relative price on nontraded goods contributes practically nothing to one’s understanding of U.S. real exchange rates. (Engel 1999, 510, italics added)

The rest of this chapter (and the book) sticks with the definition of the real exchange rate in equation 4.1—the ratio of a country’s PPP price level to the US price level, with these price levels defined by the respective GDP deflators.

Estimating Currency Valuation Using the Balassa Method Balassa (1964) was also the first to provide a method for estimating currency valuation, and his model is the basis for most estimates published since. The Balassa method is extremely straightforward. It postulates a stable functional relationship between the RER and income: 46

DEVALUING TO PROSPERITY

Table 4.1

Estimates of currency misalignments, 1960 Unadjusted RER

Misalignment (percent deviation of RER from RER*)

Country (1)

Balassa (1964) (2)

Bhalla (2007a) (3)

Balassa (1964) (4)

Bhalla (2007a) (5)

Johnson, Ostry, and Subramanian (2007) (6)

Belgium

0.80

0.67

–1

–11

17

Canada

0.93

0.95

5

6

48

Denmark

0.77

0.53

–5

–57

–12

France

0.77

0.72

–1

–3

25

Germany

0.78

0.77

–2

–5

30

Italy

0.70

0.49

5

–33

–12

Japan

0.63

0.44

1

–12

–16

Netherlands

0.67

0.49

–17

–54

–17

Norway

0.80

0.71

2

–8

23

Sweden

0.90

0.79

9

–12

29

United Kingdom

1.00

1.00

3

–33

6

United States

0.82

0.62

–1

–12

50

RER = real exchange rate Notes: Balassa (1964) estimates a linear equation between his own estimates of RER and real income y to derive RER* (predicted value of regression). Bhalla (2007a) obtains the predicted RER from the equation RER = 1.11*(1 – 0.97^y), where the RER is the measured real exchange rate and y is per capita daily income at 1996 purchasing power parity prices. Johnson, Ostry, and Subramanian (2007) use Penn World Table 6.2 data for the RER and estimate a log-log relationship between the RER and per capita income to obtain RER*; this regression is estimated for each year. See appendix A for details. Sources: Balassa (1964); Bhalla (2007a); Johnson, Ostry, and Subramanian (2007); author’s calculations.

RER = a + b*y,

(4.5)

where a is a constant and b is the response of RER to each 1 unit increase in the per capita income in constant dollars, y. The predicted value from this equation, RER*, is defined as an estimate of “equilibrium,” and the percentage deviation of the RER from RER* is the estimate of currency misalignment. Table 4.1 presents the Balassa (1964) estimations for 12 Organization for Economic Cooperation and Development countries in 1960, as well as estimates for the same year and same countries by two recent studies (Bhalla 2007a; Johnson, Ostry, and Subramanian 2007). Balassa’s adjustment for income level (with higher-income countries having higher price levels) yielded a very different picture of currency misalignment (Balassa 1964, table 1 and figure 1). Several countries were closer to the equilibrium rate, including Japan (column 4 of table 4.1). The unadjusted RER for Japan indicated a 37 percent undervaluation, but the adjusted RER MEASURING CURRENCY VALUATION 47

suggested zero misalignment. Ditto for the United States—adjusted for Balassa-Samuelson income effects, the dollar was observed to be 0.8 percent undervalued.6 Bhalla (2007a) also reaches the conclusion that Japan and the United States had equal misalignments (–12 percent for both compared with Balassa’s estimate of near 1 percent for both economies). The two real outliers were Sweden (overvalued by 9 percent) and the Netherlands (undervalued by 17 percent). Balassa (1964) showed empirically that a very simple adjustment for relative prices could change perceptions about whether a currency was under- or overvalued. The traditional method had been to use the raw unadjusted real exchange rate (column 2 or 3 of table 4.1) to assess currency valuation. The new method was to adjust for the Balassa-Samuelson effect, by correcting for per capita income levels. But how best to make this correction? Columns 4, 5, and 6 in table 4.1 each show the results of a different functional form to estimate currency misalignment. Balassa (1964) uses a simple arithmetic relationship between RER and y; Bhalla (2007a) uses a nonlinear relationship (details below); and Johnson, Ostry, and Subramanian (2007) use a log-log relationship. Both Balassa and I reach near identical estimates for Canadian dollar valuation: overvalued by 5 and 6 percent, respectively. In contrast, Johnson, Ostry, and Subramanian reach very different results: The Deutsche mark is overvalued by 30 percent (rather than the relatively fair valuation estimate reached by Balassa and myself), the Canadian dollar is overvalued by 48 percent, and the US dollar is overvalued by 50 percent.

Currency Misalignments After the pioneering work of Balassa (1964) and Samuelson (1964), research on the RER and currency misalignment proceeded in two related, yet distinct, directions. The first is to measure currency valuation and, specifically, to estimate an “equilibrium” value for the real exchange rate. The second is to apply currency valuation estimates to models explaining economic growth and, specifically, to estimate the impact on economic growth of deviations from the equilibrium RER (called misalignments or over- and undervaluations). The task of measuring currency valuations became much easier when Irving Kravis, Alan Heston, and Robert Summers (1988) made available PPP data for several countries and years.7 Researchers no longer had to construct their own constant PPP price series for each country (as Balassa had to do); the real exchange rate was provided in the Kravis-Heston-Summers data as the

6. In figure 1 of Balassa (1964, 590) the United States appears to be 1 percent overvalued after adjustment for relative income effects, but this can be attributed to the lack of precision in precomputer calculations. 7. Some years later, revisions to these data came to be known as the Penn World Tables. This book primarily uses Penn World Table 6.1, whose primary source is the 1993 ICP survey, although some estimates are made using other Penn World Tables as well as the latest 2005 ICP data.

48

DEVALUING TO PROSPERITY

ratio of the price level of each country relative to the United States. Thus, the Balassa (1964) exercise could now be replicated for a large sample of countries, which is precisely what Irving Kravis and Robert Lipsey (1986) did.8 The Balassa-Samuelson thesis—that real exchange rates increased with per capita income—was confirmed by data for a number of countries, especially developed economies including Japan. Table 4.2 summarizes the research oriented primarily toward measuring the RER and establishing its relationship to income. Also reported, when available, is the elasticity of RER with respect to income. Because most of these estimates use a simple log-log model, the elasticity is simply the estimate of the coefficient of (log) per capita income. Table 4.3 summarizes the research on RER and income, which has had a dual objective: estimate the level of currency misalignment and then relate it to economic growth. David Dollar (1992) was among the first to estimate currency valuation. He used the last 10 years (1976–85) of the Kravis, Heston, and Summers (1988) data to estimate openness or the outward orientation of economies, using the deviation between the actual and the predicted RER as an indicator of openness. This is no different than estimating exchange rate misalignment using the Balassa method. The functional form used by Dollar is quadratic—both the level of per capita income and the square of such income are in the equation. Subsequently, the Dollar measure was used by William Easterly (2005) to conduct a time series of currency valuation for several countries from 1950 to 2003. Via Easterly, Dollar’s “openness” variable became possibly the most used variable representing currency misalignment. (This “collaboration” is referred to here as the Dollar-Easterly dataset.) In Bhalla (2002b) I found changes in the real exchange rate to be a significant explanatory variable for growth acceleration—the greater the change in the real exchange rate, the higher the acceleration in the per capita growth rate. This study is among the very few that find a statistical relationship between currency misalignment and growth. Building on this and earlier work,9 I employ and report a new estimate of RER equilibrium and misalignment in Bhalla (2007a). This new measure of the equilibrium exchange rate yields very robust results that are not sensitive to outliers, to the introduction of other variables, or to the time period under study. This new measurement assumes that the underlying relationship between the RER and income is S-shaped. The prevailing Dollar-Easterly measure of currency misalignment was discarded by other authors who used a log-log model, including Maurice 8. They used a preliminary version of the Kravis-Heston-Summers 1988 data. 9. Bhalla (1992) uses productivity-adjusted RERs to signify equilibrium. The analysis supported the conclusion that the British pound was considerably overvalued at the time of the Exchange Rate Mechanism crisis in 1992, and therefore that the Germans were not at fault. Bhalla (1998a) concludes that, over the years, the Chinese renminbi had become somewhat undervalued and that this put pressure on East Asian economies to devalue as well.

MEASURING CURRENCY VALUATION 49

Table 4.2

Estimating the real exchange rate (RER): Methods and findings

Source Balassa (1964)

RER–income elasticity 0.32

Description Estimated as a sample linear equation for 12 OECD countries for 1960, RER = 0.49 + 0.00025*y, where y is Balassa’s estimate of purchasing power parity income

Kravis and Lipsey (1986)

Fundamental determinants model; RER is a function of terms of trade, etc.; panel data 1962–84 for 12 developing countries

Edwards (1989)

Fundamental determinants model; RER a function of terms of trade, etc.; panel data 1962–84 for 12 developing countries

The Economist’s Big Mac Index (1989)

0.26

Assumes there is only one good in the economy, a McDonald’s Big Mac sandwich; the price of this uniform good is compared with the price of the Big Mac in the US to obtain an estimate of RER

Bhalla (1992)

Real exchange rate equilibrium assumed to be in 1987; excess inflation and output growth (relative to US) determine new equilibrium RER levels, and hence new under- or overvaluation levels

Williamson (1994)

RER estimated using the “fundamental equilibrium exchange rate” method; equilibrium RER assumed to be that which creates a zero current account balance (now assumed to be different for different countries)

Rogoff (1996)

0.37

Estimates log-log model

Goldfajn and Valdés (1999)

Uses a cointegration approach on fundamental variables like terms of trade and size of government to determine equilibrium RER

Easterly (2001)

Accepts Dollar (1992) framework and extends the analysis to 1950–2003 (from a single point estimate of 1976–85) for a larger set of countries

Hinkle and Montiel (2001)

Comprehensive review of methods of measuring RERs

Bergin, Glick, and Taylor (2004)

0.25

Estimates the log-log model for several time periods and extends the analysis back to the 1880s; RER income elasticity range of 0.12 to 0.46 observed for the period 1880–2000

Frankel (2006)

0.37

Estimates a log-log model using Penn World Table 6.1 data

Thomas, Marquez, and Fahle (2008)

0.18

The WARP estimates US RER based on monthly trade and exchange rate data for major US trading partners

OECD = Organization for Economic Cooperation and Development; WARP = weighted average relative price

50

DEVALUING TO PROSPERITY

Table 4.3

Currency valuation and growth: Methods and findings

Source

RER–income elasticity Description

Cottani, Cavallo, and Khan (1990)

Uses dynamic least squares technique to measure real exchange rate (RER), then tests the effect of the volatility of misalignment on growth

Dollar (1992)

0.21

Estimates a quadratic model, y = a + bx + cx2, with continental dummies; y is RER and x is per capita income; varying elasticities obtained with average at 0.21; model estimated for a cross section of countries with data pooled for 1976–85

Razin and Collins (1997)

0.30

RER and misalignment determined via RER determinants (terms of trade, average education, share of trade in GDP, log income, etc.); similar framework as Cottani, Cavallo, and Khan (1990)

Bhalla (1998b)

Extends Bhalla (1992) and relates currency wars to mercantilism and the Asian crisis of 1997–98

Bhalla (2002b)

Extends Bhalla (1998a, 1998b) and estimates a model relating growth acceleration to change in currency valuation

Aguirre and Calderón (2005)

Uses dynamic least squares technique to measure RER, then tests the effect of the volatility of misalignment on growth

Aghion et al. (2006)

0.23

Estimates the effects of both the variance and level of misalignment on growth; misalignment measured as the deviation of RER from a model relating RER to its fundamental determinants

Bhalla (2007a)

0.77

“Exponential” model, which results in an S-shaped evolution of real exchange rate with respect to income RER = b0(1 – b1y) where b0 = 1.11, b1 = .97 and y is per capita income. An S-shaped relationship between the RER and income is postulated and estimated. This method yields varying elasticities; average of 0.72 is for the time period 1996 to 2009

Johnson, Ostry, and Subramanian (2007)

0.26

A regression of log RER and log per capita is estimated for each year; the elasticity reported is the average of such elasticities and is in the range of 0.09 to 0.37

Rodrik (2007a, 2008)

0.25

Same as Johnson, Ostry, and Subramanian (2007) except model is estimated for five-year periods rather than each year

Obstfeld and Kenneth Rogoff (2000), Jeffrey Frankel (2006), and Simon Johnson, Jonathan Ostry, and Arvind Subramanian (2007). While these authors estimated the RER–income elasticity based on annual cross-country data, Dani Rodrik (2007a) used five-year averages. Not surprisingly, the two estimates are fairly close, with correlation coefficients above 0.95 for most years. MEASURING CURRENCY VALUATION 51

Thus, prior to 2006, there was only one measure of currency valuation available for a long time period and for a large cross-section of countries—the Dollar-Easterly measure or some derivative thereof (such as Aghion et al. 2006 or Aguirre and Calderón 2005). Or, to be more precise, the Dollar-Easterly measure was the only one that corrected for the Balassa-Samuelson effect on the evolution of RERs. To be sure, there were alternate RER estimates (and deviations), including the estimates by the International Monetary Fund (IMF) or the Bank for International Settlements (BIS) of the real effective exchange rate (REER), which is calculated simply on the basis of differential inflation and differential changes in the nominal exchange rate. There are two important drawbacks to such REER calculations: First, they do not correct for the Balassa-Samuelson effect, and, second, and perhaps more important, they only yield estimates of changes in valuation relative to a base equilibrium year. This is very restrictive, because each country likely has its currency in equilibrium in a different year. Today, in contrast, researchers and policy analysts now have at least six easily replicable currency-valuation series: the IMF (and BIS) REER measures (with their deficiencies); the Dollar-Easterly measure (available only until 2003); and the measures by Bhalla (2007a); Johnson, Ostry, and Subramanian (2007); and Rodrik (2007a).10 The most robust methods for relating RER and income are those that can best explain diverse and important economic phenomena. Such methods should also be based on theory and should pass several “smell” tests related to their application and predictive abilities. These tests are an important part of this book.

Real Exchange Rates and Income The method for assessing currency misalignment gauges the RER as a function of income and under- or overvaluation as the deviation of RER from its predicted value. Therefore, the estimation of the proper functional relationship between RER and income is critical. What should we expect the relationship between the RER and income to be? A linear or log-linear specification has some obvious drawbacks. Balassa (1964) used a linear relationship between the RER and income for developed economies whose per capita incomes were in a narrow range; in this case, a linear specification was most likely appropriate. However, today there is enormous variation in the income levels of the nearly 100 developing economies (not to mention developed economies). In 1980, the level of per capita income ranged from PPP$22,000 in Chile to only PPP$443 in Uganda—a ratio close to 50. In contrast, in the Balassa sample the ratio was only 4 between the richest (United States) and poorest (Japan). 10. The fundamental equilibrium exchange rate measure of the Peterson Institute for International Economics (Cline and Williamson 2011) is available for selected years and selected countries; this makes it difficult to use the series in a cross-country framework.

52

DEVALUING TO PROSPERITY

Implicit in a lot of RER models is the assumption that the elasticity between the RER and income is broadly constant across a wide range of incomes (and for the log-log model that it is literally constant). If this assumption proves untrue, there could be large measurement errors in the estimations of currency misalignment, and these errors would exert a bias on any resulting effect on growth by the currency valuation, pushing them toward zero. Balassa (1964) shows that for wealthy countries this relationship is likely to be linear, even log-linear. And the theory and evidence of catch-up show that wealthier countries grow at a much lower rate than poorer, developing countries. Combining the two yields the stylized fact that wealthy countries can be expected to have a lower-than-average RER–income elasticity. Another reason to expect this low, relatively flat elasticity is to prevent the unrealistic result of the RER becoming very large once a country becomes wealthy. This is the theoretical flaw with the Balassa linear, Dollar quadratic, and Johnson-OstrySubramanian, and Rodrik log-log formulations—the RER keeps increasing with income. The same problem occurs at low levels of income—the RER is expected to be low and flat. Again, the extrapolation of the log-log relationship would imply negative RERs at very low levels of income, a theoretical impossibility. If the RER levels are expected to be flat at low and high levels of income, then an S shape is probably the best functional form. The importance of the functional form is underscored by the correlation coefficients. The Balassa estimate has a coefficient of only 0.22 with the log-log form and 0.57 with the nonlinear form. The key statistic is the elasticity of RER with respect to income, which is central to any calculation of currency valuation.

Conventional Wisdom on Income Patterns During the process of development, the levels of income and growth often trace an S shape: The growth rate is slow at first, then accelerates, and then slows once the economy becomes developed. Once this happens, per capita income plateaus, albeit at a very high level. Walt Whitman Rostow (1960) characterized this third stage of growth as the take-off stage, which corresponds to the steep slope area of the S. The final stage, when the economy has matured, is the flat portion of the S. Earlier, much earlier than Rostow, it was conventional wisdom that income levels followed a nonlinear, S-shaped path (as discussed in chapter 2). The reasoning was as follows. In the early stages of development, agriculture accounted for 70 to 80 percent of output, and its growth was around 3 percent. Movement of agricultural workers to the higher-productivity nonagricultural sector (which grew at, say, 6 percent a year) only marginally affected the overall GDP growth rate. Hence, in these early years, per capita income had a relatively flat growth pattern—the bottom end of the S. Once the agricultural share of GDP dropped below 50 percent, accounting itself dictated that the overall MEASURING CURRENCY VALUATION 53

growth would accelerate—the steep part of the S. And when the economy became developed or the share of agriculture dropped below 25 percent, the reallocation effect was about over and income growth again decelerated—the top part of the S. There are several individual country growth paths that fit the S pattern. India is one (see figure 2.1). Several researchers, including Dani Rodrik and Arvind Subramanian (2004) and Atul Kohli (2006), argue that the economic reform process in India started in the early 1980s and caused the growth rate to accelerate to above 5 percent. I examine the role of the reallocation process in increasing India’s GDP annual growth rate from around 3.5 percent in the early 1960s to 5.5 percent by the late 1980s (Bhalla 2010). Indeed, this reallocation explains almost the entire acceleration in Indian growth to above 5 percent in the early 1980s. The S-shaped pattern predicted by the RER and the growth in income is suggestive of what is termed, in jargon, an “exponential form with one asymptote.” Estimations of this S-shaped functional form generate exceptionally powerful explanatory results—R2 equal to 0.87 compared to an R2 equal to about half that amount with a log-log model (discussed below). The estimated nonlinear relationship for the S-shaped curve is as follows: RER = 1.11*(1 – 0.971y), Number of observations = 2,496, R2 = 0.86,

(4.6)

where y is income per capita per day in 1996 PPP dollars.11 The time period of estimation is annual data for 1996–2009. The starting year is 1996, the base year for estimates of PPP (in the Penn tables). The equation yields an estimate of currency misalignment with respect to the PPP dollar. This estimate is given by the difference between the actual RER and the predicted value of the RER (RER*). How this misalignment with respect to the PPP dollar is converted into misalignments with respect to the US dollar is detailed in chapter 7. Figure 4.1 plots the predicted RER and actual RER for 2011 against per capita income, per equation 4.6. If the outliers are removed, the relationship between the RER and (log) income per capita is nonlinear and predominantly S-shaped. At low levels of income, the country in question is likely to be overly dependent on agriculture and not very involved in international trade—in other words, it is likely to be a closed economy. The exchange rate, nominal or real, is unlikely to play a large role; the elasticity of RER with respect to income is close to zero. Once a country advances to middle income status, with income

11. Nominal PPP 1996 income for years after 2000 are obtained from IMF (2011) and World Bank (2010) and linked to the Penn World Tables. Conversion into real values is done via the US GDP deflator. Twenty-five outliers (estimated according to the Hadimavo method) are excluded from the estimation. Inclusion of the outliers changes the first coefficient to 1.10 from 1.11 and does not affect the second coefficient.

54

DEVALUING TO PROSPERITY

Figure 4.1

Real exchange rate and income per capita, 2011: An S-shaped relationship

real exchange rate 1.8 Actual RER Predicted RER

1.6

che dnk

1.4 1.2 1.0

zmb

jam

aus swe aut fin franld jpn bel deuirl can esp ita nzl grc gbr usa prt isr

cri bra hrv hun lbnmex chl svn cze nam ury rom est svk ltu lva pol kor lbr per tur zafbwa gtmjor tza mng col ecu oan hnd bgr eri alb pan slv mkddom civ mysarg ben mrt mdg ginbol blr pry tgo sen khm tha egy caf tjk mli cmr png idn swz mar ken bih chn gnb mwi gha ner bfauga hti lao mdaphllka arm ukr tun mus pak gmbnpl nic vnm ind uzb bgd eth moz geo sle kgz bdi

0.8 0.6 0.4 0.2 0 400

4,000

40,000 PPP per capita income, 1996 prices (dollars)

Notes: The predicted real exchange rate (RER) is obtained from the equation RER = 1.11*(1 – 0.97^y), where the RER is the measured real exchange rate and y is per capita daily income at 1996 purchasing power parity prices. See table B.1 in appendix B for country abbreviations. Sources: See Bhalla (2007a) and appendix A for further details.

per capita about $1,000 (corresponding to approximately PPP$4,000), it can be expected to increasingly engage in international trade. This is the muchdiscussed “openness effect.” At this stage, exchange rates and competitiveness begin to matter. For most products, there is a flat global demand curve, and small margins of advantage can become crucial for export success. This is the stage when income elasticity is likely to rise. Given that the elasticity for developed economies is about 0.3, RER–income elasticity can be expected to peak at about 1 and then to decline at higher levels of income. Some obvious patterns emerge from the cross-country analysis in figure 4.1. Sub-Saharan African countries are distinguished by overvaluation—the actual values are considerably above the predicted RER line. Countries of East Asia are distinguished by being some distance below the predicted line—the currencies of these countries are undervalued, with the distance away from the line measuring the magnitude of currency valuation, under and over. Sub-Saharan Africa has problems of slow growth; East Asia has problems of fast growth. The conclusion is self-evident: Before reaching developed-country MEASURING CURRENCY VALUATION 55

status, countries with overvalued exchange rates (above the predicted RER line) are generally poorer, while countries with undervalued exchange rates (below the line) are generally richer.

Alternative Models One feature of the relationship between the RER and per capita income is that the RER converges to some flat value at high levels of per capita income. This follows an S-shaped pattern, with the RER staying at a low level for a long period during development and then rising toward its convergence level. By construction, the S-shape equation 4.5, y = b1*(1 – b2 y), converges to b1 when y becomes very large (as long as b2 is less than 1, which it is). The coefficient b1 is the average convergence level of RER for wealthy countries. It is estimated at 1.11—that is, as incomes rise beyond a certain level, the equilibrium RER does not change much. The convergence level is at an annual income level of about PPP$33,000 per capita. This property is not shared by the popular log-log model, in which estimates of the RER can reach infinity at high levels of income and sink to minus infinity at low levels. The log-log equation yields the following estimate for the same set of observations: Log(RER) = –1.81 + 0.36*log(y), Number of observations = 2,496, R2 = 0.46.

(4.7)

The coefficient on log y, 0.36, is constant across all values of per capita income. In the S-shape model above, the elasticity varies with income—flat at first, then increasing and large, and then flat again. In the log-log model, the RER elasticity (0.36) is the same for Ethiopia (at a per capita income level of PPP$1.101 in 2011 at 1996 prices) and the United States (per capita income level of PPP$35,680 in 2011). In the S-shape model, the RER income elasticity is 0.94 (Ethiopia) and considerably lower (0.17) for the United States. By forcing the RER elasticities to be the same for widely different levels of income, the log-log model introduces a measurement error. One estimate of the magnitude of this error is suggested by the vast difference in R2 produced by the two models. The log-log model yields a synthetic R2 (calculated as the R2 or variation in RER from equation 3.7) of 0.58, considerably lower than the 0.86 obtained for the S-shaped curve. Equations 4.6 and 4.7 capture the essence of the difference between the method used here and that predominant in the literature. The relationship is between identically measured variables, RER and y, but the functional form and the elasticity estimates are different. The constant elasticity in the log-log model can be considered as a weighted average of the “true” elasticity along different levels of per capita income as indicated by the S-shaped curve.

56

DEVALUING TO PROSPERITY

Figure 4.2

Elasticity of real exchange rate (RER) to income per capita, by income group, 1950–2010

RER–income elasticity (percent) 1.2 Maximum elasticity Mean elasticity Minimum elasticity

1.0 0.8 0.6 0.4 0.2

> 35,000

> 20,000 & ≤ 35,000

> 15,000 & ≤ 20,000

> 10,000 & ≤ 15,000

> 8,000 & ≤ 10,000

> 5,000 & ≤ 8,000

> 3,000 & ≤ 5,000

> 1,500 & ≤ 3,000

> 1,000 & ≤ 1,500

> 750 & ≤ 1,000

> 500 & ≤ 750

> 200 & ≤ 500

0

annual per capita income group (constant PPP dollars) Notes: Countries classified by income per capita, 1996 purchasing power parity (PPP) prices; five-year data for 1950–2011; elasticity obtained from RER = 1.11*(1 – 0.97^y). Source: Bhalla (2007a) dataset extended to 2011.

Accurate Measurements It is worthwhile to carefully consider the RER–income elasticity calculations, since they are key to understanding the debate on currency valuation. The Balassa-Samuelson formulation emphasizes that a higher equilibrium RER is part and parcel of the development process. But the nature of this upward path is ambiguous. If the RER-income relationship is indeed S-shaped, then imposing an average elasticity is incorrect. Tables 4.2 and 4.3 outline the difference this makes. The log-log models almost all produce a RER–income elasticity of about 0.25–0.35. The Johnson, Ostry, and Subramanian (2007) log-log model, estimated for each year, shows an average of 0.26 with a minimum maximum range of 0.09–0.38. Rodrik’s (2007a) panel estimate yields an elasticity of 0.25. Figure 4.2 plots the minimum, mean, and maximum elasticity of the RER to income per capita by income group for the S-shape model. Except for the last two income groups,

MEASURING CURRENCY VALUATION 57

above PPP$20,000, there is no overlap between the Bhalla (2007a) elasticities and those yielded by the methods, let alone a close one. This means that most models produce elasticities that miss the big picture for all but the wealthiest countries. The elasticities for more than 90 percent of the world’s population are around 0.7, if not higher, but most models estimate these to be below 0.3. One clear implication is that currency valuations for most developing economies, especially the fast-growing ones, are vastly underestimated by the conventional log-log models. Equivalently, the log-log models underestimate predicted RERs for developed economies, and therefore the currency valuations estimated by these models for developed economies are biased toward showing overvaluation. This mismeasurement has consequences for policy. The conclusion that currency valuation does not matter for growth—all too common in the literature—may spring from large measurement errors in estimating the predicted RER. If the true relationship between income and the RER is S-shaped, but if the estimated relationship is linear in logs, then there will be large measurement errors in estimations of the predicted real exchange rate and therefore in estimated currency valuations (the deviation between the actual and predicted RERs). One of the basic theorems of econometrics is that the presence of measurement errors biases any estimated relationship toward zero. In other words, the conventional result obtained by most authors that currency valuation does not matter, or that it matters only weakly, is neither conceptual nor real, but may really be due to measurement errors. Whether this is the case is assessed through a large number of “smell” tests of the different methods.

Different Measures of Currency Valuation Ordinarily, measurement errors and how they may bias results are topics of conversation only between academic economists. But in this case, the topic has a real-life application, especially to policy. Over the years, and especially since about 2002, there has been a continuing debate about whether certain currencies were overvalued or undervalued, especially the Chinese renminbi. After reviewing all the evidence, Steven Dunaway and Xiangming Li (2005, abstract) conclude: The number of studies attempting to estimate the “equilibrium” real value of China’s currency has proliferated in recent years…. These studies have sought to establish whether or not a significant part of China’s competitive prowess can be attributed to foreign exchange value of the renminbi. Unfortunately, no consensus has emerged because the studies yield a very wide range of estimates. [This] paper looks at a sample of these studies, with estimates of undervaluation ranging from zero to nearly 50 percent. It attributes the wide variation in these estimates to the influence of such factors as the different methodologies used, explanatory variables, subjective judgments of the various researchers in deriving their results, and instability in underlying economic relationships, especially in a rapidly developing economy like China. 58

DEVALUING TO PROSPERITY

The debate on how to properly measure currency valuation is similar to another debate that arose around the same time (2002) involving the poverty estimates of the World Bank (see Chen and Ravallion 2001). Again, different estimates of consumption (national accounts or surveys) and different estimates of PPP exchange rates (consumption or income) yielded poverty estimates that were, once again, all over the map. For example, for India, poverty levels ranged from as low as 13 percent of the population to as high as 45 percent for the same year and using the same definition of poverty (income below $1.08 per capita per day in 1993 prices) (see Bhalla 2007a). The challenge was the same: What to do when confronted with widely varying estimates for the same phenomenon? The solution was to subject the various methods and estimates to a battery of robustness and smell tests. The ones that came out relatively unscathed were most likely to have the fewest measurement errors and were to be preferred. The solution to the currency valuation challenge may be the same.

Smell Tests Smell tests are not rigorous statistical tests. Instead, they are meant to give a flavor for what seems right and what seems unrealistic. Smell tests take several different forms, such as different estimates for the same country at a point in time, different patterns of change for different countries at different points in time, and so on. Currency valuation estimates are given for a small set of developing and developed economies in table 4.4 for 1970–2011. Three methods are used for the estimates: the quadratic functional form of Easterly (2005), the nonlinear form of Bhalla (2007a), and the log-log form of Johnson, Ostry, and Subramanian (2007). For Brazil, all three methods yield broadly similar estimates for the different years. For India, the quadratic and log-log forms yield similar estimates for 1985 (undervaluations of 23 and 18 percent, respectively), but the nonlinear form shows a currency overvaluation of 131 percent in 1985. In 2000, both the quadratic and the nonlinear models show the Korean won undervalued, while the log-log model shows it being overvalued. The nonlinear and the log-log model also have vastly different estimates for China: In 1985, the log-log model shows an undervaluation of 23 percent, while the nonlinear method suggests an overvaluation of 106 percent. Table 4.5 shows currency valuation estimates for 2011 for a different set of countries and for two additional methods. The first is from William Cline and John Williamson (2011), who estimate fundamental equilibrium exchange rates (FEERs).The second is The Economist’s Big Mac Index, which assumes that the consumption basket consists of only one good, a McDonald’s Big Mac sandwich. Differences in the price of this uniform good are taken to be indicative of differences in prices of all goods and services and hence of currency misalignments. For a number of countries, the Cline-Williamson estimates are similar to MEASURING CURRENCY VALUATION 59

60 DEVALUING TO PROSPERITY

Table 4.4

Currency valuations for selected countries, 1970–2011 (percent) Easterly (2005)

Country

Bhalla (2007a)

1970

1985

1995

2000

2011

Brazil

n.a.

–31

35

n.a.

n.a.

China

Developing countries

Country

1970

1985

1995

2000

2011

Brazil

14

–25

40

–6

72

Developing countries

n.a.

n.a.

n.a.

n.a.

n.a.

China

477

106

–1

–18

–44

India

7

–23

–49

–53

n.a.

India

208

131

24

–12

–26

Korea

8

–12

23

–10

n.a.

Korea

27

–4

8

–21

–31

Developed countries

Developed countries

Germany

–16

–25

36

–13

n.a.

Germany

–16

–16

60

3

34

Japan

–30

3

111

66

n.a.

Japan

–33

–5

94

55

40

United Kingdom

–21

–22

7

4

n.a.

United Kingdom

–28

–19

14

10

24

United States

n.a.

n.a.

n.a.

n.a.

n.a.

United States

5

9

–15

1

–7

1970

1985

1995

2000

Brazil

–12

–30

28

2

51

China

52

–23

–39

–32

India

1

–18

–37

–41

–19

–7

19

5

Johnson, Ostry, and Subramanian (2007) 2011

1970

1985

1995

2000

2011

Germany

25

18

83

39

42

–47

Japan

–4

34

122

108

49

–51

United Kingdom

5

11

30

58

31

–27

United States

61

47

12

29

3

Developing countries

Korea

Developed countries

n.a. = not available Notes: For each study, currency valuation is the percentage deviation for real exchange rate from its predicted value. See text for details of construction of each estimate. Source: Bhalla (2007a) dataset extended to 2011.

Table 4.5

Currency valuation estimates, 2011 (percent)

Country

Bhalla (2007a)

Cline and Williamson (2011)

Johnson, Ostry, and Subramanian (2007)

The Economist’s Big Mac Index

Brazil

72

2

51

52

Chile

1

–5

0.1

–2

China

–44

–29

–47

–44

34

–6

42

21

India

–26

–16

–51

–53

Japan

40

–6

49

0

Korea

–31

–11

–27

–14

Russia

4

–5

3

–34

–11

–38

–11

–10

–7

–6

3

12

Germany (euro area)

Singapore United States

Notes: A negative value signifies undervaluation. The estimates of currency valuation for the United States for Bhalla (2007a) and Big Mac Index are obtained as the negative of the weighted average of currency valuation for the 37 countries included in the US Federal Reserve’s Broad Index. The weights are the trade shares with the United States. For Cline and Williamson (2011), the estimates are the percentage difference between the actual exchange rate for 2011 and the fundamental equilibrium exchange rate (FEER). For the United States, their “target change” estimate is reported.

the Bhalla (2007a) estimates; for example, the US estimates are almost identical, at a nearly 7 percent undervaluation. For Germany and Japan, the two diverge quite considerably: Cline-Williamson estimates indicate that these currencies are undervalued, while the Bhalla (2007a) method reports them to be overvalued. The Big Mac Index yields broadly the same estimates as Bhalla (2007a), which is mirrored by the fact that the correlation coefficient between the two for 2011 for a broader sample of 46 countries is 0.73. In contrast, the correlation coefficient between Bhalla (2007a) and the Cline-Williamson estimates for a sample of 42 countries is only 0.33.

How Divergent Are the Estimates? Figures 4.3 and 4.4 track the correspondence, or lack thereof, between the three different functional forms. The comparison with the quadratic form is for 1980, since the Dollar (1992) formulation of currency valuation was for 1976–85. As noted, Easterly (2005) took these estimates and extended them backward to 1960 and forward to 2003. The charts comprise four quadrants: Quadrants I and III represent the same sign for valuation as indicated by the two estimates; quadrants II and IV represent the opposite signs. The solid line in the graph is a 45q line, and any observation on or close to the line represents near equivalence between the methods. For example, in figure 4.3, Sweden is on the line, suggesting that in 1980, both the Dollar-Easterly and Bhalla (2007a) methods considered the Swedish krona to be overvalued by about 50 percent. MEASURING CURRENCY VALUATION 61

Figure 4.3

How accurate are the 1980 currency valuation measures?

Dollar and Easterly (percent) 100 civ

IV

caf

80

gab

60

dza arg bol

gmb zwe

swe

40

mus

ner

I

cog

cmr

lso pry mar hnd bwa jam che fin irl isl aus dnkdeu dom jor pan gbr nld egy ury sur bel oan fra jpn aut slv chlecu isr nor cri png bra kor nzl esp cyp ita lux ven tur prt phl fji irn bhr idn mlt mys tto sgp col zaf mex per tha can nic syr

tgo sen rwa

sle

mdg

grc

20 0 usa

–20 –40

sdn bdi

ken

gin hti ind pak npl

eth

bgd

–60 lka

–80

III

II

–100 –50

0

50

100

150

200 Bhalla (percent)

Notes: The figure presents figures for (log) valuations. If the two valuation indices are within ±30 percent of each other, then they are marked with a square symbol. The 45˚ line measures equality between the two estimates. Countries marked with a circle symbol suggest that at least one estimate is far from the (unobserved) true value. See table B.1 in appendix B for country abbreviations. Sources: Dollar (1992); Easterly (2005); Bhalla (2007a).

Countries marked by squares are those for which the two estimates are within the “tolerance” level of ±30 percentage points. The countries marked by circles are the countries for which there is a wide divergence. Particularly worrisome are countries in quadrants II and IV, for which the sign of valuation of Bhalla (2007a) is opposite to the sign of the comparator (Dollar-Easterly or Johnson, Ostry, and Subramanian). Evidence that the Dollar-Easterly estimates of currency valuation may be deeply flawed is provided by the observations that lie significantly outside the ±30 percent range. For example, in figure 4.3, for the period 1976–85, India was one of the most protected economies in the world; Bhalla (2007a) estimates that the Indian rupee was overvalued by 206 percent; Dollar-Easterly estimates are that it was undervalued by 6 percent. Analogously, Sri Lanka has the most undervalued currency (49 percent) according to Dollar-Easterly estimates, but Bhalla (2007a) estimates consider it to be overvalued by 52 percent. Similar contrasts can be found in figure 4.4, between the Bhalla (2007a) and

62

DEVALUING TO PROSPERITY

Figure 4.4

How accurate are the 2011 currency valuation estimates?

Johnson, Ostry, and Subramanian (percent)

125

IV

yem

I

cog zar

zmb brn

che kwt nga eri norirn dnk afg lby lbr ago aus ven srb mng uaegab swe nzl jam tgo nld mdg fra bel bra esp ita jpn tza grc aut gnq fin bhr deu irllux civ can caf saulbngrd prt gbr isl hnd tcd sdn sur vut ben isr cyp mlthrv rom mrtmli nam gtm hun gnb mex lca jor bol djilsosen bfa aze cri per omn rus qatdzaecu usa tjk ken ner svn chl com zafcol syr bhs hti khm gha cze mwi lva fji slv ltu est sgp kir cmr pol ury kna svk bwa turmkd gin bgr pan bdi syc slb alb idn png pry tto dom atg cpv lao zwe rwa mar wsmpak kor dma uga mdv egy kaz phl mys brb mda ton npl gmb arg blz guy oan hkg bih swz vct lkatmp tha nic btn mmr ukr arm chn vnm sle eth blrtun ind moz bgd geo stp

75

25

–25

–75 III –125 –150

mus

–100

II

uzbkgz

–50

0

50

100

150

200

250

300

350

Bhalla (percent) Notes: See figure 4.3 for details and table B.1 in appendix B for country abbreviations. Sources: Johnson, Ostry, and Subramanian (2007); Bhalla (2007a).

the Johnson, Ostry, and Subramanian (2007) estimates for 2011. The constant elasticity of the log-log functions of Johnson, Ostry, and Subramanian leads to some anomalies; in particular India, a much poorer country with a current account deficit, has a larger undervaluation level (51 percent) than China, which is three times richer (47 percent) and has a large current account surplus. Bhalla (2007a) estimates consider India’s and China’s currencies to be undervalued by 26 and 44 percent, respectively. Another anomaly is the Philippine peso, which the Johnson, Ostry, and Subramanian estimates consider undervalued by 30 percent, but the Bhalla (2007a) estimates consider overvalued by 8 percent. Figure 4.5 plots the Bhalla (2007a) estimates of currency valuation for selected countries for 2011. The most misaligned currencies are those of China, Hong Kong, and Taiwan, each of which is undervalued by 40 percent or more. Countries with currencies that are overvalued by 40 percent or more include Australia, Brazil, and Japan.

MEASURING CURRENCY VALUATION 63

Figure 4.5

China Taiwan Hong Kong Argentina Thailand Korea Malaysia India Singapore United States Turkey Chile Russia Philippines South Africa Indonesia Israel United Kingdom Mexico Canada Japan Australia Brazil

–60

Real exchange rate currency valuations for selected countries, 2011 –44 –41 –39 –39 –35 –31 –30 –26 –11 –7 –4 1 4 8 9 10 16 24 26 30 40 63 72 –40

–20

0

20

40

60

80 percent

Notes: Currency valuation is defined as the (log) difference between the predicted and actual real exchange rate (RER). The RER is defined as the ratio of the purchasing power parity (PPP) exchange rate obtained from the Penn World Tables and the nominal dollar exchange rate. The predicted RER is obtained from the equation RER = 1.11*(1 – 0.97^y), where the RER is the measured real exchange rate and y is per capita daily income at 1996 PPP prices. Sources: Bhalla (2007a) extended dataset. See text and appendix A for details.

Income, Currency Valuation, and Growth: A Review A number of papers use the Dollar-Easterly data and reach the same seemingly robust conclusion that currency valuation was theoretically important but empirically insignificant in explaining growth in developing countries.12 Even when the currency valuation coefficient was significant, the authors note several associated problems. Easterly (2005) finds that the “worth” or significance of currency valuation depended on the inclusion of outliers. Acemoglu et al. (2003) find that the importance of currency misalignment is a function of other variables that were introduced into the model and generally find that after controlling for the role of institutions, the effect of currency valuation

12. Prominent among these papers are Easterly (2001, 2005), Easterly and Levine (2002), Acemoglu et al. (2003), and IMF (2005).

64

DEVALUING TO PROSPERITY

disappears. IMF (2005) finds an effect of currency valuation on growth but that its magnitude is small and its significance fragile. Other studies using different methods find strong statistical significance for both currency misalignment and the instability (variance) in such misalignment.13 Ofair Razin and Susan M. Collins (1997) estimate misalignment as the deviation between the actual and predicted real exchange rates, with the latter obtained not from an equation relating RER only to income per capita but from an equation relating RER to its several assumed determinants such as income and terms of trade. They find misalignment to be an important determinant of growth but one that is dependent for significance on high values of misalignment. Several recent studies that use data other than the DollarEasterly data also arrive at the same result: Currency undervaluation does have a positive, albeit weak, effect on growth.14 In short, the conclusions based on new estimates of currency misalignments are not much different from those based on older estimates: Currency valuation is only sometimes an important determinant of economic growth. However, as Michael Woodford (2009) notes, the weak relationship is very likely biased upward—that is, if correction is made for the incorrect specification, there may not be any relationship between currency valuation and per capita growth. Woodford rightly states that regressions that have the mean value of currency valuation as an independent variable in determining growth have an upward construction bias.15 Income per capita is both on the independent, right side of the equation and on the dependent, left side. Accommodating Woodford’s criticism, we are back where we started: Currency undervaluation may be an important determinant of growth, but most of the existing currency valuation estimates are biased upward, and therefore one cannot assess with any degree of confidence whether currency undervaluation (or overvaluation) has any significant effect on growth. Parallel studies have also estimated the impact on growth of economic openness and economic freedom. These three determinants—openness, freedom, and currency misalignment—are related either directly or indirectly through their empirical proxies. For example, economic freedom is represented by Gerald Scully (1988) and World Bank (1991) as the black-market premium on currency (the divergence between a market-determined and actual exchange rate), which can easily be interpreted to represent currency misalignment, particularly currency overvaluation. Notably, most attempts to relate currency valuation to growth have used

13. Cottani, Cavallo, and Khan (1990) analyze 1960–83 and Dollar (1992) examines 1976–85. 14. Prasad, Rajan, and Subramanian (2007) and Rodrik (2008). 15. Bhalla (2007a) escapes the Woodford criticism because the variable representing currency valuation is initial valuation; the growth analyzed is subsequent to this initial value. The average annual change in valuation as an independent regressor, and this variable does suffer from construction bias. In this book, the problem with change in valuation is appropriately addressed.

MEASURING CURRENCY VALUATION 65

variables involving overvaluation—black-market premium, lack of openness, and currency overvaluation itself. That is because most focus on the determinants of a lack of growth. Given the strong growth in many developing economies over the last few decades, especially in emerging-market economies, this bias has lessened. As discussed, and as documented in the remainder of this book, the story of the past 35 years may just be that currency undervaluation leads to higher growth and not just that currency overvaluation leads to lower growth. The measurement problems outlined in this chapter underscore the need to rigorously test the links between currency valuation and growth to properly assess the proposition that countries can undervalue their way to prosperity. At its essence, the proposition is that currency undervaluation affects growth via investment. Chapter 5 provides the theoretical and empirical foundation for an examination of the link between currency valuation and investment.

66

DEVALUING TO PROSPERITY

5 The Yin and Yang of Investment

Simplicity is the ultimate sophistication. —William Gaddis, Anselm in The Recognitions (1955) The common element among all the high-growth Asian economies of the globalization era—China, India, and Vietnam—is increasing and high rates of investment, or capital formation. This is entirely predictable. But what distinguishes these economies from high-growth economies from previous eras, such as those in the former Soviet Union in the early phases of the Cold War, is the efficiency of investment. One important reason investment has been more efficient in these economies is that it was facilitated by real interest rates comparable to the cost of capital in the developed economies. Another important reason is currency undervaluation. Heuristically, it is easy to see the connection. A cheaper currency means lower (international) costs of production. Lower costs mean higher profitability. Higher profitability means higher investment. And higher investment means higher (more efficient) growth. This chapter subjects these links to some tests that make clear the connections between currency valuation, investment, and growth.

Currency Valuation and Investment Growth is a function of capital, labor, and productivity. An accounting identity relates higher investment to higher growth, but what is the role of currency misalignment? It is not often mentioned among the many factors underlying investment. Bradford DeLong and Lawrence Summers (1991) highlight the important contribution made to growth by plant and equipment investment, and a large volume of literature emphasizes the importance of foreign direct investment. The argument is that such investments have a higher rate of return than the average rate of capital formation and thereby affect growth directly. This leads to a prior question: What determines investment? The obvious answer is profitability, both absolute (the rate of return is higher than the cost 67

of capital) and relative (the rate of return is higher than the return on the nextbest alternative investment). Therefore, if currency valuation plays a role in determining investment, it must be via profitability. Currency valuation directly affects the cost of labor, an important input to the production process. Today, the cost of capital is nearly the same for most foreign investors, although as Gregory Clark (2007) documents, this was also the case during the 19th century. So what makes one country more appealing than another to foreign investors? Several features can make investing in one economy more attractive (relative to others and in absolute terms), such as a lack of bureaucracy and red tape, high-quality workers, a favorable tax regime, among others. However, a dominating influence is likely to be the relative cost and profitability of the investment. If the currency is overvalued, the costs will be relatively higher, and profitability and foreign investment will be relatively lower, other things equal. In this way, currency undervaluation can play a direct role in decisions about which economies do and do not get foreign investment. The same logic applies to domestic investment. Consider the experiences of China and India. China has received far more foreign direct investment than India between 1980 and 2011. For China, foreign investment has constituted 2.5 percent of GDP, and the ratio of investment to GDP has averaged 40 percent during this period; for India, the comparable numbers are 0.4 and 26 percent, respectively. Why the difference? The explanations cite many factors, including red tape, authoritarianism, a bumbling democracy, and barriers to entry for foreign investors. All of these are important factors of the differential in investment, and some are even crucial. But there are other determinants. When the exchange rate is more than competitively priced, as appears to have been the case for China during most if not all of the previous decade,1 there is a large cost/productivity advantage that pulls in extra investment and generates extra growth. The level of undervaluation implies a certain equilibrium level of investment and therefore growth. The attractiveness of the undervaluation to investors is balanced against other considerations, such as real interest rates, bureaucratic costs or delays, the overall investment environment, and the extent of corruption. If investment in the economy is still profitable, then investment flows in, aided, of course, by the cost advantage afforded by currency undervaluation. The net result is that investment and economic growth are both negatively related to currency valuation. (Recall that a negative currency valuation indicates undervaluation.) If a currency is undervalued, the costs of investment are lowered and investment (capital formation) is higher. This higher rate of investment increases the rate of growth. The virtuous circle linking currency undervaluation and higher growth is complete. The link now must be verified, theoretically and empirically.

1. Chapter 14 explains that, although China is about three times richer than India, Chinese wages, in dollar terms, have until very recently been substantially lower than Indian wages.

68

DEVALUING TO PROSPERITY

Investment Impact of Currency Devaluation Two assumptions underlie the model linking currency valuation, investment, and growth. First, the average cost of labor in any country is proportional to the average income per capita in that country (with both measured in current dollars). The averages disguise a lot of factors, including income inequality and the use of unskilled versus skilled labor, among others, but the average cost of labor can be expected to be representative of the underlying real cost of labor. Profits depend on the productivity of labor relative to its cost. One aggregate measure of labor productivity is average income per capita, measured in current purchasing power parity (PPP) dollars—dollars defined in terms of a common currency and unaffected by short-term fluctuations in the exchange rate. This is the second assumption, that productivity is measured in current PPP dollars. To summarize, costs are measured in current dollars, and productivity is measured in current PPP dollars. The investment rate (investment as a percent of GDP) is inversely proportional to the ratio of costs and productivity: Investment D(Cost of labor/Productivity of labor), with cost measured in current US dollars.

(5.1)

Substituting for costs and productivity (measured in current PPP dollars), Investment = –k (GDP per capita in US dollars)/(GDP per capita in PPP dollars), where k is the proportionality constant, which is negative—higher labor costs lead to less profit and less investment; higher labor productivity leads to more profit and more investment. Translating into local currency, Investment = –k (GDP per capita in local currency/XRUS)/(GDP per capita in local currency/XRppp), where XRUS and XRppp are the exchange rates with respect to the dollar and PPP dollar, respectively. With GDP per capita in local currency cancelling out (it is both in the numerator and denominator), the equation reduces to Investment = –k (XRppp /XRUS).

(5.2)

The real exchange rate (RER) is equal to the ratio of price levels, and, by definition, the ratio of price levels in any economy with respect to the United States is the ratio of PPP exchanges rates to US dollar exchange rates. The former is what a dollar really buys in any country other than the United States; the latter is what a dollar really buys in the United States as expressed in local currency via the exchange rate. (This follows from the assumption that $1 THE YIN AND YANG OF INVESTMENT 69

equals PPP$1.) Therefore, the RER is equal to the ratio of price levels, and this is equal to the ratio of the PPP exchange rate and the dollar exchange rate (XRppp/XRUS). Hence, investment is inversely proportional to RER, or Investment = –k × RER.

(5.3)

This is the base investment equation, which would hold true in the absence of Balassa-Samuelson effects. But we expect that income or labor productivity increases with the equilibrium level of RER. The modification required of equation 5.3 is simple: Instead of investment being a function of RER, it is a function of RER adjusted for Balassa-Samuelson effects, or RER*. The key insight here is that RER* is not constant but increases with income per capita. Therefore, a more accurate reflection of reality is that investment is not a function of the RER, nor a function of the equilibrium RER, but rather a function of the deviation between the two. When RER is lower than RER*, other things equal, profitability is higher, and therefore investment is higher. This deviation is described in chapter 4 as the currency valuation, or CV (see equation 4.2): Investment = –k × CV.

(5.4)

The Balassa-Samuelson time effects are a reflection of differences in productivity growth. Thus, if costs increase (due to an increase in the RER or in wages) and productivity increases in the same proportion, then the profitability of an enterprise is unaffected. However, an equilibrium valuation means that all investment adjustments have taken place and there is a “neutral” level of profits. The initial currency valuation sets the potential for extra profitability at a particular level, and investment reaches the equilibrium level in a short period of time. Only further depreciation in the RER can generate extra investments, profits, and growth. In a symmetric fashion, appreciation in the RER would mean higher labor costs, lower profitability, lower investments, and lower growth. Hence, the initial level of currency valuation matters (it affects the initial equilibrium investment rate), and the change in valuation matters (it affects additional investments). If lags are now introduced into this process, the investment equation becomes: Investmentt = f(CVt, CVt–1, CVt–2,…, CVt–n).

(5.5)

The traditional method for modeling the lagged process contained in the above equation is to replace the valuation series by its mean.2 This method assumes that all currency valuation levels have the same impact on investment, or in a reduced form equation, on economic growth. 2. This method is followed by Easterly (2005); Johnson, Ostry, and Subramanian (2007); and Rodrik (2007a).

70

DEVALUING TO PROSPERITY

It can also be argued that initial currency valuation matters. Assume two countries, the first with a 100 percent initial overvaluation and a zero percent final overvaluation. The second country starts with an overvaluation of zero and moves to 100 percent. Both countries average the same amount of valuation, namely an overvaluation of 50 percent. The traditional method would indicate that both countries should be similarly affected in terms of investments. But it is more likely that the first country will receive increasing amounts of investment as it becomes more competitive, and the second will receive decreasing amounts as it becomes less competitive. So a very different reality emerges once the change in currency valuation is introduced into the model. The econometric reasoning behind the differential effects is as follows. Consider two specifications of the effect of currency valuation on investment. In the first case, the total effect CV* is a linear combination of CV in the present period and CV lagged one period: CV* = [a1 × CVt ] + [a2 × CVt–1].

(5.6)

In the second case, CV* is a function of CV in the present period and the difference in CV from the previous period: CV* = [c1 × CVt ] + [c2 × (CVt – CVt–1)].

(5.7)

In equation 5.7, CV* is a function of the initial level of CV and the change in CV. But this equation is simply a linear transformation of equation 5.6 with a1 equal to (c1 + c2) and a2 equal to –c2. In the traditional formulation, a1 and a2 are forced to be equal; with a lag structure, the two coefficients can differ. The expectation is that both the initial level (designated iCV) and the average change (designated dCV) will have a negative sign in determining the investment rate. A lower initial currency valuation level is expected to increase investment because of higher profitability, other things equal; a depreciating real currency (dCV) is expected to also yield to higher profitability, higher investment, and therefore higher growth. Thus, the relevant equation for determining investment, It , is: It = a + b × iCV0 + c × dCVt + d × Zt + et ,

(5.8)

where iCV0 is the initial level of undervaluation, dCVt is the average change in undervaluation from initial time period t0 to the final time period t, Z is a vector of other determinants (e.g., real interest rates, tax rates, and corruption costs), and e is the error term. This specification allows for lagged effects of real exchange rate changes on investment to be different from initial effects. Equation 5.8 emphasizes the role of both the initial level of valuation and the change in this level. This is one of the major innovations in this study of currency valuation and growth. By relating valuation to investment, empirical estimations of investment do not suffer from any type of the construction bias

THE YIN AND YANG OF INVESTMENT 71

noted by Michael Woodford (as described in chapter 4). If investment is related to currency valuation, then there should be strong empirical support for the conclusion that currency valuation is an important determinant of growth.

Investment as the Channel of Influence Figure 5.1 plots the observed relationship between currency valuation and the share of investment in GDP for eight selected countries for 1960–2011. The values for currency valuation and investment shares are averages for each fiveyear period (with 2010 and 2011 being part of the “five years” starting in 2005). The data for each country are sorted by the magnitude of the currency valuation; the corresponding level of investment is shown for that five-year period. At least for these eight countries, there is a pronounced negative relationship between the two. The sudden bumps in investment for some countries (Germany, Japan, and the United States) are due to the fact that the data are not sorted in chronological order. For the developing countries, there are virtually no bumps, but instead a fairly sharp and straightforward negative relationship. As emphasized by theory, and regardless of time or space (country), the share of investment appears to increase with an increase in currency undervaluation. A cross-section panel of growth and growth-related data for more than 130 countries is used to analyze the relationship between currency valuation and investment. In addition, a smaller sample is also considered, consisting of 27 countries—the G-193 plus eight additional countries (Chile, Hong Kong, Israel, Malaysia, Philippines, Singapore, Taiwan, and Thailand). Given that there is no construction bias between the share of investment in GDP and currency valuation, table 5.1 reports on results for three separate models: the mean undervaluation during the period, the initial currency valuation for each five-year period, and both initial and change in currency valuation. To a surprisingly robust degree, currency valuation is shown to affect the rate of investment. The coefficient of currency valuation is extremely robust, and its magnitude is centered around −0.05—for each 10 percent sustained level in (log) undervaluation, investment shares rise by 0.5 percent of GDP. For the less noisy G-27 sample, the impact is almost double—each 10 percent of initial undervaluation leads to a 0.9 percentage point increase in the investment share. The change in valuation also matters and leads to a large bang for the buck—an increase in undervaluation of 10 percentage points in any period leads to an increase in investment share of almost 1.2 percentage points. Figures 5.2 and 5.3 show the results of adding variables to the model. There is a pronounced negative relationship between investment shares and initial currency valuation and a significant negative relationship between investment shares and the change in such valuations. This chapter documents a strong empirical relationship between investment and the real price of a currency. The empirical results are consistent 3. That is, G-20 minus the European Union.

72

DEVALUING TO PROSPERITY

Figure 5.1

Investment and currency valuation: A strong relationship, 1960–2011 Germany

China investment (percent of GDP)

investment (percent of GDP)

44

27

39

25

34

23

29

21

24 19 –70

19 –20

30

80

130

180

17 0

5

10

15

20

currency valuation

30 25 20 15

50

35

investment (percent of GDP)

35

0

30

Japan

India investment (percent of GDP)

10 –50

25

currency valuation

100

150

currency valuation

38 36 34 32 30 28 26 24 22 20 –30

–10

10

30

50

currency valuation (continues on next page)

with the theoretical model outlined above. Both domestic and foreign investors respond to a depreciated real exchange rate, and foreign investors are likely to have a larger reaction. Foreign investment often has associated externalities, including technology transfer, governance improvements, and more. Hence, there is reason to believe that currency valuation exerts an independent influence on growth, in addition to the effects on growth occurring via investment. The chapter started by documenting the larger flow of investment to China than to India. It ends by positing that some, perhaps a great deal, of the differential investment rates between the two economies, and therefore the differential in their GDP growth rates, may have been due to their different exchange rate policies. THE YIN AND YANG OF INVESTMENT 73

Figure 5.1

Investment and currency valuation: A strong relationship, 1960–2011 (continued) Korea

United Kingdom

investment (percent of GDP)

investment (percent of GDP)

38

20.5

33

19.5

28

18.5

23

17.5

18

16.5

13 –40

–20

0

20

40

15.5 –30

–20

–10

0

10

20

currency valuation

currency valuation Thailand

United States

investment (percent of GDP)

investment (percent of GDP)

40

20.5

35 19.5

30 25

18.5

20 15 –70

–20

30

80

currency valuation

17.5 –21

–16

–11

–6

–1

4

currency valuation

Notes: Investment is for all countries for the period 1960–2011, subject to data availability; true period for each country, 1960–2011, subject to data availability. For each country data are sorted according to the estimate of currency valuation; positive value signifies overvaluation. Currency valuation is defined as the (log) difference between the predicted and actual real exchange rate (RER). The RER is defined as the ratio of the purchasing power parity (PPP) exchange rate obtained from the Penn World Tables and the nominal dollar exchange rate. The predicted RER is obtained from the equation RER = 1.11*(1 – 0.97^y), where the RER is the measured real exchange rate and y is per capita daily income at 1996 PPP prices. Source: Author’s calculations.

74

DEVALUING TO PROSPERITY

Table 5.1

Effects of currency valuation on investment Currency valuation

Fixed effect

Mean

Initial

Average change

R2

Number of observations

0.4997

993

Selected sample Model 1

–0.059***

Model 2

–0.052***

0.4736

991

Model 3

–0.063***

–0.12***

0.4811

991

–0.09***1

–0.13***

0.6226

248

G-27 countries

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. All models are based on five-year datasets for 1960–2011. The models include the log of initial per capita income and time and country dummies. 2010 and 2011 are included in the data for years beginning with 2005. The sample includes 132 countries. See appendix A for details. The G-27 sample includes the G-20 minus the European Union, plus Chile, Hong Kong, Israel, Malaysia, Philippines, Singapore, Taiwan, and Thailand. Source: Bhalla (2007a) dataset extended to 2011.

Figure 5.2

Investment growth compared with initial currency valuation

investment growth (percent of GDP) bwa

20

lby ner per alb

15 10 5 0 –5 –10 –15

tgo

alb sen tha idn mng gab mys nor npl mrtsgp idn mdg swz ind lao grc zmbaze hnd dza jam ven vnm mysben zmb jpn egy bgr chn mrtmli kor erisgp jpn bdi tkm bfa che swe che nic egy moz syr nic gha mar gmb grc bgd per hkg arm gha mys isr civ mar vnm gab hun sen blr jor mwi hnd bgr bel chn bgd fin caf jam nga idn pan zaf omn lbn nam npl dnk prthun aus nic gmb gnb bgd egy cmr eth lka chl cfra iv gnb dza civ uae tto tgo jpn chn svk can kaz sen nplidn turuzb fin mli zaf mex sau gtm mar chl tto phl nga nor tur pak sgp lbn omn jor ginjor bdi uae cog png bgr eri jam rom tun cog lka cribra arg nor dza bdi mar gab gmb tun bol cog ken chn mwi tto hnd pry civ mdg chn chn arg fin ury phl jam tjk gbr ken zaf ind ind lka pan eth deu aut hkg cmr pry dza tha mrt npl kor pry hnd esp mex pol gbr cmr ita png nzl oan isr ven lby slv irlben bwa bwa dnk khm gin gtm tur mus bra bih pol npl nld lkachl gha cze nor sgp prt gtm bfa rom nam ven mrt mda ghacaf mli fin kor mar lva usa bgd irl grc kwt sgp dom png bra zmb mwi ner chl hkg ben uga ken arg ecu est mwi che esp slvpak usa pry zaf vnm ita pry ecu deu mus svn ind png irl syrcol mar lby swe mli colhun nld svn cog jam per tha swe gbr ury pol ind phl isr zaf bwa gab irl tto col nor oan ltu png arg col gmb tjk hkg svn khm mexdeu slvnld png moz col idnlkasyr isr sen ken dom ven irl ury syr tun tto ner aus sgp slv nld bol tur cze dom lao nzl cmr cri can pan gtm chl civ phl slv lva isr pak deu gbr oan fin usa can bfa moz aut esp uga kwt bgd col mus aut isr tto dnk aut prt sau jam bfa cri bol fra rom bel dom cog kwt tun mus gin ury nzl hrv syrpak kgz lka zmb mli syr ita aus auspak aze fra bra rus ben ken mkd cri mex mrt ken ury tur nic gin ben dom hrv kor nzl kor chl gtm hkg deu lbn ner svk uae bol tgo mys oan bel mkd est grc bfa usa omn ukr prt dnk lbr yem geo chl slv sle ven ita ukr ltu dom ecu nzl slv can tha kor bwa arg jpn pak swz rus esp phl usa dnkjpn nic rus hun esp gbr prt kgz kgz mar ita caf ury nam bel uga cri hun dom swz col uae bgd cri bra mex arg bwa aus per can mkd omn moz sau swz kwt fra kor aus khm hkg aut ita pak bfa pak pan sau phl omn gha cri cmr esp dnk egy caf tun nld ecu bdi bol pry mdg bdi tur tto ven can phl caf che usa hrv svk prt bel grc che bfa gbr ecu usa cridza cribra bra mrt tgo belitaaut gbr mda deu egy gtm sau gbr tun mex can hun kwt frafra sau nam ken egy nam sle bra fin tgo dom prt aut bra civ zaf fra arg dom bol zmb jpn nic per mda ltu usa gha png bol uzb nzl nic per col swe pol mex nzl moz ita bel ind bdi ner ken slv swe can mex mwi ury ner swe aus arg cog nld kor tgo nzl mng pan sen tha aus phl sau mys esp jam swe tun swe ner gmb gtm fra sen mrt idn thaeth can pol che jpn cmr ven jor mwi phl nic pry dnk syr hnd col ita bdi gtm bol est grc bwa ben ben tgo ven kwt jor tun nor bgd syr jpn mex arg ecu hnd mli gtm nld rom cog bgd sen oan cmr bgr pak gbr arm caf dnk dom gnb sle deu hkg swz ben isr uga ury hkg cri bih mli dza autnldche grc aus mex hkg hnd tha usa nor uzb slv tha bel alb sle blr swz png irl tha eth zmb nic tgo gha zaf egy bdi kor isr egymng grc mdg oan ben irl ury gmb dza mdg cmr grc zaf blr alb bwa pry lva gab civ gin lbn esp png chn gbrsyrind gtm caf tun lka per ken npl khm civ uae mys sgp fin zmb pry eth slv dnk cze irl omn che fin mys gnb lby ner pol gmb svk isr tha cmr cog lby chl pry ttotur ben arm deu zaf gmb gmb idn dza ind sle fin zmb tur tur ven bra isr syr gab jor hnd tgo dza per fin egy pak swz gin bfa chl rom mli col ner kaz ind omn tto mys caf jor mwi hun hnd uae nerhkg tjknga ngagha phl ken caf arg sen jam svn mysbfa chn jpn nam nor zaf pernpl civ lka mwi mrt dzasen hnd mrt bwa mus idn hun cog nor che ind prt ven png bgrmys lbn per mng chl egy tgo norlkagrc vnm mli civ mdg bgd bdi tkm tto moz mar gab senkor sgp gab pan jpn gha chn zmbsgp mar sgp npl jam mar bgd npl bgr jam alb aze mrt vnm idn lao nic eri idn chn alb bwa lby

–20 –100

sle caf

–50

0

50

100

initial currency valuation coefficient = –.06271402, (robust) standard error = .00762864, t = –8.22 Notes: The model relates investment growth to initial currency valuation. An added variable or partial regression plot shows the effect of adding a variable to an existing model; it relates the residuals of the dependent variable Y on other determinants X to the residuals relating the new variable (Xadd) to X. For country abbreviations, see table B.1 in appendix B. Source: Author’s calculations.

THE YIN AND YANG OF INVESTMENT 75

Figure 5.3

Investment growth compared with average change in currency valuation

investment growth (percent of GDP) bwa

20

tgo

lby

15

gab

alb

alb ner per

tha senswz idn mng bgr idnjam mys benaze zmb dza sgpmdg mrt nor ven sle grc npl mli mrt kor mys lao nic bgr egy hnd vnm tkm ind zmbjorsyr eri nicper blr sgp arm gha mar nganic caf egy idn mwi mys bdi grc vnm isr civ hkg bgd jam nam bfa pak egygnb tto cmr jor gmb mar gnb pan jam swe ken rom uaesen kaz bdi zaf omn sen hun hkg zaf tgo chl hnd gtm gmb gin chn che lbn mli dza bel civ chn pry hun jpn jpn bgd tursvk png gha tun gmb tjkjam cog uae bdi eth lbn cog hnd kor ken eri bgd gab esp dza omn mex che caf sau aus bol mar prt phl civ tto chn idn fra mrt lka npl tur ven gin gab dza can tun oan fin cmr zaf mwi chl khm bra ner sgp nor cri mdg fin bwa cmr isr bfa pan esp ven bgr arg mrt col slv tto mda jor slv npl bwa mwi mus ken irl hnd nam rom pol nld pry pak nga tha png nor phl kor uzb jpn pry cog lka lva mex mwi pry kwt npl chn dnk aut civ civ sgp zmb mex hkg col bra mex oan mar npl bih nic sgp deu svn chn tur hun png png gtm ecu cri pol cze ury fin pol fin gha arg gbr vnm ben arg bgd bra prt nzl sgp tjk lby usa chl caf cog png jpn tun ind pry deu est bfa ken mus dom usa irl tha isr aus dnk deu gab isr gha aut ury dom zaf nld slv mli che col phl dom phl kor bwa lka lva tto usa ben lka lby irl ven rom eth ita syr bel chl sle usa isr png gin grc mus nzl chl svn ecu dnk aut can khm swe nzl omn hrv caf gbr gbr bgd tto tha phl mar nld moz uga mar irl gtm ury can turpakind zaf svn nor gmb per ind jpn ken gtm dnk aus i nd dom col cri sen khm usa pan che irl swe ita mkd kwt gbr ecu ita chl ben kwt espesp can gbr pak rus hun tun chn arg cri prt tto hkg ury arg col bgd chl bol deu kgz bfa fin slv uga ukr brakgz jam cze nld ury lbn moz bra bol uga gtm mys ltu mngdeu mkd sau che mar ita ner per cri col tur bel ben caf swz mli cmr isr caf fra prt bel grc slv syr nzl bel rus bfa egy phl gbr lka mli tto bwa geo lbr yem ury fra nam sau esp pol usa ury prt fra gin ltu egy cog svk aus jpn dom kwt ven oan aut mus gbr ury dnk nld dnk aus kor aus sau nzl sle ita dom swe hrv cri syr lka ltu zmb phl bwa ner mrt hkg slv col bolbol bra bel dom ita aut swz ner bol ven bdi ukr gtm arg ita kor fra uzb can bdi oan hkg omn zaf jam dom lao tun ita gha mex prt kgz pak grc est cmr fra nzl syr can aze nam sle arg tgo slv bra ken mrt eth gmb fra hrv rus slv per arg syr can moz dnk pan gtm ken pry ven swe nld sau che uae zmb uae jam lka ecu syr usa caf idn bra pak tur swz ind nzl png swe cri nam mkd cmr swe aut dnk est slv bfa irl gha sen aus nic civ pol aut syr ind ecu mdg sau dza syr gbr mda hkg ita mex bra sen col bol can tgo dom pry cog fra sau kwt usa hnd cri cri ben esp mwi prt mrt bdi pan hun oan tun fin alb mda omn deu bol gtm grc isr hnd chn col kor ecu bwa tgo caf swe ken che ind eth svk swe uga pak hun tha lbn aus gbr nor cog cmr mex deu fin gtm swz mex ury dza phl mli pak per zaf moz hnd pak grc fi n ner uzb phl usa ven pak nic egy nic chn idn aut sen gbr bih alb npl nld bgd mex bfa tgo nzl mdg gmb tun jpn svk nld ben mex nicnor dza aus jpn kwt ner kor tha cze usaarg mys tun tha irltgo fin eth tha can mys cri zmb ben arm fin mli egy syr mys omn zaf blr gha bgd tun ttobdi gnb bel uae jpn dom esp chl grc hnd tur tha deu rom jor hun grc gha per gmb bwa arg coglka tgo bel swz turben ner zmb bdi mdg lby mwi pry oan ury hkg kor jor isr civ png mli zmb tur gab png hkg che gnb pry gin tgo cmr sgp esp egy nor ind blr lby tha gtm nga bfa dnk ben jor bfaciv irl ner arm idn mys bgr prt mys isr uae nic lva bgd hkg indisr lka swz gab ner gmb gmb mwi gnb sle idn sen che nor svn hnd sle sen caf ken civ omn cmr jor dza dza pry mng grc hun bra slvkhm colzafzaf chl cog egy dza per nga npl mrt civkaz gin mdg mrt nor caf chl npl polhndchn mwi mng phl tgo ven mli egy tto tjk png per mar mus rom gha jpn npl nam jam mar ken gab gabtto lbn vnm sgp venbdi lao pansgp bwa mar tkm korjam sen zmb mys sgp bgd chn idn moz vnm aze alb idn jam mrt bgr eri nic lby bwa alb moz

10 5 0 –5 bgr

–10 –15

per

–20 –25

–15

–5

5

15

25

average change in currency valuation coefficient = –.12048782, (robust) standard error = .03814495, t = –3.16 Notes: The model relates investment growth to the average change in currency valuation. An added variable or partial regression plot shows the effect of adding a variable to an existing model; it relates the residuals of the dependent variable Y on other determinants X to the residuals relating the new variable (Xadd) to X. For country abbreviations, see table B.1 in appendix B. Source: Author’s calculations.

76

DEVALUING TO PROSPERITY

6 Is the Real Exchange Rate Endogenous?

Can analysis be worthwhile? Is the theater really dead? —Paul Simon, The Dangling Conversation There is no doubt that investment affects growth, and chapter 5 demonstrates how investment is affected by currency valuation. So a strong case can be made for at least an indirect link between currency valuation and growth. This chapter addresses objections to the existence of links between currency misalignments and economic variables of concern that spring from a simple belief that it is not really possible for currency misalignments to exist for any reasonable length of time. The objection is legitimate. Changes in the relative valuation and/or competitiveness of different economies comes about through changes in the nominal exchange rate. But how can changing a nominal variable, the exchange rate, affect a real event such as economic growth? There is a related but narrower question: How can changing the nominal exchange rate affect the real exchange rate (RER)? Let us say a country wants to be more competitive and so it devalues its currency by changing the nominal exchange rate. If the country is at full employment, then there will be excess demand (whether it is domestic or foreign is immaterial), and this excess demand will lead to higher inflation, which will lead to higher wages. This will erase the initial competitive advantage. Therefore, the change in the nominal variable (the exchange rate) does not affect real events (competitiveness). How often economies are at full employment is not clear. The world changed after the 2008 crisis. In any case, precious few developing economies have ever been at full employment. Therefore, the theoretical foundation of the argument that you cannot affect the real exchange rate is based on an unrealistic empirical assumption (full employment). In addition, it is often the case that, whether there is full employment or not, wages may not increase by the full amount of change in the nominal exchange rate. If they do not, the result is a real currency devaluation. And there is a third possibility, namely, that there is a fixed exchange rate (as in China from 1994 to 2005), which precludes 77

any change in the nominal exchange rate. Does that imply that a change in the real exchange rate is impossible? No. If inflation is higher in the economy with the fixed rate, then the real exchange rate will appreciate; if inflation is lower (for example, Hong Kong), then it will depreciate. In addition, if productivity growth is higher, other things equal, then the real exchange rate will depreciate. Say that the initial level of the fixed rate is undervalued, then it will move farther away from the equilibrium rate toward greater and greater undervaluation with each succeeding period of higher productivity growth (for example, China since 1994).

An Endogenous Real Exchange Rate: The Theory The theory that the RER is endogenous—that is, determined by the system itself—says that any attempt to depreciate the currency will be thwarted by economic factors set in motion by the very policy of depreciation. At the time it is initiated, a devaluation immediately causes domestic goods and services to become cheaper and imports to become more expensive. This leads to excess demand, which leads to more inflation and a rise in the domestic price level that is approximately the same as the initial depreciation. The initial price advantage that accrued to domestic producers is eliminated. According to theory, devaluation is not a road to prosperity but a circular path back to the starting point. The net effect of devaluation on the RER is zero. Ronald McKinnon and Gunther Schnabl (2006, 6) summarize the endogenous RER argument rather forcefully: In a world where many countries peg their nominal dollar exchange rates, changes in these nominal pegs (as in the case of China) could be considered a legitimate right-hand side or “exogenous” variable. But then relative monetary policies…must be altered to sustain any such nominal changes—either easy money and inflation in the U.S. associated with the nominal depreciation, or tight money and deflation in the foreign country whose currency appreciates in nominal terms. With the passage of time, the macroeconomic upshot could then be little or no change in real exchange rates.

While intuitively appealing, the endogenous RER argument is dependent on several assumptions and may be deeply flawed, at least empirically. It assumes that all economies are at full employment at all times, obviously a strong assumption. But the real problem with the endogeneity argument is that it just fails to hold true empirically. This is documented here using several tests.

The Impossible Trinity Real exchange rate orthodoxy is a restatement of the Impossible Trinity, which concludes that it is impossible to simultaneously target the exchange rate (leg 1) in the presence of free capital flows (leg 2) and in pursuit of an independent monetary policy (leg 3). 78

DEVALUING TO PROSPERITY

The most common problem facing a country trying to maintain a depreciated RER is the need to deal with large capital inflows. Countries have tried various policies to counter the appreciation effect of such flows. Many have used direct capital controls, including Chile in the 1980s, Malaysia in 1998, and Thailand in 2007. China and India long had ultra-strict controls on capital flows and today have only strict controls. But the Impossible Trinity contends that these controls cannot fully close the tap, and so higher inflation will inevitably follow an increase in capital flows, bringing the onset of Dutch disease.1 Raising domestic interest rates (leg 3) is unlikely to be fully effective. Hence, the RER will be relatively unaffected by the various policy decisions. According to the Impossible Trinity, countries can choose either a marketdetermined appreciation of the nominal exchange rate and low inflation, or a controlled exchange rate and higher inflation. In the former case, the RER appreciates and there is low inflation. In the latter, there is higher inflation and this forces the RER higher too. Hence, it is pointless for policymakers to attempt to maintain an undervalued exchange rate. If the exchange rate were initially overvalued, the result would be the reverse: There would be slower growth and slower inflation (or deflation) and the RER would depreciate. The arguments for and against the Impossible Trinity and endogenous RER theories can be effectively settled only by looking at actual experiences of exchange rate policy change and relative inflation/deflation. To do this, economists have conducted time-path tests and a variety of formal tests of the endogenous RER hypothesis. There is the “unit root” test of whether the RER is mean-reverting and therefore different from a “random walk.” There is the “half-life” calculation, which estimates, if there is a unit root, the number of years before 50 percent of the shock (change) in the RER is “recovered.” Simply put, however, what all these tests seek to do is determine whether inflation follows an exchange rate change over an extended period of time.

Test 1: Does Inflation Follow Devaluations Step by Step? Underlying the assertion that a nominal exchange rate change does not affect a real variable is the assumption that the RER is endogenous—determined within the system. No one argues that the RER stays fixed; what is in dispute is whether policy can affect the RER. The endogenous RER argument holds that higher (relative) inflation follows devaluations, and symmetrically lower (relative) inflation follows appreciations (the reference economy is the United States). The theory can easily be tested, and several authors have done so. Ilan Goldfajn and Rodrigo Valdés (1999, 229) concluded that “in most cases large 1. Dutch disease occurs when increased capital inflows (or revenues from natural resources) cause the currency to appreciate and thereby make exports more expensive in foreign markets and decrease competitiveness.

IS THE REAL EXCHANGE RATE ENDOGENOUS? 79

and medium appreciations are reversed with nominal devaluations” rather than through higher domestic inflation. In his earlier study of the RER in developing economies, Sebastian Edwards (1989, table 7.1) examined 39 devaluations between 1962 and 1982 and found that only six countries did not achieve a real devaluation in response to a nominal devaluation; most of the time (33 instances), inflation after the event was less than the initial devaluation. The simple average of the real devaluation achieved is 75 percent, which is a considerable deviation from the theory. There have been a number of time-series tests pertaining to the endogeneity of the RER.2 Four stylized facts emerge from these studies. First, in hyperinflation economies, there is almost complete pass-through—that is, inflation follows devaluation rather quickly. In such instances, however, the cause and effect are not easily separated. Second, in developed economies, the RER does appear to have a unit root—that is, there is a systematic tendency for the RER to revert back to the prechange level. Third, 50 percent of the reversion to the prior mean for these developed economies (the half-life) takes place over an average of four years. Fourth, for nonhyperinflation developing economies, the RER pattern generally does not have a unit root—that is, nominal changes assumed to be temporary changes in exchange rates become permanent. In other words, the RER can be affected by policy. There are sound theoretical explanations for these empirical results. There are several reasons why inflation will not follow a devaluation, especially in developing economies. Invariably, such economies have a large amount of labor slack. First there is an unlimited supply of labor; then there is underemployment; then there is catch-up. Add to those the impact of globalization, and the endogenous RER argument loses any residual effect. Catch-up means that a skilled wage level in developing economies is often only one-fifth to onehalf the corresponding level in developed economies. As a result, workers may not demand a commensurate wage increase when there is a devaluation. This is most likely what happened in East Asia following the 1997–98 crisis. Even though devaluations were on the order of 40 percent, inflation was barely in double digits during the first year after the crisis, and two years later inflation rates were below the precrisis levels.3

Test 2: Do Real Exchange Rates Revert to the Mean? If inflation follows depreciation and deflation follows appreciation, then the real exchange rate is mean reverting, showing no distinct trend, rather than a random walk, which would show some trend over time. Broad patterns of the behavior of real exchange rates shed light on the likely accuracy of the argument 2. Edwards (1989), Frankel and Rose (1995), Rogoff (1996), and Taylor and Taylor (2004). 3. This is not true for Indonesia, but even in this extreme outlier case (devaluation of log 123.7 percent in 1997), inflation rates hit a high of only 56.2 percent in subsequent years. In 2006, the inflation rate in Indonesia was at 12.5 percent, almost at the precrisis inflation level of 9 percent.

80

DEVALUING TO PROSPERITY

that the RER is endogenous. Figure 6.1 charts the RER and the Bhalla (2007a) estimate of the predicted RER for seven regions and two countries (China and India) between 1960 and 2011. The predicted RER rises over time for most regions, and the steepness of the path is determined both by the initial condition (how rich the region was in 1960) and by its rate of growth per capita. The distance between the solid (actual) and dotted (predicted) lines is the undervaluation or overvaluation of the currency. If actual is higher (as for sub-Saharan Africa), the currency is overvalued; if lower, the currency is undervalued. As figure 6.1 shows, the RER for developed economies has stayed close to equilibrium; Eastern Europe shows a reasonable increase in the RER. South Asia, excluding India, remains mildly overvalued, as does the Middle East and North Africa region. Latin America shows large swings but few trends. The gap between predicted and actual (currency undervaluation) has, in recent years, been the most extensive for China. Direct tests of mean reversion for a large number of countries fail to support the Impossible Trinity. If that proposition were true, over a sufficiently long period, say 20 years or more, the change in the RER should be approximately zero. That is, domestic inflation should be approximately equal to the sum of nominal currency depreciation plus US inflation.4 This is manifestly not the case. None of the formal models that test for changes in the RER correct for the Balassa-Samuelson effect; they are strictly oriented toward the conventional measure of RER change, which equates excess change in relative price levels to the magnitude of the relative change in currency values. If there were a BalassaSamuelson effect, then the RER would increase with growth, especially with fast-paced growth. If a nominal devaluation in a fast-growing economy were to become a real devaluation, then this would be strong proof that the endogenous RER argument is invalid and that the Impossible Trinity exists only in theory.

Test 3: Predictions and Some Spectacular Errors There have been several spectacular devaluations over the past decade or so. Another test of the endogenous RER theory is to examine the behavior of inflation in economies that have experienced such devaluations. The expectation is that inflation would follow the path cleared by the devaluation. Example 1: United Kingdom. In September 1993, the pound fetched 3 Deutsche mark; within days after “Black Wednesday,” it fetched 2.2, a devaluation of more than 30 percent. The half-life of RER convergence is estimated to 4. Inflation in all countries is measured by the GDP deflator. This is only a matter of convenience, because the GDP deflator is available for a larger number of countries and is likely to be a more reliable indicator of inflation for many developing economies than the conventional consumer price index (CPI) measure of inflation. In any case, none of the results are affected by the use of the GDP deflator rather than the CPI.

IS THE REAL EXCHANGE RATE ENDOGENOUS? 81

Figure 6.1

Evolution of actual and predicted real exchange rates (RER), 1960–2011

82

Latin America

Developed economies

China real exchange rate

real exchange rate

0.7

0.6

1.2

Actual

0.6

Actual

0.5

0.5

1.1

Actual

1.0

0.4

0.9

0.3

0.8 0.2 Predicted

0.1

0.3

0.7 Predicted 0.6

East Asia

India

0.45

0.5

196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

0

real exchange rate

Predicted

196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

0.4

South Asia

real exchange rate

real exchange rate

0.7

0.45

0.6 Actual

0.35

0.5

Actual

Actual

0.35

0.4 0.25

0.25

0.2

0.05

Predicted

0.15 Predicted

0.1 0

0.05 196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

Predicted

196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

0.15

0.3

196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

DEVALUING TO PROSPERITY

real exchange rate

Russia and Eastern Europe

0.7 0.6

Sub-Saharan Africa real exchange rate

real exchange rate

0.65

Middle East and North Africa real exchange rate

Actual

Predicted

0.8 Actual

0.7

0.55

0.6

0.5

0.45 0.5

0.2 196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

IS THE REAL EXCHANGE RATE ENDOGENOUS? 83

0.3

0.35

Predicted

0.4

0.25

0.3

0.15

0.2

Predicted

196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

Actual

196 0 196 3 196 6 196 9 197 2 197 5 197 8 198 1 198 4 198 7 199 0 199 3 199 6 199 9 200 2 200 5 200 8 201 1

0.4

Notes: The actual RER is a five-year moving average; the predicted RER is a weighted equilibrium exchange rate for each region (China, India, and the United States are not part of the regional averages). Source: Bhalla (2007a) dataset extended to 2011.

be about four years—that is, 15 percent of the 30 percent devaluation should be made up by excess inflation in four years. But the excess of British inflation over German inflation for the four years from 1993 until 1996 was only 2 percentage points, and it did not reach 30 percent until 2010, 17 years after the devaluation. Example 2: Mexico. During a matter of days in December 1994, the peso depreciated against the dollar by close to 60 percent (from a predevaluation level of 3.4 pesos to $1, to 5.6 pesos to $1). For the five years following the devaluation, there was a real depreciation, but by the sixth year (2000), excess inflation had caught up and the peso was back to its predevaluation level in real terms. The Mexican example seems to perfectly fit the endogenous RER forecast, at least until the fifth year. Since 2000, the Mexican peso has appreciated, and in 2011, it averaged 12.4 pesos to $1. Example 3: Thailand. The examples from the Asian crisis spectacularly refute the endogenous RER argument. The Thai baht was 25 to $1 in June 1997; it averaged 41.13 to $1 in 1998 and 44.5 to $1 in 2001. In 2011, it averaged 30.4 to $1. On average, the exchange rate has appreciated at an annual rate of 2.3 percent since 1998. Thai inflation from 1999 until 2011 averaged 2.5 percent a year, compared with 2.2 percent for the United States—despite a big devaluation, inflation rates about the same as the United States! Example 4: India. In 1991 India undertook major economic reforms, and in 1993 it moved to a managed exchange rate regime. Inflation prior to the reforms and through 1993 was in the range of 8–9 percent. The rupee depreciated between 1994 and 2007, from 30.5 to $1 in 1993 to a peak of 48.6 to $1 in 2002, and by an annual average during these 14 years of (log) 2.2 percent. In the United States inflation was also 2.2 percent a year. The Indian inflation rate should have been about 4.4 percent a year, given the currency depreciation, but the actual rate was 5.2 percent—some 4 percentage points less than the historical average and a paltry 0.9 percent a year higher than the rate predicted by the endogenous RER theory. What these examples and the various tests reveal is that there is relatively little in the literature to substantiate the conclusion that the real exchange rate is endogenous or that the Impossible Trinity theorem holds. The general acceptance of these theories is therefore somewhat of a mystery. The “political economy” behind acceptance of the theory might be revealing. If indeed it is the case that the real exchange rate is not endogenous, then countries can willfully devalue without fear of inflation consequences. And thereby steal an advantage from their competitors. How far this “cheapen yourself to beggar thy neighbor” policy can go, and how important it is to reverse it, is examined in chapter 14.

84

DEVALUING TO PROSPERITY

Passive Devaluation It turns out that a large part of the real currency devaluation that occurs in developing economies is the result not of active but of passive policy. A conventional, official measure of the real exchange rate is published on a regular basis by the International Monetary Fund (IMF). This official rate is often called the “real effective exchange rate,” or REER. The Bank for International Settlements (BIS) provides its own time-series data of REER for several countries.5 These indices represent the purchasing power of each currency relative to a benchmark economy and base year. One benchmark is the US dollar; another is a weighted basket of currencies or economies, that is, with trade shares as weights for the currency rates and inflation rates of partner countries (as done by the BIS). There are two important components in the calculation of the REER: how much more inflation there is in the domestic country than the benchmark economy, and how the currency valuation has changed with respect to the benchmark. An increase in inflation increases the REER; a currency devaluation decreases the REER. The REER can also be thought of as changes in the costs of production, relative to a base year. Given the official status of the IMF and the BIS, many discussions of currency misalignment especially in the media and within investment banks, start and stop with reference to their REER indices. But these official indices only partially capture the relative costs in an economy. If productivity growth were the same in a given economy as in the United States, then the REER would be an appropriate cost measure. But productivity differentials are not the same, and developing economies’ productivity growth is likely to be higher than in the United States. When such “excess” productivity growth occurs, the costs of production decline by an equivalent (percentage) amount. Excess productivity growth is equivalent to a devaluation in the absence of any increase in domestic inflation—that is, a real devaluation. Thus, there are two distinct components to any change in the RER. The direct effect (RERdirect) is the popularly measured deviation of a currency value from its REER—the IMF and/or BIS calculation. By giving zero weight to relative productivity changes, these official REER measures fail to capture what could be the most important component of changing costs and changing competitiveness—the indirect effect on currency valuation via differences in productivity growth. This can be termed the “standing still” component of a real devaluation (RERsstill). These two components of valuation, direct and indirect, are additive in the logs and therefore allow for a decomposition of the real change in currency values as follows (all changes are log changes denoted by the prefix d): dCVt = CVt – CVt–1

(6.1a)

5. The BIS has released REER data for more than 60 countries for 1994–2010.

IS THE REAL EXCHANGE RATE ENDOGENOUS? 85

dREERt = REERt – REERt–1

(6.1b)

dCVdirect = dREERt = dXRt + dPGDPt − dPGDPt, usa

(6.1c)

dCVsstill + dCVdirect = dCV,

(6.1d)

where XR is the nominal exchange rate with respect to the United States and dXR is positive when the domestic currency appreciates, and PGDP and PGDPusa are the GDP deflators for the domestic country and the United States, respectively.

Changes in Currency Valuation, 1980–2011 Table 6.1 shows RERdirect and RERsstill for individual countries. The total change in RER is just the sum of the two components (because the components are reported as log percentages). Table 6.2 reports the same computation for different regions. Recall that the direct effect ignores productivity changes, and the difference between the direct and total change reflects the attempts by governments to keep their currencies undervalued by intervening in the foreign exchange markets (accumulating reserves). Estimates of changes are provided for two periods, 1980–94 and 1995–2011. The results reinforce the importance of accounting for trends in relative productivity. Table 6.1 can be read as follows: Total real depreciation for 1980–2011 in China was –230 percent. Of this, almost one-third was due to the net effect of currency change and excess inflation; two-thirds was due to standing still. India also shows a parallel, but considerably less real devaluation than China, –161 percent. But this is where the similarity ends. The RER change in China was accompanied by a marked improvement in the current account surplus, from near-balance in 1980, to a peak level of 11 percent in 2007, and an average level of 2.4 percent for the full 32-year period. In contrast, India has consistently had a current account deficit, which averaged –1.3 percent of GDP for the same period. As shown in table 6.2 for 1995–2011, the average direct effect for East Asia is a cumulative appreciation of (log) 36 percent, which garners attention and elicits comments about the region’s appreciating currencies. However, the total effect during the same period is a large real depreciation of log 41 percent or 34 percent in conventional terms. For South Asia the divergence in the two estimates is even greater: The direct effect is a log 20 percent appreciation (mostly due to higher inflation) while the total effect is a log 47 percent real depreciation. The developed economies, as expected, show very little real exchange rate change, and the contribution of the standing-still component is small. Some other important conclusions can be drawn from these tables. First, the standing-still component accounts for a larger share of the changes for countries with pronounced currency intervention policies and strong 86

DEVALUING TO PROSPERITY

Table 6.1

Direct and indirect (standing still) estimates of real currency devaluations, 1980–2011 (in logarithms) 1980–94

Country

1995–2011

1980–2011

RERdirect

RERsstill

RERdirect

RERsstill

RERdirect

RERsstill

Residual change (1)

Standing still (2)

Residual change (3)

Standing still (4)

Total (3 + 4)

Residual change (5)

Standing still (6)

Total (5 + 6)

Argentina

–27

5

–43

–18

–61

–70

–13

–83

Australia

–17

–2

55

–7

48

38

–9

29

IS THE REAL EXCHANGE RATE ENDOGENOUS? 87

Brazil

–7

1

56

–18

38

49

–18

31

Canada

–9

1

35

–4

31

25

–2

23

Chile

–39

–19

39

–28

11

0

–47

–47

China

–127

–58

57

–102

–45

–70

–160

–230

France

–10

1

9

–1

8

–1

–1

–2

0

–8

–5

0

–5

–6

–8

–14

Germany Greece

4

4

29

–10

19

34

–6

28

24

–36

–42

3

–39

–18

–33

–51

India

–63

–44

20

–72

–52

–44

–117

–161

Indonesia

–43

–55

50

–39

11

7

–95

–88

Israel

18

–19

13

–13

0

31

–33

–2

Italy

12

–8

25

4

29

37

–5

32

Hong Kong

Japan

48

–11

–28

2

–26

20

–9

11

Korea

16

–61

–15

–23

–38

1

–84

–83

(continues on next page)

88 DEVALUING TO PROSPERITY

Table 6.1

Direct and indirect (standing still) estimates of real currency devaluations, 1980–2011 (in logarithms) (continued) 1980–94

1995–2011

1980–2011

RERdirect

RERsstill

RERdirect

RERsstill

RERdirect

RERsstill

Country

Residual change (1)

Standing still (2)

Residual change (3)

Standing still (4)

Total (3 + 4)

Residual change (5)

Standing still (6)

Total (5 + 6)

Malaysia

–38

–27

–2

–21

–23

–40

–48

–88

Mexico

13

–4

17

–12

5

30

–16

14

Philippines

–3

13

8

–32

–24

4

–19

–15

Russia

33

–4

100

–23

77

133

–27

106

–55

29

57

3

60

2

32

34

26

–32

–2

–14

–16

24

–46

–22

Saudi Arabia Singapore South Africa Taiwan Thailand Turkey

6

6

22

–15

7

28

–9

19

57

–84

–41

–24

–65

–9

–83

–92

–7

–57

–6

–27

–33

–13

–85

–98

–82

–18

46

–37

9

–36

–56

–92

United Kingdom

2

–7

16

–4

12

18

–11

7

United States

0

10

0

8

8

0

18

18

Notes: The direct effect is the conventional estimate of change in the real effective exchange rate (REER), due to exchange rate fluctuations and inflation differences. The indirect effect (columns 2, 4, 6) refers to the change in the RER due to the differences in productivity growth, the “standing still” effect. Source: Bhalla (2007a) dataset extended to 2011.

Table 6.2

Different estimates of (log) changes in currency valuation, 1980–2011 Estimate of real exchange rate (RER) change (log percent) Direct only

Region

1980–94

Developed economies

6

1995–2011

1980–2011

4

10

East Asia

–73

36

–37

Russia and Eastern Europe

–14

93

79

–9

33

24

–49

40

–9

Latin America Middle East and North Africa South Asia

–58

20

–38

Sub-Saharan Africa

–49

22

–26

All regions

–13

24

11

Direct + standing still 1980–94 Developed economies East Asia Russia and Eastern Europe Latin America Middle East and North Africa South Asia

1995–2011

1980–2011

5

6

11

–125

–41

–166

3

60

62

–9

15

6

–52

20

–32

–101

–47

–148

Sub-Saharan Africa

–28

–10

–37

All regions

–22

–3

–25

Notes: The regional classifications follow those of the World Bank. The direct effect refers to the “traditional” currency change and inflation differential computation of real exchange rate effect. “Direct plus standing still” refers to the total change in valuation, which is the log sum of the direct effect plus the Balassa-Samuelson indirect effect that incorporates differences in productivity growth. Data for individual countries are weighted by GDP to obtain regional averages. Source: Bhalla (2007a) dataset extended to 2011. See text and table 6.1 for the definitions of direct and indirect (standing still) components of changes in currency valuation.

accumulation of foreign reserves, namely, the developing, especially East Asian, economies. Second, the developed economies are generally characterized by a small real change during 1995–2011. Both Greece6 and the United States have had large current account deficits, and both show large appreciation, 28 and 18 6. Since January 1998, the countries of the European Union have a common currency, the euro. The differences in the real exchange rate change among the Euro countries for the post-1997 period arise out of different rates of inflation, different rates of per capita income growth, and different RER-income elasticities.

IS THE REAL EXCHANGE RATE ENDOGENOUS? 89

percent, respectively. Germany shows a net 14 percent depreciation since 1980. Japan has had an 11 percent overvaluation. The fact that these two countries have shown a divergent trend in their economic growth and current account surpluses may not be entirely coincidental. Third, between 1980 and 1994 China experienced the fastest and largest real devaluation in the postwar period. Over the next 17 years (1995–2011), the Chinese renminbi further depreciated, in real terms, by 45 percent. This counteracts assertions that the Chinese renminbi appreciated by close to 20 percent during 2005–11 (from 8.3 yuan to $1 to 6.6 yuan to $1); in fact, China enjoyed a cumulative real depreciation of close to 6 percent during this period!

Is China a Currency Manipulator? The Chinese renminbi was heavily overvalued in 1980 and needed a large real devaluation. By 1994, it was one-sixth of its real value in 1980 and was the sixth most undervalued currency.7 There were no objections to this devaluation. Indeed, international institutions, scholars, and policymakers all recommended such a change. But the magnitude of the real devaluation was unexpected. Starting in 1994, and lasting until 2005, China maintained a fixed exchange rate with the dollar, an exchange rate pegged at the undervalued rate of 8.3 yuan to $1. And China’s policy began to be questioned, precisely because of the standing-still depreciation and accompanying large changes in the current account balance. This standing-still policy has increased the undervaluation of the renminbi to the extent that by 2011 it was one of the most undervalued currencies, second only to Mauritius outside of the former Soviet Union economies. Some observers do not consider the Chinese currency to be undervalued; and some who do believe that China should proceed as usual. Among them are Ronald McKinnon and Gunther Schnabl (2006, 277), who state: “In the new millennium, history is repeating itself, but now China bashing is superseding Japan bashing.”8 It is worth noting that the last time the United States defined any country as a currency manipulator was when it defined China as such in 1994. In 1994, the renminbi was undervalued by only 12 percent versus a 44 percent undervaluation today, but the US Treasury no longer defines China as a currency manipulator. In its 2005 report, the Treasury said that reaching judgments about whether countries manipulate the rate of exchange between their currency and the United States dollar for purposes of preventing effective balance of payments adjustments or gaining unfair competitive advantage in international

7. Among nonoil and non–former Soviet Union and East European economies. 8. It gets a bit confusing because many of the critics who fear a Japan repeat also believe that the real exchange rate, and therefore the under- or overvaluation of a currency, does not affect economic growth.

90

DEVALUING TO PROSPERITY

trade is inherently complex, there are several indicators that can provide helpful information in considering this question. However, no single indicator, or set of indicators, in and of itself, can establish that a specific economy has met the technical requirements for designation under the Act, and the context in which these questions are assessed varies with individual country circumstances. (US Treasury 2005, 1; emphasis added) .

In Summary Over the past 30 years, the world has rapidly globalized and integrated. This integration implies that the currencies of the fast-growing countries should have appreciated; instead, the opposite has happened. The hypothesis that inflation inevitably follows devaluation by an equal amount was emphatically rejected in this chapter. There is still a possibility of the endogeneity of real exchange rate change, but this endogeneity suggests a positive, not negative, relationship between the RER and income growth. Fast income growth should be associated with an appreciating RER, at least according to the stylized fact of Balassa-Samuelson. What has been documented is that there is a negative relationship—that is, fast-growing economies are observed to have a depreciating RER. That this has occurred not through endogeneity but active policy action is indicated by the fact that the countries have intervened and accumulated foreign exchange reserves in the process of keeping their exchange rate cheap (undervalued). In brief, the real exchange rate is not endogenous.

IS THE REAL EXCHANGE RATE ENDOGENOUS? 91

7 Rashomon Rules: US Dollar, Euro Dollar…

Take nothing on its looks; take everything on evidence. There’s no better rule. —Charles Dickens, Great Expectations Akiro Kurosawa’s 1950 movie Rashomon documents the tenuous relationship between perception and reality: The same reality has different perceptions and is therefore a different reality for different people. Today, oftentimes, and especially in the social sciences, one can see the Rashomon effect in action. For example, are global imbalances because of China’s undervalued currency or because of America’s fondness of consumption? If Rashomon rules, you can take your pick. Imbalance is the operative word for the world economy. There are imbalances between some countries’ savings and others’ deficits; imbalances in growth; and imbalances in debt. Of all the imbalances, current account imbalances may be the most problematic. They have the potential to be and have the very recent history of creating global crises—two in just the last four years, the US subprime mortgage crisis of 2008 and the European debt crisis of 2010–11. It is assumed that current account imbalances reflect the relative “priceyness” of the exchange rate. But how much of the world’s imbalances can really be explained by movements in exchange rates? Previous chapters have outlined the role of currency valuation, which is often a matter of policy, in determining investment and economic growth. A new nonlinear method of measuring currency misalignment was introduced, tested against the existing alternatives, and shown to be an important determinant of investment in both developed and developing economies. This chapter is the third in a series of smell tests of both that measure of currency valuation and its effect on economic variables. The tests presented here may be the most stringent. If the exchange rate does have an effect on current account imbalances, then the twin US and European crises are tailor-made to be a test of whether currency valuation is 93

correctly measured. The current account imbalances in the United States and in the European Union are contentious. There has been a rich tradition of arguing for a change in the value of the dollar to combat high and rising US current account deficits. This is somewhat more difficult than a run-of-the-mill devaluation because the US dollar is a reserve currency. And there is the question of just how much a change in the value of the dollar could affect the US current account balance. In Europe, the imbalances are largely between the very strong economies at the core of the euro area (Austria, Belgium, Finland, France, Germany, Luxembourg, and the Netherlands) and the five euro area economies now in crisis (Greece, Ireland, Italy, Portugal, and Spain). Discussions about the survival of the euro often center around the inability of the crisis economies to address their current account imbalances through adjustments in the exchange rate. A broader issue is the role of the valuation of the euro in creating these current account imbalances to begin with. Previous chapters have demonstrated that currency valuation has an effect on current account balances. It follows that the more accurate the currency valuation measure, the greater its power or statistical significance in explaining trends in the current account balances of the United States and the European Union (or, more accurately, the euro area). Are there differences in the explanatory power of the various currency valuation alternatives? For example, there are several indices of the real value of the US dollar. How well do the different estimates explain the recent declines and improvements in the US current account balance? How do the different estimates compare with the officially sanctioned benchmark estimate of the US dollar, the Federal Reserve’s Broad Index?1 The first-order effect of currency valuation is on current account imbalances. This chapter conducts smell tests of whether currency valuations adequately explain the US and euro area imbalances. If they do, then the smell tests will be complete and we can turn toward evaluating the role that currency misalignments play in affecting growth, the main focus of this book.

US Current Account Deficit In 2005 and 2006, the United States had a current account deficit that exceeded 6 percent of GDP. In the mid-1990s, Thailand had a current account deficit of about that magnitude and that led to the collapse of the Thai baht in 1997 and to the Asian crisis of 1997–98. The financial crisis of 2008–09 drew considerable attention to the US current account deficit and to the question of whether it led to the global financial crisis of 2008–09. This question has been, and will continue to be, the focus of considerable research. Obviously, there are many explanations for the US current account deficit. US saving rates are too low (the definitional corollary of the current account 1. This monthly series has been available since January 1973. The index is a weighted basket comprising the currencies of the 37 major economies that trade with the United States.

94

DEVALUING TO PROSPERITY

being in deficit). The economy has structural flaws, such as a lack of skilled labor and a surplus of unskilled labor. Health care costs are rising faster than growth, and the population is aging. These and other factors all play a role. But the concern here is to explore the extent to which one particular factor, the value of the US dollar, explains the current account imbalance in the United States. This is not the usual focus of discussions about currency valuation measures. Few studies address how particular currency valuation methods affect estimates of the equilibrium exchange rate and how that in turn affects the US current account.2 This may be the result in part of the fact that the dollar is used to define the real exchange rates for other currencies (it is the numeraire in the RER calculation). Because $1 equals PPP$1, the dollar’s RER is always assumed to be 1. This is how the whole system of accounts is constructed. However, as shown below, there is a method to extract the equilibrium value of the dollar, and therefore a measure of its undervaluation or overvaluation, and to estimate its effect on the US current account.

Valuations of the Dollar The Broad Index has been used extensively to explain the behavior of US exports and imports. For example, Martin Baily and Robert Lawrence (2006) use it to help explain a proxy for the US current account, the log of the ratio of exports to imports. Charles Thomas, Jamie Marquez, and Sean Fahle (2008) construct a parallel Broad Index for the real effective exchange rate of the dollar which they call the weighted average relative price (WARP) index. It is useful to review the procedures used for calculating these two widely used dollar indices. The Broad Index uses trade weights to construct a simple real exchange index for each month and therefore each year. The national consumer price index (CPI) is used as a deflator. The WARP index combines “bilateral relative prices into an aggregate measure for the United States,” using a “weighted geometric mean where the weights vary over time and reflect each country’s importance in US trade” (Thomas, Marquez, and Fahle 2008, 5). The WARP estimate is extremely data intensive, as it employs monthly/ quarterly trade, inflation, and exchange rate data for a large proportion of the Broad Index countries. One disadvantage of both these measures is that they yield estimates of overvaluation and undervaluation not on an absolute basis but relative to 2000 or another chosen base year. For example, both methods would yield a valuation in, say, 2011 that is relative to its value in 2006. That means that a judgment can be made on whether the dollar is undervalued or overvalued in a particular year only by referring to other information, such as the US current account.

2. The singular and important exception is the fundamental equilibrium exchange rate (FEER) method of John Williamson (1994). See Cline and Williamson (2011) for a discussion of how the US current account can be expected to behave with changes in exchange rates. Unfortunately, no time series of FEER estimates is available.

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 95

Both the Broad and WARP indices estimate the value of the dollar directly from data on exchange rates and inflation. In Bhalla (2007a) I develop an index for the dollar that is considerably indirect and is structurally different from both the Broad and the WARP indices. First, it uses annual, not monthly, data on exchange rates. Second, instead of the CPI it uses the GDP deflator as an index for inflation. Third, it uses data from Penn World Table 6.1 on PPP exchange rates (1996 base) and PPP income. Fourth, it estimates a nonlinear regression relating the RER (the ratio of the PPP exchange rate to the nominal exchange rate against the dollar) and real per capita PPP income. This regression is estimated for more than 100 countries for 1996–2009 (equation 4.6), and forward and backward forecasts are obtained through model projections. Finally, the predicted value from this equation yields the predicted RER (RER*) for each country against the PPP dollar. Deviations between the actual and predicted RER show how much each currency, including the US dollar, is undervalued or overvalued with respect to the PPP dollar. However, the valuation of primary interest is against the actual US dollar and not the PPP dollar, a currency that no one has ever seen or traded. By construction, the valuation of each currency against the US dollar is the (log) difference between each country’s valuation against the PPP dollar (equation 4.6) and the US dollar’s valuation against the PPP dollar. The latter is not equal to 1 but is the predicted value given income per capita. For example, in 2011, the valuation of the Chinese renminbi against the dollar (log –57.5 percent) is the difference between its valuation against the PPP dollar (CVppp = –62.7 percent) and the dollar’s valuation against the PPP dollar (CVppp = –5.2 percent).3 Adjusting the valuations of other currencies to account for undervaluation or overvaluation of the dollar is a big improvement over existing models. Johnson, Ostry, and Subramanian (2007), Rodrik (2007a), Dollar (1992), and Easterly (2001) do not make such an adjustment. Now we estimate the valuation of the dollar. The procedure starts from the definition that the valuation of the dollar is the mirror image (negative) of the valuation of all currencies against the dollar. Thus, the misalignment of the dollar against the 37 currencies in the Broad Index is the negative of the weighted average, with trade share as weights, of the misalignment of these 37 currencies. This yields a dollar valuation of –7 percent in 2011. Table 7.1 lists the estimated value of the dollar for selected years from 1960 to 2011, using 2000 as the base year. According to the WARP index, the dollar was just as undervalued in 2000, when the current account deficit was 4 percent, as in 2006, when the deficit hit its peak at 6.1 percent. Also reported are the estimates according to Johnson, Ostry, and Subramanian (2007) and The Economist’s Big Mac Index. The Economist does not publish its estimate of how much the dollar is misaligned; it only publishes the valuation of other currencies against the dollar based on the price of the Big Mac sandwich. 3. These estimates are in log terms. A log –57.5 percent for China translates as –43.7 percent in conventional terms.

96

DEVALUING TO PROSPERITY

Table 7.1

Dollar valuation by different measures, 1960–2011 Currency valuation index (2000 = 100)

Year

Broad Index

WARP Index

The Economist’s Big Mac Index

Bhalla (2007a)

Johnson, Ostry, and Subramanian (2007)

US current account balance (percent of GDP)

1960

n.a.

n.a.

n.a.

87.8

123.9

0.5

1973

95.2

87.5

n.a.

85.9

94.8

0.6

1985

118.1

105.1

n.a.

106.9

113.8

–2.8

1995

83.2

80.9

62.9

83.8

86.7

–1.5

2000

100.0

100.0

100.0

100.0

100.0

–4.2

2006

92.4

100.0

93.4

99.3

88.5

–6.0

2007

88.0

n.a.

93.9

96.1

83.5

–5.1

2008

84.3

n.a.

76.5

91.4

79.2

–4.7

2009

87.8

n.a.

112.0

100.5

88.3

–2.7

2010

83.8

n.a.

111.5

97.0

85.1

–3.2

2011

79.3

n.a.

104.3

91.8

79.7

–3.1

n.a. = not available; WARP = weighted average relative price Notes: See tables 4.2 and 4.3 for definitions of the various currency valuation estimates. All currency valuation indices are indexed to 100 for the year 2000. An increase in the index signifies appreciation. Broad Index is the real exchange rate index for 37 major US trading partners. WARP Index is defined in Thomas, Marquez, and Fahle (2008). Sources: Penn World Table 6.1 (Heston, Summers, and Aten 2002); Maddison (2003); World Bank, World Development Indicators; and IMF, World Economic Outlook database. See appendix A for details.

Therefore, the Big Mac Index for the US dollar has been computed on the same basis as my estimate—the negative of the weighted value of the 37 Broad Index currencies. There is wide variation. For 2011, my index has the dollar at a level 8 percent lower than in 2000; the Johnson, Ostry, and Subramanian index has the dollar 21 percent lower and the Big Mac Index shows the dollar level as 4 percent higher. Table 7.2 reports a correlation matrix for the selected dollar estimates. The Johnson, Ostry, and Subramanian (2007) estimates are out of line with the other three indices. It is also noteworthy that the very direct WARP estimates are very closely correlated with my extremely indirect estimates—the correlation level is 0.99, and this close correspondence is also brought about by the regressions explaining the behavior of the US current account (reported below). The official Broad Index has correlation levels of 0.79, 0.76, 0.94, and 0.44, respectively, with the Bhalla, WARP, Johnson-Ostry-Subramanian, and Big Mac indices. However, the real test is not the correlations in the levels per se, but rather how well these indices explain the US current account.

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 97

Table 7.2

Correlation between different currency valuation measures for the US dollar, 1993–2011 (2000 = 100)

Measure

Bhalla (2007a)

Johnson, Ostry, and Subramanian (2007)

Thomas, Marquez, and Fahle (2008)

The Economist‘s Big Mac Index

Bhalla (2007a)

1.00

Johnson, Ostry, and Subramanian (2007)

0.64

1.00

Thomas, Marquez, and Fahle (2008)

0.99

0.43

1.00

The Economist‘s Big Mac Index

0.81

0.09

0.89

1.00

Broad Index

0.79

0.76

0.94

0.44

Broad Index

1.00

Note: See text for definition of various currency valuation indices.

Currency Valuations and the US Current Account Deficit Trade is primarily a function of a country’s level of development (income per capita) and the exchange rate. The demand for any good is a function of income and price. And the price, as far as trade is concerned, is nothing but the exchange rate. The current account is the balance of exports and imports of all traded goods and services. Currency values directly affect these three variables— current account, exports, and imports. In order to explain the US current account, Baily and Lawrence (2006) use a related variable, net exports, defined as the (log) of the ratio of US exports to imports.4 For estimates of the RER— the price of the dollar—Baily and Lawrence use the Broad Index: log(Exports/Imports) = a + b × L1.log(CV) + c × L2.log(CV),

(7.1a)

where L1 and L2 are the lags of 1 and 2 years, respectively, and CV is the Broad Index for the dollar. The equation is estimated for the years between 1978 and 2005. The sum of the coefficients (b + c) is an estimate of the elasticity of net exports; if the value is less than –1 (their estimate is –1.15), then a devaluation can help improve the US current account. Alternatively, the following equation relates the share of current account in GDP to currency valuation: (Current Account Balance/GDP) = a + b × L1.log(CV) + c × L2.log(CV) (7.1b)

4. The current account balance, Baily and Lawrence (2006) convincingly argue, is equivalent to the ratio of the two constituents of the current account balance—gross exports of goods and services minus gross imports of goods and services.

98

DEVALUING TO PROSPERITY

Table 7.3

Explaining the US current account balance: Log (exports/imports), 1978–2006 Index of currency valuation Adjusted R2

Number of observations

0.67

29

0.96

29

–0.41 ) (–1.2)**

0.36

29

–0.74*** (–5.9)**)*

0.92

29

Variable

Lag 1

Lag 2

Broad Index

–0.37 (–1.4))

–0.79** (–2.6)**)

WARP Index

–0.79*** –0.37***) (–5.24)*** (–10.1)***)

Johnson, Ostry, and Subramanian (2007)

–0.55* (–1.7)*)

Bhalla (2007a)

–0.51*** (–4.6)***)

WARP = weighted average relative price Note: Broad Index is the real exchange rate index for 37 major US trading partners. WARP Index is defined in Thomas, Marquez, and Fahle (2008). Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Numbers in parentheses are t statistics.

Table 7.4

Explaining the US current account balance: Percent of GDP, 1978–2006 Index of currency valuation

Variable

Lag 1

Lag 2

Adjusted R2

Number of observations

Broad Index

–2.43) (–0.5)0

–9.63*) (–1.8)3*

0.39

29

WARP Index

–4.96*) (–1.9)6*

–9.84***) (–3.6)***4

0.83

29

Johnson, Ostry, and Subramanian (2007)

–4.74) (–1.0)4

–5.13 ) (–1.1) 3

0.19

29

Bhalla (2007a)

–6.0**) (–2.3)**

–10.21***) (–3.4)1***

0.82

29

WARP = weighted average relative price Note: Broad Index is the real exchange rate index for 37 major US trading partners. WARP Index is defined in Thomas, Marquez, and Fahle (2008). Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Numbers in parentheses are t statistics.

Equation 7.1b is equivalent to equation 7.1a, except for the construction of the dependent variable. Table 7.3 presents estimates for the Baily and Lawrence model for four different measures of the RER for the period 1978–2006 and the dependent variable as the (log) ratio of US exports to US imports. Table 7.4 uses the conventional dependent variable, the current account balance as a share of GDP. The Johnson, Ostry, and Subramanian (2007) method is chosen to be RASHOMON RULES: US DOLLAR, EURO DOLLAR… 99

representative of the alternate log-log method of estimating currency valuations. In addition to the Broad Index, Bhalla (2007a), and Johnson, Ostry, and Subramanian (2007), the tables also present estimates based on the data-intensive WARP Index method (Thomas, Marquez, and Fahle 2008). The WARP Index explains changes in net exports for the United States to an extraordinarily large degree—96 percent of the variation for 1978–2006 is explained just using the WARP currency valuation measure. This is considerably more than the explanatory power obtained by Baily and Lawrence (2006) using the Broad Index (67 percent). In sharp contrast, the Johnson, Ostry, and Subramanian (2007) method performs very poorly, explaining only 26 percent of the variation. Somewhat surprisingly, and reassuringly, the highly indirect Bhalla (2007a) index performs almost as well as the WARP method—explaining 90 percent of the variation. For purposes of forecasting the current account, net export/import elasticity is most meaningful. Both the method used by Baily and Lawrence (2006) and the WARP method used by Thomas, Marquez, and Fahle (2008) produce near-identical elasticities of 1.16 and 1.15. This suggests that for each 1 percent decrease in the trade-weighted value of the dollar, exports will increase by 1.16 or 1.15 percent, and imports will decline by 1 percent. The Johnson, Ostry, and Subramanian (2007) method suggests that devaluation of the dollar will be counterproductive for net exports, with the elasticity, at 0.89, less than 1— that is, a 1 percent devaluation of the dollar will lead to a 1 percent decline in imports and a 0.9 percent increase in exports. My method yields the highest elasticity of net exports with respect to price: 1.25. And as already mentioned, the variation explained is above 90 percent. Both are consistent with the other findings and the theme of this book that countries can undervalue their way to prosperity. Figure 7.1 plots the value of the dollar since 1973 using three measures— the Broad Index, WARP Index, and Bhalla (2007a). There is an expectation that the WARP method closely parallels the official Broad Index; the real surprise is how close my very indirect estimate tracks the valuations produced by these other direct indices. Figures 7.2 and 7.3 plot the actual and predicted values for the two definitions of the current account—the actual balance and the log ratio of (exports/imports). Regardless of the definition, three years (2005, 2006, and 2007) appear to be exceptional for the US current account. Several observations emerge. First, as expected, the ratio of log (exports/ imports) is a very good proxy for the current account deficit. Second, there is a very close fit between the actual and predicted (log) ratio of exports to imports. The fit is especially remarkable for the years from 1995 to 2003. Third, the correlation between Bhalla (2007a) and the WARP Index is nearly perfect (0.9710) for the period for when both datasets are available (1971–2006). Fourth, the out-of-sample predictions of both for the postestimation period are reassuring. The WARP data extend only until 2006 (and so predictions are available only until 2007). The Bhalla (2007a) method captures the out-ofsample reality till 2011 extremely well. The numbers for the dependent variable 100

DEVALUING TO PROSPERITY

Figure 7.1 Three estimates of the real value of the US dollar, 1973–2011 index (2000 = 100) 115 110 105 100 95 90 85 WARP Index Bhalla (2007a) Broad Index

80

5 197 7 197 9 198 1 198 3 198 5 198 7 198 9 199 1 199 3 199 5 199 7 199 9 200 1 200 3 200 5 200 7 200 9 201 1

197

197

3

75

WARP = weighted average real price Note: The three indices have been indexed to 100 for the year 2000. Broad Index is the real exchange rate index for 37 major US trading partners. WARP Index is defined in Thomas, Marquez, and Fahle (2008).

(log of exports to imports) compared with the Bhalla (2007a) predicted values for 2007–11 are as follows: in 2007, –0.34 versus –0.33; in 2008, –0.33 versus –0.32; in 2009, –0.22 versus –0.27; and in 2011, –0.23 versus –0.27. Thus, reassuringly, the models (and nonofficial estimates of the value of the dollar) capture the turning points, and in particular, the recent sharp improvement in the US current account. At about –3 percent, the US current account in 2009 and 2011 is back to the levels prevailing in the late 1990s.5 This extended discussion of the US current account balance is meant as a smell test, and the results reinforce the conclusion that the Bhalla (2007a) method captures currency valuation to a significantly greater degree than other available measures. The poor performance of the popular log-log model supports the hypothesis that it contains gross measurement errors and that these errors may have been responsible for the popular but erroneous conclusion that currency valuation does not affect economic growth. Even so, this extended smell test tells us only about the misrepresentation with respect to

5. It should be noted that this improvement is not due to the decline in the price of oil. Exclusion of oil shows just as sharp an improvement in the US current account, a point also noted by Lawrence (2009).

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 101

Figure 7.2

Explaining the US current account deficit, 1990–2011

percent of GDP –0.5

–2.0

–3.5

–5.0

–6.5 1990

Actual current account balance Predicted, WARP Index Predicted, Bhalla (2007a) 1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

WARP = weighted average real price Note: Dependent variable is the share of US current account balance as a share of GDP. WARP Index is defined in Thomas, Marquez, and Fahle (2008). Source: IMF, World Economic Outlook Database.

the dollar. While highly unlikely, it is theoretically possible that the Johnson, Ostry, and Subramanian (2007), Rodrik (2007a), or Easterly (2005) measures may better represent the reality for other currencies.

Smell Test—The Euro and the European Imbalance An equally important smell test is the extent to which the various currency valuation measures for the euro explain European current account imbalances. For the last two years, 2010 and 2011, policymakers and financial markets have been fixated on the sovereign debt crisis in the euro area—specifically, in Greece, Ireland, Italy, Portugal, and Spain. Each of these economies faces twin deficits, in the fiscal balance and the current account balance. The current account deficits were much larger in the years leading up to the crisis (2005–07), and fiscal deficits were larger in the years after the crisis (2008–11). The euro area provides a natural experiment for testing the efficacy of currency valuation estimates—a common currency shared by a large number of economies. It was documented above that the nonlinear Bhalla (2007a) 102

DEVALUING TO PROSPERITY

Figure 7.3

Explaining US current account imbalances, 1978–2011

log 4.95 Actual log (exports/imports) Predicted, WARP Index Predicted, Bhalla (2007a) 4.85

4.75

4.65

4.55

197 1978 9 198 0 198 1 198 2 198 3 198 4 198 5 198 6 198 7 198 8 198 9 199 0 199 1 199 2 199 3 199 4 199 5 199 6 199 7 199 8 199 9 200 0 200 1 200 2 200 3 200 4 200 5 200 6 200 7 200 8 200 9 201 0 201 1

4.45

WARP = weighted average real price Notes: The general model under consideration for the figure is log(Vx/Vi) = α + β1 × L1.CV + β2 × L2.CV, where the dependent variable is the log ratio of exports to imports for the United States and CV is the respective currency valuation measure. WARP Index is defined in Thomas, Marquez, and Fahle (2008).

measure does remarkably well in explaining the US current account imbalances. Its valuation of the euro does well in explaining the persistent differences in the current account balances of various euro area economies. The euro forces countries to have the same nominal exchange rate, but their various inflation rates and levels of income per capita mean that they have different currency valuation levels. The valuations are estimated with reference to Germany, not the United States. I test the ability of these different valuation levels to help explain the differences in the current account balances of the core euro area economies and the crisis economies. The euro was introduced in January 1999 and was initially adopted by 12 countries. In 1998, these countries had levels of income per capita that ranged from $11,638 in Portugal to $26,668 in Germany. This wide disparity meant that differential inflation and productivity growth paths would mean a different competitive value of the euro for each country. Even if all 12 countries were competitive at their pre-euro exchange rate levels, at least according to the evidence and analysis presented earlier, these countries would no longer be competitive after adoption of the common currency. It is just not possible for RASHOMON RULES: US DOLLAR, EURO DOLLAR… 103

a single currency to be competitive for all nations…unless this single currency depreciated in real terms. This is exactly what happened, and why the debt problems were not immediately apparent. The initial exchange rate for the euro was €1.18 to $1. Greece adopted the euro on January 1, 2001. From 2003 until about 2006, the euro averaged €1.25 to $1. In 2007, however, the euro averaged €1.37 to $1, and the European debt crisis was born. In 2010, there were 16 countries using the common currency.6 Also in 2010, two years after the global financial crisis, the euro exchange rate started to assume center-stage. And a reasoned, and reasonable, argument began to take shape. The common currency could not survive unless the euro exchange rate was competitive for all the euro area economies, not just the competitive economies at the core but also those on the periphery. Why wasn’t this argument made until 2010? Table 7.5 contains the answer. The data are presented for seven core euro area countries (Austria, Belgium, Finland, France, Germany, Luxembourg, and the Netherlands) and five countries on the periphery, the crisis economies of Greece, Ireland, Italy, Portugal, and Spain. Data for both groups of countries are aggregated to give results for two synthetic economies. Data are presented for selected economic variables for three periods: period 1, the eight years before the introduction of the euro (1991–98); period 2, the six years of relatively benign euro values (1999–2004); and period 3, the four years of high euro values (2005–08). When the euro was introduced in January 1999, the average income of the peripheral countries was two-thirds that of the core countries. For the previous seven-year period, currency valuation was a near neutral 3 percent overvalued for the core countries and 20 percent undervalued for the peripheral countries. Their current accounts were in balance. The value of the euro continuously declined from €1.18 to $1 in January 1999, to a low of €0.83 to $1 in October 2000. The average value in 2003 was €1.13 to $1. The euro’s decline from its introductory value of €1.18 to $1 helped all the euro area economies to prosper. The annual growth rate for the peripheral countries accelerated to 2.5 percent during 1999–2004 from 1.9 percent for 1991–98. Their current accounts deteriorated but were still at manageable levels of only –2.3 percent of GDP. However, starting in 2005, the euro began to steadily appreciate and climbed to its record €1.58 to $1 in July 2008. This euro appreciation can be considered the equivalent of the US subprime mortgage crisis for the peripheral economies. The data strikingly confirm the hypothesized relationship between the value of the euro and the differential impact on the current account balances of the peripheral and core euro area countries. Changes in exchange rates and currency valuation directly impact the current account.7 6. Estonia, the most recent of the current euro area members, adopted the euro on January 1, 2011. 7. The economics behind these developments in the euro currency union and its implications for the individual member countries are entertainingly described by Martin Wolf (2010) in his rendition of the grasshopper and ant fable.

104

DEVALUING TO PROSPERITY

Table 7.5 The euro area: Current account balances and related indicators, 1991–2008 Core economies

Peripheral economies

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 105

Currency valuation versus Germany (percent) (1)

Current account balance (percent of GDP) (2)

Average growth in GDP (percent) (3)

Currency valuation versus Germany (percent) (4)

Current account balance (percent of GDP) (5)

Average growth in GDP (percent) (6)

1991–98

2.7

0.4

1.8

–19.8

–0.1

1.9

1999–2004

2.2

1.8

1.7

–12.9

–2.3

2.5

2005–08

1.6

3.4

2.1

–0.2

–5.8

1.7

Period

Notes: The peripheral economies are Greece, Ireland, Italy, Portugal, and Spain; the core economies are the other seven countries that orginally adopted the euro in 1999 (Austria, Belgium, Finland, France, Germany, Luxembourg, and the Netherlands). The currency valuation estimates are obtained using the income per capita and the exchange rate prevailing at the time of their adoption of the euro in January 1999 (except for Greece, which joined the euro area in January 2001). Individual country inflation rates are used to estimate the nominal and real exchange rates. Sources: Columns 1 and 4 are author’s calculations, see text; other columns from IMF, World Economic Outlook Database.

A formal analysis of the euro area current account balances can be conducted as follows.8 The equation or relationship between currency valuation and the current account balance for a single country is easily generalized for groups of countries. Briefly, the hypothesized relationship is as follows, with CAB representing the weighted current account balance (as a proportion of GDP) for time t and region 1 or 2 (peripheral or core countries, respectively), and CV representing the weighted currency valuation (relative to the value of the Deutsche mark): CAB1t = a1t + b × CV1t

(7.2a)

CAB2t = a2t + b × CV2t

(7.2b)

Taking the difference in the two equations, one obtains: CAB1t – CAB2t = (a1t − a2t ) + b × (CV1t – CV2t)

(7.2c)

Or, dCAB = CAB1t – CAB2t = (a1t − a2t ) + b × dCV

(7.2d)

Equation 7.2c represents a fixed region effects model. It is estimated by regressing not the levels but rather the first differences (equation 7.2d). The difference model allows for the control of effects that are particular to a country or region (for example, Germany as a perennial surplus country possibly due to its high taste for savings, other things equal). A simple regression for annual data for the period from 1993 to 2011 between dCAB and dCV yields a correlation of −0.90 and a coefficient on dCV of −0.31; for the longer period, 1980–2011, the correlation is still large at –0.79 and the coefficient is –0.259 (table 7.6). These are large effects; these estimates indicate that if the peripheral economies were to depreciate their currencies relative to the core countries by just 10 percent, their weighted current account imbalance would improve somewhere between 2.5 and 3.1 percentage points— that is, they would have current account deficits of about 1 to 2 percent, rather than 4 percent. Figure 7.4 plots the differences between the current account balances of the peripheral and the core economies against the differences in their levels of currency valuation. It is a near perfect fit, and the strength of the correlation underlines the importance of currency valuation as a determinant of current account imbalances and provides a plausible explanation for the troubles facing the peripheral economies after the strong appreciation of the euro.

8. See Bhalla (2011) for a detailed analysis. 9. Both regressions exclude the outlier Deutsche mark peak year of 1995.

106

DEVALUING TO PROSPERITY

Table 7.6

Currency valuation and current account balance for the euro area, 1980–2011 Difference between the current account balances of peripheral and core economies

Dependent variable Difference in currency valuation (core minus peripheral)

1993–2011

1980–2011

–0.31***

–0.25***

Time dummy, 1995

3.79***

2.38***

R2

0.83***

0.64***

19******

32******

Number of observations

Notes: The peripheral economies are Greece, Ireland, Italy, Portugal, and Spain; the core economies are the other seven countries that orginally adopted the euro in 1999 (Austria, Belgium, Finland, France, Germany, Luxembourg, and the Netherlands). Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Currency valuation of all countries is estimated with reference to Germany. Source: Bhalla (2007a) dataset extended to 2011; see table B.1 in appendix B.

Figure 7.4

Does currency valuation matter for current account balance? Core and peripheral economies of the euro area, 1993–2011

difference in current account balance (percent of GDP), core/peripheral 2007

10

2006 2008

8

2010

2005 2009

6

2004

2011 2003

4

2002 2001 1999

2000

2

1998

0

1997 1993 1996

1995

1994

–2 –5

0

5

10

15

20

25

30

35

40

45

difference in currency valuation (percent), core/peripheral Notes: Euro area economies are aggregated into two groups. Peripheral economies are the five weaker economies, Portugal, Ireland, Italy, Greece, and Spain. Core economies are the seven “strong” economies of Europe, including Germany, France, and the Netherlands. All data are aggregated for the two groups; the valuation estimates are weighted by GDP. Source: Author’s calculations.

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 107

Whither the Dollar? A critical aspect of any solution to current account imbalances is a change in the value of the exchange rate. As emphasized, this is not a perfect recipe, since there are other determinants of imbalances, but the exchange rate carries a heavy burden. The US dollar is a reserve currency, and after the end of the dollar’s direct convertibility into gold in 1971, it is a difficult proposition for the United States to revalue or devalue. It has to “talk” the dollar down, suggesting and cajoling the countries with current account surpluses to allow their currencies to appreciate against the dollar. Not surprisingly—especially given the evidence presented in this book that an undervalued currency is good for growth—the surplus economies are loath to do so. When an irresistible force meets an immovable object, the result is a standstill. In economic parlance, that means an economic crisis. After the global crisis in 2008, and the euro area debt crises since then, there are justifiable cries that this should not be allowed to happen ever again. For this to come true, some adjustments will be necessary to prevent severe current account imbalances. William Cline and John Williamson have published annual estimates of currency change needed for an orderly transition toward a more balanced world since 2008. Cline and Williamson base their estimates on Williamson’s (1994) fundamental equilibrium exchange rate (FEER) model, which basically equates current account surpluses and currency appreciation. In contrast, my estimated currency valuations (Bhalla 2007a) yield different combinations for the same country. For example, both the Indian rupee and the Vietnamese dong are deemed to be undervalued, yet both these countries have reasonably large current account deficits. The Japanese yen and Swedish krona are both considered to be overvalued, yet both these countries have reasonably large current account surpluses. I propose a simple method to evaluate the essential elements of a new currency order—one that would likely lead to currency peace, not wars. The process uses data on current account balances and the Bhalla (2007a) currency valuation estimates and involves two steps. First, it is noted that the causal relationship between currency valuation and the current account balance is robust and stable. As shown in chapter 3, a simple regression between the two yields, for several five-year periods between 1965 and 2011, an average coefficient of –0.04—that is, each 25 percent level of currency undervaluation, other things equal, is equivalent to or consistent with a 1 percentage point increase in the current account balance. This equivalence allows for the combination of diverse data on the current account and currency valuation into a single unified measure of the current account balance for each country. This adjusted current account balance is equal to the existing balance minus 0.04 times the (log) percentage of the currency valuation (the more negative the valuation, the higher the expected balance). This adjusted balance basically levels the playing field and allows for the estimation of equilibrium exchange rates. There will be several paths for the 108

DEVALUING TO PROSPERITY

transition from a misaligned to an aligned state. Obviously, surplus countries will adjust their currencies upward, and deficit countries will adjust their currencies downward. This rule is explicit in the Williamson (1994) model, except for the fact that his FEER is based on actual surpluses whereas the recommendation here is that movement should follow the path of an adjusted surplus. Another departure from the FEER is that there is no explicit target for the current account balance; the only recommendation or conclusion is about the direction of the currency’s movement and its approximate magnitude. The second step involves estimating the magnitude of the required currency change given the scale of the adjustment needed in the current account. The recommended adjustment path (arbitrary but realistic and consistent with the underlying economics) is that all countries with a positive adjusted surplus should appreciate their currencies by (log) 20 percent or one-fifth of the adjusted surplus; all countries with a negative adjusted current account balance should depreciate their currencies by one-fifth. In both cases, the maximum adjustment is capped at (log) 35 percent. In mixed cases—for countries whose currencies are overvalued but also have an adjusted current account surplus, or whose currencies are undervalued and have a adjusted currency deficit—the currency change is capped at a maximum of plus or minus (log) 10 percent. Calculating the adjusted current account surpluses will indicate what changes will be required in the valuations of various currencies. Then it is possible to calculate the combined long-term effects of these currency changes on the US current account. This involves estimating the elasticity of US exports and imports. As noted, Baily and Lawrence (2006) and Thomas, Marquez, and Fahle (2008) obtain an elasticity of US net exports of about 1.15; my method yields a higher elasticity, 1.25. Here I adopt a conservative estimate of the elasticity, 1.15—that is, for each 1 percent appreciation of the foreign currency, US exports increase by 1.15 percent and imports decrease by 1 percent (the reverse occurs in the event of a depreciation). In 2011, the United States had a deficit of $720 billion in the trade of goods; China accounted for close to 40 percent of this deficit, or $290 billion. These 2011 magnitudes are put in perspective by noting that the past 1980s “culprits” Germany and Japan had a joint net surplus in 2011 of $47 billion and $58 billion, respectively—only a third of the amount of China alone. Table 7.7 shows in rank order countries with trade surpluses with the United States. Column 2 reports on the aggregate current account balance, and columns 3 and 4 document the actual and adjusted trade balance with the United States. Column 5 is the currency adjustment required using the method outlined above. This adjustment is capped at (log) 35 or 41.9 percent. Column 6 is the exchange rate as of September 30, 2011, and column 7 is the adjusted exchange rate.10 Columns 8 and 9 complete the exchange rate picture

10. The euro countries’ exchange rate is reported according to their individual currencies before their adoption of the euro in 1999 for the original 12 countries and thereafter for other euro area members.

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 109

110 DEVALUING TO PROSPERITY

Table 7.7 Toward an adjustment of the US dollar, 2011

Country (1)

Current account balance (percent of GDP) (2)

Trade balance with the United States (billions of US dollars)

Actual (3)

Adjusted (4)

Exchange rate per US dollar

Currency adjustment required (percent) (5)

Actual as of September 30, 2011 (6)

Equilibrium exchange rate per US dollar

Adjusted (7)

Bhalla (2007a) (8)

Cline and Williamson (2011) (9)

3.69

5.09

China

5.2

290

79

41.9

6.56

4.62

Mexico

–1.0

66

88

–4.5

11.80

12.40

14.80

11.60

Japan

2.5

58

52

2.9

80.40

78.10

112.75

76.00

Germany

5.0

47

32

10.1

1.38

1.25

1.85

0.67

–3.3

35

102

–10.4

0.97

1.08

1.27

0.94

Italy

–3.5

18

24

–11.5

1,370.00

1,548.00

1,941.39

0.67

Taiwan

11.0

15

–15

41.9

29.00

20.40

17.11

22.8

India

Canada

–2.2

15

18

–4.8

47.30

49.80

35.13

41.00

Thailand

4.8

14

–1

38.3

32.30

23.40

20.94

27.00

Korea

1.5

13

–3

15.7

1,082.00

935.0

750.04

979.00

Indonesia

0.2

12

12

–0.5

8,674.00

8,717.00

9,553.67

7,554.00

Malaysia

11.3

11

–7

41.9

3.36

2.37

2.34

2.35

France

–2.7

10

17

–9.8

4.64

5.14

6.68

0.67

Israel South Africa

0.3

9

9

–0.7

3.68

3.71

4.26

3.29

–2.8

2

3

–7.4

6.97

7.53

7.57

7.38

Philippines

1.7

1

1

3.5

45.8

44.30

49.56

38.2

Argentina

–0.3

–5

–7

8.7

4.16

3.83

2.56

4.00

Chile United Kingdom Turkey

0.1

–6

–6

0

474.00

474.00

480.57

454.00

–2.7

–6

4

–8.5

0.61

0.67

0.76

0.58

–10.3

–10

–1

–39.6

1.60

2.70

1.69

1.98

Singapore

19.8

–12

–35

41.9

1.24

0.88

1.10

0.90

Brazil

–2.3

–13

–5

–10.6

1.62

1.81

2.78

1.65

Australia

–2.2

–16

–12

–9.8

0.95

1.05

1.55

0.98

5.4

–31

–50

41.9

7.79

5.49

4.76

5.92

Hong Kong

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 111

Notes: All computations are based on average data for 2011; for example, in 2011 the average value of the euro was 1.41 dollars for 1 euro. For the EU economies, currency undervaluation is based on the exchange rate benchmarked to January 1999, or the time of euro adoption. Sources: Bhalla (2007a) dataset extended to 2011; Cline and Williamson (2011); author’s computations.

by reporting on the target or equilibrium values of the exchange rate according to my method and that of Cline and Williamson (2011), respectively. The two approaches have the same goal—identifying the set of exchange rate changes that will help reduce the US current account toward an equilibrium level. Among nonoil exporters, five countries appear to be in need of the most (maximum) appreciation—China, Taiwan, Malaysia, Singapore, and Hong Kong. This method leads to several broad observations about the US current account. First, as noted above, the US dollar was fairly valued in 2011, but the United States still had a trade deficit of more than $500 billion. Second, if the target changes do take place, the US dollar will depreciate about 6.3 percent from 2011 levels. This should lead to an improvement in the current account of about $220 billion, to a level of around $280 billion, which in 2011 prices would be about 1.5 percent of GDP. In sum, with the recommended adjustments, the US current account deficit would be a very reasonable 2 percent of GDP.11 The above analysis is based on direct calculation of the currency adjustments required for various economies. In the past, the dollar has not necessarily moved according to the calculus outlined here. According to equation 7.1, a coefficient of around 0.15 for currency change should be expected. Given the magnitude of the constant term, about –3.75, an 8 percent change in dollar value would translate into a current account deficit of (–3.75 + 0.15 × 8), or about 2.6 percent of GDP—almost 1 percentage point higher than the targeted currency change. In other words, a targeted currency change is likely to lead to a larger change in the US current account deficit. This is not entirely surprising. What was appropriate for a US current account balance in 1985 or even 1995 is different from what is appropriate today. The dollar will likely follow a different trajectory, perhaps even including an appreciation against the yen and the euro, and a depreciation against the currencies of the emerging-market economies, especially in Asia. Let us briefly consider the appropriate adjustments for some specific currencies. The renminbi, at 4.7 yuan to $1, is nearly identical to the Cline and Williamson (2011) estimate.12 As expected, Asian emerging-market economies are in need of the greatest exchange rate adjustments. The yen does not need to change much because Japan’s share of the US trade imbalance, or of world trade in general, is not as high as it used to be. But overall the yen should depreciate, because it is presently overvalued. The euro also needs a large adjustment; 11. In 2009, a recession year, the US current account deficit was $370 billion, or 2.6 percent of GDP. This occurred despite an appreciation of the dollar by 6 percent over 2008 levels. This suggests that the net export elasticity assumed above (elasticity of 1.15) is reasonable. 12. The differences with Cline and Williamson (2011) arise with respect to Japan and Korea, among others. I estimate the equilibrium value for the yen at 105, while Cline and Williamson estimate 76. For Korea, my model estimates an equilibrium of 742, while Cline and Williamson estimate a more depreciated level of 979.

112

DEVALUING TO PROSPERITY

it was overvalued by more than 30 percent in 2011. The required adjustment is an aggregate 4 percent devaluation—but note that it is different for different European countries.13 For example, for Greece the valuation change is from a 28.4 percent overvaluation to a 2.5 percent undervaluation; for Germany, from 33.4 percent overvaluation to 42 percent overvaluation (recall that the direction of the adjustment is dictated by the size of the adjusted surplus and so this would be a 9 percent appreciation). For the euro, Cline and Williamson recommend a movement from 1.39 to 1.53, or a 10 percent appreciation, which suggests that the Cline and Williamson estimate is influenced by the stronger countries in the euro area. It is also worth noting that the equilibrium real exchange rate for the euro is observed to be very close to parity, based on the observation that when the euro was at 1.31 in March 2011, it was 31 percent overvalued. However, as noted above, the euro is a composite of a diverse group of nations and what is equilibrium for Germany is not equilibrium for Greece. The currencies in need of the greatest adjustment are the currencies of the Asian emerging-market economies, specifically, China, Hong Kong, Malaysia, Singapore, and Taiwan. These all require adjustment of at least 42 percent. Excluding China, these currencies have been deeply undervalued for a long time but have not caused a global imbalance problem. China is different, because of the size of its worker population and its extraordinary growth rate. What the analysis reveals is that an appreciation of the Asian currencies in general, and the Chinese renminbi in particular, is absolutely essential for redressing both the US current account imbalance and the global imbalances.

13. The euro countries’ exchange rate is reported according to their individual currencies before 1999, e.g., Deutsche mark and French franc.

RASHOMON RULES: US DOLLAR, EURO DOLLAR… 113

8 Currency Valuation and Growth

Cause and effect, means and ends, seed and fruit cannot be severed; for the effect already blooms in the cause, the end preexists in the means, the fruit in the seed. —Ralph Waldo Emerson, on Compensation, Essays and English Traits This chapter examines the empirical relevance of currency valuation to growth for various countries over time, using various econometric methods, and as affected by the presence of institutions. It is worth reviewing the findings of the analysis so far. Currency valuation estimates are able to explain investment behavior over the past 50 years for a broad range of countries. Somewhat unexpectedly, very indirect estimate of the US dollar (Bhalla 2007a) explains most of the variation in the US current account, performing as well as the direct WARP estimate and better than the official estimate, the US Federal Reserve’s Broad Index. My measure of euro currency valuations also robustly explains the differential performance of current account balances in the euro area economies since the adoption of the common currency in January 1999. In fact, these results show that the adoption of a single currency and the overvaluation of the euro after 2005 created deep problems for the poorer economies on the periphery, problems that may have contributed in large part to the euro area crisis of 2010–11. The ability of currency valuations to explain these current accounts is only an intermediate output, however. The big question is still whether currency misalignments play a role in affecting the fortunes, and misfortunes, of nations. This is the subject of the next several chapters. As mentioned throughout, the question is not exactly new. For almost 20 years, this has been the focus of formal econometric testing by scholars and by institutions such as the International Monetary Fund. The findings have been mixed, at best. The profession is willing to accept the conclusion that currency overvaluation hurts growth. It is loath to accept the opposite, that currency undervaluation helps growth. In other words, it is not acceptable to assert that countries can devalue their way to prosperity. 115

As a consequence, researchers moved away from the question of whether currency undervaluation helps growth and searched for other explanations. This (re)search seems to have confirmed that a primary reason some countries are rich and many others are poor is because the rich countries had “better” institutions, specifically Western-style institutions. If the presence or absence of such institutions can explain differential living standards and growth rates for a broad range of countries, across time and geography, then we can reasonably consider the growth “puzzle” solved. This is the task of chapter 11. In the meantime, this chapter is concerned with a comprehensive examination of the empirical relevance of currency valuations.

Overview Let’s first briefly recap the conclusions in the literature, perhaps with a bit of repetition. The purpose is to provide a background perspective for the somewhat striking results of this analysis, which confirm a very strong relationship between currency valuation and growth. The literature is replete with discussions of how currency overvaluation hurts growth, but what little exists about how undervaluation helps growth is tentative and contentious. In addition, there is much less econometric testing of the latter hypothesis than appropriate for such an admittedly controversial and important proposition. Furthermore, the existing evidence is suspect because of the presence of measurement errors. There is more of a need for tests that affirm the relationship between currency valuation and growth than for evidence about the accuracy of various measurements or estimates of currency valuation. This chapter aims to fill some of the gaps. It does so by conducting several tests of the significance and robustness of various currency valuation methods. It also assesses various growth models for all countries and for only developing countries, over both short and long time periods. The chapter includes some tests of symmetry: If currency valuation helps growth by a certain amount, then is it the case that overvaluation hurts growth by the same amount? The existence of such symmetry would help prove the importance of currency valuation. Additional tests explore whether deep undervaluation (overvaluation) helps (hurts) more than standard magnitudes of currency valuation, or whether there are diminishing returns. Finally, the chapter conducts some stringent robustness tests on the model that demonstrates that countries can undervalue their way to prosperity. This chapter excludes several topics that are covered elsewhere in the book: the period from 1870 to 1938 (chapter 12) and the role of institutions in generating growth and in shaping currency valuation policy (chapter 11).

116

DEVALUING TO PROSPERITY

Econometric Models of Growth Lacking sufficient natural experiments to serve as growth laboratories, economists have adopted a multipronged framework for analysis that includes such methodologies as case studies of individual countries, historical analysis as far back as the 19th century, and cross-section panel data analysis for developed and developing economies. The goal is to rigorously assess all the determinants of growth that are mentioned in the literature, from democracy and demography to openness and fiscal deficits, and to evaluate them individually and simultaneously.

A Benchmark for Analysis The various models can be evaluated within one simple benchmark model of growth. Following the convergence model of Robert Barro and Xavier Sala-iMartín (1992), the basic model relates per capita growth to (log) initial income. For example, if the period under investigation is 1980 to 2011, the variable to be explained is average rate of growth during this 32-year period. However, this growth rate depends on how wealthy the particular country was in the base or initial year, 1980. If the country were already quite rich (for example, the United States), then the prospects for high growth rates would be dim. If the country were poor (for example, China), the prospects for high growth rates would be bright. Why? Because China could borrow technology and expertise from the United States (and other advanced economies), while the United States would have to invest and discover new technologies in order to grow. This initial disadvantage is modeled by using the (log of) initial income per capita in 1980 (the base year) as an important determinant of growth for the subsequent period, 1981–2011. This specification from Barro and Sala-iMartín (1992) is the econometric procedure for incorporating the important phenomenon of catch-up.

Levels of Econometric Sophistication Developing growth models has been something of an industry, and so there is now broad consensus on the variables with a legitimate claim to be considered as possible determinants of growth. One way to test the robustness of currency valuation variables is to assess their importance when almost all these other variables are included in the regression equation. Two papers are useful in this regard: The first, by Barry Bosworth and Susan Collins (2003), is an exhaustive study of growth determinants; the second, by Eswar Prasad, Raghuram Rajan, and Arvind Subramanian (2007), extends the Bosworth and Collins specification by adding the current account surplus (as a percent of GDP) to the list of the right-hand-side variables. Addition of the current account surplus is controversial.

CURRENCY VALUATION AND GROWTH 117

The list of growth determinants in Bosworth and Collins (2003) includes the usual suspects: convergence variables; factor inputs in the form of education; exogenous variables such as share of worker population; and policy variables such as the degree of openness in the economy, fiscal deficits, and institutions.

Data and Sample Selection The approach taken here is to assemble for a large group of countries1 the data on real income per capita and the real exchange rate between 1950 and 2011. To these, subject to availability, information is added on other determinants of growth such as fiscal deficits, investment shares, dependency ratios, life expectancies, and more. This cross-country panel dataset (Bhalla 2007a) is one of the largest assembled to explore questions about economic growth. If this large dataset yields robust findings, then currency valuation can be considered to be a major determinant of economic growth. Appendix A describes the definitions, sources, and methods for the data collation. Not all the data are available for all countries for all years between 1950 and 2011 (data are more complete from 1960 onward). The primary structure of the analysis is a panel of countries for five years. Data for longer periods are sometimes analyzed (for example, for the entire 62 years, or for the globalization period, 1980–2011). (Appendix table A.1 documents the country composition.) There are three levels of analysis, which can be considered tests for statistical significance at increasing levels of difficulty. Ordinary least squares (OLS) regressions allow even weak relationships to appear robust. This estimation technique is eschewed by most but undertaken by all as a reference point. The analysis moves from OLS estimation to instrument variables (IV) regressions— two-stage least squares. IV regressions are necessary to control for simultaneity, and economists have been quite creative in assembling identification variables. For example, to estimate a simultaneous relationship between income and institutions, variables that affect institutions but not the level of income can be classified as instruments for institutions. Several variables significant in OLS regressions fail at the IV threshold. The third level of tests is dynamic panel estimation, a method pioneered by Manuel Arellano and Stephen Bond (1991). This method, which fills the important gap when an identification variable is unknown, uses the lagged values of the levels and the first differences of time-dependent variables as instruments. The use of lagged values makes it very difficult statistically for variables to pass tests of significance. Therefore, a variable that does pass the Arellano-Bond test is likely to be extremely robust—in this case, a likely determinant of growth. 1. Data are gathered for 180 countries, and these are narrowed to about 130 for the analysis. See appendix B for a list of the sample countries.

118

DEVALUING TO PROSPERITY

Tests of Simple Growth Models The various regressions reported throughout this book use the basic Occam’s razor principle, which states that the simplest explanation is preferred.2 The basic model has as the determinants of growth for each five-year period only the log of the initial income per capita and the initial value of currency valuation. The lagged value of the average change in currency valuation is added to this basic model. In addition to these three determinants, time and country “dummy variables” are included. This model is estimated for different definitions of the dependent variable, for different samples of countries, and so forth. The time framework is as follows. Country data are aggregated for fiveyear periods starting in 1950 (1950–54 is the first period, 1955–1959 the next, and so on). The last period actually comprises seven years, 2005–11. Given the need to account for initial values of currency valuation, and the avoidance of construction bias, each five-year period has the initial values for the first year, with the growth rates of income per capita for the subsequent four years. In models where the dependent variable is other than growth in income per capita (e.g., savings or investment as a share of GDP), the average over the entire period is used since there is no construction bias link with the dependent variable. This is best described using a detailed example. Assume that the five-year period under consideration is 1980 to 1984. The initial value of income per capita and currency valuation is for 1980. Average growth in income per capita (the dependent variable) and the average change in currency valuation are for the four years 1981–84. The one-period lag on average change means that in a regression explaining growth for 1981–84, the average change is for the four years 1976–79. Not representing the change in currency valuation by its lagged value would mean introducing a construction bias since growth between 1981 and 1984 would be regressed on the change in valuation between 1981 and 1984.

Role of Real Income Several estimates are available for purchasing power parity (PPP) income. The variations among them primarily concern changes in the base year of the PPP survey. The data from the United Nations’ International Comparison Program (ICP) surveys are processed and improved upon and released under the banner of Penn World Tables. At present, there are four separate Penn World Table estimates of income per capita. The first comprehensive estimate used 1996 as

2. The principle is that “entities must not be multiplied beyond necessity.” Occam’s razor is also known as the principle of parsimony: “When competing hypotheses are equal in other respects, the principle recommends selection of the hypothesis that introduces the fewest assumptions and postulates the fewest entities while still sufficiently answering the question” (http://onlinephilosophyclub.com/occams-razor.php).

CURRENCY VALUATION AND GROWTH 119

the base year (Penn Table 6.1); the second was for the survey year 2000 (Penn Table 6.2); and the third was based on the ICP survey for 2005 (Penn Table 6.3). Penn Table 7.0, released in May 2011, also uses 2005 as the base year but updates the data through 2009. Controversy persists with respect to the ICP Survey for 2005, which is problematic because it somewhat inexplicably estimates the Chinese price level in 2005 to be more than 40 percent higher than the aggregate price level previously assumed. This translated into a GDP level estimated at 40 percent below the estimates prevailing the day before the ICP data were released! Such a wide discrepancy looks even more unlikely when juxtaposed against the corresponding estimate for the Indian price level, which was (possibly not coincidentally) 40 percent higher than the prevailing estimates (with GDP 40 percent lower). This coincidence led several scholars to question the accuracy of the ICP data for 2005.3 So, the ICP authorities undertook a new survey in 2011. Until the results of this new ICP survey are available, possibly in 2013, international organizations such as the IMF and World Bank have decided to accept as accurate the ICP 2005 data. The immediate effect, as mentioned, is that China’s GDP is estimated at 40 percent lower. According to the Penn World Table 6.1, which is used in this book, China’s PPP GDP exceeded that of the United States in 2007. According to the new method, that day of reckoning is shifted to about 2016. According to the new PPP data, in 2011 the gap between US and Chinese income was about PPP$4 trillion—with the United States at PPP$15 trillion and China at PPP$11.3 trillion. With the gap standing at about 25 percent, an excess growth in income per capita each year of 5 percent for China would bring its income into convergence with US income in about five years. Table 8.1 shows the deep divergence between the 1993 and 2005 ICP surveys for selected countries. The currency valuation for China declines with the drop in its GDP estimate. The currency valuation is –9 percent in 2011 according to the 2005 ICP, compared with –44 percent in the 1993 ICP. At face value, this suggests that there is no Chinese currency undervaluation problem, nor has there ever been! For the Brazilian real, it was undervalued by more than twice as much as the renminbi in 2005—and with a real exchange rate of 2.4 to $1!

Test 1: Overall Results The results are strikingly robust. No matter what definition is used for the real exchange rate or what the sample selection, the currency valuation coefficients are strongly significant. Furthermore, the magnitude of the coefficients changes little when other determinants are added to the equation.

3. Bhalla (2008b) extends these data back to the 1950s and concludes that if the ICP 2005 data were right, then all Asians would have been dead in 1950! Angus Deaton and Alan Heston (2010) also find the data problematic.

120

DEVALUING TO PROSPERITY

Table 8.1

Comparison of currency valuation estimates for selected countries, 1970–2011 (percent) 1993 ICP Survey

Country

1970

1985

2005

2005 ICP Survey 2011

1970

1985

2005

2011

Developing economies Brazil

15

–25

–3

72

4

–32

–22

39

China

477

106

–42

–44

1,100

386

–10

–9

India

208

131

–23

–26

345

226

14

8

Korea

27

–4

–19

–31

14

–15

–21

–24

–16

–16

31

34

–38

–36

12

20

Developed economies Germany Japan

–33

–5

25

40

–35

–10

19

34

United Kingdom

–29

–19

30

24

–38

–30

19

12

5

9

0

–7

16

20

7

–6

United States

ICP = International Comparison Program Notes: For each study, currency valuation is the percentage deviation for real exchange rate from its predicted value. A negative value indicates undervaluation; a positive value indicates overvaluation. See text for details of construction of the estimates in each survey. Source: Bhalla (2007a) dataset extended to 2011.

More than 40 regression results are reported using this basic growth model in tables 8.2 through 8.6. The first set of regressions reported in table 8.2 (panel A) is for models without a construction bias (following Woodford 2009) in the measurement of the variables. The next set of regressions (panel B) follow the conventional specification with the biases present—that is, using the mean currency valuation (not initial) and the change in valuation (not lagged). The coefficients of the misspecified construction bias model can be considered an upper bound to reality. For the complete sample of more than 180 countries, using the average currency valuation, the model has a coefficient of –0.008. That means that for each 10 percent of initial valuation (the negative sign indicates undervaluation; a positive sign would indicate overvaluation), there is an extra 0.08 percent of growth per capita. Excluding small countries produces a somewhat larger coefficient, and the final sample of 130 (excluding small countries and countries with problematic data) has a coefficient for initial currency valuation of –0.012. The regressions that include the construction bias—the model commonly estimated in the literature—are reported in panel B of table 8.2. The difference between the two panels is as follows. In models with no construction bias, both growth in income per capita and the change in currency valuation are variables measured for the four years following the initial year, which is the year for which initial currency valuation is reported. The change variable is lagged one period. In models with a construction bias, all variables (except CURRENCY VALUATION AND GROWTH 121

122 DEVALUING TO PROSPERITY

Table 8.2

Estimating the impact of currency valuation on the growth of income per capita, 1950–2011 Coefficients

Row

Sample selection

Initial income per capita (log)

Currency valuation Initial

Average change

Adjusted R2

Number of countries

Number of observations

Panel A: Without construction bias 1

All sample

–3.42***

–0.008***

–0.035***

0.28

181

1,376

2

Population > 1 million

–3.39***

–0.01***0

–0.024**0

0.42

148

1,159

3

Selected sample

–3.25***

–0.012***

–0.032***

0.45

133

1,074

4

Selected sample with change in valuation not in equation

–3.25***

–0.011***

0.34

134

1,168

5

Row 3 excluding Russia and Eastern Europe

–3.19***

–0.013***

0.43

107

985

–0.031**0

Panel B: With construction bias 6

All sample

–2.73***

–0.018***

–0.14***0

0.43

182

1,510

7

Population > 1 million

–2.49***

–0.017***

–0.12***0

0.45

149

1,267

8

Selected sample

–2.44***

–0.019***

–0.09***0

0.45

134

1,168

9

Selected sample with change in valuation not in equation

–2.72***

–0.021***

0.35

134

1,168

10

Row 8 excluding Russia and Eastern Europe

–2.52***

–0.021***

0.46

107

1,072

–0.11***0

Notes: Data are for a panel of countries, for five-year periods between 1950 and 2011. Dependent variable is growth in income per capita. All regressions have time and country dummies and log intial income per capita. Average change in currency valuation is inserted as a one-period lag in the panel A regressions. Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Source: Bhalla (2007a) dataset extended to 2011.

initial income) are represented by their means. Interestingly, the coefficient of interest, currency valuation, has approximately the same coefficient in both models. But the change in currency valuation has a larger coefficient in the biased model, about –0.09 compared with –0.03 in the correctly estimated model. Rows 4 and 9 compare the models without any change in currency valuation. This is the model estimated in the literature. The results show that the presence of construction bias nearly doubles the impact of currency valuation—a coefficient of –0.021 compared with the no-bias estimate of –0.011. There is no difference in statistical significance. The strong result from this preliminary analysis is that currency valuation is an important and significant determinant of growth. The coefficients of both exchange rate valuation variables are very strongly significant. The coefficient of the initial valuation is robustly centered around –0.01—that is, each 10 percent valuation adds 0.1 percent of extra per capita GDP growth per year. The coefficient for the average long-term change in currency valuation is robustly centered around –0.03. It is likely that introduction of other determinants of growth into the basic model reduces the contribution of currency change; we turn now to examine whether this occurs, as well as the importance of these various determinants.

Test 2: Cross-Section Panel Data, Two-Stage Least Squares Another test is to estimate the model for long periods. Table 8.3 presents the results for cross-country panel data for three such periods, 1960–79, 1980– 2011, and 1960–2011. Estimation of such models reintroduces the problem of construction bias for the average change in currency valuation. Use of lagged values is no longer a solution; two-stage least squares estimation is preferred. Following the model by Andrew Rose and Saktiandi Supaat (2007), the fertility rate is used to instrument for the change in currency valuation. The assumption is that the larger the decline in fertility, the larger the move toward currency undervaluation. First-stage results confirm this expectation. Another possible instrument is the change in reserves as a share of GDP. Regardless of the instrument chosen, the results are robust. Initial valuation has a coefficient between –0.01 and –0.02, and the change in valuation has a coefficient between –0.4 and –0.7. The final column in the table reports an acceleration model for the two periods. The dependent variable in this regression is the change in the growth rate of income per capita for the two periods; the other variables are changed accordingly. Data for both periods are available for 97 countries. The results are nearly identical to the two-stage least squares results, with the coefficient of the initial currency valuation at –0.024 and the coefficient for the change at –0.377.

CURRENCY VALUATION AND GROWTH 123

124 DEVALUING TO PROSPERITY

Table 8.3 Two-stage least squares regression results Instrument variables Fertility rate

Change in reserves (percent of GDP)

1960–79

1980–2011

1960–2011

1960–79

1980–2011

1960–2011

Acceleration model

Initial currency valuation

–0.005

–0.022***

–0.021***

–0.001

–0.014***

–0.011***

–.024***

Change in currency valuation

–0.35**

–0.7***

–0.69***

–0.150

–0.46***

–0.35***

0.24

0.38

Independent variable

Adjusted R2 Number of observations

96

135

0.13 135

0.130

0.43

91.00

134.00.00

0.27 134.00

–.377*** 0.51 97.00

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Rate of growth in income per capita is the dependent variable and change in currency valuation is instrumented, first with the fertility rate and second with change in reserves as a percent of GDP. The final column reports the results for acceleration in growth (growth during 1980–2011 minus growth during 1960–79) with the corresponding acceleration in the independent variables, namely, the difference of initial currency valuation in the two periods and the difference in average change in currency valuation. Source: Bhalla (2007a) dataset extended to 2011.

Test 3: Dynamic Panel Estimation Tests As noted throughout, currency valuation variables pose problems related to construction bias, endogeneity, and/or omitted variables. Correlations can be built in both directions by the underlying dynamics of growth. For example, an argument can be made that a devaluation (a lowering of the real exchange rate) is a response to certain underlying macrophenomena that are bad for growth. For example, higher inflation can lead to overvaluation under a fixed exchange rate regime, and this would lower growth. By setting the prices “right,” a devaluation could enhance growth in the future and thus yield the correlation between a lower currency valuation and higher growth. In contrast, Balassa-Samuelson considerations suggest a positive relationship between currency appreciation and growth: As a country becomes richer, its price level rises, thereby raising its real exchange rate. How can these divergent possibilities be handled econometrically to model an average “genuine” effect? The generalized method of moments (GMM) pioneered by Arellano and Bond (1991) allows for the efficient estimation of the impact effects of variables that are potentially endogenous, and it does so without recourse to finding appropriate instruments for the endogenous variables. The Arellano-Bond method allows all determinants to be endogenous by using the lagged values of time-dependent variables. Table 8.4 contains the results for each definition of PPP income. As discussed, the 2005 ICP Survey affects the level of currency valuations, but, strikingly, it does not affect their relationship with growth. Two sets of results are shown, one for fixed effects models and one for models based on the Arellano-Bond method. Two conclusions emerge. First, the definition of income does not matter to the results—currency valuation and the average change in valuation both remain strongly significant. Second, the more accurate Arellano-Bond method shows a larger effect; the coefficient on initial valuation is two to three times higher in absolute magnitude (about –0.02 to –0.03), and the coefficient on change is about three to four times higher (about –0.07 to –0.12).

Test 4: Effect of Other Determinants In cross-country regressions, the magnitude and significance of a variable often are dependent on what other determinants are included. One configuration may suggest that a particular variable (say, currency valuation) is important, and another may suggest that it is not. This makes it important to conduct sensitivity tests before asserting the importance of any determinant. One such test is to include several determinants in the same regression. The regressions above have constantly tested for bias, and the results are encouraging: All the tests suggest that currency valuation and changes in currency valuation are both strong determinants of growth. The results for tests including additional determinants are reported in table 8.5. As always, the right-hand variables are CURRENCY VALUATION AND GROWTH 125

Table 8.4

Currency valuation and growth: Different econometric methods Initial currency valuation

Change in currency valuation

Bhalla (2007a)

–0.018***

–0.094***

753

Penn Table 6.2

–0.006***

–0.074***

753

Penn Table 6.3

–0.029***

–0.116***

753

Penn Table 7.0

–0.013***

–0.016***

753

2005 ICP Survey

–0.027***

–0.072***

735

–0.019***

–0.074***

Model specification

Number of observations

Arellano-Bond models (selected sample)

Average Fixed effects (selected sample) Bhalla (2007a)

–0.012***

–0.032***

1,075

Penn Table 6.2

–0.009***

–0.025***

1,074

Penn Table 6.3

–0.009***

–0.023***

1,075

Penn Table 7.0

–0.01****

–0.02****

1,074

2005 ICP Survey

–0.009***

–0.039***

1,057

–0.01****

–0.028***

Average ICP = International Comparison Program

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. For the fixed effects model, the change in currency valuation is entered as a one-period lag. The Arellano-Bond (1991) model is a consistent generalized method of moments (GMM) estimator for the parameters of this model. Source: Bhalla (2007a) dataset extended to 2011.

log of initial income per capita and time and country dummies. Row 1 is the basic model with the two currency valuation variables; row 2 adds life expectancy; row 6 adds the initial worker population ratio to the model reported in row 2, and so on. Each variable is added according to its initial value for each five-year period.4 Several of the additional variables are statistically significant as determinants of growth, and several are not. The nonsignificant variables include life expectancy, education, and fiscal deficits. The significant variables include the initial worker population ratio, initial middle class, initial fraction of trade, financial openness, and three of the six components of the World Bank governance index. The last regression includes three prominent determinants, trade, financial openness, and demography. The presence of the additional determinants does not diminish the relevance of currency valuation—at least one of the

4. Inclusion of the mean for these variables would be improper because for many of them (for example, fiscal deficits, trade, and current account balance) there is likely to be an interdependence with growth.

126

DEVALUING TO PROSPERITY

Table 8.5

Determinants of growth, fixed effects model, 1950–2010 Coefficients Change in currency valuation (lagged)

Row

Model specifications

Initial currency valuation

1

Select sample 133 countries

–0.012***

–0.032***

2

Log (life expectancy)

–0.012***

–0.029***

1.58

0.37

1,009

3

Mean education years

–0.013***

–0.026***

–0.152

0.35

1,003

Added variable

Adjusted R2

Number of observations

0.36

1,075

4

Share of middle class in population (percent)

–0.011***

–0.03***0

0.017*

0.37

1,009

5

General government deficit (percent of GDP)

–0.01***1

–0.028***

0.041

0.33

555

CURRENCY VALUATION AND GROWTH 127

6

Initial share of population 15–64 years (percent)

–0.01****

–0.035***

0.191***

0.39

1,015

7

Financial openness, Chinn and Ito (2008) index

–0.019***

0.003***

0.223**

0.42

787

8

Initial share of trade (percent of GDP)

–0.011***

–0.028***

0.039***

0.38

958

9

World Bank governance indicator 1

–0.007

–0.068***

1.04*

0.48

382

10

World Bank governance indicator 2

–0.006

–0.068***

2.51***

0.50

382

11

World Bank governance indicator 3

–0.006

–0.07***8

1.07***

0.49

382

12

World Bank governance indicator 4

–0.006

–0.07***8

0.54

0.47

382

13

World Bank governance indicator 5

–0.005

–0.07***8

0.57

0.47

382

0.99

0.47

382

0.43

768

14

World Bank governance indicator 6

–0.004

–0.07***8

15

Add demography, openness, and trade

–0.017***

–0.0008**

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Dependent variable is growth rate of income per capita. Sample includes 133 countries. Five-year panel data for 1950–2011. All models include (log) initial income per capita and time and country dummies. Source: Bhalla (2007a) dataset extended to 2011.

two valuation variables is significant in every equation. The most surprising, and encouraging, result is the robustness of the coefficients and statistical significance of the currency valuation variables. No matter what the specification, the coefficient of initial undervaluation is “stuck” at about –0.015.5

Test 5: Importance of Outliers or Extreme Observations Country selection does not matter; the period does not matter; the regression technique makes no difference. What remains to be tested before we can accept the conclusion that currency valuation is an important determinant of growth across all countries? We must test whether outliers or extreme observations may be driving the results, as concluded by William Easterly (2005). He found that a few extreme observations or outliers negated the significance of currency valuation in his models of growth, and his conclusion has been widely accepted. This makes testing the importance and relevance of outliers very important. The first test of outliers is to eliminate the extreme observations, by excluding the top and bottom 1 and 5 percent of the initial currency valuations. The second and third tests involve estimating the impact of initial currency valuation in narrow and broad ranges. These results are reported in table 8.6. No matter what the test, the result is the same in statistical significance and magnitude. Figure 8.1 illustrates the frequency distribution of initial undervaluation and its average change for developed and developing economies. The estimates seem to be broadly, evenly, perhaps even normally distributed, especially for the change in currency valuation. The residual plot reported in figure 8.2 suggests that the relationship between currency valuation and economic growth is systematic and pronounced. It is also revealing that the extreme observations on either side of the mean (the fitted line) form parallel lines with the center. This suggests that overvaluation is likely to hurt as much as undervaluation is likely to help. The relationship is symmetrical.

Test 6: Evaluating Regional Trends Table 8.7 reports on the averages for the different regions and several selected countries during the globalization period, 1980–2011. This is a descriptive test. Note the close correspondence between growth and currency valuation, especially for developing economies. Several of the fast-growing Asian economies show high rates of decline in the valuation measure; this increasing undervaluation is one reason for their higher-than-average (“miracle”) growth.

5. Interestingly, this estimate, between about 0.01 and 0.02, is nearly identical to the estimate of mean undervaluation found by most authors in studies of the effects of currency misalignment on growth.

128

DEVALUING TO PROSPERITY

Table 8.6

Do outliers matter? Coefficients of currency valuation Initial

Change (lagged)

Dummies of initial undervaluation < –10

–0.014***

> –10 and < 10

–0.033***

> 10

–0.011***

–0.033***

Exclusion of outliers At 1 percent

–0.012***

–0.034***

At 5 percent

–0.009***

–0.041***

Percentiles of initial undervaluation 1 (0 to 10)

–0.01

2 (10 to 25)

–0.023**

3 (25 to 75)

–0.01***

4 (75 to 90)

–0.013***

5 ( > 90)

–0.011***

Average of initial undervaluation

–0.033***

–0.023***

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. All regressions have time and country dummies and log initial per capita income. Sample includes 133 countries. Five-year panel data for 1950–2011. Currency valuation data organized according to the percentile (ranked from most negative [undervalued] to most positive [overvalued]) for each five-year period. Source: Bhalla (2007a) dataset extended to 2011.

Seven countries of the Association of Southeast Asian Nations (ASEAN-7) are in a cluster, with a low initial valuation (average of 30 percent) and a high negative rate of change (–2.4 percent per year).6 A slow-growing East Asian economy, the Philippines, has the same average level of initial valuation (30 percent), but a much smaller rate of improvement—an average of –1 percent a year—compared with the average for the ASEAN-7 (–2.4 percent) or the average for China (–7.5 percent). The different rates of change from comparable initial levels may explain most of the difference in the growth performance of Southeast Asian economies: The Philippines’ average growth rate of GDP per capita is 1 percent over 31 years, 1980–2011, and this lags far behind the 4.7 percent rate achieved by the ASEAN-7.7 6. The ASEAN-7 economies are Hong Kong, Indonesia, Korea, Malaysia, Singapore, Taiwan, and Thailand. 7. According to most models of growth, the Philippines should have participated in the economic miracle, with its very high levels of education, English-speaking population, and high female labor force participation rate. What happened? The answer is currency overvaluation. The East Asian

CURRENCY VALUATION AND GROWTH 129

Figure 8.1

Distribution of currency valuation, 1950–2011 (log percent) a. Developed countries

density

300

250

200

150

100

50

0 –215

–137

–59

19

97

175

253

currency valuation density 400 350 300 250 200 150 100 50 0 –48

–36

–24

–12

0

12

24

36

48

average percent change in currency valuation

130

DEVALUING TO PROSPERITY

Figure 8.1

Distribution of currency valuation, 1950–2011 (log percent) (continued) b. Developing countries

density

250

200

150

100

50

0 –109

–64

–19

26

71

116

161

206

251

currency valuation density 1,000

800

600

400

200

0 –47

–35

–23

–11

1

13

25

37

49

average percent change in currency valuation Notes: The range for the level of currency valuation for both developed and developing countries is between –250 and 250; the range for the average percentage change in currency valuation is between –50 and 50. A total of 68 country-year observations lying outside the ranges indicated are exluded from the figures.

106

Source: Bhalla BOOK TITLE(2007a).

CURRENCY VALUATION AND GROWTH 131

Figure 8.2

How does currency valuation impact growth?

marginal impact on growth 15

omn

10

5

0

–5

–10

irl arg

bwa jpn hkg hkg bdioan jor sgp kor hkg oan sgp fin sgp sgp kor sgp jpn tun nor lbn jpn blr usa arm hkg sgp sleciv cog usa isr fin mar deu civ oan sgppng nld mar koroan tto aus eth lva ita can bol prt esp chn nor isr can dnk nzl dnk can aus swe prt bra nor itausa gab aut chl gbr usa che mus irl hnd hnd aus bel deu ven aut tto dnk bdi fra isr jpn fra usa swe jpn sweche tha deu ury fin svn usa dnk fra bel swe oan aut esp bel aus isr jpn bwa fin irl grc grc gbr fra esp nor jpn tto tto nld uae prt uzb mys dnk aut nld gab nor aut che esp tto usa gtm kaz hkg can che deu swe tha dnknldkwt bel fra deu gbr deu bol muschechesvknzl grc kor dom fra aut dnk bel gbr omn ven uga oan jpn dnk geo swz npl bel hrv fra ita ita nzl can usa ita aut grc aus fra fra can aut chl bwa jam lva kwt pol can prt aut omn est che svn ita can oan che swz fin mus nld aut aus aze swe can tto est aus deu ita ita pry usa nor mus bel per arm kaz ttobel ita fra irlisr fra swe kor nld ken svn nor sau esp dnk caf nzl prt swe bwa isrmys kor nld esp chn irl ben hun che swe grc fin nor deu cmr aus dom jpn gbr che isr sle gbr chl ben aze pak usa omn npl mus nldbih nzl tun nam gab chl deu nor prt irl dnk idn per grc uzb aut ury ken gha hun nor gbr isr mys ita dom kor ury che prt uae gbr fin rus nzl tun deu jam dza png aushun chl namtur bra chl deu mys arg dnk nor tun ita oandom bel sau syr fin gab arg ltu gbr esp grc ury isr prt hkg pan hkg irl lao phl omn lbn arg swe pan swe est nzl nld kaz geo slv chl idn irl mus prt isr tur hun lbr gab pol grc mex aze kwt tkm mys pak dza nzl lao esp bwa pan tjkargprt mexcri chn egy dza oan hnd mys jam cze zaf chn fin bra tha mex mex civ syr hun uzb jam nam irl per bra nic pry tun lva chl bwa zaf tha mkd alb joroan mex mex bih syr tureth ven tto nam slv vnm tkm tha phl mex bra zmb dom argzaf mdg tto moz gtm cri cri zaf col ccol ribgr turnam arg pry bwa tgo bwa uga ven tha pan ltucaf irl egy dom gtm tur idn pry czeidn zmb cze per sen lka isr oan grc bra ury tur egy lbr tur chn tza zmb prt mus zaf sen tjktur esp mex hun idn tto ven cog idn omn cmr tza ecu pak mli zaf cri nam mkd col ner per pan slv gab cri tun gbr mar sau gtm jam tur alb tur yem jam ind ind lka colnam per tun rus slv pol pak mkd blr col ven bra gtm pan jor dom mda kgzblrmar cri kor yem gtm mys vnm mrt mar rom jam ind irl lbn idn egy cog arm moz bra ven mwi mar ner hnd phl swz mda dom egy jor jpn gab mar cri mys egy bgd bol egy khm nzl zafausrus mar bwa pan chl gin phl arg gmb ind arg slv arg chl gha ven civ tun col phl egy pan phl ury gin col tto egy moz phl pry gin pak dza swz sau mdg pry kor kor col jor lkakenmrt ind ken slvzaf turind crisau dom gin kor mar per zaf jorpry pol pry dzagtm slveriury gtm tha eri ben tkm ind gab usa dom nam sle gnb moz bgr nic swz lka png ltu jpn pry col vnm chn nic ecu bol nic gtm gtm chn mrt svk mli sle ecu phl slv vnm jam syr bol dza ben bgd hnd fin slv khm alb tha tha png gbr bol syr nam cmr mex mrt khm swz idn mwi mar gabsvk hun phl tun ury mar dza bra ven cri sau hkg syrtgo moz lka mli bra col uga bol egy ner cog chl civ npl mus khm npl ind mli dza syr cog ecu bgr ury kwt lka pak hnd nic per gmb lao lao nga romtgo uga jam bdi lka dza jor bgd syr npl png dom sle bgr lby lbn tjk png sen jam vnm rom ken bra sen hnd ven phl mng chn rom arg ind caf pan cmr lka ben hrv hnd syr ukr bol ner civ hkggtm gmb gha npl mys pak lao ken gha jor jorlkazmb gha bfa cri pak zmb sgp hrv gha idn pry mwi mrthnd gin pryper lka lka nzl png mrt lby chn bfa jam bgd civsgp nga sen bfa musbgd cog pak sen ken mrt mng pan caf bfa mli phl mng col npl uae per cog ukr bgd gmb chn mus eth nld kgzukr senbfa gnb moz cog tgo bdi bol hnd ken nic kwt civ chn nld bgd bdi rom pan gmb ecu tgo mdg mus eth eth ben slv gmb slvven nzlury mdg npleth bfa pak per bol eth sle civ ind ben hnd tgo nga lby gnb png mwi zmb bfa ken tgo eri jor pol cmr gha mli ner zmbcan sen pak sen mrtmli cmrbgd ben kgz npl cafgmb ken tha grc fin zaf ngazmb alb tha svk gmb mda bdi mdg nic cmr egy cmr mli cmr nga indcafmli etheth zmb bdi cog mex mwi mozgmb tgo caf ner lby ner esp gnb mex caf nic png mys zmb mrt bgr bgd belben bdi ethbdi mwi bel can nic espgnb mng ner grc sletgo nic ner nga sle moz moz caf sle

–15 –100

0

100

200

average initial currency valuation Notes: The vertical axis is the residual from a fixed effects regression; the residual includes the fixed effect for each country. Data are for 1980–2011, segmented into five-year periods. Initial currency valuation corresponds to the average initial currency valuation measures for the first year of each five-year period starting in 1980. Source: Bhalla (2007a) dataset extended to 2011.

Test 8: Are There Diminishing Returns? Is there a point beyond which further changes in valuation fail to help growth? That is, are there diminishing returns to under- or overvaluation? Alvaro Aguirre and César Calderón (2005) find nonlinearity; indeed, they find that beyond a small level (20 percent), increases in currency undervaluation hurt growth. This is analogous to the finding that outliers have an impact and that the impact is asymmetric.

economies are prominent in having large negative changes in undervaluation and large levels of investment. The Philippines is the one outlier, with a low level of investment and, on average, an overvalued currency. The consistently overvalued Philippine peso discouraged both foreign and domestic investment, which reduced growth.

132

DEVALUING TO PROSPERITY

Table 8.7

Regional growth and regional currency valuation, 1980–2011 Currency valuation (percent)

Region/country Developed economies

Initial

Change

Average growth in per capita GDP (percent)

14

0

1.6

Germany

22

–0.6

1.4

Japan

29

0.9

1.5

United States East Asia

–8

0.8

1.7

16

–3.3

3.8

ASEAN-7

30

–2.4

4.7

China

35

–7.5

7.6

–12

–3.3

3.2

12

–1.0

1.1

Malaysia Philippines Thailand Russia and Eastern Europe Latin America Brazil Chile

–8

–3.2

4.0

–23

0.6

2.9

23

–0.4

0.8

4

–0.5

0.9

–1

–2.3

2.7

Middle East and North Africa

20

–1.7

0.7

South Asia

46

–3.8

3.1

43

–5.9

4.5

99

–1.6

0.5

India Sub-Saharan Africa

ASEAN = Association of Southeast Asian Nations Notes: Data are segmented into five-year periods. Initial currency valuation corresponds to the average initial currency valuation measures for the first year of each five-year period starting in 1980. Negative measure indicates undervaluation; positive measure indicates overvaluation. Source: Bhalla (2007a) dataset extended to 2011.

Growth effects being positive and significant for real undervaluation levels up to 11 percent…for undervaluation levels between 11 and 20 percent, growth effects are positive although statistically not different from zero. Finally, growth impact of undervaluation becomes negative for undervaluation higher than 20 percent—and significant for undervaluation higher than 30 percent. (Aguirre and Calderón, 2005, 19 and figure 2)

I do not find this asymmetry. Table 8.6 presents two different tests of the Aguirre and Calderón hypothesis and results. The first test classifies the range of observations into three categories: valuation levels below –10 percent, between –10 and +10 percent, and higher than +10 percent. The coefficients for the first and third categories are nearly equal, significant, and of the same sign; interestingly, the coefficient for the second category is not significant. A CURRENCY VALUATION AND GROWTH 133

more elaborate test classifies initial valuation levels into five categories, with the boundaries chosen according to the percentile of valuation, with the lowest percentile being the most undervalued. Note that, by definition, both the low percentiles (extreme undervaluation) and high percentiles (extreme overvaluation) are outliers, so this classification also helps test the outlier hypothesis. The results show near constancy of the initial valuation coefficient at about –0.01 for the largest coefficient and –0.023 for the 10–25 percentile category (that is, for countries with greater undervaluation). This is somewhat opposite to the Aguirre and Calderón (2005) result that larger undervaluation hurts growth, as does larger overvaluation. The estimates indicate that the economies with highly overvalued currencies—those with an average valuation near 100 percent, like most economies in sub-Saharan Africa—grew about 1 percentage point less per year.

Conclusion This chapter examined the empirical relevance of currency valuation to growth for various countries over time, using various econometric methods. The basic law of economic growth is that the economic productivity of individuals is overwhelmingly affected by incentives. More capital investments will be made when there are fewer constraints on production and enterprise. When individuals are productive, they will enjoy a higher level of income and, collectively, the economy will grow faster. The most significant result from the various tests conducted in this chapter is that the price of the currency is an important economic incentive—the cheaper the currency, the greater the incentive to produce, with the results being higher profits, higher investment, and faster growth. The sensitivity analysis suggests that currency valuation retains its significance no matter what variables are added to the growth model, either individually or in combination. The significance of currency valuation is not due to the presence of outliers. This conclusion is therefore robust for a variety of specifications and using various econometric tests: Currency valuation is shown to have a symmetric effect on growth. Unlike earlier studies, this result holds for all ranges and not just for extreme values of misalignment. Countries with overvalued currencies grow at a slower rate; those with undervalued currencies grow at a faster rate, other things equal. There is little asymmetry or nonlinearity in the results. This result informs the analysis throughout the remainder of this book.

134

DEVALUING TO PROSPERITY

9 Policy Failures and Growth Miracles

No testimony is sufficient to establish a miracle, unless the testimony be of such a kind that its falsehood would be more miraculous than the fact which it endeavors to establish. —David Hume, An Enquiry Concerning Human Understanding The worth of any policy or variable is the extent to which it can explain success and failure. The previous chapters show how currency undervaluation can help explain investment and growth. However, some nagging questions remain regarding the conclusion that a country can undervalue its way to prosperity. First, currency undervaluation does not explain high growth in various studies that are based almost exclusively on the Dollar-Easterly measure of currency valuation (see chapter 4). Second, some models that did support the relationship between currency undervaluation and growth—focusing on export-led growth strategy as an indirect and somewhat inefficient form of currency undervaluation—did so only when outliers were included in the model, calling into doubt the role of currency valuation as a determinant of growth. Third, deliberate currency undervaluation, even assuming it worked, is often considered subservient to the real determinant of high growth: effective institutions. And finally, Michael Woodford (2009) shows that there is “construction bias” in traditional measures of currency valuation. Previous chapters present evidence that addresses these points. Most important, chapter 4 discusses how traditional measures of currency undervaluation contain significant measurement errors, and how this makes the noise-to-signal ratio of these measures high. It is no surprise, therefore, that the use of these measures has led to an underestimation of the importance of undervaluation as a determinant of growth. When we correct for a large part of the measurement error through the use of a better functional form between two given variables—in this case, the real exchange rate and income per capita— we see that exchange rate policy does matter. And it matters even more when policy is represented by two variables rather than one—both the initial level of currency undervaluation (to signal the relative cheapness of investment) and 135

the change in the level of valuation (to signal the directional change in the profitability of investment). Both variables affect investment behavior and do so with differing impact. Through these relationships, we see that currency undervaluation affects growth via investment. This chapter subjects the link between currency valuation and growth to further tests. In particular, I examine whether currency misalignments play a role in growth changes.

Growth Successes and Failures There have been some conspicuous growth successes and failures, and four reports analyze these in some detail. The first is the pioneering study by the World Bank (1993) on the East Asian “miracle.” The second is the article by Ricardo Hausmann, Lant Pritchett, and Dani Rodrik (2005) documenting the determinants of growth acceleration in a sample of more than 80 countries.1 The third is by Leszek Balcerowicz and Stanley Fischer (2006), who use a case study approach to examine growth in several developed and developing economies, and in particular the transition economies of Eastern Europe. The fourth, from the Commission on Growth and Development (2008) of the World Bank, identifies13 countries whose growth performance could be defined as extraordinary—that is, an average annual GDP growth above 7 percent for a continuous period of 25 years. Members of this distinguished club include Japan from 1950 to 1983, Brazil from 1950 to 1980, and Thailand from 1960 to 1997. Another contribution to this group is not a study but the hypothesis that growth slows once a country’s income per capita enters a “middle income range” (see Gill and Kharas 2007; Eichengreen, Park, and Shin 2011). This is often referred to as the middle income trap. An alternative explanation is that growth slows partly because of advanced catch-up. For example, in 2000 at an income per capita of close to $1,000, China had a catch-up growth advantage of approximately 1.5 percent per year. In 1980, this catch-up factor was 2 percent, but by 2011, it had declined to 0.8 percent. The implication is that China’s annual growth rate can be expected to be lower by 1 percent per year than the growth rates from just a decade ago. In other words, catch-up may help explain the middle income trap. Another factor, besides advanced catch-up, may be contributing to declining growth rates and therefore to the perception that there exists a middle income trap. With accelerated growth, there is an increase in capital flows, which causes the currency to become either overvalued or less undervalued. In either event, the GDP growth rate will tend to decline, as described in chapter 6. Overall, then, the middle income trap can be said to exist only if there remains a systematic tendency for growth rates to slow after controlling for catch-up and overvaluation effects. 1. Also worthy of mention is Rodrik (2003a), which provides background material for Hausmann, Pritchett, and Rodrik (2005).

136

DEVALUING TO PROSPERITY

Explanations for Failure The World Bank (1993) report attributes the East Asian miracle primarily to these countries’ pursuit of an export-push strategy geared toward opening up their economies (trade liberalization) and allowing their exchange rates to remain competitive. This policy is described by the World Bank (1993, 115) as having maintained for Hong Kong “a stable and at times slightly undervalued exchange rate vis-à-vis the US dollar”; and for the high-performance Asian economies (HPAE) as having “avoided the severe depreciation that beset sub-Saharan Africa and Latin America” (p. 114). The overall assessment is that several HPAE governments used exchange rate policies to offset the adverse impact of trade liberalizations…. [S]ome used deliberately undervalued exchange rates to assist exporters. In these instances, exchange rate policy and fiscal and monetary tools to carry it out became a part of an overall exportpush strategy. (p. 125)

In essence, the report concludes that growth miracles come about via export-led growth, a view endorsed by many economists, prominent among them Béla Balassa (1978) and Anne Krueger (1990). But the report suggests that export-led growth is possible without any undervaluation of the currency. It is worth quoting this somewhat radical conclusion in some detail: The export-push strategy appears to hold great promise for other developing economies. Fortunately, many powerful instruments of export promotion are not only within the institutional capacity of many developing economies but remain viable in today’s economic environment. Creating a free trade environment for exporters, providing finance and support services for small and medium-size exporters, improving trade-related aspects of the civil service, aggressively courting export-oriented foreign direct investment, and focusing infrastructure on areas that encourage exports are all attainable goals that are unlikely to provoke opposition from trading partners. Indeed, some or all of these have been part of the export push in Indonesia, Malaysia, and Thailand. These three economies, the most recent participants in the “economic miracle,” may show the way for the next generation of developing economies to follow export-push strategies. (World Bank 1993, 25)

Several scholars criticized the World Bank study on grounds of reverse causation—that is, higher trade and openness per se may have had little to do with the miracle of growth. Rodrik (2003b) discusses about a dozen growth miracles, among them some outside of East Asia and China such as Botswana and Mauritius. The study’s conclusion is simple: Growth miracles occur because of “institutions”: Institutions that provide dependable property rights, manage conflict, maintain law and order, and align economic incentives with social costs and benefits are the foundation of long-term growth. This is the clearest message that comes across from the individual cases. (Rodrik 2003b, 10)

POLICY FAILURES AND GROWTH MIRACLES 137

Hausmann, Pritchett, and Rodrik (2005) do not study economic miracles but rather cases of extreme growth acceleration from a reasonably low base. They use three criteria: Growth per capita is rapid (above 3.5 percent per year); growth accelerates by at least 2 percent per year for at least eight years; and output after the growth acceleration period exceeds the pre-episode peak. There is a conflict in the study between the hypothesis that exchange rate depreciation matters and the empirical finding that it does not. The abstract states that “growth accelerations tend to be correlated with increases in investment and trade, and with real exchange rate depreciations,” but a later passage states, “If we look at these same variables during the eight-year growth acceleration, instead of just around the start of the process we find similar results except for the real exchange rate…. [R]eal exchange rate changes are no longer statistically different from zero” (p. 317). The authors are quite candid about their failure to find any positive correlations: Growth accelerations tend to be highly unpredictable: the vast majority of growth accelerations are unrelated to standard determinants and most instances of economic reform do not produce growth accelerations…. [S]ustained and unsustained growth accelerations tend to be triggered by different conditions.... [M]ost growth accelerations are not preceded or accompanied by major changes in economic policies, institutional arrangements, political circumstances, or external conditions…. [I]t would appear that growth accelerations are caused predominantly by idiosyncratic, and often small-scale, changes. The search for the common elements in these idiosyncratic determinants—to the extent there are any—is an obvious area for further research. (Hausmann, Pritchett, and Rodrik 2005, 303, 327–28)

Balcerowicz and Fischer (2006, 4) conclude that in Europe and the transitional economies of Eastern Europe “reliance on market forces within an open economy in a stable macroeconomic environment, with assured property rights, are the keys to rapid economic growth.” A recent addition to the literature by the Commission on Growth and Development (2008) reaches the same conclusions as the four mentioned here. Possibly reflecting the fact that the document was written by a committee, this voluminous and exhaustive study makes no mention of real currency depreciation as a determinant of growth. What did make possible the exceptional growth witnessed in 13 countries from Botswana to Oman to Japan to China over the last 50 years? According to the report, there were five major factors: The successful countries “fully exploited the world economy…maintained macroeconomic stability…mustered high rates of saving and investment… let markets allocate resources…[and] had committed, credible, and capable governments” (p. 21). The Gill and Kharas (2007) and Eichengreen, Park, and Shin (2011) studies focus on growth deceleration, or the middle income trap. They explain growth failures, or slowdowns, as follows.

138

DEVALUING TO PROSPERITY

Rather than having to pioneer new technologies, late-developing countries can import knowhow from abroad. They can reap productivity gains simply by shifting workers from underemployment in agriculture to export-oriented manufacturing, where those imported technologies are utilized. With young generations engaged in saving enjoying higher incomes than elderly dissavers, they are able to finance high levels of investment…. Periods of high growth in late-developing economies do not last forever. Eventually the pool of underemployed rural labor is drained. The share of employment in manufacturing peaks, and growth comes to depend more heavily on the more difficult process of raising productivity in the service sector. A larger capital stock means more depreciation, requiring more saving to make this good. As the economy approaches the technological frontier, it must transition from relying on imported technology to indigenous innovation. (Eichengreen, Park, and Shin 2011, 1)

So what works? What can we conclude from the plethora of studies on the determinants of growth successes (and failures)? Taken at face value, these studies suggest that little is really known about growth, growth acceleration, or growth miracles. A stable economic environment (noted as one cause of high growth) is more likely an outcome than a determinant. Increased openness is a possible factor, but there is the problem of identification. In the absence of clear determinants, the explanation often centers on the presence or absence of Western-style institutions, particularly institutions dealing with property rights. Chapter 11 tests the conclusion that institutions positively affect growth, and finds it to be lacking—confidence in institutions is highly misplaced. Somewhat surprisingly, currency valuation is not explicitly discussed in any of these key studies on the factors for growth success or failure, although it often lurks in the background. For example, it is completely absent from the Commission on Growth and Development (2008) report. And while currency valuation is mentioned by Hausmann, Pritchett, and Rodrik (2005), they conclude that its contribution to sustained growth is fragile, at best. Chapter 8 exhibited how currency valuation can explain systematic differences in growth rates. The remainder of this chapter explores whether currency valuation can explain growth accelerations and decelerations. Specifically, what level, or pattern, of currency undervaluation is indicative of a possibility of future growth acceleration?

Structural Breaks in Growth This section reports on several econometric tests of growth acceleration, which are identical to those in Hausmann, Pritchett, and Rodrik (2005). They identify a growth break year—the year of a change in the trend of growth per capita— and determine whether the seven-year period after the break is significantly different from the seven years before. This is the definition used to estimate growth performance for the episodes include in the above five studies.2

2. Both Balcerowicz and Fischer (2006) and the Commission on Growth and Development (2008)

POLICY FAILURES AND GROWTH MIRACLES 139

In addition, currency valuation is used to identify an additional type of break year, when the trend of currency valuation changes either toward undervaluation or toward overvaluation. The former is expected to have a positive effect on subsequent growth, the latter a negative effect. A simple formulation of trend change is when the three-year average of the level of valuation moves below the 33rd percentile of such averages. A break from above, toward undervaluation, is termed a negative break. A break from below, toward overvaluation, is identified according to the 50th percentile of the three-year average and is termed a positive break. (The data used to identify the break years are from Bhalla 2007a.) The third type of break year is identified in the spirit of the Commission on Growth and Development (2008) study—namely, when a country’s growth rate per capita begins to achieve a doubling in income levels in the minimum period of time (again using the dataset from Bhalla 2007a). For example, China’s income per capita in purchasing power parity (PPP) terms has increased by 11 times since 1979. The first doubling took 12 years and ended in 1991; the second doubling took 10 years and ended in 2001; the third doubling took only 7 years and ended in the middle of 2008.3 Therefore, the minimum period required for China to double its income per capita is seven years, and the break year is 2001. (The relationship between currency valuation and the minimum doubling time is explored below.) The fourth type of break year is the year of maximum currency devaluation since 1950. According to the hypothesis that the real exchange rate is endogenous (see chapter 6), such devaluations cannot really work. An alternate hypothesis is explored here—namely, that such nominal devaluations also bring about a real devaluation and that such real devaluations help growth. The fifth type of break year is identified to test the middle income trap hypothesis. The primary example is Brazil. Between 1970 and 1980, while the rest of the world was caught in the grips of stagflation, Brazil grew at an average rate per capita of 5.8 percent per year. In 1980, Brazil’s income per capita was PPP$6,380 in 1996 prices (or PPP$17.5 per day). Between 1981 and 1990, when the rest of the world was booming, income per capita in Brazil grew at an average annual rate of –0.26 percent. While other explanations are possible, this drastic slowdown is most often cited as proof of the growth trap that awaits economies as income per capita exceeds $6,000 in 1996 prices, or about PPP$13,000 in 2011 prices. Here, middle income is defined to be income per capita per day between PPP$16 and PPP$32 in 1996 prices, or income per capita per year between PPP$12,700 and PPP$25,400 in 2011 prices. The break year is thus identified as the year when the three-year average rate of income per capita enters this range. use small datasets, especially if countries from Eastern Europe or the former Soviet Union are excluded. 3. In 2008, China’s income per capita was PPP$8,704, in 1996 prices. In 2011, it was estimated at PPP$11,310. In 1979 income per capita was PPP$1,023; hence, 2011 income levels were almost 11 times the 1979 income levels.

140

DEVALUING TO PROSPERITY

A total of nine break years are considered: three for growth decelerations (one of which is for a currency overvaluation and two for middle income trap) and six for growth accelerations. These are used to test for the existence of a middle income trap and to see if currency valuations predict the future course of growth. This provides a rigorous test of whether, under varied circumstances, currency valuations matter for growth.

Explaining Growth Acceleration What determines the difference in growth rates in the before and after acceleration periods? Table 9.1 presents the results on currency valuation changes and growth for each study and type of break year. Column 3 predicts the direction of growth change. As discussed, several datasets were constructed based on the existence of a growth spurt (e.g., Hausmann, Pritchett, and Rodrik 2005; Bhalla 2007a for income doubling). These yield an average growth acceleration of 4.7 percent per year. The dataset used to test the middle income trap (Bhalla 2007a) had an observed average growth decline of –3.8 percent. Four break years were identified with datasets constructed using the data in Bhalla (2007a), for currency undervaluation, currency overvaluation, the middle income threshold, and the year of maximum currency devaluation. The first three of these show almost no change in growth. The maximum devaluation break year does yield a high growth acceleration of 1.8 percent, but this is to be expected in part because the devaluations most likely occurred as a result of large current account deficits, a circumstance not normally associated with growth. What is of interest is whether the change in currency valuation is associated with a change in growth rates. The growth rate in any seven-year period, t, can be represented as follows: Gt = D + E1*(iYt) + E2 × (iCVt) + E3 × dCVt+ E4 × Zt + Ht; and Gt–1 = D + E1 × (iYt–1) + E2 × (iCVt–1) + E3 × dCVt–1 + E4 × Zt–1 + Ht–1

(9.1)

where G is the growth rate of income per capita for the six years subsequent to the initial break year, iY is the log of income per capita in the initial year of the seven-year period, CV is the currency valuation in the initial year, dCV is the mean change in currency valuation, and Z is a vector of other determinants of growth. The difference in the two equations, Gt – Gt–1, is the acceleration in growth rates. If the break year is t, then the “before” dataset is for (t – 8), and the acceleration for the before dataset is from (t – 15) to (t – 8); the “after” dataset is from (t – 8) to (t + 7).4 If the coefficients on the currency valuation variables are significant, then this will be yet another robust confirmation of the hypothesis that currency valuation is a serious determinant of economic growth rates. 4. The break year is not included in either the before or after computations.

POLICY FAILURES AND GROWTH MIRACLES 141

142 DEVALUING TO PROSPERITY

Table 9.1

Currency valuation and growth accelerations and decelerations, 1950–2011 (percent) Currency valuation

Type of growth break (2)

Expected change in growth rates (3)

Before (4)

After (5)

Change (6)

Acceleration in growth (7)

Hausmann, Pritchett, and Rodrik (2005)

Growth acceleration

Positive

68.5

57.6

–10.9

3.7

Balcerowicz and Fischer (2006)

Growth acceleration

Positive

32.3

24.2

–8.2

6.5

Commission on Growth and Development (2008)

Growth acceleration

Positive

108.4

66.6

–41.8

3.0

Study (1)

Bhalla (2007a)

Income doubling

Positive

39.4

29.4

–10.0

5.6

Growth deceleration

Negative

–11.2

–4.1

7.1

–3.8

Bhalla (2007a)

Negative (toward undervaluation)

None

12.5

–4.7

–17.2

0.7

Bhalla (2007a)

Positive (toward overvaluation)

None

21.4

35.8

14.4

0.6

Bhalla (2007a)

Middle income threshold

None

8.2

0.1

–8.1

–2.1

Bhalla (2007a)

Year of maximum currency devaluation

None

59.8

35.6

–24.2

1.8

35.3

24.9

–10.4

1.9

Eichengreen, Park, and Shin (2011)

Average

Notes: Growth breaks identified using the dataset in Bhalla (2007a). The seven-year average currency valuation enters the bottom third percentile of the seven-year average. Income doubling refers to the number of years taken by an individual country to double its income per capita, 1950–2011. The first four studies yield an average acceleration of 4.7 percent per year.

Table 9.2 shows the results after pooling the data for the nine datasets. Four sets of results are presented for different combinations of sample selection and growth break years. The results are surprising, especially given the literature. Currency valuation is very significant in explaining growth accelerations and decelerations. For 487 different break years for more than 100 countries over the past 60 (1950–2011), each 10 percent change in currency valuation affects growth by –0.2 percent. The impact is fairly constant across datasets. The impact of the acceleration of change in valuation varies between –0.1 and –0.24. Results for the tests on the middle income trap are reported in table 9.3. The first row presents the conventional fixed effects model used throughout the book. A middle income dummy has a positive sign, although the coefficient is not significant. This test suggests that there is little evidence that the growth in income per capita stalls once an economy gets into this zone. Brazil’s experience in the 1980s is not common, let alone universal. The other two regressions report results for growth acceleration for the two different middle income samples. Both show very sharp effects for currency valuation. The hypothesis stated at the start of this chapter was that the middle income trap may involve an identification problem—that successful growth might cause currencies to become overvalued and that this overvaluation may cause the observed growth slowdowns. But there does not seem to be any evidence for this. The significance of currency undervaluation for growth acceleration shown here stands in sharp contrast to the results of Hausmann, Pritchett, and Rodrik (2005). They use the Dollar-Easterly measure and find currency valuation to be insignificant. A change in the functional form relating the real exchange rate to income per capita considerably improves the importance and significance of currency valuation in explaining growth. This is further affirmation of the conclusion that measurement errors most likely cause a misinterpretation of the role of currency undervaluation in determining growth. Chapter 4 suggested that a useful way to look at the real exchange rate was as a ratio of productivity to costs. If a currency is undervalued, this ratio is increased because of a lowering in the denominator (costs), which makes investments more profitable, which increases investment and thereby increases growth. According to the results presented in this chapter, growth accelerates when policy makes investments more profitable. The break years identified according to changes in currency valuation (positive and negative) do not reflect changes in growth rates but rather perceived improvements or deteriorations in relative competitiveness, with competitiveness measured by the change in the percentile ranking of currency valuation. No growth acceleration was expected or observed (the average acceleration was only 0.6 or 0.7 percent per year). Yet, currency undervaluation is significant: Countries that allowed costs to decline via a currency depreciation or the prevention of a currency appreciation reaped the fruits of higher growth. It is also noteworthy that break years chosen on the basis of growth show similar results. POLICY FAILURES AND GROWTH MIRACLES 143

144 DEVALUING TO PROSPERITY

Table 9.2

Explaining growth accelerations, different definitions, 1950–2011 Coefficients Difference in currency valuation

Row

Model specifications

1

All sample

Initial income per capita –9.84***

Initial –0.020***

Average change

R2

Number of observations

–0.16***0

0.47

487

2

Selected sample

–9.87***

–0.013***

–0.099***

0.47

396

3

Row 2 excluding Russia and Eastern Europe

–9.73***

–0.023***

–0.11***0

0.45

367

4

Row 3 and only currency valuation breaks

–11.67***

–0.03****

–0.24*00*

0.61

100

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. The dependent variable for all model specifications is acceleration in growth rate of income per capita. For each of the different classifications of countries according to a break year, the model relates the acceleration in growth of income per capita to the difference in initial income per capita, the difference in initial currency valuation, and the difference in the average change in currency valuation. See text for details. Source: Bhalla (2007a) dataset extended to 2011.

Table 9.3

Is there a middle income trap? Coefficients Currency valuation

Row

Type/model

POLICY FAILURES AND GROWTH MIRACLES 145

Dependent variable

Initial

Change (lagged)

Middle income

R2

Number of observations

Per capita income growth

–0.013***

–0.029**

0.36

0.43

961

1

Selected sample

2

Middle income (Eichengreen, Park, and Shin [2011] definition)

Acceleration in per capita income growth

–0.064***

–0.160**

0.37

33

3

Middle income (defined in text)

Acceleration in per capita income growth

–0.04****

–0.130**

0.46

48

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Data for row 1 are compiled in five-year intervals with the last five-year observation for 2005–11. See appendix A for details. The fixed effects model includes time and country dummies. Rows 2 and 3 relate acceleration in growth rate of income per capita with the difference in initial currency valuation and the difference in the rate of change in currency valuation. Middle income defined as income per capita between PPP$12,700 and PPP$25,400 in 2011 prices. Source: Bhalla (2007a) dataset extended to 2011.

Figure 9.1

Added variable plot: Growth acceleration and initial level of currency valuation

acceleration in per capita income growth 10 gnb bwakwt bwa lbn jor vnm jpn lbn jpn oan oan tha bra tto kor chl irl jpn mar syr kor arg mus kwt sgp dom jpn chn mar grc oan gha chn kor espoan esp dom chn egy chl bwa pak prt mus chn prt tha vnm grc pak esp prt cmr uga irl ind mys pry grc chl prt tto chn pan eth nic gnb ury per egy gnb hkg moz argcoguae uga mli irlury vnm mys tha bel mozethind syr per isr esp png mys grc moz pan fin pan fin fra panbgd chl can ury per mys dza ita usa mwi isr indbwa mys mwi can hkg bel syr esp sgp lka lka cmr lka prt gbr sgp dza col mex lao arg aut aus irlirl nor can pak tun tur mli dom phl irl nld turphl mar dza zaf ben nor irl tun chl mex jam chlpng dnk zaf sen bel lby bgd col npl mus slv mrt nzl che cri dnk gab gab pry swe egy arg mwi arg mex gbr egy nld col lka mus fin nor isr aut cri eth npl gha oan tto ita dnk bel egy col mex oan swe esp bol nor col irl tur ner deu fra zaf bfa gbr deu col marthabwa nzljpndnk aus che ecu ven usa aut kor aus nor ita ken ind bgd mdg bel phl gbr nld nam nor bwa autjpn dza fra mex nam aus usa pakgtm bra chl fra bra sgp fin mys ecu bra deu nzl prt gbr gha pry isrbel chl sgp nam dom venidngin cri usa mys tun jpn ury grc civ tun nld bra mex bol slv che hkg cog caf thagab ken hnd syr dza swz kwt tur isr gtm itaven omn civ ven tto fin grc esp jorven nga pan tto zaf zmb tgo cri png ury arg caf nic moz gmb syr sle uae nam gab tza bol gab kwt chn

5

0

vnm

–5

–10 –100

–50

0

50

100

difference in initial currency valuation coefficient = –.02318916, (robust) standard error = .00566043, t = –4.1 Notes: For each of the nine different classifications of countries according to identified break years, the model relates the acceleration in growth of GDP per capita to the difference in (log) initial income per capita, the difference in initial currency valuation, and the difference in the average change in currency valuation. Plot is based on equation reported in table 9.2, row 3. See text for details and appendix table B.1 for country abbreviations. Source: Bhalla (2007a) dataset extended to 2011.

Figures 9.1 and 9.2 are the partial scatter plots for the all-inclusive sample (row 1 and the last two regressions in table 9.2, respectively). It is evident that outliers do not determine the significance. Regardless of whether the method has means (figure 9.1) or initial value and change (figure 9.2), the broad impact is apparent: Residuals are equally distributed on either side of the trend line. The policy that best promotes growth can be called export-led growth, or openness to trade, or exchange rate undervaluation—the name does not matter, the results do. This conclusion is contrary to most, if not all, recent evaluations of the causes of growth, although it is consistent with the arguments of Balassa (1978) and some studies before 1990.

146

DEVALUING TO PROSPERITY

Figure 9.2

Added variable plot: Growth acceleration and average change in currency valuation

acceleration in per capita income growth 10 gnb chn bwa uga

5

chn uga syr vnm

0

gha

–5

bwa jor kwt

lbn jpn jpn lbn arg tto mar oan oanirlkor gha jpn tha bra jpn egy grc mar chnsgp syr kor esp chn chlkwt muscmr dom chn dom pak vnm esp oan prt kor pak oan tha grc bwa pry prtind esp egy mus mys irlchltto moz gnb irl chl hkg grc per per prt mozpan nic eth mli gnb chntha moz mys syr vnm ury fin bel cog uae vnm ury png grc ind dza per mys pan dza esp bgd usa pan ury fin mwi pan sgp chl fra mys gbr eth mex ita can sgp mys cmr mwi ind bel isr irlaus irlisrarg mli col hkg nor nor dom can irl lka lka chl bwa aut pak tunchl mar esp prt zaf mus mex arg tun swe nld ben phl slv irl mrtnor dza zaf bgd tur dnk sen pry dnk egy mex esp png col nld mus phl argswe lao nzlgbr gab cri lby jam mwi npl bel belche bgd tur cri mex lka colkor dnk ita col oan isr oan irldeu aut gab gha eth aut fra fin zaf jpn npl nor fra tto che dnk aut deu ita aus col arg marner gbr jpnaus egy bol nor bfa gbr bel usa tur tha ind nld egy dza bwa bra usa ken aus nzl chl ecu sgp nor mex deu fra fin bel bwa sgp bra mdg nam col bra phl prt hkg ecu gtm mys gbr nzl mys nld ven pak usa jpn isr bol tun grc idn tun ury chl kwt slv pry civbra cog dom cri nam nam tha syr dza gtm turgab isrcaf che gin ven hnd fin ven mex esp tto civ swz ita omn arg venurytto ven nga panken grc caf zmb zaf tgo cri jor png moz syr nic uae gab gmb sle nam bol gab tza kwt

–10 –30

–20

–10

0

10

20

difference in average rate of change in currency valuation coefficient = –.1086147, (robust) standard error = .02657861, t = –4.09 Notes: For each of the nine different classifications of countries according to break years identified, the model relates the acceleration in growth of GDP per capita to the difference in (log) initial income per capita, the difference in initial currency valuation, and the difference in the average rate of change in currency valuation. Plot is based on equation reported in table 9.2, row 3. See text for details and appendix table B.1 for country abbreviations. Source: Bhalla (2007a) dataset extended to 2011.

POLICY FAILURES AND GROWTH MIRACLES 147

10 Mercantilism and Miracles

Fair is foul, and foul is fair. Hover through the fog and filthy air. —William Shakespeare, Macbeth The evidence presented so far is compelling in demonstrating that currency undervaluation generates extra growth. This ranks it alongside the other two factors identified in the literature as underpinning growth—investment and openness. Taking this evidence as a reference point, this chapter stretches the analysis in a related dimension: What are the effects of currency undervaluation at the extreme? Overvaluation at the extreme is easy to judge and to criticize—policymakers are hurting their own economy. But the results of extreme undervaluation can mirror those of mercantilism—the scourge of those who believe in markets and free trade. This raises the question of whether extreme currency undervaluation should be considered mercantilist. To address this question, we first define mercantilism and then offer a measure of mercantilism and test it to determine how well mercantilism pays.

Defining Mercantilism Webster’s Dictionary defines mercantilism as “an economic system developed during the decay of feudalism to unify and increase the power and especially the monetary wealth of a nation by strict governmental regulation of the entire national economy, usually through policies designed to secure an accumulation of bullion, a favorable balance of trade, the development of agriculture and manufactures, and the establishment of foreign trading monopolies.” We can modernize this definition by referring instead to the accumulation of US dollars (instead of gold bullion) and a stable exchange rate (for a favorable balance of trade). This updated definition suggests that reserve accumulation and mercantilism are closely associated. However, reserve accumulation is an effect, not a cause of mercantilism. 149

Reserve accumulation is the result of currency interventions meant to prevent the appreciation of the currency. This suggests that currency undervaluation should figure prominently in any index of mercantilism. But there can be instances when an undervalued currency does not lead to a surplus on the trade account. One prominent example is India, which has consistently had a current account deficit of about 2 percent. This shows that exchange rate undervaluation cannot be the only component of mercantilism. A second important component follows from the definition—the size of the current account surplus (correlating to the accumulated stores of bullion). But a current account surplus also can be caused by factors other than mercantilism, such as a cultural taste for high savings. Therefore, the size of the surplus and the magnitude of currency undervaluation individually are incomplete indicators of mercantilism. Together, however, they do form a balanced indicator of mercantilism. Specifically, countries that have a current account deficit and an undervalued exchange rate are not as mercantilist as countries that have both an undervalued exchange rate and a current account surplus. Incorporating this balance is the Borda rank,1 which is a simple average of the ranks of the two components. For each year, countries in the sample are ranked according to the degree of undervaluation as measured by the Bhalla (2007a) method, with the most undervalued currency given the rank of 1. Analogously, countries are ranked according to the size of their current account surplus (as a percent of GDP), with the highest current account surplus given the rank of 1. The two ranks are then averaged for each year and for each country. This average rank (Borda rank) is the proposed index of mercantilism. Table 10.1 reports the mercantilism rankings and associated variables for 1998–2011, the years following the Asian crisis. The mercantilism rankings are presented for the three periods 1990–2011, 1990–97, and 1998–2011. For the 14 years from 1998 to 2011, several East Asian economies are in the top 10, with Malaysia, Singapore, Taiwan, Thailand, Hong Kong, and China at the top, as the most mercantilist economies. After the start of the Asian crisis, Malaysia reacted by closing the capital account, at least temporarily, and by taking charge of the exchange rate. It has prevented the ringgit from appreciating in nominal terms since the crisis—the exchange rate averaged 3.9 to $1 in 1998, a year after the crisis, and in November 2011, the exchange rate was about 3.15 to $1. The currency has been undervalued by more than 50 percent during the last decade; for 2011, the undervaluation was about 41 percent. Note the low mercantilism rankings for the East Asian economies prior to the onset of the crisis. China’s mercantilism ranking was 24 for 1990–97,

1. Named after the 18th century French mathematician and political scientist Jean Charles de Borda, who in 1770 devised a method for evaluating the superior among n alternatives, each of which has an ordinal ranking. His method has been applied to choosing winners in elections: The first choice receives the highest rank (1), the second choice receives 2 and so on. The person with the highest average rank (in this case the lowest score) is the winner.  

150

DEVALUING TO PROSPERITY

Table 10.1

Mercantilism rankings for selected countries, 1990–2011 Mean values for selected indicators (1998–2011)

Country

Current account surplus (percent of GDP)

Currency valuation (percent)

Per capita GDP growth (percent)

Rankings, 1998–2011

Current account surplus

Mercantilism rankings

Currency valuation

Per capita GDP growth

1990–2011

1990–97

1998–2011

Developing economies Malaysia

13.0

–50.1

2.4

4

11

63

10

66

1

Singapore

18.9

–20.1

3.0

2

33

51

2

3

3

MERCANTILISM AND MIRACLES 151

Taiwan

6.9

–39.8

3.4

11

18

36

3

13

4

Thailand

4.7

–49.2

2.5

23

10

39

21

77

6

Hong Kong

8.2

–30.7

2.8

8

29

45

6

28

7

China

4.7

–42.8

9.2

19

17

1

7

24

8

Korea

3.2

–31.9

3.4

28

21

34

11

35

11

Indonesia

2.3

–31.5

2.3

31

24

46

16

43

13

Saudi Arabia

13.8

2.1

0.5

12

49

117

23

54

14

Botswana

6.1

–16.1

4.2

25

38

27

15

25

18

Chile

0.3

–24.4

2.5

44

28

55

25

47

19

Argentina

1.1

–41.7

2.6

38

23

37

38

74

21

India

–0.9

–23.8

5.7

47

31

5

35

64

22

Egypt

0.2

–13.0

2.5

41

41

71

22

19

30

Philippines

1.4

–2.7

2.1

35

48

73

59

89

32

(continues on next page)

152 DEVALUING TO PROSPERITY

Table 10.1

Mercantilism rankings for selected countries, 1990–2011 (continued) Mean values for selected indicators (1998–2011)

Country

Current account surplus (percent of GDP)

Rankings, 1998–2011

Mercantilism rankings

Currency valuation (percent)

Per capita GDP growth (percent)

Current account surplus

Currency valuation

Per capita GDP growth

1990–2011

1990–97

1998–2011

–2.7

–20.3

2.1

67

33

75

32

21

40

1.0

6.3

1.7

37

62

84

60

85

42

Developing economies (continued) South Africa Israel Brazil

–1.4

5.9

1.9

62

57

78

65

58

61

Mexico

–1.5

17.5

1.7

58

70

89

74

64

73

5.4

23.1

1.3

17

83

108

31

23

39

Developed economies Netherlands Japan

3.2

26.3

0.5

22

79

127

45

53

44

Germany

3.3

21.7

1.3

27

77

101

55

71

46

United States

–4.2

0.4

1.2

85

52

107

57

20

77

United Kingdom

–2.2

19.0

1.1

66

72

110

76

56

81

Spain

–5.2

15.7

1.9

93

71

82

84

61

92

Notes: The mercantilism ranking is the average rank of current account surplus and currency valuation for the period. Countries are listed by the mercantilism ranking for 1998–2011. Source: Bhalla (2007a) dataset extended to 2011.

but has been 8 since 1997. The mercantilist rankings generally pass the smell test. The top mercantilist rankings are dominated by most members of the ASEAN-7.2 Among developed economies, the Netherlands tops the list during all three periods. Chapter 12 documents how the Dutch guilder was one of the most undervalued currencies in the late 19th century; it is interesting to note that the Netherlands continues its mercantilist tendencies.3

Mercantilism and Growth Some argue that there is no need for concern about the practice of currency undervaluation because the countries practicing it are being irrational by giving away underpriced goods. For example, American consumers benefit from the availability of cheap Chinese goods. But this argument overlooks what a mercantilist economy may gain in exchange for what it loses by selling its goods cheap. One such gain could be extra growth. We now test this hypothesis. The mercantilism index is computed in a very indirect manner; it is an average of rankings obtained from the magnitude of current account surpluses (easily observed) and currency undervaluation (not so easily observed). As such, the mercantilism index likely contains a large measurement error at least concerning the currency misalignment. If the model is estimated with the initial mercantilism ranking, then any test of the index is biased against success because of the measurement errors involved—the initial five-year values of the two variables are first transformed into rankings, then the rankings of two different components are averaged, and finally a rank of rankings obtained. A realistic expectation is that the mercantilism indices are not significant in cross-country panel regressions. Table 10.2 presents the results for the mercantilism model and for the conventional model for reference. Despite the built-in disadvantages outlined above, mercantilism shows a significant and strong effect on growth—the initial level of mercantilism is strongly significant and the change in mercantilism is significant, at the 1 percent level, once countries of the former Soviet Union and Eastern Europe are excluded. The overwhelming yet simple conclusion is that mercantilism pays. Now we attempt to assess how well it pays.

Miracle Economies One test of the benefits of mercantilism is whether it is associated with miracle growth. A short-term miracle is defined as the shortest period required to double income per capita; a long-term miracle is the highest residual (and absolute) growth, where the residual attempts to eliminate the positive effects of policies involving currency valuation and other natural 2. ASEAN-7 are Brunei, Indonesia, Malaysia, the Philippines, Singapore, Thailand, and Vietnam. 3. The Dutch guilder was replaced by the euro in January 1999. As explained earlier, separate currency valuations have been estimated for the European economies after 1998.

MERCANTILISM AND MIRACLES 153

154 DEVALUING TO PROSPERITY

Table 10.2

Mercantilism and growth, 1950–2011 Mercantilism index

Model specification

Initial

Change

Adjusted R2

Number of observations

Number of countries

1

All countries

–0.016***

–0.016***

0.340

1,088

135

2

Excluding Russia and Eastern Europe

–0.015***

–0.051***

0.344

999

109

Basic model

Currency valuation

3

All countries

–0.012***

–0.037***

0.354

1,088

135

4

Excluding Russia and Eastern Europe

–0.013***

–0.038***

0.352

999

109

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Dependent variable is growth in income per capita; all models have (log) initial income per capita and time and country dummies. Source: Bhalla (2007a) dataset extended to 2011.

Table 10.3

Growth in GDP and factors of production, 1960–2011 (percent)

Labor

Capital

Total factor productivity growth

Developed economies

1.1

3.4

0.5

3.0

East Asia

2.1

7.9

1.9

7.2

Russia and Eastern Europe

0.5

n.a.

n.a.

2.3

Latin America

3.0

3.6

0.4

3.8

Middle East and North Africa

2.4

4.5

1.0

4.5

South Asia

2.4

5.4

1.6

5.7

Sub-Saharan Africa

3.2

3.9

0.5

3.9

Total

2.0

5.7

1.3

5.3

Region

GDP

n.a. = not available Notes: Total factor productivity growth is calculated on the basis of the share of capital and labor being 0.57 and 0.43, respectively. Source: Bhalla (2007a) dataset extended to 2011.

determinants such as demography and the rate of investment. An economy that invests a lot can grow at a higher rate, but that does not necessarily make it a miracle economy. We can define miracle growth in relation to normal, and potential, growth. Technological change is an important component of trend growth, and the most common name for this effect is total factor productivity growth (TFPG). As the name implies, this is the component of observed actual growth that is not the result of the three known factors of production (land, labor, and capital). Each of these factors can be refined using knowledge about its quality, and the most important quality difference comes via education. Earlier chapters discussed two components of potential growth, reallocation and catch-up. The former is about growth that occurs as a result of the reallocation of labor from agriculture to industry; the latter is about growth resulting from the adoption of technology that is closer to the frontier. The first occurs in the early stages of development, and the second occurs in the later stages. The contribution of reallocation and catch-up is between 1 and 2 percent per year, and both are part of TFPG. In the early stages of development, TFPG is about 10 to 20 percent of total growth; when a country becomes developed, the TFPG contribution to total growth increases to 60 to 80 percent. By definition, growth is the sum total of contributions from capital, labor, and TFPG.4 Table 10.3 reports the historical averages. Expected world GDP

4. Estimates of TFPG used in the computations are borrowed from Bhalla (forthcoming); TFPG is obtained from a traditional growth model with capital and labor as inputs. An all-country Cobb-

MERCANTILISM AND MIRACLES 155

growth is about 5 to 5.5 percent per year (4 percent being the total contribution of labor and capital and 1 to 1.5 percent the TFPG contribution). For developing economies, the higher contribution of capital yields average GDP growth of about 6 percent per year. For developed economies, an annual rate of growth of GDP of about 3 percent can be considered normal. Anything higher is better than normal. These averages are for the post-1960 period. For the post-1980 period, developing countries grew somewhat faster (primarily because of accelerated growth in the two populous nations, China and India) and the developed economies considerably slower. Can we use the rate of GDP growth per capita to determine the beginning of miracle growth? Statistically, during the postwar period, excluding oilexporting economies and countries with very small populations, 57 countries have experienced, at least once, a 10-year growth rate in GDP per capita above 5 percent per year (table 10.4). When the cutoff is set to growth rates of 6 and 7 percent, the number of countries falls geometrically to 25 and 7, respectively. With a total of 720 episodes, that translates into a probability of less than 3.5 (at 6 percent) and 1 percent (at 7 percent). In order to define a miracle as a low-probability black swan event, a growth rate in GDP per capita of 7 percent per year seems reasonable; to make it a more ordinary “white swan” event, rates between 6 and 7 percent can be used. The Commission on Growth and Development (2008) identifies 13 countries with growth rates above 7 percent for 25 years. This is roughly equivalent to growth in income per capita of 5 to 5.5 percent per year. As noted, there are 57 such cases of countries sustaining such rates over a 10-year period, and these include the same countries two or three times (such as China and Korea). There are several examples of fast growth, including Botswana, China, and Ireland recently, and the United Kingdom and United States during the 19th century. The common factor connecting these disparate sets of countries and circumstances is ubiquitous currency undervaluation. In fact, once the advantage of currency undervaluation has been taken into account, many black swan miracles turn into ordinary white swans—that is, their exceptional growth is largely explained by currency undervaluation. A true black swan is a country that grew at a miracle pace without the help of real devaluations. One method of defining miracle growth is to find the shortest period of time required to double income per capita. For most countries, this takes a long time, and some have barely achieved this once during the last 50 years. Excluding oil-exporting economies, there are 58 episodes when countries doubled their income per capita in between 5 and 14 years. Singapore doubled its income per capita in only 5 years; Botswana in 6; China in 7; and Japan in 8. Other countries with relatively fast income growth include Ireland (10) and India (13). In contrast, Peru took 21 years and the Philippines took 23 years. Douglas regression yields the share of capital to be 57 percent and the share of labor to be 43 percent. See Bhalla (forthcoming).

156

DEVALUING TO PROSPERITY

Table 10.4

Miracle growth, 1951–2011

Region

1951–79

1980–2011

Number of countries with decadal growth above:

Number of countries with decadal growth above:

5 percent

6 percent

7 percent

Total (1951–79)

5 percent

6 percent

7 percent

Total (1980– 2011)

Total (1950– 2011)

Developed economies

7

5

4

16

1

1

0

2

18

East Asia

6

4

4

14

11

7

5

23

37

Eastern Europe and former Soviet Union

2

2

1

5

16

10

7

33

38

MERCANTILISM AND MIRACLES 157

Latin America

4

3

1

8

5

2

1

8

16

Middle East and North Africa

4

0

0

4

3

1

0

4

8

South Asia

0

0

0

0

1

1

0

2

2

Sub-Saharan Africa

6

3

3

12

7

2

1

10

22

29

17

13

59

44

24

14

82

141

Total

Notes: Growth is per capita and measured by purchasing power parity, Penn World Table 6.1; the decade average has to exceed the level indicated to be counted as a miracle. Sources: Growth data are from Penn World Table 6.1 (Heston, Summers, and Aten 2002); author’s calculations.

Does currency valuation help speed up the process? Equations 10.1 and 10.2 show a simple regression. Doubling time = 19.0 + 0.084 × CV + 2.15 × dCV (17.3) (5.8) (6.1)

(10.1)

Number of observations = 78, R2 = 0.41 where the dependent variable is the number of years taken to double income per capita and the independent variables are the average level of currency valuation (CV) and average change (dCV) during the doubling period.5 Excluding the structural change economies of the Soviet Union and Eastern Europe: Doubling time = 20.4 + 0.091 × CV + 2.68 × dCV (17.3) (6.9) (5.9)

(10.2)

Number of observations = 68, R2 = 0.47 The average time required to double income per capita is 20.4 years without the aid of currency undervaluation. Each 10 percentage points of currency overvaluation increases the doubling time by 0.91 years. An average decline in currency valuation of only 1 percent per year decreases the doubling time by 2.7 years. A country grows faster with a cheap, and cheapening, currency. Figure 10.1 plots the predicted and actual time (in years) needed to double income per capita for countries that have done so in less than 19 years. Botswana, Japan, Germany, and Malaysia are among the countries close to the line. Figure 10.2 plots the relationship between the time required to double income per capita and the contribution of currency valuation to the process. The relationship is clear—it helps countries to undervalue if they want to grow faster. But other factors are at work. The period required for income doubling helps define short-term miracles. How can we identify long-term growth miracles? We can extend the growth relationship outlined in chapter 4 to identify excess growth beyond what is indicated by initial conditions, demography, investment, and currency valuation. The independent variables in the model explaining growth include the following: initial log income per capita, initial share of population in the 15–64 age group, and initial currency valuation in each five-year period. There are two additional variables, average change in currency valuation and the share of investment in GDP. The residual in the estimated fixed effect equation is indicative of relative productivity growth; the higher this country effect, the greater the miracle.

5. Sample comprises selected countries, excluding the oil-exporting economies, bad data economies, and small economies; see appendix A.

158

DEVALUING TO PROSPERITY

Figure 10.1

Doubling per capita income, helped by currency undervaluation

predicted number of years 30 idn

25

cmr tun

irl

lbn

bel khm fin

dnk lao

mex

mar

20

pan

bra kor oan tha

15

mys

eth moz

prt

nld dom fra

ita aut ind isr

deu

chl uga

vnm

jor jpn

10

egy

bwa

grc mus esp hkg

5 chn

0 5

7

9

11

13

15

17

19

actual number of years Notes: The model relates the minimum number of years taken to double per capita income and the predicted number of years. The prediction is based on a model that relates doubling years to the average level of and average change in currency valuation during the doubling period. See equation 10.2 in text. Source: Bhalla (2007a) dataset extended to 2011.

(10.3) Growth in per capita income = (0.79) + (–3.68) × LogY + (–0.018) × iCV + (–11.1) (–5.9) (–0.059) × dCV + 0.13 × Dep + 0.14 × Inv (–4.0) (3.66) (8.3) where Y is initial per capita income, iCV is the currency valuation in the first year of each five-year period, dCV the average change in currency valuation, Dep the initial worker population ratio, and Inv the share of investments in GDP. Adjusted R2 is 0.40, the number of observations is 708, and the sample is 74 countries (oil-exporting countries and those belonging to the former Soviet Union and Eastern Europe are excluded). Time period of analysis is five-year data from 1960 to 2011, with 2010 and 2011 forming the last “five-year” sample. Again, currency valuation plays a significant and consistent role in the growth process. Despite inclusion of the share of investment, the initial level MERCANTILISM AND MIRACLES 159

Figure 10.2

Added variable plot: Number of years taken to double per capita income and average value of currency valuation

number of years taken to double per capita income

30

gmb nzl bfa

20

gha mwi gbr

10

0

blr

–10

chn hkg

uga

arg mus uzb geo egy isr esp grc vnm jpn tjk jor bwa

zaf col aus usa npl can nic pry che cri ury swe dnk phl lka nldper mex bgd tur cmr belpol svk lbn pan lao fin fra pak dom mys ltu ind ita chl aut mar tun bra moz deu est prt khm idn irl eth lva tha kor oan

ken

tza

alb

–20 –7

–5

–3

–1

1

3

5

average value of currency valuation prevailing in those years coefficient = 2.1469911, (robust) standard error = .35396904, t = 6.07 Notes: The model relates the minimum number of years taken to double income per capita and the predicted number of years. The prediction is based on a model that relates doubling years to the average level of and average change in currency valuation during the doubling period. See equation 10.2 in text. This added variable plot relates number of years to average level of currency valuation. Source: Bhalla (2007a) dataset extended to 2011.

of currency undervaluation and its changes are strongly significant and show magnitudes similar to those observed earlier (–0.018 and –0.060 for the initial level and average change, respectively). Higher rates of investment help: There is extra growth of approximately 1.4 percent per year for each 10 percentage point increase in the share of investment. This fixed effects model yields coefficients of excess growth for each individual country—that is, growth effect specific to each country after incorporating the role of the known determinants of growth. The values are not estimates of productivity growth, but the fixed effect estimates are indicative of the rank order of productivity in the different economies. Since the period of estimation is large (50 plus years), this indirect estimate can be considered an estimate of the extra, miracle contribution to growth, analogous to a TFPG estimate; hereafter, an indirect estimate of TFPG. There are thus two estimates of TFPG available—a production function estimate (see note 4) and a fixed effect estimate. Average per capita income growth is a third estimate of productivity; however, this estimate is clouded 160

DEVALUING TO PROSPERITY

by the presence of variables (e.g., investment) that, while contributing to extra growth, do not contribute toward a miracle. Recall that this was the criticism of Paul Krugman (1994) and Alwyn Young (1995) about the fast growth in East Asian economies—i.e., that there was high growth there, but high growth because of high investment. Nevertheless, per capita income growth as an indicator of productivity growth has appeal, especially because it is easily computable—and widely used. The three estimates of productivity and growth are combined to form a final ranking of miracle economies. The procedure was as follows. First, a rank is computed for each indicator; second, the ranks are assigned the following weights: 25 percent for per capita growth and 37.5 percent to each of the TFPG indicators. This weighted rank is the ordering of miracle growth economies. Table 10.5 presents the results. Data are reported for the average levels of productivity growth, the ranks, as well as the mean rate of investment. The miracle is defined as exceptional growth, all things considered, and long term is defined as more than 50 years. The list is headed by Taiwan, followed closely by Ireland, Hong Kong, Israel, Finland, and the United States. Ireland has been in the news recently because of the problems in the euro area. It has an exceptional long-term growth record—mean growth of 3.5 percent per annum since 1960, compared with 2.7 percent for 5th ranked Finland and 2.1 percent for 6th ranked United States. And it did so with an average investment rate of only 20 percent compared with 26 percent for 3rd ranked Hong Kong. The fact that the United States, a highly developed (and the most advanced) economy, grew at a rate of 2.1 percent per capita per year for more than 50 years is miraculous. The comparison between the United States (ranked 6) and China (ranked 14) is revealing. The United States ranks higher because the measure is not absolute growth but exceptional growth, and exceptional growth is a matter of making do with less. Average investment rate for the United States, 19 percent per annum, is about half the average investment rate for China. China’s average rate of decline in currency valuation is the fastest among the 74 economies considered, –4.4 percent per annum, compared with –0.4 percent for the United States. Another reason China is as “low” as 14th most miraculous economy is because of its high(est) investment rate among all the countries—an average of 35.2 percent for 50 years plus, with the second highest investment rate observed for Singapore, 31.3 percent, and 29.6 percent for Japan. Two decades of lost growth lower Japan’s ranking, several notches below Portugal (rank 25 and 19, respectively). Japan’s average per capita growth rate of 3.6 percent per annum is reasonably high, but not as much a miracle because it was made possible with a very high investment rate. This has been an exploratory but revealing exercise. The analysis confirms the perception that East Asian economies were growing faster because of extra investment; however, some countries invested heavily and were efficient. Four of the top 10 miracle economies are East Asian—Taiwan, Hong Kong, Singapore, and Korea. Some not-so-surprising surprises—Chile (the best performing MERCANTILISM AND MIRACLES 161

162 DEVALUING TO PROSPERITY

Table 10.5

Country (1)

Miracle growth, 1960–2011: Results (and associated data)

Miracle rank (2)

Average investment rate (percent) (3)

Average (percent) Per capita income growth (4)

Relative productivity (fixed effect) (5)

Rank for Total factor productivity growth (6)

Per capita income growth (7)

Relative productivity (fixed effect) (8)

Total factor productivity growth (9)

Taiwan

1

25.4

5.7

4.0

0.7

2

3

17

Ireland

2

20.0

3.5

4.2

0.8

10

2

13

Hong Kong

3

25.7

5.0

3.9

0.4

5

4

27

Israel

4

22.0

2.6

3.1

1.1

26

12

6

Finland

5

24.0

2.7

3.0

1.0

21

16

9

United States

6

19.2

2.1

4.7

0.7

36

1

15

Mauritius

7

26.4

3.0

0.6

1.8

17

32

2

Singapore

8

31.3

5.7

3.2

0.3

1

10

36

Sweden

9

21.4

2.1

3.6

0.7

37

7

16

Korea

10

28.8

5.3

2.2

0.4

4

22

25

Chile

11

21.0

2.5

1.0

1.1

28

27

7

Australia

12

26.4

2.1

3.2

0.5

35

11

21

France

13

23.9

2.3

3.3

0.4

31

9

26

China

14

35.2

5.4

–2.2

1.9

3

53

1

Italy

15

22.2

2.4

2.8

0.5

30

18

20

Tunisia

16

25.6

3.3

–0.3

0.8

13

38

12

Turkey

17

19.1

2.3

0.3

1.2

32

34

5

Belgium

18

25.2

2.5

3.0

0.4

29

15

29

Portugal

19

25.4

3.1

1.5

0.4

15

26

28

Netherlands

20

23.0

2.2

3.1

0.4

34

13

30

Botswana

21

28.3

4.8

1.8

0.2

6

24

38

Uruguay

22

17.8

1.8

0.3

1.3

43

35

3

United Kingdom

23

18.2

1.9

3.4

0.3

41

8

33

Denmark

24

21.6

1.9

3.6

0.2

40

6

37

Japan

25

29.6

3.6

2.9

0

9

17

48

Austria

26

25.1

2.8

3.0

0.1

20

14

44

Argentina

27

21.5

1.7

0.9

0.8

48

29

11

MERCANTILISM AND MIRACLES 163

Greece

28

23.7

2.8

1.8

0.3

19

25

35

Germany

29

24.7

2.2

2.6

0.4

33

20

31

Spain

30

23.4

3.1

2.5

0.2

16

21

43

India

31

23.5

3.4

–2.5

1.0

12

59

8

Malaysia

32

25.6

3.6

1.0

0.1

8

28

45

Egypt

33

21.1

2.6

–1.1

0.5

25

46

19

Thailand

34

26.9

4.5

–0.6

0.2

7

42

40

Dominican Republic

35

18.5

3.1

0.6

0.1

14

31

47

Canada

36

22.2

2.0

3.6

–0.2

39

5

57

Sri Lanka

37

22.2

2.7

–2.9

0.8

23

61

14

Brazil

38

18.8

2.6

0.7

–0.1

24

30

54

Kenya

39

19.4

1.2

–2.5

1.2

60

58

4

Indonesia

40

26.4

3.5

–2.3

0.2

11

55

41

Notes: Column 6 (total factor productivity growth) is based on a production function model with average share of capital estimated to be 57 percent and average share of labor estimated to be 43 percent. The fixed effect relative productivity estimate (column 5) is from equation 10.3 in the text. Column 7, weight of 0.25, and columns 8 and 9, weight of 0.375, are combined to yield miracle rank in column 2. Source: Bhalla (2007a) dataset extended to 2011.

economy of Latin America) and Australia—are ranked highly at 11 and 12, respectively. The miracle economy of Africa, Botswana, is ranked 21st, despite the fact that its average per capita growth rate was 4.8 percent per annum—a growth rate made possible by a high investment rate, 28.3 percent, and a fifth fastest average rate of decline in currency valuation, –2.7 percent per annum.

11 Institutions versus Exchange Rate Policy

Believe those who seek the truth. Doubt those who find it. —André Gide, French author and 1947 Nobel Prize winner in literature Theories about the determinants of growth come and go. Over the years, there have been three broad trends. The first was that geography was seen to dictate the fortunes of nations. The second, made popular by John Maynard Keynes in the mid-1930s, was that governments could affect the shape of things to come through policy. This theory held sway until the end of the 20th century, when a new theory was championed—the role of institutions. The crisis of 2008 and the Great Recession that followed led to a resurrection of the Keynesian emphasis on policy. As an explanation for long-term trends in economic growth, however, this revival is likely to be short lived. And attention is likely to turn back to the role of institutions. Good institutions help facilitate business, establish fair rules of the game, and enhance competition. There is some circularity to this logic—that is, countries that are wealthy today are also likely to have had good institutions, however defined, in the past. Could it be that higher incomes result in better institutions, rather than the other way around? A voluminous amount of research in recent years has addressed this very question and has helped to break the circularity by isolating the causative role of institutions in generating higher growth.1 The pioneers in this effort were Robert Hall and Charles Jones (1999), and Daron Acemoglu, Simon Johnson, and James Robinson, who wrote the first article in a series of papers in 2001. Others have used their data and found similar results. Hence, the near consensus that institutions “rule.” Today, any model testing a potential determinant of growth either theoretically or empirically is benchmarked by the inclusion in the model of institutions. 1. This literature has exploded in the past decade, especially since 2006. Rare is a conference without some paper detailing the virtues of institutions or governance.

165

This chapter tests the importance of institutions against the main arguments of this book: that currency valuation is a major determinant of growth, that some key countries are following the policy of undervaluing their way to prosperity, and that this policy has been taken to a mercantilist extreme. This book makes the case that currency undervaluation works and works so well that policymakers cry foul when any country engages in a “currency war.” This chapter subjects that hypothesis to the test against institutions. How do institutions perform in the presence of currency valuation? And does currency valuation matter in the presence of good institutions?

The Conventional Wisdom The role of institutions in growth has been analyzed in many different ways and has involved all the important determinants of growth—geography, demography, and policy. The near universal conclusion has been that institutions matter, and that policy does not. The following is representative of this conclusion that “institutions rule.” We estimate the respective contributions of institutions, geography, and trade in determining income levels around the world, using recently developed instruments for institutions and trade. Our results indicate that the quality of institutions “trumps” everything else. Once institutions are controlled for, measures of geography have at best weak direct effects on incomes, although they have a strong indirect effect by influencing the quality of institutions. Similarly, once institutions are controlled for, trade is almost always insignificant, and often enters the income equation with the “‘wrong” (i.e., negative) sign, although trade too has a positive effect on institutional quality. (Rodrik, Subramanian, and Trebbi 2002, i)

In fact, the overwhelming weight of opinion is not only that institutions matter most, but that they may be the only thing that matters: [W]e are struck by the way that endowments and policies have no independent effect once we control for institutions, contrary to a number of stories, and that institutional quality seems to be a sufficient statistic for accounting for economic development. (Easterly 2002, 33)

In its survey of studies pertaining to the effect of institutions on growth, the International Monetary Fund concludes: The key finding from the empirical analysis…is that institutional quality has a significant impact on economic performance. This result holds whether performance is measured by cross-country difference in the level of income per capita, in growth rates, or in the volatility of growth. Specifically, improvements in institutions lead to higher incomes, stronger growth, and lower volatility. These results are quite robust and are independent of the specific measure of institutional quality adopted: similar results emerge whether

166

DEVALUING TO PROSPERITY

one focuses on political, legal, or economic institutions. Moreover, the relationships hold across all the main regions, and are not driven by one or two specific country groups…. The analysis also indicates the presence of “catchup” or convergence effects. While countries at all levels of development would benefit from stronger institutions, the impact of institutional improvement growth appears to be strongest for countries starting from a lower level of economic development. This result further emphasizes the need for institutional strengthening to be at the forefront of efforts to improve growth and reduce poverty, particularly among the low-income countries. A key question then is how to create a “virtuous circle” whereby policies are put in place to strengthen institutions, and stronger institutions help support and sustain better policies. (IMF 2003, 111–12)

The emphasis on institutions displaced the role of policy, as noted in several key studies: A striking aspect of these regressions is the relatively minor evidence of a direct role for conventional government policies. Instead, the most important determinants of growth appear to be factors that cannot be changed substantially in the short run. (Bosworth and Collins 2003, 159) The new growth literature, using both endogenous growth and neoclassical models, has generated strong claims for the effect of national policies on economic growth. Empirical work on policies and growth has tended to confirm these claims. This paper casts doubt on this claim for strong effects of national policies, pointing out that such effects are inconsistent with several stylized facts and seem to depend on extreme observations in growth regressions. (Easterly 2005, 2)

The World Bank reached a similar conclusion about institutions but held out hope that government policies can work—provided they operate within good institutions! Recent empirical research has found that, when a measure of institutional quality is included in cross-country regressions, the explanatory power of other variables, including all measures of policies, becomes negligible…. This reasoning suggests that good institutions matter more for growth than do good policies. From a syndrome viewpoint, it is easy to see that this is not an assertion that “policies don’t matter”—of course they do. Rather the question is whether good policies can be sustained and implemented in the absence of adequate public sector organizations and institutions. (World Bank 2005, 49–50)

This brief survey of the literature documents the near-universal acceptance of the conventional wisdom that policies are not important unless they operate within an improved institutional setting, and that the number one policy choice facing developing economies today, especially the poorest ones, is how best to import, introduce, and/or improve largely Western-style institutions.

INSTITUTIONS VERSUS EXCHANGE RATE POLICY 167

New Evidence It is worth reexamining the evidence in favor of institutions and its conclusions. And the best instrument for doing so is policy—specifically, policy that promotes exchange rate undervaluation. Daron Acemoglu, Simon Johnson, James Robinson, and Yunyong Thaicharoen (2003) specifically examine the effect of currency valuation on growth. They use the Dollar-Easterly measure for exchange rate valuation, which, as previous chapters show, is likely to contain large measurement errors.2 As a result, Acemoglu et al. (2003) may have mistakenly concluded that policy in the form of exchange rate valuation does not matter and that institutions do matter. In addition to reestimating their model using the currency valuation measure introduced here, various other tests are conducted of the proposition that institutions matter for growth. Testing for the role of institutions is complicated by the problem of simultaneity—institutions affect growth, and vice versa. In contrast, geography is truly exogenous, and so there is no simultaneity between geography and growth—growth does not cause rainfall or sunshine, although rainfall can affect land productivity and therefore standards of living. This is almost the same for government policy, especially over long periods. Not surprisingly, therefore, research has concentrated on identifying variables that affect one of the simultaneously related variables but not the other. In this instance, the challenge is to explain differences in current standards of living, and so there is need for a variable that affects institutions but not economic growth. Over the last decade, several instruments and measures have been identified. Each has its advantages and disadvantages. The approach taken here is to assemble a large and varied set of measures and instruments, to pair them with each other, and then to evaluate the net importance, or effect, of each institution proxy.

Institutional Measures There are two broad categories of institutions, representing the presence of a particular political or economic freedom. And there are several measures, or proxies, for these two types of institutions, or their combinations.

Political Freedom The oldest political institution measure is the Gastil index of political freedom published by Freedom House (2011). Separate indices are available for political and civil liberties from 1973 onward. In recent years, another such measure, the Polity IV dataset maintained by the University of Maryland, has gained in popularity. It has several different measures of political freedom such as the degree of autocracy and the magnitude of executive constraints, among others. 2. See chapter 4 for a brief background on Dollar-Easterly.

168

DEVALUING TO PROSPERITY

For some measures, and some countries, the data go back at least 100 years. In 1996, the World Bank also started to publish indices of governance, including subindices of such indicators as the rule of law. The argument about the importance of political freedom is straightforward and best explained by reference to a counterfactual. In a society with low levels of political freedom, the welfare of the people is a function of both how benign and how forward-looking the dictator is. The probability of having both qualities in the dictator is very small. Simply put, for every good dictator there may be 10 bad dictators. The odds for economic success must be higher in societies that can throw out bad leaders. This is also the logic behind the executive constraint measure in the Polity dataset—the presence of checks and balances. Related to the concept of checks and balances are other measures that gauge legal and bureaucratic structures.3 Different legal systems can affect the incentive structures. For example, the British system may provide greater impetus to private entrepreneurship than the French system. If the legal system is friendly to litigation, then the costs of economic transactions (the costs of doing business) may rise, and growth may decline. Analogously, when citizens have to jump through several bureaucratic hoops (and maybe bribe several officials) to start a new business, economic activity is slowed, and investment and output are likely to be lower, other things equal.

Economic Freedom The second and possibly the most important category of institutions involves those that enhance economic freedom; within this set, institutions that preserve private property rights are prominent. Individuals will neither lend nor invest if they feel that there is a high probability that their assets (investments) will be expropriated. Institutions (or laws) that help preserve property rights are expected to be conducive to private gain, investment, and growth. Freedom House started publishing an economic freedom index in 1997, and both the Heritage Foundation and Fraser Institute also publish indices of economic freedom. The original institutions index (a form of economic freedom) is the expropriation index published by the International Country Risk Guide (ICRG) (PRS Group, various years), which is used by both Hall and Jones (1999) and Acemoglu et al. (2003).

Colonial Heritage I was among the first to examine, using instrument variables, the joint relationship between political freedom and growth (Bhalla 1997a).4 The argument 3. Rafael La Porta, Florencio López-de-Silanes, Andrei Shleifer, and Robert Vishny (1998) evaluate the impact on growth of various legal institutions (Anglo-Saxon, German, and Scandinavian). 4. The 1997 publication date of the article is somewhat misleading, given that it was presented at the Nobel Symposium Conference in September 1994.

INSTITUTIONS VERSUS EXCHANGE RATE POLICY 169

is that political freedom affects growth and growth most likely increases the demand for political freedom. An “identification” variable is needed to break the simultaneity. There was reason to believe that the presence or absence of colonialism and the nature of the institutions of the colonizer in the 19th century affected the degree of political freedom in the colonized economy after World War II. Hence, colonization is used as a “perfect” identifying variable, introducing colonialism as an identifier for the institution of political liberties. Further, colonialism is partitioned into three separate categories: colonized by the British but residing in Africa, colonized by Britain but residing outside Africa, and colonized by the French. Colonial heritage has since become a popular identification variable, particularly for institutions. Colonial heritage in a slightly varied form is the basis of the “legal origin” identification variable used by La Porta et al. (1998) to classify countries by legal system. There is a strong correlation between a country’s colonial power and its legal system, but there are some exceptions, including uncolonized countries such as Ethiopia and Thailand and former British colonies such as Egypt, which ended up using the French legal system. For Acemoglu et al. (2003), the identification variable is not colonialism but a variable strongly related to colonialism—the settler mortality rate, or the rate of mortality of those who settled in the colonies. This mortality rate determined whether white colonialists settled in the colonies over the long term (as in the United States) or did not (as in Ghana and India). They examine the colonial heritage of four developed economies—Australia, Canada, New Zealand, and the United States—and consider them to be the same as, say, India. Another identification variable they use is the fraction of white settler population in about 1900. Indices of ethnic fragmentation also have been used to identify institutions. The null hypothesis is that if a society is ethnically fragmented, then it is more liable to have bad institutions and therefore lower growth. Acemoglu et al. (2003) also use the population density in 1500 as an identification variable—the larger the population density in 1500, the less likely the invaders were to settle, because of competition from the existing population. The hypothesis is that in more densely populated countries there was a lower probability that white settlers would stay for the long term and therefore a higher probability that institutions would have less capacity. These two variables (or variants thereof)—colonial heritage and ethnic fragmentation—are now standard in the literature. There are also two new identification variables—education and middle class in 1850.5

Instruments for Measuring Institutions Seven institutions were selected for analysis. Three of these are popular political institution indices: two measures from the Polity data (the overall polity 5. I explore the importance of these variables in Bhalla (forthcoming).

170

DEVALUING TO PROSPERITY

index and the index on executive constraints) and one from Freedom House (the average of political and civil liberties). Two economic institution variables were selected: the ICRG index of expropriation risk6 and the economic freedom index published by Freedom House. Finally, two composite institution indices (combining economic and political freedom) were selected: the World Bank composite index of governance and the Heritage Foundation index.7 Eight instruments were selected. Four instruments are directly or indirectly related to colonialism: colonial heritage per se (whether colonized or not), settler mortality, legal origin, and the fraction of white settler population in 1900. One additional instrument is indirectly related to colonial heritage, population density in 1500. The sixth instrument is ethnic fragmentation. The seventh and eighth instruments are the mean education levels in 1850 and the fraction of the population that was middle class in 1850.

Institutions and Growth: Revisiting the Evidence As introduced and discussed by Acemoglu et al. (2003), the econometric relationships are as follows. The level of income per capita is a function of institutions and other determinants: log Yi = D + E × CIi + G × Xi + Hi,

(11.1)

where Y is income per capita, CI is the proxy variable for the presence of current institutions, X is a vector of other covariates, and Hi is a random error term. It is very likely that CIi and Yi are strongly correlated, thus resulting in a simultaneous equation bias. However, if a variable is found that is related to CI but not to Y, then use of this instrument variable can yield unbiased estimates of the coefficient of interest, namely, E. And if E is significant, then we have proof that institutions help determine how rich or poor a particular country is at any point in time. Acemoglu et al. (2003) choose the settler mortality rate as their preferred instrument for the institution measure. In a rather detailed examination of the Acemoglu et al. (2003) settler mortality data, David Albouy (2006) offers new estimates of settler mortality for a few countries. He concludes that their finding of a significant role for institutions is extremely sensitive to their settler mortality data and to the inclusion of two additional geography variables: average temperature and the minimum monthly rainfall in a year. Albouy found their results to be fragile with respect to the statistical significance of institutions (expropriation risk) in an equation explaining levels of income per capita around 2000. 6. This is the ICRG variable used by Acemoglu et al. (2003) and others; it pertains to the level of expropriation risk in the 1980s. See appendix A for details. 7. Both the Heritage Foundation indices and the Freedom House data on political and civil liberties—the Gastil index—rank countries from good to bad from 1 upward; that is, the higher the number, the worse the index. These have been converted to conform with the notion that a higher value corresponds to better institutions.

INSTITUTIONS VERSUS EXCHANGE RATE POLICY 171

Table 11.1

Institutions and growth, 2010 and 1980–2010 Instrument = log (settler mortality rate in 19th century) Institution = Executive constraint

Model Model 1L

1.61***

Model 1G

1.04***

Model 2L

1.46***

Model 2G

0.99***

Model 3L

1.62***

Model 3G

0.98***

Currency valuation Bhalla (2007)

DollarEasterly

Government consumption

Average inflation

0.007 –0.024*** 0.015 –0.024***

–0.094*** 0.007

–0.024***

–0.202***

0.075 –0.445

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Models with subscript L have log per capita income in 2010 as the dependent variable; with subscript G, the dependent variable is the growth in per capita GDP, 1980–2010. Model 2 has share of government consumption in GDP and model 3 has average annual inflation as additional independent variables in the two-stage least squares (2SLS) regression. Source: Bhalla (2007a) dataset extended to 2011.

Acemoglu et al. (2003) separately test for the importance of institutions in explaining economic growth rather than the level of income. They consider three macroeconomic indicators as independent variables in an equation explaining growth from 1970 to1997: log (inflation), share of government consumption in GDP, and currency valuation. They find that exchange rate undervaluation is insignificant, and the wrong sign, in explaining growth in their sample of ex-colonies (Acemoglu et al. 2003, table 7). The coefficient value was 0.001 with a standard error of 0.02. Similar results are reported by IMF (2003). In the IMF models, exchange rate undervaluation is only occasionally significant in explaining growth. The IMF uses the same currency valuation variable (Dollar-Easterly) as Acemoglu et al. As pointed out earlier, this measure may contain large measurement errors and therefore is not the measure of choice to test for the relationship between institutions, macroeconomic variables, and growth. Table 11.1 outlines results parallel to those contained in Acemoglu et al. (2003, table 7). The macrovariables employed are the same: currency valuation (model 1), share of government consumption in GDP (model 2), and (log) average inflation (model 3). Each model is tested with the same institution variable (executive constraints from the Polity data) and the same identification variable ((log) settler mortality). The only additional variables in the model are the currency valuation variables from Bhalla (2007a). The Acemoglu et al. (2003) results are confirmed for the lack of significance using the Dollar-Easterly currency valuation variable. However, the nonlinear Bhalla (2007a) currency valuation variable is always significant, often at the 1 percent level of confidence, and it has nearly the same coefficient across models. These results indicate that both institutions and currency valu-

172

DEVALUING TO PROSPERITY

ation matter for growth. This result is for one combination of instruments, institutions, and sample selection. If the four developed economies—Australia, Canada, New Zealand, and the United States—are excluded from the Acemoglu et al. sample, the institution variable is rendered insignificant at even the 10 percent level of confidence in regressions involving the Bhalla (2007a) currency valuation variable. The results are striking in rejecting the “institutions rule” theory. Of 12 models tested, the institution variable is not statistically significant in two models; it is statistically significant at the 10 percent level in five models, at the 5 percent level in four models, and at the 1 percent level in only one model. In contrast, the Bhalla (2007a) currency valuation variable is always significant, and often at the 1 percent level. In no specification does the Dollar-Easterly real exchange rate variable even approach significance. However, this may not be a proper test of the impact of currency valuation since the variable was entered in a form involving construction bias. Tests without this bias are reported below. A complete evaluation of the hypothesis that only institutions matter, that only currency valuations matter, or that both matter, involves a series of tests using various combinations of institutions, instruments, and sample sizes. In addition, models are evaluated with two different, but related, variables. The level of the standard of living at a point in time is simply the cumulative growth over a long period of time. Consequently, two different variables are used to represent development in a country: the level of income per capita in a particular end year (2010), or the rate of growth per capita over a long period (1980 to 2010). In addition, three different sample selections are made: first, the Acemoglu et al. (2003) 64-country dataset; second, data for 136 countries (excluding small countries and those with data problems, see chapter 2); and third, the developing-country dataset, which is an 88-country subsample of the 136 countries. For each institution index and each instrument, nine models are estimated. The first four models pertain to regressions where the dependent variable is the “terminal” or 2010 level of (log) income per capita. For these level regressions, the first model (model 1) is the base regression, that is, log(Y) = a + b × Institution, estimated as an instrument variable regression with one selected instrument; the second model adds latitude, the third model adds temperature and average rainfall (but no latitude), and the fourth model adds latitude, temperature, and rainfall. With growth as the dependent variable, two models comprise the estimation of the growth model without (model 5) and with initial currency valuation (model 6). Growth Y ’ = a + bo × Y0 + b × Institution + c × initial currency valuation, (11.2) where Y0 is log income per capita in 1980, Y ’ is growth per capita for 1980– 2010, and initial currency valuation is the level in 1980. To this model 6, three

INSTITUTIONS VERSUS EXCHANGE RATE POLICY 173

other models are added as for the level regressions reported above—latitude, temperature and average rainfall, and latitude plus temperature plus rainfall (models through 9). This accounts for a total of nine models. Four of the models (6 through 9) involve initial currency valuation. To each of these four models, mean change in currency valuation is added, with the average change in the worker population ratio as an instrument (models 10 through 13). The goal is to comprehensively evaluate the several indices of institutions and the several measures of instruments. Separate regressions are run for each separate match of institutions and instruments; in the end, results are evaluated for the entire set of models/regressions both with and without the presence of currency valuation variables. There are thus a total of 13 models, 7 institutions, 8 instruments, and 3 datasets, for a total of 2,184 regressions. In addition, the Hansen overidentification (exclusion criterion) test involved pairing each instrument with another. Tables 11.2 and 11.3 present the summary results for institutions and instruments, respectively. The results are presented in terms of the percentage of models passing each stage, and three stages are reported. Stage 1 is whether the instrument was statistically significant in explaining variations in the institution in the first stage of the two-stage least squares regression. For example, in 71 percent of all models estimated, the instruments were successful in explaining the World Bank institutions index.8 In stage 2, the World Bank institutions index is significant in 49 percent of the cases. A joint pass of both stages reduces the success rate to 43 percent. This ratio is further reduced to 29 percent because of the rejection of models in which the selected instrument affects the institution selected as well as the level (or growth) of income per capita. This third and final stage is a robust test for the influence of institutions on growth. The World Bank institutions index and the economic freedom indices represented by the Fraser Institute and ICRG (expropriation risk) perform best, with an aggregate success ratio between 29 and 30 percent. The political freedom variables perform worst, with a success ratio in the low to middle teens. Among the instruments, the most successful is the share of the middle class in 1850 (32 percent) followed by colonial heritage, 24 percent (table 11.3). The mortality index introduced by Acemoglu et al. (2003) is significant in only 21 percent of cases, about the same ratio as ethnic fragmentation and the overall average. In only 22 percent of cases are institutions a significant factor in explaining growth. This is considerably less than the prior expectation that institutions rule. Some models have policy variables (currency valuation) and some do not. The presence of these policy variables—CV or initial currency valuation in 1980 and dCV or mean change in currency valuation between 1980 and 2010— further sharply dents the importance of institutions (table 11.4). In models without the policy variable, the institutions variable is significant in 43 percent 8. Success is taken to be significance at the 5 percent level of confidence.

174

DEVALUING TO PROSPERITY

Table 11.2

Importance of institutions in explaining growth (pass percent)

Institution/source

Stage 1: Ordinary least squares

Stage 2: Two-stage least squares

Stages 1 and 2

Final pass percentage

48

32

29

22

78

48

45

30

All Economic freedom Fraser Institute Heritage Foundation

33

30

23

21

World Bank

71

49

43

29

International Country Risk Guide (ICRG)

66

44

40

30

Freedom House

27

17

15

13

Polity IV: Democracy

28

16

15

13

Polity IV: Executive constraints

32

21

19

17

Political freedom

Notes: Stage 1 implies that the coefficient of instrument for institution is significant in the first stage of the OLS regression; stage 2 implies that the coefficient of institution is significant in the second stage of the 2SLS regression; stages 1 and 2 imply that coefficients in both stages are significant; final pass percentage shows the percentage of models that pass the Hansen overidentification test.

Table 11.3

Importance of instruments in explaining growth (pass percent)

Instrument

Stage 1: Ordinary least squares

Stage 2: Two-stage least squares

Stages 1 and 2

Final pass percentage

All

48

32

29

22

Colonialism

54

34

33

24

Ethnic fragmentation

28

41

25

20

Legal origin

42

12

12

9

Log settler mortality

31

38

31

21

Log population density in 1500

33

16

14

13

Mean years of education in 1850

71

25

24

20

Percent middle class in 1850

55

39

36

32

Percent white settlers in 1900

57

30

27

21

Note: See table 11.2. Source: Bhalla (2007a) dataset extended to 2011.

INSTITUTIONS VERSUS EXCHANGE RATE POLICY 175

Table 11.4 Currency valuation versus institutions Percent of coefficients significant Currency valuation

Model type

Number

Institutions

Initial valuation (1980)

All models

4,204

79

Only institutions

1,680

43

Institutions plus initial currency valuation

1,785

18

30

739

18

68

Institutions with initial plus change in currency valuation

Change in valuation (1980–2011)

73

Notes: Institutions are instrumented by the different instrument variables outlined in the text and table 11.3; average change in currency valuation instrumented by log (change in share of worker population). Source: See text.

of the cases; but when these models include currency valuation, the percent of institutional variables passing a significance test drops to only 18 percent—not as universally strong or robust as claimed. Among instruments, the settler mortality variable does not allow institutions to pass even the first stage if policy variables are included. Neither of the two political institutions variables, executive constraints or political and civil liberties, is significant when policy variables are included (these results are not shown). Table 11.5 shows two important results for the currency valuation policy variable. First, for a very large set of models, at least one of the two currency valuation variables (initial level of valuation or average annual change in valuation from 1980 to 2010) is significant. Second, the coefficients do not vary much across the instruments or the institutions chosen to be in the model. Broadly speaking, the coefficient is –0.01 for CV and –0.76 for dCV. The coefficients are very similar to those obtained by a variety of different methods. As an illustration of the large set of regressions/models estimated, table 11.6 reports the specific results for two samples—the Acemoglu et al. (2003) 64-country sample and a developing-country sample; for two models—the level model (with just institution as a right hand side variable) and the growth model (with institution plus log initial income per capita plus currency valuation variables); and for two institutions—expropriation risk and executive constraints. In each case, the instrument used is log settler mortality, from Acemoglu et al. (2003). For average change in currency valuation, the level of fertility is used as an instrument variable. Summarizing the results: The institution variable is more significant in the 64-country sample than in the developing-country sample, and the introduction of the currency valuation variables often renders the institution variables insignificant. 176

DEVALUING TO PROSPERITY

Table 11.5

Average values and significance of currency valuation in the presence of institutions Number of models

Models

Average value, currency valuation

Success ratio (percent)

Initial

Change

Models with only initial currency valuation Coefficient not significant

2,463

61

–0.004

Coefficient significant

1,590

39

–0.015

Models with both initial and change in currency valuation Neither coefficient significant

330

22

–0.013

–0.830

Both coefficients significant

927

61

–0.013

–0.820

56

4

–0.015

–0.580

199

13

–0.011

–0.825

–0.012

–0.760

Only initial currency valuation significant Only change in currency valuation significant Total models/average valuations

5,565

Source: See text.

The chapter started with the proposition that there was near universal acceptance of the result that institutions rule. It ends with the result that not only ruling is out of the question but also the presence of institutions as an important determinant of growth is open to serious question. The result is not that institutions do not matter. Of course they do, and they matter a lot for the well-being of any society. Good institutions are very desirable. The only residual doubt is whether institutions are the result or the cause of growth. The results of this chapter suggest that they are very much the former.

INSTITUTIONS VERSUS EXCHANGE RATE POLICY 177

Table 11.6

Role of institutions in growth

178 DEVALUING TO PROSPERITY

Instruments Log (settler mortality rate in 19th century)

Middle class in 1850

Currency valuation Dependent variable

Institution

Initial

Change

Education in 1870

Currency valuation Institution

Initial

Change

Currency valuation Institution

Initial

Change

A. Institution = Risk Acemoglu et al. (2003) sample (64 countries) Log income per capita (2010)

1.02***

Growth rate per capita (1980–2010)

2.56***

0.77*** –0.011**

–0.3**9

1.22***

0.71*** –0.017***

–0.68***

0.59***

–0.023***

–0.88***

–0.009***

–0.21***

–0.011***

–0.63***

–0.035***

–1.05***

Developing-country sample (58 countries) Log income per capita (2010)

1.39***

Growth rate per capita (1980–2010)

1.09***

–2.53***

4.53*** –0.018**

–0.79**

–0.44***

–0.028***

–1.19***

2.6****

B. Institution = Executive constraints Acemoglu et al. (2003) sample (64 countries) Log income per capita (2010) Growth rate per capita (1980–2010)

1.07*** –0.74***

0.74*** –0.031**

–1.31**

0.84***

–0.017**

–0.85**

0.052**

0.65*** –0.010***

–0.59***

0.53***

–0.026***

–1*****)

–0.06***

Developing-country sample (58 countries) Log income per capita (2010)

1.28***

Growth rate per capita (1980–2010)

0.91***

0.53***

0.25***

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Instrument for change in currency valuation is the level of fertility; see Rose and Supaat (2007).

12 Currency Undervaluation: A Time-Tested Policy for Growth

Study the past, if you would divine the future. —Confucius, Analects of Confucius Currency valuation matters for growth. The evidence offered has pertained to the period since World War II and especially since 1960. The hypothesis that currency valuation matters has been subjected to many tests, and the result is the same. A legitimate question therefore is, If currency undervaluation has worked so well during the past 50 years, has it also worked during previous years? Present-day economic growth patterns are well explained by the simple fact that high-growth countries keep their currencies weak rather than strong. However, until the late 19th century, all world currencies were convertible to silver, and after that, to gold. With currencies fixed against gold, could countries devalue their way to prosperity? They could. President Franklin D. Roosevelt, against the advice of his economists, gradually devalued the dollar against gold in the mid-1930s. He believed this would help the United States emerge from the Great Depression, and the policy may indeed have helped.1 If such currency undervaluation could help to explain other historical phenomena, that would lend credibility to the hypothesis that devaluation has long been a path toward prosperity. The previous chapter examined the role of institutions in explaining why developed economies are much wealthier than developing economies—that is, why developed economies are rich.2 This chapter addresses some narrower questions. For example, what was the pattern of real exchange rate (RER) movements in the 19th century? Did countries with undervalued exchange rates

1. This is entertainingly and convincingly documented by Ahamed (2009). 2. Bhalla (2007a) takes up this issue of major divergence in some detail.

179

grow faster before World War I (1870–1913) and before World War II (1913–38)? Some analysts attribute the growth differential to the existence of high import tariffs in developed economies (see Clemens and Williamson 2002). This is equivalent to saying that currency overvaluation helps growth—the very opposite of the argument made in this book. Was the prewar period that different? Both theory and empirics suggest that actions against trade are bad for growth. Currency undervaluation is a subsidy to exports and a tax on imports. Currency undervaluation decreases the domestic costs of production in terms of international currencies, and an import tariff increases such costs. And that is the likely reason why tariffs are considered a negative for growth, and currency undervaluation is considered positive. According to Anne Krueger: One of the reasons for the success of the countries adopting outward-oriented trade strategies is that an export orientation imposes a discipline and a set of constraints on all economic policies that prevent the adoption of very many measures severely antithetical to growth; and second, the extent of liberalization and smooth functioning of markets that is consistent with rapid economic growth increases with the level of output per capita. (Krueger 1990, 110)

This chapter also examines an intriguing exchange rate question: What explains the fixing of the Japanese yen at 360 to $1 after World War II (besides the whim of General Douglas MacArthur)? And was this exchange rate deeply undervalued and therefore consequential in helping Japan grow at a miracle pace over the following 30 years?

Nineteenth-Century Exchange Rates Until the middle of the 19th century, most countries were on a silver standard, moving near the end of the century to the gold standard. Under the silver and/ or gold standard, the real exchange rate, conventionally defined, could not change over time.3 This meant that exchange rates adjusted to levels consistent with purchasing power parity (PPP). However, the initial or ex ante exchange rates with respect to silver or gold were set by the monetary authorities in the individual countries. The former colonial countries in Latin America set their own rates, but the rates for colonized countries in Asia and Africa were set by the colonial governments. This sets up a natural experiment for testing the hypothesis of whether the exchange rate was more overvalued in the colonies than in the colonizing home countries. Table 12.1 shows income per capita and the RER for countries for which data are available for the historical period. The Penn World Tables data start

3. Unless one country was on the silver standard and the other on the gold standard during the late 19th century. The price of silver halved between 1875 and 1900, while the price of gold remained the same. This meant that countries such as India, which moved from a silver to a gold standard, should show a sharp devaluation; this did not happen. Indeed, with the RER staying constant and a decline in relative income per capita, the RER moves toward more, not less, overvaluation.

180

DEVALUING TO PROSPERITY

in 1950. For years prior to 1950, the PPP exchange rate was computed on the basis of excess inflation between the country and the United States.4 For several countries, the earliest such year is 1870; for some (especially Asian economies), consumer price data are unavailable for years before 1913. When data are available, the RER can be computed as the ratio of the PPP and the dollar exchange rate. Currency valuations for these years are obtained using the same method as for other years. It is the deviation between the actual and predicted RER, and the latter is obtained from the same equation in chapter 4, with Y being income per capita: RER = 1.11 × (1 – .971Y ).

(12.1)

This backcasting is liable to introduce a fair amount of measurement error—that is, an equation estimated for 1996–2009 is being used to project equivalent values more than 100 years earlier. All countries are found to be overvalued with respect to the PPP dollar. If a similar bias creeps in for all countries, then the undervaluation or overvaluation with respect to the dollar should contain relatively fewer measurement errors. And since it is valuation against the dollar that is of interest, the backcasting may not be as problematic as first believed. For the United States, the valuation for the historical period is obtained in the same manner as for other years—the negative of the weighted average of valuation against the dollar for the Broad Index countries. Besides the RER, the table documents the valuation estimates for the two years 1870 and 1950. The final column lists the average annual growth rate of income per capita for the 80-year period from 1870 to1950. Four stylized facts emerge from even a cursory perusal of table 12.1. First, for each country there is near constancy in the RERs between the late 18th or early 19th century and the start of the postwar period. India and China have virtually identical RERs for 1870 and 1950, as do Australia and France. Second, there is not much difference during the 19th century in the RERs of the rich and poor countries; for example, the RER for India in 1870 is 0.55, the same as for the Netherlands. England, the wealthiest country in the world in 1870, has nearly the lowest RER, at 0.47. England’s RER is slightly higher than that of Finland, 0.43. The developing economies were considerably poorer in 1870, and having the same RER as 1950 implies that the exchange rates in

4. These data have been assembled from various sources; see appendix A for details. For most countries, the consumer price index (CPI) is used as the inflation index prior to 1950; after 1950, the GDP deflator is used as an index whenever available. The simple principle followed in data construction was to use the GDP deflator first, and to use the CPI data only when the GDP deflator was not available. For example, the Indian CPI (base 1913 = 100) increased from a level of 69 in 1870 to 364 in 1950. The US GDP deflator showed an increase from 8.28 to 17.61 during the same period (base 1996 = 100). The PPP exchange rate for India in 1950, according to the Penn World Tables, is 2.65. Hence, the estimated PPP exchange rate for India for 1870 is 2.65 × 69 × 17.61/(8.28 × 364) or 1.07.

CURRENCY UNDERVALUATION: A TIME-TESTED POLICY FOR GROWTH 181

Table 12.1

Real exchange rate and currency valuation, 1870–1950

Real exchange rate Country

1870

1950

Currency valuation against the US dollar (percent)

Average growth in income per capita (percent)

1870

1950

1870–1950

Developed economies Australia

0.66

0.58

–75

–45

1.02

Belgium

0.59

0.69

–53

2

0.88

Canada

0.68

0.82

–16

–9

1.82

Denmark

0.56

0.51

–39

–52

1.55

Finland

0.43

0.68

–8

17

1.65

France

0.69

0.71

2

15

1.29

Italy

0.56

0.48

3

–1

1.06

Japan

0.25

0.34

–30

17

1.20

Netherlands

0.55

0.45

–73

–51

0.97

Norway

0.64

0.61

1

–16

1.73

Portugal

1.66

0.51

156

59

0.95

Spain

0.83

0.44

34

22

0.74

Sweden

0.87

0.70

29

–12

1.75

United States

1.00

1.00

33

–18

1.70

United Kingdom

0.47

0.55

–97

–36

0.97

Argentina

0.70

1.19

–9

54

1.67

Brazil

0.28

0.97

18

150

1.07

China

0.45

0.48

68

184

0.02

India

0.55

0.56

82

176

0.19

Mexico

0.68

0.36

60

–2

1.57

Developing economies

Notes: For estimation misalignments against the US dollar, see text and chapter 4. Sources: See table 3.1 and appendix A.

1870 were even more overvalued than in 1950, when they were also overvalued. India and China were both overvalued against the dollar in 1870, and even more so 80 years later. The Western and Latin American countries generally had undervalued currencies in 1870. The countries in these regions grew at 1.66 and 1.32 percent per year, respectively. In contrast, India and China had average overvaluation levels of 82 and 68 percent, respectively, in 1870, and so it is no surprise that their growth rates were also considerably lower, at 0.32 percent per year. 182

DEVALUING TO PROSPERITY

According to the currency valuation growth model, India and China should have annual growth rates approximately 1 percentage point less than those of the independent Western and Latin American countries. The level of initial valuation and the change in valuation accounts for about 0.8 percentage point of the decreased growth for India and China; that is, almost the entire lower growth in these two countries is explained by exchange rate policy. Summarizing, the most undervalued currency in the world in 1870 was the pound sterling. At 97 percent, its misalignment far exceeded that of its colony, Australia (–75 percent), or the “jewel in the crown,” India (82 percent). I examine this aspect of colonial policy in some detail in Bhalla (forthcoming). For the moment, it suffices to note that, perhaps not coincidentally, the poorer countries had higher overvaluation levels and grew more slowly than the advanced economies.

Tariffs and Growth Did currency valuation have any effect on output growth in the years prior to 1913 and between 1913 and 1938? The hypothesis being addressed is whether tariffs and/or the exchange rate practices in the different economies had an effect on growth. Table 12.2 presents data for selected countries. A cursory reading of the data suggests that changes in currency valuation and in growth are causally related. This is formally tested below. The basic model is the one used earlier—growth of income per capita is a function of (log) initial income, the average tariff rate, the level of initial currency valuation, and the average change in currency valuation. The model is estimated for three time periods: 1870–1913, 1913–38, and the pooled data from 1870–1938 (with a dummy variable representing the two periods). Table 12.3 presents the results. Columns 1 to 3 are regression results for 1870–1913 but without (log) initial income per capita; the next three columns add this variable. The first column reports the O’Rourke (2000) regression—10 developed economies, with data from 1875–1910 in five-year intervals. Each 10 percent higher tariff level adds 0.06 percent to the growth rate. The second column is for 43 countries for which data on tariffs are available. The coefficient on tariffs is no longer significant. The third column reports the results for the 23 countries for which both tariff and valuation data are available. Exchange valuation is significant (and with the same sign as for the late 20th century—that is, currency undervaluation helps growth). The coefficient on tariffs becomes (barely) significant at the 10 percent level. The next three columns repeat the same models, but with initial income per capita included in the regression. The coefficient on tariffs is very unstable and, more often than not, insignificant. The weakness of the relationship between tariffs and growth has been well documented by Douglas Irwin (2002). Table 12.4 adds data for 1913–38. The same results are obtained, with the average tariff rate not significant but currency valuation (whether initial level or change in level) always significant. CURRENCY UNDERVALUATION: A TIME-TESTED POLICY FOR GROWTH 183

 Table 12.2 Tariffs and currency valuation, 1870–1913 Initial currency valuation, 1870 (log percent)

Change in currency valuation, 1870–1913 (log)

GDP growth per capita, 1870–1913 (in PPP, percent)

Country

Average tariff (percent)

Argentina

0.6

–9

18

2.5

Australia

1.3

–75

18

1.1

Belgium

0.4

–53

14

1.0

Brazil

1.8

18

23

0.3

Canada

0.4

–16

19

2.2

China

1.1

68

11

0.1

Denmark

0.2

–39

7

1.6

Finland

0.5

–8

35

1.4

France

0.1

2

9

1.4

India

1.2

82

17

0.5

Indonesia

n.a.

a

78

10

0.8

Italy

0.2

3

11

1.2

Japan

1.0

–30

6

1.5

Mexico

–0.5

60

21

2.2

Netherlands

1.0

–73

4

0.9

Norway

1.2

1

12

1.4

a

Philippines

n.a.

1

8

1.1

Portugal

0.5

156

24

0.6

Spain

0.3

34

37

1.2

Sweden

0.2

29

10

1.4

Switzerland

n.a.

a

–61

11

1.6

United States

–1.1

33

23

1.8

1.6

–97

5

1.0

United Kingdom

n.a. = not available; PPP = purchasing power parity a. Values are for 1913. Source: Bhalla (2007a) dataset extended to 2011.

These historical experiments are very supportive of the argument that currency undervaluation helps growth. It is striking that the same model helps explain developing-country growth in the late 19th century and in the late 20th century (covering the period when both groups of countries were in the early stages of development). Even the coefficients of the currency valuation variables (initial undervaluation and change in undervaluation) are comparable, if not nearly identical.

184

DEVALUING TO PROSPERITY

CURRENCY UNDERVALUATION: A TIME-TESTED POLICY FOR GROWTH 185

Table 12.3 Tariffs, currency valuation, and growth, 1870–1913 Model Variable

1

2

3

Initial income (per capita) Average tariff rate

0.06*( (2.20)*

0.02) (1.30)

Currency valuation (mean)

0.04***) (1.70)***

4

5

0.36**) (2.80)**

–0.11** (–0.5)**)

0.01*)) (1.50)*)

0.02**) (1.80)**

–0.01***) (–4.2)***0

6 –0.26) (–1.3) 0.01) ) (0.90)

–0.01** (–3.5)**)

Initial currency valuation (1980)

–0.01** (–4.1)

Average change in currency valuation (percent)

–0.71***) (–5.3)

Number of observations 2

Adjusted R

70

43***)

23*))))

43**))

23**)

20**)))

0.07*)

0****

0.23***)

0.12**)

0.27**

0.42**))

Notes: The tariffs are as calculated by O’Rourke (2000). Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. t statistics are in parentheses. Source: Bhalla (2007a) dataset extended to 2011.

Table 12.4 Tariffs, currency valuation, and growth, 1913–38 Model Variable

1

Initial income (per capita) Average tariff rate

–0.01) (–1.7)0

3

4

0.31*) (2.00)*

2

–0.15*** (–0.8)***)

–0.06*** (–0.3))))))

–0.01 (–1.1))))

0.02*** (2.50)*))

0**))) (0.20)**

Currency valuation (mean)

–0.01*** (–4.3)***)

Initial currency valuation

–0.00*** (–2.7)***)

Average change in currency valuation (percent)

–0.55*** (–6.1)***)

Time dummy (1913–38 = 1) Number of observations 2

Adjusted R

–0.05) (–0.3)))

–0.39** (–1.5))*)

–0.49*** (–1.5)**))

–0.69*** (–2.0)***)

65**))

65**)))

31**)))))

28**))))))

0.08)

0.13*)

0.27**)

0.36***

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Numbers in parentheses are t statistics. See text for details. Source: Bhalla (2007a) dataset extended to 2011.

The Yen Exchange Rate in 1950 After World War II, General Douglas MacArthur set the Japanese exchange rate at 360 yen to $1. Before the war, in 1938, the rate was only 3.56 yen to $1. Table 12.5 documents the evolution of the RER for Japan from 1890 to1960, along with its valuation and the level of income per capita. These data suggest at least two stylized facts. First, to a rather surprising degree, given that Japan is believed to have been a preeminent currency undervaluer, Japan appears to have consistently played by the rules, at least by the rules authorized by Béla Balassa (1964) and Paul Samuelson (1964). In 1950, at an exchange rate of 360, and an RER of 0.37, the Japanese currency was overvalued by 24 percent. The Deutsche mark was overvalued by 32 percent. In contrast, the dollar and the pound were undervalued by 10 percent and 36 percent, respectively. The Dutch guilder, in keeping with its historical tradition of consistent undervaluation (a result also found by Balassa 1964), was undervalued by 51 percent, the fourth most undervalued currency that year.5 Postwar exchange rates were set under the Bretton Woods Agreement, and given that the rates for both Germany and Japan were set at

5. The most undervalued currency in 1950 was that of Mauritius (61 percent), followed by Switzerland (60 percent) and Denmark (52 percent). Australia and the United Kingdom followed the Netherlands in the undervaluation sweepstakes, along with South Africa, Uruguay, Norway, the United States, Canada, and Mexico.

186

DEVALUING TO PROSPERITY

Table 12.5

How the rate of 360 yen to $1 was selected in 1950 Exchange rate

Purchasing power parity (PPP)

Year

US dollars

Real exchange rate (RER)

Undervaluation (percent)

Daily income per capita (1996 PPP dollars)

1890   

0.30

1.19

0.25

93

3.21

1913  

0.75

2.03

0.37

101

4.40

1938 

1.03

3.56

0.29

24

7.78

1950  

121.70

361.00

0.34

17

6.10

1960 

156.80

360.00

0.44

–12

12.50

Note: See text for the definition of the variables. The PPP exchange rates for years prior to 1950 are obtained by linking Japanese and US inflation to the PPP exchange rate obtained from Penn World Table 6.1 for 1950. RER is the ratio of the PPP and the US dollar exchange rate. Source: Bhalla (2007a) dataset extended to 2011. See appendix A for details.

overvalued levels, it appears that undervaluation went hand-in-hand with victory in the war. The RER for Japan shows a steady increase with income. In 1890, Japanese income per capita was PPP$3.2 per day, and the RER was 0.25. In 1913, income per capita was PPP$4.4 per day, and the RER had increased, in the spirit of the Balassa-Samuelson effect, to 0.37. Then the wars intervened, hyperinflation occurred, and by 1948 Japan’s RER had risen to 0.49. Given these developments, the US administrators chose the RER rate based on the level before the fluctuations caused by the wars and the Great Depression, choosing the exchange rate prevailing in 1913, the last clean observation. In that year, the RER was 0.37. Imposing this RER on the PPP exchange rate in 1949 of PPP$132.3 produced an exchange rate of 360, which was the rate chosen and fixed.6 The strong results reported in the previous chapters were produced from analysis pertaining to the developing economies of today. It is reassuring to note that the same method, the same S-shaped relationship between RER and income per capita, helps to explain the growth pattern among countries that were developing more than 100 years ago. It is apparent that countries have been devaluing their way toward prosperity for many years.

6. The data for price levels and exchange rates for the pre-1950 period for Japan (and several other countries including the United States) were kindly provided by Alan Taylor.

CURRENCY UNDERVALUATION: A TIME-TESTED POLICY FOR GROWTH 187

13 Economics of the Yen and the Renminbi

If we choose, we can live in a world of comforting illusion. —Noam Chomsky (2001) The dollar, the euro, the yen, and the renminbi. The markets swirl around currency values 24 hours a day, 7 days a week. When traders, investors, or policymakers are asked to name the most egregious currency imbalances in (recent) history, they almost universally mention the yen in the early 1980s. A second example, sometimes offered only after much prodding, is the euro. The European Union cannot have a single currency adequately adjusted to the disparate rates of development among the 27 EU members or the 17 euro area members. What about the Chinese renminbi? American officials would answer that, yes, the Chinese currency is undervalued. Others would not be so forthcoming. However, in the last year or so, government officials from countries as disparate as Brazil, South Africa, and India have eschewed any sense of political solidarity with China as another emerging-market economy and have agreed that China’s currency is too cheap to ignore.

Japan in the 1980s and China in the 2010s In the 1980s, the most common threat to the economic hegemony of the United States was Japan, which was feared for its growth and the competitiveness of its industry. There was an outpouring of research evaluating the Japanese threat. The dollar was deemed too strong, and the yen and the Deutsche mark too weak. Robert Lawrence’s 1984 book, Can America Compete? was a must-read. In 1985 came the G-5 Plaza Agreement, which increased the value of the yen more than 100 percent in a matter of a few years. At its peak, in 1991, Japan’s income was 43 percent of US GDP. Japan’s growth slowed to a trickle thereafter, and starting in the mid-1990s begins the story of Japan’s lost decades. In 2011, China is the new, and much larger, economic threat to the United States. In 2008, China’s GDP was higher than US GDP on the basis of 189

existing purchasing power parity (PPP) estimates. In 2007, China’s GDP was PPP$13.5 trillion and US GDP was PPP$14.1 trillion. In 2008, China’s GDP was 12 percent higher at PPP$15.1 trillion, and US GDP was 2 percent higher at PPP$14.3 trillion. By the strangest of coincidences—of the type that give rise to conspiracy theories—the United Nations and World Bank issued new PPP estimates at just this time. These two agencies set the standards for such estimates, and organizations such as the International Monetary Fund (IMF) and Organization for Economic Cooperation and Development (OECD) dutifully use these estimates in their research and policy analysis. The new PPP data paint a much different picture of the world than the previous estimates—actually, not the whole world, just Asia. China was participating for the first time in the formal price-gathering process of the UN–World Bank International Comparison Program (ICP) price survey. There were good reasons to believe that the new estimates of China’s GDP would be different, but the difference was shocking—China’s GDP was lowered by 42 percent. The 2008 GDP for China was not PPP$14.1 trillion but PPP$8.2 trillion. Several reasons were given. It wasn’t that output was misestimated earlier; it was that the price level was 42 percent higher than previously estimated. Interestingly, the new PPP estimates for India were also almost identical to the downgraded estimates for China. The reason was the same: The price level in India was higher by the same amount as the price level for China (actually, the price level in India was higher by 39 percent, not 42 percent). For this increase, however, there was no explanation. India’s statistical system had been producing three consumer price surveys and a wholesale price index every year and a widely publicized consumer expenditure (and price) survey every five years.1 The ICP price survey did not match any of these Indian price surveys; not only did they not match, they were widely different. So by a change in measurement the momentous occasion when China’s GDP surpassed that of the United States was postponed by eight years. According to IMF estimates, in 2016 China’s GDP, in current PPP prices, will be PPP$18.7 trillion and US GDP will be PPP$18.3 trillion (IMF 2011). Notwithstanding such statistical wars, there is some basis to think that Japan and China are comparable. Chapter 10 compared Japan in the early 1980s and China during the 2000s in terms of mercantilism. Chapter 7 estimated the currency adjustments necessary to redress global imbalances and found that the appreciation needed for the renminbi was at historically high proportions, indicating that the global system was seriously out of equilibrium. It is useful, and necessary, to recall that in the mid-1980s, at the time of the Plaza Accord, the US current account was only about 2 to 3 percent of GDP. At that time, there was near unanimity on the need for a structural reversal and for the yen to appreciate significantly. And adjustments that were made by the G-3 central banks (Germany, Japan, and United States) represented a major structural change, a revaluation by an order of magnitude. 1. The National Sample Surveys of over 125,000 households in both rural and urban India.

190

DEVALUING TO PROSPERITY

The situation with China could not be more different. It is very likely that the renminbi is more undervalued by an order of magnitude than at the peak of the undervaluation of the yen or the Deutsche mark in the 1980s. China’s current account surplus is also significantly higher than that of either Germany or Japan at the peak of their respective experiments with mercantilism. Most important, China’s growth rate, relative to its own trend, is also significantly higher. Indeed, China hit its peak growth rate just a few years ago. In comparison with an average overvaluation of the yen by 9 percent during 1976–85, and an average overvaluation of the Deutsche mark by 20 percent during the same years, China has had an average undervaluation of 41 percent during the last 10 years. Despite these factual imbalances, the protests against China’s exchange rate have been relatively muted. Unlike the nominal exchange rate, the real exchange rate is a virtual variable. Economists and policymakers can, and do, hide behind nominal changes and claim that currency valuations reflect economic reality. The Chinese renminbi appreciated by about 20 percent between 2005 and 2011, yet, as shown in chapter 6, in real terms, it has actually depreciated by a small amount over the same period (due to the large standing-still effect). The virtual standing-still real exchange rate movement is the reality; movements in the nominal exchange rate are not that informative. Yet most experts have ignored the real movements in the exchange rate. The situation was somewhat similar in the 1960s when Béla Balassa (1964) and Robert Samuelson (1964) presented their papers as an answer to the anythinggoes environment then prevailing among economists and other experts.2 The fact remains that the major economies of the world collaborated in the mid-1980s to address the cheap value of the yen. Those who are calling now for a similar revaluation of the renminbi need to show, at a minimum, that the renminbi in the normal years preceding the crisis of 2008 was at least as cheap as the yen was before the Plaza Accord in 1985. Second, and equally important, is the need to document that an appreciation of the renminbi will not be as disruptive to China’s and world growth as the appreciation of the yen was to Japan’s growth after 1985.

Déjà Vu? It is likely that Japan’s growth rate collapsed in part because of the Plaza Agreement of 1985, which in three short years catapulted the yen from an average exchange rate of 240 yen to double that level (in real terms; the nominal exchange rate was cut in half). Will a quick revaluation of the renminbi lead to a similar implosion in China’s growth? And can the world afford to lose its major growth engine in the face of slow growth in the advanced economies? Interestingly, according to my measure, the yen was not really so under2. Samuelson is particularly caustic against his opponents; his 1964 article is a must-read for anybody interested in the contemporary debate on Chinese currency undervaluation.

ECONOMICS OF THE YEN AND THE RENMINBI 191

valued at the time of the Plaza Agreement in 1985. It was undervalued by only 4.7 percent compared to an undervaluation of 31 percent before the dollar float in 1971 (when the exchange rate was 360 yen to $1). This result is controversial, but it is supported by the evidence presented by Jeffrey Sachs (1981) and Robert Lawrence (1987), and by other prevailing estimates of undervaluation.3 The comparison between Japan in the 1980s and China today has nearly all the attributes of a perfect natural experiment. Some sense of the similarities and differences can be gleaned from a comparison of five-year periods preceding the respective crisis years of 1985 and 2007. Japan (1981–85) and China (2002–06) provide a test for several hypotheses: Both are large, exportoriented economies, both are Asian, both have high savings rates, both are alleged to have undervalued currencies,4 both are accumulating reserves, and both are accused of engaging in “currency manipulation.” There is another obvious comparison: Both are causing the United States to have correspondingly large current account deficits. In 1984, the US current account had moved to a historically high deficit of 2.4 percent of GDP from near balance for most of the preceding decade. And in 1998, US current account deficits started moving much higher after staying at comfortable 1 to 2 percent levels during most of the 1990s. In both 2005 and 2006, the US current account deficit reached 6 percent of GDP. The Japan-China comparison is conducted at two levels; the experience of the two economies is examined during the respective five-year periods of currency misalignment (1981–85 for Japan and 2002–06 for China) and for the longer 15-year period (1970–85 for Japan and 1991–2006 for China). Japan’s starting year is 1981 and China’s is 2002, both of which were recession years. Table 13.1 documents this comparison. (Unless noted otherwise, the figures in the text represent comparison for the five-year periods.)

Currency Misalignment The starting positions for the two countries are very, very different. In 1981, the yen was overvalued by 15 percent; in 2002, the renminbi was undervalued by 31 percent. The yen stayed overvalued except for the years between 1982 and 1985, and during these exception years it was undervalued by no more than 6 percent. In striking contrast, the renminbi has been continuously undervalued since 1994, and was undervalued by 31 percent in 2002, by 45 percent in 2006, and has been outside the extreme 25 percent band for the last 10 years—the same time that the US current account balance began to sharply deteriorate. 3. Note that both the Johnson, Ostry, and Subramanian (2007) and Rodrik (2007a) measures have the yen overvalued in 1985 by 35 and 69 percent, respectively. 4. The term “alleged” is used because, as documented later in this chapter, the conventional wisdom is that the Chinese renminbi has not been undervalued.

192

DEVALUING TO PROSPERITY

Table 13.1

Japan and China: No déjà vu 5-year period Japan 1981–85

Indicator

China 2002–06

15-year period Japan 1970–85

China 1991–2006

Growth (percent) GDP

3.3

11.9

4.0

9.7

GDP per capita

2.6

11.4

3.0

8.9

Exports

7.3

27.4

13.4

23.6

Imports

0.3

24.6

13.4

23.9

Reserves

2.1

38.1

23.9

29.9

Currency valuation

–3.9

–6.2

2.9

–4.9

2.9

–5.8

2.6

–0.8

26.0

646.0

20.8

290.0

1.9

4.7

1.0

2.4

Currency valuation (percent)

–0.1

–39.0

2.2

–15.7

Currency valuation (Johnson, Ostry, and Subramanian 2007) (percent)

24.6

–42.4

17.6

–36.6

Savings (percent of GDP)

30.9

46.1

33.2

42.2

Investment (percent of GDP)

29.1

41.5

32.6

39.8

Mercantilism (ranking in 1985, Japan and 2006, China)

14.0

3

Mercantilism (average)

12

14

21

Currency valuation (Johnson, Ostry, and Subramanian 2007) Average levels Reserves (billions of US dollars) Current account balance (percent of GDP)

11

Notes: See text for definition of mercantilism index. A negative currency valuation means undervaluation. Source: Bhalla (2007a) dataset extended to 2011.

The cumulative effect of currency valuation on income levels is documented for the 15-year periods (1970–85 for Japan and 1991–2006 for China) in figures 13.1 and 13.2, respectively. The amount of appreciation of the yen is noteworthy, as is the amount of depreciation of the renminbi. The differences in income growth are also worth noting. At the end of the 15 year period, in 1985, income per capita in Japan was 55 percent higher; at the end of the corresponding period, China’s income per capita was 256 percent higher. The currency change may tell the story: The yen appreciated by close to 30 percent during this period, and the renminbi depreciated by a cumulative 53 percent. In other words, its value in 2006 was less than half (47 percent) of its real value just 15 years earlier.

ECONOMICS OF THE YEN AND THE RENMINBI 193

Figure 13.1

Japan: Evolution of per capita income and currency valuation, 1970–85

index (1970 = 100) 170 Currency valuation 160 150 140 130 120

Per capita income

110 100 90 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 Notes: A negative value signifies undervaluation. Both income per capita and currency valuation are indexed to 100 in the beginning of the 15-year period, 1970. Source: Bhalla (2007a) dataset extended to 2011.

GDP and Export Growth Income per capita grew at a rate of only 2.6 percent per year for Japan but at a level almost four times higher for China, at 8.9 percent. Japanese exports grew at 7.4 percent per year (relative to world median growth of –0.3 percent during 1981–85).5 In contrast, Chinese exports grew at over 27 percent, some 13 percentage points above the corresponding median global rate during 2002–05.

Current Account Surplus Japan’s current account surplus averaged just 1.9 percent of GDP, in contrast to 5.0 percent for China. In subsequent years (after 2005), China’s current account surplus significantly increased; the surplus (as a fraction of GDP) averaged over 10 percent during 2006–08. This surplus number is large for

5. The world calculation refers to about 90 countries with populations greater than 1 million and excluding the major oil-exporting economies.

194

DEVALUING TO PROSPERITY

Figure 13.2

China: Evolution of per capita income and currency valuation, 1991–2006

index (1991 = 100) 350

300

250

200 Per capita income 150

100

Currency valuation

50

0 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Notes: A negative value signifies undervaluation. Both income per capita and currency valuation are indexed to 100 in the beginning of the 15-year period, 1991. Source: Bhalla (2007a) dataset extended to 2011.

any non-oil-exporting economy and certainly for the world’s second largest economy in PPP terms. When Singapore, Taiwan, Korea, or even Japan pursued a policy of currency undervaluation, the world could absorb the repercussions without significant disruption or imbalances. But when a country with more than one-fifth of the world’s population practices a beggar-thy-neighbor policy, it can cause other economies to have substantially lower growth than they would otherwise.

Reserve Accumulation The accumulation of foreign reserves by both Japan and China seems excessive, but more so for China. China’s reserves averaged $646 billion a year between 2002 and 2006,6 a level some 31 times the mean. In comparison, Japan’s 6. In March 2007, China’s reserves exceeded $1.2 trillion and were thus almost twice the average over the preceding five years.

ECONOMICS OF THE YEN AND THE RENMINBI 195

reserves, at $21 billion, were only about 4.5 times the average for all countries during 1981–85. During its peak year, Japan’s reserves accounted for 2 months of imports; China’s reserves were adequate to cover more than 15 months of imports.

Mercantilism Undervaluation and current account surpluses are often associated with mercantilism, but there are enough exceptions to suggest that evaluation of mercantilism (currency manipulation) is not straightforward. Between 1981 and 1985, both Germany and Japan were among the top 15 mercantilist countries. Despite having peak current account surpluses today, both countries have a low mercantilism ranking for the period 2002–06. India’s mercantilism ranking for the two periods is 64 and 34. China’s ranking in 1981–85 is 59 and in 2002–06 is 11. In later years China moves toward even higher levels of mercantilism.7

China Is Different Overall, with all the data examined—export growth, GDP growth, current account surpluses, mercantilism indices, historical tendencies—there is only one conclusion to be drawn: The Chinese renminbi is much more undervalued today than the yen was in 1985 at the time of the Plaza Agreement. In 1985, the yen was undervalued by only 5 percent; in 2011, and for seven previous years, the renminbi has been undervalued by more than 40 percent. If the initial conditions of China and Japan are to be compared, then the beginning of the Chinese appreciation, from a level of 6.6 yuan to $1 in 2011, is comparable to a value of 335 yen to $1 in 1985. But the cheapest yen exchange rate in nominal terms was 360 yen to $1, a level that prevailed from the end of World War II until 1970. Stated differently, an exchange rate equivalent to the yen in 1985 would be 4.71 yuan to $1 in 2011. It is unlikely that anyone would have objected to the Plaza Agreement if the exchange rate was 335 yen to $1. Or that if the world considered the yen to be too cheap at a rate of 239 yen to $1 in 1985, it should consider 4.71 yuan to $1 to be too cheap in 2011. Therefore, the existing rate of 6.6 yuan to $1 can be considered a generous gift from the world to China. The above analysis suggests that the view that China is in a similar position to Japan in the mid-1980s is misplaced and inappropriate. In the main, there are two reasons. First, the initial conditions are immensely different. Japanese GDP growth at the time of its currency appreciation was sharply lower than its own historical average and lower than the average level of Chinese growth

7. Hong Kong manages to have a single-digit ranking in both periods. South Korea does not appear to have played unfairly in either period (ranking of 44 and 23 for the two periods, respectively). In contrast, Taiwan was one of the most mercantilist countries during 2002–06.

196

DEVALUING TO PROSPERITY

in the 2000s. Second, real deflation and slower growth in Japan started in the late 1980s, when the yen became overvalued by more than 35 percent, far beyond the range of tolerance. For China to feel the same overvaluation pressure on growth, the renminbi would have to appreciate to less than 3 yuan to $1 over the next decade, an unlikely event. Even if the nominal appreciation is a gradual 10 percent per year over the next 10 years,8 as now appears possible, the renminbi will still be about 25 to 30 percent undervalued after a decade, in 2021. The point simply is that China’s rate of currency appreciation has to be significantly faster due to Balassa-Samuelson productivity considerations— that is, faster by 5 to 7 percent per year to make any dent in its deeply undervalued status. Regarding current account imbalances, it is noteworthy that China’s current account surplus has begun to decline from the peak levels of more than 10 percent of GDP achieved just a few years ago. For the period 2006–11, China’s current account surplus averaged 7.2 percent of GDP, which is comparable to the 9 percent average (weighted by GDP) for oil-exporting economies. For India, the average was –1.9 percent. The Chinese surplus numbers are unlike anything the world has ever seen, certainly in the second (and possibly the first) largest economy in the world in PPP terms. Why the difference? The intellectual and policy debates concerning the Chinese currency valuation have been very different from the corresponding debates about the yen. In the mid-1980s, there were relatively few academic articles and few commercial enterprises questioning the need for the yen to revalue by the magnitude that it did. In contrast, during the last decade, the number of articles that question the need for a revaluation of the Chinese currency far exceeds the articles that say it should. Yin-Wong Cheung, Menzie D. Chinn, and Eiji Fujii (2007) state “there is little statistical evidence that the RMB is undervalued,” while a few years earlier, IMF economists Steven Dunaway and Xiangming Li (2005) concluded that there was no basis for any conclusion about undervaluation: The number of studies attempting to estimate the “equilibrium” real value of China’s currency has proliferated in recent years as the country’s presence in world markets has grown. These studies have sought to establish whether or not a significant part of China’s competitive prowess can be attributed to the foreign exchange value of the renminbi. Unfortunately, no consensus has emerged because the studies yield a very wide range of estimates. (Dunaway and Li 2005, 1)

Tables 13.2 and 13.3 summarize the opposing views on the Chinese currency valuation over the last decade. The conventional wisdom until at least 2008 was that the renminbi was only 5 to 15 percent undervalued, if at

8. It is worth emphasizing that the 2003 Goldman Sachs BRICs report assumes, in its rather bullish forecasts, a real appreciation of the Chinese renminbi at the rate of 3.5 percent per year. One of the flaws in the report is that it did not recognize the dependence of Chinese high growth on a depreciating real currency.

ECONOMICS OF THE YEN AND THE RENMINBI 197

198 DEVALUING TO PROSPERITY

Table 13.2 The big debate: The Chinese renminbi is not undervalued

Author

Currency misalignment (percent)

Comments

2002 Xiaopu Zhang

6 to 12

“The RMB was slightly overvalued in 1996 and then more overvalued to 12 percent in 1997 and 1998…resulting in only 6 percent overvaluation in 1999.”

Norbert Walter

–20

“The nominal US Dollar/Chinese Yuan reflects an undervaluation of the [yuan] of some 20 percent…. Little may be achieved for China and the global economy by a stronger CHY.”

Stephen J. Jen

Close to 0

“The [euro], not the [renminbi], is the problem...arguments for revaluation based on perceptions that the [renminbi] is overvalued are largely flawed in my view. Further, it is unclear why China should ‘import’ deflation to ‘save the rest of the world’.”

Close to 0

“China does not compete on the basis of an undervalued currency, but mainly in terms of labor costs, technology, quality control, infrastructure, the improved human capital of its work force, and a passion for and commitment to reform.”

2003

2004 Stephen Roach

Barry Eichengreen

–5

“Relative to its average between the middle of 1996 and the middle of 2002, the RMB was undervalued on a real effective basis (weighted relative to the relative labor costs of its principal trading partners) by only about 5 percent.”

Barry Bosworth

–40

“The [renminbi] is undervalued on a PPP basis, but the PPP standard provides very weak guidance as to the appropriate exchange rate for low income countries…consideration of macro economic balance suggests little or no undervaluation. There is no evidence of a growing surplus in the most recently available trade data. In fact, the most recent information suggests a declining current account balance….”

2005 Yin-Wong Cheung, Menzie D. Chinn, and Eiji Fujii Jim O’Neilla Steve Hanke

–6

–10.5 a

ECONOMICS OF THE YEN AND THE RENMINBI 199

Michael Funke and Jorg Rahn Steven Dunaway and Xiangming Li

“In 2003 the [renminbi] was more than one standard error—but less than two standard errors—away from predicted value, which in the present context is interpreted as the “equilibrium” value. In other words, by the standard statistical criterion that applied economists commonly appeal to, the [renminbi] is not undervalued (as of 2003) in a statistically significant sense.” “Yuan is undervalued by about 10.5 percent.”

 Close to 0

“China would be foolish to consider revaluing. Should China revalue the yuan by 25 percent, it would lead to 20 percent deflation.”

–12 

“The estimated degree of undervaluation at the end of the sample period turns out to be 12 percent, again suggesting that claims of misalignment of the renminbi have been exaggerated.”

Close to 0

“Data problems—in the specification of both dependent and explanatory variables—remain substantial and will only be solved with time. It is also only in time that key underlying economy relationships determining a country’s exchange rate may stabilize sufficiently that they might be estimated with a reasonable degree of confidence. But then again because of the complexity of these relationships, being able to accurately explain movements in exchange rates may remain an elusive objective. In the meantime, estimates of the level of a country’s real exchange rate should continue to be treated with great caution.”

Close to 0

“This common presumption of renminbi undervaluation is wrong, and its appreciation need not reduce China’s trade surplus but would cause serious deflation in China.”

Jonathan Anderson

–10 to –15

“We looked at all the evidence and concluded that the renminbi was probably 15 to 20 percent undervalued in the first half of 2005—which means 10 to 15 percent undervalued today (April 2006).”

Eswar Prasad

Close to 0

“Despite these pressures..., the real effective exchange rate of the renminbi is now below its recent peak in 2002 (largely due to the US dollar’s depreciation against other major currencies).”

2006 Ronald McKinnon and Gunther Schnabl (2006) 2007

(continues on next page)

200 DEVALUING TO PROSPERITY

Table 13.2 The big debate: The Chinese renminbi is not undervalued (continued)

Author Yin-Wong Cheung, Menzie D. Chinn, and Eiji Fujii

Currency misalignment (percent) Close to 0

Comments “We find that, once sampling uncertainty and serial correlation are accounted for, there is little statistical evidence that the RMB is undervalued, even though the point estimates usually indicate significant misalignment.” “We find that the estimated misalignment detected in our previous study disappears completely with this new data set.” “After conducting various robustness checks, we conclude that although the point estimates indicate the RMB is undervalued in almost all samples, in almost no case is the deviation statistically significant, and indeed, when serial correlation is accounted for, the extent of misalignment is not even statistically significant at the 50 percent level.”

2011 Quingyuan Du and Shang-Jin Wei

Close to 0

“Empirically, those economies with a high sex ratio tend to have a low real exchange rate, beyond what can be explained by the Balassa-Samuelson effect, financial underdevelopment, dependence ratio, and exchange rate regime classifications. Once these factors are accounted for, the Chinese real exchange rate is estimated to be undervalued by only a relatively trivial amount.”

a. Michael Schneider and Bronwyn Curtis, “Hanke Says China Would Be ‘Foolish’ to Let Yuan Rise,” Bloomberg News, May 20, 2009, www.bloomberg.com (accessed on April 2, 2012).

Table 13.3 The big debate: The Chinese renminbi is undervalued

Author

Currency misalignment (percent)

Comments

1995 Surjit Bhalla

–20

“Currently, the “fair” value of the yuan appears to be around 7 suggesting that the yuan is undervalued by at least 20 percent.”

–15

“The Chinese yuan is today undervalued with respect to the dollar by about 10 to 15 percent.”

1998

ECONOMICS OF THE YEN AND THE RENMINBI 201

Surjit Bhalla (1998b) 2002 Yongxiang Bu and Rod Tyers

–5 to –12

“Undervaluation is between 5 and 12 percent in 1998.”

Surjit Bhalla (2002b)

–40

“The Chinese renminbi is today undervalued by as much as 40 percent.”

Liang Hong

–15

“We consider three scenarios based on the assumption that the renminbi is undervalued on a real effective basis by 15 percent.”

2006 Big Mac Index

–59

Jeffrey Frankel

–35

“The Yuan was undervalued by ~35 percent in 2000, and is by at least as much as that today. Typically across countries, such gaps are corrected halfway, on average, over the subsequent decade.”

2007 Morris Goldstein

–30 to –40

“The sad truth is that the [renminbi] is now grossly undervalued—on the order of 30 percent or more against an average of China’s trading partners and 40 percent or more against the US dollar—and that the appreciation of the RMB that has taken place to date against the dollar is considered inadequate to make a real dent in this huge surplus.” (continues on next page)

202 DEVALUING TO PROSPERITY

Table 13.3 The big debate: The Chinese renminbi is undervalued (continued) Currency misalignment (percent)

Author C. Fred Bergsten

 –40

Comments “An increase of 40% in the [renminbi] and other Asian currencies against the dollar would reduce the US global current account deficit by about $150 billion per year.”

2009 Dominique Strauss-Kahna Martin Wolf

b

“The facts are clear, the renminbi is undervalued and really significantly undervalued.” “A country’s exchange rate cannot be a concern for it alone, since it must also affect its trading partners.... So, whether China likes it or not, its heavily managed exchange rate regime is a legitimate concern of its trading partners. Its exports are now larger than those of any other country. The liberty of insignificance has vanished.”

2010 Paul Krugmanc

“China’s policy of keeping its currency, the renminbi, undervalued has become a significant drag on global economic recovery. Something must be done.”

Nicolas Sarkozyd

“The biggest distortion to global competition today is our currencies…. France cannot accept that the euro and Europe fall victim to the undervaluation of certain currencies.”

George Sorose

“The global imbalances have continued to increase. Notably, China continues to run a very big current account surplus. That is one reason why an appreciation of the renminbi would be desirable. The task of correcting those imbalances hasn’t yet begun to be addressed.”

Barack Obamaf

“One of the challenges that we’ve got to address internationally is currency rates and how they match up to make sure that our goods are not artificially inflated in price and their goods are artificially deflated in price.”

Guido Mantegag

“We are experiencing a currency war—devaluing currencies artificially is a global strategy.”

a. “China yuan ‘significantly undervalued’—IMF chief,” Reuters, January 26, 2009, www.reuters.com (accessed on April 2, 2012). b. Martin Wolf, “Why China’s Exchange Rate Policy Concerns Us,” Financial Times, March 2009. c. Paul Krugman, “Taking on China,” New York Times (global edition), March 14, 2010. d. “Sarkozy: Euro, Europe Can’t Be Victim of Undervalued Currencies,” Wall Street Journal Digital Network, January 2010. e. As quoted in an interview with Market Watch, “Surface Tension: George Soros on China,” Wall Street Journal Digital Network, February 2010. f. Address to the senators of the US Democratic Party, White House, Washington, January 2010. g. Eunkyung Seo, “G-20 Officials Seek Currency Policy Compromise,” Bloomberg, September 28, 2010.

all. There was hardly any currency expert at any investment bank that claimed it was more.9 Some academics claimed that high undervaluation levels were produced using faulty methodology. Only a few academics such as Jeffrey Frankel (2006) and Barry Eichengreen (2007), some industry lobbyists, and economists at the Peterson Institute for International Economics claimed that the renminbi was severely undervalued. The situation now is not much different. There are some warnings that the situation is grave. There have been complaints about currency wars (directed at the Chinese) from Brazil’s finance minister, Guido Mantega, which have been echoed by senior officials in South Africa and India. There is Paul Krugman (2010) worrying about how the cheap Chinese currency is hurting growth in the United States. From the other side comes the same story, but with new evidence that purports to show that the renminbi is not undervalued. The Economist claimed that if corrections were made to its Big Mac Index for the level of income per capita, then the renminbi would be overvalued by 3 percent. They fail to note that, according to this new calculation, there would be only five countries with undervalued exchange rates, including India, versus 31 overvalued exchange rates, including the euro. Actually, with the adjustment, India would be the second most undervalued currency! Interestingly, without these new adjustments, the Big Mac Index has the renminbi undervalued by 44 percent in 2011, almost identical to my estimate. Only India (53 percent) and Hong Kong (52 percent) were more undervalued than China. Quingyuan Du and Shang-Jin Wei (2011) argue that because of a bad gender ratio (more young males than females), the natural response of the rational economic system is for the real exchange rate to depreciate and for current account surpluses to be maintained. “Once these factors are accounted for, the Chinese real exchange rate is estimated to be undervalued by only a relatively trivial amount” (Du and Wei 2011, abstract). A leading expert on exchange rate and development economics, John Williamson—who coined the term Washington Consensus and created the concept of fundamental equilibrium exchange rates (FEERs)—was sufficiently provoked by this surprising position to state, somewhat boldly and despairingly: Even in what seem to many of us to be absolutely unambiguous cases, like the renminbi peg, one finds economists with Nobel prizes and others paid inordinate sums by investment banks and at least part of the IMF staff prepared to assert that they do not know whether the renminbi is undervalued or not. If one then tries to take the analysis back a step, to search for exchange rates that will achieve an agreed set of current account targets, one finds that it is similarly impossible to secure agreement on what current account targets should be. Unless and until this sort of intellectual laissez-faire leads to the world ending up in a new depression, it seems hopeless to imagine that it can be changed. (Williamson 2008, 150)

9. Dooley, Folkerts-Landau, and Garber (2003) believe that currency undervaluation is an explicit development strategy of China, but they refrain from divulging the calculations that lead to this conclusion.

ECONOMICS OF THE YEN AND THE RENMINBI 203

This quote was written before the Great Recession of 2008; Williamson was prescient in seeing that such laissez-faire intellectual opinions could lead to a major disaster. All the data available—on export growth, GDP growth, current account surpluses, mercantilism indices, historical tendencies—point to only one possibility: The Chinese renminbi is much more undervalued today than the yen was at the time of the Plaza Agreement. This statement says nothing about the absolute degree of undervaluation; it may be the case as several have argued that the renminbi is fairly valued in real terms. It only refers to the relative undervaluation—that is, if the yen was undervalued in 1985, then the renminbi is much more undervalued in 2006. Given this reality, why the differential treatment of China? One explanation is the possibility of conflicts of interest. In the 1980s, the Japanese economy was relatively closed to imports, foreign direct investment, and mergers and acquisitions. There was not much business income there for American firms, other than via exports of goods to Japan, and these were restricted by Japan. The current situation in China is different. There may be a belief among US firms and investment banks that they cannot profit if they are critical of the currency regime. This is consistent with the view of major US firms, which believe that investing in China is different, and considerably more profitable, than investing in Japan in the early 1980s. Robert Lawrence reaches a similar conclusion: Japanese firms were rivals for leading U.S. firms in key industries such as electronics and automobiles, and the closed Japanese market was seen as giving them an unfair advantage. Moreover, the barriers to the Japanese market were often opaque and not covered by GATT [General Agreement on Tariffs and Trade] rules. In addition, the fact that Japan developed with a market that was closed to foreign investment meant that few U.S. firms had a strong interest in maintaining the U.S.-Japan trading relationship and opposing protection against Japan. By contrast, today China accounts for a large share of the U.S. trade deficit. While labor-intensive sectors in the United States have experienced job losses, in many cases the goods the United States buys from China are no longer produced locally. Thus relatively fewer U.S. firms feel threatened. (Lawrence 2007, 17)

However, the reluctance of policymakers and analysts to take on the Chinese authorities in bilateral and multilateral circles that characterized the early years of this decade has changed. In fact, the IMF now stands at the forefront of institutions demanding that China revalue its currency. There is renewed acceptance of the old idea that a large current account surplus is just as much of a problem as a large current account deficit. Leading government officials in several European countries and in some developing countries (Brazil and India) have now joined the United States in asking for China to act on its currency. Why this change of heart? You can blame, or give credit to, the Great Recession, which prompted everyone to (correctly) realize that globalization has made us utterly interdependent. More than that, however, there is the perception of an important change within China itself. It no longer shocking 204

DEVALUING TO PROSPERITY

for Chinese officials and academics to argue that the present and future are different from the past as far as Chinese exchange rate policy is concerned. At the onset of the Asian crisis in 1997—a crisis whose origins most likely stemmed from the large Chinese devaluations during 1989–94—China, in a Machiavellian gesture of solidarity with the United States and the global economy, said it would not devalue in retaliation against the East Asian countries’ devaluations. Later, China argued that since it was a poor country with high unemployment,10 it needed an undervalued currency to grow swiftly. Next it argued that despite huge trade surpluses, its financial system was very weak and could not endure a currency adjustment. It then acknowledged that it did need to increase consumption among the poor but could not appreciate the currency because that would hurt the farming sector. The likely result, of course, is the opposite—the farming sector would benefit from increased incomes via an appreciation. Today it appears that the leading countries realize that they must hang together or they will hang separately. The well-coordinated fiscal response to the 2008 crisis, led by China, was symbolic of this newfound collaboration. In fact, China itself has indirectly hinted that its currency may be more than marginally undervalued. This augurs well. It is also likely that once the crisis is past, the Chinese currency will revert back to its path of nominal appreciation by 5 to 10 percent per year, which translates to a real revaluation of 3 to 5 percent a year.11

Revisiting Paul Samuelson What are policymakers to do given the multitude of academic and investment bank experts who say either that the renminbi is only marginally undervalued or that China’s unprecedented growth is only tangentially related to its exchange rate undervaluation? It is relevant to revisit what Paul Samuelson said in the early 1960s, when there was similar talk of imbalances concerning the US dollar. One difference between the United States of 1960 and China of 2011 is that by the Balassa calculation, the US dollar was fairly valued in 1960; by the same Balassa calculation, the renminbi was massively undervalued in 2011. Even though the imbalances in 1960 may pale in comparison with those of today, the laws of economics have not changed. What Samuelson wrote in 1964 about the world economy (and the United States) is eerily applicable to the world economy (and China) now. Therefore the quotes below from Samuelson (1964) are followed by comments, labeled “Almost Samuelson (2011),” that point out the symmetry. 10. By implication, the rest of the developing and developed world must have a milder unemployment problem, other things equal. 11. In mid-2011, China announced a change in its exchange rate policy, to allow appreciation. Whether there is any real change in the exchange rate (all puns intended) remains to be seen.

ECONOMICS OF THE YEN AND THE RENMINBI 205

„ „ „

„

„

„

„

„

Samuelson (1964, 153–54): “My own diagnosis of the dollar problem….” Almost Samuelson (2011): “My own diagnosis of the renminbi problem....” Samuelson (1964, 153–54): “The dollar has been somewhat overvalued in this last decade. This does not imply that we should depreciate. It does imply that economists everywhere would prefer, if they could rerun history, that the 1949 depreciations abroad had been somewhat less sharp.” Almost Samuelson (2011): “The renminbi has been somewhat undervalued in this last decade. This does not imply that it should appreciate. It does imply that economists everywhere would prefer, if they could rerun history, that the 1990–93 renminbi depreciations had been somewhat less sharp.” Samuelson (1964, 153–54): “The overvaluation has hampered a high employment policy at home; it has unduly limited America’s freedom to spend abroad in an efficient manner.” Almost Samuelson (2011): “The renminbi undervaluation has promoted a high employment policy at home; it has unduly limited Chinese freedom to spend at home in an efficient manner.” Samuelson (1964, 153–54): “The productivity improvements abroad since 1949…have not yet been matched by commensurate rises in foreign money wages relative to ours.” Almost Samuelson (2011): “The productivity improvements in China since 1990…have not yet been matched by commensurate rises in domestic wages relative to foreign wages.”

„

Samuelson (1964, 153–54): “Our overvaluation has had one effect that some will deem a virtue: it has kept pressure on our price levels. This anti-inflation benefit has been dearly bought in terms of unemployment, excess capacity, slow growth, and low domestic profits.”

„

Almost Samuelson (2011): “Chinese undervaluation has had one effect that some will deem a virtue: it has kept pressure on domestic price levels. This pro-inflation cost has not been much since as benefit the Chinese economy has high employment, near full capacity, fast growth, and high domestic profits.”

„

Samuelson (1964, 153–54): “Our overvaluation has helped to redistribute our disproportionate share of world gold, thus providing the miracle nations of Europe and Japan with needed secular increases in liquidity.” Almost Samuelson (2011): “The Chinese undervaluation has helped China to accumulate a disproportionate share of world reserves.”

„

206

DEVALUING TO PROSPERITY

„

„

„

„

Samuelson (1964, 153–54): “Our overvaluation has put some upward pressure on foreign price and cost levels. By voluntary currency appreciation, the surplus countries could choose to offset this.” Almost Samuelson (2011): “Chinese undervaluation has put some downward pressure on foreign price and cost levels.” Samuelson (1964, 153): “Overvaluation pushes American capital abroad, and in turn is intensified by foreign investment. These are secondary reactions to the technological miracles of growth abroad.” Almost Samuelson (2011): “Undervaluation of the renminbi pushes American capital into China, as well as investment from all over the world; this is in reaction to the miracle of high growth in China, and low growth everywhere else.”

ECONOMICS OF THE YEN AND THE RENMINBI 207

14 Changing Times, Changing Views

Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. —John Kenneth Galbraith, Economics, Peace, and Laughter (1971) Currency undervaluation has mattered significantly for enhancing economic growth for many years and for many countries, large and small—for England and the Netherlands in the late 19th century to China and its neighbors and trading partners in the late 20th century. While economists and international policymakers have endlessly debated the topic, many countries have pursued a policy of undervaluation with demonstrable results. Currency undervaluation helps growth (equivalently, currency overvaluation hurts growth). Sometimes, perhaps often, this higher growth has meant larger current account surpluses. When it does, it means that what is good for a part is not good for the whole. And when that part is a large country, the costs to other economies, and the world economy, can be extensive. There is, therefore, a conflict. Selfish (read domestic) reasons suggest that an economy can increase economic growth by undervaluing its currency. But international reasons suggest that other countries will not, should not, allow their own growth to be lower as a result of someone else’s selfish beggar-thyneighbor policy. If such a conflict exists, then history should (and does) provide evidence of the following sequence of events. First, there should be a tendency for countries to pursue policies that enhance currency undervaluation—“just to be more competitive, my dear.” Other countries, especially the ones that are hurt, will put pressure for balance to be restored. Most often, this suggestion will fall on deaf ears and the result will be an international crisis. And the resolution of these crises will provide reason, and evidence, about what is good, and desirable, for the world economy and for the fault-line economies. This chapter provides several pieces of evidence that sauce for the goose is not sauce for the gander. As you read this chapter, remember the following crises, and how they were resolved. Large-scale devaluation of the Chinese renminbi between 1989 and 1993 meant that China could grow faster, and did. 209

But this meant that East Asian economies became less attractive destinations for investment, their export growth slowed, their currencies became overvalued (or considerably less undervalued), and their growth rate slowed. Was this all resolved without substantial cost? If only. The result was the East Asian crisis, growth rates plummeting to negative in East Asia, and, postcrisis, a substantial real weakening of these currencies relative to the Chinese renminbi. The first few years of the new century were witness to the “standing still” component of Chinese devaluation. Without any change in the official exchange rate, low inflation and excess productivity growth in China (excess relative to US productivity growth) meant that China’s exchange rate became ultracompetitive. China’s current account surplus jumped to near 10 percent of GDP and at the same time US current account imbalance became close to negative 6 percent. A coincidence? Obviously not. The exports of China’s manufacturers were bought by American consumers. The US dollar had to devalue in order to restore some balance. But the US is no ordinary economy—it is a “reserve” currency. The result of the much-needed US devaluation was a financial crisis whose magnitude exceeded any other since the Great Depression.1 Just as the world began to recover, it was hit by the European crisis of 2010 and 2011. The cause was eerily similar to the China-US experience pre-2008. Except it was for a political union with a common currency, a currency that was manifestly uncompetitive for one set of countries and overly competitive for another set led by Germany. The joint US-China external account was/is near balance. The euro area current account was/is near balance. If proof was needed that appearances can be deceiving, the two imbalanced balances would be it! It is no one’s case that exchange rate levels are the only cause of recent economic crises, or even a primary cause. There are several causes but the influence of many does not diminish the importance of some. There are reasons why the level of the exchange rate is an important factor. It affects the overall domestic economy. It affects trade with other countries. The world has become inextricably linked, and this means that the exchange rates of all countries cannot be simultaneously undervalued. Besides the adding-up constraint, there is the problem of current account imbalances. Currency undervaluation increases the growth rate and increases the current account balance. Simply put, undervaluation increases own exports, but given that these exports will have to be bought by someone else, own undervaluation increases someone else’s imports. These imbalances can go on for some time, but will need to be resolved. That defines a cycle, and a dénouement. First, the bait. Since currency undervaluation offers such sweet deliverance, there are incentives for individual countries to practice it. But all cannot follow such policies. So, second, the consequence. Countries hurt by the beggar-thy-neighbor policy react but often can do so only via a crisis. 1. To be sure, the US dollar appreciated in the immediate aftermath of the great financial crisis. But its value in 2011, approximately 7 percent undervalued, is in stark contrast to its 2 percent overvalued level in 2009.

210

DEVALUING TO PROSPERITY

Currency Wars Brazilian Finance Minister Guido Mantega used the term “currency war” when he complained about the cheap value of the renminbi, which he said gave China an unfair trade advantage: “We’re in the midst of an international currency war, a general weakening of currency. This threatens us because it takes away our competitiveness.”2 A previous period of extreme real devaluations by China has been cited as one possible cause of the Asian crisis in 1997–98 (Bhalla 1998a, 1–2; 1998b).3 These devaluations left previously stable and successful East Asian exporting economies unstable and unsuccessful. These economies later devalued their currencies in an attempt to preserve their competitiveness vis-à-vis China. If most countries have their currencies fixed, and overvalued, there is a consumer loss all around, but there is little possibility of a crisis—or a currency war. In fact, the currency wars started even earlier. As early as 1994, China was declared to be a “currency manipulator” by the US Treasury, the last time the United States has labeled any country as such. A currency war results because not every country can practice undervaluation. Every country may desire a cheap currency, but not all can simultaneously achieve it. In many ways, currency wars are a zero-sum game. Countries can grow faster, but not without causing damage to the growth performance of other economies. Can we predict what countries would do when faced with threats to their competitiveness and/or to their growth? The analysis in this book provides some indication. First, the affected countries would attempt to regain their economic strength, primarily by allowing their exchange rates to be determined less by the market and more by policy in order to bring their exchange rates to competitive levels. Witness the policy move during late 2011 by the Swiss central bank to fix a parity rate for the franc at 1.2 euros. There have been similar interventions by other central banks around the world to prevent their currencies from becoming too strong. And there has been pressure on China to revalue its currency at a rate faster than 5 percent per year. (As noted in chapter 13, a real appreciation of less than 5 percent per year will not have much of an impact in redressing global imbalances.)

Why Currency Undervaluation? The undervaluation story begins with the knowledge that cheaper wages for laborers with the same level of productivity will attract investment, especially foreign investment. This is made clear by a comparison of wages in China and India from the early 1990s to the present. This is the micro effect of exchange rate manipulation. The macro effects are abnormal profits in the form of the

2. Quoted in Financial Times (September 27, 2010). 3. See Bhalla (1998a) for an early analysis of currency wars and mercantilism.

CHANGING TIMES, CHANGING VIEWS 211

large accumulation of foreign reserves—to levels much greater than what is needed for self-insurance against economic crisis or speculative attacks. The cycle progresses from depressed real wages, to faster growth, to the accumulation of reserves. Thus, there is excess growth, and excess savings, for some countries. If the countries in question are large, there is the “benefit” of low real interest rates for the overvalued others. Lower borrowing costs induce high and excess consumption because the borrowers and lenders are different actors—that is, different economies. Soon there is a crisis, which first affects consumers drowning in debt and later banking systems struggling with bankruptcies. The 2008 crisis was the wake-up call that the system needed changing. The aftermath of the crisis, characterized by slow and jobless growth, is likely a precursor for an evolved world order with less major currency undervaluation, fewer surpluses, and higher shared global growth. The next few pages document the evidence on the evolution, consequences, and eventual resolution of imbalances over the last 30 years.

Evidence of Malfunctions and Imbalances Wages in India and China A lot of the recent attention to global imbalances has centered on the US current account and later the imbalances in Europe. Until very recently, most of the complaints concerned the renminbi’s undervaluation against the dollar. Developing economies were growing, and growing fast, and so there seemed little reason to complain. But there was. India, a much poorer country than China, had higher wages than China—substantially higher wages after accounting for educational differences. Income per capita was equal in the two economies in 1985. Today, Chinese incomes are between twice and three times the Indian level (based on PPP dollars and US dollars, respectively). In 1985, the average educational attainment in China was 5.7 years, compared with 3 years in India (Barro and Lee 2010). In 2010, China’s educational level had increased to 8.9 years, and India’s to 5.4 years. The relative education level of Chinese males to Indian males has stayed constant at around 1.5. As a result, throughout the last 25 years, Chinese wages should have been higher than Indian wages by at least 50 percent if not nearly double.4 In 2010, with Chinese income per capita at two to three times the level of Indian incomes, wages in China also should have been two to three times wages in India. To summarize, Chinese wages relative to Indian wages should begin at around 1.5 in 1985 and be anywhere between 3 and 4 in 2010. Table 14.1 lists wage and related data on China and India from various sources for selected years between 1985 and 2010. Two sources indicate that 4. This assumes that the quality of education in the two economies is near equal. Indeed, it is not, with the quality of education in China substantially higher.

212

DEVALUING TO PROSPERITY

Table 14.1

China and India: Education, income, and wages, 1985–2010 Relative wages

Relative income per capita a

b

Bargain et al. (2008)e

Relative education c

d

Male

Female

n.a.

n.a.

CHANGING TIMES, CHANGING VIEWS 213

Year

US dollars

PPP dollars

Male

Female

UNIDO

1985

0.97

1.08

1.47

2.77

0.37

EIU

n.a.

1990

0.88

1.07

1.34

2.49

0.42

0.21

n.a.

n.a.

1993

1.61

1.38

1.41

2.40

n.a.

0.43

0.72

0.72

1995

1.51

1.42

1.45

2.35

0.57

0.46

n.a.

n.a.

2000

1.99

1.51

1.52

2.21

1.35

0.82

n.a.

n.a.

2004

2.32

1.89

1.49

2.06

1.37

1.05

1.00

1.03

2007

2.58

2.11

1.47

1.98

n.a.

0.73

n.a.

n.a.

2010

3.16

2.29

1.47

1.92

n.a.

0.99

n.a.

n.a.

n.a. = not available a. Income per capita is in current US dollars. b. Income per capita is in current purchasing power parity (PPP) dollars with a 1996 base. c. Monthly wages are from United Nations Industrial Development Organization (UNIDO) and are manufacturing wages in US dollars at 2007 prices. d. Wage data from the Economist Intelligence Unit (EIU) are monthly, in current US dollars for the manufacturing/business sectors. e. Monthly wages from Bargain et al. (2008) are expressed in PPP dollars with a 2000 base. Notes: All figures presented here are ratios of China:India. Sources: Bargain et al. (2010); Barro and Lee (2010) education dataset; UNIDO figures from Yang, Chen, and Monarch (2010); Economist Intelligence Unit; author’s calculations.

Chinese wages reached parity with Indian wages in 2004; the third that such wages were 40 percent higher in China. All the sources indicate that until the mid-1990s, Chinese wages were half the level of Indian wages. Stated differently, throughout most of China’s miracle growth period, Chinese wages, adjusted for productivity, were close to one-fourth the comparator wages in the next largest economy with a vast pool of underemployed labor.5 If relative wages are engineered to stay low, a country can gain a competitive advantage over its competitors. A micro imbalance can generate macro imbalances in the form of higher savings, current account surpluses, and higher growth.

Reserve Accumulation in Countries Practicing Undervaluation A side-effect of currency undervaluation is accumulation of foreign exchange reserves. Many contend that the developing economies, especially China, are foolish to accumulate such large reserves.6 By keeping its exchange rate undervalued, the argument goes, China is actually losing money. Why? Because China and other countries like it have a higher GDP growth rate and a much higher return on capital. By accumulating dollars and investing in US Treasuries, these countries lose on two scores, first, through a lower return on capital and, second, because of the devaluation of the dollar. The argument is logical but requires several reality checks. First, by keeping the exchange rate fixed, countries can avoid the loss due to appreciation or dollar depreciation. Second, and more significantly, currency undervaluation brings about higher economic growth and higher accumulation of foreign exchange reserves—a possible imbalance for one’s trading partners. Table 14.2 presents a cost-benefit analysis for reserve accumulation by India and China over the past 11 years, from 2000 until 2011. The table reports data for three years, 2000, 2007, and 2011. The average change in currency valuation is calculated for the years between the first and second year (for example, the change for year 2007 is the average of changes between 2002 and 2007). The gains to intervention occur through higher GDP growth. The higher growth is the impact of the initial currency valuation and the changes in such valuation. By keeping the exchange rate undervalued (“competitive”), China, India, and other countries like them have been able to grow at faster rates. The costs are in the form of lower interest returns for reserves and a lower return through dollar depreciation; the benefits include higher employment, more industrial profits, and a higher GDP growth rate. For the entire 11-year period, Indian GDP growth is about 0.5 percent 5. It is correct, but another story, that most of India’s errors in labor markets have much more to do with self-inflicted wounds than with the wounds dealt by foreigners. 6. However, Dooley, Folkerts-Landau, and Garber (2003, 2005) and some others offer convincing arguments for the self-interest of countries such as China to continue undervaluing their currencies and accumulating foreign exchange reserves.

214

DEVALUING TO PROSPERITY

Table 14.2

How costly is foreign reserve accumulation? 2000

Indicator GDP (billions of dollars) Reserves (billions of dollars) Exchange rate (local currency/US dollar)

2007

India

China

India

480.00

1,200.00

38.00

170.00

2011 China

India

China

1,150.00

3,490.00

1,810.00

6,760.00

270.00

1,500.00

280.00

3,181.00

44.90

8.30

41.30

7.61

47.30

6.56

–13.00

–19.00

–29.80

–59.80

–29.80

–57.50

–2.40

–5.80

0

0.50

Average difference in 10-year government bond yields (minus United States)

3.00

–0.50

5.00

0.50

Average annual appreciation

1.10

1.20

–2.10

2.00

Average of reserves (percent of GDP)

0.17

0.30

0.18

0.50

Total loss (percent of GDP) per year

0.70

0.20

0.50

1.20

0.22

0.33

0.56

1.45

Level of undervaluation (log percent) Average change in currency valuation (log percent per year) Loss from accumulation of reserves

CHANGING TIMES, CHANGING VIEWS 215

Gains from undervaluation (via higher GDP growth) Percent gain in annual GDP from initial valuation (coefficient –0.015) Percent gain in annual GDP from change in valuation (coefficient –0.10)

0.27

0.79

0

–0.05

Total gain in GDP from currency valuation, percent per year

0.49

1.12

0.56

1.40

–0.20

0.90

0

0.20

Benefit-cost (percent per year)

Notes: The gain/loss figures in 2007 and 2011 refer to the periods 2001–07 and 2007–11, respectively. For blank cells, the conditions reported are not applicable. Each 10 percent undervaluation leads to 0.1 percent extra GDP growth. Each 1 percent average real depreciation over a period of time leads to extra 0.10 percent annual GDP growth. Sources: IMF (2006); Penn World Table 6.1 (Heston, Summers, and Aten 2002); author’s calculations (see appendix A).

higher per year; for China, GDP growth is about 1.25 percent higher per year. However, higher interest rates in India make the net gains near zero for the entire 11-year period (actually there is a small loss); for China, the net gains average nearly 1 percent per year during the earlier period, 2000 to 2007. After that, low interest rates in the United States and currency appreciation in China limited the net gains from currency intervention to only 0.2 percent per year. The calculations reveal two sharp conclusions. First, as revealed by the preferences of several Asian economies, currency undervaluation is a very potent policy for delivering extra growth, and likely more potent than other options. A large and near impossible 5 percentage point decline in the fiscal deficit is likely to yield (at most) an extra 0.5 percent of GDP growth; this is achieved by a country just by undervaluing by 15 percent in the initial year and by depreciating the nominal exchange rate by about 3 to 5 percent each year. The second conclusion is that the party cannot last forever. In China’s case, the net gains from currency intervention have declined to only 0.2 percent per year—a sharp decline from the previous seven years. If China’s currency continues to appreciate, the net gains will decline toward zero and may even become negative.

Low Real Interest Rates for the Others The accumulation of reserves can lead to excess savings, and these excess savings may lead to low interest rates, which may fuel excess consumption, which may lead to crisis. Figure 14.1 documents the trend in real long-term (10-year) bond yields since 1950. Real government bond yields of six countries, Canada, France, Germany, Italy, the United Kingdom, and the United States, were weighted by GDP (in dollars) to generate a world bond yield. A real rate of 3 to 5 percent is broadly considered historical and normal. This rate peaked at 7 percent during the anti-inflation era of the 1980s (note the persistence of negative real rates during the high-inflation 1970s). Normality resumed in the late 1980s and rates stayed in the range of 4 to 5 percent until the Asian crisis in 1997. The low of 1.2 percent occurred in 2005, with the next two years averaging rates of 1.8 and 2 percent, respectively. There were long and variable lags, but we know what happened in 2008. Note the anomaly: The US recession started in December 2007, at a time when real bond yields were the lowest of the postwar period excluding the high-inflation years of the 1970s. These low yields may have been the result of acute imbalances in the current account, which were generated by competitive devaluations. (See also discussion in Wolf 2008.)

The Seductive Appeal of Currency Undervaluation An undervalued and declining real exchange rate means that costs decline relative to productivity, which enhances profitability and attracts foreign investment. This means more technology transfer, more efficient production, and higher growth for the country that is able, or allowed, to practice currency undervaluation. 216

DEVALUING TO PROSPERITY

Figure 14.1 World real interest rate, 1950–2011 percent 7 6 5 4 3 2 1 0 –1 –2 –3 1950 1954 1958 1962 1966 1970 1974 1978 1982 1986 1990 1994 1998 2002 2006 2010 Notes: The world real interest rate is the average of the real interest rates of Canada, France, Germany, Italy, United Kingdom, and United States, weighted according to dollar GDP. Source: Global Financial Database, www.globalfinancialdata.com/Databases/databases.html (accessed in January 2012).

The pattern of changes in aggregate currency valuation levels is revealing. Table 14.3 documents the trend in mean valuation levels on a global basis as well as for countries that have overvalued and undervalued currencies. For both categories, the levels are aggregated using world trade share as weights. The numbers suggest that world currencies have moved steadily to a net negative valuation over the course of the last 50 years. In the 1960s, in the aggregate, there was a 3 percent undervaluation, which moved to a 10 percent overvaluation for the next three decades. In just one decade after the 1990s, there was a sharp decline in aggregate currency valuation to a historical low of –11 percent. This was caused by the increase in the aggregate undervaluation level to 20 percent in the 2000s from nearly one-fourth that level during the previous three decades. Not only did the undervaluation level increase; the overvaluation aggregate level declined from 20 percent in the 1970s, 1980s, and 1990s to a historical low of only 9 percent in the 2000s. By the time the phrase “currency war” CHANGING TIMES, CHANGING VIEWS 217

Table 14.3 World crisis foretold through world currency valuations World currency valuation Net world currency valuation (percent)

World current account balance as percent of world GDP

Growth in world income per capita (percent)

Mean for overvalued currencies

Mean for undervalued currencies

1960s

13.4

–15.9

–2.5

0.0

4.1

1970s

20.0

–8.4

11.6

0.5

2.9

1980s

18.5

–6.4

12.2

0.9

2.3

1990s

15.5

–5.0

10.5

0.8

2.5

9.1

–19.8

–10.7

1.3

3.7

12.7

–23.7

–11.0

1.1

4.1

Decade

2000s 2010–11

Notes: Countries are grouped according to their positive or negative value of currency valuation. For each group (positive and negative) the world currency valuation is the weighted average of individual country valuations (weights according to share in world trade). The net currency valuation column is essentially the addition of the positive and negative world currency valuations. The “surplus” for the world is estimated by summing the absolute value of the current account balance for each country and dividing by 2. See text and Blanchard and Milesi-Ferretti (2009) for details. Sample of countries excludes oil-exporting countries and countries of Eastern Europe and part of the former Soviet Union. Source: Bhalla (2007a) dataset extended to 2011.

was first being murmured, the event itself was at least a decade old. The data confirm the hypothesis that action begets reaction—countries tried to be proactive with regard to ensuring the competitiveness of their currency, especially after China practiced deep undervaluation and became ultracompetitive. If the aggregate movement was toward undervaluation, then the aggregate current account balance should also have worsened. It did. Chapters 3 and 7 documented that there was a consistent negative relationship between a country’s current account balance and its currency valuation level—that is, the greater the undervaluation level, the higher the current account balance. This coefficient was robustly centered around –0.04. That means that each 10 percentage point increase in individual country undervaluation increases its current account balance by 0.4 percent of GDP. On an aggregate global basis, the current account is always in balance— world exports are equal to world imports. In order to help isolate the determinants of global imbalances, following Olivier Blanchard and Gian Maria Milesi-Ferretti (2009), the absolute value of each country’s current account balance is aggregated. Table 14.4 presents the results for a simple regression, with world current account balances as percent of world GDP as the dependent variable and average world negative currency valuation as the only independent variable. The equation is estimated for 32 years, from 1980 to 2011. The coefficient on valuation is –0.031, and the R2 is 0.64. There are three outlier years, 2006, 2007, and 2008. A dummy variable for these years decreases the coefficient to –0.025 and the R2 increases to 0.73. Figure 14.2 plots the share of this aggregate in world GDP against the mean undervaluation level for the coun218

DEVALUING TO PROSPERITY

Table 14.4

Effect of negative currency valuation on world current account balances, 1980–2011 Coefficients

Row

Dependent Variable

Negative currency valuation

1

Current account balance (percent of GDP)

–0.031***

2

Current account balance (percent of GDP)

–0.025***

Time dummies, 2006/2007/ 2008

0.33***

R2

Number of observations

0.64

32

0.73

32

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. The world current account balance is the sum of the absolute value of the current account balance (in current US dollars) divided by 2 for each country. Countries are grouped according to their positive or negative value of currency valuation. For each group, (positive and negative) the world currency valuation is the weighted average of individual country valuations (weights according to share in world trade). The estimated regression equation is: World current account balance (% GDP) = β0 + β1 (negative currency valuation). In row 2, the same equation is estimated with a time dummy for the years 2006, 2007, and 2008. Source: Bhalla (2007a) dataset extended to 2011.

tries with undervalued currencies.7 The graph confirms that one reason for the large increase in global imbalances in the 1990s was that more countries were practicing undervaluation and/or making their currencies even cheaper. Note that this evidence is for the world, and it is striking in similarity to the evidence of individual countries. When an individual country currency becomes more undervalued, it improves its current account balance—an individual country economic good. When a large amount of world output is produced under the auspices of currency undervaluation, the world current account balance worsens—a world economic bad.

One Country’s Ceiling Is Another Country’s Floor— Evidence on Stolen Growth China’s growth has been somewhat miraculous, but as shown in previous chapters, a fairly large part of the “excess” growth over what it should have been was the result of one policy—deep currency undervaluation. This was first accomplished between 1978 and 1993 by means of the fastest and largest real devaluation anywhere, anytime. The value of the renminbi in 1993 became close to one-third of its real value in 1977. Over the subsequent 17 years, from 1994 to 2011, the Chinese currency further depreciated in real terms to close to one-half its real 1993 value, or close to one-sixth its 1977 value. Size does matter, especially with regard to global imbalances. If a small or 7. Oil-exporting countries and those of the former Soviet Union and Eastern Europe are excluded from analysis.

CHANGING TIMES, CHANGING VIEWS 219

Figure 14.2 World current account balances affected by currency valuations, 1980–2011 world current account balance (percent of world GDP) 2005

1.5

1.3

2004

2010

2000

2003

2009

1986 1999

2002

1.1

2001

1985

1998

1988 1989

1984

0.9

19941997 1990 1992 1993 1983

0.7

0.5 –25

1987

1996 19821995 1981 1991 1980

–20

–15

–10

–5

0

curency valuation of countries with currency undervaluation Notes: The world current account surplus is the sum of the absolute value of the current account balance (in purchasing power parity dollars) divided by 2 for each country. Oil-exporting economies and countries of the former Soviet Union and Eastern Europe are excluded from analysis. Countries are grouped according to their positive or negative value of currency valuation. For each group (positive and negative) the world currency valuation is the weighted average of individual country valuations (weights according to share in world trade). See notes to table 14.4 for the estimated regression function. Source: Bhalla (2007a) dataset extended to 2011.

even a medium-sized country achieves miraculous growth by undervaluing its currency, it will upset some neighboring countries, particularly its competitors, but it will have only a marginal impact on the global economy. It is another matter altogether when the country that achieves miraculous growth through large currency devaluations accounts for more than one-fifth of the world’s population. As noted throughout this book, undervaluation is a very old model for pursuing growth that was practiced in the 19th century by some of the wealthiest countries in the world today, including the United Kingdom, the Netherlands, and to a lesser extent the United States. But what is new about China practicing undervaluation is its sheer size. There is also the question of how much growth is subtracted from other countries when a large country practices large real devaluations. It may be true that China’s growth helped neighboring countries supply extra inputs, much like England’s colonies benefited from supplying cheap imports to England in the 19th century. But the counterfactual may be more important—what would 220

DEVALUING TO PROSPERITY

the growth rate of these countries have been if China (or colonial England) were not practicing extreme undervaluation? The currencies of their neighbors and competitors would have been cheaper, and those economies would have had higher growth rates. This most likely was the story behind the Asian crisis of 1997–98. Table 14.5 is an initial attempt to evaluate the effects of the changes in currency valuations by a large country or region on the growth of individual countries. Two periods are considered: 1980 to 2011 and 1995 to 2011. The change in average currency valuation of each region is tested for its effects on the growth rate of individual countries within the context of the five-year panel data used throughout this book. Three large regions are identified: China, the euro area (the original 12 countries), and the United States. Each column contains the results of a fixed effect regression excluding the country or countries belonging to the selected region. For neither period does the valuation of the euro have any effect on other countries’ growth. For 1995 to 2011, the average annual rate of change in the value of the US dollar was 0.83 percent per year. This appreciation had a positive effect on other countries’ growth—a more expensive dollar hurt US growth but helped the growth rate of other countries. The effect of the change in China’s exchange rate is opposite. For both periods the coefficient is significant and negative. Each 1 percentage point increase in the annual rate of change in depreciation of the Chinese renminbi leads to a 0.21 percentage point decline in an average country’s growth rate. This is the strongest proof that currency undervaluation, especially for a large country, is a beggar-thy-neighbor policy. The result for the US dollar is the same—a movement toward strengthening of the US dollar hurts US growth but is helpful to other countries growth.

Breaking the Cycle The lessons from the crises of the last several years are quite apparent. Real devaluations on the part of one country beget other nations to follow similar beggar-thy-neighbor policies. Since all countries cannot simultaneously engineer an undervalued currency, it is inappropriate and unfair for one or a few countries to be allowed to engineer large real devaluations. The cycle must be broken. There are encouraging signs of change. The very fact that currency wars are being discussed openly by senior policymakers is welcome. As is the admission by China that it sees the future course of its currency to be appreciation. So too is the recognition by China that the consumption share of its GDP must increase from the abysmally low levels of about 35 percent. And the recognition that a weaker dollar relative to Asian currencies is in the interests of US and global growth. And, finally, the recognition that part of the euro area’s problem is the overvaluation of the euro for the poorer states on the periphery of Europe. The world has gone through an extensive financial crisis over the last four years, 2008 through 2011. The great financial crisis and the euro area crisis have made us all remember our interdependencies, and the events after World CHANGING TIMES, CHANGING VIEWS 221

222 DEVALUING TO PROSPERITY

Table 14.5

Is there any evidence of stolen growth? 1980–2011 China

Euro area

1995–2011 United States

China

Euro area

United States

Currency valuation Initial (Lagged) change

–0.017***

–0.022***

–0.02***0

–0.047***

–0.026***

–0.013**

0.005***

0.009***

–0.004***

–0.022***

–.008***

–0.04***

Region dummy effect China

0.077***

Euro area

0.21**** 0.037***

United States

0.06**** –0.045***

–0.33***

Means and computations Average change in region (percent per year)

–7.22*0**

0.19***o

0.71***0

–3.65****

0.55****

0.83***

Net effect on growth (obtained as the product of mean and change coefficient)

–0.56*0**

0.01***0

–0.03***0

–0.77****

0.03****

–0.27***

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1. Data are compiled in five-year intervals with the last five-year observation comprising years 2005 to 2011; see appendix A for details. A positive coefficient for region dummy means that currency depreciations in the host country have a negative effect on the growth rates of other countries. Between 1995 and 2011 China’s currency valuation registered an average depreciation of 3.65 percent per year; this yields a gross decline in other countries’ average growth rate of 0.77 percent (product of 0.21 and 3.65). Or each 1 percentage point increase in China’s undervaluation leads to a 0.21 percentage point decline in the growth rate of other countries. Source: Bhalla (2007a) dataset extended to 2011.

War I and preceding World War II. Growth is of the essence to every economy; when world growth falters, there is political upheaval, war, and other consequences. It is emphatically not in the interest of individual countries to have large world imbalances. And that is the major reason to be hopeful about the future—change is coming, change we can hardly believe in.

CHANGING TIMES, CHANGING VIEWS 223

15 Conclusion

Experience seems to most of us to lead to conclusions, but empiricism has sworn never to draw them. —George Santayana, Character and Opinion in the United States Just a few years ago, the overwhelming view among experts was that the efficacy of affecting overall GDP growth through currency undervaluation was questionable, even for such seemingly obvious candidates as China. The underlying basis for the policy of intervening in currency markets to keep the price of a currency cheap was considered to be doomed to failure on both theoretical and empirical grounds. Blaming China for undervaluing its currency in pursuit of growth was considered a severe case of envy. Large current account surpluses were considered a sign of virtue, just as large current account deficits were a sign of vice. Times have changed. And the world may have the Great Recession to thank for some illumination. It is no longer so controversial to state that a country can change its level of competitiveness by changing its nominal exchange rate. A country can devalue its way to prosperity, sometimes at the expense of other nations. That is indeed the major conclusion of this book. Some stylized facts. Currency undervaluation can typically add about 0.5 percent to GDP growth per capita. And currency overvaluation can typically subtract the same amount. That the other really important determinant of growth rates—catch-up—adds about the same is testament to the efficacy of the currency undervaluation model.

Currency Undervaluation Affects Investment and Generates Growth How does currency undervaluation produce extra growth? It works via investment, a finding backed by strong empirical support. Devaluation decreases the

225

dollar costs of production1 and increases profitability; investors respond by increasing the investment rate, which propels the growth rate higher, and the virtuous cycle continues. Since undervaluation also helps productivity growth (investments are more technology-intensive), fast growth is not the result of extra factor inputs alone, as (mistakenly) alleged to have been the case for East Asia. The analysis in this book subjected the currency undervaluation–growth model to a wide variety of sensitivity and robustness tests, across a wide variety of time periods (historical from 1870 to 1938, and contemporary from 1980 to 2011, and for periods in between), using a wide variety of estimation techniques, and through joint testing with a wide range of variables (including health, education, openness, geography, initial conditions, fiscal deficits, and more). The end result is the same: Currency undervaluation matters for growth. The term currency undervaluation is nearly synonymous with export-led growth, and export-led growth has been projected by many to be the primary policy for development (Balassa 1978, Krueger 1990). What is different about these results if this same conclusion has been reached before? What is different is that this is the first analysis that supports the theory with empirical results that currency undervaluation matters. Existing measures of currency valuation are flawed by measurement errors; the measurement used here corrects for such errors and shows that currency valuation is an obvious explanation of high growth.

The Dual of Currency Undervaluation Is Currency Overvaluation This book has provided considerable evidence that currency undervaluation helps growth—and equivalently that currency overvaluation hurts growth. If something has worked for a part does not mean that it should work for the whole. Sauce for the goose is not sauce for the gander. For every country benefiting from currency undervaluation, there is an example of a country hurt by currency overvaluation. This is definitional, and the evidence supports the existence of this equivalence. Not every country is allowed the luxury of currency undervaluation. A well-functioning world abhors imbalances; a crisis-prone world welcomes such deficits. Thus, readers of this book should not interpret the evidence that countries should devalue their way to prosperity. They can, but only if the world allows them to do so. The evidence suggests that small countries can practice undervaluation for extended periods; that countries are even encouraged to decrease the overvaluation levels of their currencies; that countries can devalue their way to prosperity as long as their current account is in deficit. But 1. As noted, currency undervaluation does increase the dollar costs of imported inputs. However, the reduction in cost of domestic resources of production (primarily labor) is likely to swamp any effect of increased costs of imported inputs.

226

DEVALUING TO PROSPERITY

being large, having a deeply undervalued currency, and having current account surpluses are a recipe for individual success and a world disaster.

The Real Exchange Rate Can Be Influenced by the Nominal Exchange Rate A prominent objection to the proposition that real devaluations can lead to faster growth—besides the absence of empirical results—has been on theoretical grounds, namely, that such devaluations cannot be achieved via policy because the real exchange rate is endogenous. In other words, devaluations would lead to excess demand, which would lead to excess inflation, which would wipe out the initial advantage of lower costs. The failure of the predicted excess inflation to materialize in most countries that have undervalued currencies points out that the real exchange rate is not endogenous in practice, despite the theory. Example after example—from the United Kingdom’s “Black Wednesday” devaluation to the remarkable series of devaluations undertaken by China between 1978 and 1993—demonstrates that excess inflation does not necessarily follow a nominal devaluation. China devalued the renminbi by a cumulative 201 percent between 1980 and 1995, yet its inflation rate between 1996 and 2011 exceeded that of the United States by barely 1.2 percent per year, or a cumulative amount of only 21 percent over 16 years. The theory that the real exchange rate cannot be affected by nominal exchange rate changes would predict an excess inflation level of more than 200 percent for that same period. Over the last three decades, and perhaps even before, the nature of inflation transmission changed; globalization has meant that international effects are more pronounced than those pertaining to domestic demand. Thus, the presumed automatic linkage between domestic demand and inflation has weakened. This has an important implication. Countries’ inflation rates have converged, and inflation due to excess domestic demand does not follow, which means that a nominal currency devaluation can and does become real.

The Real Exchange Rate Can Be Influenced by Standing Still The common perception is that a policy directive is required in order to achieve a real devaluation. Not so. Indeed, more often than not, developing economies have achieved a substantial real devaluation, and a substantial reduction in costs, simply by keeping the exchange rate fixed. An important stylized fact of development is the Balassa-Samuelson effect, according to which a currency has a tendency to appreciate with fast growth. In recent years, this effect has been thwarted by countries intervening to keep their exchange rates “competitive.” China took the lead and other, especially Asian, countries followed. Such interventions to keep the currency cheap, in the absence of excess inflation, have led the Asian currencies to become even more undervalued. This “standing still” real depreciation is a large part of the currency story in the wake of the Asian crisis of 1997–98—and one of the important conclusions of this book. CONCLUSION

227

Mercantilism Is Alive and Well Closely related to the proposition that currency undervaluation helps growth is the controversial linking of such currency policy to the practice of mercantilism. This book defines mercantilism as the coexistence of a large undervalued currency and a large current account surplus. Using the average of these two measures, the most mercantilist country in the last decade has been Malaysia, followed by Singapore, Taiwan, Thailand, Hong Kong, and China.

There Are Parallels between 1870 and 1950 The analysis of historical 19th century data yielded some surprising results on the real exchange rate in developed and developing economies. First, it was found that for several developing economies, the real exchange rate in 1950 was the same as in 1870 and that the rate in 1870 was heavily overvalued. Second, the colonial powers, particularly England and the Netherlands, had the most undervalued exchange rates in the late 19th century. Perhaps not coincidentally, their growth rate per capita was also higher than average.

Institutions Don’t Rule In recent years, the views of economists and policymakers worldwide have converged upon the belief that Western-style institutions are the key to longterm growth. There has been a small minority of detractors, and the results of this book firmly support this minority. Using several key variables for institutions, and several instruments for these institutions (as required by simultaneity economics), institutions proved to be significant determinants of growth or levels of development in fewer than 23 percent of the large number of cases examined. Geography does not explain growth differences either. What does? Again, the answer is currency undervaluation, whose variables are significant in more than 70 percent of the samples considered for a range of different countries, specifications, instruments, and proxies for institutional development.

A Postcrisis Realignment Among the other conclusions that can be drawn from the analysis in this book, one of importance pertains to the forecast for the value of the dollar, the flip side of currency devaluation by other countries. Currency misalignments have consequences; large misalignments have large consequences. Regardless of what may have caused the 2008 financial crisis, the fact remains that it caused a great upheaval. And if a similar crisis can be prevented in the future, then the citizens of the world should do so for the benefit of all. In fact, something can be done. Currency misalignments should be corrected as a necessary first step toward a more harmonious economic and 228

DEVALUING TO PROSPERITY

political future. One of the important findings of this book is that the global misalignment has little to do with either Europe or Japan. The major realignment of the dollar has to come with respect to Asia (excluding Japan), and especially China. China has a misalignment of its currency as large as 45 percent, and a misalignment that is increasing by 3 to 5 percent per year (due to the standing-still argument). Never has a currency been so misaligned for so long, much less in so large and important an economy. Add to this the need of China’s Asian neighbors, including Japan, to stay competitive, and the problem becomes worse. What are the chances that China will begin to revalue its currency? They are very high, despite the fact that the change in the value of the renminbi since June 2010 has been small. The need for a postcrisis currency realignment is something all global players recognize as inevitable—as inevitable as the Balassa-Samuelson effect. Before 2008, the world was different—there was no such alignment of natural interests because we did not realize how interdependent the global economy was. Now we do, and the future will be different.

CONCLUSION

229

Appendices

Appendix A Data and Methods

Real Income Penn World Table data (Version 6.1) are available from 1950 onward and form the basis of the tests of various hypotheses. For years prior to 1950, real income data are available from Maddison (2001), and the two datasets are linked with 1996 as the base. Taylor (1996) tests for the evolution of the real exchange rate (RER) for 23 countries for 1884–1996.1 These data consist of price data based on the consumer price index (CPI) or GDP deflator and nominal exchange rate data. The Taylor data were supplemented for India, China, and some other Asian countries. This provides data on nominal exchange rate and price indices for years prior to 1950. In an analogous fashion to the Western countries, these data were also linked to the 1950 purchasing power parity (PPP) exchange rate. The same formula relating the RER to real income is used to obtain the predicted RER for years prior to 1950. As a result, for some countries real income, the RER, and predicted RER (after accounting for Balassa-Samuelson effects) are available for years prior to 1950. See tables 4.1, 4.2, and 4.3 for an exposition and examples. The base source of per capita income data is Penn World Table Version 6.1, which contains incomplete panel data for nominal and real income per capita in PPP international prices for 1950–2000.2 The gaps in country-specific real income data from the Penn World Table are filled in by using data from Maddison (2001).3 The linkage factor used is the ratio of the Penn World Table 1. The author thanks Alan Taylor for making these data available. 2. Available at http://pwt.econ.upenn.edu/php_site/pwt_index.php (accessed on April 30, 2010). 3. Statistics on world population, GDP, and GDP per capita, for years 1 AD to 2008 are available at www.ggdc.net/maddison (accessed on April 30, 2010).

233

and the Maddison data for 1996. The latter are available for selected years from 1 AD until 2003. This 1996 base income per capita series is then extended to 2009 by using data on real GDP growth per capita from the IMF (2011).

Population Country population data were obtained from the US Census Bureau (www. census.gov) and Maddison (2001). The Census Bureau’s population figures for 1950–2025 are linked backward using country, region, and world population data from Maddison from 1 AD. Where Maddison data for individual countries are not available, regional population growth rates are used to compute the missing population data. Linking the two sets of data provides estimated population data for individual countries for selected years between 1 AD and 1950 and for each year thereafter until 2025.

Real Exchange Rate The RER is defined as the ratio of the PPP exchange rate to the dollar exchange rate. RER data since 1950 are obtained from the Penn World Table and updated by the World Bank’s World Development Indicators and the IMF’s World Economic Outlook datasets. Data prior to 1950 on the constituents of the RER—inflation and nominal exchange rates—are taken from various sources, especially Taylor (1996).

Equilibrium Real Exchange Rate (RER*) This is defined via a nonlinear equation whose values are as follows: RER* = 1.11 × (1 – .97Y), where Y is income per capita per day in 1996 PPP prices.

Undervaluation and Overvaluation (UV) Defined as the log ratio 100 × (RER/RER*); a negative value indicates undervaluation, a positive value indicates overvaluation. The one period change in UV is defined as dUV.

Average Change in UV and dUV Defined as the change per year in the magnitude of valuation; a negative sign value indicates that the currency has moved from a more expensive valuation to a less expensive valuation. It can mean that a currency is becoming less overvalued or that it is becoming more undervalued.

234

DEVALUING TO PROSPERITY

Tariffs Tariff data are collected from many sources. Data for the years after the 1980s are collected from the World Bank’s World Development Indicators database. Data prior to 1950 are collected from O’Rourke (2000).

Labor Force Defined as the percent of the population between the ages of 15 and 64 in the labor force. The source is the World Bank’s World Development Indicators database.

Capital Stock The Nehru and Dhareshwar (1993) series is used until 1990; from 1990 onward, investment (deflated by GDP deflator) and depreciation of 5 percent are used to extend the series.

Total Factor Productivity Growth This is estimated using a conventional Cobb-Douglas model with country dummies (fixed effects model) and a time dummy for the years 1973–83.

Distribution of Income (Quintile) The primary source of the income distribution data is the United Nations University’s World Income Inequality (WIDER) 2008 database. These panel data range from 1950 to 2002. Data prior to 1950 are obtained from Bourguignon and Morrison (2002), who report income distribution quintiles for 33 countries and several regions for 1820–1992. When individual country data are not available, regional income distribution data are used. The pooling of the two sets allows the construction of individual country quintile data from 1820 to 2003. Income distribution data for years prior to 1820 are assumed to be the same as 1820; for years after 2003, they are assumed to be the same as 2003. This assumption makes possible an income distribution (in quintiles) series from 1 AD to 2025.

Distribution of Income (Percentiles) Quintile data are converted into percentile data according to the modified Kakwani procedure outlined in Bhalla (2002a).

Middle Class Defined as the proportion of people whose income per capita is equal to or higher than the weighted average of poverty lines in developed economies. APPENDIX A 235

This weighted average is observed to be $8.20 per capita per day in 1996 PPP prices, or PPP$11.2 in 2011 international prices. The middle class extends to 10 times this initial value.

Geography Four geography variables obtained from Parker (1997) are used: latitude, mean temperature, minimum rainfall, and average rainfall. Variables on the proportion of surface land in tropical areas are obtained from Bosworth and Collins (2003).

Institutions Property Rights Average protection against risk of expropriation (from the International Country Risk Guide index) is used as an index of property rights. These data for the 1980s are obtained from databases developed by Sachs (2003); Acemoglu, Johnson, and Robinson (2001); and Albouy (2006). These data were previously used by Knack and Keefer (1995) and were organized in electronic form by the IRIS Center (University of Maryland). The original compiler of this data is Political Risk Services.

Executive Constraints Data are obtained from Polity IV datasets available at www.systemicpeace.org/ polity/polity4.htm.

Political Liberties This database is obtained from the Freedom House datasets available at www. freedomhouse.org.

Colonial Heritage These data are from Bhalla (1997) and expanded from information available from a variety of general sources.

Governance Data on control of corruption, rule of law, political stability and absence of violence, voice and accountability, government effectiveness, and regulatory quality are obtained from Kaufmann, Kraay, and Mastruzzi (2005).

236

DEVALUING TO PROSPERITY

Table A.1

Country composition in Bhalla (2007a), gross country panel dataset, 1950–2011 Developed economies plus

Total

Developing economies

Russia and Eastern Europe

OECD plus

Asia

Row

Sample

A

All countries with data on population and income

204

153

28

23

39

B

All countries with data on population and income and real exchange rates

181

130

28

23

31

C

Countries with population < 1 million and data on population and exchange rates

34

32

0

2

11

D

Oil exporters and population >1 million

22

17

4

1

0

E

Selected sample (all minus B)

137

89

27

21

18

OECD = Organization for Economic Cooperation and Development Notes: The “developed plus” countries include the “OECD plus” and the former Soviet Union and East European countries. The “OECD plus” countries include Australia, Canada, West European countries, Japan, and the United States.

Instruments Several variables are used as instruments for the estimation of equations relating institutions to economic growth. Variables on settler mortality in the 19th century, urbanization and population density in the 16th century, and the share of European population in 1900 are obtained from Acemoglu, Johnson, and Robinson (2001). Data on the middle class in the mid-19th century are the author’s computations; data on educational attainment in the late 19th century are obtained from Morrisson and Murtin (2003).4 Table A.1 outlines the construction of the Bhalla (2007a) dataset, which is used throughout the analysis to gauge currency valuation. Table B.1 in appendix B lists the sample of countries used in the analysis along with values for per capita income, growth, and currency undervaluation.

4. The author thanks Christian Morrisson and Fabrice Murtin for making their data available.

APPENDIX A 237

Appendix B Bhalla (2007a) Dataset Extended to 2011

239

240 DEVALUING TO PROSPERITY

Table B.1

Country data, 2011 Annual income

Country

World Bank country code

Population (millions)

PPP, 1996 base (billions of dollars)

Current (billions of US dollars)

Per capita, PPP, 1996 base (dollars)

Annual growth per capita, 1980–2011 (percent)

Currency valuation (percent)

Exchange rate Local currency to US dollar

PPP, 1996 base

2011

Average log change per year, 1980–2011

Albania

alb

3.7

28

13

7,690

2.0

99.3

43.2

13.4

–0.8

Algeria

dza

35.0

298

183

8,518

0.8

79.4

45.2

36.6

–0.4

Argentina

arg

42.0

884

435

21,049

1.4

4.2

2.0

–38.5

–3.3

Armenia

arm

3.0

26

10

8,643

1.0

382.0

116.1

–27.3

4.9

Australia

aus

21.8

932

1,507

42,656

2.0

0.9

1.5

63.3

0.9

Austria

aut

8.2

303

425

36,907

1.8

9.8

12.8

38.1

0.1

Azerbaijan

aze

8.4

125

69

14,892

2.1

0.8

0.6

19.3

7.5

Bangladesh

bgd

153.0

537

115

3,511

3.3

73.6

13.9

–9.2

–4.7

Belarus

blr

9.6

248

58

25,873

2.9

4,244.0

1,739.0

–50.5

6

Belgium

bel

10.4

380

529

36,412

1.6

28.5

38.2

42.9

–0.2

Benin

ben

9.3

17

7

1,849

1.0

464.1

204.6

293.0

–1.6

Bolivia

bol

10.0

46

24

4,538

0.3

6.8

3.4

96.1

0.4

Bosnia and Herzegovina

bih

4.6

55

18

11,877

2.7

1.4

n.a.

–30.3

–1.6

Botswana

bwa

2.0

31

16

15,896

4.0

7.1

4.1

–11.2

–2.7

Brazil

bra

202.0

2,477

2,518

12,251

1.2

1.6

1.6

71.6

1.6

Bulgaria

bgr

7.1

94

54

13,306

0.3

1.4

0.7

–6.8

–1.8

Burkina Faso

bfa

16.3

27

10

1,629

1.6

464.0

174.8

263.0

–2.5

Burundi

bdi

9.9

7

2

717

–1

1,287.0

278.5

361.0

–1.2

Cambodia

khm

15.1

48

13

3,188

3.0

4,051.5

1,618.4

112.0

4.8

Cameroon

cmr

19.7

63

26

3,188

0.3

487.4

179.7

100.0

–1.6

Canada

can

34.3

1,397

1,759

40,774

1.5

1.0

1.2

30.4

0.8

Central African Republic

caf

4.9

6

2

1,278

–2.1

459.5

194.6

447.0

3.2

Chile

chl

16.9

303

243

17,933

3.0

474.5

340.3

1.3

–1.8

China

chn

1343.0

20,137

6,988

14,992

7.9

6.6

2.4

–43.7

–7.0

Colombia

col

45.9

438

321

9,531

1.6

1,903.9

1,097.4

22.4

–0.7

Congo

cog

4.3

13

15

2,952

1.1

465.1

606.0

651.0

0.3

Costa Rica

cri

4.4

46

40

10,587

1.2

515.0

336.6

29.9

–0.6

Cote d’Ivoire

civ

19.8

47

24

2,355

–1.6

465.9

234.5

301.0

0.5

Croatia

hrv

4.5

71

64

15,768

0.9

5.4

4.4

29.6

2.8

Czech Republic

cze

10.2

275

220

26,984

1.5

17.3

13.0

–10.0

1.9

Denmark

dnk

5.5

211

349

38,167

1.4

5.2

8.7

75.4

0.7

Dominican Republic

dom

9.9

105

54

10,603

3.2

39.4

18.2

–7.3

–2.9

APPENDIX B 241

Ecuador

ecu

15.1

94

65

6,237

0.3

1.0

0.5

52.7

–0.7

Egypt

egy

88.6

681

232

7,685

2.7

5.7

2.2

–0.6

–1.5

El Salvador

slv

6.4

43

23

6,627

0.5

8.8

4.2

37.7

1.2

Eritrea

eri

6.0

6

3

936

–0.7

15.5

8.5

819.0

0.9

Estonia

est

1.3

27

23

21,083

2.3

11.1

7.5

–10.4

5.4

Ethiopia

eth

91.1

134

31

1,470

1.8

16.8

2.9

85.7

–3.5

Finland

fin

5.3

202

271

38,414

2.0

4.2

5.5

38.0

–0.1

France

fra

63.2

2,070

2,808

32,762

1.3

4.6

6.1

44.0

–0.1

(continues on next page)

242 DEVALUING TO PROSPERITY

Table B.1

Country data, 2011 (continued) Annual income

Country

World Bank country code

Population (millions) 1.6

Currency valuation (percent)

Exchange rate

PPP, 1996 base (billions of dollars)

Current (billions of US dollars)

Per capita, PPP, 1996 base (dollars)

Annual growth per capita, 1980–2011 (percent)

18

17

11,555

–0.1

462.9

Local currency to US dollar

Gabon

gab

Gambia

gmb

1.9

4

1

2,060

0.5

29.1

Georgia

geo

4.6

62

14

13,414

–0.5

1.7

Germany

deu

82.2

2,815

3,629

34,230

1.5

1.4

Ghana

gha

24.8

65

39

2,605

1.7

Greece

grc

10.8

277

312

25,741

Guatemala

gtm

13.8

78

47

5,652

Guinea

gin

11.0

45

5

Guinea-Bissau

gnb

1.6

1

Haiti

hti

9.2

25

Honduras

hnd

8.1

Hong Kong

hkg

Hungary

hun

India

PPP, 1996 base 453.2

2011

Average log change per year, 1980–2011

79.6

1.2

6.8

91.2

–1.3

0.5

–53.4

5.2

1.7

34.1

–0.4

15,084.7

5,723.7

129.0

–5.8

1.2

240.7

275.9

42.4

1.1

0

7.8

4.7

98.9

1.0

4,068

0.5

7,298.4

2,745.6

62.7

–2.6

1

885

1.2

502.6

159.0

469.0

–2.8

7

2,748

2.1

39.0

15.0

130.0

–1.8

26

17

3,189

0.1

18.9

10.0

185.0

0.2

7.1

377

247

52,961

3.7

7.8

4.8

–38.9

–1.6

9.9

181

148

18,405

1.5

189.9

154.0

16.4

1.5

ind

1,188.0

7,860

1,843

6,618

4.7

47.3

12.2

–25.8

–5.0

Indonesia

idn

246.0

1,838

834

7,482

3.5

8,673.7

3,685.0

10.1

–2.9

Ireland

irl

4.3

179

222

41,717

3.6

0.6

0.7

32.0

–0.7

Israel

isr

7.1

203

245

28,575

2.0

3.7

3.7

15.8

–0.2

Italy

ita

58.0

1,740

2,246

29,980

1.1

1,370.1

1,704.5

41.7

0.8

Jamaica

jam

2.9

15

15

5,060

0.2

87.5

65.2

167.0

0.2

Japan

jpn

126.0

4,586

5,855

36,285

1.7

80.4

105.2

40.2

0.6

APPENDIX B 243

Jordan

jor

7.0

48

28

6,793

0.7

0.7

0.4

68.4

–1.4

Kazakhstan

kaz

15.5

360

180

23,180

2.0

146.8

78.6

–31.5

14.8

Kenya

ken

41.1

79

36

1,924

0.5

82.3

29.6

206.0

–1.8

Korea

kor

49.2

1,611

1,164

32,726

5.3

1,082.2

681.4

–30.7

–2.6

Kuwait

kwt

2.9

80

171

27,663

–0.7

0.3

0.5

101.0

1.8

Kyrgyzstan

kgz

5.5

31

5

5,682

–0.9

48.0

8.2

–44.9

7.7

Laos

lao

6.9

24

8

3,432

2.8

7,967.0

2,600.6

59.9

–4.8

Latvia

lva

2.2

38

27

17,207

1.1

0.5

0.3

–3.8

13.0

Lebanon

lbn

3.9

45

41

11,340

1.1

1,518.9

1,228.5

56.2

0.4

Liberia

lbr

3.9

8

1

1,988

0.3

1.0

n.a.

426.6

0.8

Libya

lby

6.6

29

76

4,412

–3.0

1.3

1.1

232.0

1.9

Lithuania

ltu

3.5

58

43

16,481

0.7

2.4

1.6

–5.4

11.2

Macedonia

mkd

2.1

19

10

9,061

1.2

51.6

24.4

7.0

1.0

Madagascar

mdg

21.9

24

9

1,086

–1.0

2,208.9

1,008.7

576.0

–0.3

Malawi

mwi

15.5

20

6

1,309

1.3

158.4

47.0

263.0

–2.4

Malaysia

mys

26.6

476

248

17,901

3.3

3.4

1.6

–30.4

–2.5

Mali

mli

13.4

24

11

1,794

1.1

464.3

195.2

279.0

–1.3

Mauritania

mrt

3.5

8

4

2,222

–0.4

286.7

127.5

229.0

0.2

Mauritius

mus

1.3

36

11

27,714

4.2

30.2

9.1

–64.9

–2.6

Mexico

mex

114.0

1,486

1,185

13,068

0.8

11.8

8.6

25.5

0.2

Moldova

mda

4.3

23

7

5,276

–1.2

11.7

3.8

12.2

9.0

Mongolia

mng

3.1

9

9

2,941

1.9

1,221.4

846.8

289.0

–4.4

(continues on next page)

244 DEVALUING TO PROSPERITY

Table B.1

Country data, 2011 (continued) Annual income

Country

World Bank country code

Morocco

Annual growth per capita, 1980–2011 (percent)

Currency valuation (percent)

Exchange rate Local currency to US dollar

Average log change per year, 1980–2011

Population (millions)

PPP, 1996 base (billions of dollars)

Current (billions of US dollars)

mar

34.4

246

102

7,139

1.9

8.0

3.2

6.6

–2.3

Mozambique

moz

22.5

58

12

2,574

1.7

31,631.9

5,689.5

17.8

–3.3

Namibia

nam

2.2

18

13

8,180

1.0

7.2

4.7

59.5

–0.4

Nepal

npl

29.3

74

18

2,529

2.5

72.5

17.8

62.8

–2.2

Netherlands

nld

16.8

605

858

35,896

1.6

1.6

2.1

45.3

–0.1

New Zealand

nzl

4.3

125

169

28,907

1.3

1.2

1.6

50.2

1.0

Nicaragua

nic

6.1

16

7

2,629

–1.4

22.5

5.2

45.6

4.5

Per capita, PPP, 1996 base (dollars)

PPP, 1996 base

2011

Niger

ner

15.7

21

6

1,368

–0.3

464.6

154.5

287.0

–1.4

Nigeria

nga

144.0

207

247

1,435

–0.4

152.9

98.4

622.0

–1.8

Norway

nor

4.7

198

479

42,182

1.9

5.5

10.0

88.3

1.0

Oman

omn

3.2

95

67

29,669

2.6

0.4

0.3

0.3

–1.8

Pakistan

pak

178.0

655

204

3,674

2.5

81.5

25.6

59.6

–2.5

Panama

pan

3.4

46

30

13,233

2.0

1.0

0.5

–8.3

–1.7

Papua New Guinea

png

6.3

28

11

4,449

0.2

2.8

1.0

43.6

0

Paraguay

pry

6.7

54

22

8,009

0.9

4,317.6

1,870.7

5.5

–1.4

Peru

per

30.1

292

168

9,699

1.2

2.9

1.8

32.5

0.9

Philippines

phl

99.7

614

216

6,153

1.0

45.8

16.3

8.2

–0.5

Poland

pol

38.4

748

532

19,453

2.3

2.8

1.8

–10.9

0.4

Portugal

prt

10.8

242

Romania

rom

22.1

206

Russia

rus

139.0

2,700

Saudi Arabia

sau

29.7

529

Senegal

sen

14.0

Sierra Leone

sle

6.7

Singapore

sgp

242

22,516

1.8

185

9,312

3.7

1,885

19,460

1.5

560

17,788

–1.5

36

15

2,566

12

2

1,724

4.7

263

266

55,504

142.2

141.5

28.9

0.8

2.9

2.0

52.0

–2.0

28.6

21.4

3.8

5.5

3.7

3.5

35.4

0.4

0.8

465.5

200.5

181.0

–1.5

0

4,316.4

770.4

76.0

–3.6

4.1

1.2

1.1

–11.4

–0.6

APPENDIX B 245

Slovakia

svk

5.4

141

97

26,036

1.8

21.3

14.5

–18.2

1.8

Slovenia

svn

2.0

59

52

29,334

1.9

169.9

139.9

–4.7

–0.4

South Africa

zaf

47.6

664

422

13,939

0.8

7.0

4.5

8.5

0.1

Spain

esp

40.6

1,189

1,536

29,307

2.0

117.9

145.7

42.2

0.4

Sri Lanka

lka

21.7

153

59

7,045

3.5

113.8

36.2

–14.8

–2.1

Sudan

sdn

44.1

140

63

3,160

2.3

276.6

152.0

159.0

–2.8

Swaziland

swz

1.3

10

4

7,905

0.5

7.3

2.4

–16.7

–0.5

Sweden

swe

9.1

349

572

38,380

1.7

6.2

9.1

53.1

–0.4

Switzerland

che

7.6

312

666

40,773

0.9

0.9

1.8

117.0

1.4

Syria

syr

22.3

140

65

6,294

1.3

46.4

23.1

51.9

–1.4

Taiwan

oan

23.7

862

505

36,338

5.0

29.0

n.a.

–41.0

–3.1

Tajikistan

tjk

7.6

26

7

3,355

–1.9

4.6

2.0

120.7

7.3

Tanzania

tza

41.8

44

23

1,062

0.8

1,465.1

639.6

558.0

–2.8

Thailand

tha

66.7

913

339

13,677

4.3

32.3

12.5

–35.2

–3.1

Timor-Leste

tls

1.2

7

1

5,740

1.4

1.0

n.a.

–5.0

0.9

Togo

tgo

6.4

8

4

1,213

–1.4

465.8

221.4

531

1.4

(continues on next page)

246 DEVALUING TO PROSPERITY

Table B.1

Country data, 2011 (continued) Annual income PPP, 1996 base (billions of dollars)

Current (billions of US dollars)

Per capita, PPP, 1996 base (dollars)

Annual growth per capita, 1980–2011 (percent)

Currency valuation (percent)

Exchange rate Local currency to US dollar

2011

Average log change per year, 1980–2011

Trinidad and Tobago

tto

1.1

30

22

27,408

2.2

6.3

3.9

–26.1

–1.7

Tunisia

tun

10.7

144

49

13,431

2.5

1.4

0.4

–44.0

–2.9

Turkey

tur

76.6

917

763

11,976

2.5

1,600,000.0

872,876.3

–4.1

–1.9

Turkmenistan

tkm

5.2

74

24

14,232

0.5

n.a.

3,088.8

n.a.

7.8

United Arab Emirates

uae

3.5

101

358

28,835

–1.5

3.7

4.8

53.4

1.2

United Kingdom

gbr

63.1

2,094

2,481

33,172

1.7

0.6

0.7

23.7

–0.4

United States

usa

313.0

15,351

15,065

48,988

1.7

1.0

1.0

–6.9

0.6

Uganda

uga

34.6

63

16

1,835

3.6

2,605.5

650.7

126.0

–12.5

Ukraine

ukr

45.1

495

163

10,971

–0.6

8.0

2.6

–34.6

7.5

Country

World Bank country code

Population (millions)

PPP, 1996 base

Uruguay

ury

3.5

68

49

19,213

2.0

18.5

11.9

–12.8

–1.4

Uzbekistan

uzb

27.9

220

44

7,886

1.3

1,750.0

340.7

–51.4

3.8

Venezuela

ven

27.7

283

310

10,206

–0.3

4,288.8

4,322.9

109.0

0.7

Vietnam

vnm

90.3

477

122

5,277

4.9

20,180.7

5,167.0

–10.9

–8.2

Yemen

yem

24.3

31

37

1,258

–0.8

220.1

247.8

1,463.0

3.4

Zambia

zmb

12.3

22

18

1,758

0.1

4,726.9

4,446.6

777.0

1.0

n.a. = not available; PPP = purchasing power parity Source: Author’s dataset, referred to as “Bhalla (2007a) dataset extended to 2011” in the text and tables.

References

Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. The Colonial Origins of Comparative Development: An Empirical Investigation. American Economic Review 91 (December): 1369–401. Acemoglu, Daron, Simon Johnson, James A. Robinson, and Yunyong Thaicharoen. 2003. Institutional Causes, Macroeconomic Systems: Volatility, Crises and Growth. Journal of Monetary Economics 50, no. 1: 49–123. Ahamed, Liaquat. 2009. Lords of Finance. London: William Heinemann. Aghion, Philippe, Philippe Bacchetta, Romain Ranciere, and Kenneth Rogoff. 2006. Exchange Rate Volatility and Productivity Growth: The Role of Financial Development. Swiss Finance Institute Research Paper Series 6-16 (May). Zurich: Swiss Finance Institute. Aguirre, Alvaro, and César Calderón. 2005. Real Exchange Rate Misalignments and Economic Performance. Central Bank of Chile Working Paper 315 (April). Santiago, Chile: Central Bank of Chile Albouy, David. 2006. The Colonial Origins of Comparative Development: An Investigation of the Settler Mortality Data. CIDER Paper C04-138. Center for International and Development Economics Research, University of California, Berkeley. Anderson, Jonathan. 2007. The Real Case for Revaluation. Asian Focus (July 13). UBS Investment Research. Arellano, Manuel, and Stephen Bond. 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Review of Economic Studies 58: 277–97. Baily, Martin Neil, and Robert Z. Lawrence. 2006. Competitiveness and the Assessment of Trade Performance. In C. Fred Bergsten and the World Economy, ed. Michael Mussa. Washington: Peterson Institute for International Economics. Balassa, Béla. 1964. The Purchasing Power Parity Doctrine: A Reappraisal. Journal of Political Economy 72, no. 6: 584–96. Balassa, Béla. 1978. Exports and Economic Growth: Further Evidence. Journal of Development Economics 5, no. 2: 181–89.

247

Balcerowicz, Leszek, and Stanley Fischer, eds. 2006. Living Standards and the Wealth of Nations: Success and Failures in Real Convergence. Boston, MA: MIT Press. Bargain, Olivier, Sumon K. Bhaumik, Manisha Chakrabarty, and Zhong Zhao. 2008. Earnings Differences between Chinese and Indian Wage Earners, 1987–2004. Discussion Paper 3284. Bonn: Institute for Study of Labor. Barro, Robert, and Jong Wha Lee. 2001. International Data on Educational Attainment: Updates and Implications. Oxford Economic Papers 53, no. 3: 541–63. Barro, Robert, and Jong Wha Lee. 2010. A New Data Set of Educational Attainment in the World. NBER Working Paper 15902. Cambridge, MA: National Bureau of Economic Research. Barro, Robert J., and Xavier Sala-i-Martín. 1992. Convergence. Journal of Political Economy 100, no. 2: 223–51. Bergin, Paul, Reuven Glick, and Alan M. Taylor. 2004. Productivity, Tradability, and the Great Divergence. University of California, Berkeley (March). Available at http://elsa.berkeley.edu (accessed on April 30, 2010). Bergsten, C. Fred. 2007. The Dollar and the Renminbi. Statement before the Hearing on US Economic Relations with China: Strategies and Options on Exchange Rates and Market Access, Subcommittee on Security and International Trade and Finance, Committee on Banking, Housing and Urban Affairs, United States Senate, May 23. Bernanke, Benjamin S. 2005. The Global Saving Glut and the U.S. Current Account Deficit. The Sandridge Lecture, Virginia Association of Economists, Richmond, Virginia, March 10. Bhalla, Surjit S. 1980. The Measurement of Permanent Income and Its Applications to Savings Behavior. Journal of Political Economy 88, no. 4 (August). Bhalla, Surjit S. 1992. ERM Mechanism: The Germans are Not at Fault. Goldman Sachs. Photocopy (September). Bhalla, Surjit S. 1994. Freedom and Economic Growth: A Virtuous Cycle? Paper presented at the Nobel Symposium on Democracy’s Victory and Crisis, Uppsala, Sweden, September. Bhalla, Surjit S. 1995. China and the Yuan: Medium Term Prospects. Developing Trends (April). New York: Deutsche Bank AG. Bhalla, Surjit S. 1997. Freedom and Economic Growth: A Virtuous Cycle? In Democracy’s Victory and Crisis, ed. Axel Hadenius. Cambridge, UK: Cambridge University Press. Bhalla, Surjit S. 1998a. Chinese Mercantilism: Currency Wars and How the East Was Lost. ICRIER Working Paper 45. New Delhi: Indian Council for Research on International Economic Relations. Bhalla, Surjit S. 1998b. Domestic Follies, Investment Crises: East Asia Lessons for India. In Asian Contagion: The Causes and Consequences of a Financial Crisis, ed. Karl Jackson. London: Perseus Book Group. Bhalla, Surjit S. 2002a. Imagine There’s No Country: Poverty, Inequality and Growth in the Era of Globalization. Washington: Institute for International Economics. Bhalla, Surjit S. 2002b. Trade, Growth and Poverty: Re-examining the Linkages. Paper presented at World Bank-ADB Fourth Asia Development Forum, Seoul, Korea, November. Bhalla, Surjit S. 2007a. Second Among Equals: The Middle Class Kingdoms of India and China. Manuscript (May). Oxus Research and Investments, New Delhi. Bhalla, Surjit S. 2007b. There Are No Growth Miracles. Paper presented at the International Monetary Fund, India Department, Washington, August. Bhalla, Surjit S. 2008a. Economic Development and the Role of Currency Undervaluation. Cato Journal 28, no. 2: 313–40. Bhalla, Surjit S. 2008b. Most Asians Dead in 1950. Business Standard, August 23.

248 DEVALUING TO PROSPERITY

Bhalla, Surjit S. 2010. Indian Economic Growth 1950-2008: Facts & Beliefs, Puzzles & Policies. In Essays in Honor of Montek Ahluwalia, eds. Shankar Acharya and Rakesh Mohan. New Delhi: Oxford University Press. Bhalla, Surjit S. 2011. Euro and the Yuan: Different Peas in the Same Pod. Comparative Economic Studies 53, no. 3 (September): 355–82. Bhalla, Surjit S. Forthcoming. The Old Order Is Fading: Challenges from the Middle Class Kingdoms of China and India. Oxus Research and Investments, New Delhi. Blanchard, Olivier, and Gian Maria Milesi-Ferretti. 2009. Global Imbalances: In Midstream? IMF Staff Position Note SPN/09/29 (December 22). Washington: International Monetary Fund. Bosworth, Barry. 2004. Valuing the RMB. Paper presented to Tokyo Club Research Meeting, February. Bosworth, Barry P., and Susan M. Collins. 2003. The Empirics of Growth: An Update. Brookings Papers on Economic Activity 2. Washington: Brookings Institution. Bourguignon, François, and Christian Morrisson. 2002. Inequality among World Citizens: 1820– 1992. American Economic Review (September). Bu, Yongxiang, and Rod Tyers. 2002. China’s Equilibrium Real Exchange Rate: A Counterfactual Analysis. Working Paper in Economics and Econometrics 390 (February). Canberra: Australian National University. Cassel, Gustav. 1921. The World’s Money Problems. New York: E. P. Dutton and Co. Chen, Shaohua, and Martin Ravallion. 2001. How Did the World’s Poorest Fare in the 1990s? World Bank Policy Research Working Paper 2409. Washington: World Bank. Chenery, Hollis, and Moises Syrquin. 1975. Patterns of Development, 1950–1970. New York: Oxford University Press. Cheung, Yin-Wong, Menzie D. Chinn, and Eiji Fujii. 2005. Why the Renminbi Might Be Overvalued (But Probably Isn’t). Proceedings (September). Federal Reserve Bank of San Francisco. Cheung, Yin-Wong, Menzie D. Chinn, and Eiji Fujii. 2007. The Overvaluation of Renminbi Undervaluation. Journal of International Money and Finance 26, no. 5: 762–85. Chinn, Menzie D., and Hiro Ito. 2008. A New Measure of Financial Openness. Journal of Comparative Policy Analysis: Research and Practice 10, no. 3: 309–22. Chomsky, Noam. 2001. 9-11: Was There an Alternative? Open Media Books. Clark, Gregory. 2007. A Farewell to Alms: A Brief Economic History of the World. Princeton, NJ: Princeton University Press. Clemens, Michael A., and Jeffrey G. Williamson. 2002. Why Did the Tariff-Growth Correlation Reverse after 1950? NBER Working Paper 9181. Cambridge, MA: National Bureau of Economic Research. Cline, William, and John Williamson. 2011. Estimates of Fundamental Equilibrium Exchange Rates, May 2011. Policy Briefs in International Economics 11-5. Washington: Peterson Institute for International Economics. Commission on Growth and Development. 2008. The Growth Report: Strategies for Sustained Growth and Inclusive Development. Washington: World Bank. Cottani, Joaquin A., Domingo Cavallo, and M. Shahbaz Khan. 1990. Real Exchange Rate Behavior and Economic Performance in LDCs. Economic Development and Cultural Change 39, no. 1: 61–76. Deaton, Angus, and Alan Heston. 2010. Understanding PPPs and PPP-based National Accounts. American Economic Journal: Macroeconomics 2, no. 4: 1–35. DeLong, Bradford, and Lawrence Summers. 1991. Equipment Investment and Economic Growth: How Robust Is the Nexus? Quarterly Journal of Economics 106, no. 2: 444–502.

REFERENCES 249

DeLong, Bradford, and Lawrence Summers. 1993. How Strongly Do Developing Countries Benefit from Equipment Investment? Journal of Monetary Economics 32, no. 3: 395–415. Dollar, David. 1992. Outward-Oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976–1985. Economic Development and Cultural Change 40, no. 3: 523–44. Dooley, Michael P., David Folkerts-Landau, and Peter Garber. 2003. An Essay on the Revived Bretton Woods System. NBER Working Paper 9971. Cambridge, MA: National Bureau of Economic Research. Dooley, Michael P., David Folkerts-Landau, and Peter Garber. 2005. International Financial Stability: Asia Interest Rates and Dollar (October). London: Deutsche Bank. Du, Quingyuan, and Shang-Jin Wei. 2011. Sex Ratios and Exchange Rates. NBER Working Paper 16788 (February). Cambridge, MA: National Bureau of Economic Research. Dunaway, Steven, and Xiangming Li. 2005. Estimating China’s “Equilibrium” Exchange Rate. IMF Working Paper 05/202. Washington: International Monetary Fund. Easterlin, Richard A. 1981. Why Isn’t the Whole World Developed? Journal of Economic History 41, no. 1 (March): 1–19. Easterly, William. 2001. The Lost Decade: Developing Countries’ Stagnation in Spite of Policy Reform 1980–1990. Journal of Economic Growth 6, no. 2: 135–37. Easterly, William. 2002. The Elusive Quest for Growth: Economists’ Adventures and Misadventures in the Tropics. Cambridge, MA: MIT Press. Easterly, William. 2005. National Policies and Economic Growth: A Reappraisal. In Handbook of Economic Growth, eds. P. Aghion and S. Durlauf. Amsterdam: Elsevier. Easterly, William, Michael Kremer, Lant Pritchett, and Lawrence Summers. 1993. Good Policy or Good Luck? Country Growth Performance and Temporary Shocks. Journal of Monetary Economics 32, no. 3: 459–83. Easterly, William, and Ross Levine. 2002. Tropics, Germs, and Crops: How Endowments Influence Economic Development. NBER Working Paper 9106. Cambridge, MA: National Bureau of Economic Research. Edwards, Sebastian. 1989. Real Exchange Rates, Devaluation, and Adjustment: Exchange Rate Policy in Developing Countries. Cambridge, MA: MIT Press. Edwards, Sebastian. 1992. Trade Orientation, Distortions, and Growth in Developing Countries. Journal of Development Economics 39, no. 1 (July): 31–57. Eichengreen, Barry. 2004. Chinese Currency Controversies (April). Berkeley: University of California. Eichengreen, Barry. 2007. China’s Exchange Rate Regime: The Long and Short of It (March). Berkeley: University of California Eichengreen, Barry, Donghyun Park, and Kwanho Shin. 2011. When Fast Growing Economies Slow Down: International Evidence and Implications for China. NBER Working Paper 16919. Cambridge, MA: National Bureau of Economic Research. Engel, Charles. 1999. Accounting for U.S. Real Exchange Rate Changes. Journal of Political Economy 107 (June): 507–38. Frankel, Jeffrey A. 2006. On the Yuan: The Choice between Adjustment under a Fixed Exchange Rate and Adjustment under a Flexible Rate. CESifo Economic Studies 52, no. 2: 246–75. Frankel, Jeffrey A., and Andrew Rose. 1995. A Panel Project on Purchasing Power Parity. NBER Working Paper 5006. Cambridge, MA: National Bureau of Economic Research. Fraser Institute. 2011. Economic Freedom of the World. Available at www.fraserinstitute.org. Freedom House. 2011. Freedom in the World 2011. Available at www.freedomhouse.org. Friedman, Benjamin M. 2005. The Moral Consequences of Economic Growth. New York: Alfred A. Knopf.

250 DEVALUING TO PROSPERITY

Friedman, Milton. 1962. Capitalism and Freedom. Chicago: University of Chicago Press. Funke, Michael, and Jorg Rahn. 2005. Just How Undervalued Is the Chinese Renminbi? World Economy 28, no. 4: 465–89. Gerschenkron, Alexander. 1962. Economic Backwardness in Historical Perspective. Cambridge, MA: Harvard University Press. Gill, Indermit, and Homi Kharas. 2007. An East-Asian Renaissance: Ideas for Economic Growth. Washington: World Bank. Goldfajn, Ilan, and Rodrigo O. Valdés. 1999. The Aftermath of Appreciations. Quarterly Journal of Economics 114, no. 1: 229–62. Goldstein, Morris. 2006. The IMF as Global Umpire for Exchange Rate Policies. In C. Fred Bergsten and the World Economy, ed. Michael Mussa. Washington: Peterson Institute for International Economics. Goldstein, Morris. 2007. A (Lack of) Progress Report on China’s Exchange Rate Policies. Working Paper 07-5. Washington: Peterson Institute for International Economics. Hall, Robert E., and Charles I. Jones. 1999. Why Do Some Countries Produce So Much More Output per Worker than Others? Quarterly Journal of Economics 114, no. 1: 83–116. Harrison, Ann. 1996. Openness and Growth: A Time-Series, Cross-Country Analysis for Developing Countries. Journal of Development Economics 48: 419–47. Hause, John C. 1971. Ability and Schooling as Determinants of Lifetime Earnings, or If You’re So Smart, Why Aren’t You Rich? American Economic Review 61, no. 2: 289–98. Hausmann, Ricardo, Lant Pritchett, and Dani Rodrik. 2005. Growth Accelerations. Journal of Economic Growth 10, no. 4: 303–29. Hayek, Friedrich von. 1944. The Road to Serfdom. Chicago: University of Chicago Press. Helpman, Elhanan. 2005. The Mystery of Economic Growth. New Delhi: Academic Foundation. Heritage Foundation. Various years. Index of Economic Freedom. Available at www.heritage.org. Heston, Alan, Daniel Nuxoll, and Robert Summers. 1994. The Differential Productivity Hypothesis and Purchasing-Power Parities: Some New Evidence. Review of International Economics 2, no. 3: 227–43. Heston, Alan, Robert Summers, and Bettina Aten. 2002. Penn World Table Version 6.1. Philadelphia: Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania. Heston, Alan, Robert Summers, and Bettina Aten. 2006. Penn World Table Version 6.2. Philadelphia: Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania. Hinkle, Lawrence E., and Peter Montiel. 2001. Exchange Rate Misalignment: Concepts and Measurement for Developing Countries. New York: Oxford University Press. Hong, Liang. 2002. Chinese Renminbi: Whose Problem Is It? Goldman Sachs Economics Flash. New York. IMF (International Monetary Fund). 2003. World Economic Outlook: Growth and Institutions (April). Washington. IMF (International Monetary Fund). 2005. World Economic Outlook: Globalization and External Balances (April). Washington. IMF (International Monetary Fund). 2006. World Economic Outlook: Financial Systems and Economic Cycles (September). Washington. IMF (International Monetary Fund). 2011. World Economic Outlook: Slowing Growth, Rising Risks (September). Washington.

REFERENCES 251

Irwin, Douglas A. 2002. Interpreting the Tariff-Growth Correlation of the Late 19th Century. American Economic Review 92, no. 2: 165–69. Jen, Stephen J. 2003. Three Objective Perspectives on the RMB Debate (July 24). New York: Morgan Stanley Equity Research. Johnson, Simon, Jonathan Ostry, and Arvind Subramanian. 2007. The Prospects for Sustained Growth in Africa: Benchmarking the Constraints. IMF Working Paper 07/52. Washington: International Monetary Fund. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2005. Measuring Governance Using CrossCountry Perceptions Data (August). Washington: World Bank. Knack, Stephen, and Philip Keefer. 1995. Institutions and Economic Performance: Cross-Country Tests Using Alternative Institutional Measures. Economics and Politics, no. 7: 207–27. Kohli, Atul. 2006. Politics of Economic Growth in India, 1980–2005. Economic and Political Weekly (April 1). Kravis, Irving B., and Robert E. Lipsey. 1986. The Assessment of National Price Levels. NBER Working Paper 1912. Cambridge, MA: National Bureau of Economic Research. Kravis, Irving B., Alan W. Heston, and Robert Summers. 1988. Real GDP Per Capita for More than One Hundred Countries. Economic Journal 88, no. 350: 215–42. Krueger, Anne. 1990. Asian Trade and Growth Lessons. American Economic Review 80: 108–111. Krugman, Paul. 1994. The Myth of Asia’s Miracle. Foreign Affairs (November/December). Krugman, Paul. 2010. Taking on China. New York Times (global edition), March 14. Kuznets, Simon. 1955. Economic Growth and Income Inequality. American Economic Review 45 (March): 1–28. Landes, David. 1998. The Wealth and Poverty of Nations: Why Some Are So Rich and Some So Poor. New York: W. W. Norton. La Porta, Rafael, Florencio López-de-Silanes, Andrei Shleifer, and Robert Vishny. 1998. The Quality of Government. Cambridge, MA: Harvard University. Photocopy (August). Lawrence, Robert Z. 1984. Can America Compete? Washington: Brookings Institution. Lawrence, Robert Z. 1987. Imports in Japan: Closed Markets or Minds? Brookings Papers on Economic Activity 2. Washington: Brookings Institution. Lawrence, Robert Z. 2007. US Trade Policy: The Exception to American Exceptionalism? In Power and Superpower: Global Leadership in the 21st Century (June). Washington: Center for American Progress and the Century Foundation. Lawrence, Robert, Z. 2009. The US Economy and Its Trade Patterns. Paper presented at Athens Seminar Series, Athens, Greece, July. Lewis, Arthur W. 1955. The Theory of Economic Growth. London: George Allen & Unwin. Levy-Yeyati, Eduardo, and Frederico Sturzenegger. 2007. Fear of Appreciation. Policy Research Working Paper Series 4387. Washington: World Bank. Lucas, Robert E., Jr. 1990. Why Doesn’t Capital Flow from Rich to Poor Countries? American Economic Review 80, no. 2: 92–96. Maddison, Angus. 2001. The World Economy: A Millennial Perspective (June). Paris: Organization for Economic Cooperation and Development. Maddison, Angus. 2003. The World Economy: Historical Statistics. Paris: Organization for Economic Cooperation and Development McKinnon, Ronald, and Gunther Schnabl. 2006. China’s Exchange Rate and International Adjustment in Wages, Prices, and Interest Rates: Japan Déjà vu? CESInfo Economic Study 52, no. 2: 276–303.

252 DEVALUING TO PROSPERITY

Montiel, Peter J., and Servén, Luis. 2008. Real Exchange Rates, Saving and Growth: Is There a Link? Policy Research Working Paper 4636. Washington: World Bank, Development Research Group, Macroeconomics and Growth Team. Morrisson, Christian, and Fabrice Murtin. 2003. History and Prospects of Inequality among Europeans. Paris School of Economics. Photocopy. Myrdal, Gunnar. 1968. Asian Drama: An Inquiry into the Poverty of Nations, Volumes 1–3. Penguin Books. Nehru, Vikram, and Ashok Dhareshwar. 1993. A New Database on Physical Capital Stock: Sources, Methodology and Results. Revista de Análisis Económicá 8, no. 1 (June): 37–59. Obstfeld, Maurice, and Kenneth Rogoff. 2000. The Six Major Puzzles in International Macroeconomics: Is There a Common Cause? In NBER Macroeconomics Annual 2000, eds. Ben Bernanke and Kenneth Rogoff. Cambridge, MA: MIT Press. O’Rourke, Kevin. 2000. Tariffs and Growth in the Late 19th Century. Economic Journal 110, no. 463: 456–83. Parker, Philip M. 1997. National Cultures of the World: A Statistical Reference. Cross-Cultural Statistical Encyclopedia of the World 4. Westport, CT: Greenwood Press. PRS (Political Risk Services) Group. Various years. International Country Risk Guide. East Syracuse, NY. Polity IV Project. n. d. Political Regime Characteristics and Transitions, 1800–2008. Available at www. systemicpeace.org/polity/polity4.htm. Prasad, Eswar S. 2007. Is the Chinese Growth Miracle Built to Last? China Economic Review 20, no. 1: 103–23. Prasad, Eswar, Raghuram Rajan, and Arvind Subramanian. 2007. Foreign Capital and Economic Growth. Brookings Papers in Economic Activity 1. Washington: Brookings Institution. Pritchett, Lant. 2001. Where Has All the Education Gone? World Bank Economic Review 15, no. 3: 367–91. Razin, Ofair, and Susan M. Collins. 1997. Real Exchange Rate Misalignments and Growth. In The Economics of Globalisation: Policy Perspectives from Public Economics, eds. A. Razin and E. Sadka. Cambridge, UK: Cambridge University Press. Rigobon, Roberto, and Dani Rodrik. 2004. Rule of Law, Democracy, Openness, and Income: Estimating the Interrelationships. NBER Working Paper 10750. Cambridge, MA: National Bureau of Economic Research. Roach, Stephen. 2004. The Scapegoating of China. Global Economic Forum (July 14). Morgan Stanley. Available at www.morganstanley.com (accessed on February 14, 2010). Robinson, Sherman. 1976. A Note on the U Hypothesis Relating Income Inequality and Economic Development. American Economic Review 66, no. 3 (June): 437–40. Rodrik, Dani. 2003a. Growth Strategies. NBER Working Paper 10050. Cambridge, MA: National Bureau of Economic Research. Rodrik, Dani, ed. 2003b. In Search of Prosperity: Analytic Narratives on Economic Growth. Princeton, NJ: Princeton University Press. Rodrik, Dani. 2006. Goodbye Washington Consensus, Hello Washington Confusion? A Review of the World Bank’s Economic Growth in the 1990s: Learning from a Decade of Reforms. Journal of Economic Literature XLIV (December): 973–87. Rodrik, Dani. 2007a. The Real Exchange Rate and Economic Growth: Theory and Evidence. Harvard University Working Paper (July). Cambridge, MA: Harvard University. Rodrik, Dani. 2007b. One Economics, Many Recipes. Princeton, NJ: Princeton University Press. Rodrik, Dani. 2008. The Real Exchange Rate and Economic Growth. Brookings Papers on Economic Activity 2. Washington: Brookings Institution.

REFERENCES 253

Rodrik, Dani, and Arvind Subramanian. 2004. From “Hindu Growth” to Productivity Surge: The Mystery of the Indian Growth Transition. IMF Working Paper 04/77. Washington: International Monetary Fund. Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi. 2002. Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Development. NBER Working Paper 9305. Cambridge, MA: National Bureau of Economic Research. Rogoff, Kenneth. 1996. The Purchasing Power Parity Puzzle. Journal of Economic Literature XXXIV (June): 647–68. Romer, Paul. 1986. Increasing Returns and Long-Run Growth. Journal of Political Economy 94: 1002–37. Rose, Andrew K., and Saktiandi Supaat. 2007. Fertility and the Real Exchange Rate. NBER Working Paper 13263. Cambridge, MA: National Bureau of Economic Research. Rostow, Walt Whitman. 1960. The  Stages  of Economic  Growth: A Non-Communist Manifesto. Cambridge: Cambridge University Press. Sachs, Jeffrey D. 1981. The Current Account and Macroeconomic Adjustment in the 1970s. Brookings Papers on Economic Activity 1: 201–82. Washington: Brookings Institution. Sachs, Jeffrey D. 2003. Institutions Don’t Rule: Direct Effects of Geography on Per Capita Income. NBER Working Paper 9490. Cambridge, MA: National Bureau of Economic Research. Sachs, Jeffrey D., and Andrew Warner. 1995. Economic Reform and the Process of Global Integration. Brookings Papers on Economic Activity 1: 1–118. Washington: Brookings Institution. Sala-i-Martín, Xavier. 1997a. I Just Ran Two Million Regressions. American Economic Review 87, no. 2: 178–83. Sala-i-Martín, Xavier. 1997b. I Just Ran Four Million Regressions. NBER Working Paper 6252. Cambridge, MA: National Bureau of Economic Research. Samuelson, Paul A. 1964. Theoretical Notes on Trade Problems. Review of Economics and Statistics 46, no. 2: 145–54. Scully, Gerald W. 1988. The Institutional Framework and Economic Development. Journal of Political Economy 96, no. 3: 652–62. Smith, Adam. 1776. The Wealth of Nations. London: Penguin Books. Solow, Robert M. 1970. Growth Theory: An Exposition. Oxford, UK: Clarendon Press. Swan, T. W. 1956. Economic Growth and Capital Accumulation. Economic Record 32: 334–61. Taleb, Nassim Nicholas. 2010. The Black Swan, 2d ed. New York: Penguin. Taylor, Alan M. 1996. International Capital Mobility in History: Purchasing-Power Parity in the Long Run. NBER Working Paper 5742. Cambridge, MA: National Bureau of Economic Research. Taylor, Alan M., and Mark P. Taylor. 2004. The Purchasing Power Parity Debate. NBER Working Paper 10607. Cambridge, MA: National Bureau of Economic Research. Thomas, Charles P., Jamie Marquez, and Sean Fahle. 2008. Measuring U.S. International Relative Prices: A WARP View of the World. International Finance Discussion Papers 917 (January). Washington: Board of Governors of the Federal Reserve System. US Treasury. 2005. Appendix: Analysis of Exchange Rate Policies Pursuant to the Act in Report to Congress on International Economic and Exchange Rate Policies. Washington. Vandenbussche, Jerome, Philippe Aghion, and Costas Meghir. 2004. Growth, Distance to Frontier and Composition of Human Capital. IFS Working Paper 04/31. London: Institute for Fiscal Studies. Wacziarg, Romain, and Karen Horn Welch. 2003. Trade Liberalization and Growth: New Evidence. NBER Working Paper 10152. Cambridge, MA: National Bureau of Economic Research. Walter, Norbert. 2003. Is the Chinese Currency, the Renminbi, Dangerously Undervalued and a Threat to the Global Economy? The International Economy (Spring).

254 DEVALUING TO PROSPERITY

Williamson, John. 1990. Latin American Adjustment: How Much Has Happened? Washington: Institute for International Economics. Williamson, John, ed. 1994. Estimating Equilibrium Exchange Rates. Washington: Institute of International Economics. Williamson, John. 2003. Exchange Rate Policy and Development. Washington: Peterson Institute for International Economics. Photocopy. Williamson, John. 2008. Exchange Rate Economics. Working Paper 2. Washington: World Bank on behalf of the Commission on Growth and Development. Wolf, Martin. 2008. Fixing Global Finance. Baltimore: Johns Hopkins University Press. Wolf, Martin. 2009. Why China’s Exchange Rate Policy Concerns Us. Financial Times, March. Wolf, Martin. 2010. The Grasshopper and the Ant—A Modern Fable. Financial Times, May 25. Woodford, Michael. 2009. Is an Undervalued Currency the Key to Economic Growth? Comment on Dani Rodrik’s The Real Exchange Rate and Economic Growth in Brookings Papers on Economic Activity (2008). Discussion Paper 0809-13. New York: Columbia University Department of Economics. World Bank. 1991. World Development Report: The Challenge of Development. New York: Oxford University Press. World Bank. 1993. The East Asian Miracle: Economic Growth and Public Policy. World Bank Policy Research Report. Washington. World Bank. 2005. Economic Growth in the 1990s: Learning from a Decade of Reform. Washington. World Bank. 2010. World Development Indicators. Washington. Yang, Dennis Tao, Vivian Chen, and Ryan Monarch. 2010. Rising Wages: Has China Lost Its Global Labor Advantage? Pacific Economic Review 15, no. 4: 482–504. Young, Alwyn. 1995. The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience. Quarterly Journal of Economics 110, no. 3: 641–80. Zhang, Xiaopu. 2002. Equilibrium and Misalignment: An Assessment of the RMB Exchange Rate from 1978 to 1999. Working Paper 127 (February). Stanford, CA: Center for Research on Economic Development and Policy Reform, Stanford University.

REFERENCES 255

Index

Africa, growth in geography and, 15–17, 16t miracle, 164 agriculture, labor reallocation and, 19–21, 20f Asian financial crisis (1997–98) causes of, 2, 5, 210, 211 endogenous RER, 84 mercantilism and, 150 savings and, 35 “standing still” since, 227 Association of Southeast Asian Nations (ASEAN-7), 153n. See also specific country growth rates, 129 mercantilism rankings, 153 Australia, 63, 164, 170, 181, 183 Balassa-Samuelson effect, 44–46 China-Japan comparison, 191 confirmation of, 49, 186 correction for, 52, 81 currency valuation using, 46–48, 47t investment and, 70 RER-income calculations, 56–57 “standing still” effect, 227 Bank for International Settlements (BIS), 52, 85 beggar-thy-neighbor policy, 209–10, 219–21, 222t benchmark growth models, 117 Bernanke, Ben, 35 bias, 58–59, 65, 135 Big Mac Index, 59–61, 60t

Chinese renminbi valuations, 203 US dollar valuations, 96–97, 97t, 98t black-market premium, 65–66 black swan growth. See miracle growth borrowing costs, 212, 216, 217f Botswana, 164 Brazil, 59, 63, 140 Bretton Woods system, 29 BRICS, 197n Broad Index, 95–102, 97t–99t, 100f–103f, 115 Canada, 170 Can America Compete? (Lawrence), 189 capital cost of, 68 data on, 235 as factor of production, 18–19, 134, 155t, 155–56 capital controls, 6–7, 78–84 catch-up, 25–26, 136 Chile, 72, 79 China capital controls, 79 currency valuations, 59, 63, 86, 90, 112–13 current account, 37–39, 38t, 191, 210 devaluation strategy, 2, 8, 227 changes in, 221–23, 229 as currency manipulator, 90–91, 211, 219–21, 225 debates concerning, 197–204, 198t–202t GDP, 189–90

257

Great Recession (2008) and, 204–205 growth breaks in, 140 catch-up, 136 miracle, 161, 219 historical context, 12, 181–83, 182t investment in, 68 versus Japan (1980s) (See China-Japan comparison) mercantilism, 150, 228 reserve accumulation, 214–16, 215t saving rate, 35 wages, 8–9, 211–14, 213t China-Japan comparison, 90, 189–207 conclusions, 196–205 historical context, 189–91 measures, 191–96, 193t currency misalignment, 192–93, 194f, 195f current account surplus, 194–95, 197 GDP and export growth, 194 mercantilism, 196 reserve accumulation, 195–96 colonial heritage, 12, 169–71, 183 data and methods, 236–37 Commission on Growth and Development, 138–40 comparative advantage, 21, 77 Confucianism, 17 Confucius, 179 consumer price index (CPI), 44, 46, 81n, 95–96, 181n, 233 cross-section panel data, 123, 124t culture, 17 currency adjustments, requirements, 112–13 currency manipulator, China as, 90–91, 211, 219–21, 225 currency misalignments, 48–52, 50t, 51t China-Japan comparison, 192–93, 194f, 195f postcrisis realignment, 228–29 currency overvaluation Chinese renminbi, 90 defined, 234 growth effects, 4–6, 116, 180, 226–27 currency undervaluation as beggar-thy-neighbor policy, 209–10, 219–21, 222t breaking cycle of, 221–23 changing view of, 209–10, 216–19 Chinese renminbi (See China) current account and, 39–40, 40t defined, 33n, 41n, 234 versus export-led growth, 23, 226

258 DEVALUING TO PROSPERITY

global use of, 8 growth effects, 4–5, 9, 23–24, 116, 209, 225–26 malfunctions, 212–16 mechanisms, 6, 211–12 historical context, 179–87 tariffs, 183–85, 184t–186t 19th century exchange rates, 180–83, 182t, 228 yen exchange rate (1950), 186–87, 187t inflation and, 79–80 investment impact of, 69–71, 211–12, 225–26 mercantilism rankings by, 150–53, 151t–152t passive, 85–86 trade-oriented use of, 7, 23–24 world GDP, 218–19, 220f currency valuation. See also specific country Balassa-Samuelson effect, 46–48, 47t current account and (See current account) growth effects, 7–8, 115–16 econometric models (See econometric growth models) historical context, 8, 86–90, 87t–89t investment and (See investment) measurement of, 5–6, 41–66 overview, 64–66 currency misalignments, 48–52 equilibrium exchange rates, 41–42 errors in, 58–59, 65, 135 real exchange rates, 42–48, 52–58 smell tests, 59–63 overvaluation (See currency overvaluation) savings and, 5, 34–35, 36t trade-oriented use of, 7 undervaluation (See currency undervaluation) currency wars, 1, 211, 217–19 current account. See also specific country China-Japan comparison, 194–95, 197 currency undervaluation and, 39–40, 40t global imbalances, 7, 93–113 growth effects, 5–6, 37–39, 38t, 209 mercantilism and, 149–50 policy, 24–25 savings, 5, 34–35, 36t, 94 world GDP, 218, 219t data and methods, 233–37 dataset, 240t–246t institutions, 236 instruments, 237, 237t population, 234

real exchange rate, 234–36 real income, 233–34 demography, 26, 234 diminishing returns, 132–34 Dollar-Easterly measure, 49–52, 59–64, 135 currency valuations, 61–63, 62f–64f growth patterns, 143 institutions, 168, 172–74 Dutch disease, 79, 79n dynamic panel estimation, 118, 123, 124t East Asian miracle growth, 136, 161 econometric growth models, 117–18 cross-section panel data, 123, 124t diminishing returns, 132–34 dynamic panel estimation tests, 125, 126t other determinants, 125–28, 127t outliers, 128, 129t, 130f–132f overall results, 120–23, 122t real income, 119–20, 121t regional trends, 128–29, 133t structural breaks, 139–43, 142t, 144t time framework, 119 econometrics investment equations, 70–71 measurement bias in, 58 economic freedom, 65–66, 169, 236 economic growth. See growth economic inequality, 17–18 economic openness, 22, 65–66, 137 education, 18–19, 25 externality effect of, 19n wages and, 212–14, 213t elasticity income, 57f, 57–58 trade, 109 employment, full, 77–78 endogenous real exchange rate, 9, 77–91, 227 Impossible Trinity, 6–7, 78–84 theory on, 78 equilibrium exchange rates (EERs) defined, 234 measurement of, 6, 41–42, 49 errors, measurement currency valuation, 58–59, 65, 135 Impossible Trinity, 81–84 ethnic fragmentation, 170, 171 European debt crisis (2010–11), 1, 93, 102, 210 European Union. See also specific country currency valuations, 89n, 103–104, 105t, 115 current account, 93, 102–106, 107f, 107t exchange rate(s) equilibrium, 6, 41–42

fixed, 8, 77–78 fundamental equilibrium, 52n, 59, 95n, 108, 203 19th century, 180–83, 182t, 228 nominal, 9, 23–24, 77, 227, 233 policy, 4–5, 23–24 real (See real exchange rates) executive constraints, 236 export-led growth, 4, 23, 194, 226 export-push strategy, 137 externalities education, 19n investment, 72 factor accumulation, 18–19 factor reallocation, 19–21, 20f financial crises Asian (See Asian financial crisis (1997–98)) global (2008) (See Great Recession (2008)) Finland, 161 fiscal policy, 24–25 fixed effects model, savings, 35, 36t fixed exchange rate, 8, 77–78 foreign direct investment. See investment foreign exchange reserves. See reserve accumulation France, 181 Fraser Institute, 169, 174 freedom economic, 65–66, 169, 236 political, 168–69, 236 Freedom House, 168–69, 171, 171n, 236 full employment, 77–78 fundamental equilibrium exchange rates (FEERs), 52n, 59, 95n, 108, 203 Gastil index of political freedom, 168–69, 171n GDP China, 189–90 export growth and, 194 factors of production, 155t, 155–56 income per capita, 158, 159f, 160f miracle growth and, 156, 157t share of investment in, 72, 73f–74f share of trade in, 22 US, 99t, 99–100, 189–90 world currency undervaluation, 218–19, 220f current account balances, 218, 219t GDP deflator, 44, 81n, 96, 181n, 233 geography, 2–4, 228 data, 236 theories about, 15–17, 16t, 165 Germany, 37–39, 38t, 59, 72, 90

INDEX 259

global financial crisis. See Great Recession (2008) Goldman Sachs BRICS report, 197n gold standard, 180, 180n good fortune, 28t, 28–29 governance, 65–66, 168–69, 236 government intervention. See policy Great Recession (2008) causes of, 2, 93 Chinese renminbi before, 191 currency valuations before, 217–18, 218t growth before, 1–2 policy and, 8 postcrisis realignment, 228–29 transformational effects, 1, 8, 204–205, 212, 221–23 Greece, 89, 113 growth. See also specific country acceleration, 136, 141–47, 146f, 147f catch-up, 25–26, 136 currency valuation and, 7–8, 115–16 (See also currency valuation) econometric models (See econometric growth models) determinants of, 2–4, 11–31 (See also specific determinant) theories on, 15–29 failure of, 136–39 before Great Recession (2008), 1–2 historical context, 11, 13–15, 14t, 16t, 136 mercantilism and, 153, 154t miracle (See miracle growth) principles of, 5, 13, 18, 25 S shape of, 53–55, 55f structural breaks in, 139–43, 142t, 144t theories on, 5, 12–13 Heritage Foundation, 169, 171, 171n high-performance Asian economies (HPAE), 137 historical context currency valuations, 8, 86–90, 87t–89t undervaluation, 179–87 world, 217–18, 218t growth, 11, 13–15, 14t, 16t, 136 Japan (1980s), 90, 189–96 tariffs, 183–85, 184t–186t 19th century exchange rates, 180–83, 182t, 228 yen exchange rate (1950), 186–87, 187t Hong Kong currency valuation, 63, 113 investment rate, 72 mercantilism, 150, 228

260 DEVALUING TO PROSPERITY

miracle growth, 161 human capital, 18–19 identification variable, 170 impact coefficient, 15 Impossible Trinity, 6–7, 78–84 income, 52–58, 64–66 accurate measurements, 56–58, 57f alternative models, 56 conventional patterns, 53–55, 55f data and methods, 233–35 econometric growth models, 119–20, 121t income per capita historical context, 181–83, 182t institutions and, 171 Japan (1980s), 193, 194f miracle growth and, 158, 159f, 160f wages and, 212–14, 213t India capital controls, 79 colonial heritage, 170, 183 currency valuation, 59, 63, 203 current account, 37–39, 38t endogenous RER, 84 fiscal policy, 24n growth in geography and, 17 income patterns, 54 labor reallocation theory, 20f, 20–21 historical context, 12, 181–83, 182t investment in, 68 poverty levels, 59 PPP estimates, 190 reserve accumulation, 214–16, 215t wages, 8–9, 211–14, 213t Indonesia, 80n Industrial Revolution, labor reallocation during, 19–21, 20f inequality, 17–18 inflation currency devaluation and, 79–80, 227 exchange rate and, 6–7, 78–84 price measures, 44, 46, 81n, 95–96, 181n, 233 institutions, 2–4, 165–78 conventional wisdom on, 26, 166–67 data and methods, 236 growth effects, 8, 26, 171–78, 172t, 175t–178t, 228 instruments for measuring, 170–71 measures, 168–71 miracle growth and, 137 new evidence on, 168

instrument variables, 21, 118 interest rates, 212, 216, 217f International Comparison Program (ICP), 43–44, 48n, 119–20, 121t, 190 International Country Risk Guide (ICRG), 169, 171, 174, 236 International Monetary Fund (IMF), 29, 52 fiscal deficit data, 25 on institutions, 166–67, 172 PPP estimates, 190 REER data, 85 World Economic Outlook, 234 investment, 6, 67–76 as channel of influence, 72–73, 73f–76f, 75t currency undervaluation and, 69–71, 211–12, 225–26 determinants of, 67–68 Ireland, 161 Israel, 72, 161 Japan currency valuation, 59, 63 current account, 37–39, 38t investment rate, 72 miracle growth, 161 post-war fixed exchange rate, 8 in 1980s (See China-Japan comparison) yen exchange rate (1950), 186–87, 187t Keynes, John Maynard, 3–4, 29, 165 knowledge, 18–19, 25 Korea, 37–39, 38t, 59, 161 Krueger, Anne, 137, 180 Krugman, Paul, 161, 203 Kuznets, Simon, 17 Kuznets Curve, 18 labor, 18–19 cost of, 68, 69 data, 235 as factor of production, 155t, 155–56 reallocation of, 19–21, 20f labor force participation rate (LFPR), 26 labor productivity, 69 law of one price, 41–42 legal origin, 170–71 Lehman Brothers, 5 Lewis, Sir Arthur, 5, 12, 13, 18, 19, 25 MacArthur, Douglas, 8, 180, 186 Malaysia, 72, 79, 113, 150, 228 Malthus, Thomas, 27 Marx, Karl, 27 mean, RER reversion to, 80–81, 82f–83f measurement errors

currency valuation, 58–59, 65, 135 Impossible Trinity, 81–84 mercantilism, 8, 149–64, 228 China-Japan comparison, 196 defined, 149–50 growth and, 153, 154t miracle economies, 153–64 rankings, 150–53, 151t–152t Mexico, 84 middle class, 27–28, 235–37 middle income trap, 136, 140–41, 143, 145t Mills, John Stuart, 27 miracle growth, 8, 136, 153–64 defined, 155, 156, 161 factors of production in, 155t, 155–56, 160–61 GDP growth per capita, 156, 157t identification of, 158–59 mercantilism and, 153 reasons for, 137–38, 219 monetary policy, 6–7, 78–84 Netherlands, 153, 181, 228 New Zealand, 170 nominal exchange rate, 9, 23–24, 77, 227, 233 Occam’s razor, 119, 119n openness, 22, 65–66, 137 ordinary least squares (OLS) method, 35, 36t, 118 Organization for Economic Cooperation and Development (OECD), 190 outliers, 128, 129t, 130f–132f passive devaluation, 85–86 Penn World Tables, 48n, 96, 119–20, 181, 233 Philippines, 63, 72, 129n Plaza Agreement (1985), 189, 191, 196, 204 policy exchange rate, 4–5, 23–24 fiscal, 24–25 growth effects, 2–4, 165, 168, 227 institutions and, 174–76 monetary, 6–7, 78–84 trade, 4, 21–22 Washington Consensus, 4, 7, 24, 29–31, 30t, 203 political economy, 84 political freedom, 168–69, 236 Polity IV dataset, 168–71, 236 population data, 234 population density, 170–71, 237 poverty, 27, 58–59 price measures, 44, 46, 81n, 95–96, 181n, 233 price ratio, real exchange rates as, 45–46

INDEX 261

productivity, 134, 143, 226 investment and, 211 labor, 69 total factor productivity growth, 155t, 155–56, 160–61, 162t–163t, 235 profitability, 67–68, 226 property rights, 236 Protestant work ethic, 17 purchasing power parity (PPP) Balassa-Samuelson effect, 45 data, 233 econometric growth models, 119–20 equilibrium exchange rate in terms of, 41–42 estimates, 190 historical context, 180–81 investment and, 69–70 US dollar valuation, 96 real effective exchange rate (REER), 52, 85 real exchange rates (RERs) currency valuations, 63, 64f data and methods, 233–34 defined, 45, 46, 234 endogenous, 9, 77–91, 227 Impossible Trinity, 6–7, 78–84 theory on, 78 income and, 52–58 investment and, 69–70 measurement of, 6, 42–49, 50t, 51t, 85 policy effects on, 23–24 as price ratio, 45–46 “standing still,” 85–86, 87t–89t, 90, 191, 210, 227 real income, 233–34 regional trends currency valuations, 86–90, 89t, 221, 222t econometric growth models, 128–29, 133t religion, 17 reserve accumulation, 149–50 China-Japan comparison, 195–96 currency undervaluation and, 214–16, 215t Roosevelt, Franklin D., 179 Sachs, Jeffrey, 22, 192, 236 Samuelson, Paul 1964 article, 8, 191n, 205–207 Balassa-Samuelson effect (See BalassaSamuelson effect) savings, 5, 34–35, 36t, 94 scatter plots of growth acceleration, 146, 146f, 147f settler mortality rate, 170–71, 176, 237 silver standard, 180, 180n Singapore, 72, 113, 150, 161, 228

262 DEVALUING TO PROSPERITY

smell tests, 7 current account balances, 93–113 divergent estimates, 61–63, 62f–64f methods, 59–61, 60t, 61t Smith, Adam, 12 Solow, Robert, 13 S shape of growth, 53–55, 55f “standing still” effect, 85–86, 87t–89t, 90, 191, 210, 227 structural breaks in growth, 139–43, 142t, 144t subprime mortgage crisis (US), 2 Summers, Lawrence, 28, 67 Taiwan, 63, 72, 113, 150, 161, 228 tariffs, 183–85, 184t–186t, 235 technology, 2–4 labor reallocation and, 19–21, 20f Thailand, 72, 79, 84, 150, 228 The Theory of Economic Growth (Lewis), 12, 13 total factor productivity growth (TFPG), 155t, 155–56, 160–61, 162t–163t, 235 trade balance, 109–12, 110t–111t, 210 trade elasticity, 109 trade liberalization, 22, 65–66, 137 trade policy, 4, 21–22 two-stage least squares method, 123, 124t United Kingdom, 81, 84, 181, 227, 228 United Nations International Comparison Program (ICP), 43–44, 48n, 119–20, 121t, 190 World Income Inequality (WIDER) database, 235 United States colonial heritage, 170 currency valuation, 89 (See also US dollar) current account, 7, 37–39, 38t, 93–95, 97t currency valuations and, 98–106 historical context, 192 trade balance and, 109–12, 110t–111t, 210 GDP, 189–90 historical context, 181–83, 182t investment rate, 72 miracle growth, 161 saving rate, 94 subprime mortgage crisis, 2 US Census Bureau, 234 US dollar investment and, 69–70 as reserve currency, 94, 108 valuations of, 95–97, 97t, 98t, 112, 115 value forecasts, 7, 108–13, 228–29 US Federal Reserve, Broad Index, 95–102, 97t–99t, 100f–103f, 115

US Treasury, 90–91, 211 wages, 8, 211–14, 213t Washington Consensus, 4, 7, 24, 29–31, 30t, 203 The Wealth of Nations (Smith), 12 weighted average relative price (WARP) index, 95–100, 97t–99t, 100f–103f, 115 white settler population, 170–71 wholesale price index, 44, 46 women, labor force participation rate of, 26 World Bank, 29 East Asian growth study, 136, 137

economic freedom, 65 governance indices, 169, 171 on institutions, 167 institutions index, 174 poverty estimates, 58–59 PPP estimates, 190 World Development Indicators, 25, 234, 235 World Development Report, 22 World Income Inequality (WIDER) database, 235 yen exchange rate (1950), 186–87, 187t

INDEX 263