Global Credit Review 9789814412643, 9789814412636

Citation preview

GLOBAL CREDIT REVIEW Volume 2

World Scientific

8531_9789814412636_tp.indd 1

21/8/12 2:26 PM

Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE and Risk Management Institute National University of Singapore 21, Heng Mui Keng Terrace, Level 4 Singapore 119613

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.

GLOBAL CREDIT REVIEW Copyright © 2013 by World Scientific Publishing Co. Pte. Ltd. and Risk Management Institute, National University of Singapore All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publishers.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the Publishers.

The responsibility for facts and opinions in this publication rests exclusively with the authors and their interpretations do not necessarily reflect the opinion of the publishers.

ISBN 978-981-4412-63-6

In-house Editor: Agnes Ng

Typeset by Stallion Press Email: [email protected]

Printed in Singapore.

Agnes - Global Credit Review.pmd

1

9/3/2012, 3:33 PM

b1423

Global Credit Review Volume 2

3 September 2012 1:42 PM

Message from the Editor

T

he Risk Management Institute (RMI), a university-level research institute at the National University of Singapore, strives to become a financial risk knowledge centre where scholars, regulators and industry professionals gather to advance cutting edge research. By advancing the knowledge frontier and transferring know-how to the community, we aim to contribute to the professional practice of financial risk management. In line with this objective, this second annual volume of the Global Credit Review targets finance professionals, policy makers and academics with an interest in credit markets. The Global Credit Review provides an overview of the most important developments in global credit markets and the regulatory landscape, covers theoretical and empirical research on credit ratings and credit risk, and reports on recent findings and evolutions of the RMI Credit Research Initiative (RMI-CRI). Events in the previous 12 months have led to tumultuous consequences for global credit markets and the

months ahead are likely to be volatile and challenging as well. The European debt crisis and subsequent bailouts have severely hit investor confidence, heightened risk aversion, and increased uncertainties in credit markets. To date, the eurozone is still struggling to regain its stability and credibility while being in a state of ‘slow burn’ that is likely to continue for a long time. The main challenges for the eurozone are likely to be political instability and lack of decision making which will ultimately pose the biggest threat for future global recovery. In the UK, further fiscal austerity and a weakening labor market add to the negative effect of the eurozone and put downward pressure on its economy. In the US, the fiscal deficit remains large with foreign holdings of US government debt hitting a record high. In addition to the European threat, the US fiscal deficit severely limits its policy options to boost economic growth for the US. The recent EU and US slowdown has reverberated and spill-over effects are felt globally. Asia is struggling with the forces of growth and inflation concurrently. Emerging

GLOBAL CREDIT REVIEW VOLUME 2

1201001.indd iii

iii

9/3/2012 2:10:50 PM

b1423

Global Credit Review Volume 2

markets continue to grow, but at a slower pace due to their need to rely more on domestic consumption rather than exports. Financial markets seem more immune to the European threat with market indices such as VIX no longer in sky high territory. However, financial juggling alone cannot be a panacea for global economic ills. While austerity measures have been put in place, whether there is sufficient discipline and determination to execute such measures remains doubtful due to the varying interests of stakeholders; hence, significant risks remain in the global credit markets. In this edition of the Global Credit Review we provide some critical analysis and new insights to address the challenges ahead. This second volume of the Global Credit Review begins with an opinion piece by Dr. David Rowe entitled ‘Credit Markets: Retrospect and Prospect’. In a second article ‘An Improved Regulatory Framework for Credit Rating Agencies?’ Mr. James Weston provides a critical analysis of the most important attempts at CRA regulation that have recently been undertaken by authorities worldwide. This is followed by an article on ‘Stress Testing’ by Mr. Noel D’Cruz and Dr. Davide Crippa. In a fourth article, ‘MegaBanks’ Self-Insurance with Cocos: A Work in Progress’, Prof. George von Furstenberg provides a framework for a more meaningful use of cocobonds. We continue

iv

1201001.indd iv

3 September 2012 1:42 PM

with an article by Prof. Emmanuel Mamatzakis and Mr. Panos Remoundos ‘What are the Driving Factors Behind the Rise of Spreads and CDS of Eurozone Sovereign Bonds?’ Next is a review article on ‘Measuring Distance-to-Default for Financial and Non-Financial Firms’ by Prof. Jin-Chuan Duan and Mr. Tao Wang. This is followed by an updated ‘NUSRMI Credit Research Initiative Technical Report’, which offers technical details on implementation and performance of the current RMI-CRI corporate default prediction model. We conclude this second volume of the GCR with ‘A Lead-Lag Investigation of RMI PD and CRA Ratings’. This final article provides a more qualitative check on the RMI PD in which we compare the behaviour of RMI PD to the rating actions of external credit rating agencies such as Moody’s and S&P for some well-known default cases. For a more elaborate discussion and update on our default predictions produced thus far, we refer readers to our Quarterly Credit Review. We hope you enjoy reading this second volume and welcome your feedback on the Global Credit Review and the RMI-CRI. Prof. Jin-Chuan Duan Director, NUS-RMI Cycle & Carriage Professor of Finance NUS Business School

MESSAGE FROM THE EDITOR

9/3/2012 2:10:50 PM

b1423

Global Credit Review Volume 2

21 August 2012 3:29 PM

Contents

Message from the Editor Credit Markets: Retrospect and Prospect David Rowe

b1423_FM.indd v

iii 1

An Improved Regulatory Framework for Credit Rating Agencies? James Weston

11

Stress Testing Noel D’Cruz and Davide Crippa

39

Mega-Banks Self-Insurance with Cocos: A Work in Progress George von Furstenberg

53

What are the Driving Factors behind the Rise of Spreads and CDS of Euro-Sovereign Bonds? Emmanuel Mamatzakis and Panos Remoundos

79

Measuring Distance-to-Default for Financial and Non-Financial Firms Jin-Chuan Duan and Tao Wang

95

NUS-RMI Credit Research Initiative Technical report RMI staff

109

A Lead-Lag Investigation of RMI PD and CRA Ratings RMI staff

169

8/30/2012 9:43:48 AM

b1423

Global Credit Review Volume 2

21 August 2012 3:29 PM

This page intentionally left blank

b1423_FM.indd vi

8/30/2012 9:43:48 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:03 AM

Credit Markets: Retrospect and Prospect

INTRODUCTION

A

Dr. David M. Rowe President, David M Rowe Risk Advisory, LIC; Senior Advisor, Kamakura Corporation [email protected]

s recently as the late 1980s, credit risk management was largely confined to microunderwriting of individual credit extensions. As early as the 1960s, work by Ed Altman and others introduced formal quantitative techniques into the analysis of the credit quality of individual firms. Later work drew on: • • •

•

actuarial techniques the history of credit rating transitions implications of option pricing theory in the context of limited liability corporations, and reduced form regression analysis and hazard rate analysis of default history against both firm-specific and macro-economic variables.1

These techniques introduced a new level of quantitative rigor into what had long been an almost purely judgmental process. Even so, analysis of portfolio characteristics and the impact of diversification was rare. The lack of such analysis had few

practical consequences, since there were limited available means for a bank to reshape its credit exposures in any case. Beginning in the 1990s, instruments arose to lay off and take on credit exposure of varying kinds. This made the application of modern portfolio analysis to bank balance sheet management more than a theoretical curiosity, it made it a practical and competitive necessity. Today credit risk management faces three broad issues. •

•

First, growing complexity of structured instruments has not been matched by regulations and incentives to maintain, organize and distribute the massive quantity of detailed data needed to analyze them effectively. In this context, the market fell back on overly simplistic aggregate valuation tools, such as the Gaussian Copula “Model”, that have proven to be seriously flawed. Second, the application of quantitative micro-analysis of individual obligors has not been

GLOBAL CREDIT REVIEW VOLUME 2

1250001.indd 1

1

8/30/2012 9:31:48 AM

b1423

•

Global Credit Review Volume 2

matched by corresponding attention to macroeconomic factors that drive covariability of firms across a diversified credit portfolio. This is an especially serious problem in the context of analyzing major stress scenarios. Third, the growing application of quantitative techniques gave rise to a problem of “Two Cultures”. There often was, and still is, a palpable lack of understanding and effective communication between traditional judgmental credit analysts and their more quantitatively oriented colleagues. This failure of communication and understanding has hampered the insights of both groups.

How practitioners address these issues will determine the effectiveness of credit risk management in coming years.

I. RETROSPECT 1.1 The Transformation of Credit Risk Management Up until 50 years ago, traditional credit risk management dominated the way banks thought about the potential failure of obligors to meet their commitments. This approach essentially involved detailed micro-analysis of: • • • • •

a company’s financial history and current status the size and prospective growth of its market its management its competitors barriers to entry into its market such as — — — — — —

tariffs patents economies of scale an established service network brand recognition alternate technology, etc.

1.2 The Altman Z-Scores2 In the early 1960s, a then young academic named Ed Altman turned his attention to improving the rigor of credit risk analysis. Focusing on accounting statements, 2

1250001.indd 2

30 August 2012 9:03 AM

he developed a weighted average of five financial ratios that he called a Z-score.3 The five ratios and the weights were chosen to maximize the discriminatory power of the resulting index in predicting default over the one and two years following the date of the statements. In effect, Altman brought rigorous statistical techniques to the task of defining a credit quality index. In practice, a Z-score was generally supplemented by more qualitative factors such as those cited above. For nearly 50 years, various manifestations of the Altman Z-score have continued to play an important role in fundamental credit analysis.4

1.3 Beyond a Purely Micro Focus Through much of the 1980s, despite the types of quantitative advances developed by Altman and others, if you asked a banker how his institution controlled its credit risk, the answer tended to be something like, “We only make good loans.” The mindset and discipline that this approach involved should not be underestimated. Careful attention to the details on the ground is an essential part of the process of controlling credit losses. What this approach ignores, however, is that the financial strength of some obligors will be affected very differently from that of others in the face of any given economic situation. In effect, a purely micro-underwriting approach ignores the central insight of portfolio theory that diversification often can reduce risk (measured as the volatility of value) without lowering expected return. To a large extent, this lack of attention to portfolio issues was a consequence of the buy-and-hold model of banking combined with the regional fragmentation of the industry. Banks tended to be captive to the industrial structure of their service areas but there were limited ways to reshape the composition of their credit exposures. In this environment, defining a preferred exposure profile might have been an interesting theoretical exercise but it had little practical value.

1.4 Upheavals Strain the Traditional Bank Business Model Beginning in the early 1970s, several forces began to undermine this long established model of how banks

CREDIT MARKETS: RETROSPECT AND PROSPECT

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

conducted their business. The significant economic volatility of this era, driven in part by two huge spikes in the price of oil in 1973–74 and 1979, put significant stress on the illiquid structure of bank balance sheets. A variety of instruments arose to allow banks to transfer their assets to each other and to non-bank holders and to reshape their exposures by synthetic means. These instruments included collateralized securities based on retail assets such as home mortgages, auto loans and credit card balances. In the commercial lending arena, an active market for trading whole loans developed and was followed by the introduction of Collateralized Loan Obligations (CLOs) backed by portfolios of commercial loans. The early 1990s saw the beginning of single name Credit Default Swaps (CDSs) as well as basket CDS structures that could be customized to meet the hedging needs of a specific end-user very effectively.5 These innovations enabled a major shift in the business model of banks from originate-and-hold to originate-and-distribute. Of necessity, it also focused attention on the portfolio dimension of credit risk management. The risk of any collateralized pool depended not just on the average credit quality of the underlying obligations but on how likely they were to experience simultaneous credit weakness. Furthermore, these innovations offered banks the ability to restructure the composition of their credit exposures across regions and industries. Suddenly, application of quantitative modern portfolio concepts to the management of bank balance sheets was not just a theoretical curiosity; it had become a practical and competitive necessity. This gave rise to several additional approaches to estimating credit quality.

1.5 The Actuarial Approach The actuarial approach to portfolio default estimation was at the core of the CreditRisk+ model introduced by Credit Suisse Financial Products in 1997. This approach treats estimation of the default distribution of a portfolio of credit risky assets as analogous to the mortality distribution of a population of people of different ages and health conditions. The probabilities of default are drawn from historical survival statistics for corporations of various ratings. The method does not

30 August 2012 9:03 AM

reference financial statement data or equity market valuations. The assumptions behind the model are that the probability of default in any sub-period, say a month, is the same as any other sub-period of equal length. It also assumes that the number of defaults in one period is independent of the number of defaults in any other period. Given these assumptions, the probability distribution of the number of defaults during a given period is well approximated by a Poisson distribution. It is recognized that defaults are sensitive to the general state of the economy. This can be incorporated into the method by making the mean number of defaults a stochastic variable linked to an indicator of economic conditions. This can be extended further by linking the variability of default rates to multiple background factors tied to specific industries. The mean default rates for specific obligors are then estimated as linear functions of these factors. One drawback of the actuarial approach is that it only addresses default risk and ignores the impact of rating downgrades. Nevertheless, this is broadly consistent with the traditional historical cost treatment of banking book assets. It also assumes that exposure to each obligor is known and fixed, which requires treating committed but unused credit lines as having known exposure at default.

1.6 Credit Migration Approach Just as Ed Altman’s analysis was a rigorous extension of traditional approaches to credit risk assessment, so the credit migration approach builds on historical data for credit ratings. A transition matrix displays all rating classes in the headers for both the columns and the rows. The elements of this matrix indicate the probability that an obligor starting a period with a rating corresponding to the row will end the period with the rating corresponding to the column. The largest probabilities tend to lie along the diagonal, indicating the high likelihood that a firm’s rating will be unchanged during the period. In its simplest form, this approach makes the aggressive assumption that the probability of a firm migrating to another rating is the same for all firms in a given rating class. For multi-period analysis it is GLOBAL CREDIT REVIEW VOLUME 2

1250001.indd 3

3

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

possible to introduce momentum factors for one or more periods if the data are available to support this level of detail. In this case, entities that migrated in the previous period or periods are deemed to have different future transition probabilities than those with ratings that have been stable. In addition, it is possible to apply different transition matrices depending on the projected state of the economy. Extending the transition approach to modeling multiple holdings requires some means of imposing correlations on the migration behavior. The approach implemented by CreditMetrics involves simulating the value of each firm’s assets against a grid that maps simulated asset values to corresponding credit ratings. This mapping preserves the migration probabilities of the applicable transition matrix. Historical correlations among each firm’s equity value changes are used as proxies for asset correlations and these are imposed on the simulation process. Future cash flows are then discounted in each simulation based on rates appropriate to the credit rating implied for each instrument in that scenario. Repeating this simulation many times produces an estimated distribution of future portfolio values from which a credit value-at-risk estimate can be derived.

1.7 The Merton Model In 1974, Robert Merton pointed out that the legal structure surrounding a limited liability corporation implies that debt holders have effectively written a put on the assets of the firm to the benefit of the equity holders. The strike price for this put is the book value of the liabilities. If the market value of the unleveraged assets falls below the book value of the liabilities, the equity holders have the option to “put” the assets to the debt holders. This effectively limits the downside loss of the equity holders while leaving them with unlimited upside potential, which is identical to the payoff of an asset owner with a put option. Unfortunately the market value of the unleveraged assets is not observable. The market value of the equity can be observed, but it combines the value of the excess assets (i.e., total assets less the book value of the liabilities) and the value of the implicit put option on the

4

1250001.indd 4

30 August 2012 9:03 AM

assets. In addition, however, it is possible to observe the market consensus implied volatility of the value of the equity based on the option market. From these two sources, it is possible to tease out estimates for both the level and the volatility of the value of the corresponding assets. Taken together, a) the empirical distribution for asset values based on history, b) the estimated current level and volatility of asset values and c) the book value of debt provide a basis for estimating an expected default frequency. Furthermore, since the asset values are estimated explicitly, their observed correlations can be calculated directly rather than imputing correlations based on changes in equity values.6 The Merton Model approach represents a significant departure from traditional credit analysis techniques. Rather than examining fundamentals directly, the intent is to extract the implication of the combined analysis of the market as it is manifested in a firm’s stock price. Initially, traditional credit analysts were almost universally skeptical of attempts to deploy the Merton Model in practice. While that skepticism has softened in recent years, it remains quite common. Nevertheless, most balanced assessments deem the approach to be broadly successful. Proponents argue that the model captures credit deterioration in specific entities much sooner than traditional credit analysis tools. Skeptics counter that market-based assessments such as the Merton Model produce too many predictions of deterioration that fail to materialize. The biggest shortcoming of the Merton approach is that it is a purely statistical technique based on the history of equity values and equity option prices. Historical default probability estimates have no direct link to specific firm financial characteristics other than the level and maturity of liabilities or to the influence of macroeconomic events. Forward looking simulations are driven by the volatility and correlation assumptions imposed on the stochastic behavior of changes in the market value of assets. It has no means of distinguishing the differential impact of alternative macroeconomic scenarios. This makes it ineffective for scenario evaluation and stress testing, which are increasingly important forms of analysis both to satisfy regulatory demands and for internal assessment of vulnerability to potential extreme events.

CREDIT MARKETS: RETROSPECT AND PROSPECT

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

1.8 Beyond the Merton Model Various analysts have extended the basic Merton Model in ways that are broadly consistent with its initial framework. Most of these continue to rely, however, on implicit estimation from observable bond prices, credit default swap prices, option prices or some combination of these. This leaves them vulnerable to the criticism that they are of limited value for scenario analysis and stress testing. An alternate approach is to derive the full term structure of default probabilities by explicit estimation using a historical default database. This alternate approach was first implemented on a sustained basis by Robert Jarrow and Donald van Deventer in 2002.7 The approach used in deriving default probabilities from historical data employs a hazard rate modeling estimation procedure using logistic regression. Estimated default probabilities P[t] are fitted to a historical database with both defaulting and non-defaulting observations and a list of explanatory variables Xi. Chava and Jarrow (2004) note that a logistic regression is the maximum likelihood estimator when trying to predict a dependent variable that is either one (i.e., in the default case) or zero (in the “no default” case). The explicit equation form used is: n    P[t ] = 1/ 1 + exp  − α − ∑ β i Xi [t ]  .     i =1

This reduced form approach can employ any variable that improves the quality of default prediction, including Merton default probabilities if they have explanatory power. This means that the reduced form approach can never be worse than the Merton Model because the Merton Model results can always be an input. The explanatory Xi values in this equation also can include the inputs to a traditional Altman Z-score. In this sense, the reduced form/logistic regression approach draws on the preceding work of both Altman and the several variations of the Merton Model. Investigating the contribution of relevant macroeconomic variables to the determination of firm-specific default probabilities is a logical and consistent extension to this basic estimation approach.

30 August 2012 9:03 AM

Relative to earlier methods, this process produces improved ordinal ranking of companies for grouping into defaulting and non-defaulting categories at various horizons. Adjustment of default probabilities up or down, preserving the ordinal ranking, assures that the default probabilities estimated are consistent with the likelihood of default as revealed by the actual number of failures observed over the historical sample. Default/bankruptcy predictions are an issue of academic interest. Substantial methodological advancement has been made in recent years. For example, Duffie et al. (2007) model defaults using the doubly stochastic Poisson processes with explicit consideration of the censoring effect arising from other exits such as mergers and acquisitions. Duan et al. (2012) devise a forward-intensity approach to model defaults and other exits. Once the best estimates of historical probabilities of default (PDs) are derived using all empirically useful micro and macro factors, a straightforward step is to estimate a “reduced-reduced form” equation to explain as much of the historical movement in the PDs as possible based on macroeconomic factors alone. The difference between the resulting predicted values and the historical PD estimates can be viewed as companyspecific idiosyncratic risk that is uncorrelated across companies in the universe under consideration. These reduced-reduced form PD equations are ideal for evaluating the expected impact of hypothetical stress scenarios. They also can be used as the basis for Monte Carlo simulations conditional on a given macro-economic scenario. In this case, the idiosyncratic components can be simulated on an uncorrelated basis, since the macro-economic and industry factors have been accounted for in the structure and parameters of the reduced-reduced form relationship.

II. PROSPECT 2.1 Structured Securities: The Failure of Top-Down Pricing In some extreme cases, quantitative credit analysis has become almost exclusively macro oriented and

GLOBAL CREDIT REVIEW VOLUME 2

1250001.indd 5

5

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

effectively detached from the micro details of the underlying obligations and their obligors. The most serious example of this is the use of Gaussian Copulas to evaluate different tranches of Collateralized Debt Obligations (CDOs). These instruments often have tiered loss tranches designed to attract a variety of investors with a wide range of risk/reward profiles into the debt markets. Everyone understands that default correlation is central to the distribution of total credit losses in a CDO. Nevertheless the casual, even simplistic manner in which correlation has been used in quoting prices for these instruments should give one pause. Rather than building on the characteristics of the actual underlying instruments in a portfolio, each tranche is priced on the basis of one pair-wise correlation across all names. Not only that, the single common correlation used for all names is different for different tranches, leading to what is known as the correlation smile. In effect, no single consistent stochastic structure is ever able to explain the price for all components of this type of instrument. Once fully understood, this is an anomaly of breathtaking proportions. In truth, the Gaussian Copula “Model” is not a model at all. It was simply a handy way for traders to communicate with each other. The model has only a very tenuous link to the characteristics of the underlying collateral through the average default rate. The covariability is treated only by implication from the prices of each traunch and even here the treatment is internally inconsistent.

2.2 Structured Securities: The Transition to Bottom-Up Pricing Transitioning to a bottom-up approach to pricing structured securities is both an analytical and an infrastructure challenge. At the heart of the problem is that the complexity of credit instruments has increased dramatically in the past 25 years. This complexity has advanced on two fronts. First collateralized securities have been created based on an ever wider range of underlying obligations. What began as a way to package large commercial obligations or home mortgages subject to strict and inflexible underwriting standards 6

1250001.indd 6

30 August 2012 9:03 AM

has expanded to include auto loans, revolving credit card debt, trade receivables and even such things as future movie royalties. What has lagged far behind is an infrastructure to provide ready access to all relevant risk related data on the underlying obligations. The second source of complexity is the wide variety of ways in which customized payment waterfalls are constructed. What is needed is an analytical tool to access the details of the payment waterfall and a means to assess how any given structure will allocate the available cash under different stress scenarios. In large measure the failure of markets to address these two problems is not surprising. The dual forms of complexity combined with the absence of adequate data and the associated analytical tools to evaluate their implications have fostered ever greater opacity in credit markets. This inevitably works to the advantage of large sell-side firms. In a crisis, these firms themselves can fall victim to this opacity (consider Bear Sterns, Lehman Brothers, RBS and others). Nevertheless, on a day-in and day-out basis, opacity clearly supports wider bid-offer spreads that serve to enrich those who make markets in these instruments. It is hardly surprising that sell-side firms oppose reforms to bring greater transparency to these markets with all the political pressure that their financial clout can command. What is surprising is how passive buy-side firms have been in accepting this situation as an unavoidable state of nature. It can be argued that until recently the cost and availability of computer information storage, processing power and communication capacity presented significant obstacles to addressing this problem. Today those obstacles have largely disappeared. A system where the underlying details of every individual mortgage in a MBS (such as up-todate information concerning payment status, geographically related comparables, original and current loan-to-value ratios and much more) along with the cash flow structure of the security and the implications of pre-existing defaults or repayments, could be maintained in a coherent database available to market participants. The main obstacle to this is resistance to divulging information that is deemed to convey competitive advantage. Technology can create and maintain greater transparency in these markets if the

CREDIT MARKETS: RETROSPECT AND PROSPECT

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

buy-side, regulators and the general public can muster the collective will to demand it.

2.3 Market-Driven Transparency How could such a facility become a standard feature of the markets for complex financial products? As Adam Smith would have said, we will not accomplish this by appealing to “the benevolence of the butcher, the brewer” or the investment banking executive. The dramatic improvement in transparency that technology now makes possible will only be fully realized and effectively maintained through appeals to self-interest. In addition to regulatory pressure, establishing such a system will require several things. First it will require a well heeled insurgent organization with little or no stake in the current market arrangements to underwrite the technical development of such a system. The Credit Research Initiative developed by Risk Management Institute at NUS in 2009 is providing such a facility in the domain of corporate credit risk. More specifically, through an easy-to-use web portal, the PDs for nearly 50,000 firms are available for users who can give evidence of their professional qualifications to ensure that they will not misuse the data. General users without global access are restricted to a list of 2,200 firms. Full transparency is obtained by documenting the methodology and operational implementation in a technical report that is accessible to all users. Second it will require participation commitments from a core group of buy-side firms that would stand to benefit from the greater transparency, lower risk and sharper pricing that such a system would create. Finally, it will require commitment from some aspiring second-tier sell-side firms that would stand to benefit from a first mover advantage by being an early participant in such a transformative arrangement and the big increases in trading volume it would create. Essential to the success of such an arrangement will be establishing sufficient trading volume and associated liquidity to assure investors that they can transact in reasonable volume without significant impact on prevailing prices. Marketcore,8 an intellectual property company, has designed a patented business method to achieve this goal. It is centered on provision of timelimited transaction credits to liquidity providers. These

30 August 2012 9:03 AM

credits provide either discounts on future trades or privileged access to the uniquely valuable detailed data such a system makes available. In essence, the business method leverages the most valuable commodity such a system creates, namely the consistently organized detailed data on the complex securities being traded, to solve the key challenge that any new trading system faces, namely building reliable liquidity. The stars are well aligned to support such a development. One indication of this is that the first such transformation is actually in initial operation. LexisNexis has collaborated with the Council of Insurance Agents and Brokers (CIAB) and Marketcore to create the LexisNexis Insurance Exchange.9 It is initially focused on property and casualty policies but it has plans to expand into life and health as well as reinsurance. Since a similar mechanism would be equally applicable to various heterogeneous credit and derivative instruments, this might just be the beginning of a much broader market transformation. If this transformation materializes, it will result in more robust and resilient credit markets. Such a structure would allow a wide variety of analysts to track and evaluate these securities based on reliable empirical data rather than on marketing hype or on complex top-down analytic techniques that are largely out of touch with the actual underlying collateral. In the end, such a structure would provide many opportunities even for those sell-side firms that will resist it the most. A more transparent market built on access to reliable and up-to-date detailed data will generate demand for new and innovative hedging instruments that these firms are so well equipped to provide. Given the broad social benefits that flow from more efficient allocation of savings into real investments with the best return, we all should work to realize this vision.

2.4 Portfolio Dynamics: From Simple Correlations to Structural Analysis Covariability is also an issue for aggregate bank portfolio analysis. Following the traditional Markowitz model, most statistical approaches to credit risk assessment rely on historical correlations to evaluate the implications of covariation in credit quality across the portfolio through time. The problem with this GLOBAL CREDIT REVIEW VOLUME 2

1250001.indd 7

7

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

approach is that it only captures the average pattern of covariation over the available sample period. Movements in credit quality are typically driven by a combination of macro-economic events and idiosyncratic factors specific to each company. In this context, most of the covariation across individual companies is explained by common sensitivities to macroeconomic factors. A simple correlation approach only reflects the impact of the average historical business cycle. The problem is that different business cycles are driven by different forces and any one past cycle or hypothetical future scenario may diverge considerably from the historical average. An approach that simulates portfolio behavior using only past average covariation is bound to miss important aspects of any specific potential scenario. It is for this reason that I believe greater use of structural linkages, albeit in reduced form such as in the approach by Jarrow and Chava (2004), Duffie et al. (2007) or Duan et al. (2012), is the best available analytical method in an increasingly uncertain world.

2.5 Behavioral Obstacles and the Danger of Two Cultures As this paper illustrates, credit risk analysis has become increasingly quantitative and technical over the past several decades. Much of this analysis has been viewed with varying degrees of skepticism by traditional credit analysts. In a very real sense this mirrors a cultural problem that C.P. Snow described in his 1959 essay entitled The Two Cultures and the Scientific Revolution. In it Snow highlighted the often willful lack of communication between scientists and literary intellectuals.10 In all too many cases, Snow argued, formal training compounded inherently different mindsets to produce a nearly complete lack of understanding and communication across these two cultures. Scientists, he found, often had little interest in or exposure to imaginative literature. On the other side, literary intellectuals often treated their realm as the whole of culture, blithely oblivious to the scientific edifice of the physical world as “in its intellectual depth, complexity and articulation, the most beautiful and wonderful collective work of the mind of man”. 8

1250001.indd 8

30 August 2012 9:03 AM

A similar problem afflicts the practice of modern finance, namely the split between “quants” and the larger community of traditional finance managers. Quantitative techniques and statistical risk management are little more than opaque black boxes for all too many general financial executives. What is more, those who do understand the technical details often have limited insight into broader structural and behavioral issues. They also have little incentive to make their work more transparent to outsiders since this would undermine the “mystique” that surrounds their skill set. In some cases, a lack of technical insight has little or no serious consequences. After all, few of us can understand the technical mechanics of a modern automobile but that does not inhibit our ability to drive. In the case of financial management, however, the impact of Two Cultures can be serious indeed. This is primarily because running a financial institution demands a constant series of large and small decisions under uncertainty. Such decisions can never be effective if they are made mechanically. Effective decisions must reflect experience and judgment conditioned by the available empirical evidence. As finance has become ever more complex and quantitative, the communications gap between its Two Cultures has become ever more consequential. Most senior bank managers are unable to weigh the subtle details of modern quantitative finance and few state-of-the-art quants are well equipped to assist them (even if they were motivated to do so). The weakness of the Gaussian Copula model discussed previously is a case where the existence of Two Cultures was an obstacle to effective risk management. If more general business executives had fully grasped the utter inadequacy of the analytical framework on which this huge market was based, it is possible that more firms would have acted sooner to pull back from the brink as the sub-prime sector of this market approached the point of collapse. Unfortunately I have no magic answer to the Two Cultures problem. The number of people with the background to feel genuinely comfortable in both cultures will continue to be limited. Recognizing their contribution as a bridge to facilitate communication across the organization and to raise the level of insight on both

CREDIT MARKETS: RETROSPECT AND PROSPECT

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

sides of the cultural divide is a step in the right direction. Offering opportunities where representatives from both cultures can interact on substantive issues, such as senior policy committees, will also help. Beyond this, just raising awareness of the potential dangers from miscommunication and lack of insight across the groups can be helpful when important decisions depend on considerations from both perspectives.

4

5

III. CONCLUSION Credit risk management has been transformed beyond recognition over the past 50 years. That transformation has accelerated over the past 20 years with the introduction of capital market instruments to transform and transfer credit exposure among market participants. Competitive success will require that firms develop effective working relationships across traditional qualitative credit analysts and their newer more quantitative associates. It will also require greater attention to portfolio dynamics and the impact of macro-economic factors to gain maximum advantage from the instruments that allow firms to reshape the composition of their credit risk exposure. Finally, collective attention is essential from the financial industry and regulatory institutions to establish regulations and incentives to create, organize, maintain and distribute the ever growing mountain of detailed data needed for effective bottom-up valuation and risk analysis of the increasingly varied and complex instruments available in the market. Lacking the raw material for sound analysis, there is little institutions can do but wait for the next crisis driven by irrational enthusiasm that goes unchallenged by empirically grounded insights.

NOTES 1

2

3

Other nonlinear analytical tools, such as support vector machine and neural network have also been applied to default analysis. For a more comprehensive account of statistical tools for default analysis, readers are referred to Duan and Shrestha (2011). This and the following five sections draw heavily on Rowe and Day (2007a). Altman’s original ratio was tailored to public industrial companies and involved the following five financial

6

7

8

9

10

30 August 2012 9:03 AM

ratios: (A) EBIT/Total Assets, (B) Net Sales/Total Assets, (C) Market Value of Equity/Total Liabilities, (D) Working Capital/Total Assets and (E) Retained Earnings/ Total Assets. The Z-score was defined as: Z-score = 3.3 × A + 0.999 × B + 0.6 × C + 1.2 × D + 1.4 × E. Indeed Dr. Altman himself, now just past 70, continues his research and remains a widely quoted expert on credit risk issues. For further background on the history of these developments, see Rowe and Day (2007b). The estimation method described here is a volatility restriction method. A comprehensive discussion on the pros and cons of this and other estimation methods are available in Ericsson and Reneby (2005) and Duan and Wang (2012). Please note that I am a Senior Advisor to Kamakura, the main commercial provider of this type of analysis. My association with the firm has only strengthened my conviction that this approach is the most effective available means of deriving an empirically based translation of macro events to their micro credit implications. See http://www.marketcore.com/index.php. In the interest of full disclosure, I should note that I am a Senior Advisor to Marketcore as well as Kamakura. See http://www.businessinsurance.com/article/20120101/ NEWS04/301019988 and http://blogs.lexisnexis.com/ insuranceexchange/2012/02/10/getting-left-behind-iscloser-than-you-think/ Snow was a trained scientist who also wrote imaginative literature. As such, he was uniquely qualified to assess the problem of The Two Cultures.

REFERENCES Chava, S. and R. Jarrow (2004), Bankruptcy Prediction and Industry Effects. Review of Finance, 8, pp. 537–569. Duan, J.-C. and K. Shrestha (2011), Statistical Credit Rating Methods. Global Credit Review, pp. 43–64. Duan, J.-C. and T. Wang (2012), Measuring Distance-toDefault for Financial and Non-Financial Firms, Global Credit Review, pp. 95–108. Duan, J.-C., J. Sun, and T. Wang (2012), Multiperiod Corporate Default Prediction — A Forward Intensity Approach. Journal of Econometrics, forthcoming (DOI: 10.1016/j.jeconom.2012.05.002). GLOBAL CREDIT REVIEW VOLUME 2

1250001.indd 9

9

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

Duffie, D., L. Saita, and K. Wang (2007), Multi-period Corporate Default Prediction with Stochastic Covariates. Journal of Financial Economics, 83, pp. 635–665. Ericsson, J. and J. Reneby (2005), Estimating Structural Bond Pricing Models. Journal of Business, 78, pp. 707–735. Rowe, D. and T. Day (2007a), Credit Modeling Innovations. The RMA Journal, pp. 34–38.

10

1250001.indd 10

30 August 2012 9:03 AM

Rowe, D. and T. Day (2007b), Credit Risk Management’s 25-Year Transformation. The RMA Journal, pp. 40–46. Snow, C.P. (1959), The Two Cultures and the Scientific Revolution, Cambridge University Press.

CREDIT MARKETS: RETROSPECT AND PROSPECT

8/30/2012 9:31:49 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:03 AM

An Improved Regulatory Framework for Credit Rating Agencies? INTRODUCTION

C James Weston* Research Analyst, Risk Management Institute National University of Singapore [email protected]

* The author would like to thank Elisabeth Van Laere and Zhang Liming for their useful comments and suggestions.

redit Ratings Agencies (CRAs) have long been the subject of criticism from market participants and legislators alike for good reason: prior to the Global Financial Crisis of 2007– 2008, a ratings debacle occurred every three years.1 Hence, the effective supervision of CRAs has become a major consideration for regulators worldwide. Excessive regulatory reliance on privately produced and opaque credit ratings, and ineffective management of conflicts of interest contributed to the near collapse of the global financial system in the late2000s. Initially given high investment grade ratings by Fitch, Moody’s and S&P, multi-notch downgrades of thousands of residential mortgagebacked securities (RMBS) and collateralized debt obligations (CDO) in 2007 was a primary cause of the credit crunch that fueled the collapse and bailouts of large financial institutions worldwide. Starting in early2009, regulators in both the EU and the US committed to increase and enhance regulatory oversight of CRAs

and the quality of the ratings process. CRAs themselves have argued against increased regulation, claiming their reputation is at stake; credit market participants would only approach a CRA for a rating if their view and reputation carried credibility with investors, thus creating an effective self-regulatory mechanism. An escalation of the European Sovereign Debt crisis in late 2011 led to another batch of wide ranging multi-notch downgrades from the Big Three CRAs, although this time European sovereigns and banks bore the brunt of rating downgrades. Deteriorating government finances and a reduction in governmental willingness or ability to bail out large financial institutions a second time in five years, were consistently cited by CRAs in ratings commentaries. The downgrades drew the ire of politicians and market commentators, (see Box 1) as rating changes in many cases lacked correlation with market indicators of creditworthiness, such as bond spreads and credit default swap (CDS) prices. Many market participants believe that markets price credit risk changes long before

GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 11

11

8/30/2012 9:31:17 AM

b1423

Global Credit Review Volume 2

ratings are announced by CRAs. But ECB President Mario Draghi and other commentators noted that in many recent cases, eurozone sovereign bond prices moved in the opposite direction to credit rating downgrades. For example, after S&P downgraded France on January 13, yields on the nation’s bonds decreased by 4.5 bps the following day.2 In addition, several large corporate defaults during the past year have allowed market participants to gauge the accuracy of CRA ratings. For example, the sudden bankruptcy of MF Global in October, 2011 occurred while the company was still rated above investment grade by Fitch, Moody’s and S&P. According to testimony by a Moody’s representative to a US congressional subcommittee, Moody’s was unaware that the $6.3 bn bet which brought down MF Global was proprietary in nature until October 21,3 less than a fortnight before MF Global

Box 1 — S&P draws criticism across the globe. Standard & Poor’s bore the brunt of rating agency criticism over the past 12 months, as a number of high profile mistakes affected the firm’s credibility. Several announcements made by S&P during 2011 were derided for inaccuracy, mistakes, bad timing and the effect the announcements had on markets. The market reaction that was triggered by S&P’s incorrect rating actions illustrates the power CRAs have in the markets despite their past mistakes. S&P downgraded the US sovereign rating to AA+ from AAA on August 7, 2011, citing increased uncertainty about fiscal consolidation, and the effect this could have on the country’s medium- and long-term debt obligations.4 The US treasury accused S&P of miscalculating the 10-year forecast of US debt by $2 trillion. S&P denied making an error, but subsequently changed to another economic scenario that led to a 2021 debt load $2 trillion smaller than S&P’s original forecast.5 The SEC is conducting a broad examination of the downgrade, with S&P citing the country’s rising debt burden and increased political risk in their defense of the downgrade. The downgrade caused a worldwide selloff in global stock markets, with the S&P 500 experiencing its sharpest 1-day decline in 3 years on August 8, the first 12

1250002.indd 12

30 August 2012 9:03 AM

trading day after the downgrade. The SEC has subpoenaed several hedge funds in relation to trades on August 5, and interviewed S&P staff. The SEC believes that S&P’s pending US rating action had been disseminated to a number of financial institutions before the public announcement.6 On November 10, 2011, S&P mistakenly sent a message to a number of subscribers that suggested the French sovereign credit rating had been downgraded from AAA.7 In the following days the yield spread between French 10-year bonds and 10-year German bunds reached the highest level since the founding of the eurozone, as rumors persisted that the mistaken release was a sign that an actual downgrade was imminent. The company is under investigation by the French stock market regulator Autorité des marchés financiers, while the French government is a staunch supporter of measures to reduce the influence of CRAs.8 S&P’s mistake also confirmed the European Commission’s belief that CRAs require greater oversight.9 Several days later, S&P published the wrong rating when it released a review of the Brazilian sovereign credit rating on November 17, 2011.10 The headline read that the Brazilian sovereign rating had been upgraded to BBB-, the grade the country already had, when in fact S&P was upgrading the country’s sovereign rating to BBB. The body of the release correctly stated that Brazil was being upgraded to BBB. The mistaken headline inferred Brazil had been upgraded to investment grade from non-investment grade, confusing local markets and creating volatility in currency markets. On December 5, 2011, S&P warned that 15 eurozone nations faced impending negative ratings changes, including France and Germany. The firm subsequently downgraded 13 eurozone nations on January 13, with France and Austria losing their AAA ratings. Both announcements were derided for bad timing, as the potential downgrades were announced in the middle of tense European Financial Stability Fund (EFSF) negotiations, and S&P confirmed the downgrades during fragile discussions over a Greek bond exchange.11 European Commissioner for Internal Market and Services Michael Barnier expressed concerns about the timing of S&P’s announcements. He believed S&P’s ratings movements were rushed and did not make an adequate assessment of the progress made to solve the eurozone sovereign debt crisis.12

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

filed for bankruptcy on October 31. Moody’s oversight is damning, as information regarding the eurozone trades was found within MF Global’s most recent annual and quarterly reports, implying a major gap in Moody’s analysis. Conversely, American Airlines, Eastman Kodak and Switzerland– based Petroplus had all been rated well-below investment grade by the major CRAs long before each company declared bankruptcy. Mistakes made by Moody’s and S&P during the past year and CRAs’ past failings galvanized a renewed regulatory effort for increased competition in the industry and greater oversight of CRAs, particularly in the EU. Increasing CRA disclosures and reducing conflicts of interest through rating assignment systems are recurrent themes throughout regulations proposed by various regulatory bodies in the last year. To facilitate greater investor understanding, regulators worldwide have also focused on standardizing rating symbols and definitions. An endorsement and equivalence framework outlined in the EU CRA regulation has motivated a number of smaller jurisdictions to align their CRA supervisory frameworks with the EU CRA Regulation.13 At the same time, regulators in some jurisdictions outside the EU and the US (e.g., South-East Asian countries) have refrained from introducing CRA regulations until there is some commonality in CRA regulations worldwide. CRA blunders have also focused the attention of market commentators on another important part of supervisory framework changes, reducing the quasiregulatory role of CRAs through the reduction of regulatory references to ratings. In November 2010, after the Basel-based Financial Stability Board (FSB) requested that national regulators and international standard setters reduce their reliance on credit ratings, G20 leaders agreed to reduce regulatory references to credit ratings. During the past year, CRAs have vocally supported this move, as it would reduce the perceived overinfluence of ratings and also reduce mechanistic market behaviors when credit ratings change.14 Leaders in emerging economies have long been critical of the perceived power CRAs have over investment decisions due to regulatory references, and more specifically, the objectivity of the primarily US-based Big Three CRAs.15

30 August 2012 9:03 AM

However, little progress has been made in this regard. In November 2011, the FSB released a report outlining the progress various jurisdictions and international bodies had made in removing regulatory references to credit ratings, noting that efforts remain at an early stage.16 When the FSB released its report in November, the US had made the most progress in this regard, primarily in response to the statutory requirements of the Dodd–Frank Act of 2010. Since the report was released, US regulatory agencies have released a final rule that will remove references to credit ratings from bank regulatory capital requirements in the US, while the SEC has removed references to credit ratings from a number of securities registration forms. As the US is the first large jurisdiction to propose and implement replacements for credit ratings in legislation and regulations, it provides an interesting example of possible alternative measures of credit risk and rules that could emerge worldwide. In the following section, we explore the progress the US has made in removing references to credit ratings in current regulatory frameworks, as well as other changes during the past year to sections of the Dodd–Frank Act applicable to CRAs. The remainder of this article explores proposed and implemented changes in CRA regulatory frameworks in the EU and other jurisdictions. In addition, we explore potential and actual changes in competitive dynamics in the increasingly global credit rating industry.

I. CRA REGULATORY DEVELOPMENTS IN THE US 1.1. The Dodd–Frank Act and CRA Compliance The US Dodd–Frank Act of 2010 endowed the SEC with greater decision making authority in regulating CRAs, known as Nationally Recognized Statistical Rating Organizations (NRSROs) in US regulations. The Act aimed to revamp the credit rating industry by introducing a series of rules pertaining to compliance, internal controls, ratings quality, disclosures of information, management of conflict of interests, and end-user protection. Many of the GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 13

13

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

regulatory measures in the Dodd–Frank Act that are applicable to NRSROs were due to come into effect by June 2012. However, a number of proposed measures have been delayed or are still in commentary phases, and some are not due to come into effect until the end of 2012. For instance, the establishment of the SEC’s Office of Credit Ratings was delayed until June 2012, due to budgetary constraints and a lack of resources. Despite this, SEC staff released their first annual report on NRSROs’ US-based businesses required under the Dodd– Frank Act in September 2011. The report’s findings illustrate the degree to which NRSROs have complied with the Dodd–Frank regulations currently in force and the previous Credit Rating Agency Reform Act of 2006.17 The report identified an even greater trend towards the much derided issuer-pays model, as two NRSROs which have traditionally relied on the subscriber-pays model have recently made changes to their business structure and placed greater emphasis on the issuerpays model. The two US NRSROs transitioning to the issuer-pays model are Kroll Bond Ratings, which has rated a number of structured products in the past year, and Morningstar, which has traditionally rated stocks and mutual funds. Morningstar acquired the NRSRO Realpoint in 2010, and recently began publishing bond ratings. The report found that since a previous public report in 2008, NRSROs have made notable efforts to address disclosure and ratings process issues, as they have become subject to higher legal and regulatory oversight since 2008. Also, the report asserted that NRSROs have improved their management of conflicts of interest, outside of conflicts inherent to current business models, but have ongoing problems with employee security ownership. The SEC staff believe NRSRO policies in this area could be improved, with smaller NRSROs having the weakest employee security ownership policies, even allowing key staff to own securities related to rating actions they oversee. Smaller agencies also appear to have ongoing problems in the areas of public disclosure, internal supervisory controls and the resources available to the designated compliance officer, a role required under Dodd–Frank. In April 2012, the SEC 14

1250002.indd 14

30 August 2012 9:03 AM

announced charges against issuer-paid ratings agency Egan-Jones for material misrepresentations in submissions to the SEC, and for failing to ensure analysts were free from conflict of interest problems arising from employee security ownership. Egan-Jones countersued in June to move the matter to federal courts, accusing the SEC of pursuing action over matters that have not materially affected rating actions, and a bias towards the conflicted issuer-paid business model. Egan-Jones also alleged the SEC was diverting focus from more important matters, including investigating the Big Three CRAs over their failed securitization ratings. Despite policy and procedural changes, one of the Big Three NRSROs, which remained anonymous in the SEC report, allowed limited disclosure of a pending rating action before public release. Another unnamed larger NRSRO erroneously applied its ratings methodology to a number of asset-backed securities, and later placed a majority of these ratings on review.18 To avoid similar occurrences, the SEC has proposed that NRSROs must make an announcement in the public domain whenever a significant error is identified in an NRSRO’s credit rating methodology. The SEC is yet to clarify who will define the term “significant error”. S&P expressed concern that this proposal would affect their rating decisions. S&P believes that the SEC would effectively be substituting its own judgment for the rating agencies’ if the SEC were to decide what a significant rating error was.5

1.2. Removal of Statutory References to Credit Ratings and NRSROs (Dodd–Frank 939A) One of the most far-reaching provisions under Dodd– Frank stipulates that references to ratings issued by NRSROs be removed. More specifically under Section 939A of the Dodd–Frank Act, every federal agency is required to remove any reference or reliance on credit ratings issued by NRSROs and substitute such references with an alternate standard of credit worthiness. Despite moves by several US federal agencies to remove statutory references to ratings, as discussed below, substantial legislative changes are still needed to completely remove the quasi-regulatory role of CRAs in US legislation and statutes.

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

The SEC has made the most progress in this area, adopting new rules on the usage of Form F-3 and Form S-3, referred to as ’short forms.’ The changes remove references to credit ratings in these documents, with all changes becoming effective by December 31, 2012.19 Short forms allow companies to expedite the process of issuing securities; issuers seeking to use these forms previously required an investment-grade rating from a NRSRO on securities it wished to sell to the public. The credit rating requirement has been eliminated under the new rules and replaced with four new tests (see Box 2), of which one must be satisfied for an issuer to use the short forms. A 3-year grandfather provision implies that issuers seeking to use the short forms will still be able to qualify using their credit rating for the near future. Furthermore, if an issuer’s internal rules stipulate that an asset-backed security (ABS) requires a certain rating, the ABS will still require an investment-grade rating when registering using these forms.20 The changes to short forms also remove the references to information about credit ratings from forms Box 2 — Criteria replacing ratings on SEC short forms. Four new criteria have replaced investment-grade credit rating requirements on SEC short forms.21 Issuers will be able to register securities using Forms F-3 and S-3, and use Forms F-4, S-4 and Schedule 14A by reference, if they meet one of the following criteria: 1) The issuer has completed primary offerings of at least $1 bn in non-convertible securities, other than common equity, for cash under the Securities Act in the past 3 years. 2) The issuer has at least $750 m of non-convertible securities outstanding, excluding common equity. 3) The issuer is a wholly-owned subsidiary of a well known seasoned issuer (WKSI). Rule 405 of the Securities Act (1933) defines a WKSI as an issuer that meets short-form registration requirements and has an outstanding global market value of at least $700 m. Asset-backed issuers and investment companies are not included in the definition of a WKSI. 4) The issuer is majority owned by a real estate investment trust that qualifies as a WKSI.

30 August 2012 9:03 AM

F-4, S-4 and Schedule 14A. These three SEC filings previously incorporated credit rating information by referencing short form requirements. Forms F-4, S-4 and Schedule 14A are filings used to register securities connected to mergers, in particular exchanges of securities during business combinations. SEC rule changes also eliminated the use of Form F-9, used by Canadian issuers using Canadian GAAP to issue securities in the US, which referenced credit ratings. Canadian companies will now have to register securities issuances using Form F-10, which does not contain any references to NRSROs or credit ratings.

1.3. Decreasing Reliance on External Ratings in US Bank Risk Management and Regulatory Capital Requirements22,23 The US Office of the Comptroller of the Currency (OCC), the Federal Reserve and the Federal Deposit Insurance Corporation (FDIC), collectively referred to herein as ‘the federal agencies’, released a notice of final rulemaking in June 2012 that aims to remove references to credit ratings in the prevailing US version of the Basel Accords. US banks are currently required to calculate the amount of capital they must hold against their risk-weighted assets using credit ratings; from January 1, 2013, credit ratings will be replaced with alternative measures. One of the key problems under the current regulations is a mechanical reliance on credit ratings; there is no gradual increase in riskweightings as credit ratings move from one rating grade to another, so a position’s risk-weighting can change suddenly and dramatically. A sudden change in a credit rating, so-called ‘cliff risk’, can lead to a large increase in a bank’s risk-weighted assets under the current regulations. The federal agencies’ proposal attempted to solve this problem by complementing credit rating alternatives with market based indicators. However, ‘cliff risks’ largely remain, as the federal agencies have retained the bracket like system for risk-weights; a security can still suddenly move from a risk-weight of 20% to a risk-weight of 100%. At this stage, only the 30 largest US banks and bank holding companies would have to comply with the proposed incoming rules.24 However, US GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 15

15

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

regulators could use a similar framework in future rules that also apply to smaller regional banks. Market participants noted that unless the federal agencies acknowledge that ratings are a suitable alternative for smaller banks, US community and regional banks could face significantly higher cost burdens due to increased due diligence requirements. Commentators also believe that these changes may give regulators too much discretion to criticize a bank’s risk-weighting calculations. The three federal agencies proposed the following alternative approaches to measure the risk of trading book assets. 1.3.1. Sovereigns Under the changes, banks will have to weight the debt of sovereign nations using the Country Risk Classifications (CRCs) published by the OECD, which are currently used to calculate the interest rates on export credits and are more focused on currency convertibility risk than on credit risk. Banks would also be able to weight the debt of Public Sector Entities (PSE), such as states and other political subdivisions, based on the CRC assigned to the home country of the PSE. The federal agencies believe that assessments of creditworthiness produced by international organizations such as the OECD are free from conflict of interest problems, as the OECD does not receive fees for the CRCs it produces. Some market participants believe that CRAs pressure sovereign governments into paying for ratings by maintaining relatively lower unsolicited ratings on nations that do not actively engage CRAs. This may not address ratings bias concerns expressed by the leaders of BRIIC economies (Brazil, Russia, India, Indonesia and China) and other nations, as the OECD currently does not include any BRIIC nations and the CRC methodology involves analysis of political and other risk factors which require substantial subjective analysis. The federal agencies expressed similar concerns in their request for comment, as OECD and eurozone members classified as high-income by the World Bank are automatically assigned the most favorable CRC. This could result in debt issued by non-OECD members receiving 16

1250002.indd 16

30 August 2012 9:03 AM

higher risk-weights, while debt issued by OECD nations could receive more favorable risk-weighting. For example, Greece, Hungary and Portugal each have a current CRC of 0 which is the same as the US; at the date of publication each of these countries had credit ratings below investment grade from all the major ratings agencies. Usage of CRCs could also result in large changes in assigned risk-weights (see Table 1). To alleviate this problem, banks will have to apply the highest risk-weighting factor of 12% to any sovereign which has restructured or defaulted on payments in the previous five years. However, this clause would only occur when a default or restructuring occurs and does not account for the probability of a sovereign default or restructuring; risk-weightings for lower rated sovereigns would likely be too low under this proposal. It is also unclear whether an international bail-out would constitute a default in the proposal. Moreover, various market participants remain unconvinced by the use of CRCs and pointed to an OECD statement that explicitly states that CRCs are not sovereign risk classifications and should not be compared with a sovereign rating from an NRSRO.25 Furthermore, the federal agencies expressed concern that CRCs may misclassify sovereign risk, as a large amount of quantitative information used to calculate CRCs is only available on a quarterly or annual basis. In order to reduce this problem, the federal agencies have suggested that credit default swaps (CDS) or bond spreads could be used to complement the usage of CRCs in the risk-weighting process. However, CDS and bond prices reflect more than credit risk. In particular, the liquidity premium incorporated into the market price of sovereign bonds and CDSs could distort the usefulness of marketbased determinates of sovereign risk; low liquidity in a number of sovereign credit markets could make relying on such information difficult. The federal agencies decided not to complement the use of CRCs with these market-based indicators for similar reasons in the final rule published in June 2012. In addition, the federal agencies were concerned that the usage of such indicators would require further investigation before being the included in a final rule.

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:18 AM

b1423

Table 1.

Global Credit Review Volume 2

30 August 2012 9:03 AM

CRCs and credit ratings: implied risk-weightings differ for a number of economies.

CRCs and credit ratings: implied risk-weightings differ for a number of economies Rating implied risk-weighting CRC implied risk-weighting factor Economy CRC Fitch Moody's S&P factor (%) (%) Czech Republic 0 0.0 0.25-1.6 A+ A1 AAFrance 0 0.0 0.0 AAA Aaa AA+ Germany 0 0.0 0.0 AAA Aaa AAA Greece 0 12.0 12.0 CCC C CCC Hungary 0 0.0 8.0 BB+ Ba1 BB+ Iceland 0 0.0 0.25-1.6 BBBBaa3 BBBIreland 0 0.0 0.25-1.6 BBB+ Ba1 BBB+ Israel 0 0.0 0.25-1.6 A A1 A+ Italy 0 0.0 0.25-1.6 AA3 BBB+ Poland 0 0.0 0.25-1.6 AA2 APortugal 0 0.0 8.0 WD Ba3 BB Slovak Republic 0 0.0 0.25-1.6 A+ A2 A Slovenia 0 0.0 0.25-1.6 A A2 A+ South Korea 0 0.0 0.25-1.6 A+ A1 A Spain 0 0.0 0.25-1.6 BBB Baa3 BBB+ Hong Kong 1 0.25-1.6 0.0 AA+ Aa1 AAA Taiwan 1 0.25-1.6 0.0 A+ Aa3 AAChina 2 0.25-1.6 0.0 A+ Aa3 AASaudi Arabia 2 0.25-1.6 0.0 AAAa3 AABrazil 3 0.25-1.6 0.25-1.6 BBB Baa2 BBB India 3 0.25-1.6 0.25-1.6 BBBBaa3 BBBRussia 3 0.25-1.6 0.25-1.6 BBB Baa1 BBB Bulgaria 4 8.0 0.25-1.6 BBBBaa2 BBB Croatia 5 8.0 0.25-1.6 BBBBaa3 BBBKazakhstan 5 8.0 0.25-1.6 BBB Baa2 BBB+ Argentina 7 12.0 8.0 B B3 B Belarus 7 12.0 8.0 NR B3 BLebanon 7 12.0 8.0 B B1 B Ukraine 7 12.0 8.0 B B2 B+ Venezuela 7 12.0 8.0 B+ B2 B+ Sovereign debt positions with a risk-weighting between 0.25 and 1.6% are also weighted using the maturity of the obligation All CRCs and credit ratings are correct as of June 15 2012.

1.3.2 Corporates26 Under the federal agencies’ original proposals, banks would have been allowed to use an internal riskweighting method that uses market-based information and historical accounting information to assign riskweightings to publicly listed, non-financial institutions. The methodology would make use of indicators such as the obligor’s leverage, cash flow and stock price volatility. Securities issued by corporates with low leverage and stock volatility could receive a relatively low risk-weighting, depending on the securities’ remaining tenors. Privately traded companies could receive a relatively high, static risk-weighting factor under the methodology. The federal agencies’ original proposal also involved comparing a corporate bond spread to several key CDS indices to determine a risk-weight. The problem with this is that it only removes direct references to NRSROs. The corporates that make up a

CDS index are still determined based on credit ratings. In addition, the federal agencies suggested that 1-year averages of market spreads would be used in order to smooth volatility. This would reduce ‘cliff risk’, but would reduce the informational benefits of using market-based measures. The latter could be addressed by using an exponentially weighted average. However, due to concerns about the feasibility and efficacy of market indicators to determine the risk-weighting factors of corporate debt positions, the federal agencies decided not to use the aforementioned methodologies. Under the federal agencies’ final rule, US banks will have to determine risk-weighting factors for corporate debt based on the OCC’s proposed definition of an ‘investment-grade’ security. In order to meet the Dodd–Frank requirements, the OCC plans on changing its current definition of investment-grade in order to remove references to NRSROs. At present, GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 17

17

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

to meet the OCC’s definition of investment-grade, a security must be rated investment-grade by each NRSRO that rates the security. Under the new definition, a security will be considered investment-grade if the security issuer has an adequate capacity to meet the financial commitments under the security until the security matures. To meet this definition, banks will have to show that the risk of default by the obligor is low and that the full repayment of principle and interest is expected to occur in a timely manner. In order to meet these criteria, banks will be expected to make use of internal risk ratings, default statistics, and other appropriate sources of information on the security, including external credit ratings. Banks will also have to confirm that the spread to US Treasuries and default risk is low and consistent with bonds of similar credit quality and maturity. This could result in a security rated investment-grade by an NRSRO not meeting the OCC’s new definition, and vice-versa. To alleviate concerns about disparity of investment-grade designations between banks, the federal agencies declared that their ongoing supervision of banks would address any major differences in investment-grade designations between lenders. Under the final rule, an investment-grade exposure may receive a risk-weighting factor from 0.5% to 4%, while a non-investment-grade exposure would receive a risk-weighting factor of 12%. This approach will only apply to the debt of issuers with outstanding public instruments; debt issued by private corporations must be assigned an 8% riskweighting factor. Because the financial health of corporations is generally a factor of the prevalent economic conditions, banks will not be allowed to assign a corporate debt position a risk-weighting factor lower than that which corresponds to the CRC of the issuer’s country of incorporation. 1.3.3. Banks and credit unions The federal agencies believe the differences in balance sheet structure between financial and nonfinancial corporates makes it difficult to calculate the risk ratings of financial institutions using the aforementioned methodology. With this in mind, the 18

1250002.indd 18

30 August 2012 9:03 AM

federal agencies decided that banks must assign debt exposures to other banks and credit unions a risk-weighting based on the CRC of the entity in question’s respective sovereign of incorporation. In the event the respective sovereign defaults or the sovereign has defaulted in the past five years, banks must assign the debt exposure a risk-weighting of 12%. Covered bond-like positions will not receive a risk-weighting lower than that corresponding to the CRC of the respective sovereign. 1.3.4. Securitizations The federal agencies’ final rule includes simplified adaptations of the Basel II advanced approaches to assign specific risk-weighting factors to securitization positions, instead of the current reliance on credit ratings to assign these risk-weighting factors. Commentators believe the proposed rules regarding securitizations do not adequately distinguish between differences in the underlying assets of securitizations nor effectively account for credit enhancements that help mitigate the potential risk of some securitizations. A detailed discussion of related calculations is beyond the scope of this article.

1.4. The Al Franken Amendment and CRA Business Models27–34 The Al Franken amendment was aimed at addressing conflicts of interest in the issuer-pays business model for structured products. Politicians and market commentators claim that NRSROs placed revenue and market share before ratings quality as demand for structured finance ratings and associated fees increased significantly in the early and mid-2000s. The Al Franken amendment would require an SEC sponsored board to assign the task of issuing initial ratings of structured financial products to NRSROs on a quasi-random basis and make annual assessments of ratings accuracy. It was removed from the final version of the Dodd–Frank Act and replaced with Section 939F, which mandates the SEC to investigate alternative business models for NRSROs that prevent issuers of structured products from selecting the NRSRO to perform initial ratings. The Al Franken amendment,

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

also known as Section 15E(w) of the Securities Exchange Act as proposed by Dodd–Frank 939D, was the main focus of an SEC study into alternative business models for NRSROs, which was released for commentary in May 2011. Senators Al Franken and Roger Wicker, as well as the American Federations of State, County and Municipal Employees were among the commentators supporting the implementation of the Section 15E(w) rules. These proponents believe this system is the only model that will limit conflicts of interest in the ratings business and eliminate ratings shopping. The latter refers to the practice of issuers soliciting multiple NRSROs for the highest possible rating and threatening to take business elsewhere to gain a better initial rating for new issuances. Other market participants claimed that any alternative business model for NRSROs would simply shift the source of conflict of interest problems from issuers to other parties. The American Securitization Forum (ASF) and Securities Industry and Financial Markets Association (SIFMA) replied to the request for comment with very negative views of the Section 15E(w) provisions. The ASF reflected the concern of many commentators and the SEC itself that the existence of the board would create the perception that the US government endorses the ratings provided by assigned NRSROs, encouraging even greater reliance on ratings from NRSROs and thus increasing moral hazard. However, Senators Franken and Wicker commented that the amendment would strictly state that ratings that are produced under the assignment process are in no way approved or certified by the US government. They also pointed out the clear distinction between eliminating reliance on ratings and reducing overreliance on ratings, the latter being a key goal of the Dodd–Frank Act. Due to the complex nature of structured transactions, the Senators believe that credit ratings are still a valuable form of assessment for these products. SIFMA believes that the implementation of the Section 15E(w) system would be an unprecedented instance of government control in private financial markets. According to DBRS, a government entity forcing two private parties to deal with each other could be a violation of the Fifth Amendment of the

30 August 2012 9:03 AM

United States Constitution.35 Political intrusion is also a key concern of the ASF, particularly in the wake of S&Ps downgrade of the US sovereign rating in August. The ASF is concerned about the board’s assessment of NRSRO rating accuracy, specifically because ratings assignments would depend on these reviews. Board members would not be immune to political pressure or conflicts of interest, and funding concerns could impact the board’s decision to assign issuers to certain NRSROs, as the board would be funded from levies on NRSROs and issuers. In addition, the board may propose performance measures that diverge from the market’s assessment. An assessment based on the number of downgrades following an initial rating would lead to more conservative ratings, and NRSROs would have an incentive to avoid downgrading securities. The alternative to this would involve the board analyzing transactions to determine the accuracy of ratings, effectively doing the same thing as an NRSRO, and creating pressure for NRSROs to use methodologies similar to those used by the board. This pressure could lead to a loss of diversity in opinions between ratings agencies, reluctance for NRSROs to innovate, and a subsequent decline in ratings quality. According to the ASF, because the identity of an NRSRO would be unknown until late in the sales process of a security, investors would make the most conservative assumption about the methodology and reputation of the assigned NRSRO, effectively increasing the cost of securitizations. Related to this, many institutional investors have internal policies that only allow them to invest in securities that have been rated by certain NRSROs; securities that are not rated by the largest NRSROs would be priced significantly worse and may even be unmarketable. Commentary by Morgan Stanley identified similar concerns that issuers would be disadvantaged if assigned a NRSRO without sufficient resources to adequately assess their securities.36 Senators Franken and Wicker pointed out that the assignment process would not be completely random: NRSROs would only be assigned to transactions that they have the institutional capacity and expertise to rate. The amendment simply eliminates issuer preference as one of the criterions for assignment. GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 19

19

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

Both Moody’s and S&P expressed concern that the Section 15E(w) rules would reduce competition and innovation in the rating of structured products. NRSROs would be incentivized to do the least amount of work to remain in the lottery system, as participating NRSROs would be guaranteed a steady flow of revenues. S&P also believes the system would reduce the incentive to publish public commentaries, which currently serve as a way for NRSROs to increase their reputation and awareness amongst issuers and investors. 1.4.1. SEC outlines different business models for CRAs In the SEC’s request for comment on the possible implementation of the Al Franken amendment, the SEC also asked for commentary on several proposed alternative business models for CRAs (see Box 3). The commission identified three additional possible business models for CRAs, outside of the issuer-pays and the model proposed under the Al Franken amendment. In Moody’s response to the SEC proposals, the rating agency identified the conflicts of interests that a range of parties have in the ratings process, inferring that no business model is free from conflicts of interest.

Box 3 — Proposed alternatives to the issuer-pays model. Stand-alone Model. Under such a model, NRSROs would receive compensation from issuers, investors and other participants over the life of the product. A transaction fee would be levied on participants at issuance and during secondary market transactions, with NRSROs being compensated from these fees. Designation Model. Under this model, issuers would provide information on securitizations to all interested NRSROs, and the issuer would pay a rating fee to a third-party administrator. Investors would designate which NRSROs would receive the fee they pay, with the size of each investor’s fee proportional to the size of their investment. Investor and User Funded Model. Two proposed models that are primarily investor funded involve either investors paying for a transactions rating, or investors establishing their own NRSRO. 20

1250002.indd 20

30 August 2012 9:03 AM

Issuers are motivated to obtain the highest rating, and sustain that rating, to reduce their cost of borrowing and increase security prices. Short-selling investors encourage negative rating actions, as this places downwards pressure on assets they may have shorted. Shortselling investors may also favor unexpected and more negative rating movements. Long investors may prefer lower ratings before they purchase an asset, to increase the risk premium that investors can demand from an issuer. Long investors want ratings to either be maintained or raised. Governments are motivated to protect nationally or systemically important issuers, by placing pressure on ratings firms to maintain the ratings of national champions or sovereign ratings, to limit the effect of rating ceilings on local companies. 1.4.2. SEC exemptions a sign that issuer-pays model will endure In a sign that the SEC may favor the continuation of the issuer-pays model, Kroll Bond Rating Agency and Morningstar Credit Ratings each received an exemption from an SEC rule preventing an NRSRO from issuing or maintaining a rating on an entity that provides more than 10% of an NRSRO’s revenue, in September 2011 and March 2012 respectively.37,38 Both companies are transitioning from a subscriberbased business model to an issuer-pays business model. Kroll and Morningstar believed that in changing business models, there was a possibility they could source more than 10% of net revenues from a single customer, as ratings fees paid by issuers are relatively higher than fees paid by the companies’ existing subscribers. This move could be interpreted as a soft approval of the issuer-pays model by the SEC.

1.4.3. Existing alternatives to the Al Franken Amendment Respondees to the SEC’s request for comments on alternate business models for CRAs were broadly in favor of maintaining the issuer-pays model for US structured products, as they believe conflict of interest problems in the rating of structured products have largely been addressed by efforts NRSROs have made

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

to manage and disclose potential conflicts, and Rule 17g-5 of the Exchange Act which came into effect on June 1, 2010. Rule 17g-5 requires NRSROs that rate structured products under the issuer-pays model to make information they receive from issuers of structured products available to all other NRSROs, allowing other NRSROs to make unsolicited or voluntary ratings. Such ratings from other NRSROs should make it more difficult for issuers to exert influence over the NRSROs they hire, as NRSROs that compromise their rating standards to gain business should be exposed by unsolicited ratings from other NRSROs. In addition, other rules prohibit NRSROs from advising on the legal structure, assets or liabilities of an issuer or a security, and require NRSROs to include a description of representations and warranties of a security they rate, compared to similar transactions. Both the ASF and SIFMA believe that these rules help reduce informational asymmetries and conflicts of interest in the rating process for structured products. However, Moody’s is concerned that the voluntary ratings that are supposed to arise under Rule 17g-5 could act as a substitute for increased transparency in the structured product market. Moody’s believes that the lack of publicly available information regarding securitizations is the main cause of conflicts of interest in this market, and that issuers still retain a large amount of control over the dissemination of information under Rule 17g-5. S&P identified several related problems that have prevented NRSROs from making full use of the transactional data available under Rule 17g-5. Issuers have taken the unusual step of deeming all information provided to NRSROs confidential in order to comply with Rule 17g-5, preventing NRSROs from using any information that is not available to the public in ratings rationale, and preventing the public disclosure of transactional peculiarities designed to game ratings methodologies. In addition, S&P observed several issuers instructing trustees not to respond to queries from NRSROs regarding rated transactions since Rule 17g-5 became effective. These problems may ultimately result in even less transparency in structured finance ratings. Despite these hindrances, CRAs began openly criticizing the ratings assigned to several deals by their competitors in the past year. In March 2012, Moody’s described ratings assigned to a single-tranche subprime

30 August 2012 9:03 AM

auto-debt securitization by DBRS and S&P as inappropriate. DBRS assigned the bonds a AAA rating, while S&P rated the bonds AA.39 In the same month, Fitch criticized the ratings assigned to the senior tranches of a RMBS by DBRS and S&P, which each gave the top four tranches of the RMBS a AAA rating. Fitch said the top-four tranches did not meet its credit enhancement rules for high quality mortgage collateral.40 Access to structured transaction data will also help NRSROs improve existing models; a key defense CRAs cited for their focus on market share in the structured finance sector in the mid-2000s was the informational benefit of receiving performance data from every deal. Moreover, the Big Three NRSROs believe that issuers should make all information provided to NRSROs under Rule 17g-5 requirements available to the general public as this alone would significantly increase transparency in the issuer-pays business model. This would also allow investors to review NRSRO ratings of structured finance products and encourage academic efforts in the structured finance field. It is interesting to note that an academic effort to offer a plausible “public good” alternative was launched by the Risk Management Institute (RMI) of the National University of Singapore in July 2009. RMI’s Credit Research Initiative offers daily default predictions and currently covers over 35,000 listed corporate entities. This initiative is discussed further in Section IV below. Curiously the SEC study on alternate CRA business models overlooked this effort as a possible substitute to the current status quo. If more information related to structured finance transactions is made available to the public, non-profit undertakings to assess structured finance products similar to RMI’s effort in the corporate sphere will be possible.

1.5 Reinstating NRSRO Protection Under the Securities Act of 1933 In July 2010, incoming changes under the Dodd– Frank Act removed NRSROs’ exemption to expert liability when rating information is included in a prospectus or registration statement. The changes also required issuers to obtain consent from NRSROs to use ratings in registration statements. The Big Three NRSROs subsequently withheld permission for GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 21

21

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

issuers to publish their credit ratings in bond issuance documentation, sending bond markets into turmoil on July 20, 2010. US issuers have historically included ratings in their prospectuses, and ABS issuers are mandated by SEC regulations to disclose whether a sale of ABS is dependent on a minimum rating.41 In light of this, the SEC issued a ‘no-action’ letter exempting ABS issuers from reporting ratings and minimum requirements in registration statements. The letter was extended indefinitely in November 2010.42 In addition, a bill that would restore NRSROs’ exemption from expert liability was proposed by Representative Steve Stivers in 2011 and was advanced to the House of Representatives and the Senate for review in July 2011.43 Passage of the bill would allow issuers to include credit ratings in registration statements without the express consent of NRSROs once more.44

1.6. Standardization within the Ratings Process45 In order to meet requirements under Section 939(h) of the Dodd–Frank Act, the SEC requested commentary on the standardization of credit ratings in December 2010. The SEC is seeking to address differences in the meaning of credit ratings across ratings agencies; current credit rating terminology can represent probability of default, expected loss or distance-to-default depending on the rating agency in question. For example, S&P ratings represent the pure probability of default, with the firm issuing a loss-given default rating separately for issuers with lower ratings. In comparison, Moody’s notches each rating they issue to reflect loss-given default.46 The SEC also asked if the market conditions under which ratings are evaluated or revised should be standardized. The SEC suggested that standardized market stress conditions in which credit ratings are evaluated should include changes in key economic indicators and financial market declines. This proposal is aimed at addressing criticisms of NRSROs’ ability to promptly revaluate ratings in times of market stress. Ratings agencies had rated Lehman Brothers at least A before the investment bank declared bankruptcy on September 15, 2008 and had failed to downgrade MF Global from investment grade until a week before the brokerage firm declared bankruptcy on October 31, 2011. 22

1250002.indd 22

30 August 2012 9:03 AM

In addition, the SEC suggested that under standardized economic stress conditions, a quantitative correspondence between credit ratings and a range of default probabilities and loss expectations would be useful. More nuanced credit information will itself be a major step forward in credit rating reform. Finally, the SEC has raised the possibility of standardizing credit rating terminology across asset classes and issuers, but believes the idiosyncrasies of several asset classes and limits on the quality and accuracy of data may create difficulties in achieving this goal. These proposals are very similar to EU proposals discussed below.

II. CRA REGULATORY DEVELOPMENTS IN THE EU 2.1. ESMA Becomes the EU CRA Regulator The European Securities and Markets Authority (ESMA) was given exclusive responsibility for the supervision of CRAs in the European Union (EU) from July 21, 2011. On October 31, ESMA announced that the Big Three CRAs and DBRS, a smaller Canadian CRA, were compliant with the EU Regulation on Credit Rating Agencies (EU CRA Regulations), and are now registered as CRAs in the EU. The registration process involved assessment of the CRAs’ independence, governance and transparency, with ESMA noting that its future supervision of CRAs, along with new regulatory requirements, will contribute to an improvement in the quality of credit ratings. ESMA also announced during the past year that it considers the CRA supervisory frameworks in Argentina, Australia, Brazil, Canada, Hong Kong, Japan, Mexico, Singapore, and the US as stringent as the EU CRA Regulations.47 This allows CRAs to endorse credit ratings issued in these jurisdictions, enabling credit ratings to be used for regulatory purposes in the EU. This endorsement mechanism effectively constrains EU financial institutions and a number of institutional investors to investments with endorsed credit ratings. Japan remains the only jurisdiction with a CRA supervisory framework that the European Commission has determined equivalent to the EU CRA Regulations. In April, ESMA provided

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:18 AM

b1423

Global Credit Review Volume 2

the technical advice necessary to the European Commission for an equivalence decision for the CRA regulatory regimes in Australia, Canada and the US.48 ESMA released its first annual report on the Supervision of Credit Rating Agencies on March 22, 2012.49 It is the first such examination since the amendment of Regulation No 1060/2009 (CRA 1) by Regulation 513/2011 (CRA 2), and allowed ESMA to review the extent to which CRAs have complied with CRA 1 and 2, and the registration process. The report was based on ESMA’s on-site inspections at CRAs during December 2011, and on ESMA’s analysis of ratings actions between November 1, 2010 and October 31, 2011, with the examination limited to Fitch, Moody’s and S&P. ESMA focused on sovereign, bank and covered bond ratings actions in its examination, as ESMA believed that given current market conditions, these ratings classes merited the most attention. ESMA’s major finding was that CRAs had failed to sufficiently record details of internal meetings, specifically the discussions and voting results in Rating Committees that were used by CRAs to determine rating actions. Detailed recording of internal discussions would allow CRAs to show ESMA they are meeting compliance regulations involving the application of ratings methodologies and management of conflicts of interest in the ratings process. Although ESMA identified substantial improvements in the composition and clarity of CRA methodologies, it noted that a number of CRA methodologies still required consolidation into a single identifiable document. Improvements in this regard would assist external users of credit ratings in due diligence activities. ESMA also expressed concern that high turnover in certain lines of business at one or more CRAs impaired analytical capacity and ratings quality. ESMA lauded CRAs for improvements in governance and control functions. This was largely due to EU CRA registration requirements stipulating the need for independent directors charged with monitoring compliance and internal controls at CRAs. This is in clear contrast to the governance and internal control functions at CRAs’ US offices, where the SEC found internal control functions and roles insufficiently defined and under resourced.

30 August 2012 9:03 AM

Finally, ESMA was highly critical of CRAs’ IT resources and procedures. In particular, ESMA noted a high degree of automation and limited controls in the ratings publication processes at the Big Three CRAs, and limited oversight of the publications function by analytical teams. ESMA believes this could increase the risk of publication errors, similar to S&P’s mistaken email to subscribers in November indicating a downgrade of the French sovereign rating. ESMA also expressed concern about the extensive use of external service providers in CRA IT projects, increasing the risk that confidential information could be disseminated to the market. On February 1, 2012, ESMA began publishing statistics on rating activity and performance of registered CRAs on a publicly accessible website.50 The goal of the website is to improve the transparency and clarity of information available on rating movements, and to protect investors by providing information on the performance of past ratings from CRAs registered in the EU.

2.2. A Third Round of EU CRA Regulations51 On November 15, 2011 the European Commission, led by the European Commissioner for the Internal Market and Services, Michel Barnier, released a nascent proposal outlining the most intrusive regulatory environment for CRAs currently under consideration. Important changes centered on adjustments in rating outlooks, which are now covered by rules that currently only cover rating movements. CRAs would have to disclose the time horizon in which the change in credit ratings will occur once they place an entity’s credit rating on watch or outlook for a possible rating change. The proposal also involves a number of more targeted regulatory proposals, outlined below. Problem: An overreliance on credit ratings, leading to “cliff” effects in financial markets.52 Proposed solutions: • Reduce reliance on large CRAs by promoting alternatives, such as internal ratings models, and require financials to compare external ratings with internal credit analysis and choose the least favorable assessment. Financial institutions and EU legislative GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 23

23

8/30/2012 9:31:19 AM

b1423

•

Global Credit Review Volume 2

and regulatory bodies are required to make their own assessments of credit risk under this proposal, and not rely completely or mechanically on credit ratings when assessing creditworthiness. Many large institutional investors already perform similar analysis when making investment decisions. Require increased information disclosures from companies offering structured products, which would allow financial entities and institutional investors to perform their own risk assessments, and enable more accurate unsolicited credit ratings. This proposal is similar to the formerly discussed Rule 17g-5 of the US Securities Exchange Act introduced in July 2010, under which issuers are required to make information about structured products disclosed to one NRSRO available to all NRSROs. The European Parliament (EP) approved these measures on June 19 and inserted an amendment that requires the removal of regulatory dependencies on credit ratings from EU laws. The EP also mandated that financial institutions regulated in the EU will be prohibited from automatically selling assets in response to credit rating downgrades.53

Problem: Significant contagion effects caused by changes in sovereign debt ratings.

•

24

1250002.indd 24

Proposed solutions: Mandate that CRAs publish ratings changes after close of EU trading, to minimize market disruption and ensure all investors have sufficient time to react to ratings movements. This proposal was broadly motivated by a perceived mistiming of sovereign debt ratings during the height of the European Sovereign Debt crisis. CRAs will have to inform a sovereign entity of any impending rating or outlook change during working hours, a full working day before publication. This would give a sovereign entity sufficient time to assess the validity of the rating movement and present any objections to the relevant ratings firm. CRAs will also have to publish full research reports on sovereign ratings changes, and will have to disclose the quality and any limitations on the accuracy of data used in determining sovereign ratings alongside such publications. The European

•

•

30 August 2012 9:03 AM

Parliament approved this proposal in full on June 19 and inserted amendments requiring each CRA to prepare and publish an annual timetable for the release of sovereign ratings. Require sovereign debt ratings to be reviewed on a more frequent basis, at least every six months, to help eliminate the lag between major political and economic developments and ratings changes. Extend ESMA power to scrutinize rating methodologies, and vest ESMA with the power to restrict or ban sovereign debt ratings changes temporarily in exceptional circumstances. The EC placed this proposal on hold in late November, claiming it requires more technical work, after much criticism from market participants.54 For example, Moody’s claimed that a restriction on ratings changes for any sovereign by ESMA would cause a panicked sell off in that nation’s debt and severely disrupt capital markets.55 Defending the proposal, Michel Barnier said the rule was aimed at preventing CRAs from announcing sovereign ratings changes during the negotiations of international aid programs. Downgrades of the Greek sovereign rating by the Big Three CRAs during bailout discussions in July 2011 increased market volatility significantly, increasing the yield on Greek government debt significantly and compromising the negotiation effort.

Problem: Limited choice and competition in the credit rating market. Proposed solutions: • Encourage the emergence of small and medium CRAs, and promote greater networking between these institutions, to facilitate the emergence of new CRAs that can effectively compete with the Big Three US-based CRAs. More competition may cause larger CRAs to compete on price and informational quality, which may partly address conflict of interest problems inherent to the issuer-pays model. Moreover, a number of smaller CRAs with certain specialties already compete with the Big Three and are the preferred choice for certain issuers and transactions. A.M Best remains the primary provider of credit ratings for

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:19 AM

b1423

•

•

Global Credit Review Volume 2

insurance companies, while credit ratings from the Japan Credit Rating Agency or Rating and Investment Information are almost always required when a company issues yen-denominated securities. In May, a spokesman for European banks suggested CRAs should no longer receive privileged information. A standardised package would place all providers of credit opinion on equal footing. Require CRAs to disclose pricing of ratings beyond the disclosures in annual transparency reports, which are required under Regulation 1060/2009. CRAs would be required to provide aggregate information on actual fees charged for different asset classes. This would allow issuers to more easily compare the difference in prices for ratings and ancillary services between CRAs. CRAs would also be required to ensure that fees are not discriminatory, charging all issuers a similar price for identical services. Oversight in this area by ESMA would help eliminate the risk of hidden pricing practices which are not in the interest of investors and help eliminate conflict of interest problems. The EC stated that such regulation would need to be carefully drafted to avoid collusion between CRAs. Harmonize ratings scales to allow investors to compare ratings between CRAs more easily and increase investors’ understanding of ratings. Under this proposition, all CRAs would need to make disclosures explaining the correspondence between current ratings scales and the harmonized ratings scale. The European Association of Corporate Treasurers (EACT) believes that this could reduce the informational content of current ratings, due to differences in methodologies between CRAs and the meaning of ratings between CRAs,56 as discussed in Section I. Furthermore, CRAs also publish distinctly different types of ratings for companies and sovereigns. For instance, banks receive separate ratings for their deposits and bonds.

Despite market participants and politicians calling for the EU to sponsor a new CRA to compete with the

30 August 2012 9:03 AM

Big Three, the EC deemed such a move ineffective due to high start-up costs and concerns that an EU-backed CRA would distort competition due to the public nature of the proposed agency. A compromise between legislators will see EU institutions carry out their own “creditworthiness assessments” of sovereign nations. On March 22, 2012, Moody’s released a research paper written by Moody’s CEO, Ray McDaniel, suggesting that governments concerned about sovereign downgrades should establish competing firms.57 He opined that several governmental bodies around the world already have both the analytical expertise and recognition to offer views on sovereign credit risk; the only obstacle is political will. A public sector voice in credit markets would stimulate debate amongst credit analysts and is the best alternative to stifling CRAs’ opinion through regulation, according to Ray McDaniel. Instead of establishing an EU-backed CRA, the EU may require all CRAs to communicate each newly issued rating to a central party, such as ESMA, charged with publishing ratings and related information in the form of a European Rating Index (EURIX). Information would be freely available to all investors through an online portal and would help small and medium CRAs gain visibility and reputation over time. The EACT also expressed concern about the EURIX proposal, as it would reduce diversity of opinion represented by the current ratings system. The EC proposed that ratings from different CRAs would be published alongside the average EURIX rating, and CRAs would still be able to release ratings through their own websites. Despite this, the EACT believes the clout EURIX may have over non-institutional investors could overshadow the differing informational content of ratings from the various CRAs. Problem: Insufficient rights of redress for users of ratings suffering losses due to a CRA issuing an inaccurate rating that infringes the EU CRA Regulation. Proposed solutions: • Increase the civil liability of CRAs towards end users in national courts where a CRA breaches GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 25

25

8/30/2012 9:31:19 AM

b1423

Global Credit Review Volume 2

the EU CRA Regulations. Any gross breach of these regulations that has an impact on rating outcomes would allow investors to seek damages against the infringing CRA. However, the burden of proof of damages in such cases would still rest with investors. Investors with the capability to complete their own risk assessments would also have redress under these changes. Michel Barnier indicated that S&P’s miscommunication to a small group of investors of an impending downgrade of the French sovereign rating on November 10, 2011 was an example of serious misconduct or gross negligence that could allow investors to seek damages from a CRA.58 The yield of French 10-year bonds increased 27 bps in trading the same day, and the miscommunication fuelled speculation that S&P was preparing to downgrade the French sovereign in the near future. In the weeks that followed, the yield spread between 10-year French government bonds and 10-year German bunds reached a record high of 190 bps. CRAs argued against policy changes in this area, suggesting that this would increase reliance on credit ratings, as investors would have increased sources of redress if their investments did not perform. The EACT believes these changes would increase barriers to entry for new CRAs,56 as the insurance costs and higher capital levels needed to offer ratings under the proposed regulatory environment could become prohibitive. CRAs expressed concern that this would result in an increase in frivolous claims from investors. Despite these objections, the EC believes that if there were no policy change in this area, there would be an inconsistent level of liability for CRAs between Member States, and CRAs would seek the application of laws in Member States where their civil liabilities are smaller. The European Parliament approved and clarified this proposal on June 19, 2012, requiring that investors seek redress under the civil law of their country of residence if losses arise from a CRA’s infringement of the EU CRA Regulation. Problem: Potentially undermined independence of CRAs due to conflicts of interest arising from the issuer-pays model, ownership structure and long tenure of the same CRA. 26

1250002.indd 26

30 August 2012 9:03 AM

Proposed solutions: Prevent CRAs from issuing or maintaining credit rating where persons who have significant influence over the CRA also have a significant interest in an entity that requests a rating from the CRA, unless the CRA undertakes significant public disclosures. Furthermore, such persons should not provide consultancy and advisory services to a rated entity they have a significant interest in. CRAs will also be prohibited from compensating and evaluating an employee involved in the rating process by the amount of revenue derived from rating activities performed by that employee. • To ensure continual independence in the rotation process, and prevent conflicts of interest, any member or shareholder who holds more than 5% of a CRA would be prevented from holding 5% or more in any other CRA, unless the CRAs are part of the same corporate group. This should prevent further consolidation within the CRA industry in Europe; the wave of mergers and acquisitions between 1997 and 2000 that allowed Fitch to become the third largest global CRA markedly reduced competition in the industry. • CRAs will also be required to rotate lead analysts, so that analysts are not involved in rating the same entity for more than four years. This rule is also aimed at preventing senior analysts from moving to a different CRA and rating the same entity for successive three-year periods. Interestingly, CRAs are required to perform similar rotations of junior analysts under current rules, but CRAs can apply for an exemption to this rule if a subsidiary has less than 50 staff, or a large dependence on a certain language skill. For example, five of Fitch’s European subsidiaries are currently exempted from rotation requirements under current rules.59 • Issuers will be required to change the CRAs they engage with every three years, or every year if an issuer has ten consecutive issuances rated by the same CRA in one year. Issuers of products that require ratings from two CRAs, either by law or corporate strategy, may retain one of the agencies for a maximum of six years. Once an issuer is required to change the CRA they engage, they will •

AN IMPROVED REGULATORY FRAMEWORK FOR CREDIT RATING AGENCIES?

8/30/2012 9:31:19 AM

b1423

Global Credit Review Volume 2

not be able to engage the same CRA for a paid rating the next four years, and the previous CRA must pass a handover file containing relevant rating information on the issuer to the incoming CRA. This rotation policy would not apply to sovereign ratings, a majority of which are unsolicited. CRAs will still be able to capitalize on their knowledge to produce unsolicited ratings on entities they are prevented from rating. But market participants fear that ratings from CRAs prescribed by the rotation will receive the greatest attention in the media, as they will have the greatest access to non-public company specific information. Fitch expressed concern that the rotation would further entrench the dominance of Moody’s and S&P, as companies may choose to retain only one of these two agencies for the 6-year period. Fitch is uniquely positioned to comment on this issue, as the firm has invested over $1bn during the past 20 years establishing itself has a viable alternative to Moody’s and S&P.60 Fitch also believes issuers would select the ratings firm it engages for the secondary rating based on rating results, compounding the ratings shopping problem. Fitch included a diagram of how it believes the rotation mechanism would work in their commentary paper (see Table 2).61 The diagram shows how smaller CRAs would be marginalized by the rotation mechanism, as issuers attempt to maximize the period in which they are rated by S&P and Moody’s. One of the more vocal proponents of the rotation mechanism was DBRS, which believes that mandatory CRA rotation by issuers will increase the amount and diversity of credit rating opinions and reduce what DBRS views as concentration risk in the industry.62 Members of the UK Treasury Select Committee suggested DBRS may be motivated by the larger market share it would receive under the rotation proposal. DBRS defended its support of the rotation mechanism by outlining its long-term plans to provide an alternative choice to the Big Three CRAs; the rotation mechanism is simply a method that would allow DBRS to fulfil that goal. Overall, market participants expressed concerns similar to those about the Al Franken amendment

30 August 2012 9:03 AM

currently under consideration in the US. The Al Franken amendment, discussed in Section I, would assign the rating of structured financial products to CRAs on a quasi-random basis. The rotation mechanism proposed by the EC would cover all ratings. The primary concern regarding the two regulatory mechanisms is the degree to which implementation would affect the efficiency of capital markets, as issuers could have trouble selling or face higher yields on bonds without ratings from the Big Three CRAs. Another key concern is the artificial market volatility the rotation mechanism could create. Ratings could change whenever an issuer is forced to rotate between CRAs, due to differences in methodologies and perspectives amongst CRAs, increasing instability in European and global capital markets. In addition, market participants and CRAs believe that the rotation mechanism would disadvantage EU-based issuers in global capital markets, as EU-based issuers may not always have two ratings from CRAs that are globally recognized. Following significant criticism of the rotation proposal, the European Parliament voted to scrap most of the rotation proposal on June 19. The rotation plan will be limited to re-securitised debt, such as collateralized debt obligations (CDOs), and issuers will only be required to rotate CRAs every five years.63 If an issuer sources ratings from more than two CRAs for an issuance, the rotation rule will not apply. Viability concerns were a key reason why the plan was toned down: The ECB and several EU finance ministers believe the sector currently lacks the critical mass necessary to adequately support a rotation mechanism for all types of ratings.64 As Fitch’s illustration in Table 2 shows, European issuers would need four additional CRAs with similar expertise to the Big Three in the first five years of operation if the full rotation mechanism was introduced at some point in the future. Problem: Insufficiently sound credit rating methodologies and processes. •

Proposed solutions: CRA 1 and 2 currently require CRAs to use credit rating methodologies that are rigorous, systematic, continuous, and subject to validation based on historical experience, including back-testing. Under proposed amendments, CRAs would need GLOBAL CREDIT REVIEW VOLUME 2

1250002.indd 27

27

8/30/2012 9:31:19 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:03 AM

Table 2. Rotation illustration. Rotation Illustration

Contracted Cooling off period

Infrequent Issuers ( PiCS*. If the floor price is binding (add *), ∆N1cCS* = PAC/PiCS*, P1c ≤ PiCS*.

(2a)

The lower stock-price limit has a high probability of being binding when conversion actually occurs. For example, CS issued cocos on February 24, 2011 with a floor price on conversion of USD20 per share. This was somewhat less than half the stock price, Pi, of USD46 on the date the cocos were issued. Since then CS has repeatedly closed under USD22 per share. However, its cocos were far from triggering in late 2011 because (CET1/RWA)% was in double digits rather than approaching 7%. Thus if these cocos were ever to trigger in the next five or ten years, they would do so most likely when the share price is well below USD20. The pre-specified minimum, PiCS*, then would be binding, arresting any death spiral.14 The second alternative to FR uses the stock market price at the time of cocos issue, Pi, as LBG has done. The number of shares to be issued in conversion under this method is: ∆N1cLBG = PAC/Pi. GLOBAL CREDIT REVIEW VOLUME 2

1250004.indd 59

(2)

(3) 59

8/30/2012 9:34:23 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:05 AM

Table 1. 52-wk low-high stock prices to 12/01/11, volatility (V) & Pi on date of cocos issue. Low Stock Price BOC.AT (€)

High Date

Price

When COCOS issued Date

Clos. Price

Date

0.5

11/30/2011

3.58

2/18/2011

2.21

5/18/2011

21.18

11/23/2011

47.63

2/18/2011

46.13

2/24/2011

LLOY.L (£)

0.2164

11/23/2011

0.9195a

2/25/2011

0.5414b

12/1/2009

LYG/4 ($)

0.3325

11/23/2011

1.12

2/18/2011

NA

NA

CS ($)

c

Implied V. d

Historical V.

BOC.AT

CS

LYG

NA

71.10%

86.80%

70.38%

48.80%

57.60%

Glossary BOC.AT (€) CS ($)

Bank of Cyprus Public Company Ltd., Athens Stock Exchange Credit Suisse Group AG ADS, 1 ADS rep. 1 Registered Share, NYSE

LLOY.L (£)

Lloyds Banking Group PLC, London Stock Exchange

LYG ($)

Lloyds Banking Group PLC ADR, 1 ADR rep. 4 Ordinary Shares, NYSE

Sources: Yahoo Finance (historical stock prices), Ameritrade (volatility). Notes: aThis intraday outlier was recorded before the NYSE opened on a high-volume day on the LSE on which the stock opened and closed at £0.63 in London. b The specified Pi calc. w. volume-weighted data for 11–17/11/09 was £0.592093 after being adjusted for a rights issue that took effect on 11/27/11. c For trailing 1-year volatility only one observation is refreshed per trading day. Implied volatility, however, is a function only of current option prices and much more variable. Example LYG: 77.2% reported 12/01, but 86.8% 12/02/11. d Historical or statistical volatility = standard deviation of log rates of change in daily closing prices, multiplied by square root of N = 252 for 12/01 2010 to 2011.

Both the floor price under the CS method, PiCS*, and the market price around the time of issue, Pi, applied under the LBG method, are announced or observed at the time of issue, years before the time of conversion: PiCS* so far has been set at about half of Pi. Just as the low of CS has already been more than 50% below its stock price at the time of cocos issue, the price of LBG’s stock (LLOY.L) has fallen steeply from Pi of 53.2093 pence on volume-weighted average for November 11–17, 2009, shortly before the date of cocos issue, to a closing low of 21.64 pence two years later as shown in Table 1. Hence CS and LBG cocos would have been triggered within 9 months (CS) to two years (LBG) if the trigger had been set at 50% of the price of their shares at the time of cocos issue.15 Wiping out the leading cocos so soon after origination would have defeated both their 60

1250004.indd 60

longer-term insurance purpose and undermined the ability to sell new cocos to investors. Price-based triggers would have been unsafe.

4.3. Redistribution from Conversion under the Different Methods Gains and losses from conversion are distributed between cocos holders, pre-existing shareholders, and other bond holders.16 In so far as cocos are issued as a substitute for senior unsecured regular debt and not for common equity, having go-cocos on the balance sheet to convert outside of bankruptcy will necessarily improve the position of the other bondholders by making their claims less risky. As between go-cocos holders and existing shareholders, better terms of conversion for one group must come at the expense of the other although

MEGA-BANKS’ SELF-INSURANCE WITH COCOS

8/30/2012 9:34:23 AM

b1423

Global Credit Review Volume 2

near the point of non-dilution both groups may still gain from conversion due to bankruptcy avoided. If P1cCS is the applicable conversion price, it is in theory equal to the market price, P1c , established after the conversion announcement. Small differences may arise from the practical need to determine P1cCS, and hence the number of shares to be issued in conversion, prior to the act of conversion itself though not before its announcement has hit the market. When conversion is at P1cCS, cocos holders obtain the full principal amount which may differ from the market value of their claim before there was a clear prospect of conversion. In the more likely event that P1c < PiCS* and the market price ends up below the floor price, cocos holders upon conversion get a value of less than PAC in the form of common shares. The outcome would be qualitatively the same if the stock market price after the announcement of conversion, P1c, were less than the stock price, Pi, recorded at the time when a cocos was issued under the LBG method of conversion. Indeed, empirically P1c so far has been less than half of Pi.17 Table 1 provides some useful indications of stock price relations for three cocos issuers and the volatility of their share prices. While application of the proposed FR method would keep P1cBV the same as P0cBV, book value of tangible common equity on the verge of conversion is likely to be about twice as high as market value per share for struggling banks that had issued cocos in better times.18 Observed outcomes therefore point to cocos holders salvaging 40% (LBG) to 50% (FR) of PAC from conversion. There are two possibilities for cocos holders to do better. Both arise under the CS method: (i) When the floor price is binding but considerably less than twice the market price at conversion, so that P1c < PiCS* 2P1c .

(4)

(b) The net value of tangible common equity per share at the time of conversion may come to be chosen as the preferred conversion price, particularly for long-term cocos. It marks the point of non-dilution when specified as the conversion price under the FR method. This book value per share before and after conversion tends to be around twice as high as the market price per share of common stock after mandatory conversion, so that: P0cBV = P1cBV ≈ 2P1c .

(5)

(c) The implication of equations (4) and (5) is that the conversion price specified under the LBG method is higher than the unchanging book value per share under the FR method of conversion which, in turn, is higher than the floor price under the CS method which is above the market price per share upon conversion: ∞ > Pi > P0cBV > PiCS* > P1c ≥ 0.

(6)

The limited experience with cocos available so far suggests that the LBG method will be anti-dilutive, the CS method with floor price dilutive, and the FR method neutral or non-dilutive. Cocos holders may expect to recoup less than half of the principal, say 40%, under the LBG method, about 50% under the FR method, and about 80% if conversion occurs under the CS* method. The latter percentage assumes that P1c post-conversion is on average about 20% below the minimum price per share set under the CS*conversion method and 60% below the conversion price set under the LBG method. Under all of these methods, cocos

30 August 2012 9:05 AM

holders may expect to receive stock valued at less than the face value of their claim at conversion though the extent of the shortfall may differ greatly. Write-downonly cocos, favored by UBS, are, of course, the most anti-dilutive of all, laying any of their buyers open to exploitation and excessive risk taking by existing shareholders.25 Such cocos can be represented as convertible at a share price of infinity in the scheme above (Berg and Kaserer, 2011), thus yielding no shares from conversion at all. If the market price of shares upon conversion fell to zero instead, cocos holders also would be left empty-handed but existing shareholders would share their lot.

9.2. CDS-Pricing of Cocos Using the recovery rates, R, of 0, 0.4, 0.5, and 0.8 as representative for the debt write-off only, LBG, FR, or CS* conversion method respectively, the problem of pricing cocos can be represented as putting a value on the respective single-asset Credit Default Swap (CDS) of the same reference entity. Such a CDS purchase at the price of fixed periodic premium payments conditional on survival of the cocos would free the cocos from exposure to losses from conversion provided mandatory conversion is identified as a credit event in the CDS contract. As a first approximation, the riskless rate plus the annualized premium payment could therefore be viewed as the appropriate market interest rate on cocos. Allowing for the likely illiquidity of cocos and the imperfect pricing of the credit default swaps on this reference object as well as counterparty risk would add to this constructed rate depending on the cocos issue in question. The objective is to compare the extra yield required on cocos due to the CDS premium with the risk premium required on common equity into which they turn upon conversion. Considering a 10-year cocos with an end-of-quarter fixed premium payment of K/4 conditional on survival through that quarter, a discount factor curve, DFt, and a survival curve representing the declining probability of survival through any of the t quarters, PSt, must still be specified to estimate the fixed quarterly premium payment. Such payments at the quoted annual rate of K, paid quarterly at K/4, cease from the GLOBAL CREDIT REVIEW VOLUME 2

1250004.indd 69

69

8/30/2012 9:34:23 AM

b1423

Global Credit Review Volume 2

quarter with the credit event, i.e., upon conversion. Their present value, conditional on survival of the cocos to the end of any of the 40 quarters, at par must be equal to that of (1−R) times the conditional probability of default, CDt, occurring during any of the t = 1 to 40 quarters.26 The par-value condition means that the present value of the premium leg is the same as that of the default leg so that there is no upfront payment associated with this exchange of contingent payments. Thus the equation to be solved for K with summation over all 40 quarters is: 40

0.25K ∑ (DFt PSt) t=1

40

= (1−R) ∑ (DFt CDt). t=1

(7)

The “riskless” discount curve DFi was estimated using the SNACDiscountCurve Function in the toolkit of the Cross-Assets Analytics Provider Numerix. It requires cash quotes and interest-rate swap quotes as inputs. The inputs used here are the 1-month, 3-month, and 6-month Eurodollar deposit rates and the 2- through 5-year, 7-year, and 10-year interest rate swap rates for February 1, 2012 from the Federal Reserve’s H.15 release on selected interest rates. To allow for sensitivity testing to the level of interest rates, the square of the resulting USD.SNAC.DF.curve values was taken. This implied roughly doubling all interest rates implied in the quarterly discount curve over the 10-year horizon. For instance, the discount factor after 20 quarters was 0.951466 and 0.905288 after squaring, implying annualized interest rates of 1.00% and 2.01%, respectively, while the discount factor after 40 quarters was 0.820425 and 0.673097 after squaring, implying interest rates of 2.00% and 4.04%. The specification of the cocos survival curves involves first a choice of the endpoints reached after 40 quarters. These were represented by the survival rates at maturity of 90%, 80%, 70%, or 60%. Cumulative default rates, in the present context actually cocos conversion rates, are the complement of the cocos survival percentages. For some background, global corporate issues initially rated B had an average cumulative default rate of 30% after 10 years while those initially rated BB, still below investment grade, experienced just under 20% default by that 70

1250004.indd 70

30 August 2012 9:05 AM

time (S&P, 2011, p. 4). Comparability is limited because go-cocos as opposed to goner-cocos are to help prevent default. Hence conversion rates on gococos should be higher than their participation rates in bankruptcy resolutions. S&P data more directly applicable to cocos show that the cumulative default rates on the issues initially rated below investment grade rise at a strongly decreasing rate as the time horizon lengthens, while the opposite pattern of increasing at an increasing rate applies to the investment-grade issues. In other words, seasoning is salutary for those entities which start out with poor ratings but survive the initial years of high mortality while the passage of time is detrimental to those which start out with top ratings and next to no defaults only to experience growing reversals later on. Equation 8 captures these different patterns of the survival curves through suitable assignment of values of s, such as 0.75 and 1.25 on either side of unity, in the exponent −λts. The decline to the specified endpoints at PS40 is fastest in the initial years with s = 0.75; it starts most slowly with s = 1.25. PSt[λ (PS40, s), s] = exp(−λts).

(8)

Four alternative values of the (conversion) hazard rate λ, which is also the first-quarter default rate in the CDS terminology here adopted, are calculated with each of the 3 values of s from equation (8) for (cocos) survival rates, PS40, of 60%, 70%, 80% or 90% at T = 40. Thus 12 values of λ are found in all. These range from a low of 0.001047 with a 90% survival endpoint and s = 1.25 to a high of 0.0321165 with a 60% endpoint and s = 0.75. Corresponding to the 12 survival curves that result there are 12 conditional default rates27 in any quarter t which are generated simply as: CDt = PSt-1 − PSt, where PS0 = 1 and t = 1 ... 40. (9) There is one dimension of sensitivity testing that can be dropped before turning to the presentation of results. Empirically it was found that changing the discount factors by squaring, as described, affected the values of K that satisfied equation (7) hardly at all. The intuition behind this finding is most easily explained for s = 1 in continuous time. The time derivative of the downward sloping survival curve (8)

MEGA-BANKS’ SELF-INSURANCE WITH COCOS

8/30/2012 9:34:23 AM

b1423

Global Credit Review Volume 2

with sign reversed is equal to the default rate conditional on survival at that point in time, and the slope of the default rate is equal to the second derivative of the survival curve with sign reversed. If s = 1, the result is that the slope of the survival curve used in the premium leg is simply −λPSt and the slope of the default curve used in the default leg on the other side of equation (7) is −λCDt. Since the time variation of these two elements that are multiplied by a discount factor on each side of equation (7) is the same, changing the discount factors that apply to them equally on both sides of this equation has symmetric multiplicative effects so that K does not change. If s≠1 the interest sensitivity of K is still very low. Hence the results that follow are all based on the single actual discount-factor curve, DFt, estimated with data for February 1, 2012. To establish the competitiveness of cocos with common equity, the CDS premium on cocos and the premium on common equity will both be measured as premium rates above the AAA (S&P and Fitch) or Aaa (Moody’s) corporate bond rate. The notes to Table 3 explain that before assessing the competitiveness of the premium rates for covering default risk, two adjustments to the equity premium are required if what remains of it is to be comparable to the estimated CDS risk premiums on cocos. First, when estimated over 10 years rather than 30 years, the equity premium of 5.5%28 is raised because the riskless rate falls by 25 bps when represented by 10- rather than 30-year U.S. Treasury bonds. Secondly the equity premium is lowered by about 105 bps on account of illiquidity and factors other than default risk which raise the interest rate on AAA-rated corporate debt (whose effective maturity is around 10 years) compared with 10-year Treasuries. The net result is an equity premium over the yield on AAA corporate bonds that is required to be about 4.7% to cover default-related risks alone which are specifically addressed by purchasing CDS for cocos.

9.3. Key Findings and Conclusions Table 3 shows that, except for debt write-off-only types of cocos issued with R = 0 by high-risk companies for which the cocos survival rate over 40 quarters is only 60%, the premium required on cocos is always less than 4.7% of the notional reference amount.

30 August 2012 9:05 AM

Table 3. The annualized fixed premium leg in % of notional for a par value CDS for combinations of structural parameters = 0.75, 1, or 1.25 for survival curves with endpoints PS40 of 0.90, 0.80, 0.70 or 0.60 with recovery rates R = 0, 40%, 50% or 80%. R=0

R = 0.4

R = 0.5

R = 0.8

s = 0.75

PS40 = 0.9 PS40 = 0.8 PS40 = 0.7 PS40 = 0.6

1.08 2.30 3.72 5.41

0.65 1.38 2.23 3.24

0.54 1.15 1.86 2.70

0.22 0.46 0.74 1.08

s = 1.00

PS40 = 0.9 PS40 = 0.8 PS40 = 0.7 PS40 = 0.6

1.06 2.24 3.58 5.14

0.63 1.34 2.15 3.08

0.53 1.12 1.79 2.57

0.21 0.45 0.72 1.03

s = 1.25

PS40 = 0.9 PS40 = 0.8 PS40 = 0.7 PS40 = 0.6

1.04 2.19 3.47 4.94

0.62 1.31 2.08 2.96

0.52 1.09 1.74 2.47

0.21 0.44 0.69 0.99

Notes: The Market Risk Premium (MRP) used in 2011 by finance professionals for estimating the required rate of return to equity is reported in Fernάndez, Aguirreamalloa, and Corres (2011) as averaging 5.5% with a standard deviation of 1.7% across estimates. In the United States this premium may be calculated as an addition to the yield on 30-year Treasury securities, viewed as free of default risk. The Required Equity Premium (REP) should not differ greatly from the Expected Equity Premium (EEP) over such a long horizon. What then does Table 3 suggest as an adequate yield on 10-year go-cocos and would that yield be competitive with common equity for which they become a perfect substitute when triggered? Two adjustments need to be made before the cocos risk premiums shown above can be evaluated. First the equity premium is a little wider when measured against 10-year, rather than 30-year Treasuries as the yield on 30-year Treasuries on average has been 25 bps (with a standard deviation of 0.38 bps) higher since February 1977. Secondly, even without taking account of default risk, as a debt instrument of perhaps severely limited liquidity 10-year go-cocos would still require a higher yield than 10-year Treasuries. Although bonds originally rated AAA by S&P are not entirely free from default risk, their 10-year cumulative default rate averages only 0.79 percent in their (2011, p. 31) cohorts. Hence almost the entire average spread of 105 bps found in the data described above between AAA-rated seasoned corporates and 10-year Treasuries is attributable to factors other than default risk that would also raise the yield required on cocos. To make the add-ons for loss-of-value risk commensurate and hence comparable

for common equity and go-cocos, the equity premium as well as the CDS premium on go-cocos are measured and interpreted as add-ons to the AAA/Aaa rate on corporate bonds. The entries in the 4 columns of results for different values of R are proportional to (1 – R). For instance, from the first to the third column, (1 – R) falls from 100% to 50%, reducing the results by half.

Lowering the finance experts’ estimate of the equity premium by one standard deviation, 1.7%, from its mean to 3.8% and again adding 25 bps and subtracting 105 bps, leaves 300 bps to compare with the CDS GLOBAL CREDIT REVIEW VOLUME 2

1250004.indd 71

71

8/30/2012 9:34:24 AM

b1423

Global Credit Review Volume 2

premium rates on cocos. Even if one took away another full percentage point by arguing that a premium in excess of 2% over the yield required on AAA-rated debt might not be tolerated by go-cocos issuers, all the combinations except those shown in italics in Table 3 would remain viable. Thus gococos, unlike goner-cocos that can be associated with R = 0,29 in most cases would be significantly less costly than issuing a supplementary common-equity buffer front-up. The rationale for making this further subtraction could be that 2% is still almost twice as high as the historical (geometric) average spread between Moody’s Baa- and Aaa-rated seasoned corporate bonds.30 Then issuing debt securities at more than 2 percentage points above the AAA or Aaa yield might put them at risk of being viewed as barely at or below investment-grade. At the same time the deleveraging of financial institutions in recent years may have lowered the required rate of return on equity by about 1 percentage point along Modigliani-Miller lines,31 producing an equivalent reduction in the equity premium that serves as comparator. Cocos, upon conversion, would themselves contribute to deleveraging. Overall Table 3 clearly shows that the higher the expected terminal survival rate PS40 and the greater the recovery rate R, the more qualified a financial institution would be to issue cocos at a premium over the AAA rate that would be well below that on common equity measured compatibly. Compared with these two parameters, PS40 and R, the market level and term structure of the discount factor curve and the assumed time structure of the survival curve, determined by s, hardly matter to the results in Table 3. The expected recovery rate is largely governed by the choice of the method of conversion, with recovery of 40%, 50%, and 80% of the principal value of cocos associated with the LBG, fair-rule, and CS methods of conversion, respectively. The terminal survival rate up to 10 years is a function of the amount of risk taken on by the firm compared with its ability to absorb losses. The latter is determined in part by the amount of go-cocos it has issued and the height of their triggers. These are the factors which cocos issuers principally should consider to assess the appropriate level of the interest margin over the 72

1250004.indd 72

30 August 2012 9:05 AM

AAA/Aaa rate on corporates to be offered on cocos to cover their conversion risk.

NOTES 1

2

3

4

5

To reduce the failure rate and to discourage further growth of banks that are already too big, such global banks will be required to hold extra capital initially of between 1% and 2.5% of Risk-Weighted Assets (RWA) under Basel III. For the details see von Furstenberg (2011a, p. 41 and 2011b, p. 14). The FSB (2011, Annex) in November 2011 named 29 financial institutions, including 17 headquartered in Europe, 8 in the United States, and 4 in East Asia as important to the world’s financial system. They were identified in BCBS (2011c, p. 5) as Global Systemically Important Banks, G-SIBs, by five equally-weighted indicators: Global activity across multiple jurisdictions, size, interconnectedness with the financial system, lack of substitutability (alternative supplies or suppliers) for some of the critical infrastructure services provided, and complexity of operations making them difficult to unwind. To avoid windfalls for the holders of lower-trigger cocos from the unexpected issuance of cocos with higher triggers, the latter should always be issued, or at least announced, before the former. Furthermore, the amount outstanding of each type of issue which the firm intends to maintain even after individual issues have matured should be outlined for investors as a matter of the firm’s policy. Lavishing windfalls on some classes of existing cocos holders denies existing shareholders the benefits of lower-cost financing that would otherwise have been available. Credit Suisse, but not Rabobank, has followed the correct sequencing in this regard. CS issued high-trigger (7%) cocos before turning to low-trigger (5%) cocos while Rabobank issued cocos with a higher trigger (8%) than the high-trigger (7%) cocos already outstanding. OSFI’s (2011) following assertion of unchecked powers is disconcerting in this regard: “Canadian authorities will retain full discretion to choose not to trigger [cocos] notwithstanding a determination by the Superintendent that a [Deposit Taking Institution] has ceased, or is about to cease, to be viable. Under such

MEGA-BANKS’ SELF-INSURANCE WITH COCOS

8/30/2012 9:34:24 AM

b1423

6

7

8

9

Global Credit Review Volume 2

circumstances, the DTI’s creditors and shareholders could be exposed to losses through the use of other resolution tools or in liquidation.” For the opposite choice — making “no recommendations” on go-cocos, or their design, while strongly advocating goner-cocos — see ICB (2011, pp. 102–103). The width of this range is due in part to the risk-weight, which is the ratio of RWA to total assets, TA, differing greatly between countries and institutions. For recent cocos issuers, RWA/TA at the end of the third quarter of 2011 was 37.8% for Lloyds Banking Group, only 19.8% for Credit Suisse, but 66.5% for the Bank of Cyprus Group. Such risk-weights alone are insufficient indicators of risk even for otherwise comparable banking institutions. For instance, UBS had a risk weight of only 11% (Basel II) on its total assets at the end of 2007 and shortly before, unlike CS, it needed to be bailed out by the Swiss government. On the other hand, as of 09/30/2011, Rabobank (2011, p. 19), which, like Nomura Holdings, is an issuer of the debt write-offonly type of cocos, had a 91.8% ratio of “risk-weighted exposure” to the principal amount of assets according to its consolidated financial statements. Yet Rabobank, an institution for banking cooperatives, has retained a top rating from all major credit agencies for years. The New York Times of January 13, 2012 notes the utter “lack of comprehension” evident from the transcripts of the 2006 meetings of the U.S. Federal Open Market Committee in this regard. The average monthly stock price at the open for the IYF iShares Dow Jones U.S. Financial Sector declined relentlessly from $114.50 in March 2007 to $82.07 in March 2008 and then to $32.13 in March 2009, a decline of 72%. If, instead of using regulatory capital ratios based on book values, some fraction of the trailing average monthly share price at the time of cocos issue had been specified as the trigger, the jointly suffered stock market convulsions typical of financial crises, and not the recapitalization needs of individual firms, would drive cocos conversion. Basing both the trigger- and the conversion-price on the latest market price of shares would give markets that are severely unsettled by a financial crisis even more capacity to rattle the financial system. Elsewhere in von Furstenberg (2011b, pp. 17–19) I have spelled out 8 other reasons for rejecting conversion triggers based

10

11

12

13

14

15

30 August 2012 9:05 AM

on stock prices or stock price indexes for the financial sector. The company announced on May 20, 2011 that only EUR 890 mn of the up to 1.342 bn Enhanced Capital Securities (CECS) authorized for Bank of Cyprus in March were actually issued. Taking advantage of the enhancement feature that combines the characteristics of a regular convertible bond with those of a contingent convertible issue, some of these CECS were converted by their holders into common shares during the first optional conversion period of September 1–15, 2011 at a share price of EUR 3.30 on the upside. Banks rightly are discouraged from investing in cocos by having to “deduct such investments from their Common Equity Tier 1 in accordance with the treatment of common stock investments under Basel III” (BCBS, 2011a, p. 26). von Furstenberg (2011b, p. 5) provides a record of such advocacy. See also Berg and Kaserer (2011). While a cap could equally well be placed directly on the maximum number of shares that may be issued in conversion, as the BCBS (2011a, p. 26) has proposed, all existing “CS” cocos specify a floor price instead. Death spirals are unlikely under the FR method of conversion because that method does not reference stock market prices at all. Short sales of shares that do not produce an increase in the number of shares outstanding or automatically raise funding and collateral requirements when stock prices fall may speed needed price adjustments but do not cause death spirals. Under the LBG and CS-min methods, the number of shares to be issued in conversion is fixed at the time of cocos issue except for an adjustment for dilution for instance through rights issues. Death spirals are conceivable only under the unconstrained CS method, a favorite with cocoskeptics which has never been applied in practice and should be avoided. De Spiegeleer and Schoutens (2011, pp. 26–28), using a method similar to the credit derivatives method, have associated an implied stock market trigger price with the LBG accounting trigger of 5% of core tier 1 capital. They estimated that implied price as 22.5 pence which is close to the market price of LBG in November 2011 and equal to 38% of the conversion price of 59.21 p. specified at the time of cocos issue (see Table 1). GLOBAL CREDIT REVIEW VOLUME 2

1250004.indd 73

73

8/30/2012 9:34:24 AM

b1423

16

17

18

19

20

74

1250004.indd 74

Global Credit Review Volume 2

Because switching to alternative conversion methods changes the distribution of net benefits, it is unhelpful to call for conversion to harm all stakeholders, “managers, shareholders and bondholders” for the sake of “enhancing market discipline” (Ötker-Robe et al., 2011, p. 14). It is difficult to see why cocos should ever be allowed to be issued if their conversion harmed all stakeholders instead of conveying net benefits on at least some of them. Cocos are recapitalization tools that are beneficially used in emergencies. Bank of Cyprus in effect imposed a minimum market price per share of EUR1.25 (80% of which is the minimum conversion price of EUR1) if the CECS it issued on 5/18/2011 should have to be converted. The market price of its shares two days before the cocos issue was €2.61. This price fell to less than €0.60 in November 2011 when conversion of cocos became imminent. Tables 1 and 2 of von Furstenberg (2011b, pp. 32–33) provide further details on the terms and conditions of large cocos issues. Taking two still rather challenged U.S. G-SIBs as examples, for Citigroup the book value (of tangible common equity) per share at the end of 2011:Q2 was $48.75 compared with a closing price of $41.64 on 06/30/11. Citigroup’s stock closed at $25 on 11/21/11. This was about half of its extrapolated book value per share at that time. For Bank of America the corresponding three values were $12.65 for book value per share and $10.96 and $5.49 for the two closing prices. Hence book value was again twice the market value per share on the latter (11/21/11) date. Bank of Cyprus, which actually was on the verge of converting about 70% of the €860.4 million cocos outstanding at the end of 2011:Q3, had a tangible net value of common equity per share of €1.70 when the closing price was €1.17. Conversion, if occurring at the end of the second quarter rather than at the time of the trigger event earlier, would have raised common equity from 5% of RWA to 7.8%, or by 56%, rather than from 7% of RWA to 9.8%, or by 40%. The ICB’s final report (2011, p. 86) states, “equity is by far and away the best form of loss-absorbing capacity,” without discussing these opposing arguments. The reasons given for the ICB’s conclusion are unconvincing. It claims, for instance, that more equity and less debt mitigate the debt overhang problem. But so do more

21

22

30 August 2012 9:05 AM

cocos and less other debt. It argues that equity cannot run away, but neither can long-term cocos, and it forgets that go-cocos, which it does not favor, turn into equity before anyone would want to run. It still has the specter of destabilizing death spirals (p. 102) cast a shadow over cocos, although such spirals would have to be based on irrationality and inefficient markets under all methods of conversion that have been applied so far. For cocos to be the reason why investors panic when a bank’s reported capital ratio sinks to a level that would require remedial attention with or without cocos would be particularly perverse because the conversion of gococos helps with timely recapitalization just when issuing new equity in other ways would prove difficult. In a crisis, individual banks have only the options of cutting costs and dividends and trying to reduce leverage and RWA if they cannot recapitalize through new stock issues from cocos conversion. Attempting to reduce the size and risk-weighting of their assets will be much more urgent if there are no go-cocos to convert. This makes a collapse in share prices in view of a reported decline in capital ratios much more likely without go-cocos than with them. Title 26 of the Internal Revenue Code, Subtitle A, Chapter 1, Sec. 163(k)(1) (1)–(3)(A)–(C) first denies the interest deduction for a disqualified debt instrument defined as “indebtedness of a corporation which is payable in equity of the issuer or a related party or equity held by the issuer.” However, Revenue Ruling 2002–31 then reminds that under Section 163(l) indebtedness shall be treated as payable in equity only if “a substantial amount of the principal or interest is required to be paid or converted, or at the option of the issuer or a related party is payable in, or convertible into, such equity” and “the indebtedness is part of an arrangement which is reasonably expected to result in a transaction described [above]” and that there is “substantial certainty” that any conversion options involved will be exercised. Since cocos conversion is neither unconditionally required nor optional, and since the triggering of conversion is a lowprobability event, cocos are not disqualified debt instruments as defined in the above Section. At least since 2004 there has been some disagreement on whether the non-contingent comparator should or could be fixed-rate debt that is convertible, rather than nonconvertible, into stock.

MEGA-BANKS’ SELF-INSURANCE WITH COCOS

8/30/2012 9:34:24 AM

b1423

23

24

25

26

27

Global Credit Review Volume 2

Zähres (2011, pp. 10–11) explains the difficulties rating agencies have had in rating cocos. Such a rating is necessary for a debt instrument to be included in a bond price index. Fitch (2009, p. 2) adopted a formula approach by which cocos generally are rated at least three notches below the issuer default rating, IDR. Thus differences in cocos triggers, conversion terms, and undergirding by higher-trigger cocos in the financing mix get short shrift. von Furstenberg has estimated elsewhere (2011b, pp. 13–16) that a cocos buffer equal to 3% of total assets, TA, would have been sufficient to recapitalize the largest U.S. banks on average during the 2007–2009 financial crisis. If RWA/TA = 0.3, the buffer percentage corresponding to 3% of TA would be 10% of RWA. Methodologically similar estimates, but in percent of RWA rather than TA, are given in BCBS (2010a). Logutenkova (2011) reported on 11/17/2011 that [the new] CEO Ermotti and CFO Naratil said that “[t]he bank [UBS] still prefers contingent instruments that are not dilutive to shareholders, such as bonds that are written down when triggered.” This leaves the false impressions that (1) write-down cocos are merely not dilutive, rather than strongly anti-dilutive, for existing shareholders and (2) cocos that actually convert to common equity are necessarily dilutive. It may be helpful to think of writedown cocos as equivalent to cocos that convert at an assumed share price of infinity, therefore yielding no shares at conversion, while cocos holders would come to own all but an infinitesimally small fraction of the shares outstanding upon conversion if the conversion price were assumed to be asymptotically equal to zero. Berg and Kaserer (2011, p. 6) have emphasized that, in the latter case, cocos holders can have a net gain from the adversity triggering a high-trigger conversion. They favor that method nonetheless even though it would turn seniority rules upside down and give cocos holders undue risk-taking incentives and control. The calculation is simplified by assuming that the time of valuation and default coincide with one of the quarterly payment dates on the CDS schedule and that the quarterly discount factor and survival rates stretching 10 years into the future are deterministic, with alternative parameter values assigned for sensitivity testing. The unconditional default and hazard rate at time t is defined as Dt = (St-1 − St)/St-1. However, it is that rate

28

29

30

31

30 August 2012 9:05 AM

conditional on survival, CDt = Dt ( St-1) = St-1 − St that is needed for present purposes since the CDS contract pays off the default leg in any quarter with the probability of the reference entity experiencing default in that quarter. “Over a longer [1995–2010] time horizon, financial firms have tended to achieve ROEs of 11% [for nonbank financials] −12% [for banks], which is close to the average for non-financial corporations” (BIS, 2011, p. 81). Subtracting 6.3%, which is the average yield on Moody’s Aaa-rated corporates for this period, leaves 4.7% for the former and 5.7% for the latter although Fernάndez, Aguirreamalloa and Corres (2011) rightly caution that even long-run returns on equity over some riskless rate may not simply be equated to the required equity premium. If R were treated as stochastic, zero recovery from equity obtained by conversion could also be identified as complete loss of equity value associated with bankruptcy. This is a possibility which even having go-cocos on the balance sheet cannot exclude entirely. However, only investors in goner-cocos should expect R = 0 as the most likely outcome from the start. Table 3 then shows that low-trigger gonercocos would be so expensive that there would be little if any demand for them. The condition R = 0 thus can be associated, but with very different implications, both with debt write-off-only cocos which can be placed only by financial institutions of the highest quality and with goner-cocos that bear a high risk of ending up worthless. The source of the historical Aaa and Baa corporate bond yields back to January 1919 is the Federal Reserve Bank of St. Louis via wikiposit.org/uid?FRED.AAA or BAA. Because the yield difference between Baa- and Aaa-rated bonds appeared to be lognormally, rather than normally, distributed, the geometric mean of 1.04 percentage point, rather than the arithmetic mean of 1.20% is reported in the text for monthly data through January 2012. While deducing an equity risk premium of 5.3% for a panel of 54 international banks, an unsigned contribution in the ECB’s Financial Stability Review of December 2011, pp. 128–130, finds substantial, though less than full, M-M effects ranging from 41% to 78% of the predicted full effect. GLOBAL CREDIT REVIEW VOLUME 2

1250004.indd 75

75

8/30/2012 9:34:24 AM

b1423

Global Credit Review Volume 2

REFERENCES Allen and Overy LLP (2011), Global Tax Practice: Tax Treatment of Additional Tier 1 Capital under Basel III. Bank of Cyprus Group (BOC) (2011), Interim Condensed Consolidated Financial Statements for the Nine Months ended 30 September 2011. Basel Committee on Banking Supervision (BCBS) (2010a), Calibrating Regulatory Minimum Capital Requirements and Capital Buffers: A Top-Down Approach. Bank for International Settlements (BIS), October. BCBS (2010b), Basel III: A Global Regulatory Framework for More Resilient Banks and Banking Systems. BIS, December, revised June 2011. BCBS (2011a), Global Systemically Important Banks: Assessment Methodology and the Additional Loss Absorbency Requirement. Consultative Document, BIS, July. BCBS (2011b), Basel III Definition of Capital — Frequently Asked Questions. BIS, October. BCBS (2011c), Global Systemically Important Banks: Assessment Methodology and the Additional Loss Absorbency Requirement. Rules Text, BIS, November. Berg, T. and C. Kaserer (2011), Convert-to-Surrender Bonds: A Proposal of How to Reduce Risk-Taking Incentives in the Banking System. Unpublished draft, February. BIS (2011), 81st Annual Report. Submitted in Basel on June 26. Calomiris, C.W. and R.J. Herring (2011), Why and How to Design a Contingent Convertible Debt Requirement. http://ssrn.com/abstract=1815406. Chant, J.F. (2011), Strengthening Bank Regulation: OSFI’s Contingent Capital Plan. Financial Services e-brief, C.D. Howe Institute, May 27. De Spiegeleer, J. and W. Schoutens (2011), Pricing Contingent Convertibles: A Derivative Approach. Department of Mathematics, Katholieke Universiteit Leuven, March 18. European Banking Authority (EBA) (2011a), EBA Recommendation on the Creation and Supervisory Oversight of Temporary Capital Buffers to Restore Market Confidence. London: EBA/REC/2011/1, London: December 8. EBA (2011b), 2011 EU Capital Exercise. 76

1250004.indd 76

30 August 2012 9:05 AM

Fernάndez, P., J. Aguirreamalloa and L. Corres (2011), US Market Risk Premium Used in 2011: A Survey with 5,731 Answers. IESE Working Paper WP-918, May. Financial Stability Board (FSB) (2011), Policy Measures to Address Systemically Important Financial Institutions, November 4. FINMA (2011), Addressing ‘Too Big to Fail’, the Swiss SIFI Policy, June 23. Fitch (2009), Rating Hybrid Securities. December. Goodhart, C.A.E. (2010), Are CoCos from Cloud CuckooLand? Central Banking, 21, pp. 29–33. Hammer, V. and J. Bush (2011), The Taxation of DoddFrank. Tax Notes, Special Report, July 11. Hammer, V., S. Chen and P. Carman (2011), United States: Tax Treatment of Contingent Convertible Bonds. Derivatives & Financial Instruments, May/June, pp. 97–106. Hymas, J. (2011), OSFI Targets Bond Investors. Advisor. CA, May 13. Independent Commission on Banking (ICB) (2011), Final Report: Recommendations. September. KPMG (2011), An Introduction to the Tax Implications of the Dodd-Frank Wall Street Reform and Consumer Protection Act. Washington National Tax and Americas’ FS Center of Excellence, May. Logutenkova, E. (2011), UBS Plans to Issue Contingent Capital, Chief Ermotti Says. Bloomberg, November 17. Ötker-Robe, I., A. Narain, A. Ihyina and J. Surti (2011), The Too-Important-to-Fail Conundrum: Impossible to Ignore and Difficult to Resolve. IMF Staff Discussion Note SDN 11/12, May 27. Office of Superintendent of Financial Institutions (OSFI) (2011), Advisory: Non-Viability Contingent Capital. August. Pazarbasioglu, C., J. Zhou, V. Le Leslé and M. Moore (2011), Contingent Capital: Economic Rationale and Design Features. IMF Staff Discussion Note SDN/11/01, January 25. PriceWaterhouseCoopers (2009), Debt Restructurings and Bankruptcy: Accounting, Tax and FAS 109 Considerations. Tax Accounting Services. Rabobank (2011), Disclosure Statement for the Nine Months Ended 30 September 2011. Rabobank Nederland. Standard & Poor’s (S&P) (2011), 2010 Annual Global Corporate Default Study and Rating Transitions.

MEGA-BANKS’ SELF-INSURANCE WITH COCOS

8/30/2012 9:34:24 AM

b1423

Global Credit Review Volume 2

U.S. Securities and Exchange Commission (SEC) (2002), Acceleration of Periodic Report Filing Dates and Disclosure Concerning Website Access to Reports. File No. S7-08-02. von Furstenberg, G.M. (2011a), Contingent Capital to Strengthen the Private Safety Net for Financial Institutions: Cocos to the Rescue? Discussion Paper Series 2: Banking and Financial Studies, No. 01/2011, Deutsche Bundesbank.

30 August 2012 9:05 AM

von Furstenberg, G.M. (2011b), Concocting Marketable Cocos. HKIMR Working Paper No. 22/2011, Hong Kong Institute for Monetary Research, July, http://ssrn. com/abstract=1895984. Zähres, M. (2011), Contingent Convertibles: Bank Bonds Take on a New Look. Deutsche Bank Research, Financial Market Special, EU Monitor, 79, May 23.

GLOBAL CREDIT REVIEW VOLUME 2

1250004.indd 77

77

8/30/2012 9:34:24 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:05 AM

This page intentionally left blank

1250004.indd 78

8/30/2012 9:34:24 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:05 AM

What are the Driving Factors Behind the Rise of Spreads and CDS of Eurozone Sovereign Bonds? A Panel VAR Analysis

INTRODUCTION

R Prof. Emmanuel Mamatzakis School of Business, Management and Economics, University of Sussex, UK [email protected]

Panos Remoundos Deputy Manager, Financial Products Sales and Head, International Markets, Alpha Bank, Athens [email protected]

ecent events of the eurozone sovereign debt crisis could look like a modern version of ancient tragedy. The eurozone sovereign debt crisis has its origin back in mid-2007, yet it was in December 2009 that it burst, leading spreads over Bund and Credit Default Swaps (CDS) to unprecedented levels. It all started with the Greek tragedy that slowly spread out to other eurozone countries, such as Ireland and Portugal. The 10-year Greek sovereign spread over Bund was 120 basis points (bps) in September 2009. Alas, Greek sovereign spread reached 400 bps by January 2010. In May 2010, the month Greece signed the memorandum of understanding regarding policy conditionality with the IMF, ECB and EU, the spread reached 1287 bps. Since then the Greek spread has climbed even further despite the Private Sector Involvement (PSI) in March 2012

that resulted in a large debt haircut of close to 75% in net present value terms (53.5% in nominal terms). The incident that signaled the eurozone sovereign debt crisis can be traced back to early summer of 2007, when the sub-prime crisis rattled global financial markets. However, the second half of 2007 was still early days as the 10-year sovereign bond yield of Irish bonds was actually lower than its German equivalent. In late 2008, right after the collapse of Lehman Brothers, Irish sovereign yields rates started climbing contrary to the German ones that stayed at low levels, as a result of investors’ reduced appetite for risk and preference towards quality sovereign bonds. The sovereign yield spreads between Greek and German bonds also increased during the same period, and in the third quarter of 2010, the 10-year Greek government bond spread surpassed the psychological level of 1000 basis points. Prior to 2008’s tectonic events, the market GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 79

79

8/30/2012 9:33:20 AM

b1423

Global Credit Review Volume 2

perception that a eurozone member could default1 was out of the question. But the stratospheric Greek spreads since late 2010 revealed the magnitude of the crisis and the market’s appreciation that Greece faced a very high risk of default. Recent episodes of the sovereign debt crisis in the eurozone show that the worst may not be behind us, as Spain and Italy have been facing rising costs of borrowing. Naturally, the European sovereign debt crisis is a manifestation of economic and market forces at work. To gain a deeper understanding of the transmission mechanism, we need to analyze the data using a well grounded methodology. Duffie and Singleton (1999), and Hull and White (2000b) provide evidence that there are arbitrage relationships between CDS and credit spreads for a given level of maturity. Moreover, Duffie (1999) was one of the first in the credit risk literature to suggest that arbitrage between CDS prices and credit spreads result in a cointegrated relationship. Similar findings for the difference between the CDS and EU corporate spreads are reported by Norden and Weber (2004) and Zhu (2006). In a similar way, Blanco et al. (2005) show that in the US there is a long run relationship between corporate bonds and CDS. They argue in further detail that there is a long run steady state relationship between CDS and corporate spreads in the US that in fact dictates a cointegrated relationship between the two.2 We contribute to the existing literature by applying panel-VAR methodology (P-VAR hereinafter) for the first time on CDS and sovereign spreads. The main advantage of P-VAR compared to a simple VAR is that the latter involves only the time dimension and thus lacks the cross sectional dimension that is crucial for credit risk analysis. As a result, the full information context of responses is lost. P-VAR on the other hand allows examination of the underlying responses of CDS and spreads of eurozone sovereign bonds to innovations in a plethora of variables across time and countries by identifying the full transmission mechanisms at play. To this end, contagion could also be taken into account. The results of our study suggest that liquidity, credit risk and flight to quality drive both bond spreads and CDS spreads of five years maturity for Greece, Spain, Portugal, Italy and Ireland in recent 80

1250005.indd 80

30 August 2012 9:05 AM

years. Moreover, in current illiquid market conditions spreads will continue to follow a steep upward trend with certain adverse financial stability implications, in addition to the negative feedback effect from counterparty credit risk through the balance sheets of the private and official financial sectors. For example, in March 2012 the PSI went through for Greek sovereign debt with substantial success if one considers the close to 96% participation. The outcome of the PSI was however detrimental to Greek commercial banks, requiring large recapitalisation, which was then provided by ECB, IMF and other eurozone member states as part of the second Greek bailout within a two-year period. The rest of this article is structured as follows: Section 1 presents a brief description of the debt crisis in the eurozone as well as an overview of the related literature, while Section 2 outlines some stylized facts of how the CDS market operates. Section 3 discusses the data and the methodology employed. Section 4 presents the results and finally Section 5 discusses the implications of the results.

I. RESEARCH MOTIVATION In this section we relate the timeline of the financial/ sovereign crisis with the evolution in credit spreads and eurozone yields and provide an overview of related academic studies.

1.1. Brief Description of Financial/ Sovereign Debt Crisis Timeline & Credit Spread Evolution From the beginning of the financial crisis until the end of 2011, five distinct phases of the financial/sovereign debt crisis can be identified. At the start of the second half of 2007 the subprime crisis was evident, as several US financial institutions (e.g., Bear Stearns, Countrywide Financial, New Century Financial) alerted the investment community with problems in their subprime loans portfolio or even filed for bankruptcy protection. The first phase, situated between June 2007 and September 2008, marked the stage of the

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:23 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:05 AM

The fourth phase of the financial/sovereign crisis lasted from October 2009 to September 2010. Since October 2009, rising distinctive sovereign risk had resulted in large differences among eurozone member states, with the yields on specific sovereign bonds approaching record highs. Following December 2009, the Greek sovereign debt crisis came to the forefront; an evolution that pushed yield spreads and CDS spreads to unprecedented levels (see Figure 2). And while initially it appeared as if the Greek tragedy would expand to other eurozone countries, markets soon seemed to perceive the Greek case as a unique one. As tensions surged again in early Q2 2010, European authorities along with the IMF created a financial rescue mechanism with a nominal firepower of €750 bn to tackle the crisis. Moreover, the ECB stepped into the secondary bond market and bought Greek, Irish and Portuguese government bonds in order to improve liquidity in the peripheral debt markets and stem the yields’ upward trend. This ECB operation was named the Securities Market Program (SMP). Finally, the last phase3 of the financial/sovereign debt crisis emerged with Ireland (in the fourth quarter of 2010) and Portugal (in spring 2011) applying for EU/IMF financial support, the private sector involvement negotiations for the adjustment

financial crisis build-up. Throughout this period, spreads remained within a relatively narrow, although widening, range before exploding on the event of the Lehman Brothers bankruptcy (15/9/2008). During the second phase, from October 2008 to March 2009, there were fears of systemic failure due to the collapse of Lehman Brothers, and sovereign spreads started diverging markedly. With the exception of German Bunds, eurozone government bond yields moved sharply above the 10-year swap yield, as problems in the banking sector spilled over to sovereign balance sheets. The spread between the three-month dollar Libor (or Euribor) and the overnight indexed swap rate (see Figure 1), a barometer of the reluctance of banks to lend, increased to the highest in history in the aftermath of the Lehman Brothers collapse but declined in the following months as a result of government responses globally. Moving to the third phase, in the period from April to September 2009, the global systemic response allowed credit spreads to converge, although at higher levels. As financial spillovers were contained and systemic risk subsided, sovereign bond yields returned back to the level of the 10-year swap yield, especially for those countries in which they had increased considerably in the previous phase.

3M LIBOR/OIS and EURIBOR/OIS Spreads

basis points

360 320 280 240

USD

200 160 120

EUR (solid line)

80 40

Figure 1.

1/12

10/11

7/11

4/11

1/11

10/10

7/10

4/10

1/10

10/09

7/09

4/09

1/09

10/08

7/08

4/08

1/08

10/07

7/07

4/07

0

Three month Libor/Euribor OIS spread evolution, 2007–2011.

Source: Bloomberg. GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 81

81

8/30/2012 9:33:23 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:05 AM

Greece only 1,200

10,000 9,000

1,000

8,000 7,000

800

6 000 6,000 5,000

600

4,000 400

3,000 2,000

200

1,000 1 000 0

Portugal Ireland Netherlands Greece (right hand scale)

Figure 2.

Spain Austria France

12/11

11/11

9/11

10/11

8/11

7/11

6/11

5/11

4/11

3/11

2/11

1/11

12/10

11/10

9/10

10/10

8/10

7/10

6/10

5/10

4/10

3/10

2/10

1/10

0

Italy Belgium Germany

CDS spreads evolution, 2010–2011.

Source: Bloomberg.

Figure 3.

ECB sovereign bond buying program, 2010–2011.

Source: Bloomberg.

of the Greek public debt burden and the extension of the sovereign crisis to core eurozone markets such as Italy, Spain, Belgium and France. As a result the ECB started buying massive quantities of Italian and Spanish government bonds in order to prevent yields from rising to unsustainable levels. Note that the ECB has bought €213 bn of eurozone country bonds from the secondary markets up to early January 2012, however with limited effectiveness in stemming the rise in bond yields (see Figure 3). 82

1250005.indd 82

The sovereign debt crisis was about to spin out of control. Political developments in late 2011 in Greece and Italy resulted in a grand coalition and technocratic governments respectively. Both governments were led by former EU officials, ECB’s Vice President Papademos in Athens and EU Commissioner Monti in Rome. The new governments have yet to prove that they have the ability to implement the necessary structural reforms, succeed in fiscal consolidation and bring their economies back to solid growth in order to restore investors’ confidence.

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:24 AM

b1423

Global Credit Review Volume 2

1.2. Sovereign Debt Crisis & Government Bond Yields Evolution in the Eurozone In July 2007, the yield on the 10-year maturity Irish sovereign bond was lower than the yield on a comparable German one. By early 2009 Irish government bond yields started to surge, German rates remained grounded if not falling as investors looked for safety. The yield spread between Irish and German 10-year bonds rose rapidly in January 2009 to over 200 bps from about 40 bps before Lehman Brothers collapsed, and exceeded 800 bps in June 2011. The size of the spread and, particularly, the underlying swing within a 30-month period was particularly visible for Ireland and Greece. However, since the second half of 2011, the Irish sovereign spread has been declining as opposed to the Greek spread that has remained at stratospheric highs. By now, spreads have increased substantially throughout the whole eurozone. Sovereign spreads are considered to be financial markets’ traditional measure of a country’s default risk. While even at the elevated levels, the perceived probability of default remains relatively low for most of the countries, markets are demanding higher risk premia for certain countries (see Table 1). These aforementioned developments have emerged after several years of tranquility in the sovereign bond markets of the eurozone. In the run-up to the establishment of the European monetary union, interest rates fell considerably, mainly in the peripheral countries. Whilst Spain and Italy had to pay more than 10% interest on 10-year bonds in the middle of the 1990s (see Figure 4), this rate approached 5% by 1999. Moreover, 10-year Greek government bond yield fell to 3% in the middle of the 2000s from approximately 8% in the late 1990s. It is also noteworthy that differences among eurozone sovereign bond yields were remarkably tight from 1999 till mid-2008. The collapse of eurozone governments’ borrowing costs in the first half of the 2000–2010 decade can mainly be attributed to investors’ perception that the ECB would remain committed to the ultimate task of keeping inflation at low levels and that the Stability Pact would enforce sound fiscal policies along with the belief that the eurozone

30 August 2012 9:05 AM

Table 1. European countries default probabilities. 5-year CDS Level (in basis points)

5-year Default Probability (%)

Belgium

331

25

France

227

18

Germany

108

9

Greece

8670

99

Ireland

715

46

Italy

520

37

Netherlands

128

11

Spain

430

31

Austria

225

18

As of 11/1/2012

Finland

83

7

1110

63

98

8

Denmark

135

11

Sweden

74

6

Portugal United Kingdom

Source: Bloomberg (recovery rate assumption: 40%)

countries would allow no default by any member state (Commerzbank, 2011). The IMF considered the stability and convergence of spreads a hallmark of successful financial integration within the eurozone. Currently, financial market participants consider the instability and divergence in yields that have been taking place since 2010 as threats to the foundations of the eurozone. The abovementioned increases in specific eurozone countries’ borrowing costs worsen their fiscal standing and raise debt sustainability issues (e.g., Italy, Portugal and Spain). Despite the fact that a high or unsustainable debt burden and fiscal vulnerabilities are common for countries that face an increasing debt servicing cost, there is no common ground on the appropriate remedy. The US prioritizes a growth stimulating approach through unsterilized money creation from the Federal Reserve’s printing machine and a mild medium-term fiscal retrenchment, whereas the eurozone countries’ response is a vigorous fiscal austerity approach accompanied by timid refinancing practices such as the European Financial Stability Facility (EFSF) and the European Financial Stability Mechanism (EFSM). The ECB launched additional GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 83

83

8/30/2012 9:33:24 AM

Global Credit Review Volume 2

30 August 2012 9:05 AM

%

b1423

34 31 28 25 22

Greece

Ireland

Portugal

19

Spain

Italy

German

16

France

13 10 7 4

Figure 4.

10/11

10/10

10/09

10/08

10/07

10/06

10/05

10/04

10/03

10/02

10/01

10/00

10/99

10/98

10/97

10/96

10/95

10/94

10/93

10/92

10/91

1

Eurozone countries 10-year bond yields evolution, 1991–2011.

Source: Bloomberg.

policy measures in order to improve conditions in the capital and money markets of the eurozone, like the Securities Market Program (SMP) in mid-2010 and the extension of the Long Term Refinancing Operations to 3 years in late 2011. The European Financial Stability Facility (EFSF), the European Financial Stability Mechanism (EFSM) and the IMF financial support created a €750 bn rescue fund with the intention to assist eurozone governments which faced problems in accessing sovereign debt markets (Deutsche Bank, 2010). These measures only helped to temporarily improve sentiment in eurozone sovereign debt markets.4 Since early 2010, eurozone sovereign yields have shown unparalleled volatility. In March 2009, the spread between the yield on a 10-year Greek government bond and the yield on a German Bund of equivalent maturity was 280 bps, by September 2009 it had receded close to 120 bps while in January 2010 it approached again 400 bps. In April 2010, the spread reached 670 bps only to climb even higher to the level of 1287 bps in May 2010, the month a joint initiative of the IMF, the EU Commission and the ECB was signed along with the memorandum of understanding regarding policy conditionality. The 10-year Greek government bond yield spread over bund surpassed the 1000 bps mark in early April 2011 before flying to the stratospheric level of 3360 bps by the end of December 2011. Likewise, the credit default swap spread, the 84

1250005.indd 84

premium investors are prepared to pay to insure the same sovereign Greek bond against a credit event, followed a trajectory similar to the spread. Note that although spreads of other eurozone countries also experienced an upward trend, no other sovereign bond spread has followed a similar trajectory as the Greek one.

1.3. Literature Review Reinhart and Rogoff (2009), after studying the major financial episodes of the last 200 years, found that in the aftermath of financial crises, the real value of government debt bursts, rising an average of 86% in the major post–World War II episodes. The main cause of public debt explosion is not the widely cited costs of bailing out and recapitalizing the banking system, but the collapse in tax revenues that governments suffer in the wake of deep and prolonged output contractions, as well as the often ambitious countercyclical fiscal policies aimed at mitigating the downturn. Despite the aforementioned evidence, to this day there is limited evidence (e.g., Fontana and Scheicher, 2010) of what has happened in the market with reference to the eurozone sovereign debt crisis. Financial market theory advocates that CDS spreads and corporate bond spreads for the same entities are bound by no-arbitrage conditions (Fontana and Scheicher, 2010). Allowing for differences in liquidity, assuming that the maturity of the corporate debt equals that of

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:24 AM

b1423

Global Credit Review Volume 2

the CDS and that the CDS counterparty is a risk-free entity, an investor who obtains a corporate bond and in parallel purchases protection in the CDS market is hedged against the default of this particular firm. If the no-arbitrage assumption between the two markets holds, then the CDS should equal the corporate bond yield spread (Fontana and Scheicher, 2010). Norden and Weber (2004) and Zhu (2006) report a long-term relationship between the two credit markets for corporate European entities. The majority of previous studies verified that there exists a long run (cointegrating) relationship between the two markets for most of the cases examined. However, the abovementioned long run relationship between the two markets does not exclude the possibility of arbitrage opportunities in the short run. Levin et al. (2005) suggest that in the short run though, market frictions could result in non-zero CDS-bond spread basis. The authors argue that market frictions stem from systematic and idiosyncratic factors. Forte and Pena (2006) reveal that stock markets prelude in price discovery from either bond or CDS markets. Concentrating on the corporate CDS and bond markets, Doetz (2007) examines the price discovery in these two markets in a time-variant context. The results show that although the CDS market dictates the price discovery process, its contribution was reduced significantly in 2005 following General Motors and Ford’s downgrade to below investment grade. Fontana and Scheicher (2010) show evidence of ‘limits of arbitrage’ between bonds and CDS (see also Shleifer and Vishny, 1997) during the stormiest periods of the credit crunch (late 2008 onwards and spring 2010). Moreover, the authors argue that liquidity preference along with obstacles in seizing arbitrage opportunities would act in such a way so as to separate pricing in the CDS market and the sovereign bond market. Fontana and Scheicher (2010) point out that the crisis had a negative impact on both market and funding liquidity. Bhanot and Guo (2010) and Fontana (2010) show similar findings for the basis5 between corporate bond spreads and the corresponding CDS during the crisis. In general, many market sections also observed the breakdown of a rather stable pricing relationships prior to the crisis

30 August 2012 9:05 AM

(Mitchell and Pulvino, 2010 and Krishnamurty, 2010).

II. THE MARKET OF CREDIT DEFAULT SWAPS AND THE PRICING OF CREDIT RISK In this section we briefly describe how the credit default swap market operates and how it relates to credit spreads. Furthermore, we document in some detail the academic literature related to credit risk pricing (see Jarrow and Turnbull, 1995, Jarrow et al. 1997 and Duffie and Singleton, 1999). In a credit default swap transaction, the seller of the protection agrees to pay the default payment to the buyer of the protection in the case that a default event takes place before maturity of the contract. In the case that no default event takes place before maturity, the protection seller has no obligation to deliver a payment. The buyer of the protection pays a periodic premium, typically quarterly, to the seller over the life of the contract, or until default or maturity, whichever takes place first. The annualized fee is usually referred to as the credit default swap price. The default payment is either repayment at par against physical delivery of a reference asset (physical settlement) or the notional amount minus the post-default market value of the asset under consideration, determined by a credit event auction (cash settlement). Physical delivery as well as cash settlement is one of the two dominant forms of settlement in the market. Initially physical delivery used to be the settlement method of choice. In physical settlement, when a credit event occurs, the buyer delivers the reference asset to the seller, in return for which the seller pays the face value of the delivered asset to the buyer (Chouldhry, 2006). The contract may specify a number of alternative assets (called deliverable obligations) that the buyer can deliver. When more than one deliverable obligation is specified, the buyer will invariably deliver the cheapest asset on the list of eligible assets; this provides the concept of the cheapest-to-deliver (CTD) option, which is an embedded option afforded by the protection buyer. As a result of the rapid development of the CDS market, the notional amount of CDS contracts exceeded the notional amount GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 85

85

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

of deliverable bonds. This created problems on the implementation of the method and directed the industry to develop a cash settlement approach (Chernov, 2011). In the cash settlement option, the protection seller pays the buyer the difference between the nominal amount of the default swap and the final (market) value of the reference asset, determined by means of a credit event auction. This last value can be viewed as the recovery value of the asset6 (Carbori, 2011). In the latter option, an auction settlement process determines cash payouts following a credit event by factoring in the amount submitted for physical delivery. At the end of the auction, all CDS are cash settled. A CDS contract is considered to be triggered if during the period of the protection, an event that materially affects the cash flows of the reference debt obligation occurs. Credit events are strictly defined by a standard ISDA 2003 agreement. Major cases of a credit event include instances of a reference entity filing for bankruptcy, being dissolved or becoming insolvent. Other credit events can include failure to pay, restructuring, obligation acceleration, repudiation, and moratorium. The abovementioned events excluding restructuring are not contentious, although the evolving ISDA documentation has dropped events such as obligation default or acceleration and repudiation or moratorium in some jurisdictions since they have been deemed subsumed by events of bankruptcy and failure to pay. However, restructuring is considered the most controversial of credit events. The ISDA documentation defines the restructuring credit event as being triggered in cases that include reduction in the rate of interest or amount of principal payable (i.e. “haircut”); deferral of payment of interest or principal (which would include an extension of maturity of an outstanding obligation); subordination of the obligation; and change in the currency of payment to a currency that is not legal tender in a G7 country or a AAA-rated OECD country. Another important element of the definition of restructuring is that the event has to occur in a form that binds all holders of the “restructured” debt. The most important problem is that not all deliverable assets necessarily become due and payable should restructuring occur and it is conceivable that some deliverable 86

1250005.indd 86

30 August 2012 9:05 AM

obligations will be cheaper than others. This is likely to be particularly acute where deliverable assets include very long-dated or convertible bonds that often trade at a discount to shorter-dated straight bonds. As a result of the above, there will be a non-negligible probability of a restructuring that falls short of making all debt due and payable and where some obligations trade at a substantial discount to others, then a physically-settled CDS price also contains a cheapest-to-deliver option and is not a pure measure of credit risk. There is a large and growing literature on the pricing of credit risk, in which two approaches dominate. The first one is the structural model, which is based on the value of the firm and usually derived from Merton (1974). In this class of models, default occurs when the process describing the value of the firm hits a given boundary. Black and Cox (1976), Longstaff and Schwartz (1995) are two important references. Pierides (1997) employs structural models to the pricing of credit derivatives and also used reducedform or intensity-based models. He argues that the timing of default is better specified in terms of a hazard rate. Jarrow and Turnbull (1995), Jarrow et al. (1997) Duffie and Singleton (1999), Das and Sundaram (1998), Duffie (1999) and Hull and White (2000a, 2000b) propose reduced-form models for the pricing of credit derivatives. Earlier surveys on structural and reduced-form approaches are surveyed Lando (1998) and Schönbucher (2000). Credit risk pricing continues to attract the attention of researchers with many newer pricing models being proposed.

III. DATA AND METHODOLOGY 3.1. Data Selection We use weekly closing data for CDS spreads and generic sovereign bond yields collected from Bloomberg. Our sample period is 4 January 2008 to 23 December 2011, reflecting data availability. The series of concern is 5-year CDS denominated in US$ for Greece, Ireland, Italy, Portugal and Spain. We focus on the countries that are at the epicentre of the recent sovereign bond crisis and chose the 5-year tenor as it is the most liquid in CDS transactions and

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

we accordingly select government bond yields with the same maturity. The sovereign spread for Greece, Ireland, Italy, Portugal and Spain at time t (Srdt) is measured as the difference between secondary-market yield on the country’s 5-year bond and the swaps yield. We also include the CDS for 5-year maturity, which reflects the insurance premium paid by the market’s participants against default. The following factors are included to examine their impact on sovereign CDS spreads as well as on sovereign bond spreads. First, we consider factors that influence credit risk. As the Merton model (1974) suggests, changes in the risk-free rate are in general negatively related to credit spreads. A rising risk-free rate decreases the present value of the expected future cash flows of sovereign bonds and the price of the put option also decreases. As a euro-wide homogeneous proxy we use the Euribor 3-month rate. Second, we include the corporate CDS premium (iTraxx) as a measure of aggregate credit market developments. The iTraxx provides information regarding the Main Investment Grade index. The premium on this CDS index should also contain a proxy for investors’ overall appetite for credit risk. Third, we include a proxy of a country’s public debt. In structural models of sovereign credit risk (Gapen et al., 2005), leverage, defined as the ratio of sovereign debt to assets, is a major risk factor. This risk factor is also acknowledged in a fiscal policy perspective as the EU’s Stability and Growth Pact aims to cap a country’s total debt at 60% of its GDP. As a proxy we use a country’s total outstanding bonds relative to its GDP. This choice of variable is motivated by data availability, as the amount of bonds outstanding is available in Bloomberg on a monthly frequency. Outstanding bonds measure the impact of fiscal imbalances on the sovereign debt crisis. We expect a positive relation between the level of sovereign debt and CDS spreads. If the sovereign bonds market has an elastic demand, the size of sovereign debt reflects market liquidity because a larger bond market generally contributes to lower transaction costs. However, if overall supply of new issuance exceeds existing demand, then there could also be an adverse impact on bond market liquidity. We expect

30 August 2012 9:05 AM

the second effect to be primarily relevant for bond spreads. Fourth, we also include the spread between Euribor and Eurepo. This variable is expected to have a positive impact on the spreads. When the repo rate is lower than the Euribor, it is costly to implement a positive basis trade that implies short selling the underlying bond obtained via repurchase agreement and selling protection.

3.2. Methodological Framework of the P-VAR An important drawback of examining the underlying factors of the sovereign bond spreads and CDS using standard OLS is the endogeneity bias. A way of dealing with this criticism is to employ a dynamic panel data (DPD) analysis, which uses a GMM estimator. In the empirical section of this paper we report both simple panel estimation and DPD estimation. In addition, this study goes a step further and opts for a more flexible framework, which is the panel-VAR analysis. Essentially all our variables, discussed above, are entered as endogenous variables in the panel-VAR to address the causality among them. Another advantage of the panel-VAR is that it examines the underlying dynamic relationships compared with the static functional form of a standard fixed effects model. We examine the underlying causality between the basis (bsit), which is defined as the difference between sovereign spread for Greece, Ireland, Italy, Portugal and Spain at time t (Srdt) and the CDS at time t (CDSt), and market specific variables. In the first stage, a first order 4 × 4 panel-VAR model is chosen: Xit = µi + ΦXit−1 + ei,t, i = 1, …, N, t = 1, …, T

where Xit is a vector of four random variables (i counts for the ith country and takes values up to 5), including, ‘basis’ (bsit ), the corporate iTraxx (iTrit ), the market specific variable EURIBOR (EUborit ) and debt measured as outstanding bonds over GDP, (Dit). Thus, Φ is a 4 × 4 matrix of coefficients, µi is a vector of m individual effects and ei,t are iid residuals. GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 87

(1)

87

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

IV. EMPIRICAL RESULTS

The panel-VAR takes the following form:

4.1 A Simple Random Effect Panel Estimation

bsit = a10 + b11bsit −1 + b12 iTrit −1 + b13 Dit −1 + b14 EUborit −1 + e1i,t iTrit = a20 + b21bsit −1 + b22 iTrit −1 + b23 Dit −1 + b24 EUborit −1 + e2i,t

(2)

Dit = a30 + b31bsit −1 + b32 iTrit −1 + b33 Dit −1 + b34 EUborit −1 + e3i,t EUborit = a40 + b41bsit −1 + b42 iTrit −1 + b43 Dit −1 + b44 EUborit −1 + e4i,t The moving averages (MA) form of the above model sets bsit, iTrit, Dit and EUborit equal to a set of present and past residuals e1i; e2i; e3i; and e4i; from the panelVAR estimation.7

bsit = µ 10 + ∑ b11 j e1it − j + ∑ b12 j e2it − j j

j

+ ∑ b13 j e3it − j + ∑ b14 j e4it − j + e1i,t j

j

iTrit = µ 20 + ∑ b21 j e1it − j + ∑ b22 j e2it − j j

j

+ ∑ b23 j e3it − j + ∑ b24 j e4it − j + e2i,t j

(3)

j

Dit = µ 30 + ∑ b31 j e1it − j + ∑ b32 j e2it − j j

j

Prior to presenting the panel-VAR estimation, we report simple panel estimations. Table 2 reports some descriptive statistics of the main variables of our model. The high volatility of the variables for market conditions is striking as reflected in the high standard deviations in the Euribor 3M, Itraxx, and the spread between the Euribor and Eurepo. Note that following Fontana and Scheicher (2010) we opt for the Euribor 3M to account for the risk-free rate. We expect that the risk-free rate has a negative impact on spreads, as an increase in risk-free rate will decrease the present value of the expected future cash flows. To take into account market perception on credit risk, we use the iTraxx Main Investment Grade index. We use a measure to account for fiscal sustainability issues proxied by the total outstanding bonds relative as percentage to GDP. To this end, outstanding bonds measure the impact of fiscal imbalances on the basis. We employ outstanding bonds instead of government debt, because the former is recorded in high frequency in line with the other variables included in our model. Table 3 provides empirical evidence of a random effect regression of the basis on Euribor 3M, iTraxx outstanding bonds as a ratio to GDP and spread (defined as Euribor-Eurepo). Overall, the significance of the regression is rather poor with the impact of both the Euribor-Eurepo Table 2. Market conditions and sovereign debt in the eurozone. Outstanding Debt/GDP

+ ∑ b33 j e3it − j + ∑ b34 j e4it − j + e3i,t j

j

EUborit = µ 40 + ∑ b41 j e1it − j + ∑ b42 j e2it − j j

j

+ ∑ b43 j e3it − j + ∑ b44 j e4it − j + e4i,t j

j

Using the above panel-VAR, individual heterogeneity in the levels is ensured by introducing fixed effects in the model, denoted µi. 88

1250005.indd 88

30 August 2012 9:05 AM

Mean Standard Deviation

Euribor 3M

Spread

Itraxx GRE POR

ES

IT

IRL

2.018

57.34

117.085 0.703 0.426 0.294 0.495 0.254

1.58

35.776

35.323 0.012 0.005 0.002 0.003 0.021

Minimum

0.635

20.4

60.969 0.498 0.305 0.206 0.433 0.065

Maximum

5.381

184.3

216.868 0.989 0.607 0.416 0.632 0.598

Note: Spread is the difference between Euribor 3M and Eurepo, Itraxx counts for the corporate spread rate of EU main investment grade, the debt variable captures a country’s (namely Greece (GRE), Portugal (POR), Spain (ES), Italy (IT) and Ireland (IRL)) total outstanding bonds relative to its GDP.

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:25 AM

b1423

Table 3.

Euribor 3M

Global Credit Review Volume 2

30 August 2012 9:05 AM

Random effect panel regression for basis.

Table 4. Dynamic panel data regression for basis.

Coef.

Std. Err.

Z

P > |z|

Coef.

0.030304

0.054622

0.55

0.618

bs(-1)

−0.0564

Std. Err.

Z

P > |z|

0.2266

−0.25

0.803

Spread

0.140923

0.137924

1.02

0.382

Euribor 3M

−0.6612

0.3206

−2.06

0.033

iTraxx

−0.00013

0.000298

−0.44

0.687

Spread

−1.779

1.0854

−1.64

0.101

Debt

−0.02275

0.016005

−1.42

0.25

iTraxx

0.00024

0.0007

0.33

0.738

Con.

0.542409

0.108786

4.99

0.016

Debt

0.000023

0.00069

0.03

0.0973

Con

0.4654

0.1514

3.07

0.002

16.27

Prob > chi2

0.0227

2

R

0.4629

Note: The Random Effect GLS estimation is used and the sample covers the period Q1 2009 to Q4 2011. The regression of the basis is: bsit = α + β1EUbort + β2 (EUbor-Eurepo)t + β3iTr + β4D bs is the difference between the spread of 5-year maturity for sovereign debt and the CDS of similar maturity, Spread is the difference between Euribor and Eurepo with a maturity of 3 months, iTraxx corporate CDS premium (iTraxx), and finally debt is a country’s total outstanding bonds relative to its GDP. The sample includes the following countries: Greece, Ireland, Portugal, Spain and Italy.

spread and the Euribor 3M positive but not statistically significant. The impact of debt and iTraxx on the basis is negative, but again of no statistical significance.

4.2. Dynamic Panel Data Analysis A common criticism on simple random effect panel regression analysis refers to the static nature of the analysis and possible issues of endogeneity. To address this criticism we now employ a DPD analysis, which uses a GMM estimator (Arellano and Bover, 1995). Table 4 reports the empirical results of the DPD analysis. In contrast with the results of the static panel, the Euribor-Eurepo spread now has a negative impact on the basis. This implies that in the short run it could be rather costly to apply a negative basis trade, which is to buy sovereign bond and buy CDS. However, this impact is not statistically significant. Similar results are reported for the impact of the basis with one lag and Euribor 3M on the basis, but again the results are not statistically significant. On the other hand, the impact of debt on the basis is positive and statistically significant, implying that fiscal imbalances would raise the spread for sovereign debt in the eurozone.

Wald chi2(7)

Note: The Dynamic Panel Data regression is based on Arelano and Bover estimation and uses quarterly observations from Q1 2009 to Q4 2011. The regression equation takes the form: bsit = α + β1bsit−1 + β2EUbort + β3 (EUbor-Eurepo)t + β4iTr + β5D bs is the difference between the spread of 5-year maturity for sovereign debt and the CDS of similar maturity, Spread is the difference between Euribor and Eurepo with maturity of 3 months, iTraxx corporate CDS premium (iTraxx), and finally debt is country’s total outstanding bonds relative to its GDP. The sample includes the following countries: Greece, Ireland, Portugal, Spain and Italy.

The empirical evidence of the DPD analysis shows that there is a link between the basis and outstanding debt, which captures the impact of government debt. Moreover, government debt increases the basis and thus the cost of sovereign debt in the five member states of the eurozone, Greece, Portugal, Ireland, Spain and Italy. The reported results show that improving fiscal balances could play a role in the current sovereign debt crisis. Moreover, fiscal adjustment would improve markets’ expectations and close the gap in the basis. However, a drawback of both the static and the DPD analysis is that the statistical significance of the estimations is quite limited. In addition, both the static and dynamic panel analysis do not allow the disentangling of the impact of underlying shocks on the sovereign debt crisis of the eurozone. In the next section we report the results of the panel-VAR analysis.

4.3. Panel-VAR Estimations As a first step in the panel-VAR estimation we make a choice regarding the optimal lag order j for the righthand variables in the system of equations (Lutkepohl, 2006). We employ the Arellano-Bond GMM estimator for the lags of j = 1, 2 and 3.8 We select the lag order GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 89

89

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

of one based on the Akaike Information Criterion (AIC), confirmed also by the Arellano-Bond AR test. To test for autocorrelation, more lags are added. Also, the Sargan test reports that for lag order one, the null hypothesis cannot be rejected and thus the VAR model should be of order one. Note that the lag order of one reserves the degrees of freedom and preserves valuable information, given the low time frequency of some of the data. In addition, we run normality tests for the residuals using the Shapiro–Francia W-test and the results do not show violation of normality.9 The impulse response functions (IRF) derived from the unrestricted panel-VAR are reported in Figure 5. The plots show the response of each variable in the panel-VAR, ‘basis’, iTraxx (iTr), Euribor 3M (EUbor) and outstanding debt (D ), to its own innovation and to the innovations of the other variables. The first row shows the response of ‘basis’ (bs) on a one standard deviation shock in iTr, EUbor and D. It is clear from the graph that the response of 324.0969

‘basis’ to iTr is positive for most of the period, reaching a peak after two periods and converging towards equilibrium thereafter. Note that iTraxx (iTr) takes into account market perception over credit risk. Moreover, this variable is a credit spread that compensates investors for unexpected loss due to credit events. Thus, it is a measure of aggregate credit market developments and shows investors’ overall appetite for credit risk. The reported impact of iTr in Figure 5 clearly shows that one standard deviation shock in iTraxx positively affects the basis and this, in turn, feeds back into the sovereign debt crisis by further increasing sovereign debt spreads over time. Similarly, a shock either on EUbor and/or D asserts a positive impact on ‘basis’. Recall that we use as a measure of fiscal sustainability the total outstanding bonds, D. The IRFs show that fiscal imbalances increase the basis and contribute to higher spreads for sovereign euro debt. Also the impact of fiscal

85.3412

-1.8e+02

84.1151

-40.9899 0

response of bs to bs shock

s

-0.3043

response of EUbor to bs shock

s

6

0.3936

-1.1531

response of iTr to bs shock

s

6

s

response of D to bs shock

Figure 5.

response of D to EUbor shock

0

6

6

s

6

response of iTr to d shock 3.7e+09

-9.8e+08 s

6

-0.6622 s

1.5e+09

0

s

response of EUbor to d shock

response of iTr to iTr shock

-1.8e+09 0

0

0.3730

0

6

1.2e+09

-3.5e+09

6

response of EUbor to iTr shock

response of iTr to EUbor shock

2.5e+09

s

-0.1941 0

6

s

-0.1494 0

0.1731

-0.0078 0

6

0.0000

response of EUbor to EUbor shock

0.4541

s

response of bs to d shock

-0.0513 0

0

6

0.0000

0.0000 6

s

response of bs to iTr shock

0.2920

s

-1.5e+02 0

6

response of bs to EUbor shock

0.0000

0

167.8535

-17.3721 0

6

s

30 August 2012 9:05 AM

-2.6e+09 0

s

6

0

response of D to iTr shock

s

6

response of D to d shock

Impulse response function (IRF) for basis (bs), iTr, EUbor and D.

Note: bs counts for the basis, which is the difference between the spread and the CDS, EUbor is the Euribor 3M, iTr counts for the iTraxx and D is the outstanding debt.

90

1250005.indd 90

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

imbalances is bigger in magnitude compared to the impact of iTraxx and Euribor 3M. However, the impact of D has a very short term as it crosses the zero line within two periods and becomes insignificant thereafter. One should interpret this result with some caution, though it is evident that market conditions, measured, e.g., by iTraxx, have an impact on spreads and thus on the basis that persist over a longer time than the impact of fiscal imbalances. Of course, fiscal imbalances matter for the eurozone sovereign debt crisis, yet market conditions also play a role and could ultimately be the factor that dominates over time. Along these lines, we observe from Figure 5 that the impact of a one standard deviation shock on EUbor has a similar positive impact to iTraxx on the basis. Note that the coefficient of EUbor captures the impact of the risk-free rate on the basis. Note, the higher the EUbor, the higher the basis, and therefore the higher the spread for sovereign eurozone debt. Once more, market conditions, as measured by the risk-free rate, persist over a longer time than the impact of fiscal imbalances, whilst the response of the ‘basis’ to one standard deviation shock in the iTraxx (iTr) also lasts for a long period. Table 5 presents the variance decomposition (VDC) estimations. These results are consistent with the IRFs. More specifically, close to 1% of the forecast error variance of ‘basis’ after 10 years is explained by iTraxx. Note however that outstanding debt has the dominant contribution of close to 16%, in the variation of basis. Furthermore, Euribor 3M Table 5. Variance decompositions (VDCs) for 1 lag of basis, Itraxx, Euribor 3M and Debt. S

iTraxx

Euribor 3M

Debt

basis

10 0.776505

0.008763

0.055136

0.1595

iTraxx

10 0.056447

0.750145

0.120463

0.0729

0.390184

0.302408

0.1285

Euribor 3M 10

basis

0.17881

Debt

10 0.247488

0.03368

0.127624

0.5912

basis

20 0.775178

0.009923

0.055376

0.1595

itraxx

20

0.06233

0.720926

0.135091

0.0816

Euribor 3M 20 0.160606

0.433296

0.28055

0.1255

20 0.246447

0.037461

0.127977

0.5881

Debt

Note: S defines the periods ahead of VDCs.

30 August 2012 9:05 AM

explains 5.5% of the variation of basis efficiency. Overall, the VDC analysis confirms the importance of iTr to basis. During crises, banks are undercapitalised (Duffie, 2010) which leads to arbitrage opportunities. The recent credit crunch shows that illiquid markets contribute to high costs of holding sovereign bonds due to possible high haircuts (Mitchell and Pulvino, 2009). Deteriorating market liquidity would unavoidably lead to high sovereign bonds spreads and CDS. The above dynamics of the sovereign ‘basis’ as depicted by IRFs and VDCs could suggest that a shift towards pessimism has taken place during the crisis, which could also be the outcome of an ongoing process of unwinding fiscal imbalances that the eurozone member states fail to face. As it appears, enhancing fiscal adjustment could reverse this spiral, and thereby improve market expectations over the fiscal sustainability.

V. CONCLUSION Duffie (2010) argued that high CDS for corporate bonds in the US after the credit crunch persisted due to severe depletion of capital that in turn causes large distortions in arbitrage, and to a lesser extent, this high CDS was due to counterparty risk or default risk. In recent months we have witnessed, in addition to Greece and Ireland, also Spain and Italy to be within the cyclone’s reach. In the present conditions of debt crisis, markets are short of capital and banks are undercapitalised. This depletion of capital poses a great challenge for the eurozone sovereigns with large fiscal imbalances and in turn results in high cost to hold such sovereign bonds due to high haircut in the repo markets (Mitchell and Pulvino, 2009). Thus, deteriorating market’s liquidity appears to be the driving force behind high sovereign spreads and CDS. In this paper we show for the first time in the literature using a Panel-VAR analysis that the ‘basis’, the difference between the spread and the CDS of sovereign debt in certain countries of the eurozone, namely Greece, Portugal, Ireland, Italy and Spain, is positively affected by shocks both in fiscal imbalances and market conditions, proxied by iTraxx and Euribor. Impulse response functions show that the response of the ‘basis’ to shocks in market condition GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 91

91

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

shocks persist for longer periods compared to fiscal imbalances as measured outstanding debt, though variance decomposition of the ‘basis’ show that the latter has the larger contribution. Note, however, that the static and dynamic panel analysis show weak evidence in terms of statistical significance, indicating that further empirical analysis is warranted and that current results should be treated with caution. Market conditions have deteriorated in recent months, aggravating further the sovereign debt crisis in the eurozone. As Duffie (2010) argued in periods of credit crisis, liquidity is harder to obtain. In turn, illiquid market conditions have implications on financial stability. In this respect, there is a dangerous spiral. All these factors feed back to the ‘basis’. A better understanding of the basis dynamics should be helpful to the execution of monetary policy.

NOTES 1

2

3

92

1250005.indd 92

The introduction of the euro in 2001 resulted in convergence in the spreads of sovereign debts across the eurozone. Both economic and financial integration was supposed to be at play in the early days of the euro. However by the end of 2010 divergences across member states of the eurozone emerged. External shocks, namely the US credit crunch, had a detrimental effect in the widening of sovereign spreads in the eurozone. In the case of Greek sovereign debt crisis, domestic factors, other than common external factors, contributed to very high spreads. As a result, high sovereign spreads for some eurozone member states persist. CDS is an insurance premium that the holder of a corporate or sovereign debt pays against the event of default. To define a credit event and thus credit default event is not a very straightforward exercise. There are several events that could be recognized as default such as bankruptcy, failure to pay, obligation default or acceleration, repudiation or moratorium (for sovereign entities), restructuring. Based on the definition of the International Swaps and Derivatives Association (ISDA) default could also be a credit event in case the interest rate or the principal paid at maturity are lowered or delayed. This, in fact, is the recent case for the Greek sovereign debt. Our data covers the period till December 2011, thus we do not include a recent phase of the Greek debt crisis

4

5

6

7

8 9

30 August 2012 9:05 AM

such as the credit event of restructuring of sovereign debt in March 2012. We do not include the provision of approximately 1 trillion euro of long term (3 years) refinancing operations from the ECB, especially for Spain and Italy, since our sample covers the period till December 2011. The basis is measured as the difference between credit spread and the CDS for similar maturity. In a basis trade, it is typical that investors hedge by combining a bond position with a CDS trade so as to exploit price misalignments (Fontana and Scheicher, 2010). Note that for a notional value of 1 the seller pays the loss given default LGD = (1−RR), where RR is the recovery rate of the reference asset. Following Love and Zicchino (2006) all data are forward mean-differenced using the Helmert procedure. Standard errors of the impulse response functions are calculated and confidence intervals generated with Monte Carlo simulations. Results are available upon request. The results do not show violation of the normality. Panel-VAR results are available upon request.

REFERENCES Ackert, L. F. and M. D. Racine (1999), Stochastic Trends and Cointegration in the Market for Equities. Journal of Business and Economics Statistics, 51, pp. 133–143. Adrian, T., E. Etula and H. S. Shin (2010), Risk Appetite and Exchange Rates, NY Fed, Staff Report No. 361. Amato, J. and E. Remolona (2003), The Credit Spread Puzzle. BIS Quarterly Review, December, pp. 51–63. Arellano, M. and O. Bover (1995), Another Look at the Instrumental Variable Estimation of Error-Components Models. Journal of Econometrics, 68, pp. 29–51. Basel Committee on Banking Supervision (2006), International Convergence of Capital Measurement and Capital Standards: A Revised Framework, Comprehensive Version, June. Bhanot, K. and L. Guo (2010), Types of Liquidity and Limits to Arbitrage — The Case of Credit Default Swaps. Mimeo. Black, F. and J. C. Cox (1976), Valuing Corporate Securities: Some Effects of Bond Indenture Provisions. Journal of Finance, 31, pp. 351–367.

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

Beber, A., M. W. Brandt and K. A. Kavavejc (2009), Flightto-Quality or Flight-to-Liquidity? Evidence from the Eurozone Bond Market. Review of Financial Studies, 22, pp. 925–957. Berndt, A. and I. Obreja (2010), Decomposing European CDS Returns. Review of Finance, 14, pp. 189–233. Blanco, R., S. Brennan and I. W. Marsh (2005), An Empirical Analysis of the Dynamic Relationship between Investment-Grade Bonds and Credit Default Swaps. Journal of Finance, 60, pp. 2255–2281. Brenner, R. J. and K. F. Kroner (1995), Arbitrage, Cointegration, and Testing the Unbiasedness Hypothesis in Financial Markets. Journal of Financial and Quantitative Analysis, 30, pp. 23–42. Buiter, W. (2010), Sovereign Debt Problems in Advanced Industrial Countries. Citi Global Economics View, April. Campbell, J. and G. Taksler (2003), Equity Volatility and Corporate Bond Yields. Journal of Finance, 58, pp. 2321–2349. Capuano, C., J. Chan-Lau, G. Gasha, C. Medeiros, A. Santos and M. Souto (2009), Recent Advances in Credit Risk Modeling. IMF Working Paper 09/162. Carbori, A. (2011), The Sovereign Credit Default Swap Market. Banca D’ Italia Working Paper No. 821. Chernov M., A. Gorbenko and I. Makarov (2011), CDS Auctions, November, London School of Economics Working Paper. Choudhry M. (2006), The Credit Default Swap Basis. Bloomberg Professional. Commerzbank (2011), Economic Insight, November 23. Das, S. R. and R. K. Sundaram (1998), Of Smiles and Smirks: A Term-Structure Perspective. Stern School of Business Working Paper. Deutsche Bank (2010), Understanding the EFSF. Fixed Income Special Report, October. Dieckmann, S. and T. Plank (2010), Default Risk of Advanced Economies: An Empirical Analysis of Credit Default Swaps during the Financial Crisis. Mimeo, Wharton School. Doetz, N. (2007), Time-Varying Contributions by the Corporate Bond and CDS Markets to Credit Risk Price Discovery. Discussion Paper Series 2: Banking and Financial Studies, No. 08/2007. Duarte, J., F. J. Longstaff and F. Yu (2007), Risk and Return in Fixed-Income Arbitrage: Nickels in Front of

30 August 2012 9:05 AM

a Steamroller? Review of Financial Studies, 20, pp. 769–811. Duffie, D. (1999), Credit Swap Valuation. Financial Analysts’ Journal, 83, pp. 635–665. Duffie, D. (2010), Presidential Address: Asset Price Dynamics with Slow-Moving Capital. Journal of Finance, 65, pp. 1237–1267. Duffie, J. D. and K. J. Singleton (1999), Modeling Term Structures of Defaultable Bonds. Review of Financial Studies, 12, pp. 687–720. Ejsing, J. and W. Lemke (2010), The Janus-Headed Salvation: Sovereign and Bank Credit Risk Premia During 2008–09. ECB Working Paper 1127. Elton, E. J., M. J. Gruber, D. Agrawal and C. Mann (2001), Explaining the Rate Spread on Corporate Bonds. Journal of Finance, 56, pp. 247–277. Ericsson, J., K. Jacobs and R. Oviedo-Helfenberger (2009), The Determinants of Credit Default Swap Premia. Journal of Financial & Quantitative Analysis, 44, pp. 109–132. Fleckenstein, F., A. Longstaff and H. Lustig (2010), Why Does the Treasury Issue TIPS? The TIPS-Treasury Bond Puzzle. NBER Working Paper 16358. Fontana, A. (2010), The Persistent Negative CDSBond Basis during the 2007/08 Financial Crisis. Mimeo. Fontana, A. and M. Scheicher (2010), An Analysis of Euro Area Sovereign CDS and their Relation with Government Bonds. Working Paper Series 1271, European Central Bank. Forte, S. and I. Pena (2006), Credit Spreads: Theory and Evidence about the Information Content of Stocks, Bonds and CDS. Business Economics Working Papers wb063310, Universidad Carlos III, Departamento de Economía de la Empresa. Gapen, M., D. F. Gray, C. H. Lim and Y. Xiao (2005), Measuring and Analyzing Sovereign Risk with Contingent Claims. IMF Working Paper 2005/155. Giesecke, K., F. Longstaff, S. Schaefer and I. Strebulaev (2010), Corporate Bond Default Risk: A 150-Year Perspective. NBER Working Paper 15848. Gorton, G. and A. Metrick (2009), Haircuts. NBER Working Paper 15273. Haugh D., P. Ollivaud and D. Turner (2009), What Drives Sovereign Risk Premiums? An Analysis of Recent Evidence from the Euro Area. OECD Working Paper 718. GLOBAL CREDIT REVIEW VOLUME 2

1250005.indd 93

93

8/30/2012 9:33:25 AM

b1423

Global Credit Review Volume 2

Hull J., M. Predescu and A. White (2005), Bond Prices, Default Probabilities and Risk Premiums. Journal of Credit Risk, 1 / Spring, pp. 53–60. Hull, J. C. and A. White (2000a), Valuing Credit Default Swaps I: No Counterparty Default Risk. Journal of Derivatives, 8, pp. 29–40. Hull, J. C. and A. White (2000b), Valuing Credit Default Swaps II: Modeling Default Correlations. Journal of Derivatives 8, pp. 12–22. Jarrow, R., and S. Turnbull (1995), Pricing Derivatives on Financial Securities Subject to Credit Risk. Journal of Finance, 50, pp. 53–86. Jarrow, R., D. Lando and S. Turnbull (1997), A Markov Model for the Term Structure of Credit Risk Spreads. Review of Financial Studies, 10, pp. 481–523. JP Morgan (2009), Basis Handbook. JP Morgan credit research. Krishnamurthy, A. (2010), How Debt Markets have Malfunctioned in the Crisis. Journal of Economic Perspectives, 24, pp. 3–28. Krishnamurthy, A. and A. Vissing-Jorgensen (2009), The Aggregate Demand for Treasury Debt. Mimeo. Lando, D. (1998), On Cox Processes and CreditRisky Securities. Review of Derivatives Research, 2, pp. 99–120. Lesmond, D., J. Ogden and C. Trzcinka (1999), A New Estimate of Transactions Costs. Review of Financial Studies, 12, pp. 1113–1141. Levin, A., R. Perli and E. Zakrajsek (2005), The Determinants of Market Frictions in the Corporate Market. Computing in Economics and Finance. Longstaff, F., S. Mithal and E. Neis (2005), Corporate Yield Spreads: Default Risk or Liquidity? New Evidence from the Credit Default Swap Market. Journal of Finance, 60, pp. 2213–2253. Longstaff, F., J. Pan, L. Pedersen and K. Singleton (2008), How Sovereign is Sovereign Risk? Mimeo. Longstaff, F. J. and E. Schwartz (1995), A Simple Approach to Valuing Risky, Fixed and Floating Rate Debt. Journal of Finance, 50, pp. 789–821. Lütkepohl, H. (2006), Structural Vector Autoregressive Analysis for Cointegrated Variables. AStA Advances in Statistical Analysis, 90, pp. 78–88. Manganelli, S. and G. Wolswijk (2009), What Drives Spreads in the Euro Area Government Bond Market? Economic Policy, 24, April, pp. 191–240. 94

1250005.indd 94

30 August 2012 9:05 AM

Merton, R. C. (1974), On the Pricing of Corporate Debt: The Risk Structure of Interest Rates. Journal of Finance, 29, pp. 449–470. Mitchell, M. and T. Pulvino (2010), Arbitrage Crashes and the Speed of Capital. Mimeo. Mody, A. (2009), From Bear Sterns to Anglo Irish: How Euro-zone Sovereign Spreads Related to Financial Sector Vulnerability. IMF Working Paper 108. Norden, L and M. Weber (2004), Informational Efficiency of Credit Default Swap and Stock Markets: The Impact of Credit Rating Announcements. Journal of Banking & Finance, 28, pp. 2813–2843. Pan, J. and K. Singleton (2008), Default and Recovery Implicit in the Term Structure of Sovereign CDS Spreads. Journal of Finance, 63, pp. 2345–2384. Panetta, F., T. Faeh, G. Grande, C. Ho, M. King, A. Levy, F. Signoretti, M. Tabog and A. Zaghini (2009), An Assessment of Financial Sector Rescue Programmes. BIS Paper 48. Pierides, Y. A. (1997), The Pricing of Credit Risk Derivatives. Journal of Economic Dynamics and Control, 21, pp. 1579–1611. Raunig, B. and M. Scheicher (2009), Are Banks Different? Evidence from the CDS Market. OeNB Working Paper 152. Reinhart, C. and K. Rogoff (2009), The Aftermath of Financial Crisis. NBER Working Paper 14656. Schönbucher, P., (2000), A Tree Implementation of a Credit Spread Model for Credit Derivatives. Bonn Econ Discussion Papers bgse17_2001, University of Bonn, Germany. Sgherri, S. and E. Zoli (2009), Euro Area Sovereign Risk During the Crisis. IMF Working Paper 2009/22. Shleifer, A. and R. W. Vishny (1997), The Limits of Arbitrage. Journal of Finance, 52, pp. 35–55. Stulz, R. (2010), Credit Default Swaps and the Credit Crisis. Journal of Economic Perspectives, Winter: pp. 73–92. Upper, C., and T. Werne (2002), Tail Wags Dog? TimeVarying Information Shares in the Bund Market, Bundesbank, Working Paper 24/02. Zhu, H. (2006), An Empirical Comparison of Credit Spreads between the Bond Market and the Credit Default Swap Market. Journal of Financial Services Research, 29, pp. 211–235.

DRIVING FACTORS BEHIND THE RISE OF SPREADS AND CDS OF EUROZONE SOVEREIGN BONDS

8/30/2012 9:33:26 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:09 AM

Measuring Distance-to-Default for Financial and Non-Financial Firms INTRODUCTION

T Prof. Jin-Chuan Duan Risk Management Institute & Department of Finance, National University of Singapore [email protected]

Tao Wang Department of Finance, National University of Singapore [email protected]

Editorial Comment: Data for DTD* is available for all listed companies under the coverage of the Credit Research Initiative at the Risk Management Institute of NUS. They can be downloaded from www.rmicri@org.

his article reviews several empirical methodologies for estimating Distance-toDefault (DTD), a popular measure for gauging how far a limited-liability firm is away from default. We focus on the idea behind each method and discuss its strengths and weaknesses both conceptually and through the use of concrete examples. The methodological differences and implications are brought to the fore by analyzing several banks and insurance companies, which are typically of high financial leverage. We show that distortion can be substantial when an inappropriate estimation method is applied. DTD has been widely adopted by academics in financial research and extensively used in business applications by industry practitioners. The academic papers that use DTD are too numerous to list here whereas in the commercial applications, Moody’s KMV model is arguably the most prominent one. The precise definition of DTD depends on a theoretical model, particularly the seminal credit risk model of Merton (1974), which treats corporate debt as an option-like

financial instrument. Conceptually, a firm’s asset value evolves according to some stochastic dynamic and its debt will be honored when the asset value stays above the promised payment in some future time stipulated under the debt contract. Otherwise, this firm is in default and its debt holders can only recover a partial amount equal to what is left of the firm. Even though the future asset value of a firm cannot be known today, its current value serves as the natural base with which one can assess how likely the firm will default in the future. For example, when the current firm value is much higher than its promised future payments, the likelihood of default will be small simply because the firm has significant buffer to absorb losses in its asset value. This thinking underlies how we traditionally view corporate financial leverage, such as the debt-toasset ratio. A lower-leverage firm is expected to be more resilient to future losses. Since the asset value moves randomly due to external shocks, the leverage ratio alone cannot be good enough to adequately capture the notion of DTD. The trend and volatility of the asset value movements GLOBAL CREDIT REVIEW VOLUME 2

1250006.indd 95

95

8/30/2012 9:37:19 AM

b1423

Global Credit Review Volume 2

must play an important role in determining the likelihood of default, because the same level of buffer may not be sufficient to withstand potential losses when the firm’s asset value is highly volatile. Put simply, a good DTD must be a leverage ratio adjusted for trend and volatility of the firm’s asset value. We introduce the Merton (1974) model to define such a DTD. Although DTD is an appealing concept, it runs into two kinds of implementation challenges. Computing DTD requires knowing the market value of the firm’s assets and the parameters governing the asset value movements (trend and volatility). But the market value of a firm’s assets as postulated in the Merton (1974) model cannot be directly observed. Without a time series of observed asset values, it is obviously difficult to estimate the model parameters that define trend and volatility of the asset value movements. Different estimation methodologies have been proposed in the literature; for example, (1) the market value proxy method used in Brockman and Turtle (2003) and Eom, Helwege and Huang (2004), among others, (2) the volatility restriction method proposed by Jones, Mason and Rosenfeld (1984) and Ronn and Verma (1986), (3) the KMV iterative method described in Crosbie and Bohn (2003), and (4) the transformed-data maximum likelihood method by Duan (1994, 2000). We describe these methodologies and discuss their strengths and weaknesses with concrete examples. Specifically, we use financial firms to illustrate the limitation of the KMV estimation method. Because financial firms typically have a large proportion of liabilities that cannot be accounted for by the KMV estimation method (for example, policy obligations of an insurance company), the method tends to inflate asset volatility and cause a distortion to DTD. We argue that the maximum likelihood method proposed first by Duan (1994) and modified later by Duan (2010) and Duan et al. (2012) to deal with financial firms is the most appropriate and flexible method for estimating DTD. The second application challenge arises from applying DTD strictly according to the structural credit risk model as defined. The Merton (1974) model and many subsequent models along the same line are typically classified in the literature as structural credit risk models in contrast to reduced-form models. Strictly 96

1250006.indd 96

30 August 2012 9:09 AM

following the Merton (1974) model, one can obtain a firm’s default probability by directly applying the cumulative normal distribution to the negative of DTD. But the results from such a direct and consistent application of the structural model are at odds with empirical default rates. Academic researchers and industry practitioners have long realized that DTD is highly informative about defaults, but it must be used along with other variables to achieve good performance.1 Further calibration through a reduced-form model, such as logistic regression, is a must in practice. It is somewhat ironic to say that DTD, as a measure defined by a structural credit risk model, must be further calibrated by a reduced-form model to yield good empirical performance. How DTD can be intelligently applied is not within the scope of this article. We bring up this issue so that readers can have a general awareness of the limitation of applying DTD strictly in accordance with the Merton (1974) model even though DTD is built upon that model.

I. THE DISTANCE-TO-DEFAULT The Merton (1974) model assumes that firms are financed by equity, with its value at time t denoted by St, and one single pure discount bond (denoted by Dt) with maturity date T and principal F. The asset value Vt follows a geometric Brownian motion: dVt = µ Vt dt + σ Vt dWt .

(1)

Here, Wt is a standard Brownian motion. Due to limited liability, the equity value at maturity is ST = max (VT - F ,0). Therefore, the equity value at time t ≤ T by the Black-Scholes option pricing formula becomes

)

(

S (Vt , σ ) = Vt N (dt ) − e − r (T − t ) FN dt − σ T − t , (2)

where r is the instantaneous risk-free rate, N(.) is the standard normal cumulative distribution function, and

dt =

σ 2 V   ln  t  +  r + (T − t )  F  2  σ T −t

.

(3)

MEASURING DISTANCE-TO-DEFAULT FOR FINANCIAL AND NON-FINANCIAL FIRMS

8/30/2012 9:37:20 AM

b1423

Global Credit Review Volume 2

Following the Merton (1974) model, it can be shown that the probability of the company’s default at time T evaluated at time t is N (−DTDt) where the DTD at time t is defined as

DTDt =

σ 2 V   ln  t  +  µ − (T − t )  F  2  σ T −t

.

(4)

Since the standard normal distribution function is universal, the sole factor that determines the default probability is the DTD. As the formula suggests, DTD is the logarithm of the leverage ratio shifted by the expected return (µ − σ 2/ 2)(T − t), and scaled by the volatility σ T − t . Consider two firms with identical leverage ratios and volatilities, but the asset value of one is expected to increase at a faster rate than the other. We naturally expect the one with a higher expected return to be further away from default, i.e., have a larger DTD. If two firms have identical leverage ratios and expected returns, their volatilities will determine which one is farther away from default. It is evident that the conclusion depends on the sign of the numerator. If the numerator is positive, meaning that the asset value will cover the debt obligation on average, a lower volatility should make the firm less likely to default, and it indeed has a larger DTD. When the numerator is negative, the situation can be understood as the firm is on average not expected to meet its debt obligation in the future. A higher volatility will make DTD less negative, which is consistent with the intuition that the firm has a higher chance, due to a higher volatility, to get its future asset value to exceed the debt obligation. The default probability of interest is naturally the one under the probability law that governs the physical movement of the asset value; that is what the DTD formula in Equation (4) intends to capture. In the credit risk literature, the risk neutral probability often surfaces to describe a hypothetical scenario in which economic agents are risk neutral. Were economic agents indeed risk neutral, the expected return on financial investment would have to equal the risk-free rate of return. By implication, parameter µ in Equation (4) should be replaced by r under the hypothetical scenario. Computing the risk-neutral DTD and then the

30 August 2012 9:09 AM

risk-neutral default probability does not need the value of µ. Moreover, the volatility parameter, σ can be obtained by just calibrating a pricing model to the observed market price of debt and/or prices of some credit derivatives. Obviously, the risk-neutral default probability is easier to obtain. However, one should be mindful of the fact that in theory it is not the default probability to be physically experienced. Since it is the physical DTD that we are after in practice, we should have suitable estimates for both expected return and volatility. Due to the nature of diffusion models, however, parameter µ cannot be estimated with reasonable precision using high frequency data over a time span of several years, a well-known fact in the financial econometrics literature. The technical reason for this result is that parameter µ in Equation (1) is accompanied by a time factor of dt whereas parameter σ is by a time factor of dt which is implicit in dWt. Data sampled frequently is less informative about µ than σ, because dt is much smaller than dt when the value of dt is small. With the estimation precision issue in mind, it makes empirical sense to avoid using µ in the DTD estimation, particularly when DTD is only used as an input to a reduced-form model to be further calibrated to empirical default rates. Therefore, it may be advisable to use the following alternative form of DTD to reduce sampling errors:

DTD*t =

V  ln  t   F σ T −t

Note that DTD* amounts to setting µ = σ 2/ 2 in Equation (4), and its calculation does not require the value of µ.2 As shown later in this paper, estimated DTD* is much more stable than DTD.

II. ESTIMATION METHODS There are several difficulties in implementing the Merton (1974) model. First, the asset values are not directly observable, and therefore they are not available for plugging into the DTD formula directly even if the model parameters were readily available. Second, the parameters µ and σ governing the unobserved asset GLOBAL CREDIT REVIEW VOLUME 2

1250006.indd 97

(5)

97

8/30/2012 9:37:20 AM

b1423

Global Credit Review Volume 2

value process are unknown and need to be estimated. But their estimation becomes a serious challenge simply because the asset values are not directly observed. Several estimation methods are commonly applied in the finance literature and in business practice, but not enough attention is paid to their theoretical and empirical shortcomings. We introduce these methods through the use of concrete examples and discuss their limitations and shortcomings. Our discussions specifically touch upon a new methodological advancement for dealing with financial firms which factors in their somewhat unique liability structure. We use three different types of firms in our empirical illustration. They are IBM (a US industrial firm), Barclays (a British bank) and Tokio Marine (a Japanese insurance company). These three firms are used for general comparison of different estimation methods. Later, we will focus on financial firms to zero in on the difference between the KMV method and the transformed-data maximum likelihood method. For that, we use three banks (Bank of America, Barclays, and DBS) and three insurance companies (Sun Life, AXA SA and Tokio Marine) from different regions of the world so as to appreciate that the methodological impact is far reaching.

2.1. The Market Value Proxy Method It is easy to obtain the equity value of an exchange listed firm, but the same cannot be said about the asset value. A direct valuation of asset value is practically impossible, because a firm as a going concern presumably possesses intangible assets and their values are hard to determine. Adding together the market values of equity and debt to arrive at the market value of the firm makes sense conceptually, but the market value of debt is hard to come by because a typical firm will have a large portion of debt in some non-tradable forms. Thus, a hybrid approach of adding market capitalization (equity value) to the book value of liabilities became very popular in corporate finance literature. The papers using this market value proxy method to obtain firm value are too numerous to mention. If one adopts the market value proxy method, estimating the two parameters (µ and σ) becomes fairly straightforward. One can obtain a time series of, 98

1250006.indd 98

30 August 2012 9:09 AM

say, daily asset values by summing daily updated market capitalizations with quarterly updated book values of liabilities. With the time series in place, one can obtain the daily logarithmic asset returns and then compute the sample mean and standard deviation of the return as the estimates for µ and σ. Indeed, the market value proxy method has been used in the empirical credit risk literature as well, for example, Brockman and Turtle (2003) and Eom et al. (2004). However, the quality of the market value proxy method is questionable. It has been argued in Wong and Choi (2009) that such a method will produce an upward biased estimate of the asset value. In fact, the bias magnitude is directly related to how volatile the firm’s asset value is. The reason is not difficult to appreciate. When the value of a discount debt is artificially set to its par value, it has inflated the market value and the amount by which it has been inflated (the discount portion) increases with the firm’s volatility, by standard option pricing theory. Its impact on credit analysis has not, in our opinion, been fully appreciated in the literature. For example, the market value proxy makes the firm’s equity (as a call option) always in the money at the time of assessing credit risk. If other things are equal, the DTD will be biased upwards, making the estimated default probability smaller than it should be. In light of this, it is unclear as to how one should interpret the empirical findings in, say, Eom et al. (2004). To illustrate the market value proxy method, we consider three firms in different sectors and countries — IBM, Barclays, and Tokio Marine. The input variables used to estimate DTD and DTD* are given in Panel A of Table 1. Their values are as of the end of December 2011. Note that the last entry is equity volatility. It is not needed for the market value proxy method, but will be used for the method discussed in the next section. For IBM, Barclays and Tokio Marine, the values (except for equity volatility) are in million USD, million GBP and million JPY, respectively. In order to compute µ and σ, we use one year of market capitalizations on a daily basis and add quarterly updated book value of liabilities to form one year’s worth of daily asset value time series. We then compute the sample mean and standard deviation of the continuously compounded daily returns. The values for µ and σ are annualized in the typical fashion. With the

MEASURING DISTANCE-TO-DEFAULT FOR FINANCIAL AND NON-FINANCIAL FIRMS

8/30/2012 9:37:20 AM

b1423

Global Credit Review Volume 2

Table 1.

30 August 2012 9:09 AM

Different estimation methods on three types of firms. IBM (Million USD)

Barclays (Million GBP)

Tokio Marine (Million JPY)

216,724 39,843 21,915 28,506 22.44%

21,477 255,193 171,657 1,004,083 56.15%

1,371,714 1,922,395 121,673 12,095,019 31.56%

Panel B: The Market Value Proxy Method µ 14.91% σ 16.06% Asset value (12/2011) 306,988 DTD (12/2011) 8.4697 DTD* (12/2011) 7.6215

−6.94% 6.67% 1,452,410 −0.8495 0.2233

−7.53% 4.84% 15,510,801 0.3326 1.9147

Panel C: The Volatility Restriction Method µ 17.03% σ 18.19% Asset value (12/2011) 278,482 DTD (12/2111) 10.2009 DTD* (12/2011) 9.2645

−7.69% 3.45% 448,327 5.6874 7.8173

−27.51% 12.74% 3,415,782 2.0445 4.2032

Panel D: The KMV Method µ σ Asset value (12/2011) DTD (12/2011) DTD* (12/2011)

−7.90% 6.09% 359,291 −0.4710 0.8563

−26.90% 16.84% 3,352,857 1.4360 3.1174

Panel A: Input Variables Market cap Short-term debt Long-term debt Other liabilities Equity volatility

17.09% 18.51% 267,464 9.8056 8.9748

Panel E: The Transformed-Data MLE Method (with the KMV assumption) µ 13.06% −1.58% −20.33% σ 18.47% 6.91% 16.83% Asset value (12/2011) 267,464 358,293 3,352,858 DTD (12/2011) 9.6054 0.4517 1.8285 DTD* (12/2011) 8.9911 0.7148 3.1208 Panel F: The Transformed-Data MLE Method (including other liabilities) µ 10.70% −1.02% −5.08% σ 18.02% 1.54% 5.17% δ 45.78% 61.83% 60.78% Asset value (12/2011) 280,498 979,658 10,696,331 DTD (12/2011) 8.7478 0.5292 1.6400 DTD* (12/2011) 8.2132 1.1915 2.6339

parameter values in place, we then compute the DTD and DTD* for the end of December 2011 according to Equations (4) and (5). The calculation assumes the maturity equal to one year and the debt amount equal to the total liabilities at the December end. The results are reported in Panel B of Table 1.

2.2. The Volatility Restriction Method A popular way of implementing the Merton (1974) model for pricing corporate bonds and other credit

sensitive instruments is the volatility restriction method of Jones et al. (1984) and Ronn and Verma (1986). Here, we present the version by Ronn and Verma (1986) which has been widely applied in deposit insurance literature along with the Merton (1977) deposit insurance pricing model. The method has also been applied in the empirical credit risk literature such as Hillegeist et al. (2004) and Campbell et al. (2008) without recognizing that their estimation method is in essence that of Ronn and Verma (1986). GLOBAL CREDIT REVIEW VOLUME 2

1250006.indd 99

99

8/30/2012 9:37:20 AM

b1423

Global Credit Review Volume 2

The volatility restriction method uses the following two-equation system:

St = S (Vt ; σ )

σ st = σ

Vt N ( dt ) S (Vt ; σ )

(6) (7)

where S (Vt ; σ ) , N(.) and dt have already been given in Equations (2) and (3), and σSt is equity volatility. Equation (6) links the observed market capitalization to its theoretical counterpart implied by the model. Equation (7) forms a volatility restriction linking the equity volatility to the asset volatility where its righthand side can be derived by applying Ito’s lemma to the pricing formula in Equation (2). There are two unknowns in the above two equation system — Vt and σ. With the two equations, one can proceed to solve for the two unknowns.3 In pricing applications, the volatility restriction method is a calibration which can be carried out just for one time point. Once Vt and σ are available, one can compute the prices of contingent claims. For the DTD, however, one must also know µ. There are two ways of obtaining such an estimate. First, one can repeatedly solve the two-equation system to obtain asset values for many time points, and then compute the sample mean of continuously compounded returns derived from these implied asset values to obtain an estimate for µ. Second, one can solve the two-equation system only once at the time of interest, and apply the obtained asset volatility to all earlier time points to obtain the implied asset values. These implied asset values will differ from those obtained by the first method, but can similarly be used to produce an estimate for µ. Here, we apply the second approach to obtain µ. We apply the volatility restriction estimation method to produce estimates for the three firms as of the end of December 2011. We assume that the debt maturity is one year and the default point (i.e., effective debt level for triggering default) is as in the KMV method (to be discussed in the next section). Since the KMV default point is always lower than the total liabilities, this assumption alone will cause the DTD and other estimates to differ from those produced by the market value proxy method. This choice of default 100

1250006.indd 100

30 August 2012 9:09 AM

point is made to facilitate comparisons with the KMV and other methods to be introduced later. Equity volatility appears in the left-hand side of Equation (7), and its value is per usual estimated by computing the sample standard deviation of the continuously compounded equity returns over some sample period. For this, we use a one-year long time series of daily equity returns to estimate equity volatility, an input to the volatility restriction method. The equity volatilities for the three firms are given in Panel A of Table 1. For estimating parameter µ, we use the estimated asset volatility corresponding to the last data point of the one-year long time series of daily market capitalizations to solve for the entire time series of implied asset values. Our results are reported in Panel C of Table 1. It is clear that the results are quite different from those produced under the market value proxy method. This is of course not at all surprising due to at least two factors: (1) the volatility restriction method recognizes the optionality of corporate liabilities but the market value proxy method does not, and (2) the volatility restriction method has been implemented with the KMV default point assumption that only factors in parts of the total liabilities. Duan (1994) pointed out a critical methodological problem with the volatility restriction method. In essence, the volatility restriction in Equation (7) is obtained by applying stochastic differentiation to the pricing formula in Equation (2). According to the Merton (1974) model, the derived equity volatility must be a stochastic variable. Since it is not a parameter, the sample standard deviation of equity returns should not be the quantity being plugged into the left-hand side of Equation (7). Conceptually, it cannot provide an additional restriction for identification. In practice, one can still obtain estimates as shown in this section, but abusing the system could produce seriously biased estimates as was demonstrated by, for example, Ericsson and Reneby (2005) using a simulation study.

2.3. The KMV Method The KMV method powers the credit analytics service offered by Moody’s KMV. The method is described in

MEASURING DISTANCE-TO-DEFAULT FOR FINANCIAL AND NON-FINANCIAL FIRMS

8/30/2012 9:37:21 AM

b1423

Global Credit Review Volume 2

reasonable detail in Crosbie and Bohn (2003). The method is sometimes mistakenly understood to be the volatility restriction method described above.4 Crosbie and Bohn (2003) started out describing the KMV method on pages 13 and 16 as if it were the volatility restriction method but without a reference to Ronn and Verma (1986). Later on page 17, they described the actual KMV method as an iterative procedure consisting of the following steps: Step 1: Apply an initial value of σ to Equation (5) to obtain a time series of implied asset values and hence continuously compounded asset returns. Step 2: Use the time series of continuously compounded asset returns to obtain updated estimates for µ and σ. Step 3: Go back to Step 1 with the updated σ unless convergence has been achieved. The KMV implementation fixes the maturity at one year and sets the default point to the sum of the shortterm debt and one half of the long-term debt. The argument given in Crosbie and Bohn (2003) is that the KMV experience suggests that a typical firm defaults when its asset value falls somewhere between the short-term debt and the total liabilities. We implement the KMV method on the three firms as before. Again, we use daily market capitalizations along with quarterly updated debt levels in 2011 to generate the estimates. The parameter estimates and the DTDs at the year end for the three firms are reported in Panel D of Table 1. The results are very different compared with those by the market value proxy method or the volatility restriction estimation method. The KMV method represents a clear methodological improvement over the estimation methods discussed thus far. It is self-consistent in the sense that at convergence, the volatility estimate used to produce the implied asset values is also the one implied by the asset values. The KMV method has obvious limitations though. Since its updating mechanism entirely depends on using the implied asset values, one cannot use the method to obtain any unknown parameters present in capital structure. This turns out to be an important limitation for dealing with financial firms, and for which we provide some discussion in the next

30 August 2012 9:09 AM

section and present a practical and better alternative to the KMV method. In addition to the confusion in the literature mentioned earlier about what the KMV method is, there is also a great deal of misunderstanding about the statistical estimation of the Merton model. For example, Bharath and Shumway (2008) commented in page 1345 on the estimation of the Merton (1974) model with the following statement: “Since the Merton DD model is not a typical econometric model, it is not clear … how its parameters might be estimated with alternative techniques. It is also unclear how standard errors for forecasts can be calculated for the Merton DD model.” However, their characterization was inaccurate. In the next section, we will discuss the maximum likelihood technique for this class of models that already appeared in the literature in 1994.

2.4. The Transformed-Data Maximum Likelihood Estimation Method The transformed-data maximum likelihood estimation (MLE) method for models such as Merton (1974) was proposed in Duan (1994, 2000). When the firm’s asset values are not directly observable, one can express the likelihood function of the observed equity values by viewing the equity values as the transformed data where the equity pricing formula in Equation (2) defines the transformation. It should be noted that the transformation involves the unknown asset volatility. By standard transformation theory, the likelihood of observed equity values must equal the product of the likelihood of the asset values (implied by the equity values) and the Jacobian of the inverse transformation (from the equity value back to the asset value). Once the likelihood function is in place, one can apply MLE and its associated statistical inference. The transformed-data MLE method has been used in the deposit insurance and credit risk literature. For the applications in deposit insurance/banking literature, see for example, Duan and Yu (1994), Duan and Simonato (2002), Laeven (2002) and Lehar (2005). As to its applications in credit analysis, Ericsson and Reneby (2004, 2005) and Wong and Choi (2009) are some examples. GLOBAL CREDIT REVIEW VOLUME 2

1250006.indd 101

101

8/30/2012 9:37:21 AM

b1423

Global Credit Review Volume 2

The log-likelihood function based on a sample of n equity prices under the Merton (1974) model can be expressed as: L ( µ , σ ; S1 , S2 ,..., Sn ) =−

n Wˆ 2 n −1 1 n ln (2π ) − ∑ ln (σ 2 ht ) − ∑ 2t 2 2 t =2 t = 2 2σ ht n

n

t =2

t =2

− ∑ ln(Vˆt ) − ∑ ln N (d (Vˆt , σ , Ft , τ t )),

(8)

where Vˆt = g(St ; σ , Ft , τ t ), 1   Wˆ t = ln Vˆt − ln Vˆt −1 −  µ − σ 2  ht ,  2  g(.) is the inverse of the equity pricing formula in Equation (2), and ht is the length of time between two consecutive equity values. With ht, one can easily take care of missing equity values. Maximizing the loglikelihood function in Equation (8) yields maximum likelihood estimators, µˆ and σˆ . Using MLE has clear advantages, because its statistical properties are known and it gives users more than just point estimates. Maximum likelihood estimators are known to be normally distributed in an asymptotic sense. In addition, any differentiable transformation of maximum likelihood estimators is also a maximum likelihood estimator. Thus, the implied asset value can also be characterized by a sampling distribution, and in fact, any quantity of interest can be stated along with a confidence level. Following Duan (1994), the following asymptotic distributions are readily available for the two parameters and the implied asset value at time t: µˆ − µ

n  σˆ −σ 0  → N (0, An−1 )  0   ∂L ( µˆ ,σˆ ) 1  ∂µ 2 where An = −  n  ∂L ( µˆ ,σˆ )  ∂σ ∂µ

   ∂L ( µˆ ,σˆ )  ∂σ 2  

∂L ( µˆ ,σˆ ) ∂µ ∂σ

(9)

n (Vˆt − Vt ) → N (0, Ct' An−1Ct ) 0   ∂g(St; σˆ , Ft , τ t )   where Ct =   ∂σ  

102

1250006.indd 102

(10)

30 August 2012 9:09 AM

A similar result as in Equation (10) is available for other variables of interest, such as DTD and default probability. In those cases, one should recognize that the parameters affect the sampling error directly as well as indirectly through the implied asset value. Specifically, DTD at time t in accordance with Equation (4) should be viewed as a function of, Vˆt (σˆ ), µˆ and σˆ because these three variables are all subject to sampling variations. The KMV method was previously compared with the transformed-data MLE method in Duan, Gauthier and Simonato (2004). They argued in the case of the Merton (1974) model that two methods are equivalent both in principle and in implementation, but distinctively different for models involving unknown capital structure parameters. Specifically, they showed by an example of estimation using 250 daily equity values that the parameter estimates and the implied asset value time series produced by the two methods are virtually the same. For example, the KMV method converges to, µˆ = -0.025, σˆ = 0.177 , and the last implied asset value, V250, equals 0.9708. By the transformed-data MLE method, µˆ = -0.025, σˆ = 0.175, and V250 is 0.9713. They attributed the minor differences to numerical errors arising from two different numerical algorithms. It turns out that the theoretical argument of Duan et al. (2004) is flawed because a complication associated with singularity was overlooked.5 The numerical difference of two methods become more pronounced when the equity, viewed as a call option, is more out-of-the-money. The difference is evident from our analysis of three firms using the transformed-data MLE method under the same KMV assumption of maturity and default point. The results for IBM and Tokio Marine reported in Panel E of Table 1 by and large confirm the similarity conclusion, because their KMV default points are low relative to their market capitalizations.6 In the case of Barclays, the KMV default point is quite high and the divergence of two estimation methods is evident. Putting the similarity/dissimilarity issue aside, the general non-equivalence argument of Duan et al. (2004) has important implications on DTD, particularly for financial firms. Taking Barclays as an example, its default point at the end of December 2011, according to the KMV assumption, was 341 billion

MEASURING DISTANCE-TO-DEFAULT FOR FINANCIAL AND NON-FINANCIAL FIRMS

8/30/2012 9:37:21 AM

b1423

Global Credit Review Volume 2

GBP (short-term debt plus 50% long-term debt), and its market capitalization was 21 billion GBP. Barclays also had other liabilities in the amount of 1,004 billion GBP, which is huge relative to its market capitalization but is not part of the KMV default point calculation. In contrast, IBM had KMV debt (default point) of 51 billion USD, market capitalization of 217 billion USD, and other liabilities of 29 billion USD. Ignoring other liabilities of IBM may have a minimal effect, but excluding other liabilities of Barclays obviously runs a huge risk of understating its default point, underestimating its implied asset value and overestimating its asset volatility. It is worth noting one implementation issue generic to all estimation methods. Firms are dynamic entities, responding to environmental changes and also taking initiatives to grow or to consolidate. Over a sample period, large changes in a firm’s market capitalization may simply reflect scale changes but not the fundamental nature of its return per unit of assets. Failing to neutralize the effect caused by a scale change on market capitalization is likely to produce an over- or under-statement of the asset return. A scale change, however, inflates asset volatility simply because of those large but artificial moves in the asset values. Duan (2010) and Duan et al. (2012) proposed to scale the implied asset value by its corresponding book value. Consider, for example, a firm that has just doubled its asset base without making changes to any other aspect of its operations. This doubling of the asset base should in principle also double its implied asset value, market capitalization, and so on. After being scaled by its book value, however, the asset return will not exhibit an abnormal 100% jump as would happen otherwise. Obviously, if the book asset value stays unchanged throughout the sample period, such scaling has no effect as it should.

2.5. Other Liabilities and the Transformed-Data MLE Estimation Method To our knowledge, the KMV default point formula is not meant for financial firms. In fact, the research papers in the corporate default/bankruptcy prediction literature tend to exclude financial firms from analysis;

30 August 2012 9:09 AM

for example, Duffie et al. (2007). Although the reasons for exclusion are not typically given, it is quite clear from the discussion above that putting financial and non-financial firms into a common data sample requires properly treating financial firms to avoid serious distortions to empirical results. It is also clear that financial firms need a special treatment of their default points. The challenge is how to factor in other liabilities in an operationally feasible way. To account for other liabilities, a method for their inclusion into the default point was proposed and implemented in Duan (2010) and Duan et al. (2012). A haircut is applied to other liabilities much like what the KMV assumption does to the long-term debt. The difference is that the specific haircut is not a predetermined number and will be estimated using the transformed-data MLE method. This estimated haircut method has already been incorporated into the corporate default prediction system under the Credit Research Initiative at the Risk Management Institute, National University of Singapore. The objective of that initiative is to provide objective third-party credit information as a public good through offering freely accessible daily updated default probabilities on the exchange-listed firms around the globe. The specific treatment of other liabilities involves the following definition of default point (i.e., relevant debt level for triggering default): F = short-term debt + 0.5 × long-term debt + δ × other liabilities where parameter δ defines the haircut. Setting δ = 0, one is back to the KMV default point assumption. Under the above generalized default point assumption, the unknown parameters of the model increase from two (µ and σ) to three (µ, σ and δ) with the first two parameters governing the asset value dynamic and the additional one appearing solely in the capital structure. If one attempts to apply the KMV iterative procedure to estimate δ, it will become clear that δ cannot be updated by the implied asset values, because this parameter only appears in the capital structure. However, it is possible to estimate δ with the transformed-data MLE method because this parameter is part of the Jacobian of the transformation from equity value back to the asset value. GLOBAL CREDIT REVIEW VOLUME 2

1250006.indd 103

103

8/30/2012 9:37:21 AM

b1423

Global Credit Review Volume 2

The following log-likelihood function used in Duan (2010) and Duan et al. (2012) has scaled the asset value by its corresponding book value: L ( µ , σ ; S1 ,..., Sn ) = −

n −1 1 n ln(2π ) − ∑ ln(σ 2 ht ) 2 2 t =2 n  Vˆt  − ln ∑   2 t = 2 2σ ht t = 2  At  n

−∑ n

Wˆ t2

(

− ∑ ln N d (Vˆt , σ , Ft , τ t ) t =2

)

(11)

where

 Vˆ A   1  Wˆ t = ln  t t −1  −  µ − σ 2  ht , ˆ 2   Vt −1 At   and Vˆt = g(St; σ, Ft, τt), and g(.) is again the inverse of the equity pricing formula in Equation (2), At is the book asset value, and ht is still the length of time between two consecutive equity values. Panel F of Table 1 presents the estimation results on the same three firms that we have been analyzing throughout this paper. The estimates for δ range from 0.46 for IBM to 0.62 for Barclays. The two financial firms (Barclays and Tokio Marine) have remarkably Table 2.

Panel A: Input Variables Market cap Short-term debt Long-term debt Other liabilities Panel B: The KMV Method µ σ Asset value (12/2011) DTD (12/2011) DTD* (12/2011)

30 August 2012 9:09 AM

close δ. For an industrial firm like IBM, δ is small and the other liabilities are even smaller relative to the market capitalization. Therefore, using the KMV default point formula does not cause too much distortion. In fact, comparing the volatility estimates for IBM under the KMV method with this method shows a minor difference. In the case of financial firms, the impact is quite big. It is also quite clear that the KMV default point assumption overestimates the asset volatilities of financial firms. The implied asset values in the table suggest that the KMV default point assumption has the effect of lowering their values. The net effect on DTD or DTD* is naturally a mixed one, sometimes higher and other times lower, and the results in the table suggests that the default point assumption indeed affects DTDs. Since including other liabilities is meant to deal with financial firms, we further analyze its impact by considering more financial firms with two additional banks and two additional insurance companies from different regions of the world. Table 2 reports the estimation results on three banks: Bank of America, Barclays and DBS. Barclays has been presented earlier in Table 1 but is included here for easy comparisons with other banks. Similarly, we present the results of three insurance companies in Table 3 where

Banks using two estimation methods.

Bank of America (Million USD)

Barclays (Million GBP)

DBS (Million SGD)

56,355 617,218 383,517 1,038,408

21,477 255,193 171,657 1,004,083

26,976 47,696 18,940 210,572

−20.41% 9.41% 849,796 −1.6927 0.5232

−7.90% 6.09% 359,291 −0.4710 0.8563

−0.94% 25.85% 83,381 1.2948 1.4604

Panel C: The Transformed-Data MLE Method (Including Other Liabilities) µ −6.45% −1.02% σ 3.39% 1.54% δ 57.40% 61.83% Asset value (12/2011) 1,456,323 979,658 DTD (12/2011) −0.8484 0.5292 DTD* (12/2011) 1.0579 1.1915 104

1250006.indd 104

−3.88% 4.51% 67.24% 224,990 1.8795 2.7491

MEASURING DISTANCE-TO-DEFAULT FOR FINANCIAL AND NON-FINANCIAL FIRMS

8/30/2012 9:37:21 AM

b1423

Global Credit Review Volume 2

Table 3.

30 August 2012 9:09 AM

Insurance companies using two estimation methods. Sun Life (Million CAD)

AXA SA (Million EUR)

Tokio Marine (Million JPY)

Panel A: Input Variables Market cap Short-term debt Long-term debt Other liabilities

11,046 694 4,889 188,766

23,678 103,590 9,601 543,661

1,371,714 1,922,395 121,673 12,095,019

Panel B: The KMV Method µ σ Asset value (12/2011) DTD (12/2011) DTD* (12/2011)

−0.4061 0.2389 14,154 4.4863 6.3062

−0.2891 0.4078 118,351 −0.6972 0.2156

−0.2690 0.1684 3,352,857 1.4360 3.1174

Panel C: The Transformed-Data MLE Method (Including Other Liabilities) µ −0.0365 −0.0365 −0.0508 σ 0.0350 0.0784 0.0517 δ 0.5828 0.6172 0.6078 Asset value (12/2011) 122,856 460,403 10,696,331 DTD (12/2011) 1.3611 −0.0341 1.6400 DTD* (12/2011) 2.4077 0.4645 2.6339

Sun Life and AXA SA are new but Tokio Marine was previously reported. The input variables in Panel A of these two tables are given in million units of local currencies at the end of December 2011. Panels B and C reports the estimation results using the daily market capitalizations and quarterly updated financial statements in 2011. Panel B is for the KMV method whereas Panel C is for the transformed-data MLE method described in this section. The results in these tables confirm our earlier conclusion on financial firms; that is, other liabilities play an important role in their credit risk analysis. Moreover, the haircut to other liabilities in setting an appropriate default point at the end of 2011 seems to vary little in this sample of banks and insurance companies, even though they are based in various domiciles. In order to see the effect of estimation method over different phases of the macro environment, we repeat two estimation methods on Tokio Marine over the period from January 2004 to December 2011. We perform estimation once per month and use a one-year moving window of daily data. The plots in Figures 1–4 are the monthly time series of estimates (implied asset values, asset volatilities, DTDs and DTD*s) for the KMV method and the transformed-data MLE method that factors in other liabilities. The KMV

14,000,000 12,000,000 10,000,000 8,000,000 6,000,000 4,000,000 2,000,000 0

The KMV Method The Transformed-Data MLE Method(including otherlia bilities)

Figure 1. Monthly time series of implied asset values for Tokio Marine under two estimation methods.

method consistently yields lower implied asset values as shown in Figure 1 and higher asset volatilities as in Figure 2. For DTD or DTD* reported in Figures 3 and 4, the results are not as clear cut with the KMV method generating higher DTDs in the earlier period but lower DTDs in some months in the later period. Comparing Figures 3 and 4 clearly shows that DTD* is more stable over time than is DTD, confirming an earlier assertion about large sampling errors associated with the estimate for parameter µ. GLOBAL CREDIT REVIEW VOLUME 2

1250006.indd 105

105

8/30/2012 9:37:21 AM

b1423

Global Credit Review Volume 2

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

The KMV Method The Transformed-Data MLE Method (including other liabilities)

Figure 2. Monthly time series of asset volatilities for Tokio Marine under two estimation methods.

9 8

30 August 2012 9:09 AM

The transformed-data MLE method (including other liabilities) has been previously implemented to obtain DTDs in a default study of US firms by Duan et al. (2012). In that study, all exchange-listed firms are put into the data sample, which of course includes financial firms. They proposed a forward-intensity default prediction model to link defaults to some common risk factors shared by all firms, such as a benchmark interest, and individual firm attributes, such as liquidity and DTD. With the properly computed DTDs, they are able to show that their default prediction model, after being fitted to the whole data sample, can perform equally well in the financial and non-financial subsectors, suggesting that incorporating other liabilities with a haircut into the default point is a productive way of dealing with financial firms that are typically of high leverage.

7 6

CONCLUSION

5 4 3 2 1 0

The KMV Method The Transformed-Data MLE Method (including other liabilities)

Figure 3. Monthly time series of DTDs for Tokio Marine under two estimation methods.

9 8 7 6 5 4 3 2 1 0

The KMV Method The Transformed-Data MLE Method (including other liabilities)

Figure 4. Monthly time series of DTD*s for Tokio Marine under two estimation methods. 106

1250006.indd 106

In this article, we introduce a popular credit risk measure known as DTD, describe its theoretical foundation and usage, and review the ways of implementation available in the literature. We cover the market value proxy method, the volatility restriction method, the KMV method and the transformed-data MLE method. Three types of firms are used to illustrate the implementation of these methods, and through which we gain better understanding of the methods’ strengths and weaknesses. We pay special attention to financial firms (banks and insurance companies) because of their economic importance and uniqueness in capital structure. Financial firms typically have higher leverage than nonfinancial firms, and the popular KMV estimation seems ill-suited for this category of firms. Blindly applying the KMV method is shown to cause serious distortion to credit analysis. We then introduce a recent advancement in the estimation of the Merton (1974) credit risk model, a method that specifically accounts for the highleverage feature of financial firms in the transformeddata MLE framework of Duan (1994). The new approach is shown, through an analysis of three banks and three insurance companies from different regions of the globe, to differ from the KMV method in a material way. We contend that the new estimation method is superior compared with existing methods. This new

MEASURING DISTANCE-TO-DEFAULT FOR FINANCIAL AND NON-FINANCIAL FIRMS

8/30/2012 9:37:22 AM

b1423

Global Credit Review Volume 2

method is also a practical technology because it has been incorporated into the default prediction system under the non-profit Credit Research Initiative at the Risk Management Institute of National University of Singapore to generate daily updated default probabilities on exchange-listed firms around the globe.7

6

7

NOTES 1

2

3

4

5

For example, Bharath and Shumway (2008) concluded that DTD is not a sufficient statistic for default prediction. Duffie et al. (2007) and Duan et al. (2012) showed that in addition to DTD, there are other variables significantly contributing to default prediction. DTD* is similar to the DTD in the KMV method, which uses a formula, i.e., Equation (5) of Crosbie and Bohn (2003), slightly different from Equation (5) of this paper. The KMV DTD replaces ln(Vt /F ) with its approximation (Vt − F )/F. The method by Jones, Mason and Rosenfeld (1984) only uses Equation (7). The asset value is obtained by a different means. Because of the exposition in Crosbie and Bohn (2003), the KMV method has been misunderstood by some as a two-equation volatility restriction method; for example, Chapter 11 of Caouette et al. (2008) and Footnote # 8 of Eisdorfer and Hsu (2011). In fact, they, along with others such as Bharath and Shumway (2008) also failed to recognize that that the two-equation volatility restriction method was first proposed by Ronn and Verma (1986) and has been widely applied in the deposit insurance literature. Singularity arises from casting the transformed-data problem in the EM algorithm framework. One can create an incomplete-data estimation problem by viewing the observed equity price as the sum of the value produced by Equation (2) and some measurement error, and conduct a maximum likelihood estimation using the EM algorithm. Then by shrinking the measurement error, one is in effect back to the original estimation problem. There is a problem with this reasoning, however. Since the measurement error’s density function, a part of the complete-data likelihood, approaches the Dirac delta function as one shrinks the measurement error to zero (i.e., singularity), it dominates all other terms in the complete-data likelihood, and thus cannot

30 August 2012 9:09 AM

be ignored. Consequently, one cannot be certain that the KMV method produces the maximum likelihood estimate. Note that under the KMV assumption, Tokio Marine’s other liabilities are not included in the default point calculation. In other words, its equity is much more inthe-money than what it would otherwise be. Details on the non-profit Credit Rating Initiative of Risk Management Institute can be found at http://rmicri.org.

REFERENCES Bharath, S.T. and T. Shumway (2008), Forecasting Default with the Merton Distance to Default Model. Review of Financial Studies, 21, pp. 1339–1369. Brockman, P. and H.J. Turtle (2003), A Barrier Option Framework for Corporate Security Valuation. Journal of Financial Economics, 67, pp. 511–529. Campbell, J.Y., J. Hilscher and J. Szilagyi (2008), In Search of Distress Risk. Journal of Finance, 63, pp. 2899–2939. Caouette, J.B., E. Altman, P. Narayanan and R. Nimmo (2008), Managing Credit Risk: The Great Challenge for Global Financial Markets, 2nd edition, John Wiley & Sons. Crosbie, P. and J. Bohn (2003), Modeling Default Risk. Moody’s KMV technical document. Duan, J.-C. (1994), Maximum Likelihood Estimation Using Price Data of the Derivative Contract. Mathematical Finance, 4, pp. 155–167. Duan, J.-C. (2000), Correction: “Maximum Likelihood Estimation Using Price Data of the Derivative Contract”. Mathematical Finance, 10, pp. 461–462. Duan, J.-C. (2010), Clustered Defaults. National University of Singapore Working Paper. Duan, J.-C., G. Gauthier, and J.-G. Simonato (2004), On the Equivalence of the KMV and Maximum Likelihood Methods for Structural Credit Risk Models. University of Toronto Working Paper. Duan, J.-C. and J.-G. Simonato (2002), Maximum Likelihood Estimation of Deposit Insurance Value with Interest Rate Risk. Journal of Empirical Finance, 9, pp. 109–132. Duan, J.-C., J. Sun, and T. Wang (2012), Multiperiod Corporate Default Prediction — A Forward Intensity Approach. Journal of Econometrics, forthcoming (DOI: 10.1016/j.jeconom.2012.05.002). GLOBAL CREDIT REVIEW VOLUME 2

1250006.indd 107

107

8/30/2012 9:37:22 AM

b1423

Global Credit Review Volume 2

Duan, J.-C. and M.-T. Yu (1994), Assessing the Cost of Taiwan’s Deposit Insurance. Pacific-Basin Finance Journal, 2, pp. 73–90. Duffie, D., L. Saita and K. Wang (2007), Multi-period Corporate Default Prediction with Stochastic Covariates. Journal of Financial Economics, 83, pp. 635–665. Eisdorfer, A. and P.-H. Hsu (2011), Innovate to Survive: The Effect of Technology Competition on Corporate Bankruptcy. Financial Management, 40, pp. 1087–1117. Eom, Y.H., J. Helwege, and J.-Z. Huang (2004), Structural Models of Corporate Bond Pricing: An Empirical Analysis. Review of Financial Studies, 17, pp. 499–544. Ericsson, J. and J. Reneby (2004), An Empirical Study of Structural Credit Risk Models Using Stock and Bond Prices. Journal of Fixed Income, 13, pp. 38–49. Ericsson, J. and J. Reneby (2005), Estimating Structural Bond Pricing Models. Journal of Business, 78, pp. 707–735. Hillegeist, S.A., E. Keating, D.P. Cram and K.G. Lunstedt (2004), Assessing the Probability of Bankruptcy. Review of Accounting Studies, 9, pp. 5–34. Jones, E.P., S.P. Mason, and E. Rosenfeld (1984), Contingent Claims Analysis of Corporate Capital Structures: An

108

1250006.indd 108

30 August 2012 9:09 AM

Empirical Investigation. Journal of Finance, 39, pp. 611–625. Laeven, L. (2002), Bank Risk and Deposit Insurance. World Bank Economic Review, 16, pp. 109–137. Lehar, A. (2005), Measuring Systemic Risk: A Risk Management Approach. Journal of Banking & Finance, 29, pp. 2577–2603. Merton, R.C. (1974), On the Pricing of Corporate Debt: The Risk Structure of Interest Rates. Journal of Finance, 29, pp. 449–470. Merton, R.C. (1977), An Analytic Derivation of the Cost of Deposit Insurance and Loan Guarantees. Journal of Banking and Finance, 1, pp. 3–11. Ronn, E.I. and A.K. Verma (1986), Pricing Risk-Adjusted Deposit Insurance: An Option-Based Model. Journal of Finance, 41, pp. 871–895. Wong, H.Y. and T.W. Choi (2009), Estimating Default Barriers from Market Information. Quantitative Finance, 9, pp. 187–196.

MEASURING DISTANCE-TO-DEFAULT FOR FINANCIAL AND NON-FINANCIAL FIRMS

8/30/2012 9:37:22 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

NUS-RMI Credit Research Initiative Technical Report Version: 2012 update 2

INTRODUCTION

T

RMI staff article For any questions or comments on this article, please contact Oliver Chen at [email protected]

his document describes the implementation of the system which the NUS Risk Management Institute’s Credit Research Initiative uses to produce probabilities of default (PDs). As of this version of the Technical Report, these PDs cover exchange listed firms in 44 economies in Asia, Asia-Pacific, North America, Europe and Latin America. Currently, RMI covers over 35,000 listed companies. Of these, over 28,000 firms have sufficient data to release daily updated PDs. The full list of firms is freely available to users who can give evidence of their professional qualifications to ensure that they will not misuse the data. General users who do not request global access are restricted to a list of 2,300 firms. The individual company PD data along with aggregate PDs at the economy and sector level can be accessed at http://rmicri.org.

The primary goal of this initiative is to drive research and development in the critical area of credit rating systems. As such, a transparent methodology is essential to this initiative. Having the details of the methodology available to everybody means that there is a base from which suggestions and improvements can be made. The objective of this Technical Report is to provide a full exposition of the CRI system. Readers of this document who have access to the necessary data and who have a sufficient level of technical expertise will be able to implement a similar system on their own. For a full exposition of the conceptual framework of the CRI, see Duan and Van Laere (2012). The system used by the CRI will evolve as new innovations and enhancements are applied. The most substantial changes to the 2011 technical report and operational implementation of our model are (1) the default definition which now excludes covenant breaches GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 109

109

8/30/2012 9:36:32 AM

b1423

Global Credit Review Volume 2

and some default corporate actions that are specific to Taiwan (e.g., bounced checks); (2) priority of financial statements and treatment of net income, with the latter now being included on a quarterly basis when available; (3) treatment of stale market capitalization prices; (4) regrouping of economies for calibration purposes; (5) increased coverage to include Latin America and the rest of the eurozone countries; and (6) treatment of relative size. This version of the technical report provides an update on the operational implementation of the CRI and includes all changes to the system that had been implemented by July 2012. The latest version of the Technical Report is available via the web portal and will include any changes to the system that have been implemented since the publication of this version. The remainder of this Technical Report is organized as follows. The next section describes the quantitative model that is currently used to compute PDs from the CRI. The model was first described in Duan et al. (2012). The description includes calibration procedures, which are performed on a monthly basis, and individual firm PD computations, which are performed on a daily basis. Section 2 describes the input variables of the model as well as the data used to produce the variables for input into the model. This model uses both input variables that are common to all firms in an economy and input variables that are firm-specific. Another critical component when calibrating a probability of default estimation system is the default data, and this is also described in this section. While Section 1 provides a broader description of the model, Section 3 describes the implementation details that are necessary to apply given real world issues of, for example, bad or missing data. The specific technical details needed to develop an operational system are also given, including details on the monthly calibration, daily computation of individual firm PDs and aggregation of the individual firm PDs. Distance-to-default (DTD) in a Merton-type model is one of the firm-specific variables. The calculation for DTD is not the standard one, and has been modified to allow a meaningful computation of the DTD for financial firms. While most academic studies on default prediction exclude financial firms from consideration, it is important to include them given that the financial 110

1250007.indd 110

30 August 2012 9:08 AM

sector is a critical component in every economy. The calculation for DTD is detailed in this section. Section 4 shows an empirical analysis for those economies that are currently covered. While the analysis shows excellent results in several economies, there is room for improvement in a few others. This is because, at the CRI’s current stage of development, the economies all use the variables used in the academic study of US firms in Duan et al. (2012). Future development within the CRI will deal with variable selection specific to different economies, and the performance is then expected to improve. Variable selection and other planned developments are discussed in Section 5.

I. MODEL DESCRIPTION The quantitative model that is currently being used by the CRI is a forward intensity model that was introduced in Duan et al. (2012). This model allows probability of default forecasts to be made at a range of horizons. In the current CRI implementation of this model, PDs are made from a horizon of one month up to a horizon of two years. In other words, for every firm, the probability of that firm defaulting within one month, three months, six months, one year, eighteen months and two years is given. The ability to assess credit quality for different horizons is a useful tool for risk management, credit portfolio management, policy setting and regulatory purposes, since short- and long-term credit risk profiles can differ greatly depending on a firm’s liquidity, debt structures and other factors. The forward intensity model is a reduced form model in which the probability of default is computed as a function of different input variables. These can be firm-specific or common to all firms within an economy. The other category of default prediction model is the structural model, whereby the corporate structure of a firm is modeled in order to assess the firm’s probability of default. A similar reduced form model by Duffie et al. (2007) relied on modeling the time series dynamics of the input variables in order to make PD forecasts for different horizons. However, there is little consensus on assumptions for the dynamics of variables such as accounting ratios, and the model output will be highly dependent on these assumptions. In addition, the time

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:32 AM

b1423

Global Credit Review Volume 2

series dynamics will be of very high dimension. For example, with the two common variables and two firm-specific variables that Duffie et al. (2007) use, a sample of 10,000 firms gives a dimension of the state variables of 20,002. Given the complexity in modeling the dynamics of variables such as accounting ratios, this model will be diffcult to implement if different forecast horizons are required. The key innovation of the forward intensity model is that PD for different horizons can be consistently and effciently computed based only on the value of the input variables at the time the prediction is made. Thus, the model specification becomes far more tractable. Fully specifying a reduced form model includes the specification of the function that computes a PD from the input variables. This function is parameterized, and finding appropriate parameter values is called calibrating the model. The forward intensity model can be calibrated by maximizing a pseudo-likelihood function. The calibration is carried out by economy and all firms within an economy will use the same parameter values along with each firm’s variables in order to compute the firm’s PD. Subsection 1.1 will describe the modeling framework, including the way PDs are computed based on a set of parameter values for the economy and a set of input variables for a firm. Subsection 1.2 explains how the model can be calibrated.

1.1. Modeling Framework While the model can be formulated in a continuous time framework, as done in Duan et al. (2012), an operational implementation will require discretization in time. Since the model is more easily understood in discrete time, the following exposition of the model will begin in a discrete time framework. Variables for default prediction can have vastly different update frequencies. Financial statement data is updated only once a quarter or even once a year, while market data like stock prices are available at frequencies of seconds. A way of compromising between these two extremes is to have a fundamental time period ∆t of one month in the modeling framework. As will be seen later, this does not preclude updating the

30 August 2012 9:08 AM

PD forecasts on a daily basis. This is important since, for example, large daily changes in a firm’s stock price can signal changes in credit quality even when there is no change in financial statement data. Thus, for the purposes of calibration and subsequently for computing time series of PD, the input variables at the end of each month will be kept for each firm. The input variables associated with the ith firm at the end of the nth month (at time t = n∆t) is denoted by Xi(n). This is a vector consisting of two parts: Xi(n) = (W(n), Ui(n)). Here, W(n) is a vector of variables at the end of month n that is common to all firms in the economy and Ui(n) is a vector of variables specific to firm i. In the forward intensity model, a firm’s default is signaled by a jump in a Poisson process. The probability of a jump in the Poisson process is determined by the intensity of the Poisson process. The forward intensity model draws an explicit dependence of intensities at time periods in the future (that is, forward intensities) to the value of input variables at the time of prediction. With forward intensities, PDs for any forecast horizon can be computed knowing only the value of the input variable at the time of prediction, without needing to simulate future values of the input variables. There is a direct analogy in interest rate modeling. In spot rate models where dynamics on a short-term spot rate are specified, bond pricing requires expectations on realizations of the short rate. Alternatively, bond prices can be computed directly if the forward rate curve is known. One issue in default prediction is that firms can exit public exchanges for reasons other than default. For example, in mergers and acquisitions involving two public companies, there will be one company that delists from its stock exchange. This is important in predicting defaults because a default cannot happen if a firm has been previously delisted. An exception is if the exit is a distressed exit and is followed soon after by a credit event. See Subsection 2.4 for details on how this case is handled in the CRI system. In order to take these other exits into account, defaults and other exits are modeled as two independent Poisson processes, each with their own intensity. While defaults and exits classified as non-defaults are mutually exclusive by definition, the assumption of GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 111

111

8/30/2012 9:36:32 AM

b1423

Global Credit Review Volume 2

independent Poisson processes does not pose a problem since the probability of a simultaneous jump in the two Poisson processes is negligible. In the discrete time framework, the probability of simultaneous jumps in the same time interval is non-zero. As a modeling assumption, a simultaneous jump in the same time interval by both the default Poisson process and the non-default type exit Poisson process is considered as a default. In this way, there are three mutually exclusive possibilities during each time interval: survival, default and non-default exit. As with defaults, the forward intensity of the Poisson process for other exits is a function of the input variables. The parameters of this function can also be calibrated. To further illustrate the discrete framework, the three possibilities for a firm at each time point are diagrammed. Either the firm survives for the next time period ∆t, or it defaults within ∆t, or it has a nondefault exit within ∆t. This setup is pictured in Figure 1. Information about firm i is known up until time t = m∆t and the figure illustrates possibilities in the future between t = (n − 1)∆t and (n + 1)∆t. Here, m and n are integers with m < n. The probabilities of each branch are, for example: pi(m, n) the conditional probability viewed from t = m∆t that firm i will default before (n + 1)∆t, conditioned on firm i surviving up until n∆t. Likewise, p¯i (m, n) is the conditional probability viewed from t = m∆t that firm i will have a non-default exit before

Figure 1. 112

1250007.indd 112

30 August 2012 9:08 AM

(n + 1)∆t, conditioned on firm i surviving up until n∆t. It is the modeler’s objective to determine pi(m, n) and p¯i (m, n), but for now it is assumed that these quantities are known. With the conditional default and other exit probabilities known, the corresponding conditional survival probability of firm i is 1 − pi(m, n) − p¯i (m,n). With this diagram in mind, the probability that a particular path will be followed is the product of the conditional probabilities along the path. For example, the probability at time t = m∆t of firm i surviving until (n − 1)∆t and then defaulting between (n − 1)∆t and n∆t is: Prob t = m∆t[τ i = n, τ i < τ i ] = pi (m, n − 1)

n −2

∏ [1 − pi (m, j ) − pi (m, j )].

(1)

j=m

Here, τi is the default time for firm i measured in units of months, τ¯i is the other exit time measured in units of months, and the product is equal to one if there are no terms in the product. The condition τi < τ¯i is the requirement that the firm defaults before it has a non-default type of exit. Note that by measuring exits in units of months, if, for example, a default occurs at any time in the interval ((n − 1)∆t, n∆t] then τi = n. Using equation (1), cumulative default probabilities can be computed. At m∆t the probability of firm i defaulting at or before n∆t and not having

Default-other exit-survival tree for firm i, viewed from time t = m∆t.

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:32 AM

b1423

Global Credit Review Volume 2

another exit before t = n∆t is obtained by taking the sum of all of the paths that lead to default at or before n∆t: Prob t = m∆t[ m < τ i ≤ n, τ i < τ i ] =

k −1   − − p ( m , k ) [1 p ( m , j ) p ( m , j )]   . (2) ∑ i ∏ i i j=m   k=m  n −1

30 August 2012 9:08 AM

It remains to specify the dependence of the forward intensities on the input variable Xi(m). The forward intensities need to be positive so that the conditional probabilities are non-negative. A standard way to impose this constraint is to specify the forward intensities as exponentials of a linear combination of the input variables:

hi (m, n) = exp[β (n − m) ⋅ Yi (m)], While it is convenient to derive the probabilities given in equations (1) and (2) in terms of the conditional probabilities, expressions for these in terms of the forward intensities need to be found, since the forward intensities will be functions of the input variable Xi(m). The forward intensity for the default of firm i that is observed at time t = m∆t for the forward time interval from t = n∆t to (n + 1)∆t, is denoted by hi(m, n) where m ≤ n. The corresponding forward intensity for a non-default exit is denoted in h¯i (m, n). Because default is signaled by a jump by a Poisson process, its conditional probability is a simple function of its forward intensity: pi (m, n) = 1 − exp[ −∆t hi (m, n)].

(3)

Since joint jumps in the same time interval are assigned as defaults, the conditional other exit probability needs to take this into account: pi (m, n) = exp[ −∆t hi (m, n)]{1 − exp[ −∆t hi (m, n)]}.

(4) The conditional survival probabilities in equations (1) and (2) are computed as the conditional probability that the firm does not default in the period and the firm does not have a non-default exit either:

Prob t = m∆t [τ i , τ i > n + 1| τ i , τ i > n] = exp{−∆t[hi (m, n) + hi (m, n)]}.

(5)

hi (m, n) = exp[β (n − m) ⋅ Yi (m)].

Here, β and β¯ are coefficient vectors that are functions of the number of months between the observation date and the beginning of the forward period (n − m), and Yi(m) is simply the vector Xi(m) augmented by a preceding unit element: Yi(m) = (1, Xi(m)). The unit element allows the linear combination in the argument of the exponentials in equation (6) to have a non-zero intercept. In the current implementation of the forward intensity model in the CRI, the maximum forecast horizon is 24 months and there are 12 input variables plus the intercept. So there are 24 sets of each of the coefficient vectors denoted by β(0), … ,β¯(23) and β¯(0), … ,β¯(23) and each of these coefficient vectors has 13 elements. While this is a large set of parameters, as will be seen in the next part, the calibration is tractable because the parameters for each horizon can be done independently from each other, and the default parameters can be calibrated separately from the other exit parameters. Before giving the probabilities in (1) and (2) in terms of the forward intensities, a notation is introduced for the forward intensities that makes clear which parameters are needed for the forward intensity in question: H (β (n − m), Xi (m)) := exp[β (n − m) ⋅ Yi (m)].

(7)

This is the forward default intensity. The corresponding notation for other exit forward intensities is then just H (β¯(n − m), Xi(m)). So, the probability in (1) is expressed in terms of the forward intensities, using GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 113

(6)

113

8/30/2012 9:36:33 AM

b1423

Global Credit Review Volume 2

(3) for the conditional default probability and (5) for the conditional survival probability: Prob t = m∆t [τ i = n, τ i < τ i ] = {1 − exp[ −∆t H ( β (n − 1 − m), Xi (m))]} n −2

× ∏ exp{−∆t [ H ( β ( j − m), X i (m)) j=m

+ H (β ( j − m), Xi (m))]} = {1 − exp[ −∆t H ( β (n − m − 1), Xi (m))]} n −2  × exp  −∆t ∑ [ H ( β ( j − m), Xi (m)) j=m 

 + H ( β ( j − m), Xi (m))]  . (8)  This probability will be relevant in the next part during the calibration. The cumulative default probability given in equation (2) in terms of the forward intensities is then: Prob t = m∆t [ m < τ i ≤ n, τ i < τ i ]  ∑ {1− exp[ −∆t H ( β (k − m), Xi (m))]} k=m   k −1  × exp  −∆t ∑ [ H ( β ( j − m), Xi (m)) j=m    + H ( β ( j − m), Xi (m))]  .   =

n −1

(9)

This formula is used to compute the main output of the CRI: an individual firm’s PD within various time horizons. The β and β¯ parameters are obtained when the firm’s economy is calibrated, and using those together with the firm’s input variables yields the firm’s PD.

30 August 2012 9:08 AM

than the start of the economy’s dataset or they may exit before the end of the economy’s dataset. There are a total of I firms in the economy, and they are indexed as i = 1,…, I. As before, the input variables for the ith firm in the nth month is Xi(n). The set of all observations for all firms is denoted by X. In addition, the default times τi and non-default exit times τ¯i for the ith firm are known if the default or other exit occurs after time t = ∆t and at or before t = N∆t. The possible values for τi and τ¯i are integers between 2 and N, inclusive. If a firm exits before the month end, then the exit time is recorded as the first month end after the exit. If the firm does not exit before t = N∆t, then the convention can be used such that both of these values are infinite. If the firm has a default type of exit within the dataset, then τ¯i can be considered as infinite. If instead the firm has a non-default type of exit within the dataset, then τi can be considered as infinite. The set of all default times and non-default exit times for all firms is denoted by τ and τ¯, respectively. The first month in which firm i has an observation is denoted by t0i. Except for cases of missing data, these observations continue until the end of the dataset if the firm never exits. If the firm does exit, the last needed input variable Xi(n) is for n = min (τi, τ¯i) − 1. The calibration of the β and β¯ parameters is done by maximizing a pseudo-likelihood function. The function to be maximized violates the standard assumptions of likelihood functions, but Appendix A in Duan et al. (2012) derives the large sample properties of the pseudo-likelihood function. In formulating the pseudo-likelihood function, the assumption is made that the firms are conditionally independent from each other. In other words, correlations arise naturally from sharing common factors W(n) and any correlations there are between different firms’ firm-specific variables. With this assumption, the pseudo-likelihood function for a horizon of
months, a set of parameters β and β¯ and the dataset (τ, τ¯, X ) is:

1.2. Model Calibration The empirical dataset used for calibration can be described as follows. For the economy as a whole, there are N end of month observations, indexed as n = 1, …, N. Of course, not all firms will have observations for each of the N months as they may start later 114

1250007.indd 114

L
(β , β ; τ , τ , X ) =

N −
I

∏ ∏ P
(β , β ;τ i ,τ i , Xi (m)).

(10)

m =1 i =1

Here, P
(β, β¯; τi, τ¯i, Xi(m)) is a probability for firm i, with the nature of the probability depending on what

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:33 AM

b1423

Global Credit Review Volume 2

happens to the firm during the period from month m to month m +
. This is defined as: P
(β , β ; τ i , τ i , Xi (m)) = 1{t0 i ≤ m,min(τ ,τ )> m +
}
−1   × exp  −∆t ∑ [ H (β ( j ), Xi (m)) + H (β ( j ), Xi (m))] j =0   + 1{t0 i ≤ m,τ i ≤τ i ,τ i ≤ m +
}{1 − exp[ −∆t H (β (τ i − m − 1), Xi (m))]} τ i − m −2   × exp  −∆t ∑ [ H (β ( j ), Xi (m)) + H (β ( j ), Xi (m))] j =0  

+ 1{t0 i ≤ m,τ i ≤τ i ,τ i ≤ m +
}{1 − exp[ −∆t H (β (τ i − m − 1), Xi (m))]} × exp[ −∆t H (β (τ i − m − 1), Xi (m))] τ i − m −2   × exp  −∆t ∑ [ H (β ( j ), Xi (m)) + H (β ( j ), Xi (m))] j =0   + 1{t0 i > m} + 1{min(τ i ,τ i )≤ m}⋅ (11)

In words, if firm i survives from the observation time at month m for the full horizon
until at least m +
, then the probability is the model-based survival probability for this period. This is the first term in (11). The second term handles the cases where the firm has a default within the horizon, in which case the probability is the model-based probability of the firm defaulting at the month that it ends up defaulting, as given in equation (8). The third term handles the cases where the firm has a non-default exit within the horizon, in which case the probability is the modelbased probability of the firm having a non-default type exit at the month that the exit actually does occur. The expression for this probability uses the conditional non-default type exit probability given in equation (4). The final two terms handle the cases where the firm is not in the data set at month m — either the first observation for the firm is after m or the firm has already exited. A constant value is assigned in this case so that this firm will not affect the maximization at this time point. The pseudo likelihood function given in (10) can be numerically maximized to give estimates for the coefficients β and β¯. Notice though that the sample observations for the pseudo-likelihood function are

30 August 2012 9:08 AM

overlapping if the horizon is longer than one month. For example, when
= 2, default over the next two periods from month m is correlated to default over the next two periods from month m + 1 due to the common month in the two sample observations. However, in Appendix A of Duan et al. (2012), the maximum pseudo-likelihood estimator is shown to be consistent, in the sense that the estimators converge to the “true” parameter value in the large sample limit. It would not be feasible to numerically maximize the pseudo-likelihood function using the expression given in (11), due to the large dimension of the β and β¯ parameters. Notice though that each of the terms in (11) can be written as a product of terms containing only β and terms containing only β¯. This will allow separate maximizations with respect to β and with respect to β¯. The β and β¯ specific versions of (11) are: P
β (β ; τ i , τ i , Xi (m))
−1   = 1{t0 i ≤ m,min(τ ,τ i )> m +
} exp  −∆t ∑ H (β ( j ), Xi (m)) j =0   τ i − m −2   + 1{t0 i ≤ m,τ i ≤τ i ,τ i ≤ m +
} exp  −∆t ∑ H (β ( j ), Xi (m)) j =0  

× {1 − exp[ −∆t H (β (τ i − m − 1), Xi (m))]} τ i − m −2   + 1{t0 i ≤ m,τ i ≤τ i ,τ i ≤ m +
} exp  −∆t ∑ H (β ( j ), Xi (m)) j =0   × exp[ −∆t H (β (τ i − m − 1), Xi (m))]

+ 1{t0 i > m} + 1{min(τ i ,τ i )≤ m}, P
β (β ; τ i , τ i , Xi (m))
−1   = 1{t0 i ≤ m,min(τ ,τ )> m +
} exp  −∆t ∑ H (β ( j ), Xi (m)) j =0   τ i − m −2   + 1{t0 i ≤ m ,τ i ≤τ i ,τ i ≤ m +
} exp  −∆t ∑ H (β ( j ), Xi (m)) j =0   τ i − m −2   + 1{t0 i ≤ m ,τ i ≤τ i ,τ i ≤ m +
} exp  −∆t ∑ H (β ( j ), Xi (m)) j =0  

× {1 − exp[ −∆t H (β (τ i − m − 1), Xi (m))]} + 1{t0 i > m} + 1{min(τ i ,τ i )≤ m}. GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 115

(12) 115

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

Then, the β and β¯ specific versions of the pseudolikelihood function are given by:

Lβ


N −
I

(β ; τ , τ , X ) = ∏ ∏ m = 1 i =1

(

P
β

(β ; τ i , τ i , X i ( m ) )

N −
I

Thus, the β and β¯ specific pseudo-likelihood functions can be decomposed as:

Lβ
(β ; τ , τ , X ) = (13)

) ∏ ∏ P
β (β ;τ i ,τ i , Xi (m)).

Lβ
β ; τ , τ , X =

30 August 2012 9:08 AM

(


−1

∏ Lβ (
′ ) (β (
′);τ ,τ , X )


¢= 0
−1

(17)

) ∏ Lβ (
′) (β (
′);τ ,τ , X ).

Lβ
β ; τ , τ , X =

m =1 i =1


¢=0

Where With the definitions given in (12) and (13), it can be seen that: L
(β , β ; τ , τ , X ) = L
β (β ; τ , τ , X ) L
β (β ; τ , τ , X ). (14) −

Thus, L
β and L
β can be separately maximized to find their respective parameters. A further important separation is a separation by horizons. Notice that we ¯ can decompose P
β and P
β as: P
β (β ; τ i , τ i , Xi (m)) =

(


−1

∏ P β (
¢ ) (β (
¢);τ i ,τ i , Xi (m)),


¢= 0
−1

) ∏ P β (
¢) (β (
¢);τ i ,τ i , Xi (m)),

P
β β ; τ i , τ i , Xi (m) =


¢=0

Lβ (
¢ ) (β (
¢ ); τ , τ , X ) =

(

N −



∏ ∏ P β (
¢ ) (β (
¢);τ i ,τ i , Xi (m))

m =1 i =1 N −



) ∏ ∏ P β (
¢) (β (
¢);τ i ,τ i , Xi (m)).

Lβ (
¢ ) β (
¢ ); τ , τ , X =

m =1 i =1

(18) Thus, for every horizon
′, Lβ(
′)(β(
′); τ, τ¯, X) and L β(
′)(β¯(
′); τ, τ¯, X) can be separately maximized. In summary, for the current CRI implementation where the horizons are from one month to 24 months, and where there are 13 variables, a 2 × 24 × 13 dimensional maximization is turned into a 13 dimensional maximization done 2 × 24 times. This makes the calibration problem tractable. Additional implementation details on the calibration are given in Section 3.

(15)

II. INPUT VARIABLES AND DATA

where P β (
¢) (β (
¢); τ i,τ i, Xi (m))

= 1{t0 i ≤ m,min(τ i , τ i )> m +
¢+1} exp [ −∆t H (β (
¢), Xi (m))]

+ 1{t0 i ≤ m, τ i ≤τ i , τ i = m +
¢+1} {1 − exp [ −∆t H (β (
¢), Xi (m))]} + 1{t0 i ≤ m, τ i m} + 1{min(τ i , τ i )< m +
¢+1} ,

P β (
¢) (β (
¢); τ i , τ i , Xi (m)) = 1{t0 i ≤ m,min(τ i , τ i )> m +
¢+1} exp  −∆t H (β (
¢), Xi (m)) + 1{t0 i ≤ m, τ i ≤τ i , τ i = m +
¢+1} + 1{t0 i ≤ m, τ i ≤τ i ,τ i = m +
¢+1}{1 − exp[ −∆t H (β (
¢), Xi (m))]} + 1{t0 i > m} + 1{min(τ i , τ i )< m +
¢+1}.

116

1250007.indd 116

(16)

Subsection 2.1 describes the input variables used in the quantitative model. Currently, the same set of input variables is common to all of the economies under the CRI’s coverage. Future enhancements to the CRI system will allow different input variables for different economies. The effect of each of the variables on the PD output is discussed in the empirical analysis of Section 4. Subsection 2.2 gives the data sources and relevant details of the data sources. There are two categories of data sources: current and historical. Data sources used for current data need to be updated in a timely manner so that daily updates of PD forecasts are meaningful. They also need to be comprehensive in their current coverage of firms. Data sources that are comprehensive for current data may not necessarily have

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

comprehensive historical coverage for different economies. Other data sources are thus merged in order to obtain comprehensive coverage for historical and current data. Subsection 2.3 indicates the fields from the data sources that are used to construct the input variables. For some of the fields, proxies need to be used for a firm if the preferred field is not available for that firm. Subsection 2.4 discusses the definition and sources of defaults and of other exits used in the CRI.

2.1. Input Variables Following the notation that was introduced in Section 1, firm i’s input variables at time t = n∆t are represented by the vector Xi(n) = (W(n), Ui(n)) consisting of a vector W(n) that is common to all firms in the same economy, and a firm-specific vector Ui(n) which is observable from the date the firm’s first financial statement is released, until the month end before the month in which the firm exits, if it does exit. In Duan et al. (2012), different variables that are commonly used in the literature were tested as candidates for the elements of W(n) and Ui(n). Two common variables and ten firm-specific variables, as described below, were selected as having the greatest predictive power for corporate defaults in the United States. In the current stage of development, this same set of twelve input variables is used for all economies. Future development will include variable selection for firms in different economies. •

Common variables The vector W(n) contains two elements, consisting of: 1. Stock index return: the trailing one-year simple return on a major stock index of the economy. 2. Interest rate: a representative three-month short-term interest rate with the historical mean subtracted to obtain a de-meaned time series.

•

Firm-specific variables The ten firm-specific input variables are transformations of measures of six different firm characteristics. The six firm characteristics are:

30 August 2012 9:08 AM

(i) volatility-adjusted leverage; (ii) liquidity; (iii) profitability; (iv) relative size; (v) market misvaluation/future growth opportunities; and (vi) idiosyncratic volatility. Volatility-adjusted leverage is measured as the distance-to-default (DTD) in a Merton-type model. The calculation of DTD used by the CRI allows a meaningful DTD for financial firms, a critical group that must be excluded from most DTD computations. This calculation is detailed in Section 3. Liquidity is measured as a ratio of cash and shortterm investments to total assets, profitability is measured as a ratio of net income to total assets, and relative size is measured as the logarithm of the ratio of market capitalization to the economy’s median market capitalization. Duan et al. (2012) transformed these first four characteristics into level and trend versions of the measures. For each of these, the level is computed as the one-year average of the measure, and the trend is computed as the current value of the measure minus the one-year average of the measure. The level and trend of a measure has seldom been used in the academic or industry literature for default prediction, and Duan et al. (2012) found that using the level and trend significantly improves the predictive power of the model for short-term horizons. To understand the intuition behind using level and trend of a measure as opposed to using just the current value, consider the case of two firms with the same current value for all measures. If the level and trend transformations were not performed, then only the current values would be used and the two firms would have identical PD. Suppose that for the first firm the DTD had reached its current level from a high level, and for the second firm the DTD had reached its current level from a lower level (see Figure 2). The first firm’s leverage is increasing (worsening) and the second firm’s leverage is decreasing (improving). If there is a momentum effect in DTD, then firm 1 should have a higher PD than firm 2. Duan et al. (2012) found evidence of the momentum effiect in DTD, liquidity, profitability and size. For the other two firm characteristics, applying the level and trend transformation did not improve the predictive power of the model. GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 117

117

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Figure 2. Two firms with all current values equal to each other, but DTD trending in the opposite direction.

One of the remaining two firm characteristics is the market mis-valuation/future growth opportunities characteristic, which is taken as the market-to-book asset ratio and measured as a ratio of market capitalization and total liabilities to total assets. One can see whether the market mis-valuation effect or the future growth opportunities effect dominates this measure by looking at whether the parameter for this variable is positive or negative. This is further discussed in the empirical analysis of Section 4. The final firm characteristic is the idiosyncratic volatility which is taken as sigma, following Shumway (2001). Sigma is computed by regressing the monthly returns of the firm’s market capitalization on the monthly returns of the economy’s stock index, for the previous 12 months. Sigma is defined to be the standard deviation of the residuals of this regression. Shumway (2001) reasons that sigma should be logically related to bankruptcy since firms with more variable cash flows and therefore more variable stock returns relative to a market index are likely to have a higher probability of bankruptcy. Finally, the vector Ui(n) contains ten elements, consisting of: 1. Level of DTD. 2. Trend of DTD. 118

1250007.indd 118

3. Level of (Cash + Short-term investments)/Total assets, abbreviated as CASH/TA. 4. Trend of CASH/TA. 5. Level of Net income / Total Assets, abbreviated as NI/TA. 6. Trend of NI/TA. 7. Level of log (Firm market capitalization/ Economy’s median market capitalization), abbreviated as SIZE. 8. Trend of SIZE. 9. Current value of (Market capitalization + total liabilities)/Total asset, abbreviated as M/B. 10. Current value of SIGMA. The data fields that are needed to compute DTD and short-term investments are described in Subsection 2.3. The remaining data fields required are straightforward and standard. The computation for DTD is explained in Section 3.

2.2. Data Sources There are two data sources that are used for the daily PD forecast updates: Thomson Reuters Datastream and the Bloomberg Data License Back Office Product. Many of the common factors such as stock index prices and short-term interest rates are retrieved from Datastream.

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

Firm-specific data comes from Bloomberg’s Back Office Product which delivers daily update files by region via FTP after respective market closes. All relevant data is extracted from the FTP files and uploaded into the CRI database for storage. From this, the necessary fields are extracted and joined with previous months of data. The Back Office Product includes daily market capitalization data based on closing share prices and also includes new financial statements as companies release them. Firms will often have multiple versions of financial statements within the same period, with different accounting standards, filing statuses (most recent, preliminary, original, reclassified or restated), currencies or consolidated/unconsolidated indicators. A major challenge lies in prioritizing these financial statements to decide which data should be used. The priority rules are described in Section 3. The firm coverage of the Back Office Product is of sufficient quality that over 28,000 firms’ PD can be updated on a daily basis in the 44 economies under the CRI’s coverage. While the current coverage is quite comprehensive, historical data from the Back Office Product can be sparse for certain economies. For this reason, various other databases are merged in order to fill out the historical data. The other databases used for historical data are: a database from the Taiwan Economics Journal (TEJ) for Taiwanese firms; a database provided by Korea University for South Korean firms; and data from Prowess for Indian firms. With all of the databases merged together and for the 44 economies under CRI’s coverage, over 53,000 exchange listed firms are in the CRI database. This includes over 20,000 delisted firms. The historical coverage of the firm data goes back to the early 1990’s.

2.3. Constructing Input Variables The chosen stock indices and short-term interest rates for the 44 economies under the CRI’s current coverage are listed in Tables A.2 and A.3, respectively. All economies are listed by their three letter ISO code given in Table A.1. Most of the firm-specific variables can be readily constructed from standard fields within firms’ financial statements in addition to daily market capitalization

30 August 2012 9:08 AM

values. The only two exceptions are the DTD and the liquidity measure. The calculation for DTD is explained in Section 3. In the calculation, several variables are required. One variable is a proxy for a one-year risk-free interest rate, and the choices for each of the 44 economies are listed in Table A.4. Total assets, long-term borrowing and total liabilities are also required, but are standard financial statement fields and present no difficulties. Total current liabilities are also required, and due to the relatively large numbers of firms that are missing this value, proxies had to be found. The preferred Bloomberg field for this is BS_CUR_LIAB. If this is missing, then the sum of BS_ST_BORROW, BS_ OTHER_ST_LIAB and BS_CUST_ACCPT_LIAB_ CUSTDY_SEC (customers’ acceptance and liabilities/ custody securities) is used. If one or two of these are missing, zero is inserted for those fields, but at least one field is required. The liquidity measure requires different fields between financial and non-financial firms. For nonfinancial firms, the numerator of the ratio (Cash + Short-term investments) is taken as the sum of BS_ CASH_NEAR_CASH_ITEM and BS_MKT_SEC_ OTHER_ST_INVEST (marketable securities and other short-term investments). If BS_MKT_SEC_ OTHER_ST_INVEST is missing, we substitute with zero but the field BS_CASH_NEAR_CASH_ITEM is required. It was found that this sum frequently overstated the liquidity for financial firms. In place of BS_MKT_SEC_ OTHER_ST_INVEST, financial firms use the sum of ARD_SEC_PURC_UNDER_AGR_TO_RESELL (securities purchased under agreement to re-sell), ARD_ ST_INVEST and BS_INTERBANK_ASSET. If one or two of these are missing, zero is inserted for those fields, but at least one field is required. The “ARD” prefix indicates that these are “as reported” numbers directly from the financial statements. As such, for some firms these fields may need to be adjusted to the same units before adding them to other fields. Summary statistics of the firm-specific variables: DTD, CASH/TA, NI/TA, SIZE, M/B, and Sigma, with the summary statistics provided for firms grouped by economy are listed in Table A.5. GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 119

119

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

2.4. Data for Defaults The Credit Research Initiative database contains credit events of over 4,000 firms from 1990 to the present. The default events come from numerous sources, including Bloomberg, Compustat, CRSP, Moody’s reports, TEJ, exchange web sites and news sources. The default events that are recognized by the CRI can be classified under one of the following events: 1. Bankruptcy filing, receivership, administration, liquidation or any other legal impasse to the timely settlement of interest and/or principal payments; 2. A missed or delayed payment of interest and/or principal, excluding delayed payments made within a grace period; 3. Debt restructuring/distressed exchange, in which debt holders are offered a new security or package of securities that result in a diminished financial obligation (e.g., a conversion of debt to equity, debt with lower coupon or par amount, debt with lower seniority, debt with longer maturity). The more precise sub-categories of default corporate actions are listed in Table A.6. Delisting due to other reasons such as failure to meet listing requirements, inactive stock prices or M&A are counted as “other exits” and are not considered as default. However, firms that are delisted from an exchange and which experience a default event within 365 calendar days of the delisting will have an exit event reclassified as credit default. Technical defaults such as covenant violations are not included in our definition of default. The exit events that are not considered as defaults in the CRI system are listed in Table A.7. In addition to the aforementioned events, there are still cases that require special attention and will be assessed on a case-by-case basis, e.g., subsidiary default. As a general rule, the CRI does not consider related party-default (e.g., subsidiary bankruptcy) as a default event. However, when a non-operating holding parent company relies heavily on its subsidiary, bankruptcy by the subsidiary will cause a considerable economic impact on the parent company. Such cases are reviewed and final classifications made. 120

1250007.indd 120

30 August 2012 9:08 AM

The total number of firms, number of defaults and number of other exits in each of the 44 economies each year from 1992 to 2012 are listed in Table A.8. Note that the total number of firms here includes all firms where the primary listing of the shares are on an exchange in that economy and may include firms where there are too many missing data values for a PD estimate to be made. However, the number of firms listed on the CRI web portal under the tab Aggregate forecast includes firms that are domiciled in that economy and excludes firms where a PD cannot be produced due to missing data.

III. IMPLEMENTATION DETAILS Section 1 describes the modeling framework underlying the current implementation of the CRI system. It focuses on theory rather than the details encountered in an operational implementation. The present section describes how the CRI system handles these more specific issues. Subsection 3.1 describes implementation details related to data, mainly dealing with data cleaning and missing data. Subsection 3.2 describes the specific computation of distance-to-default (DTD) used by the CRI system that leads to meaningful DTD for financial firms. Subsection 3.3 explains how the calibration previously described in Subsection 2.2 can be implemented. Subsection 3.4 gives the implementation details relevant to the daily output. This includes an explanation of the various modifications needed to compute daily PD so that the daily PD is consistent with the usual month end PD, and a description of the computation of the aggregate PDs provided by the CRI.

3.1. Data Treatment Fitting data to monthly frequency: Historical end of month data for every firm in an economy is required to calibrate the model. For daily data such as market capitalization, interest rates and stock index values, the last day of the month for which there is valid data is used. For financial statement variables, data is used starting from the period end of the statement lagged by three months. This is to ensure (insofar as is possible)

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

that predictions are made based on information that was available at the time the prediction was made. Of course, for more recent data where the CRI database contains the financial statement but the period end lagged by three months is after the current day, the financial statement is used in making PD forecasts. The CRI considers financial statement variables to be valid for one year without restriction after they are first used. Currency conversions are required if the market capitalization or any of the financial statement variables are reported in a currency different than the currency of the economy. If a currency conversion is required, the foreign exchange rate used is that reported at the relevant market close. For firms traded in Asia and Asia-Pacific, the Tokyo closing rate is used; for firms traded in Western Europe, the London closing rate is used; and for firms traded in North America and Latin America, the New York closing rate is used. For market capitalizations, the FX rate used is for the date that the market capitalization is reported. For financial statement variables, the FX rate used is for the date of the period end of the statement. Priority of financial statements: As described in Subsection 2.2, data provided in Bloomberg’s Back Office Product can include numerous versions of financial statements within the same period. If there are multiple financial statements with the same period end, priority rules must be followed in order to determine which to use. The formulation and implementation of these rules is a major challenge and an area of continuing development. The first rule prioritizes by consolidated/unconsolidated status. This status is relevant only to firms in India, Japan, South Korea and Taiwan, so this rule is only relevant in those economies. Most firms in these economies issue unconsolidated financial statements more frequently than consolidated ones, so these are given higher priority. This simple prioritization can, however, lead to cases where the financial statements used switch from consolidated statements to unconsolidated statements and back again. A more complex prioritization rule is currently under development, with the intention of avoiding this situation. If, after the first prioritization rule has been applied, there are still multiple financial statements, the second

30 August 2012 9:08 AM

rule is applied. This is prioritization by fiscal period. In most economies, annual statements are required to be audited, whereas other fiscal periods are not necessarily audited. The order of priority from highest to lowest is, therefore: annual, semi-annual, quarterly, cumulative, and finally other fiscal periods. The third prioritization rule is based on filing status. The “Most Recent” statement is used before the “Original” statement, which is used before the “Preliminary” statement. The final prioritization rule is based on the accounting standard. Here, financial statements that are reported using Generally Accepted Accounting Principles (GAAP) are given higher priority than financial statements that are reported using International Financial Reporting Standards (IFRS). If an accounting standard is not indicated at all, the financial statement is not used. Financial statement entries with all other descriptors being the same but with different filing statuses will be grouped together. For each variable separately, the variable value is taken from the highest priority financial statement within the group where the value is non-null. For example, suppose two financial statement entries have the same period end, are both annual statements, are both consolidated statements, and both use the same accounting standard, but the first entry is classified as the “Most Recent” and the second is classified as the “Original” entry. Suppose the total assets and total liabilities are reported in the “Original” entry, and in the “Most Recent” entry only the total liabilities have been updated with a null value for the “Original” entry. Then, the total liabilities will be taken from the “Most Recent” entry while the total assets will be taken from the “Original” entry. This allows for the grouping of, for example, “Most Recent” and “Original” entries together because Bloomberg occasionally only updates values that change without updating other values. If the entries are not grouped, then most of the variables would have null values. One variable that needs special attention is net income. Net income is a flow variable and needs to be adjusted based on the period of the financial statement. More specifically we transform the net income GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 121

121

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

into a monthly net income by dividing the net income by the number of months that the financial statement covers. Due to the different coverage periods, several sources for the net income may be available. For example, the monthly net income can be computed from the annual net income divided by 12, the semiannual net income divided by six and the quarterly net income divided by three. When the monthly net income can be obtained from different sources simultaneously, the quarterly net income will have higher priority than any other because it covers a more recent period. Treatment of stale market capitalization prices: The market capitalization of a firm is required in a few input variables: DTD, SIZE, M/B and SIGMA. For most firms, the market capitalization is available from Bloomberg on a daily basis. A check on the trading volume of shares is used to remove stale prices. Specifically, if there are more than two consecutive days of identical market capitalization prices, subsequent identical prices are removed only if the trading volume is equal to zero. This is to avoid, for example, cases where the shares of a company are under a trading suspension but the market capitalization data is incorrectly carried forward. An exception is for Indian companies, where it is common for some companies to have market capitalizations reported only once a month with several consecutive months having identical prices and positive trading volume. These prices are very likely not to be accurate reflections of the firms’ value. So, the trading volume is not checked for Indian firms and market capitalizations are excluded after more than two repeated prices. For some firms, there are gaps in the market capitalization data provided by Bloomberg. Previously, the first recourse was to use the share price multiplied by the shares outstanding listed in the balance sheet and multiplied by an adjustment factor that Bloomberg provides to account for splits, dividends, etc. However, this data is frequently in error and using the shares outstanding as the previous available market capitalization divided by the price on that day was found to be more reliable. If the gap in market capitalization data is more than a year, then the previous computation using the shares 122

1250007.indd 122

30 August 2012 9:08 AM

outstanding from the balance sheet is again used. If there are still remaining gaps in the data, then shares outstanding from Compustat data is used. Provisions for missing values and outliers: Missing values and outliers are dealt with by a three step procedure. In the first step, the ten firm-specific input variables are computed for all firms and all months. In the second step, outliers are eliminated by winsorization. In the final step, missing values are replaced under certain conditions. The first step is to compute the input variables and determine which are missing. As mentioned previously, financial statement variables are carried forward for one year after the date that they are first used. This is generally three months after the period end of the statement. If no financial statement is available for the company within this year, then the financial statement variable will be missing. For market capitalization, if there is no valid market capitalization value within the calendar month, then the value is set to missing. For illiquid stocks, if there has been no valid market capitalization value for a firm within the last 90 calendar days, then the market capitalization is deemed to not properly reflect the value of the firm. The firm is considered to have exited with a nondefault event. Once the firm starts trading again and a new financial statement is released, the firm can enter back into the calibration. With regard to historical PD, the PD can be reported again once there are enough valid variables. With regard to the level variables, the current month and the last eleven months are averaged to compute the level. There is no lower limit on the number of valid observations. Only if all of the values are missing is the level variable considered to be missing. For the trend variable, the level is subtracted from the current month. If the current month is missing, then the trend variable is set to missing. The value of M/B is set to missing if any of the following values are missing: market capitalization, total liabilities or total assets of the firm. For the computation of SIGMA, seven valid returns over the last twelve months of possible returns are required for the regression. If there are less than seven valid returns, SIGMA is set to missing.

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:34 AM

b1423

Global Credit Review Volume 2

In this way, the eight trend and level variables plus M/B and SIGMA are computed and evaluated as missing or present. Winsorization can then be performed as a second step to eliminate outliers. The volume of outliers is too large to be able to determine whether each one is valid or not, so winsorization applies a floor and a cap on each of the variables. The historical 0.1 percentile and 99.9 percentile for all firms in the economy are recorded for each of the ten variables. Any values that exceed these levels are set to equal these boundary values. With a winsorization level and 0.1 percentile and 99.9 percentile, the boundary values still may not be reasonable. For example, NI/TA levels of nearly −25 have been observed at this stage. In these cases, a more aggressive winsorization level is applied, until the boundary values are reasonable. Thus, the winsorization level is economy and variable specific, and will depend on the data quality for that economy and variable. The applied winsorization levels different from the default of 0.1 percentile and 99.9 percentile are indicated in the tables on our web portal. A third and final step can be taken to deal with missing values. If, during a particular month, no variables for a firm are missing, then the PD can be computed. If six or more of these ten variables are missing, there is deemed to be too many missing observations and no replacements are made. If between one and five variables are missing out of the ten, the first step is to trace back for at most twelve months to use previous values of these variables instead. If this does not succeed in replacing all of the variables, a replacement by sector medians is done. The median is for the financial or non-financial firms (as indicated by their Bloomberg industry sector code) within the economy during that month. Replacement by the sector median should have a neutral effect on the PD of the firm; the firm is assessed by the other variables that it does have values for. This sector median is always performed in calibration. However, when reporting historical PD, the sector replacement is not done if it results in a relative change in PD of 10% or more where the initial PD was at or above 100bps, or an absolute change in PD of 10bps or more where the initial PD was below 100bps. Inclusion/exclusion of companies for calibration: Firms are included within an economy for

30 August 2012 9:08 AM

calibration when the primary listing of the firm is on an exchange in the economy. This ensures that all firms within the economy are subject to the same disclosure and accounting rules. There are a relatively small number of firms that are dual listed, in which two corporations listed in different exchanges operate as a single entity but retain separate legal status. In the CRI system, a combined company will be assigned to the single economy it is most associated with. An example is the Rio Tinto Group. This consists of Rio Tinto plc, listed in the UK; and Rio Tinto Limited, listed in Australia. Most of Rio Tinto’s operations are in Australia rather than the UK, so Rio Tinto is assigned to Australia. In the US, firms traded on the OTC markets or the Pink Sheets are not considered as exchange listed so are not included in calibration or in the reporting of PD forecasts. Many of these firms are small or start-up firms. Including this large group of companies would skew the calibration and the aggregate results. The TSX Venture Exchange in Canada also contains only small and start-up firms, so firms listed on that exchange are also excluded. Other examples include Taiwan’s GreTai Securities Market and Singapore’s Catalist. The challenge for markets outside of the US or Canada is that the data on whether firms are listed on the smaller markets rather than the main board is diffcult to obtain. For all economies besides the US and Canada, there is continuing work being done in the CRI system to exclude firms that are not listed on major exchanges within a country. Firms that record an exit (other than due to no trading for 90 calendar days) are not entered back into the calibration even if the firm continues to trade and issue financial statements, as can happen after firms declare bankruptcy. There are two exceptions to this exclusion. The first, determined on a case by case basis, is if the firm should be deemed to have reemerged from bankruptcy. The second exception is for all firms in China, where two situations are prevalent. The first situation is that the firm experiences few repercussions from the default and continues operating normally. The other situation is for one firm to take over a defaulted firm’s listing. This happens due to the limited supply of exchange listings. Both of GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 123

123

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

these situations can be considered as emerging from default, so the CRI system enters all of these companies back into the calibration as new companies.

30 August 2012 9:08 AM

In Merton’s model, DTD is defined as volatility scaled distance of the expected asset value under the physical measure at maturity T from the default point L:

DTDt =

3.2. Distance-to-Default Computation The distance-to-default (DTD) computation used in the CRI system is not a standard one. Standard computations exclude financial firms, but excluding the financial sector means neglecting a critical part of any economy. So the standard DTD computation must be extended to give meaningful estimates for financial firms as well. Duan and Wang (2012) provide a review of different DTD calculations with several examples for financial and non-financial firms. The description of the specialized DTD computation starts with a brief description of the Merton (1974) model. Merton’s model makes the simplifying assumption that firms are financed by equity and a single zero-coupon bond with maturity date T and principal L. The asset value of the firm Vt follows a geometric Brownian motion: dVt = mVt dt + s Vt dBt .

(19)

Here, Bt is standard Brownian motion, µ is the drift of the asset value in the physical measure and σ is the volatility of the asset value. Equity holders receive the excess value of the firm above the principal of the zerocoupon bond and have limited liability, so the equity value at maturity is: ET = max(VT − L, 0). This is just a call option payoff on the asset value with a strike value of L. Thus, the Black–Scholes option pricing formula can be used for the equity value at times t before T: Et = Vt N (d1 ) − e − r (T − t ) LN (d2 ).

(20)

where r is the risk-free rate, N(·) is the standard normal cumulative distribution function, and:

d1,2 =

124

1250007.indd 124

s T −t

( ) + (µ − )(T − t ) . σ2 2

Vt L

σ T −t

L( m , σ , d ) n  Vˆ (s , d )  n −1 1 n =− log(2π ) − ∑ log(σ 2 ht ) − ∑ log  t  2 2 t =2  At  t =2

( (

))

− ∑ log  N dˆ1 Vˆt (σ , δ ), σ , δ   

(21)

(22)

The standard KMV assumptions given in Crosbie and Bohn (2003) are to set the time to maturity T − t at a value of one year and the principal of the zerocoupon bond L to a value equal to the firms current liabilities plus one half of its long-term debt. Here, the current liabilities and long-term debt are taken from the firm’s financial statements. If the firm is missing the current liabilities field, then various substitutes for this field can be used, as described in Subsection 2.3. This is a poor assumption of the debt level for financial firms, since they typically have large liabilities, such as deposit accounts, that are neither classified as current liabilities nor long-term debt. Thus, using these standard assumptions means ignoring a large part of the debt of financial firms. To properly account for the debt of financial firms, Duan (2010) includes a fraction δ of a firm’s other liabilities. The other liabilities are defined as the firm’s total liabilities minus both the short and longterm debt. The debt level L then becomes the current liabilities plus half of the long-term debt plus the fraction δ multiplied by the other liabilities, so that the debt level is a function of δ. The standard KMV assumptions are then a special case where δ = 0. The fraction δ can be optimized along with and in the maximum likelihood estimation method developed in Duan (1994, 2000). Following Duan et al. (2012), the firm’s market value of assets is standardized by its book value At so that the scaling effect from a major investment or financing by the firm will not distort the time series from which the parameter values are estimated. Thus, the log-likelihood function is:

n

s2  V   log  t  +  r ± (T − t )  L  2 

log

t =2 n

(23) 2

  Vˆ (s , d ) At −1   σ 2  − ∑ 2  log  t × − − m ht  .  2   Vˆt −1 (s , d )   t = 2 2σ ht    At 1

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

Here, n is the number of days with observations of the equity value in the sample, Vˆt is the implied asset value found by solving equation (20), dˆ1, is computed with equation (21) using the implied asset value, and ht is the number of trading days as a fraction of the year between observations t − 1 and t. Notice that the implied asset value and dˆ1 are dependent on δ by virtue of the dependence of L on δ. Implementation of DTD computation: The DTD at the end of each month is needed for every firm in order to calibrate the forward intensity model. A moving window, consisting of the last one year of data before each month end is used to compute the month end DTD. Daily market capitalization data based on closing prices is used for the equity value in the implied asset value computation of equation (20). If there are fewer than 50 days of valid observations for the market capitalization, then the DTD value is set to missing. An observation is valid if there is positive trading volume that day. If the trading volume is not available, the observation is assumed to be valid if the value for the market capitalization changes often enough. The precise criterion is as follows: if the market capitalization does not change for three days or more in a row, the first day is taken as a valid observation and the remaining days with the same value are set to missing. The log-likelihood function given in (23) can be maximized as a three dimensional maximization problem over µ, σ and δ. After estimates for these three variables are made, the DTD can be computed from equation (22). However, with quarterly financial statements there will never be more than three changes in the corporate structure (defined in this model by L and At) throughout the year, leading to possibly unstable estimates of δ. This problem is mitigated by performing a two stage optimization for µ, σ and δ. In the first stage, the optimization for each firm is performed over all three variables. For each firm, in the first month in which DTD can be computed the optimization is unconstrained in µ and σ, while δ is constrained to being in the unit interval [0, 1]. Thereafter, at month n, the optimization is still unconstrained in µ and σ while δ is constrained to the interval [max(0,δˆn−1 − 0.05),min(1,δˆn−1 + 0.05)], where

30 August 2012 9:08 AM

δˆ(n−1) is the estimate of δ made in the previous month. In other words, a 10% band around the previous estimate of δ (where that band is floored with 0 and capped with 1) is applied so that the estimates do not fluctuate too much from month to month. It was found that this was not enough to obtain stable estimates of δ. For many firms, the estimate of δ would frequently lie on the boundary of the constraining interval. To impose greater stability, a second stage is added. At each month end, the average estimate for δ in all financial sector firms in the economy is used for every financial sector firm in the economy, meaning the optimization is only over µ and σ. The same is done for non-financial firms. In fact, the optimization can be reduced to be only over σ by using the sample mean of the log returns of the implied asset values in place of µ. Since the first stage is done to obtain a stable sector average estimate of δ, the criteria used to include a firm-month is more strict. In the first stage, a two-year window is used instead of one year, and a minimum of 250 days of valid observations of the market capitalization are required instead of 50. If a firm has less than 250 days of valid observations within the last two years of a particular month end, δ will not be estimated for that firm and that month end. It was found that the estimate of µ was frequently unstable and could lower the explanatory power of DTD. For example, suppose a firm has a large drop in its implied asset value in January 2011, so that the estimated µ is negative for the DTD calculation at the end of December 2011. If there is little change in the company in January 2012, then the drop in implied asset value in January 2011 is no longer within the observation window for the DTD calculation at the end of January 2012. There will be a large increase in the estimated µ, resulting in a substantial improvement of the DTD just because of the moving observation window. To avoid this problem, we now set µ to be equal to 2 σ /2. So in calculating DTD, the second term in the numerator of Equation (22) is eliminated. In summary, the DTD for each firm is computed using the economy and sector (financial or nonfinancial) average for δ in that month, and the estimate of σ is based on the last year of data for the firm. GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 125

125

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

Carrying out this two-stage procedure would take several months of computation time on a single PC, given the millions of firm months that are required. However, each of the stages is parallelizable. In the first stage the DTD can be computed independently between firms. In the second stage, once the sector averages of the δ have been computed for each month, the DTD can again be computed independently between firms. In the CRI system, a grid of several hundred computers administered by the NUS Computer Center is used. With this, the DTD computation can be performed for all firms over the full history of twenty years in less than one day.

3.3. Calibration Implementation: As shown in Section 1, the calibration of the forward intensity model involves multiple maximum pseudo-likelihood estimations, where the pseudo-likelihood functions are given in equation (18). The maximizations are of the logarithm of these expressions, and they are performed independently between the default parameters and the exit parameters, and between parameters for different horizons. In the notation of Section 1, the vectors of parameters β(0), … ,β (23) and β¯(0), … ,β¯(23) are independently estimated. A few input variables have an unambiguous effect on a firm’s probability of default. Increasing values of both the level and trend of DTD, CASH/TA, and NI/TA all indicate that a firm is becoming more credit worthy and should lead to a decreased PD. For large and relatively clean datasets such as the US, an unconstrained optimization leads to parameter values which largely have the expected sign. For each of DTD level and trend, CASH/TA level and trend, and NI/TA level, the default parameters at all horizons are negative. A negative default parameter at a horizon means that if the variable increases, the forward intensity will decrease (by equation (6)), so that the conditional default probability at that horizon will decrease. The one exception is the NI/TA trend variable. For some of the smaller economies and economies with lower quality datasets, an unconstrained optimization leads to the default parameters for some of

126

1250007.indd 126

30 August 2012 9:08 AM

these variables to be positive at several horizons. This leads to counter-intuitive results. For example, if the default parameters for CASH/TA are positive, a firm that increases its cash reserves, all other factors being equal, will have a PD that increases. To prevent such situations, the CRI system performs a constrained optimization with only non-positive values allowed for the default parameters associated with the level and trend of DTD, CASH/TA, and NI/TA. For this, the Matlab function “fmincon” from the Optimization Toolbox is used. The analytic gradient and Hessian are supplied and the algorithm used by “fmincon” is the trust-region-reflective optimization. Notice that at each time point and at any horizon, there are orders of magnitude more surviving firms than exiting firms. Thus, from equations (16) and (18), it can be seen that the most time-consuming part of evaluating the pseudo-log-likelihood function is the term for the surviving firms. Evaluating the forward intensity function of equation (7) can be formulated as a matrix-vector multiplication, where the rows of the matrix are the different surviving firms variables, and the vector is the vector of parameters. The matrix will typically have several hundreds of thousands of rows and does not change during the optimization (though it will change for different optimizations at different horizons). This type of problem is well-suited for a programmable graphics processing unit (GPU). The analytic expressions for the gradient and Hessian can similarly be computed efficiently on a GPU. The CRI system runs the calibrations on an NVIDIA Tesla C2050 card. For each economy, the calibrations for the default and other exit parameters for horizons up to 24 months typically require five minutes or less. Grouping for economies: There are not enough defaults in some small economies and calibrations of these individual economies are not statistically meaningful. In order to ensure that there are enough defaults for calibration, the 44 economies are categorized into groups according to similarities in their stage of development and their geographic locations. Within these groups the economies are combined and calibrated together. Starting from the May 2011 calibration, Canada and the US remain in the same calibration group, and

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

the developed economies of Asia-Pacific (Australia, Hong Kong, Japan, Singapore, South Korea and Taiwan) form another calibration group. China and India, the two major emerging economies of Asia Pacific are each calibrated as an individual group. Starting from June 2012, all 16 of the European countries covered by the CRI are in a single calibration group and the other emerging economies of Asia Pacific (Indonesia, Malaysia, Philippines and Thailand) are grouped together with the 7 Latin American countries (Argentina, Brazil, Colombia, Chile, Mexico, Peru and Venezuela) to form the calibration group “emerging markets”. All economies in these new calibrations groups share the same coefficients for all variables except for the benchmark risk-free interest rate variable. The benchmark interest rate’s coefficients will be allowed to vary, because different economies based in different currencies naturally have different dependencies on their interest rates, and the interest rate levels can differ significantly across economies. After adopting the euro, all eurozone countries use Germany’s threemonth Bubill rate as this is more reflective of monetary rather than sovereign credit conditions in each economy, which is the intent of this variable. For the period before joining the eurozone, their own interest rates are used. In addition, the benchmark interest rate is entered as the current value minus the historical month-end mean. This allows the variable to reflect its value relative to the historical average. When an economy does not have enough default events to identify a separate interest rate coefficient, the interest rate variable will be disabled for that economy by inputting a zero value for the whole time series. In fact, that is also why we de-mean all interest rate series so that setting the interest rate series of a particular economy to zero, when necessary, does not induce a bias by the base economy in the same group. Since all eurozone countries except Germany do not have enough default events prior to joining the eurozone, their benchmark interest rate is entered as zero for that period. Among the non-eurozone members of the European group, Denmark, Norway, Sweden and the UK each have separate coefficients for the

30 August 2012 9:08 AM

benchmark interest rate. Switzerland and Iceland do not use this variable for their whole history. In the Developed Asia-Pacific group, all economies have their own coefficient for the benchmark interest rate. For the North American group, both Canada and the US have their own coefficient for the benchmark interest rate. In the Emerging Markets group, there are insufficient defaults in the Latin American economies to calibrate individual economy benchmark interest rate coefficients in a statistically significant way, so all Latin American economies share the same benchmark interest rate coefficient. Each of the Asian economies in the Emerging Markets group, namely Indonesia, Malaysia, Philippines and Thailand, have their own coefficient for the benchmark interest rate. Relative Size: For the calibration dataset, the median market cap of firms in an economy for each month end includes the market cap from the last trading day of each firm in the month. If a firm does not trade in a particular month, the firm’s market cap is not included in the median. For certain economies, many firms are illiquid and the median market cap experiences large variations due to the change in composition of firms rather than the market value of the firms. Another problem is data quality at the beginning of the historical sample: if a data provider starts including the market cap for a large number of firms in one month compared to the previous, there can be a large jump in the median market cap. To avoid this problem, we use a combination of the economy’s stock index and the economy’s median market cap as the divisor in the Relative Size variable: 1. We choose a recent month where there is a more complete set of firms in the economy that have trading activity, and calculate the ratio of the economy’s median market cap to stock index value at the end of the month. 2. For each month, the divisor for the Relative Size variable of firms in the economy is taken as the month end stock index multiplied by that ratio.

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 127

127

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

3.4. Daily Output Individual firms’ PD: In computing the pseudo-loglikelihood functions in equation (18), only end of month data is needed. The data needs to be extended to daily values in order to produce daily PDs. For the level variables, the last twelve end of month observations (before averaging) are combined with the current value. The current value is scaled by a fraction equal to the current day of the month divided by the number of calendar days in the month. The earliest monthly value is scaled by one minus this fraction. The sum is then divided by the number of valid monthly observations, with the current value and the earliest monthly value counting as a single observation if either or both are not missing. Not performing this scaling can lead to an artificial jump in PD at the beginning of the month. When performing the scaling, the change in level is more gradual throughout the month. A similar procedure is done for SIGMA. Here the earliest month is not scaled, but the return from the current day to the previous month end is scaled by the square root of the fraction equal to the current day of the month divided by the number of calendar days in the month. Computing the DTD for all firms on a daily basis using the two stage process described in Subsection 3.2 would be time consuming, even on the grid. Since there should be little change to σ and δ on a day to day basis, for the daily computation of DTD these are assumed to have the same value as in the previous month’s DTD calculation. In other words, the previous month’s values for σ and δ together with the new day’s equity value are used in equation (20) to obtain the implied asset value. This implied asset value with the previous month’s values for σ and δ is used in equation (22) to obtain the new day’s DTD, with µ set to equal to σ2/2. Aggregating PD: The CRI provides term structures of the probability distributions for the number of defaults as well as the expected number of defaults for different groups of firms. The companies are grouped by economy (using each firm’s country of domicile), by sector (using the firm’s Bloomberg industrial sector code) and sectors within economies. With the individual firms’ PD, the expected number of defaults 128

1250007.indd 128

30 August 2012 9:08 AM

is trivial to compute. The algorithm used to compute the probability distribution of the number of defaults was originally reported in Anderson et al. (2003). It assumes conditional independence and uses a fast recursive scheme to compute the necessary probability distribution. Note that while this algorithm is currently used to produce the probability distribution of the number of defaults within an economy or sector, it can easily be generalized to compute loss distributions for a portfolio manager, where the exposure of the portfolio to each firm needs to be input. Inclusion of firms in aggregation: As explained in Subsection 3.1, firms are included in an economy for calibration if the firms’ primary listing is on an exchange in that economy. This is to ensure that all firms in an economy are subject to the same disclosure and accounting requirements. In contrast, a firm is included in an economy’s aggregate results if the firm is domiciled in that economy. This is because users typically associate firms with their economy of domicile rather than the economy where their primary listing is, if they are different. For example, the Bank of China has its primary listing in Hong Kong, but its economy of domicile is China so the Bank of China is included in the aggregation forecasts for China, and is included under China when searching for the individual PDs.

IV. EMPIRICAL ANALYSIS This section presents an empirical analysis of the CRI outputs for the 44 economies that are currently being covered. In Subsection 4.1, an overview is given of the default parameter estimates. Subsection 4.2 explains and provides the accuracy ratios for the different countries under the CRI cover.

4.1. Parameter Estimates With 24 months of forecast horizons, 13 variables and 6 different groups of economies, tables of the parameter estimates occupy over 20 pages and are not included in this Technical Report. They are available in the section on the technical report at the CRI web portal. In Figures B.1 and B.2, the parameter estimates

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

are from calibrations performed in July 2012 using data up until the end of June 2012. As an example, plots of the default parameters for the US are given in figures included in Figures B.1 and B.2 in Appendix B. In this part, a brief overview is given of the general traits and patterns seen in the default parameter estimations of the economies covered by the CRI. Recall that if a default parameter for a variable at a particular horizon is estimated to be positive (resp. negative) from the maximum pseudo-likelihood estimate, then an increasing value in the associated variable will lead to an increasing (decreasing) value of the forward intensity at that horizon, which in turn means an increasing (decreasing) value for the conditional default probability at that horizon. For the stock index one-year trailing return variable, most groups have default parameters that are slightly negative in the shorter horizons and then become positive in the longer horizons. When the equity market performs well, this is only a short-term positive for firms and in the longer term, firms are actually more likely to default. This seemingly counterintuitive result could be due to correlation between the market index and other firm-specific variables. For example, Duffie et al. (2009) suggested that a firm’s distance-to-default (DTD) can overstate its creditworthiness after a strong bull market. If this is the case, then the stock index return serves as a correction to the DTD levels at these points in time. The default parameters for the short-term interest rate variable are significantly positive at one- to two-year horizons for most of the economies. This is consistent with the intuition that increasing short-term interest rates typically signal increased funding costs for companies in the future, increasing the probability of default. The values at shorter horizons are varied between economies from slightly negative to significantly positive, possibly indicating different lead-lag relationships between credit conditions and the raising and cutting of short-term interest rates. DTD is a measure of the volatility-adjusted leverage of a firm. Low or negative DTD indicates high leverage and high DTD indicates low leverage. Therefore, PD would be expected to increase with decreasing DTD. Indeed, almost all of the calibrations for the different groups lead to negative default

30 August 2012 9:08 AM

parameters for the DTD level, with only China’s default parameter estimations hitting the constraint at zero for longer horizons. The ratio of the sum of cash and short-term investments to total assets (CASH/TA) measures liquidity of a firm. This indicates the availability of a firm’s funds and its ability to make interest and principal payments. As expected, for almost all economies (Indonesia being the only exception) the default parameters for CASH/TA level in shorter horizons are significantly negative. The magnitude of the default parameters decreases for longer horizons, indicating that CASH/TA level is a better indicator of a firm’s ability to make payments in the short term than the long term. The ratio of net income to total assets (NI/TA) measures profitability of a firm. The relationship between PD and NI/TA is as expected: the default parameters for NI/TA level is significantly negative for most economies and most horizons. The logarithm of the market capitalization of a firm over the median market capitalization of firms within the economy (SIZE) does not have a consistent effect on PD across different economies. For example, in the US the default parameters for SIZE level are negative for shorter horizons and positive for longer horizons, suggesting that the advantages enjoyed by larger firms, such as diversified business lines and funding sources, are a benefit in the shorter term but not in the longer term. On the other hand, in Japan the default parameters for SIZE level are negative across all horizons. These differences may reflect differences in the business environments in the respective economies. The default parameters associated with DTD Trend, CASH/TA Trend and SIZE Trend, are negative across almost all economies and horizons. The trend variables reflect momentum. The momentum effect is a short-term effect, and evidence of this is seen in the lower magnitude of the default parameters at longer horizons than at shorter horizons. The remaining trend variable is the NI/TA Trend. The current implementation of the CRI system retrieves net income only from annual financial statements. The default parameters for NI/TA Trend are constrained to be negative, but for most economies there is no clear relationship between the NI/TA Trend and the horizon. Once GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 129

129

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

NI/TA from quarterly statements can be used, this will likely be more informative. The ratio of the sum of market capitalization and total liabilities to total assets (M/B) can either indicate the market mis-valuation effect or the future growth effect. This default parameter is positive in most economies, indicating that higher M/B implies higher PD, and the market mis-valuation effect dominates. Shumway (2001) argued that a high level of the idiosyncratic volatility (SIGMA) indicates highly variable stock returns relative to the market index, indicating highly variable cash flows. Volatile cash flows suggest a heightened PD, and this finding is consistent across all economies and most horizons, with the exception of India.

4.2. Prediction Accuracy In-sample and out-of-sample testing: Various tests are carried out to test the prediction accuracy of the CRI PD forecasts. These tests are conducted either in-sample or out-of-sample. A single calibration is conducted for the in-sample tests, using data to the end of the data sample. As an example, one-year PD forecasts are made for December 31, 2000 by using the data at or before December 31, 2000 and the parameters from the calibration. These PD forecasts can be compared to actual defaults that occurred at any time in 2001. The out-of-sample analysis is done over time. The first calibration is conducted using only data up to the end of December 2000. For example, one-year PD forecasts can be made for December 31, 2000 using the data at or before December 31, 2000 with the parameters from this first calibration. These are PD forecasts that could have been made at the time, since the parameters are not based on data available after that date. This process is repeated every month. That is, the second calibration is conducted using only data up to the end of January 2001, and so on. It should be noted that for these repeated calibrations based on an expanding window of data, nothing else is changed besides the dataset. In other words, the same choice of input variables and the same choice of economy dummies within the groups are used throughout all of the calibrations. 130

1250007.indd 130

30 August 2012 9:08 AM

Some of the calibration groups have too few defaults in the period before December 2000 to be able to produce stable calibration results. If this is the case, the start date is advanced. Subsequently, if there are too few defaults after the start date to perform meaningful tests, only in-sample tests are performed for that calibration group. Out-of-sample tests are performed for (starting month of calibration in parentheses): China (12/2000), Japan (12/2003), India (12/2001), South Korea (12/2000), Developed Asia (12/2000), Emerging Markets (12/2000), North America group (12/2000), Europe group (12/2002). Accuracy Ratio: The accuracy ratio (AR) is one of the most popular and meaningful tests of the discriminatory power of a rating system (BCBS, 2005). The AR and the equivalent Area Under the Receiver Operating Characteristic (AUROC) are described in Duan and Shrestha (2011). In short, if defaulting firms had been assigned among the highest PD of all firms before they defaulted, then the model has discriminated well between “safe” and “distressed” firms. This leads to higher values of AR and AUROC. The range of possible AR values is in [0,1], where 0 is a completely random rating system and 1 is a perfect rating system. The range of possible AUROC values is in [0.5, 1]. AUROC and AR values are related by: AR = 2 × AUROC − 1. The AR and AUROC values for different horizons are available in Table B.1 of this technical report. Both in-sample and out-of-sample results are available for calibration groups where out-of-sample testing could be performed. Other calibration groups include only insample results. The in-sample AR and AUROC are computed only from the starting date of the corresponding out-of-sample tests, so that the results between insample and out-of-sample are comparable. Only economies with more than 20 defaults entering into the AR and AUROC computation are listed. The PD are taken to be non-overlapping. For example, the one-year AR is based on PDs computed on 31/12/2000, 31/12/2001, … , 31/12/2009 and firms defaulting within one year of those dates, while the two-year AR is based on PDs computed on 31/12/2000, 31/12/2002, … , 31/12/2008 and firms defaulting within two years of those dates. The AUROC values have been provided only for the purpose of comparison, if other rating systems

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

report their results in terms of AUROC. The discussion will focus only on AR. The model is able to achieve strong AR results mostly greater than 0.80 at the one and six-month horizons for developed economies. There is a drop in AR at one and two-year horizons, but the AR are still mostly acceptable. Australia, the UK and Singapore have sharp drops in AR at the two-year horizon. Hong Kong has comparatively worse AR over all horizons as compared to other developed economies. The AR in emerging market economies such as China, India, Indonesia, Malaysia, Philippines and Thailand are noticeably weaker than the results in the developed economies. This can be due to a number of issues. The quality of data is worse in emerging markets, in terms of availability and data errors. This may be due to lower reporting and auditing standards. Also, variable selection is likely to play a more important role in emerging markets. The variables were selected based on the predictive power in a developed economy, the US. Performing variable selections specific to the calibration group are expected to improve predictive accuracy, especially in emerging market economies. Finally, there could be structural differences in how defaults and bankruptcies occur in emerging market economies. If the judicial system is weak and there are no repercussions for default, firms may be less reluctant to default. The AR for the Latin American economies inside the emerging economies group are generally greater than 0.80 at horizons shorter than one year. However, these AR are for a small number of defaults. At horizons of one and six-months, out-of-sample AR are comparable to their in-sample counterparts. At horizons of one and two-years, out-of-sample AR can be substantially lower than the in-sample AR. Finally, the US has a sufficient number of financial firms and financial defaults to produce separate AR and AUROC. These are also listed in Table B.1 as outof-sample results. The financial sector ARs are actually stronger than the non-financial sector AR. This is achieved by having only minimal differences between how financial and non-financial firms are treated. The AR is a test of discriminatory power, or how well the rating system ranks firms in terms of credit worthiness. In a separate article included in the GCR

30 August 2012 9:08 AM

Volume 2, we provide a more qualitative check on the CRI PD in which we compare the behaviour of CRI PD to the rating actions of external credit rating agencies such as Moody’s and S&P for some well known default cases. Aggregate defaults: The time series of aggregate predicted number of defaults and actual number of defaults in each calibration group are also available in Figure B.3.

V. ONGOING DEVELOPMENTS The CRI can be developed along a number of directions. We now comment on obvious ones that in our view are likely to bring meaningful and measurable benefits. Besides modifications to the current modeling framework of the forward intensity, a change in modeling platform will be undertaken if another model proves more promising in terms of accuracy and robustness of results. For this type of development we also rely on the collective efforts by the worldwide credit research community to challenge and improve the existing modeling platform. The current CRI default prediction model is based on the econometric platform of modeling forward intensities developed by Duan et al. (2012). As noted by them, the forward-intensity model exhibits systematic bias in predicting longer-term defaults of the US corporate sector. In general, it overestimates (underestimates) defaults when default rates were low (high). Introducing a frailty variable to the model to capture default contagion appears to be one possibility to further improve the model. In addition, some of the likely future developments of CRI fall in the domain of further infrastructure developments at RMI. For example, by end 2012, all exchange-listed firms in all economies around the globe should be covered. Furthermore, in terms of variable selection, more experiments are needed to identify common risk factors and RMI specific attributes that are more indicative of defaults in different economies. Also in terms of grouping, further tests should be conducted, especially as new economies will be covered. It is also worth noting that all variables used thus far in the CRI implementation are the quantitative type. Soft GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 131

131

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

credit information as reflected in qualitative opinions of credit analysts may add an important dimension to future improvement. To this end, the CRI has been conducting a continuous credit analysts survey, and at this point of writing there are about 100 analysts participating in the survey. It is quite obvious that we have to expand this base of this survey in order to allow meaningful incorporation into the default prediction. ******************************************* The RMI Credit Research Initiative is premised on the concept of credit ratings as a “public good”. Being a non-profit undertaking allows a high level of transparency and collaboration that other commercial credit rating systems can not replicate. The research and support infrastructure is in place and researchers from around the world are invited to contribute to this initiative. Any methodological improvements that researchers develop will be incorporated into the CRI system. In essence, the initiative operates as a “selective wikipedia” where many can contribute but implementation control is retained. If you have feedback on this technical report or wish to work with us in this endeavor, please contact us at [email protected]

REFERENCES Anderson, L., J. Sidenius and S. Basu (2003), All your Hedges in one Basket. Risk, pp. 67–72. Crosbie, P. and J. Bohn (2003), Modeling Default Risk. Moody’s KMV technical document.

132

1250007.indd 132

30 August 2012 9:08 AM

Duan, J.-C. (1994), Maximum Likelihood Estimation Using Price Data of the Derivative Contract. Mathematical Finance, 4, pp. 155–167. Duan, J.-C. (2000), Correction: “Maximum Likelihood Estimation Using Price Data of the Derivative Contract”. Mathematical Finance, 10, pp. 461–462. Duan, J.-C. (2010), Clustered Defaults. National University of Singapore Working Paper. Duan, J.-C. and K. Shrestha (2011), Statistical Credit Rating Methods, Global Credit Review, pp. 43–64. Duan, J.-C., J. Sun and T. Wang (2012), Multiperiod Corporate Default Prediction — A Forward Intensity Approach. Journal of Econometrics, forthcoming (DOI: 10.1016/j.jeconom.2012.05.002). Duan, J.-C. and E. Van Laere (2012), A Public Good Approach to Credit Ratings: From Concept to Reality. Journal of Banking & Finance, forthcoming (DOI: 10.1016/j.jbankfin.2012.03.012). Duan, J.-C. and T. Wang (2012), Measuring Distance-toDefault for Financial and Non-financial Firms. Global Credit Review, pp. 95–108. Duffie, D., A. Eckner, G. Horel and L. Saita (2009), Frailty Correlated Default. Journal of Finance, 64, pp. 2089–2123. Duffie, D., L. Saita, and K. Wang (2007), Multi-Period Corporate Default Prediction with Stochastic Covariates. Journal of Financial Economics, 83, pp. 635–665. Merton, R. C. (1974), On the Pricing of Corporate Debt: The Risk Structure of Interest Rates. Journal of Finance, 29, pp. 449–470. Shumway, T. (2001), Forecasting Bankruptcy More Accurately: A Simple Hazard Model. Journal of Business, 74, pp. 101–124.

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

APPENDIX A Table A.1 ISO codes for economies currently covered by the CRI and the group that each economy is calibrated in. ISO Code

Economy

Calibration Group

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Argentina Australia Austria Belgium Brazil Canada Switzerland Chile China Colombia Cyprus Germany Denmark Spain Estonia Finland France United Kingdom Greece Hong Kong Indonesia India Ireland Iceland Italy Japan South Korea Luxemburg Mexico Malta Malaysia Netherlands Norway Peru Philippines Portugal Singapore Slovakia Slovenia Sweden Thailand Taiwan United States Venezuela

Emerging Developed Asia-Pacific Europe Europe Emerging North America Europe Emerging China Emerging Europe Europe Europe Europe Europe Europe Europe Europe Europe Developed Asia-Pacific Emerging India Europe Europe Europe Developed Asia-Pacific Developed Asia-Pacific Europe Emerging Europe Emerging Europe Europe Emerging Emerging Europe Developed Asia-Pacific Europe Europe Europe Emerging Developed Asia-Pacific North America Emerging

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 133

133

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table A.2 The stock indices used for each economy in computing the first common variable. Country

Stock Exchange

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP

Buenos Aires Stock Exchange Merval Index All Ordinaries Index Austrian Traded ATX Index Belgian All Shares Return Index Brazil Bovespa Stock Index S&P/TSX Composite Index SPI Swiss Performance Index Santiago Stock Exchange IPSA Index Shanghai Stock Exchange Composite Index FTSE All World Series Colombia Local Cyprus Stock Exchange General Index Cyprus Stock Exchange General CDAX Performance Index OMX Copenhagen 20 Index IBEX 35 Index OMX Tallinn OMXT OMX Helsinki Index CAC 40 Index FTSE 100 Index Athex Composite Share Price Index Hang Seng Index Jakarta Composite Index BSE Sensex 30 Index Irish Overall Index OMX Iceland All-Share Price Index Italy Stock Market BCI Comit Global Nikkei 500 KOSPI Index Luxembourg Stock Exchange LuxX Index Luxembourg Stock Exchange13 ‘Dead’ Mexico Bolsa Index Malta Stock Exchange FTSE Bursa Malaysia KLCI AEX Index OBX Price Index Bolsa de Valores de Lima General Sector Index PSEI-Philippine Stock Exchange Index PSI General Index Straits Times Index Straits Times Old Index Slovak Share Index HSBC Slovenia Dollar OMX Stockholm All-Share Index Stock Exchange of Thailand Index Taiwan Taiex Index S&P 500 Index Caracas Stock Exchange Stock Market Index

DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Period Used

9/3/2004–Present 4/2/1996–9/2/2004

1/4/1999–Present 1/2/1998–1/3/1999

1/10/2008–Present 8/31/1999–1/9/2008

*A blank Period Used column indicates that there is only a single index that is used throughout the whole period.

134

1250007.indd 134

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table A.3 The interest rates used for each economy as the second common variable. Country

Short Term Interest Rate

ARG AUS AUT

Argentina Deposit 90 Day Australia Dealer Bill 90 Day Germany 3 Month Bubill — Germany 3 Month Bubill — Andima Brazil Govt Bond Fixed Rate 3 Months Brazil CDB (up to 30 Days) Canada Treasury Bill 3 Month — Chile TAB UF Interbank Rate 90 Days China Time Deposit Rate 3 Month Colombia CD Rate 90-Day Germany 3 Month Bubill — Germany 3 Month Bubill Germany Interbank 3 Month Denmark Interbank 3 Month Germany 3 Month Bubill — Germany 3 Month Bubill — Germany 3 Month Bubill — Germany 3 Month Bubill — UK Treasury Bill Tender 3 Month Germany 3 Month Bubill — Hong Kong Exchange Fund Bill 3 Month Indonesia SBI 90 Day Indonesia SBI/DISC 90 Day India T-Bill Secondary 91 Day Germany 3 Month Bubill — — Germany 3 Month Bubill — Japan Treasury Discount Bills 3 Month Japanese Government Bond Interest Rate-1 Year Maturity Korea Commercial Paper 91 Day Germany 3 Month Bubill — Mexico Cetes 2ND MKT. 90 Day Mexico Cetes 91 Dat AVG.RET.AT AUC. Germany 3 Month Bubill — Malaysia Deposit 3 Month Germany 3 Month Bubill — Norway Govt Treasury Bills 3 Month Norway Interbank 3 Month(effective)

BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR

Period Used

1/1/1999–Present –12/31/1998 1/1/1999–Present –12/31/1998 4/3/2000–Present 10/10/1994–3/31/2000

1/1/2008–Present –12/31/2007 5/25/1993–Present 1/2/1986–5/24/1993 1/1/1999–Present –12/31/1998 1/1/2001–Present –12/31/2010 1/1/1999–Present –12/31/1998 1/1/1999–Present –12/31/1998 1/1/2001–Present –12/31/2000 7/10/2003–Present 1/1/1985–7/9/2003 1/1/1999–Present –12/31/1998 1/1/1999–Present –12/31/1998 7/10/1992–Present 9/24/1974–7/9/1992 1/1/1999–Present –12/31/1998 6/26/1996–Present 3/9/1989–6/25/1996 1/1/2008–Present –12/31/2007 1/1/1999–Present –12/31/1998 6/27/1995–Present 1/2/1986–6/26/1995 (Continued)

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 135

135

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

Table A.3 Country PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

30 August 2012 9:08 AM

(Continued )

Short Term Interest Rate Peru Savings Rate Philippine Treasury Bill 91 Day Germany 3 Month Bubill — Singapore T-Bill 3 Month Germany 3 Month Bubill — Germany 3 Month Bubill — Sweden T-Bill 3 Month Sweden Treasury Bill 90 Day Thailand Repo 3 Month(BOT) Taiwan Money Market 90 Day US Generic Govt 3-Month Yield Venezuela Overnight

Period Used

1/1/1999–Present –12/31/2008 1/1/2009–Present –12/31/2008 1/1/2007–Present –12/31/2006 5/25/1993–Present 4/25/1989–5/24/1993

*A blank Period Used column indicates that there is only a single interest rate that is used throughout the whole period.

Table A.4 The interest rates used for each economy in the DTD calculation. Country ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP

EST FIN FRA

Interest Rate Name Argentina Deposit 90 Day (PA.) Australia Govt. Bonds Generic Mid Yield 1 Year German Government Bonds 1 Year BKO Austria VIBOR 12 Month German Government Bonds 1 Year BKO Belgium Treasury Bill 1 Year Andima Brazil Govt Bond Fixed Rate 1 Year BRAZIL CDB (UP TO 30 DAYS) Canada Treasury Bill 1 Year Swiss Interbank 1 Year (ZRC:SNB) Chile TAB UF Interbank Rates 360 Days Chile TAB UF Interbank Rate 90 Days China Household Savings Deposits 1-Year Rate Colombia Government Generic Bond 1 Year Yield Colombia CD Rate 360-Dat Cyprus Treasury Bill Rate — 13 Week German Government Bonds 1 Year BKO Germany Interbank 12 Month Denmark Government Bonds 1 Year Note Generic Bid Yield Denmark Euro-Krone 1 Year(FT/ICAP/TR) German Government Bonds 1 Year BKO Spain 12 Month Treasury Bill Yield Spain Interbank 12 Month Estonia, Interest Rates, Prices, Production & Labour, Interest Rates, Deposit Rate German Government Bonds 1 Year BKO Finland Interbank Close 12 Month German Government Bonds 1 Year BKO France Treasury Bill 12 Months

Period Used

1/1/1999–Present 6/10/1991–12/31/1998 1/1/1999–Present 4/2/1991–12/31/1998 4/3/2000–Present 10/10/1994–3/31/2000

8/1/1996–Present 11/2/1992–7/30/1996 3/1/2001–Present 7/12/1993–2/8/2001 1/10/1995–Present 11/2/1990–1/9/1995 6/1/2008–Present 6/14/1985–5/31/2008 1/1/1999–Present 11/30/1992–12/31/1998 12/19/1991–11/29/1992

1/1/1999–Present 4/2/1992–12/31/1998 1/1/1999–Present 1/3/1989–12/31/1998 (Continued)

136

1250007.indd 136

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

Table A.4 Country GBR GRC HKG IDN IND IRL ISL

ITA

JPN KOR LUX MEX MLT MYS

NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

30 August 2012 9:08 AM

(Continued )

Interest Rate Name UK Govt. Bonds 1 Year Note Generic UK Govt. Liability Nominal Spot Curve 12 Month German Government Bonds 1 Year BKO Greece Treasury Bill 1 Year HKMA Hong Kong Exchange Fund Bill 12 Month Indonesia SBI 90 Day Indonesia SBI/DISC 90 Day India T-Bill Secondary 1 Year UK Govt. Liability Nominal Spot Curve 12 Month Iceland Interbank 12 Month Iceland Interbank 3 Month Iceland 90-day CB Notes German Government Bonds 1 Year BKO Italy Bots Treasury Bill 12 Month Gross Yields Italy T-Bill Auction Gross 12 Month Japan Treasury Bills 12 Month Japanese Government Bond Interest Rate-1 Year Maturity Korea Monetary Stabilization Bonds 1 Year Long Term Government Bond Yields — Maastricht Definition (Avg.) Mexico Cetes 2ND MKT. 360 Day Mexico Cete 91 DAY AVG.RET.AT AUC. Long Term Government Bond Yields — Maastricht Definition (Avg.) Bank Negara Malaysia 1 Year Govt. Securities Indicative YTM Malaysia Deposit 1 Year German Government Bonds 1 Year BKO Netherland Interbank 1 Year Norway Govt. Treasury Bills 12 Month Norway Interbank 1 Year Peru Savings Rate Philippine Treasury Bill 364 Day German Government Bonds 1 Year BKO Portugal 1-Year-LISBOR-Act/365 Day convention Singapore T-Bill 3 Month Slovak Rep. Interbank 1 Year Slovenia Treasury Bill 3 Month ‘Dead’ Sweden Interbank 1 Year Sweden Treasury Bill 1 Year Note Thailand Govt. Bond 1 Year Note Thailand Deposit 12 Month(KT) Taiwan Deposit 12 Month US Treasury Constant Maturities 1 Year Venezuela Overnight

Period Used 9/12/2001–Present 12/13/1985–9/11/2001 1/1/2001–Present 1/2/1990-12/31/2000 7/10/2003–Present 1/1/1985–7/9/2003

2/1/2000–Present 8/4/1998–1/31/2000 5/12/1987–8/3/1998 1/1/1999–Present 9/5/1994–12/31/1998 3/31/1987–9/4/1994 12/14/1999–Present 9/24/1979–12/13/1999

6/26/1996–Present 3/9/1989–6/25/1996

6/21/2005–Present 1/1/1985–6/20/2005 1/1/1999–Present 1/2/1987–12/31/1998 7/1/1997–Present 1/2/1986–6/30/1997

1/1/1999–Present 8/16/1993–12/31/1998

5/25/1993–Present 4/25/1989–5/24/1993 8/7/2000–Present 1/2/1991–8/6/2000

*A blank Period Used column indicates that there is only a single interest rate that is used throughout the whole period.

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 137

137

8/30/2012 9:36:35 AM

b1423

Table A.5

Global Credit Review Volume 2

30 August 2012 9:08 AM

Summary statistics of input variables (based on data from Jan 1991 to June 2012). DTD Level

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Min

25%

Median

75%

Max

Mean

StdDev

# Observations

−1.75 −1.41 −2.95 −2.95 −1.86 −1.13 −2.95 −0.81 0.03 −1.03 −1.27 −2.95 −2.95 −2.95 −0.30 −2.95 −2.95 −2.95 −2.95 −1.41 −1.86 −1.61 −1.74 −1.48 −2.95 −1.41 −1.41 −0.17 −1.86 −0.65 −1.86 −2.95 −2.95 −1.86 −1.86 −2.95 −1.19 −0.24 −2.47 −2.95 −1.71 −1.41 −1.13 −1.86

1.69 1.84 1.90 2.51 0.93 1.92 2.65 3.51 3.05 2.26 0.88 1.61 1.82 2.20 1.45 2.26 1.80 2.19 1.48 1.48 0.56 0.82 1.97 1.75 1.61 2.05 1.20 3.19 1.99 2.32 1.57 2.49 1.24 1.87 1.05 1.18 1.55 1.40 2.18 1.75 1.58 2.60 1.78 0.37

2.88 2.98 3.12 4.42 2.45 3.26 4.05 5.18 4.11 3.77 1.55 2.89 3.15 3.58 2.50 3.44 3.00 3.55 2.49 2.51 1.53 1.70 3.30 2.93 2.85 3.10 2.11 4.94 3.58 3.62 2.84 4.03 2.37 3.00 2.19 2.35 2.65 2.27 3.68 3.04 2.81 3.68 3.02 1.36

4.05 4.21 5.05 6.82 4.80 4.92 5.82 6.84 5.64 5.53 2.47 4.39 4.72 5.11 3.94 4.95 4.58 5.29 3.81 3.94 2.59 2.79 4.86 4.25 4.36 4.44 3.26 7.62 5.41 5.19 4.55 5.86 3.79 4.44 3.54 3.84 4.30 3.18 5.89 4.55 4.26 4.99 4.63 2.31

19.86 17.73 25.82 25.82 23.38 25.00 23.69 23.38 16.88 20.47 23.81 25.82 25.82 25.82 11.20 20.19 25.82 25.82 23.59 17.73 23.38 15.42 13.63 20.01 25.82 17.73 17.73 24.75 23.38 15.32 23.38 25.82 20.49 22.72 23.38 20.09 17.73 25.82 16.88 25.82 23.38 17.73 25.00 13.37

3.10 3.30 4.12 5.08 3.36 3.69 4.44 5.60 4.59 4.13 2.07 3.32 3.60 4.01 3.08 3.80 3.47 4.03 2.82 3.00 1.83 2.04 3.53 3.22 3.19 3.49 2.45 6.14 4.04 4.35 3.46 4.47 2.62 3.44 2.56 2.81 3.17 2.79 4.10 3.38 3.19 3.98 3.49 1.38

2.25 2.20 4.29 3.81 3.68 2.55 2.83 3.45 2.27 2.83 2.30 2.71 2.85 3.04 2.38 2.38 2.74 2.77 2.14 2.28 2.02 1.97 2.24 2.25 2.49 2.15 2.03 4.44 3.08 3.37 2.84 3.09 2.05 2.50 2.21 2.41 2.33 3.38 2.98 2.46 2.48 2.15 2.60 1.73

10999 261380 20524 27834 43815 202772 50531 26644 232944 5378 16110 166977 41748 33473 766 27101 154055 356876 52254 186450 50991 399715 9735 4274 56269 744685 251921 2797 16289 842 170119 35057 40996 10507 33701 13055 104474 776 5797 73447 84238 202372 1421457 3503 (Continued)

138

1250007.indd 138

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:35 AM

b1423

Global Credit Review Volume 2

Table A.5

30 August 2012 9:08 AM

(Continued ) DTD Trend

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Min

25%

Median

75%

Max

Mean

StdDev

# Observations

−7.83 −5.47 −8.09 −8.09 −7.83 −6.37 −8.09 −7.83 −6.13 −7.83 −8.09 −8.09 −8.09 −8.09 −2.26 −8.09 −8.09 −8.09 −8.09 −5.47 −7.83 −5.87 −6.47 −8.09 −8.09 −5.47 −5.47 −8.09 −7.67 −6.66 −7.83 −8.09 −8.09 −7.83 −7.83 −8.09 −5.47 −8.03 −5.14 −8.09 −7.83 −5.47 −6.37 −6.72

−0.54 −0.48 −0.58 −0.64 −0.45 −0.54 −0.61 −0.68 −0.57 −0.45 −0.36 −0.51 −0.52 −0.50 −0.01 −0.47 −0.48 −0.57 −0.55 −0.49 −0.31 −0.35 −0.52 −0.74 −0.58 −0.47 −0.44 −0.67 −0.46 −0.62 −0.49 −0.66 −0.44 −0.45 −0.35 −0.46 −0.48 −0.21 −0.68 −0.49 −0.51 −0.57 −0.47 −0.27

−0.02 −0.02 −0.03 −0.04 0.00 −0.02 0.00 0.04 −0.02 0.03 −0.07 −0.04 −0.01 0.01 0.12 0.03 −0.00 −0.02 −0.09 −0.01 0.01 −0.02 0.00 −0.07 −0.04 −0.02 −0.01 0.00 0.05 −0.10 −0.02 −0.03 −0.01 0.00 0.00 −0.04 −0.02 0.02 −0.12 −0.03 −0.01 −0.03 0.00 0.00

0.42 0.38 0.44 0.55 0.40 0.44 0.59 6.56 0.48 0.66 0.15 0.40 0.43 0.48 0.47 0.52 0.44 0.40 0.32 0.42 0.31 0.33 0.43 0.39 0.46 0.41 0.38 0.52 0.61 0.54 0.41 0.55 0.36 0.50 0.32 0.33 0.39 0.37 0.22 0.42 0.43 0.49 0.43 0.31

6.69 5.08 7.62 7.62 6.69 5.34 7.62 6.69 5.51 6.69 7.62 7.62 7.62 7.62 3.70 7.62 7.62 7.62 7.62 5.08 6.69 4.82 7.25 6.76 7.62 5.08 5.08 7.62 6.69 4.26 6.69 7.62 7.62 6.69 6.69 7.34 5.08 7.62 7.62 7.62 6.69 5.08 5.34 6.69

−0.05 −0.06 −0.15 −0.06 −0.15 −0.06 −0.01 −0.05 −0.07 0.08 −0.13 −0.06 −0.05 −0.01 0.26 0.02 −0.03 −0.10 −0.11 −0.04 −0.01 −0.01 −0.07 −0.18 −0.07 −0.03 −0.04 −0.11 0.08 0.01 −0.05 −0.06 −0.05 0.02 −0.02 −0.06 −0.06 0.09 −0.19 −0.04 −0.05 −0.05 −0.03 0.01

1.03 0.98 1.63 1.49 1.44 1.10 1.26 1.46 1.05 1.29 0.75 1.10 1.19 1.28 0.70 1.05 1.09 1.26 0.97 0.97 0.76 0.78 1.02 1.35 1.13 0.87 0.87 1.43 1.12 1.36 1.07 1.25 0.90 1.17 0.88 0.94 0.96 1.09 1.06 1.03 1.03 0.98 0.97 0.73

10999 261380 20521 27833 43815 202763 50531 26644 232944 5247 16104 166943 41701 33473 590 27100 153986 356876 52254 186450 50991 399485 9735 4273 56269 744685 251918 2791 16206 783 170119 35055 40933 10441 33701 13055 104474 681 5777 73434 84238 202372 1421430 3368 (Continued)

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 139

139

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

Table A.5

30 August 2012 9:08 AM

(Continued )

CASH/TA Level

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Min

25%

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0.02 0.04 0.03 0.03 0.02 0.01 0.05 0.01 0.08 0.03 0.01 0.02 0.03 0.02 0.03 0.03 0.03 0.03 0.02 0.06 0.03 0.01 0.04 0.02 0.03 0.07 0.04 0.04 0.03 0.03 0.02 0.02 0.04 0.01 0.02 0.01 0.06 0.03 0.01 0.04 0.02 0.05 0.03 0.04

Median 0.05 0.13 0.07 0.08 0.07 0.06 0.10 0.03 0.14 0.06 0.05 0.07 0.09 0.05 0.05 0.08 0.08 0.09 0.06 0.14 0.08 0.03 0.09 0.04 0.07 0.13 0.09 0.11 0.06 0.08 0.07 0.05 0.09 0.04 0.08 0.03 0.13 0.05 0.04 0.09 0.06 0.11 0.07 0.07

75%

Max

Mean

StdDev

# Observations

0.10 0.35 0.15 0.19 0.16 0.21 0.20 0.08 0.24 0.10 0.15 0.20 0.18 0.11 0.12 0.16 0.17 0.22 0.13 0.26 0.17 0.07 0.21 0.08 0.14 0.22 0.18 0.19 0.12 0.22 0.16 0.13 0.20 0.13 0.18 0.07 0.24 0.12 0.09 0.22 0.14 0.21 0.24 0.18

0.69 0.97 0.96 0.99 0.91 0.99 0.99 0.91 0.88 0.84 0.91 0.99 0.99 0.73 0.53 0.99 0.99 0.99 0.83 0.97 0.90 0.82 0.97 0.53 0.99 0.97 0.97 0.97 0.77 0.50 0.91 0.99 0.99 0.91 0.91 0.55 0.97 0.57 0.41 0.99 0.88 0.94 0.99 0.91

0.07 0.23 0.11 0.15 0.11 0.15 0.15 0.06 0.18 0.08 0.11 0.15 0.14 0.08 0.09 0.13 0.13 0.17 0.10 0.19 0.12 0.06 0.15 0.06 0.10 0.17 0.13 0.16 0.08 0.14 0.11 0.10 0.16 0.09 0.13 0.06 0.17 0.08 0.07 0.16 0.10 0.15 0.18 0.12

0.08 0.25 0.13 0.18 0.13 0.21 0.16 0.09 0.15 0.10 0.14 0.18 0.17 0.10 0.09 0.14 0.14 0.21 0.11 0.17 0.12 0.09 0.17 0.06 0.12 0.13 0.13 0.17 0.08 0.14 0.13 0.13 0.18 0.11 0.15 0.08 0.15 0.08 0.07 0.19 0.12 0.14 0.22 0.11

11240 267774 21080 29341 45844 203329 51597 27864 234610 5659 16620 170235 42980 36439 2265 27646 156142 358220 53302 190587 53184 562091 9954 4654 57508 745312 253233 3165 17301 1165 171192 35909 41551 10810 34543 13769 105424 1216 6321 73727 85166 202800 1458948 3776 (Continued)

140

1250007.indd 140

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

Table A.5

30 August 2012 9:08 AM

(Continued )

CASH/TA Trend

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Min

25%

Median

75%

Max

Mean

StdDev

# Observations

−0.33 −0.42 −0.48 −0.48 −0.33 −0.44 −0.48 −0.33 −0.30 −0.33 −0.48 −0.48 −0.48 −0.48 −0.25 −0.48 −0.48 −0.48 −0.46 −0.42 −0.33 −0.36 −0.48 −0.34 −0.48 −0.42 −0.42 −0.38 −0.29 −0.32 −0.33 −0.48 −0.48 −0.33 −0.33 −0.40 −0.42 −0.13 −0.25 −0.48 −0.33 −0.42 −0.44 −0.19

−0.01 −0.03 −0.01 −0.01 −0.01 −0.02 −0.01 −0.01 −0.02 −0.01 −0.01 −0.01 −0.01 −0.01 −0.01 −0.01 −0.01 −0.02 −0.01 −0.02 −0.01 –0.00 −0.01 −0.01 −0.01 −0.01 −0.02 −0.01 −0.01 −0.01 −0.01 −0.01 −0.02 −0.01 −0.01 −0.01 −0.02 −0.01 −0.01 −0.02 −0.01 −0.02 −0.02 −0.01

0.00 –0.00 0.00 0.00 0.00 0.00 0.00 −0.00 −0.00 0.00 0.00 0.00 0.00 0.00 0.00 −0.00 0.00 0.00 −0.00 0.00 0.00 0.00 0.00 0.00 0.00 −0.00 −0.00 0.00 0.00 0.00 0.00 0.00 −0.00 0.00 0.00 0.00 0.00 0.00 0.00 −0.00 −0.00 0.00 −0.00 0.00

0.01 0.01 0.00 0.00 0.01 0.01 0.01 0.00 0.01 0.01 0.00 0.00 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.00 0.01 0.00 0.01 0.01 0.01 0.00 0.01 0.00 0.01 0.00 0.01 0.01 0.01 0.00 0.01 0.00 0.00 0.01 0.01 0.02 0.01 0.00

0.37 0.44 0.46 0.46 0.37 0.42 0.46 0.37 0.30 0.37 0.46 0.46 0.46 0.46 0.17 0.46 0.46 0.46 0.46 0.44 0.37 0.35 0.46 0.40 0.46 0.44 0.44 0.25 0.37 0.18 0.37 0.46 0.46 0.37 0.37 0.46 0.44 0.15 0.28 0.46 0.37 0.44 0.42 0.31

00.0 −0.01 00.0 −0.00 0.00 −0.00 −0.00 0.00 −0.01 0.00 −0.00 0.00 −0.00 −0.00 −0.00 −0.00 –0.00 −0.01 −0.00 –0.00 −0.00 0.00 −0.00 −0.00 –0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 0.00 −0.01 0.00 −0.00 –0.00 −0.00 −0.00 −0.00 −0.01 −0.00 0.00 −0.00 –00.0

0.04 0.09 0.04 0.05 0.05 0.07 0.04 0.04 0.05 0.04 0.05 0.06 0.06 0.04 0.04 0.05 0.04 0.07 0.05 0.07 0.04 0.04 0.05 0.03 0.04 0.04 0.06 0.04 0.03 0.03 0.04 0.04 0.06 0.04 0.05 0.03 0.06 0.03 0.03 0.07 0.04 0.05 0.06 0.03

11240 267703 21069 29334 45807 203310 51591 27859 234588 5650 16593 170166 42969 36429 2258 27643 156028 358147 53299 190587 53184 555409 9954 4644 57502 745312 253233 3135 17275 1165 171187 35901 41536 10806 34523 13727 105406 1212 6319 73717 85166 202800 1458777 3760 (Continued) GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 141

141

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

Table A.5

30 August 2012 9:08 AM

(Continued ) NI/TA Level

ARG* AUS AUT BEL BRA* CAN CHE CHL* CHN COL* CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN* IND* IRL ISL ITA JPN KOR LUX MEX* MLT MYS* NLD NOR PER* PHL* PRT SGP SVK SVN SWE THA* TWN USA VEN*

Min

25%

Median

75%

Max

Mean

StdDev

# Observations

−0.04 −0.44 −0.50 −0.35 −0.04 −0.35 −0.50 −0.04 −0.22 −0.04 −0.50 −0.50 −0.50 −0.50 −0.09 −0.22 −0.50 −0.50 −0.50 −0.44 −0.04 −0.04 −0.50 −0.13 −0.14 −0.44 −0.44 −0.36 −0.04 −0.14 −0.04 −0.50 −0.50 −0.04 −0.04 −0.22 −0.44 −0.02 −0.06 −0.50 −0.04 −0.22 −0.35 −0.04

−0.00 −0.00 0.00 0.00 −0.00 −0.00 0.00 0.00 0.00 0.00 −0.00 −0.00 0.00 0.00 0.00 0.00 0.00 −0.00 −0.00 0.00 0.00 0.00 0.00 0.00 −0.00 0.00 −0.00 0.00 0.00 0.00 0.00 0.00 −0.00 0.00 −0.00 −0.00 0.00 0.00 0.00 −0.00 0.00 0.00 –0.00 0.00

0.00 −0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0.01 0.00 0.00 0.01 0.01 0.00 0.00 0.01 0.01 0.01 0.00 0.00 0.00 0.00 0.01 0.01 0.00 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.00 0.00 0.01 0.01 0.01 0.00 0.01 0.01 0.01 0.01 0.01 0.00 0.01 0.00 0.00 0.01 0.01 0.01 0.01 0.01

0.03 0.10 0.07 0.07 0.03 0.19 0.07 0.03 0.08 0.03 0.07 0.07 0.07 0.07 0.05 0.07 0.07 0.07 0.07 0.10 0.03 0.03 0.07 0.02 0.07 0.10 0.10 0.07 0.03 0.04 0.03 0.07 0.07 0.03 0.03 0.06 0.10 0.03 0.02 0.07 0.03 0.10 0.19 0.03

0.00 −0.02 −0.00 0.00 0.00 −0.01 0.00 0.00 0.00 0.00 −0.00 −0.00 −0.00 0.00 0.00 0.00 0.00 −0.01 0.00 –0.00 0.00 0.00 0.00 0.00 0.00 0.00 −0.00 –0.00 0.00 0.00 0.00 0.00 −0.00 0.01 0.00 0.00 0.00 0.00 0.00 −0.01 0.00 0.00 –0.00 0.00

0.01 0.05 0.02 0.01 0.01 0.03 0.02 0.01 0.01 0.01 0.03 0.02 0.02 0.03 0.01 0.01 0.02 0.04 0.01 0.03 0.01 0.01 0.02 0.01 0.01 0.01 0.02 0.04 0.01 0.01 0.01 0.02 0.03 0.01 0.01 0.01 0.02 0.01 0.01 0.03 0.01 0.01 0.03 0.01

11240 267973 21107 29384 45874 203417 51661 27900 234634 5659 16723 170624 42986 36449 2288 27658 156598 358414 53311 190589 53188 565026 9955 4670 57539 745501 256265 3207 17318 1165 171226 35940 41568 10820 34554 13841 105459 1225 6321 73761 85166 202800 1458082 3799

* Winsorization levels are at 1 and 99 percentiles instead of 0.1 and 99.9 percentiles.

(Continued) 142

1250007.indd 142

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table A.5 (Continued ) NI/TA Trend

ARG* AUS AUT BEL BRA* CAN CHE CHL* CHN COL* CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN* IND IRL ISL ITA JPN KOR LUX MEX* MLT MYS* NLD NOR PER* PHL* PRT SGP SVK SVN SWE THA* TWN USA VEN*

Min

25%

Median

75%

Max

Mean

StdDev

−0.03 −0.35 −0.30 −0.30 −0.03 −0.29 −0.30 −0.03 −0.17 −0.03 −0.30 −0.30 −0.30 −0.30 −0.30 −0.20 −0.30 −0.30 −0.30 −0.35 −0.03 −0.11 −0.30 −0.12 −0.30 −0.35 −0.35 −0.08 −0.03 −0.04 −0.03 −0.30 −0.30 −0.03 −0.03 −0.30 −0.35 −0.05 −0.06 −0.30 −0.03 −0.35 −0.29 −0.03

−0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 –0.00 −0.00 −0.00 −0.00 −0.00 −0.00 –0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 –0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 −0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 −0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 −0.00 −0.00 0.00 0.00 0.00 0.00 0.00 00.0 0.00 0.00 0.00 −0.00 −0.00 –0.00 0.00

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 –0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 –0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 –0.00 –0.00 0.00 0.00 0.00 0.00 –0.00 0.00 0.00 0.00 0.00 –0.00 –0.00 0.00 0.00

0.03 0.27 0.25 0.25 0.03 0.22 0.25 0.03 0.14 0.03 0.25 0.25 0.25 0.25 0.11 0.25 0.25 0.25 0.25 0.27 0.03 0.10 0.25 0.13 0.25 0.27 0.27 0.22 0.03 0.02 0.03 0.25 0.25 0.03 0.03 0.21 0.27 0.06 0.05 0.25 0.03 0.27 0.22 0.03

−0.00 −0.00 −0.00 −0.00 −0.00 0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 0.00 −0.00 −0.00 −0.00 −0.00 −0.00 0.00 −0.00 0.00 0.00 −0.00 −0.00 −0.00 −0.00 0.00 −0.00 −0.00 0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00 −0.00

0.01 0.04 0.01 0.01 0.01 0.03 0.01 0.01 0.01 0.00 0.02 0.02 0.02 0.02 0.02 0.01 0.01 0.03 0.01 0.03 0.01 0.01 0.02 0.01 0.01 0.01 0.03 0.01 0.01 0.00 0.01 0.01 0.02 0.01 0.01 0.01 0.02 0.01 0.00 0.02 0.01 0.01 0.02 0.00

# Observations 11240 267800 21077 29339 45841 203336 51621 27877 234628 5659 16618 170230 42985 36438 2264 27645 156149 358241 53301 190589 53188 561658 9954 4657 57519 745501 256245 3165 17295 1165 171219 35907 41552 10810 34543 13770 105448 1225 6321 73733 85166 202800 1458005 3781

* Winsorization levels are at 1 and 99 percentile instead of 0.1 and 99.9 percentile.

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 143

143

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table A.5 (Continued ) SIZE Level

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Min

25%

Median

75%

Max

Mean

StdDev

# Observations

−7.41 −6.46 −7.52 −7.52 −7.41 −5.83 −7.52 −7.41 −2.51 −5.42 −4.64 −7.52 −7.52 −7.52 −3.55 −6.36 −7.52 −7.52 −7.51 −8.79 −6.49 −5.17 −6.64 −7.52 −7.52 −9.57 −12.22 −7.52 −7.24 −4.07 −4.28 −7.52 −7.52 −7.41 −7.41 −7.52 −4.37 −6.18 −5.99 −6.03 −5.94 −7.45 −5.83 −7.41

−1.36 −1.20 −1.33 −1.37 −1.75 −1.47 −1.13 −1.04 −0.75 −1.48 −1.09 −0.26 −0.18 −1.64 −0.40 −1.72 −1.28 −1.12 −0.50 −1.52 −1.02 −1.17 −2.06 −2.04 −0.87 −0.78 −0.49 −2.50 −1.18 −0.97 −0.14 −1.85 −0.90 −1.00 −1.34 −1.95 −0.57 −0.21 −0.46 −0.52 −0.85 −0.66 −2.01 −1.46

0.27 −0.09 −0.04 0.14 −0.09 −0.19 0.14 0.11 −0.26 −0.09 −0.06 1.25 0.98 −0.19 0.42 −0.38 0.18 0.26 0.47 −0.51 0.13 0.26 −0.83 −1.11 0.28 0.26 0.30 −0.98 0.18 −0.17 0.72 −0.33 0.18 0.37 −0.23 −0.24 0.39 1.25 0.88 1.16 0.12 0.31 −0.69 −0.02

1.66 1.57 1.35 1.70 1.31 1.32 1.38 1.30 0.28 1.09 0.92 2.93 2.28 1.29 1.59 1.19 1.95 1.89 1.58 0.82 1.37 2.06 0.64 −0.20 1.70 1.54 1.36 0.15 1.51 1.01 1.82 1.16 1.43 1.81 1.08 1.34 1.62 3.13 2.41 2.82 1.24 1.32 0.75 1.19

7.17 6.97 4.50 8.04 8.19 6.01 6.31 4.30 3.66 4.33 5.95 8.04 7.33 5.31 4.74 6.40 7.67 8.04 6.40 6.97 6.17 8.33 4.55 2.76 6.31 6.97 6.97 4.33 5.05 2.31 6.44 5.99 6.65 5.54 5.15 4.68 6.97 7.96 8.04 8.04 6.53 6.66 6.01 7.67

0.20 0.34 −0.02 0.18 −0.13 −0.03 0.20 0.04 −0.17 −0.28 −0.06 1.32 1.12 −0.31 0.55 −0.26 0.42 0.49 0.58 −0.23 0.25 0.58 −0.72 −1.14 0.43 0.48 0.47 −1.13 0.08 −0.03 0.88 −0.21 0.30 0.37 −0.03 −0.46 0.62 1.68 1.22 1.25 0.31 0.41 −0.55 −0.34

2.10 2.06 1.99 2.30 2.66 2.07 1.93 1.83 0.87 1.68 1.60 2.51 1.88 2.29 1.68 1.99 2.34 2.23 1.75 1.84 1.80 2.31 1.93 1.52 1.96 1.72 1.92 2.06 1.96 1.30 1.54 2.21 1.73 2.03 1.80 2.61 1.69 2.54 2.45 2.42 1.60 1.50 2.00 2.66

12751 292540 23723 36120 55038 226939 54719 31129 255956 6789 19680 200115 46305 39969 2452 29270 184757 393113 56611 205042 61267 483134 10669 5780 60096 766833 304093 4375 19552 1629 183092 37778 45882 13225 38770 16134 114522 3468 10014 80684 94926 224664 1514407 5223 (Continued)

144

1250007.indd 144

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

Table A.5

30 August 2012 9:08 AM

(Continued ) SIZE Trend

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Min

25%

Median

75%

Max

Mean

StdDev

# Observations

−1.89 −1.60 −2.03 −2.03 −1.89 −1.91 −2.03 −1.89 −1.08 −1.34 −2.03 −2.03 −2.03 −2.03 −2.03 −2.03 −2.03 −2.03 −2.03 −1.60 −1.89 −1.65 −2.03 −2.03 −2.03 −1.60 −1.60 −2.03 −1.89 −1.23 −1.89 −2.03 −2.03 −1.89 −1.89 −2.03 −1.60 −2.03 −2.03 −2.03 −1.89 −1.60 −1.91 −1.89

−0.15 −0.17 −0.12 −0.11 −0.16 −0.16 −0.10 −0.11 −0.11 −0.10 −0.19 −0.17 −0.14 −0.10 −0.13 −0.13 −0.11 −0.15 −0.18 −0.18 −0.19 −0.22 −0.10 −0.11 −0.12 −0.12 −0.18 −0.08 −0.14 −0.08 −0.14 −0.11 −0.13 −0.14 −0.15 −0.14 −0.15 −0.06 −0.16 −0.15 −0.15 −0.14 −0.16 −0.17

−0.02 0.00 −0.01 −0.01 0.00 0.00 −0.01 −0.01 –0.00 0.00 −0.01 −0.03 –0.02 −0.00 0.00 0.00 0.00 −0.00 –0.03 −0.02 −0.03 −0.03 0.01 0.00 −0.01 −0.01 −0.03 0.01 –0.02 0.00 −0.03 0.00 0.00 −0.00 −0.01 −0.02 −0.03 0.02 −0.03 –0.00 –0.02 −0.02 −0.01 −0.02

0.11 0.18 0.09 0.07 0.15 0.17 0.09 0.09 0.10 0.10 0.16 0.09 0.09 0.10 0.14 0.14 0.12 0.13 0.13 0.14 0.14 0.14 0.13 0.12 0.09 0.09 0.03 0.12 0.09 0.08 0.09 0.11 0.13 0.12 0.14 0.08 0.09 0.14 0.07 0.14 0.12 0.11 0.13 0.12

0.20 1.85 2.26 2.26 2.20 1.84 2.26 2.20 1.14 1.93 2.26 2.26 2.26 2.26 2.26 2.26 2.26 2.26 2.26 1.85 2.20 1.96 2.26 2.26 2.26 1.85 1.85 2.26 2.20 2.05 2.20 2.26 2.26 2.20 2.20 2.26 1.85 2.26 1.98 2.26 2.20 1.85 1.84 2.20

–0.01 0.02 −0.01 −0.02 −0.01 0.00 −0.01 –0.00 0.01 0.01 −0.01 −0.05 −0.03 0.01 0.00 0.00 0.00 −0.02 −0.01 −0.00 −0.01 −0.03 0.01 0.01 –0.00 −0.01 −0.02 0.03 −0.03 0.02 −0.02 −0.00 0.00 −0.00 0.01 −0.02 −0.02 0.05 −0.06 −0.01 −0.01 −0.01 −0.02 0.01

0.30 0.38 0.27 0.26 0.37 0.36 0.24 0.22 0.20 0.22 0.35 0.34 0.27 0.28 0.33 0.27 0.29 0.34 0.33 0.36 0.36 0.36 0.30 0.31 0.25 0.22 0.34 0.30 0.26 0.25 0.26 0.26 0.33 0.29 0.33 0.26 0.26 0.31 0.29 0.34 0.28 0.25 0.33 0.43

12631 291573 23490 35808 53882 226621 54467 30708 255898 6620 19411 198678 46079 39620 2444 29223 182860 391484 56542 204824 60532 476506 10635 5687 60029 766748 303753 4190 19382 1597 183013 37733 45645 12915 38283 15794 114335 3062 9704 80601 94755 224535 1513978 5011 (Continued)

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 145

145

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table A.5 (Continued ) M/B ARG AUS* AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG* IDN IND** IRL ISL ITA JPN* KOR* LUX MEX MLT MYS NLD NOR PER PHL PRT SGP* SVK SVN SWE THA TWN* USA VEN

Min

25%

Median

75%

Max

Mean

StdDev

# Observations

0.18 0.20 0.20 0.18 0.18 0.23 0.18 0.18 0.63 0.23 0.18 0.18 0.18 0.18 0.18 0.19 0.18 0.18 0.18 0.20 0.18 0.19 0.18 0.18 0.18 0.20 0.20 0.18 0.18 0.66 0.18 0.18 0.18 0.18 0.18 0.18 0.20 0.32 0.18 0.18 0.18 0.20 0.23 0.18

0.82 0.94 0.95 0.94 0.81 0.99 0.99 0.86 1.51 0.77 0.61 1.00 0.96 0.96 0.97 1.01 0.95 0.98 0.87 0.72 0.85 0.77 1.00 1.09 0.95 0.85 0.80 0.75 0.76 0.97 0.78 1.00 0.96 0.80 0.75 0.91 0.83 0.75 0.68 1.03 0.84 0.94 1.02 0.58

1.01 1.35 1.07 1.10 1.05 1.32 1.14 1.14 2.15 1.01 0.81 1.22 1.05 1.11 1.18 1.23 1.14 1.33 1.11 1.00 1.05 1.00 1.22 1.28 1.06 1.01 0.98 0.97 1.02 1.08 1.00 1.22 1.15 1.10 1.03 1.03 1.04 0.93 0.86 1.36 1.05 1.18 1.30 0.84

1.28 2.42 1.37 1.47 1.59 2.10 1.59 1.68 3.22 1.25 1.06 1.69 1.39 1.47 1.82 1.72 1.57 2.09 1.64 1.56 1.43 1.51 1.72 1.63 1.35 1.25 1.32 1.20 1.42 1.39 1.43 1.69 1.72 1.66 1.63 1.26 1.46 1.05 1.05 2.19 1.42 1.68 2.09 1.00

157.18 14.73 77.90 77.90 787.85 60.04 77.90 787.85 38.81 787.85 50.32 77.90 77.90 77.90 43.24 77.90 77.90 77.90 77.90 14.73 787.85 10.92 45.74 77.90 77.90 14.73 14.73 9.26 8.65 15.76 787.85 77.90 77.90 29.63 787.85 47.43 14.73 3.18 8.11 77.90 77.05 14.73 60.04 787.85

1.68 2.31 1.36 1.75 15.40 2.23 1.58 2.92 2.76 2.15 1.18 1.80 1.53 1.44 1.66 1.65 1.65 2.22 1.78 1.54 1.47 1.47 1.72 1.88 1.36 1.22 1.33 1.10 1.19 1.35 1.50 1.68 1.85 1.51 7.16 1.17 1.37 0.94 0.95 2.17 1.28 1.49 2.10 9.00

5.88 2.67 1.92 4.67 86.63 3.98 2.01 23.20 2.37 27.78 2.40 3.19 2.47 1.97 1.94 2.10 3.31 4.30 3.70 1.90 8.67 1.53 2.09 5.02 2.55 0.99 1.40 0.68 0.68 0.98 8.52 1.86 3.05 1.38 57.34 0.82 1.26 0.32 0.55 3.24 0.97 1.06 3.18 61.38

11150 267583 21059 28916 45362 203058 51364 27780 233648 5610 16567 169566 42880 36140 2258 27561 155366 357753 53237 190508 53111 411752 9915 4643 57277 745007 253858 3078 17086 1165 171084 35786 41365 10757 34421 13722 105342 1212 6306 73655 85147 202797 1457333 3720

* Winsorization levels are at 0.1 and 99.5 percentiles instead of 0.1 and 99.9 percentiles. ** Winsorization levels are at 1 and 99 percentiles instead of 0.1 and 99.9 percentiles.

(Continued) 146

1250007.indd 146

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:36 AM

b1423

Global Credit Review Volume 2

Table A.5

30 August 2012 9:08 AM

(Continued ) SIGMA

ARG AUS AUT BEL BRA CAN CHE CHL CHN COL CYP DEU DNK ESP EST FIN FRA GBR GRC HKG IDN IND IRL ISL ITA JPN KOR LUX MEX MLT MYS NLD NOR PER PHL PRT SGP SVK SVN SWE THA TWN USA VEN

Min

25%

Median

75%

Max

Mean

StdDev

0.02 0.01 0.01 0.01 0.00 0.02 0.01 0.00 0.02 0.00 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.00 0.03 0.01 0.02 0.01 0.01 0.01 0.01 0.00 0.01 0.01 0.01 0.01 0.00 0.00 0.01 0.01 0.01 0.01 0.01 0.00 0.01 0.02 0.00

0.07 0.09 0.05 0.05 0.08 0.08 0.05 0.05 0.07 0.05 0.08 0.06 0.05 0.05 0.06 0.06 0.06 0.07 0.08 0.09 0.09 0.11 0.06 0.06 0.05 0.06 0.09 0.05 0.06 0.04 0.07 0.05 0.07 0.08 0.08 0.05 0.07 0.06 0.05 0.07 0.07 0.07 0.08 0.09

0.09 0.15 0.07 0.07 0.12 0.13 0.07 0.07 0.09 0.08 0.12 0.10 0.08 0.07 0.09 0.09 0.09 0.11 0.11 0.13 0.13 0.16 0.09 0.08 0.08 0.08 0.13 0.07 0.09 0.05 0.10 0.08 0.10 0.11 0.13 0.08 0.10 0.10 0.07 0.11 0.10 0.10 0.12 0.11

0.14 0.22 0.11 0.10 0.17 0.19 0.10 0.10 0.13 0.11 0.17 0.16 0.12 0.11 0.14 0.12 0.13 0.16 0.16 0.19 0.19 0.23 0.14 0.11 0.11 0.12 0.19 0.10 0.12 0.08 0.15 0.11 0.16 0.15 0.19 0.12 0.15 0.14 0.12 0.17 0.15 0.15 0.18 0.17

0.43 0.45 0.45 0.45 0.44 0.47 0.45 0.44 0.33 0.43 0.45 0.45 0.45 0.45 0.30 0.45 0.45 0.45 0.45 0.45 0.44 0.50 0.45 0.37 0.42 0.45 0.45 0.35 0.43 0.37 0.44 0.45 0.45 0.44 0.44 0.45 0.45 0.36 0.45 0.45 0.44 0.45 0.47 0.42

0.11 0.16 0.08 0.08 0.14 0.14 0.08 0.08 0.10 0.09 0.14 0.12 0.09 0.09 0.10 0.10 0.11 0.13 0.13 0.14 0.15 0.18 0.11 0.10 0.09 0.10 0.15 0.08 0.10 0.07 0.11 0.09 0.12 0.12 0.14 0.10 0.12 0.11 0.09 0.13 0.12 0.12 0.13 0.13

0.06 0.09 0.06 0.06 0.07 0.08 0.05 0.05 0.05 0.06 0.08 0.08 0.06 0.06 0.06 0.06 0.07 0.07 0.07 0.08 0.08 0.08 0.07 0.05 0.05 0.06 0.08 0.05 0.06 0.05 0.07 0.06 0.07 0.07 0.08 0.07 0.07 0.06 0.06 0.08 0.07 0.06 0.08 0.06

# Observations 11397 271566 21996 33297 47586 213387 51537 28311 237070 5792 17454 182991 43414 36225 2208 27861 167740 363765 53735 193483 53817 427485 10050 4819 56804 745701 286449 3528 17482 1377 175167 35779 41999 11091 34948 14107 107292 1903 8192 75297 89150 212379 1427608 4100

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 147

147

8/30/2012 9:36:37 AM

b1423

Global Credit Review Volume 2

Table A.6

30 August 2012 9:08 AM

Exits classified as “Defaults”. Defaults

Action Type

Subcategory

Bankruptcy filing

Administration, Arrangement, Canadian CCAA, Chapter 7, Chapter 11, Chapter 15, Conservatorship, Insolvency, Japanese CRL, Judicial Management, Liquidation, PreNegotiation Chapter 11, Protection, Receivership, Rehabilitation, Rehabilitation (Thailand 1997), Reorganization, Restructuring, Section 304, Supreme court declaration, Winding up, Work out, Other, Unknown

Delisting

Bankruptcy

Default Corporate Action

Bankruptcy, Coupon & Principal Payment, Coupon Payment Only, Debt Restructuring, Interest Payment, Loan Payment, Principal Payment, ADR (Japan only), Declared Sick (India Only), Unknown Table A.7

Exits classified as “Other Exits”. Other Exits

148

1250007.indd 148

Action Type

Subcategory

Delisting

Unknown, Acquired/Merged, Assimilated with underlying shares, Bid price below minimum, Cancellation of listing, End of When-issued trading, Expired, Failure to meet listing requirements, Failure to pay listing fees, Inactive security, Insufficient assets, Insufficient capital and surplus, Insufficient number of market makers, Issue postponed, Lack of market maker interest, Lack of public interest, Liquidated, Matured, Not available, Not current in required filings, NP/FP finished, Privatized, Reorganization security called for redemptions, the company’s request, Scheme of arrangement, Insufficient spread of holders, Selective capital reduction of the company

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:37 AM

b1423

Table A.8

Global Credit Review Volume 2

30 August 2012 9:08 AM

Number of defaults and other exits of 44 economies from 1992 to June 2012.

Economy:ARG

Economy:AUT

Defaults Year 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Active 1 1 23 87 95 79 66 71 69 51 59 61 56 61 63 67 58 58 62 58 61

#

%

0 0 0 0 0 0 1 1 0 3 8 2 0 0 0 0 0 1 0 0 0

0.00 0.00 0.00 0.00 0.00 0.00 0.94 1.08 0.00 3.41 9.88 2.67 0.00 0.00 0.00 0.00 0.00 1.49 0.00 0.00 0.00

Others # 0 0 2 18 22 31 39 21 22 34 14 12 13 4 8 10 16 8 4 9 2

Defaults

Others

%

Year

Active

#

%

#

%

0.00 0.00 8.00 17.14 18.80 28.18 36.79 22.58 24.18 38.64 17.28 16.00 18.84 6.15 11.27 12.99 21.62 11.94 6.06 13.43 3.17

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

87 101 111 118 117 118 112 108 119 113 109 108 100 99 104 106 105 99 94 83 86

0 0 0 0 1 0 0 0 0 2 0 0 0 1 0 0 2 1 1 0 1

0.00 0.00 0.00 0.00 0.81 0.00 0.00 0.00 0.00 1.39 0.00 0.00 0.00 0.85 0.00 0.00 1.71 0.88 0.88 0.00 1.12

3 9 1 2 5 6 15 17 15 29 14 21 23 18 10 11 10 13 19 16 2

3.33 8.18 0.89 1.67 4.07 4.84 11.81 13.60 11.19 20.14 11.38 16.28 18.70 15.25 8.77 9.40 8.55 11.50 16.67 16.16 2.25

Economy:AUS

Economy:BEL

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

696 829 916 956 1005 1016 1010 1072 1186 1163 1179 1225 1348 1461 1578 1752 1684 1667 1683 1677 1710

0 0 0 1 2 3 4 3 11 27 8 8 3 6 5 5 27 30 4 0 1

0.00 0.00 0.00 0.10 0.19 0.27 0.36 0.26 0.84 2.08 0.62 0.60 0.21 0.39 0.29 0.27 1.45 1.66 0.22 0.00 0.06

118 48 93 82 66 103 109 100 105 109 102 94 75 87 117 114 149 115 136 176 45

14.50 5.47 9.22 7.89 6.15 9.18 9.71 8.51 8.06 8.39 7.91 7.08 5.26 5.60 6.88 6.09 8.01 6.35 7.46 9.50 2.56

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

131 136 143 148 163 162 174 190 191 185 175 176 169 171 182 215 199 199 201 172 201

0 0 0 0 0 0 0 2 0 2 3 1 1 2 2 1 1 2 0 1 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.01 0.00 1.01 1.55 0.52 0.55 1.08 1.04 0.36 0.36 0.79 0.00 0.42 0.00

7 6 12 10 9 18 17 6 10 11 15 14 13 12 9 61 79 52 54 63 4

5.07 4.23 7.74 6.33 5.23 10.00 8.90 3.03 4.98 5.56 7.77 7.33 7.10 6.49 4.66 22.02 28.32 20.55 21.18 26.69 1.95

(Continued) GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 149

149

8/30/2012 9:36:37 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table A.8 (Continued ) Economy:BRA

Economy:CHE

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 261 273 282 258 283 320 293 273 239 260 266 260 276 340 321 324 316 307 310

0 0 0 0 0 2 3 2 2 1 3 2 1 1 0 0 0 0 0 1 0

NaN NaN 0.00 0.00 0.00 0.51 0.68 0.47 0.48 0.24 0.82 0.57 0.29 0.30 0.00 0.00 0.00 0.00 0.00 0.29 0.00

0 0 26 92 99 133 157 103 120 143 126 90 82 71 54 40 56 35 41 40 15

NaN NaN 9.06 25.21 25.98 33.84 35.44 24.24 28.92 34.29 34.24 25.57 23.50 21.39 16.36 10.53 14.85 9.75 11.48 11.49 4.62

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

143 174 177 189 210 221 227 242 261 257 249 243 237 243 248 252 250 252 247 243 250

0 0 0 0 0 1 0 0 0 2 0 2 1 1 0 0 0 0 0 1 1

0.00 0.00 0.00 0.00 0.00 0.43 0.00 0.00 0.00 0.73 0.00 0.78 0.40 0.40 0.00 0.00 0.00 0.00 0.00 0.38 0.39

30 10 20 15 15 11 14 16 12 16 17 13 11 7 15 8 14 16 17 18 5

17.34 5.43 10.15 7.35 6.67 4.72 5.81 6.20 4.40 5.82 6.39 5.04 4.42 2.79 5.70 3.08 5.30 5.97 6.44 6.87 1.95

Economy:CHL Economy:CAN Defaults Defaults

Others

Others

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

951 1159 1331 1451 1637 1782 1754 1185 1095 936 922 920 979 1030 1083 1121 1100 1027 1060 1068 1063

1 0 0 0 0 6 9 13 9 18 5 13 5 3 3 3 12 13 5 4 1

0.09 0.00 0.00 0.00 0.00 0.31 0.45 0.68 0.69 1.53 0.50 1.29 0.47 0.27 0.25 0.24 0.98 1.10 0.43 0.34 0.09

103 72 54 84 80 145 254 715 194 221 73 78 73 82 96 120 111 142 94 120 51

9.76 5.85 3.90 5.47 4.66 7.50 12.59 37.38 14.95 18.81 7.30 7.72 6.91 7.35 8.12 9.65 9.08 12.01 8.11 10.07 4.57

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 141 166 168 181 168 177 168 168 155 152 164 166 171 187 146 167 161 170 172

0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0

NaN NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.47 0.48 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 9 26 46 35 56 41 43 43 54 57 32 39 41 27 52 28 41 36 22

NaN NaN 6.00 13.54 21.50 16.20 25.00 18.81 20.38 20.28 25.71 27.27 16.33 19.02 19.34 12.62 26.26 14.36 20.30 17.48 11.34

(Continued) 150

1250007.indd 150

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:37 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table A.8 (Continued ) Economy:CHN

Economy:CYP

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

45 165 283 317 522 727 841 935 1079 1152 1204 1266 1352 1354 1372 1460 1579 1685 1995 2273 2372

0 0 1 6 10 15 34 23 26 50 48 45 109 93 66 49 35 38 23 6 0

0.00 0.00 0.35 1.85 1.88 2.02 3.88 2.40 2.35 4.13 3.77 3.37 7.33 6.33 4.37 3.05 2.10 2.17 1.11 0.26 0.00

2 0 1 1 0 2 2 1 1 8 21 23 26 22 71 98 52 30 47 62 16

4.26 0.00 0.35 0.31 0.00 0.27 0.23 0.10 0.09 0.66 1.65 1.72 1.75 1.50 4.71 6.10 3.12 1.71 2.28 2.65 0.67

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 0 0 35 43 48 58 116 137 144 134 134 139 137 137 121 109 107 81 96

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

NaN NaN NaN NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 0 0 3 0 2 2 4 6 10 21 29 22 11 8 31 26 27 50 10

NaN NaN NaN NaN 7.89 0.00 4.00 3.33 3.33 4.20 6.49 13.55 17.79 13.66 7.43 5.52 20.39 19.26 20.15 38.17 9.43

Economy:DEU

Economy:COL Defaults

Defaults

Others

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 1 50 44 46 58 47 39 49 50 53 53 55 47 50 35 42 44 41 49

0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0

NaN NaN 0.00 0.00 0.00 0.00 0.81 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 0 29 46 43 64 55 40 19 21 15 12 17 18 16 24 10 14 11 1

NaN NaN 0.00 36.71 51.11 48.31 52.03 53.92 50.63 27.94 29.58 22.06 18.46 23.61 27.69 24.24 40.68 19.23 24.14 21.15 2.00

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

400 421 573 591 621 627 723 900 1030 1027 945 886 880 909 1069 1214 1267 1236 1288 1304 1290

0 0 0 0 4 3 2 1 2 26 37 16 8 4 4 5 17 11 0 5 2

0.00 0.00 0.00 0.00 0.58 0.43 0.26 0.11 0.18 2.35 3.42 1.64 0.86 0.42 0.36 0.39 1.21 0.78 0.00 0.30 0.15

38 28 63 65 63 74 53 51 57 55 100 75 46 41 35 59 120 169 152 335 71

8.68 6.24 9.91 9.91 9.16 10.51 6.81 5.36 5.23 4.96 9.24 7.68 4.93 4.30 3.16 4.62 8.55 11.94 10.56 20.38 5.21

(Continued) GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 151

151

8/30/2012 9:36:37 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued )

Economy:DNK

Economy:EST

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

158 173 176 202 219 212 213 210 208 191 175 174 170 170 192 216 214 210 199 187 184

0 0 0 1 0 0 0 0 1 5 3 1 1 1 0 1 1 4 0 2 0

0.00 0.00 0.00 0.46 0.00 0.00 0.00 0.00 0.44 2.20 1.44 0.52 0.54 0.56 0.00 0.45 0.43 1.79 0.00 0.99 0.00

19 13 24 16 11 21 29 24 20 31 30 19 14 9 7 7 18 9 17 13 4

10.73 6.99 12.00 7.31 4.78 9.01 11.98 10.26 8.73 13.66 14.42 9.79 7.57 5.00 3.52 3.13 7.73 4.04 7.87 6.44 2.13

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 0 0 0 17 19 19 16 14 11 11 11 13 13 16 17 15 15 15 15

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

NaN NaN NaN NaN NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 0 0 0 0 1 1 4 2 3 0 0 1 2 0 0 2 1 0 0

NaN NaN NaN NaN NaN 0.00 5.00 5.00 20.00 12.50 21.43 0.00 0.00 7.14 13.33 0.00 0.00 11.76 6.25 0.00 0.00

Economy:ESP

Economy:FIN

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

147 112 238 237 264 272 231 213 212 194 206 171 161 159 162 153 142 137 139 137 137

0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 1 1 0 1 0 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.75 0.00 0.00 0.00 0.00 0.51 0.58 0.00 0.65 0.00 0.00

39 96 18 96 65 59 104 78 55 80 60 81 41 47 44 41 28 24 15 15 8

20.97 46.15 7.03 28.83 19.76 17.82 31.04 26.80 20.60 29.20 22.39 32.14 20.30 22.82 21.36 21.03 16.37 14.91 9.68 9.87 5.52

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

92 94 96 102 110 124 127 146 153 148 142 138 131 132 132 130 127 125 123 121 121

0 0 0 0 0 0 1 0 0 1 1 1 0 0 0 0 1 1 0 1 0

0.00 0.00 0.00 0.00 0.00 0.00 0.75 0.00 0.00 0.62 0.65 0.68 0.00 0.00 0.00 0.00 0.76 0.78 0.00 0.81 0.00

0 2 6 5 3 1 6 8 12 13 10 9 11 6 8 5 3 2 4 1 1

0.00 2.08 5.88 4.67 2.65 0.80 4.48 5.19 7.27 8.02 6.54 6.08 7.75 4.35 5.71 3.70 2.29 1.56 3.15 0.81 0.82

(Continued) 152

1250007.indd 152

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:37 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued )

Economy:FRA

Economy:GRC

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

627 641 692 719 781 803 814 857 919 916 870 861 841 846 911 950 902 889 847 799 828

0 0 0 0 0 1 0 0 2 9 5 4 4 5 7 7 12 6 2 2 0

0.00 0.00 0.00 0.00 0.00 0.11 0.00 0.00 0.20 0.88 0.51 0.41 0.42 0.53 0.70 0.66 1.13 0.58 0.20 0.21 0.00

72 78 103 119 116 142 193 93 97 93 113 102 106 98 75 100 150 142 168 153 39

10.30 10.85 12.96 14.20 12.93 15.01 19.17 9.79 9.53 9.14 11.44 10.55 11.15 10.33 7.55 9.46 14.10 13.69 16.52 16.04 4.50

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

90 97 162 182 196 208 231 264 308 312 309 313 313 299 285 279 274 264 256 231 224

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 2 2 6 4 5 6 8 14 19 9 10 21 16 13 18 24 30 35 19

0.00 0.00 1.22 1.09 2.97 1.89 2.12 2.22 2.53 4.29 5.79 2.80 3.10 6.56 5.32 4.45 6.16 8.33 10.49 13.16 7.82

Economy:GBR

Economy:HKG

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

1075 1188 1280 1411 1620 1718 1686 1557 1672 1683 1638 1598 1785 2035 2185 2221 2026 1824 1783 1658 1653

0 0 0 0 0 0 0 3 2 12 13 6 1 2 0 2 30 32 3 10 4

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 0.11 0.65 0.71 0.34 0.05 0.09 0.00 0.08 1.25 1.47 0.15 0.51 0.23

87 39 49 63 62 113 197 295 210 148 168 185 150 205 251 264 353 317 251 279 80

7.49 3.18 3.69 4.27 3.69 6.17 10.46 15.90 11.15 8.03 9.24 10.34 7.75 9.14 10.30 10.62 14.65 14.59 12.32 14.33 4.61

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

356 423 464 491 529 601 628 660 742 812 906 955 993 1026 1078 1167 1171 1231 1303 1359 1376

0 0 0 0 0 0 2 6 2 9 4 4 0 3 2 2 8 3 1 1 1

0.00 0.00 0.00 0.00 0.00 0.00 0.30 0.87 0.26 1.06 0.42 0.40 0.00 0.27 0.18 0.17 0.66 0.24 0.08 0.07 0.07

11 7 13 9 15 25 30 21 21 30 36 51 56 65 43 24 33 23 29 26 24

3.00 1.63 2.73 1.80 2.76 3.99 4.55 3.06 2.75 3.53 3.81 5.05 5.34 5.94 3.83 2.01 2.72 1.83 2.18 1.88 1.71

(Continued) GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 153

153

8/30/2012 9:36:37 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued )

Economy:IDN

Economy:IRL

Defaults Year 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Active 124 155 184 204 236 255 245 244 248 256 252 277 270 261 286 303 265 304 340 366 382

# 0 0 0 0 0 0 17 16 6 8 2 1 2 0 0 0 0 4 1 0 1

% 0.00 0.00 0.00 0.00 0.00 0.00 5.57 5.63 2.10 2.56 0.64 0.33 0.60 0.00 0.00 0.00 0.00 1.10 0.26 0.00 0.26

Others # 30 23 49 41 26 33 43 24 32 48 59 25 59 74 51 68 90 55 38 40 7

Defaults

Others

%

Year

Active

#

%

#

%

19.48 12.92 21.03 16.73 9.92 11.46 14.10 8.45 11.19 15.38 18.85 8.25 17.82 22.09 15.13 18.33 25.35 15.15 10.03 9.85 1.79

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

31 37 37 37 43 51 50 51 59 54 48 43 42 42 47 52 49 44 40 38 34

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

4 4 5 1 0 3 5 6 5 6 6 5 3 2 2 2 3 5 4 2 4

11.43 9.76 11.90 2.63 0.00 5.56 9.09 10.53 7.81 10.00 11.11 10.42 6.67 4.55 4.08 3.70 5.77 10.20 9.09 5.00 10.53

Economy:IND

Economy:ISL

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

1497 1844 2805 4078 4082 3014 2689 3060 2559 2265 3353 2681 2518 2520 2725 2990 3036 3136 3548 3380 3596

1 0 0 2 6 12 10 17 10 7 7 14 5 13 12 17 24 35 9 4 4

0.06 0.00 0.00 0.05 0.12 0.24 0.24 0.41 0.26 0.21 0.17 0.34 0.15 0.41 0.37 0.54 0.67 1.02 0.20 0.10 0.11

143 201 270 349 1062 1961 1471 1112 1335 1123 656 1387 923 617 480 130 531 276 908 784 73

8.71 9.83 8.78 7.88 20.62 39.32 35.28 26.55 34.20 33.08 16.33 33.98 26.78 19.59 14.92 4.14 14.79 8.01 20.34 18.81 1.99

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 0 0 26 33 50 59 58 64 55 40 32 25 25 26 11 10 8 9 9

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 1 0 0 0

NaN NaN NaN NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 13.79 6.67 0.00 0.00 0.00

0 0 0 0 0 3 3 9 17 12 17 22 13 11 5 5 14 4 4 4 1

NaN NaN NaN NaN 0.00 8.33 5.66 13.24 22.67 15.79 23.61 35.48 28.89 30.56 16.67 16.13 48.28 26.67 33.33 30.77 10.00

(Continued) 154

1250007.indd 154

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:37 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued )

Economy:ITA

Economy:KOR

Defaults

Others

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

185 181 195 209 219 218 227 252 276 279 277 257 256 262 275 295 286 276 280 269 273

0 0 0 0 2 0 0 0 0 0 1 5 2 0 0 0 1 4 0 0 1

0.00 0.00 0.00 0.00 0.85 0.00 0.00 0.00 0.00 0.00 0.34 1.75 0.75 0.00 0.00 0.00 0.32 1.33 0.00 0.00 0.36

5 8 13 14 15 28 14 7 25 17 14 23 7 15 15 13 21 20 13 22 6

2.63 4.23 6.25 6.28 6.36 11.38 5.81 2.70 8.31 5.74 4.79 8.07 2.64 5.42 5.17 4.22 6.82 6.67 4.44 7.56 2.14

Defaults

Others

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

638 647 678 705 750 1000 898 984 1122 1271 1409 1459 1476 1530 1605 1675 1704 1696 1719 1724 1703

0 0 0 0 6 53 79 26 12 16 14 11 7 8 2 0 7 6 7 2 1

0.00 0.00 0.00 0.00 0.79 4.95 7.27 2.48 1.02 1.22 0.96 0.74 0.46 0.50 0.12 0.00 0.40 0.34 0.39 0.11 0.06

1 0 0 2 3 17 109 39 43 23 28 25 53 48 10 15 34 87 91 88 41

0.16 0.00 0.00 0.28 0.40 1.59 10.04 3.72 3.65 1.76 1.93 1.67 3.45 3.03 0.62 0.89 1.95 4.86 5.01 4.85 2.35

Economy:LUX

Economy:JPN Defaults

Defaults

Others

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

2541 2621 2767 2947 3100 3218 3271 3333 3475 3582 3604 3635 3738 3821 3951 3980 3904 3773 3676 3622 3591

3 4 0 2 4 7 14 7 12 16 29 20 11 9 2 6 35 28 8 4 3

0.12 0.15 0.00 0.07 0.13 0.21 0.42 0.21 0.34 0.44 0.78 0.53 0.29 0.23 0.05 0.15 0.87 0.71 0.21 0.11 0.08

21 24 17 17 21 31 37 47 59 64 94 98 85 89 84 104 107 136 127 100 47

0.82 0.91 0.61 0.57 0.67 0.95 1.11 1.39 1.66 1.75 2.52 2.61 2.22 2.27 2.08 2.54 2.64 3.45 3.33 2.68 1.29

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

2 2 2 30 28 37 34 34 32 28 27 28 36 37 36 34 26 23 18 14 14

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.45 0.00 0.00

1 1 0 12 15 9 13 14 13 13 10 11 7 7 16 8 14 8 10 8 3

33.33 33.33 0.00 28.57 34.88 19.57 27.66 29.17 28.89 31.71 27.03 28.21 16.28 15.91 30.77 19.05 35.00 25.81 34.48 36.36 17.65 (Continued)

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 155

155

8/30/2012 9:36:38 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued )

Economy:MEX

Economy:MYS

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 87 94 105 113 108 107 103 105 93 95 99 92 96 94 86 93 99 94 95

0 0 0 0 0 1 0 1 1 1 1 2 0 0 0 0 1 1 0 0 0

NaN NaN 0.00 0.00 0.00 0.73 0.00 0.76 0.83 0.81 0.82 1.79 0.00 0.00 0.00 0.00 0.93 0.97 0.00 0.00 0.00

0 0 37 22 19 23 21 23 17 18 28 15 10 23 10 16 21 9 12 24 3

NaN NaN 29.84 18.97 15.32 16.79 16.28 17.56 14.05 14.52 22.95 13.39 9.17 20.00 9.43 14.55 19.44 8.74 10.81 20.34 3.06

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

356 405 466 523 615 702 696 701 728 729 756 819 891 963 972 946 904 897 897 903 898

0 0 0 0 0 0 14 8 8 10 7 3 2 0 5 4 12 10 11 1 0

0.00 0.00 0.00 0.00 0.00 0.00 1.92 1.11 1.07 1.32 0.89 0.36 0.22 0.00 0.49 0.40 1.23 1.06 1.18 0.11 0.00

10 0 7 2 0 2 21 14 11 18 26 20 17 27 34 61 61 37 28 27 15

2.73 0.00 1.48 0.38 0.00 0.28 2.87 1.94 1.47 2.38 3.30 2.38 1.87 2.73 3.36 6.03 6.24 3.92 2.99 2.90 1.64

Economy:NLD

Economy:MLT Defaults

Defaults

Others

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 0 0 5 6 7 7 9 9 10 10 11 11 13 14 13 12 12 14 17

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

NaN NaN NaN NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 0 0 0 0 0 1 0 2 2 3 2 2 0 3 7 3 2 2 0

NaN NaN NaN NaN 0.00 0.00 0.00 12.50 0.00 18.18 16.67 23.08 15.38 15.38 0.00 17.65 35.00 20.00 14.29 12.50 0.00

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

160 170 173 186 195 197 209 212 200 175 157 152 145 141 137 134 126 122 120 115 115

0 0 0 0 1 0 0 0 1 7 8 0 0 0 1 0 1 3 0 0 0

0.00 0.00 0.00 0.00 0.50 0.00 0.00 0.00 0.45 3.41 4.42 0.00 0.00 0.00 0.68 0.00 0.72 2.33 0.00 0.00 0.00

6 5 5 5 4 17 10 19 21 23 16 14 10 8 8 9 11 4 5 8 2

3.61 2.86 2.81 2.62 2.00 7.94 4.57 8.23 9.46 11.22 8.84 8.43 6.45 5.37 5.48 6.29 7.97 3.10 4.00 6.50 1.71 (Continued)

156

1250007.indd 156

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:38 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued )

Economy:NOR

Economy:PHL

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

81 101 115 137 159 201 216 201 195 213 205 183 198 235 261 269 242 224 223 220 214

0 0 0 0 0 0 0 0 1 3 4 4 0 0 0 0 4 6 0 1 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.44 1.21 1.64 1.75 0.00 0.00 0.00 0.00 1.39 2.24 0.00 0.43 0.00

8 2 3 2 4 12 19 28 33 32 35 42 17 23 37 47 42 38 24 13 7

8.99 1.94 2.54 1.44 2.45 5.63 8.09 12.23 14.41 12.90 14.34 18.34 7.91 8.91 12.42 14.87 14.58 14.18 9.72 5.56 3.17

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

80 108 128 157 177 186 177 183 170 165 145 161 163 164 174 179 170 189 193 206 211

0 1 0 0 0 0 1 4 0 3 7 4 6 2 0 2 1 0 0 0 0

0.00 0.80 0.00 0.00 0.00 0.00 0.49 2.01 0.00 1.45 3.41 2.02 2.80 0.99 0.00 0.97 0.49 0.00 0.00 0.00 0.00

24 16 27 15 14 21 27 12 35 39 53 33 45 37 25 26 33 18 17 12 3

23.08 12.80 17.42 8.72 7.33 10.14 13.17 6.03 17.07 18.84 25.85 16.67 21.03 18.23 12.56 12.56 16.18 8.70 8.10 5.50 1.40

Economy:PRT

Economy:PER Defaults

Defaults

Others

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

1 1 63 97 94 118 107 92 74 64 75 66 75 78 78 90 78 92 87 76 96

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 2 22 46 40 59 67 74 54 50 46 40 42 40 26 50 33 31 39 10

0.00 0.00 3.08 18.49 32.86 25.32 35.54 42.14 50.00 45.76 40.00 41.07 34.78 35.00 33.90 22.41 39.06 26.40 26.27 33.91 9.43

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

1 69 79 92 96 93 84 88 84 69 62 65 67 64 62 58 57 56 56 52 55

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.33 0.00

0 13 12 18 21 29 34 24 17 20 19 6 7 7 11 9 8 9 7 6 0

0.00 15.85 13.19 16.36 17.95 23.77 28.81 21.43 16.83 22.47 23.46 8.45 9.46 9.86 15.07 13.43 12.31 13.85 11.11 10.00 0.00 (Continued)

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 157

157

8/30/2012 9:36:38 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued )

Economy:SGP

Economy:SVN

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

177 203 233 250 272 299 318 357 426 435 448 498 574 620 661 703 695 706 708 691 688

0 0 0 1 1 1 4 2 0 2 1 1 2 4 1 0 4 13 0 0 0

0.00 0.00 0.00 0.39 0.36 0.31 1.19 0.53 0.00 0.43 0.21 0.19 0.34 0.62 0.15 0.00 0.54 1.74 0.00 0.00 0.00

11 3 4 6 8 18 14 15 18 31 34 15 9 17 25 19 41 29 36 53 16

5.85 1.46 1.69 2.33 2.85 5.66 4.17 4.01 4.05 6.62 7.04 2.92 1.54 2.65 3.64 2.63 5.54 3.88 4.84 7.12 2.27

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 0 0 0 0 75 98 94 112 103 104 108 87 77 65 67 60 69 59 58

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 1 1

NaN NaN NaN NaN NaN NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.41 0.00 1.33 1.67

0 0 0 0 0 0 1 9 34 29 36 19 22 34 28 22 15 25 11 15 1

NaN NaN NaN NaN NaN NaN 1.32 8.41 26.56 20.57 25.90 15.45 16.92 28.10 26.67 25.29 18.29 28.41 13.75 20.00 1.67

Economy:SVK

Economy:SWE

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 0 0 0 0 0 7 8 12 15 23 40 44 41 48 24 34 38 55 53 57

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

NaN NaN NaN NaN NaN NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 0 0 0 0 13 24 11 14 15 26 31 26 42 47 26 31 21 41 9

NaN NaN NaN NaN NaN NaN 65.00 75.00 47.83 48.28 39.47 39.39 41.33 38.81 46.67 66.20 43.33 44.93 27.63 43.62 13.64

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

121 144 171 184 223 270 299 338 370 362 349 337 352 389 433 501 509 496 500 494 485

0 0 0 0 0 0 1 1 1 4 7 3 1 2 0 1 2 4 2 3 0

0.00 0.00 0.00 0.00 0.00 0.00 0.31 0.27 0.25 1.01 1.82 0.82 0.27 0.50 0.00 0.19 0.37 0.75 0.38 0.56 0.00

2 2 2 0 15 28 19 27 35 31 29 26 22 11 22 14 30 32 29 34 21

1.63 1.37 1.16 0.00 6.30 9.40 5.96 7.38 8.62 7.81 7.53 7.10 5.87 2.74 4.84 2.71 5.55 6.02 5.46 6.40 4.15 (Continued)

158

1250007.indd 158

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:38 AM

b1423

Global Credit Review Volume 2

Table A.8

30 August 2012 9:08 AM

(Continued ) Economy:USA

Economy:THA Defaults

Defaults

Others

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

280 331 378 402 421 371 343 318 299 299 317 343 383 419 434 433 432 445 447 451 453

0 0 0 1 7 21 17 15 17 7 4 3 1 3 0 2 0 7 3 1 0

0.00 0.00 0.00 0.24 1.56 4.57 4.17 4.13 4.97 2.15 1.18 0.85 0.25 0.68 0.00 0.44 0.00 1.52 0.66 0.22 0.00

1 1 1 9 21 68 48 30 26 19 17 9 21 22 12 19 23 8 8 8 4

0.36 0.30 0.26 2.18 4.68 14.78 11.76 8.26 7.60 5.85 5.03 2.54 5.19 4.95 2.69 4.19 5.05 1.74 1.75 1.74 0.88

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

5334 6018 6679 7042 7594 7769 7409 7066 6776 6066 5617 5281 5254 5219 5178 5124 4832 4572 4495 4340 4310

19 25 19 20 19 53 83 87 130 194 129 85 29 37 20 25 75 95 34 33 10

0.35 0.40 0.27 0.27 0.24 0.63 0.99 1.08 1.69 2.78 2.06 1.46 0.51 0.66 0.36 0.45 1.42 1.91 0.70 0.70 0.23

107 171 279 392 405 572 873 928 788 718 506 459 377 378 377 448 364 310 305 310 119

1.96 2.75 4.00 5.26 5.05 6.81 10.44 11.48 10.24 10.29 8.09 7.88 6.66 6.71 6.76 8.00 6.91 6.23 6.31 6.62 2.68

Economy:TWN

Economy:VEN

Defaults

Others

Defaults

Others

Year

Active

#

%

#

%

Year

Active

#

%

#

%

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

236 259 291 371 442 496 572 689 787 875 982 1070 1340 1351 1377 1434 1438 1492 1591 1657 1680

0 0 0 0 0 0 4 7 8 9 7 2 6 8 3 3 6 1 0 0 0

0.00 0.00 0.00 0.00 0.00 0.00 0.68 0.99 0.99 0.99 0.68 0.18 0.44 0.56 0.21 0.20 0.40 0.07 0.00 0.00 0.00

2 1 1 0 1 3 12 8 15 21 36 19 26 66 41 33 48 22 24 40 12

0.84 0.38 0.34 0.00 0.23 0.60 2.04 1.14 1.85 2.32 3.51 1.74 1.90 4.63 2.89 2.24 3.22 1.45 1.49 2.36 0.71

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

0 7 12 15 15 47 40 39 37 30 20 25 27 29 27 24 24 26 23 31 23

0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0

NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.33 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0 0 1 3 2 17 25 18 12 12 18 10 8 7 7 8 32 21 14 15 12

NaN 0.00 7.69 16.67 11.76 26.56 38.46 31.58 24.49 27.91 47.37 28.57 22.86 19.44 20.59 25.00 57.14 44.68 37.84 32.61 34.29

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 159

159

8/30/2012 9:36:38 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

APPENDIX B: PERFORMANCE ANALYSIS

Figure B.1 Plots of US default parameters across all horizons for the Stock index one-year return, Short-term interest rate, DTD Level, DTD Trend, CASH/TA Level and CASH/TA Trend. Solid lines are the parameter estimates and dashed lines are the 90% confidence level. Horizontal axis is the horizon in months.

160

1250007.indd 160

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:38 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Figure B.2 Plots of US default parameters across all horizons for the NI/TA Level, NI/TA Trend, SIZE Level, SIZE Trend, M/B and SIGMA. Solid lines are the parameter estimates and dashed lines are the 90% confidence level. Horizontal axis is the horizon in months.

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 161

161

8/30/2012 9:36:40 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table B.1 Accuracy Ratios (AR) and Area Under Receiver Operating Characteristic(AUROC) for calibration groups and economies with more than 20 defaults in the testing set. Standard errors are reported in parentheses. AR

AUROC

Economy

1 mth

6 mth

1 yr

2 yr

1 mth

6 mth

1 yr

2 yr

North America

0.940 (0.007) 0.944 (0.008)

0.887 (0.009) 0.899 (0.011)

0.813 (0.011) 0.826 (0.014)

0.719 (0.013) 0.755 (0.019)

0.970 (0.003) 0.972 (0.004)

0.943 (0.004) 0.949 (0.005)

0.907 (0.006) 0.913 (0.007)

0.859 (0.007) 0.878 (0.010)

0.874 (0.017) 0.881 (0.017)

0.797 (0.021) 0.766 (0.022)

0.748 (0.023) 0.730 (0.024)

0.648 (0.027) 0.661 (0.030)

0.937 (0.008) 0.940 (0.008)

0.899 (0.010) 0.883 (0.011)

0.874 (0.011) 0.865 (0.012)

0.824 (0.013) 0.831 (0.015)

0.861 (0.014) 0.857 (0.017)

0.786 (0.017) 0.756 (0.021)

0.735 (0.018) 0.683 (0.024)

0.673 (0.020) 0.610 (0.028)

0.930 (0.007) 0.929 (0.008)

0.893 (0.008) 0.878 (0.010)

0.867 (0.009) 0.842 (0.012)

0.836 (0.010) 0.805 (0.014)

0.821 (0.023) 0.835 (0.031)

0.774 (0.026) 0.770 (0.036)

0.706 (0.029) 0.717 (0.041)

0.688 (0.030) 0.687 (0.048)

0.910 (0.012) 0.918 (0.015)

0.887 (0.013) 0.885 (0.018)

0.853 (0.014) 0.858 (0.021)

0.844 (0.015) 0.843 (0.024)

0.815 (0.036) 0.813 (0.038)

0.733 (0.042) 0.708 (0.046)

0.669 (0.046) 0.641 (0.053)

0.581 (0.052) 0.543 (0.061)

0.908 (0.018) 0.906 (0.019)

0.866 (0.021) 0.854 (0.023)

0.835 (0.023) 0.820 (0.026)

0.791 (0.026) 0.772 (0.031)

0.934 (0.025) 0.939 (0.027)

0.872 (0.034) 0.879 (0.038)

0.775 (0.044) 0.773 (0.053)

0.676 (0.051) 0.717 (0.061)

0.967 (0.013) 0.970 (0.014)

0.936 (0.017) 0.939 (0.019)

0.888 (0.022) 0.887 (0.027)

0.838 (0.025) 0.859 (0.031)

0.563 (0.024) 0.591 (0.027)

0.531 (0.026) 0.545 (0.029)

0.483 (0.028) 0.490 (0.032)

0.398 (0.031) 0.400 (0.038)

0.781 (0.012) 0.795 (0.013)

0.766 (0.013) 0.773 (0.014)

0.741 (0.014) 0.745 (0.016)

0.699 (0.015) 0.700 (0.019)

0.898 (0.027) 0.890 (0.029)

0.795 (0.037) 0.773 (0.040)

0.727 (0.041) 0.722 (0.044)

0.575 (0.051) 0.623 (0.060)

0.949 (0.014) 0.945 (0.015)

0.897 (0.018) 0.886 (0.020)

0.864 (0.021) 0.861 (0.022)

0.787 (0.025) 0.811 (0.030)

0.918 (0.064) 0.875 (0.082)

0.918 (0.064) 0.829 (0.094)

0.790 (0.097) 0.838 (0.101)

0.782 (0.101) 0.757 (0.129)

0.959 (0.032) 0.937 (0.041)

0.959 (0.032) 0.914 (0.047)

0.895 (0.048) 0.919 (0.051)

0.891 (0.050) 0.878 (0.065)

0.892 (0.049) 0.814 (0.064)

0.818 (0.062) 0.719 (0.076)

0.795 (0.065) 0.765 (0.074)

0.737 (0.073) 0.727 (0.088)

0.946 (0.025) 0.907 (0.032)

0.9 (0.031) 0.860 (0.038)

0.897 (0.033) 0.882 (0.037)

0.869 (0.036) 0.863 (0.044)

North America Europe Europe ADvped ADvped Emerging Emerging AUS AUS CAN CAN CHN CHN DEU DEU DNK DNK FRA FRA

In/Out of sample In Out In Out In Out In Out In Out In Out In Out In Out In Out In Out

(Continued )

162

1250007.indd 162

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:41 AM

b1423

Global Credit Review Volume 2

Table B.1

30 August 2012 9:08 AM

(Continued )

AR Economy GBR GBR HKG HKG IDN IDN IND IND JPN JPN KOR KOR MYS MYS NLD NLD NOR NOR PHL PHL

AUROC

1mth

6mth

1yr

2yr

1mth

6mth

1yr

2yr

0.875 (0.032) 0.876 (0.032)

0.808 (0.039) 0.759 (0.043)

0.733 (0.044) 0.668 (0.050)

0.646 (0.050) 0.590 (0.058)

0.938 (0.016) 0.938 (0.016)

0.904 (0.019) 0.880 (0.022)

0.867 (0.022) 0.834 (0.025)

0.823 (0.025) 0.795 (0.029)

0.678 (0.075) 0.635 (0.086)

0.479 (0.087) 0.384 (0.100)

0.437 (0.089) 0.312 (0.108)

0.360 (0.092) 0.153 (0.117)

0.839 (0.037) 0.817 (0.043)

0.740 (0.044) 0.692 (0.050)

0.718 (0.044) 0.656 (0.054)

0.680 (0.046) 0.577 (0.059)

0.717 (0.068) 0.625 (0.131)

0.708 (0.069) 0.613 (0.132)

0.555 (0.079) 0.521 (0.162)

0.588 (0.079) 0.370 (0.210)

0.859 (0.034) 0.812 (0.065)

0.854 (0.034) 0.807 (0.066)

0.777 (0.039) 0.761 (0.081)

0.794 (0.039) 0.685 (0.105)

0.657 (0.065) 0.603 (0.085)

0.666 (0.064) 0.541 (0.089)

0.578 (0.071) 0.495 (0.091)

0.478 (0.076) 0.306 (0.096)

0.828 (0.032) 0.802 (0.043)

0.833 (0.032) 0.770 (0.044)

0.789 (0.035) 0.747 (0.046)

0.739 (0.038) 0.653 (0.048)

0.906 (0.021) 0.907 (0.029)

0.866 (0.025) 0.845 (0.036)

0.822 (0.028) 0.791 (0.043)

0.782 (0.031) 0.797 (0.043)

0.953 (0.011) 0.953 (0.014)

0.933 (0.012) 0.922 (0.018)

0.911 (0.014) 0.895 (0.021)

0.891 (0.016) 0.898 (0.021)

0.887 (0.024) 0.908 (0.034)

0.786 (0.031) 0.785 (0.049)

0.737 (0.034) 0.733 (0.057)

0.692 (0.036) 0.696 (0.065)

0.943 (0.012) 0.954 (0.017)

0.893 (0.016) 0.893 (0.025)

0.868 (0.017) 0.867 (0.028)

0.846 (0.018) 0.848 (0.033)

0.857 (0.039) 0.846 (0.048)

0.795 (0.045) 0.761 (0.057)

0.743 (0.050) 0.706 (0.067)

0.693 (0.053) 0.654 (0.075)

0.928 (0.019) 0.923 (0.024)

0.898 (0.022) 0.881 (0.029)

0.871 (0.025) 0.853 (0.033)

0.846 (0.027) 0.827 (0.038)

0.806 (0.089) 0.783 (0.098)

0.735 (0.101) 0.540 (0.128)

0.791 (0.092) 0.684 (0.124)

0.743 (0.092) 0.643 (0.210)

0.903 (0.045) 0.892 (0.049)

0.867 (0.050) 0.770 (0.064)

0.896 (0.046) 0.842 (0.062)

0.871 (0.050) 0.822 (0.105)

0.944 (0.055) 0.931 (0.061)

0.952 (0.052) 0.897 (0.075)

0.829 (0.094) 0.679 (0.120)

0.731 (0.116) 0.685 (0.143)

0.972 (0.027) 0.965 (0.030)

0.976 (0.026) 0.949 (0.037)

0.915 (0.047) 0.840 (0.060)

0.865 (0.058) 0.842 (0.071)

0.641 (0.103) 0.653 (0.114)

0.600 (0.107) 0.589 (0.120)

0.569 (0.109) 0.587 (0.124)

0.647 (0.107) 0.678 (0.125)

0.820 (0.052) 0.826 (0.057)

0.8200 (0.053) 0.795 (0.060)

0.784 (0.053) 0.793 (0.062)

0.823 (0.054) 0.839 (0.062)

In/Out of sample In Out In Out In Out In Out In Out In Out In Out In Out In Out In Out

(Continued )

GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 163

163

8/30/2012 9:36:41 AM

b1423

Global Credit Review Volume 2

Table B.1

30 August 2012 9:08 AM

(Continued )

AR Economy

AUROC

In/Out of sample

1mth

6mth

1yr

2yr

1mth

6mth

1yr

2yr

0.769 (0.074) 0.781 (0.081)

0.759 (0.076) 0.760 (0.086)

0.606 (0.091) 0.566 (0.105)

0.457 (0.099) 0.393 (0.121)

0.884 (0.037) 0.891 (0.040)

0.879 (0.038) 0.880 (0.043)

0.803 (0.046) 0.783 (0.053)

0.729 (0.050) 0.696 (0.060)

SWE

0.838 (0.074)

0.761 (0.087)

0.733 (0.091)

0.624 (0.114)

0.919 (0.037)

0.880 (0.044)

0.866 (0.046)

0.812 (0.057)

In

SWE

0.849 (0.076)

0.759 (0.093)

0.726 (0.100)

0.689 (0.127)

0.925 (0.038)

0.879 (0.046)

0.863 (0.050)

0.845 (0.063)

Out

THA

0.854 (0.037)

0.812 (0.041)

0.742 (0.047)

0.679 (0.051)

0.927 (0.018)

0.906 (0.021)

0.871 (0.023)

0.839 (0.026)

In

THA

0.907 (0.053)

0.858 (0.064)

0.768 (0.084)

0.724 (0.103)

0.953 (0.026)

0.929 (0.032)

0.884 (0.042)

0.862 (0.051)

Out

TWN

0.866 (0.046)

0.781 (0.056)

0.759 (0.056)

0.697 (0.056)

0.933 (0.023)

0.891 (0.028)

0.880 (0.028)

0.849 (0.028)

In

TWN

0.928 (0.044)

0.792 (0.071)

0.664 (0.092)

0.616 (0.096)

0.964 (0.022)

0.896 (0.035)

0.832 (0.046)

0.808 (0.048)

Out

USA

0.940 (0.007)

0.888 (0.009)

0.816 (0.012)

0.722 (0.014)

0.970 (0.003)

0.944 (0.005)

0.908 (0.006)

0.861 (0.007)

In

USA

0.944 (0.008)

0.901 (0.011)

0.833 (0.015)

0.765 (0.020)

0.972 (0.004)

0.950 (0.005)

0.916 (0.007)

0.882 (0.010)

Out

USA(Fin)

0.963 (0.021)

0.933 (0.029)

0.880 (0.038)

0.834 (0.047)

0.981 (0.011)

0.967 (0.014)

0.940 (0.019)

0.917 (0.023)

Out

USA(non-Fin)

0.945 (0.009)

0.893 (0.013)

0.832 (0.016)

0.759 (0.022)

0.972 (0.004)

0.946 (0.006)

0.916 (0.008)

0.880 (0.011)

Out

SGP SGP

In Out

Note: North America includes Canada and the US. Europe includes 23 European countries. ADvped stands for Asian developed group, which includes Australia, Hong Kong, Japan, South Korea, Singapore, and Taiwan. Emerging stands for emerging group, which includes Indonesia, Malaysia, Philippines, Thailand, Argentina, Brazil, Colombia, Chile, Mexico, Peru, and Venezuela.

164

1250007.indd 164

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:41 AM

b1423

Global Credit Review Volume 2

Figure B.3

30 August 2012 9:08 AM

Performance test for the North America Group, in sample.

Figure B.4

Performance test for the Europe Group, in sample. GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 165

165

8/30/2012 9:36:41 AM

b1423

Global Credit Review Volume 2

Figure B.5

Performance test for the Asian developed group, in sample.

Figure B.6 166

1250007.indd 166

30 August 2012 9:08 AM

Performance test for the Emerging group, in sample.

NUS-RMI CREDIT RESEARCH INITIATIVE TECHNICAL REPORT

8/30/2012 9:36:43 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Figure B.7

Performance test for China, in sample.

Figure B.8

Performance test for India, in sample. GLOBAL CREDIT REVIEW VOLUME 2

1250007.indd 167

167

8/30/2012 9:36:45 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

This page intentionally left blank

1250007.indd 168

8/30/2012 9:36:47 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

A Lead-Lag Investigation of RMI PD and CRA Ratings

INTRODUCTION

I

RMI staff article For any questions or comments on this article, please contact Elisabeth Van Laere at [email protected]

n this article, we present the results of a detailed comparison between the Risk Management Institute’s (RMI) Credit Research Initiative’s (CRI) probability of default (PD) and the rating actions of external credit rating agencies (CRAs). For three well-known default cases: MF Global, Lehman Brothers and Enron, we compare the RMI PD with the rating action of credit rating agencies Fitch, Moody’s, S&P and DBRS. In doing this analysis, we have to deal with the fact that in contrast to typical ratings that use letter grades, the CRI produces more granular credit information via PDs. Hence, in order to allow for a meaningful comparison between the RMI PD and the ratings of the CRAs, we translate the long-term issuer ratings (or a comparable rating if not available) into an implied probability of default. The default probabilities per rating grade are derived from data available on ESMA’s Central Rating Repository.1 A detailed methodology is available upon request.2

In doing this comparison we should keep in mind that credit ratings from the CRAs are supposed to be a through-the-cycle measure of credit ratings and are hence expected to be relatively stable throughout time. The RMI PD on the other hand is meant to be a timely point-in-time credit risk measure that is updated on a daily basis.3 Our analysis shows that having such a timely measure provides useful and necessary information to financial markets. In all three cases RMI PD responded to declines in credit quality well before the CRAs started to revise their ratings. •

MF Global CRI data first indicated the company’s credit profile was significantly deteriorating in August 2011, before its bankruptcy filing at the end of October that year.

•

Lehman Brothers CRI data from 1995 to 2008 shows that Lehman Brothers should not have been consistently rated above investment GLOBAL CREDIT REVIEW VOLUME 2

1250008.indd 169

169

8/30/2012 9:35:56 AM

b1423

Global Credit Review Volume 2

grade by the major ratings agencies, yet the company was only downgraded below investment grade when Lehman filed for bankruptcy on September 15, 2008. •

Enron In comparison to the static ratings from the CRAs, a gradual increase in Enron’s RMI PD highlighted the upward trend in credit risk the company faced throughout 2000 before its eventual default at the end of 2001.

The lack of timeliness and accuracy in CRAs’ ratings provides a strong justification to use other alternative credit risk indicators which can act as a complement and contribute to the critical area of credit ratings in a non-biased way. It is clear that the RMI PD leads the CRAs rating actions and provides more timely information to the market.

I. THE BANKRUPTCY OF MF GLOBAL MF Global was an international brokerage firm specializing in commodities and derivatives, with a history stretching back over 200 years. The now infamous proprietary bet on eurozone sovereign debt caused the firm to file for bankruptcy. Despite roots in a traditionally safe part of the financial sector, total losses of over $570 m, incurred while the firm was a publicly listed entity between July 2007 and October 2011, underlined large failures in risk management, internal control and business strategy. However, the firm was consistently rated above investment grade by the Big Three CRAs (see Figure 1). CRI data shows that default was never a remote possibility for MF Global, especially in the final few months prior to the firm’s bankruptcy. Ultimately, concerns about off-balance sheet leverage and the company’s liquidity eventually led to a loss of client and counterparty confidence, and the firm’s Chapter 11 bankruptcy filing on October 31, 2011.

1.1. MF Global’s Rating History MF Global was originally the brokerage house of the Man Group, a UK-based alternative investment 170

1250008.indd 170

30 August 2012 9:08 AM

management business. The division traditionally earned revenues from execution and clearing fees, matched principle spreads and investing client credit balances in higher yielding securities. The company was spun off from the Man Group on July 19, 2007. The relatively low risk of MF Global’s businesses was recognized by the three largest CRAs in their initial ratings for MF Global in 2007, with the firm receiving a BBB+ rating from Fitch, an A3 rating from Moody’s and a BBB+ rating from S&P. The three ratings agencies cited MF Global’s reputation for prudent liquidity and risk management while part of the Man Group, as additional reasons for assigning MF Global investment grade ratings. Despite this, the CRAs’ opinions diverged on the prospects for MF Global. The firm received an A3 rating from Moody’s, one notch higher than what it assigned former parent Man Group as Moody’s believed the high-growth opportunities in exchange and OTC derivatives businesses would allow MF Global to leverage its specialist brokerage business model.4 Fitch’s initial BBB+ rating for MF Global was one notch below Man Group’s contemporary A- rating, as Fitch believed growing competition in the aforementioned sectors would affect MF Global’s market share.5 S&P did not assign a rating to the Man Group until September 2008. MF Global was downgraded by two of the three major ratings agencies on February 29, 2008, after reports emerged that a rogue trading incident at MF Global had resulted in a $140 mn loss. Although the firm had a history of prudent risk management practices, Moody’s and S&P expressed concerns about MF Global’s internal risk controls in ratings commentary. Accordingly, Moody’s downgraded MF Global to Baa1 from A36 while S&P downgraded the firm to BBB from BBB+. All three major ratings agencies placed MF Global on Negative Watch following the incident. An additional reason for the Negative Watches was ongoing concerns that MF Global had not made sufficient progress in changing its capital structure and liquidity profile following the divestment from the Man Group. MF Global was still relying on $1.4bn in bridge loans and had reported 50-to-1 accounting leverage in regulatory filings for the period ending December 31, 2007.

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:56 AM

1250008.indd 171

100000 Files for bankruptcy protection Reports $88.9m quarterly loss

Reports quarterly loss of $186.6m

John Corzine appointed CEO

10000

Eurozone trade receives significant media attention

Reports weak earnings for fiscal year 2009

b1423

Lehman bankruptcy

Discloses FINRA's capital treatment modification request

Reports net annual loss of $130.5m First disclosure of $6.3bn eurozone trade

Demise of Bear Stearns

Receives primary dealer status

1000

PD (bps)

$140m loss (trading incident)

Spun off from Man Group Ba2

100

BBBBaa2

Baa3

BBB B 1 Baa1

BBB+ A3

GLOBAL CREDIT REVIEW VOLUME 2

1 2/28/2007

BBB

BBB+

8/31/2007

2/29/2008

RMI 1-year PD

8/31/2008

Fitch

2/28/2009

Moody's

Figure 1.

8/31/2009

S&P

2/28/2010

Initial

MF Global Holdings.

8/31/2010

Positive

2/28/2011

Negative

8/31/2011

Stable

30 August 2012 9:08 AM

10

BBBBB+

Global Credit Review Volume 2

Issuance of $450m preferred equity and debt

171

8/30/2012 9:35:56 AM

b1423

Global Credit Review Volume 2

Following the demise of Bear Stearns in March 2008, a global easing in monetary policy began to affect MF Global’s net interest rate revenues, which made up almost a third of the firm’s revenues. A reduction in this source of revenue was offset by increased ‘execution and clearing’ revenues, as participation in futures markets increased in line with increased volatility in the middle of 2008. However, MF Global reported a net annual loss of $130.5 mn for the company’s first fiscal year ending March 31, 2008. On June 17, MF Global announced plans to issue $300 mn of convertible preferred equity and unsecured debt, in order to pay down a part of the outstanding $1.4 bn in bridge loans. Despite the announced improvement in MF Global’s capital structure, Moody’s confirmed MF Global’s rating at Baa1 with a Negative Outlook the following day as Moody’s was concerned about a decrease in net interest revenues and a lack of progress in the implementation of improved enterprise risk management following the rogue trading incident.7 Fitch affirmed MF Global at BBB+ with a stable outlook the same day, as Fitch believed that deficiencies in MF Global’s internal controls had been largely addressed, and the firm had reduced counterparty and liquidity risks through a tightening in customer financing terms. MF Global’s RMI PD history begins shortly after this period,8 on July 1, 2008. CRI’s initial PD for MF Global was 2670 bps, a result of the firm’s high idiosyncratic risk, low distance-to-default (DTD) and the $130.5mn loss for the financial year ending March 31, 2008 on June 13. In comparison, the aggregate PD for the US financial sector on July 1 was 277 bps the same day. On July 18, MF Global issued $450mn of convertible preference shares and unsecured notes, $150mn more than originally planned. Because of this improvement in MF Global’s capital structure, S&P removed the firm from CreditWatch Negative the same day. However, S&P noted that MF Global’s continuing use of high levels of leverage was a large risk to the firm. On July 17, the 1-year PD for MF Global closed at 3872 bps, down from a local high of 5675 bps on July 15. This change was largely driven by a V-shaped recovery in the company’s share price between July 8 and July 18. MF Global’s PD continued to ease during August 2008. 172

1250008.indd 172

30 August 2012 9:08 AM

1.2. Overcoming Market Forces In response to the collapse of a number of global financial institutions in September 2008, central banks around the world cut key interest rates to contain significant upheaval in global credit markets. This negatively affected MF Global’s net interest revenues. Furthermore, as large institutional clients withdrew their excess funds for the safety of FDIC insured accounts and treasuries, MF Global also experienced a minor silent run. Futures trading slowed in late-2008 further constraining MF Global’s revenues from brokerage and clearing services. The 1-year PD for MF Global had spiked following the Lehman Brothers bankruptcy on September 15, and it remained in an elevated plateau during this period, closing at 3414 bps on December 3. S&P placed MF Global on CreditWatch Negative on December 4. In their rating commentary, S&P stated that MF Global’s revenues would suffer from constrained futures trading and that any material reduction in interest revenues or cash on hand would result in a rating downgrade. Moody’s cited similar reasons in its decision to downgrade MF Global to Baa2 on January 16, 2009.9 MF Global’s 1-year PD remained in the same plateau on January 15, at 3309 bps. This prolonged peak in MF Global’s PD during the last quarter of 2008 and early 2009 was caused by decreasing DTD and a significant decline in market capitalization. In the following months, MF Global benefited from its position as a globally diversified broker with low proprietary trading activities. MF Global also applied to the Federal Reserve for approval to become a primarydealer during this period, which would allow MF Global to trade highly liquid fixed income securities directly with the Federal Reserve. MF Global’s PD declined from previous heights during this period to 556 bps on June 16, 2009, due to a reduction in the firm’s leverage (measured as the ratio of total assets to equity) to 26.8-to-1 which contributed to an increase in MF Global’s DTD. The following day, Fitch announced it was downgrading the firm to BBB from BBB+. Fitch believed that MF Global’s weak earnings in the fiscal year ending March 31, 2009, and its continued reliance on revenues from exchange traded brokerage services,

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:57 AM

b1423

Global Credit Review Volume 2

materially affected the firm’s credit outlook. At the same time, MF Global was on a Stable Outlook at Moody’s and a Negative Outlook at S&P. Throughout 2009, MF Global’s revenues continued to decline due to lower transaction volumes, and consistently low interest rates. In response, the firm dramatically increased its leverage to 41.1-to-1 in November 2009, in order to increase the volume of revenues from matched book portfolios generating low net interest revenue. This was the primary reason why Moody’s changed the firm’s ratings outlook to negative from stable on November 6; Moody’s believed higher leverage had the potential to cause funding frictions.10 Conversely, the CRI 1-year PD had significantly decreased from prior highs during 2009, falling to 595 bps on November 5. This protracted downwards PD trend was caused by an increase in MF Global’s DTD, cash and net income. Although MF Global’s leverage increased dramatically from 26.8-to-1 (reported June 10, 2009) to 37.1to-1 (reported November 6, 2009), this increased risk is not reflected in the PD at that time, as the stock was on an increasing trend. 1.2.1. A key turnaround On March 23, 2010, MF Global announced that it had appointed John Corzine as the company’s new Chairman of the Board and CEO. Mr Corzine was the CEO of Goldman Sachs from 1994 to 1999 and a former Senator and Governor of New Jersey. A key reason for his appointment was to turn around poor earnings by expanding the firm’s business beyond futures brokerage into proprietary trading. Following the announcement, Fitch placed MF Global on a Rating Watch Negative, on concerns that Mr Corzine’s sudden appointment could materially impact future managerial and strategic changes. Fitch was also concerned about the firm’s ability to generate adequate capital internally, given poor earnings in previous periods. Moody’s believed Mr Corzine’s appointment had no ratings implications, but maintained a negative ratings outlook on MF Global due to continuing use of high levels of leverage amidst weakened profitability.11 Despite CRA concerns, the market viewed Mr Corzine’s appointment favorably, as the company’s

30 August 2012 9:08 AM

share price increased 10% on March 24, the day after his appointment. Market participants believed Mr Corzine’s expertise and industry connections would allow MF Global to overcome a recent poor earnings profile. This positive perception in the stockmarket led to a decrease in MF Global’s 1-year PD, due to increasing DTD and a decrease in idiosyncratic risk. However, RMI PD was still higher than probabilities of default implied by CRA ratings. In Mr Corzine’s first annual earnings call on May 20, 2010, MF Global announced an $88.9 mn quarterly loss. In the same announcement, Mr Corzine mapped out a new firm wide strategic plan involving significant staff reductions, improvements to the company’s capital structure, and a much greater focus on product cross-selling. Despite these planned changes, Fitch maintained MF Global’s Rating Watch Negative on May 20 due to continued weak earnings performance and strategic risks.12 MF Global reported a $38.8 mn quarterly loss on November 5, 2010, due to a write-off of deferred tax assets and costs associated with capital restructuring. On November 24, S&P downgraded MF Global to its lowest investment grade rating BBB- from BBB. S&P said the downgrade followed a protracted period of weak profitability and ongoing restructuring. In addition, S&P believed that capital and liquidity levels at that time provided for only modest expansion into principal activities. S&P noted that MF Global’s investment grade rating was supported by the firm’s efforts to reduce leverage. Under Mr Corzine, MF Global’s leverage had decreased to 29.7-to-1 in the quarter ending September 30, 2010, down from a high of 41.1-to-1 a year earlier. MF Global’s 1-year PD had fallen to 46 bps on November 23, although the company’s PD surged to 116 bps on December 31. The sudden drop in MF Global’s PD on September 30 was caused by a 60% decrease in the firm’s long-term liabilities, which raised its DTD. The subsequent increase on December 31 was caused by an increase in liabilities and decrease in DTD.

1.3. Increasing Risk and Bankruptcy MF Global received Primary Dealer status from the Federal Reserve on February 2, 2011, allowing the GLOBAL CREDIT REVIEW VOLUME 2

1250008.indd 173

173

8/30/2012 9:35:57 AM

b1423

Global Credit Review Volume 2

company to trade highly liquid US securities directly with the Federal Reserve. Moody’s acknowledged that this was a credit positive for MF Global, but maintained a negative outlook on the firm due to concerns about earnings, leverage and risk management, stemming from the firm’s expanding role as a principal risk-taking dealer. The same day, Fitch removed MF Global from Rating Watch Negative, but assigned the firm a Negative Outlook reflecting uncertainty about the firm’s earning plans, and the possibility that the company’s risk profile could diverge dramatically.13 Underlining increased risk within the company’s business, MF Global’s 1-year PD had increased to 139 bps on February 1, as the firm’s DTD, net income and relative size had all decreased during January. Despite improvements in leverage under Mr Corzine, an increase in proprietary trading increased the company’s risk profile. MF Global first disclosed a $6.3bn proprietary off-balance sheet position in eurozone sovereign debt in its 2011 annual report, filed on May 20, 2011. The position was structured using repo-to-maturity trades, or repos that expire at the same time the underlying collateral matures, and allowed MF Global to move them off the company’s balance sheet. However, the company retained exposures to the market and default risk of the transactions. In addition, MF Global reported in annual filings that it had replaced its Chief Risk Officer on January 31, 2011, significantly diluting the position’s seniority and removing direct reporting requirements to the CEO and the Board of Directors.14 These disclosures should have alerted CRAs to the increasing appetite for risk at MF Global, which they had previously cited as something that may lead to rating downgrades. The company also reported a $46.5 mn loss for the quarter ending March 31, 2011 on May 20. MF Global’s 1-year PD remained elevated at 173 bps on May 23, due to decreasing DTD caused by a 7.7% decline in the company’s share price after the firm reported the aforementioned quarterly loss. Despite implicit disclosures of an increasing risk appetite at MF Global in its annual report in May, the market did not react strongly. The company’s 1-year PD remained in a narrow band between 140 bps and 190 bps during June and July 2011. 174

1250008.indd 174

30 August 2012 9:08 AM

CRI data first indicated the company’s credit profile was significantly deteriorating in August 2011; the company’s 1-year PD increased significantly during August to 237 bps on August 31, due to a marked decrease in DTD. The company’s shares had fallen 23% in the first week of August after MF Global sold $325m of bonds on August 5 with an interesting feature: bondholders would receive an extra 1% annual interest if Mr Corzine was appointed to a federal position by the US President. This ‘key man’ provision underlined Mr Corzine’s importance to the company and indicated there was a chance he could leave in the near future, placing downward pressure on the company’s share price.15 Fitch and Moody’s maintained a Negative Outlook for MF Global during this period but the firm’s rating was not revised downward, while the company remained on a Stable Outlook at S&P. In an amendment to quarterly filings on September 1, 2011, MF Global disclosed that the Financial Industry Regulatory Authority (FINRA) had asked the company to modify the capital treatment of its eurozone repo-to-maturity trades in August. To address FINRA’s request, MF Global used part of the proceeds from its $325mn August 5 bond issuance to increase its working capital. According to CRI data, MF Global’s 1-year PD increased to 256 bps on September 2, from 237 bps on August 31, due to a decrease in the company’s stock price and DTD. The firm’s 1-year PD continued to increase during September to 405 bps on September 30. In early October, CRI data gave early warning signs that MF Global was in trouble. A large decline in MF Global’s DTD and increasing idiosyncratic risk contributed to a jump in the firm’s 1-year PD to 462 bps on October 7. A week later MF Global’s 1-year PD increased overnight to 514 bps on October 13. On October 16, a prominent Wall Street Journal piece broadly disseminated FINRA’s request to the market, noting the potential perils of a higher appetite for risk under Mr Corzine.16 This was the first time MF Global’s $6.3bn proprietary eurozone trade received significant coverage in the global press; CRI’s 1-year PD for MF Global increased the following day to 569 bps as the company’s DTD decreased significantly. No rating downgrades occurred during

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:57 AM

b1423

Global Credit Review Volume 2

the following week, although MF Global remained on negative outlook at all three major CRAs. The company’s 1-year PD increased 43 bps to 612 bps during the week ending October 21. The end began on October 24 with a steep increase in the PD and a flurry of CRA downgrades. These are shown in Table 1. Moody’s downgraded MF Global to Baa3, Moody’s lowest investment grade rating, stating that a further downgrade was under consideration. Moody’s believed that the prevailing low interest rates at that time would prevent MF Global from reaching the earnings and leverage requirements that Moody’s had previously set for MF Global to avoid a downgrade. Moody’s also cited increasing concern about senior management’s ability and willingness to prudently manage increasing risk, specifically the company’s eurozone sovereign debt positions.17 On October 25, MF Global announced that it would report a record $186.6mn loss for the quarter ending September 30. This news, combined with Moody’s downgrade the day before, caused the company’s shares to almost halve on October 25. Several media outlets reported that Mr Corzine and senior executives were looking to sell part or all of the company on October 26. The same day, S&P placed MF Global on CreditWatch negative on October 26, indicating that there was a high chance S&P would downgrade MF Global below investment grade. S&P cited continuing weak earnings and increased market pressure on the company’s capital base.18 MF Global’s 1-year PD had closed at 1734 bps on October 25, a more than 1000 bps increase since October 21, due to large decreases in DTD and relative size. On October 27, both Fitch and Moody’s downgraded MF Global below investment grade. Moody’s Table 1.

Date

Fitch

30 August 2012 9:08 AM

downgraded the company another notch to Ba2, and left the company on review for a further downgrade.19 Moody’s believed the increasing appetite for risk revealed by the eurozone trades and the significant quarterly loss increased the risk that MF Global could lose customer and counterparty confidence. Moody’s noted that the new rating was supported by MF Global’s adequate liquidity profile and the transparent pricing of the firm’s other assets. Fitch downgraded MF Global two notches to BB+ and placed the company on Rating Watch Negative.20 Fitch believed that increased risk taking at MF Global had resulted in large concentrated positions which could severely impact the firm’s capital and liquidity base. On October 28, rumors abounded that MF Global had drawn down all of its credit lines to cope with a potential run on the company, and that the company’s eurozone sovereign debt trades were causing major liquidity problems for the firm. Although the risk of default on MF Global’s eurozone sovereign debt positions was relatively low due to an implicit guarantee from the European Financial Stability Facility, the key risk MF Global faced was the increasing margin calls the firm had to meet as yields increased across the eurozone in late October. The credit downgrades also triggered additional margin calls and increased collateral demands from short-term repo counterparties. At the close of trade on October 28, MF Global’s 1-year PD had jumped to 3689 bps, as the company’s DTD and relative size declined further. This change indicated that there was more than a one in three chance the company would default in the next year. Over the weekend beginning October 29, Mr Corzine and senior MF Global executives frantically searched for a buyer. In the early hours of October 31, the Wall Street Journal reported that MF Global would seek

MF Global: Credit ratings and RMI 1-year PD during final week.

Moody’s

S&P

RMI 1-year PD (bps), Previous close*

Oct 24 Downgrade, Baa3 612 Oct 25 690 Oct 26 CreditWatch, BBB− 1734 Oct 27 Downgrade, BB+ Downgrade, Ba2 2080 Oct 28 2875 Oct 31 Downgrade, C Downgrade, Caa1 Downgrade, D 3689 *Previous closes are used to avoid incorporating the market effect of new rating changes into PD data. GLOBAL CREDIT REVIEW VOLUME 2

1250008.indd 175

175

8/30/2012 9:35:57 AM

b1423

Global Credit Review Volume 2

Chapter 11 bankruptcy protection the same day, with competitor Interactive Brokers purchasing part of MF Global’s business for $1bn. According to media reports released following the firm’s bankruptcy petition, Interactive Brokers had become aware that a large amount of money was missing from segregated customer accounts in the early hours of October 31 and had cancelled all potential deals. MF Global subsequently filed for Chapter 11 bankruptcy protection.21

II. THE COLLAPSE OF LEHMAN BROTHERS CRI data from 1995 to 2008 shows that Lehman Brothers should not have been consistently rated above investment grade by the major ratings agencies, yet the company was only downgraded below investment grade when Lehman filed for bankruptcy on September 15, 2008. The failure and subsequent sale of Bear Stearns to J.P. Morgan in March 2008 had significantly increased the scrutiny of Lehman’s financial position by market participants in the months leading up to the Lehman’s September bankruptcy. Despite major debate amongst market participants regarding Lehman’s health during this period, the ratings agencies did not downgrade the bank significantly. Throughout this period CRI data showed that the risk of default at Lehman was, and had historically been, much higher than that implied by external credit ratings (see Figure 2). A study on the bankruptcy of Lehman Brothers from three years prior to its September 2008 filing has been previously conducted by Duan et al. (2012) where the authors relied on a new econometric model to produce the whole term structure of forward default probabilities. They showed quite remarkably that the term structures of forward default probabilities revealed by the model at different points of time leading up to the Lehman bankruptcy are quite consistently aligned with the actual occurrence. The RMI PD model is based on the econometric model of Duan et al. (2012). Our analysis of Lehman Brothers here is not to reassert how accurate the PD is; rather we want to shed light on the rating actions of the CRAs over this period.22 We start this analysis in 2005, 4 years before the company filed for bankruptcy. By that time Lehman 176

1250008.indd 176

30 August 2012 9:08 AM

had obtained a dominant position in the commercial securitization market and had significantly diversified its investment banking activities. The company had also increased its market position in the residential mortgage securitization business, through acquisitions of subprime lender BNC Mortgage and Alt-A loan specialist Aurora Loan Services. Lehman’s earnings were at record highs; the company had recently been upgraded by Moody’s and S&P had recently changed its credit outlook for Lehman to positive. S&P proceeded to upgrade Lehman to A+ from A on October 11, 2005, citing improvements in earnings diversification and risk management at Lehman. Between January 2004 and S&P’s rating upgrade, Lehman’s 1-year PD remained stable, increasing 3 bps to 43 bps on October 11. A decrease in DTD was offset by improved profitability and lower idiosyncratic risk.

2.1. Illiquid Positions Starting in 2007, Lehman’s inventory of residential mortgage-backed securities (RMBS) and collateralized debt obligations (CDO) positions became increasingly difficult to sell as demand for securitization began to fall at the same time that Lehman’s subprime subsidiaries were registering record loan volumes. During Q1 of 2006, Lehman’s subprime lending subsidiary BNC Mortgage was originating more than $1bn in subprime loans per month, while Aurora Loan Services was lending $3bn per month in the first quarter of 2007, becoming the largest originator of Alt-A mortgages.23 So-called ‘sticky’ inventory, mortgages in warehouse accounts and securitizations held for sale, began to build up on Lehman’s balance sheet. The collapse of two Bear Stearns hedge funds in June 2007 fuelled investor concerns about risky mortgage related bonds and loans. Despite a general market consensus that risk at broker-dealers specializing in mortgage related products was increasing, Fitch upgraded Lehman to AA- on June 28. Fitch cited Lehman’s strong financial results while volatility in key real estate businesses increased. Lehman’s 1-year PD had increased almost 60 bps to 158 bps in the six months prior to June 27, as the firm’s DTD decreased and idiosyncratic risk increased. In the

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:57 AM

1250008.indd 177

100000 Increases exposure to Californian real estate developer SunCal

Files for bankruptcy protection

Rumors about major divestiture

Shuts subsidiary BNC Mortgage

10000

b1423

Reports $3.9bn quarterly loss

Acquires a $5.4bn stake in Archstone Smith, a real estate investment trust

Global Credit Review Volume 2

Rumors about potenial privatization

Moody’s and S&P cut ratings on large number of securitizations

Announces $2.8bn loss Fed establishes PDCF

Quarterly filings reveal $79.6bn of mortgage exposures

1000

Demise of BearStearns

PD (bps)

Collapse of two Bear Stearns hedge funds

100

AH 10 A1 A+

A+ A+ GLOBAL CREDIT REVIEW VOLUME 2

AA1 2/28/2005

8/31/2005

DBRS

2/28/2006

Fitch

Moody's

8/31/2006

S&P

2/28/2007

RMI 1-year PD

Figure 2. Lehman Brothers Holdings.

8/31/2007

Initial

Positive

2/29/2008

Negative

8/31/2008

Stable

30 August 2012 9:08 AM

AH A2 A

AAL

A

177

8/30/2012 9:35:57 AM

b1423

Global Credit Review Volume 2

following weeks, the company’s PD began to increase dramatically, breaking through the 172 bps threshold on July 10 that had not been breached since 2001. Lehman’s quarterly filings released on July 10, 2007 revealed that residential and commercial mortgage assets on Lehman’s books had increased by $21.9 bn to $79.6 bn on May 31, 2007, from $57.7 bn on November 30, 2006. Lehman also announced unrealized losses of $459 mn related to declines in the value of its mortgage and mortgage-backed security holdings for the quarter ending May 31. These negative changes were masked by a record quarterly net income of $1,273 mn. Over the course of a few days in July 2007, a large number of securitizations had their credit ratings cut by Moody’s and S&P, the largest raters of structured products. The fact that Lehman’s ratings, and the ratings of other investment banks, were not adjusted in this period implies a large failure in qualitative analysis at CRAs. The high proportion of securitization downgrades should have given the ratings agencies a clue as to the ultimate quality of Lehman’s assets, as the firm’s balance sheet dramatically increased at the same time when structured products were receiving multi-notch downgrades. Although changes in the structured finance ratings are not direct inputs into RMI’s PD model, a large decline in Lehman’s stock price over this period effectively incorporates this important information into the ultimate output of the RMI PD model. Lehman’s share price declined almost 30% from July 13 to August 15, 2007. As the ratings agencies upgraded or held firm Lehman’s credit rating, the RMI 1-year PD remained above 180 bps, well above the probability of defaults implied by CRA ratings. Lehman shut BNC Mortgage on August 23, 2007, cutting 1,200 jobs 3 years after Lehman had purchased the company. Slowing demand for subprime loans and concerns about risky lending at BNC had forced Lehman’s hand.24 The closure of BNC was not followed by any ratings movements from CRAs, although DBRS reaffirmed Lehman’s rating at A (high) with a positive outlook on August 30, 2007 after the bank closed BNC, acknowledging that the closure of BNC did not materially affect Lehman’s business. DBRS also cited Lehman’s resilient 178

1250008.indd 178

30 August 2012 9:08 AM

earnings power, despite an increasingly challenging residential mortgage environment. Lehman remained on positive watch at Moody’s and stable watch at Fitch and S&P. Lehman’s 1-year PD had increased to 291 bps on August 29, due to decreasing DTD mainly caused by an almost 10% decline in the bank’s share price following the shuttering of BNC. The commercial real estate market declined even more steeply than the residential housing market in late 2007. Despite this, Lehman made several large principal investments in the commercial real estate market, acquiring a $5.4 bn stake in Archstone Smith, a real estate investment trust in October 2007, when commercial real estate already made up 6.3% of the firm’s assets. Lehman also increased its exposure to SunCal Companies, a firm that bought land in California and sought residential development approval from local governments. If development was approved, SunCal would sell the land for a large profit to homebuilders. DBRS upgraded Lehman to AA (low) on December 21, as DBRS believed Lehman’s continued international earnings diversification and the bank’s improving liquidity profile warranted an upgrade. However, DBRS acknowledged the potential risk of Lehman’s exposures to the residential and commercial real estate sectors. Lehman’s 1-year PD remained elevated at 152 bps on December 20, 2007.

2.2. The Beginning of the End Hedge fund managers began suggesting that Lehman was significantly overvaluing the real estate investments in early 2008. Shares of ArchStone’s competitors had fallen 30%, while the homebuilders that SunCal sold residentially zoned land to had written down their current holdings by around 90%. Lehman’s valuation methodologies for these investments were actively criticized by market participants during this period, who believed Lehman’s write-downs should have been even higher. Lehman first started to experience real financing difficulties in March 2008, as many market participants viewed Lehman as the next broker-dealer most likely to fail following the demise of Bear Stearns. Persistent rumors that Bear Stearns was unable to access short-term funding from Wall Street

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:58 AM

b1423

Global Credit Review Volume 2

counterparties had fuelled a run on Bear Stearns at a pace that outstripped Bear Stearns’ liquidity resources. On March 17, J.P. Morgan Chase agreed to acquire Bear Stearns, with the Federal Reserve providing $30 bn in funding. The Fed also established the Primary Dealer Credit Facility (PDCF), a new lending facility that provided liquidity to broker-dealers such as Lehman. Lehman’s CRI 1-year PD reflected the broad perception that Lehman could be the next brokerdealer to collapse, with the bank’s 1-year PD jumping more than 70 bps to 268 bps on March 14, 2008 and a further increase to 384 bps on March 17. Moody’s changed Lehman’s outlook to stable from positive on March 18, citing the bank’s exposures to commercial and residential real estate, and increasing illiquidity across financial markets. Moody’s believed Lehman’s A1 rating was supported by the Fed’s new borrowing facility, Lehman’s strong risk management history and the bank’s robust capital structure. DBRS confirmed Lehman’s AA (low) rating and stable outlook on March 19 for similar reasons. However, market participants were concerned about Lehman’s real estate related investments and the bank’s reliance on short-term funding sources, which included $7.8 bn of commercial paper and $197 bn of repurchase agreements. S&P reaffirmed Lehman’s rating at A+ in late March 2008, citing liquidity support from the Fed. However, S&P changed Lehman’s outlook to negative, taking into account the high chance of further declines in the profitability of capital market activities and Lehman’s higher exposures to the mortgage and fixed income sectors relative to peers. Fitch affirmed Lehman at AA− on April 1, 2008, but changed the bank’s outlook to negative, due to increased earnings pressure and leverage as a result of an expansion in Lehman’s real estate inventory. In the following months, Lehman’s 1-year PD never fell below 220 bps. S&P cut Lehman’s rating to A from A+ on June 2, 2008 and retained a negative outlook for the bank, citing weakness in the investment banking sector and a higher chance of write downs on assets that had lost value since Bear Stearns’ collapse. Lehman’s 1-year PD had increased 22 bps since March 17 to 406 bps on May 30, the Friday before S&P downgraded Lehman.

30 August 2012 9:08 AM

Fitch cut Lehman’s rating to A+ on June 9, 2008, due to increased earnings volatility and higher exposures to riskier assets relative to peers. Lehman’s ineffective hedges against the former were a key reason why Fitch retained a Negative Outlook on Lehman. Moody’s changed Lehman’s outlook to negative the same day for similar reasons. Fitch and Moody’s both believed Lehman’s ratings were supported by Lehman’s conservative funding structure and efforts the bank had made to decrease leverage. Lehman had announced a $6 bn equity raising exercise the same day, along with $130 bn of asset sales in a preliminary earnings report. In the same release Lehman announced a $2.8 bn loss, the first since the bank went public. Lehman’s 1-year PD had increased to nearly 570 bps on June 6, 2008, the Friday before Fitch and Moody’s rating actions, due to decreasing DTD and firm wide liquidity. Market participants again became concerned about Lehman’s short-term funding profile in June; the negative ratings actions had increased the bank’s funding costs. In the weeks that followed, Lehman’s 1-year PD reached a high of 1740 bps on July 14, 2008, due to decreasing DTD and relative market capitalization. On the same day, rumors emerged that Lehman’s CEO Dick Fuld was considering methods of privatizing the bank. Moody’s lowered Lehman’s rating to A2 on July 17 and retained a negative outlook on the bank. Moody’s believed that increased mark-to-market losses were possible in Lehman’s residential and commercial mortgage portfolios, weighing on the firm’s ability to return to profitability in the near term. The new A2 rating was supported by Lehman’s geographic earnings diversification and sound stand-alone liquidity profile. However, Moody’s reflected upon the vulnerability of Lehman’s secured financing to overall market liquidity. Lehman’s 1-year PD remained elevated at 1026 bps on July 16, 2008. A prominent Bloomberg article released on July 28 reported that Lehman’s funding costs had almost doubled since the start of 2008. The same day, Lehman’s 1-year PD had increased to 1244 bps. In the following weeks, Lehman’s 1-year PD abated to a low of 814 bps on August 6, still significantly higher than PDs implied by credit ratings. Lehman’s PD soared to 1740 bps on August 19, amidst predictions of significant Q3 writedowns and rumors that Lehman planned GLOBAL CREDIT REVIEW VOLUME 2

1250008.indd 179

179

8/30/2012 9:35:58 AM

b1423

Global Credit Review Volume 2

to sell all or part of its lucrative asset management business.25 Reports also emerged that a capital raising deal with a strategic Asian partner had fallen through. A fortnight later, Lehman’s 1-year PD had fallen to 906 bps on September 5, 2008, as rumors abounded that Lehman planned to spin-off a majority of its commercial real estate assets into a bad bank and sell part of the remaining good bank to the Korean Development Bank (KDB).26 On September 8, 2008, the following Monday, Lehman’s PD increased 257 bps to 1163 bps after analysts increased their predictions for writedowns at Lehman, and the Korean Financial Services Commission warned of the risk involved in KDB’s potential investment in Lehman. The following day, Fitch placed Lehman on Negative Watch, citing acute capital raising costs and impaired operating profitability from securitization losses. Fitch warned that multinotch downgrades could occur in the near term. S&P also placed Lehman on CreditWatch negative on September 9, for similar reasons, yet asserted that Lehman’s current liquidity and funding profiles were adequate (see Table 2). On September 10, 2008, Lehman announced a $3.9 bn quarterly loss in the bank’s Q3 earnings call, due to $7.8 bn of mark-to-market adjustments. The report highlighted reductions in residential and commercial real estate exposures and mapped out a strategic plan to increase capital and reduce balance sheet risk. DBRS downgraded Lehman the same day, to A (high) with a negative outlook from AA (low), the last major CRA to downgrade Lehman below AA. In accompanying commentary, DBRS cited two consecutive quarterly losses and a difficult operating environment that could constrain future earnings. DBRS

Table 2.

Date

DBRS

30 August 2012 9:08 AM

believed Lehman’s A (high) rating was supported by the bank’s access to the Fed’s liquidity facilities and the prospect of Lehman divesting parts of its business. Moody’s placed Lehman’s ratings on watch with direction uncertain on September 10, 2008, citing reasons similar to DBRS. In addition, Moody’s believed a potential strategic partnership would place upwards pressure on Lehman’s credit rating. On September 9, Lehman’s 1-year PD had doubled overnight to 3109 bps, although this increase could incorporate rating watches issued by Fitch and S&P the same day, as Lehman’s share price dropped dramatically. Market participants were growing concerned that potential ratings downgrades would trigger collateral demands from Lehman’s repo counterparties and affect the bank’s ability to raise funding from money market funds. On September 11, Lehman’s PD spiked significantly by 3295 bps to 6801 bps, due to decreasing DTD and relative market capitalization. On September 12, 2008, reports emerged that the Fed and the US Treasury were attempting to engineer a sale of Lehman to a number of potential suitors, although the government was avoiding options involving public money.27 News of a potential sale led S&P to change the CreditWatch on Lehman to developing from negative. The same day, Lehman’s 1-year PD increased 285 bps to 7086 bps. After potential deals with Barclays and Bank of America fell through over the following weekend, due to a lack of financial support from the US government, Lehman filed for Chapter 11 bankruptcy on the morning of September 15, 2008.

Lehman Brothers: Credit ratings and RMI 1-year PD during the bank’s final week.

Fitch

Moody’s

S&P

Sep 8 Sep 9 Negative Watch, A+ CreditWatch Negative, A Sep 10 Downgrade, AH Uncertain Watch, A2 Sep 11 Sep 12 CreditWatch Developing, A Sep 15 Downgrade, D Downgrade, D Downgrade, B3 Downgrade, SD *Previous closes are used to avoid incorporating the market effect of new rating changes into PD data. 180

1250008.indd 180

CRI 1-year PD (bps), Previous close* 906 1163 3109 3506 6801 7086

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:58 AM

b1423

Global Credit Review Volume 2

2.3. How Lehman hid Increasing Leverage and Risk from Ratings Agencies and Market Participants: The Effect of Repo 105 The Lehman Bankruptcy Examiners released a report in March 2010 that revealed some interesting balance sheet manipulation techniques Lehman used to improve reported leverage, namely the now infamous Repo 105 trades. The name reflects the interest Lehman paid to enter into these trades; a repo 105 trade involved a repurchase agreement where Lehman received $100 in cash at the beginning of the trade and pledged $105 in assets with its counterparty. Lehman would then repay the cash plus $5 to close out the trade, effectively paying 5% interest on the trade. The repo 108 trade worked in a similar manner, with Lehman paying 8% interest on such trades, but was limited to trades involving European Equities.28 The 5% and 8% numbers were significant, as these were the minimum respective premiums Lehman had to pay on Repo 105 and 108 trades to classify them as ‘true sales,’ and thus remove the assets from the balance sheet. The ‘true sale’ classification was based on an English Law interpretation of the global master repurchase agreement by UK-based law firm Linklaters. This interpretation meant that all Repo 105 and 108 trades needed to be conducted through Lehman’s UK based international subsidiary. The cash received from Repo 105 trades was used to pay down liabilities, in effect reducing the overall size of the balance sheet and the bank’s net leverage ratio.29 Internal documents from April 2007 show that a primary reason for these balance sheet management techniques was to position Lehman for a credit upgrade. According to company documents from that period, the net leverage ratio was a key measure of the company’s performance in the eyes of the CRAs. Significant reductions in Lehman’s leverage between 2003 and 2006 had resulted in several rating upgrades. In April 2007, CFO Chris O’Meara believed that a AA-rating from all major CRAs was possible with the planned balance sheet leverage ratios. The ratings agencies were becoming even more focused on bank balance sheets as credit markets began to contract, placing less emphasis on changes in income statements. As market volatility increased in late

30 August 2012 9:08 AM

2007 and early 2008, the size of the Repo 105 transactions dramatically increased (see Table 3) and the objective of the trades also changed. More specifically, Lehman was attempting to avoid a credit downgrade at all costs, as a downgrade would lead to higher debt costs and demands that Lehman post additional collateral. Lehman’s senior management prescribed net leverage ratio improvements that would show ratings agencies that Lehman was reducing leverage, but extensive use of Repo 105 trades was required to meet such targets. A majority of the assets Lehman sold under such trades were US government and agency securities, as the company’s counterparties would not accept assets of lower quality as collateral, and Lehman’s own internal accounting rules required assets used in Repo 105 trades be liquid and easily replaceable. Despite this, the trades allowed Lehman to claim higher leverage reductions in the vital reporting period after the Bear Stearns collapse in March 2008; Repo 105 trades allowed Lehman to claim a net leverage ratio 12.9% lower than actual net leverage in the firm’s filings for the quarter ending May 31, 2008. Moreover, without knowledge of the Repo 105 trades, market participants were led to believe that Lehman was reducing its leverage through sales of the most illiquid securities. CRAs focused on Lehman’s net leverage ratio, and the bank’s Repo 105 trades allowed Lehman to hide its true leverage from the CRAs. Because of the infamy and effects of the Repo 105 and 108 trades on Lehman’s balance sheet, it is interesting to investigate the effect these trades had on the RMI 1-year PD. As shown in Figure 3, ignoring the repo trades in the PD calculation results in a higher PD during 2007, and a significant spike when the company stepped up Repo 105 usage in

Table 3. Lehman Brothers: Repo 105 and leverage. Quarter ending

Repo 105 usage

Reported net leverage

Leverage without repo 105

Nov 31, 2007 Feb 29, 2008 May 31, 2008

$38.60 bn $49.10 bn $50.38 bn

16.1 15.4 12.1

17.8 17.3 13.9

Source: Lehman Bankruptcy Examiner Report 2010. GLOBAL CREDIT REVIEW VOLUME 2

1250008.indd 181

181

8/30/2012 9:35:58 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

700

600

500

1Y PD (bps)

400

300

200

100

0

Adjusted PD

Unadjusted PD

Figure 3. Adjusting for the effect of Repo 105 results in higher PD.

early 2008, as a much higher amount of debt is included in DTD calculations. The PD included in this chart reflects the situation up until August 31, 2008. According to a 2010 WSJ piece, similar balance sheet manipulation techniques are still quite common amongst Wall Street banks, so it is important for readers to take note of this effect.30

III. ENRON’S RISE AND FALL Throughout the nineties, Enron’s stock had been on a constant rise, with the company’s stock price soaring in 1999 and 2000, recording respective year-on-year increases of 56% and 87%. By December 31, 2000 Enron’s market capitalization exceeded $60 bn, four times the company’s book value; financial markets were impressed by the conglomerate Enron had become. Enron had been able to operate a questionable business model, conceal true performance and hype its stock but eventually the company was forced to file for bankruptcy in December 2001. The firm was rated investment grade by all of the Big Three credit rating agencies until four days before it filed for Chapter 11 bankruptcy protection. The sheer size and speed of the bankruptcy caught many 182

1250008.indd 182

market participants off guard and credit rating agencies were criticized for not being quick enough to identify the problems at Enron. The rating agencies have argued that the fall of Enron was expedited by the company’s extensive and constant use of aggressive accounting methods with the intent to mislead and hide unprofitable business conditions from shareholders. Although RMI’s PD for Enron was lower than the PD implied by ratings before 2000, Enron’s RMI PD had been steadily increasing since mid-1999, signaling significant deterioration in the company’s credit profile (see Figure 4). In the last months before Enron’s bankruptcy, the firm’s 1-year PD increased from 100 bps to 9598 bps right before default, indicating a much higher probability of default than implied by external credit ratings. We start our analysis in 1997 right after Jeffrey Skilling became Enron’s President and Chief Operating Officer in December 1996. Established as an American energy and commodities firm in the 1980s, the company, through its involvement in natural gas trading activities, had by 1997 transformed its business model into that of a financial trader and market maker in natural resources. From 1991 to 1997, Enron’s RMI PD exhibited minor volatility and

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:58 AM

1250008.indd 183

Files for bankruptcy Dynegy deal collapses Restates profits for past five years

b1423

SEC begins official inquiry into Enron's accounting

100000

Reports that CFO earns over $7m running SPEs

Withdraws from residential energy market in California

CC

Sole customer of Dabhol said to cancel its contract

PD (bps)

1000

Announces Q2 profit has increased 40%

Q1 earnings increase 18% on increased sales in energy trading business

B2

Rumors of profit warnings waived by company Stock price peaks at $90.56 as positive earnings continue

100

Baa3 Baa2

Baa1

BBBB CC

10

BBB-

BBB+

GLOBAL CREDIT REVIEW VOLUME 2

1 2/28/1997

8/31/1997

Fitch

2/28/1998

Moody's

8/31/1998

2/28/1999

S&P

8/31/1999

2/29/2000

RMI 1-year PD

Figure 4. Enron corporation.

Positive

8/31/2000

2/28/2001

Negative

8/31/2001

Stable

30 August 2012 9:08 AM

BBB+

Global Credit Review Volume 2

CEO Jeffrey Skilling resigns

10000

183

8/30/2012 9:35:58 AM

b1423

Global Credit Review Volume 2

remained at a low level.31 At that point the RMI PD was significantly lower than the PD implied by CRAs.

3.1. A Change in Strategy and Aggressive Accounting Practice In April 1998, Enron changed its strategy drastically when the company withdrew from the residential energy market in California. It continued on this route with a withdrawal from oil and gas production by selling its stake in Enron Oil & Gas. During that time, the company also engaged in capital intensive investments overseas, various broadband internet start-ups, and stepped up their trading activities such as EnronOnline.32 On 14 April 1999, the company announced that its first quarter earnings had beat estimates and increased 18%, based on increased sales in its energy trading business.33 Enron’s 1-year PD in 1999 remained low, at a similar level compared to previous periods. The Big Three CRAs maintained Enron’s investment grade rating and none flagged any financial weakness nor updated their credit ratings on the changed business model and the risky positions Enron undertook. Enron continued to make positive earnings announcements for the period between 1999 and 2000. More specifically, revenues grew by $10 billion from 1998 to 1999, and then jumped by another $60 billion to $100 billion in 2000, consistently beating analysts’ expectations. On August 23, 2000, Enron’s stock price peaked at $90.56. In November 2000, rumors abounded that Enron would issue a profit warning. The company dismissed these rumors on November 24, announcing that all the firm’s businesses were performing well, and that analyst earnings estimates matched the company’s expectations.34 Enron released its annual results on 22 January 2001, with earnings exceeding previous expectations, leading to a rally in the company’s stock.35 The continuing increase in Enron’s stock price was accompanied by a 150% growth in revenues during the year 2000. On the other hand, after increasing by $600 mn in 1998, net income increased by $190 mn in 1999, and advanced just $86 mn in the year 2000. The fact that Enron’s revenues and costs of goods sold significantly increased while Enron’s net income 184

1250008.indd 184

30 August 2012 9:08 AM

fell, could have been an indication that something was wrong. More specifically the company was exploiting the limitations of accounting practices in managing its financial reports. Being the first company outside the financial services industry to use mark-to-market accounting, Enron had focused on boosting revenue growth rather than cash flow and profitability. Mark-to-market accounting allowed Enron to recognize the present value of long-term trading contracts inflows, meaning it could record gains from what over time could turn out to be losses. In addition to using the mark-to-market method for its energy contracts, Enron also adopted the “merchant” model of reporting revenues for its trading activities, especially the booming volume of trading that took place over its proprietary EnronOnline trading platform. Under the “merchant model”, Enron reported the entire value of each of its trades as revenue, while the trading firms such as Goldman Sachs only reported the trading or brokerage fees as revenue using the conventional “agent model”.36 The combined effect of using the mark-to-market model and the merchant model for revenue reporting was that Enron’s revenues and costs of goods sold were boosted, while the firm’s profit before taxes did not improve. Despite the peaking share price in August 2000, the RMI PD experienced an increasing trend from March 31, 2000, to 20 bps on December 29, 2000. This increase was mainly caused by decreasing DTD with a 50% increase in the firm’s liablities offsetting the increase in Enron’s market capitalization. In comparison to the static ratings from the CRAs, RMI CRI’s gradual increase in PD highlighted the upward trend in credit risk the company faced throughout 2000. During that period, credit rating agencies continued to overlook the misleading accounting practices Enron had adopted as the company continued to recognize revenue aggressively. While the ratings for Enron by S&P and Fitch remained stable, on March 23, 2000, Moody’s upgraded Enron’s rating from Baa2 to Baa1, in part due to the sustainability of Enron’s operating cash flow.37 Furthermore, Moody’s stated that Enron’s increased earnings were a result of the firm’s strong market position in the wholesale North American energy market and that this was the main reason for maintaining its investment grade rating.38

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:59 AM

b1423

Global Credit Review Volume 2

3.2. Special Purpose Entities Enron also relied heavily on structured finance transactions by setting up Special Purpose Entities (SPE). The Financial Accounting Standards Board (FASB) required only 3% of an SPE to be owned by an outside investor for the SPE not to be consolidated as a subsidiary. Enron took advantage of this rule, which allowed the company to increase leverage without recognizing liabilities on the balance sheet, thus artificially inflating returns on assets. Credit rating agencies have little control over the accounting standards a firm adopts and rely heavily on external parties such as Enron’s auditor, Arthur Anderson, to certify the accuracy of the reported financial statements.39

3.3. Bankruptcy Up until July 2001 Enron continued to release positive news to the markets, with the firm reporting a 35% year-on-year growth in net income for the second quarter of 2001 on July 13, 2001.40 Many external parties remained positive about the company. After a meeting between Enron’s chairman Ken Lay and vice president Dick Cheney on US energy policy in May 2001, the latter released a report with positive recommendations on Enron.41 Despite these positive news releases, Enron’s share price was starting to decline. On February 6, 2001, Enron’s PD reached 30 bps, exceeding the implied ratings from each of the Big Three CRAs for the first time. From that point until September 2001, the RMI PD continued to increase and remained at a significantly higher level than CRA ratings. The upwards trend in Enron’s PD was mainly caused by decreasing DTD induced by a decline in the company’s share price and increased leverage. The negative evolution in share price continued as negative news about the company began to spread like wildfire. In May 2001 it was announced that the sole customer of Dabhol, Enron’s giant power plant in India, was planning to cancel its contract. By August 2001 Enron’s stock price had declined 53% from a 52-week high of $90.56 in August 2000 to $42.93 on August 14. On the same day Enron’s CEO Jeffrey Skilling resigned and the founder of the

30 August 2012 9:08 AM

company Ken Lay was forced to take control. Enron’s PD increased 23 bps the following day to 240 bps, and it continued to increase, reaching 396 bps on August 31. The following working day, the company’s PD spiked 50 bps following a large increase in Enron’s idiosyncratic risk. Despite a large decline in the company’s share price and Jeffery Skilling’s resignation, the rating agencies reiterated the fundamental strength of Enron’s energy and trading business due to the firm’s longstanding operations. S&P confirmed Enron’s investment grade ratings on October 16, 2001 and acknowledged that the company’s strong operating cash flows would be sufficient to sustain short term debt and near term capital expenditure. In addition, $2.35bn in proceeds from the recent divestiture of subsidiary Portland General Electric was expected to improve Enron’s liquidity profile. Enron’s PD remained heightened at 396 bps on October 15, down from highs a fortnight before, due to a recovery in the firm’s DTD and a decrease in idiosyncratic risk. Moody’s and Fitch had made no moves to change their ratings for Enron up to this point, although Moody’s placed the firm on negative watch on October 16. On October 16, 2001, Enron’s management reported recurring earnings in Q3 2001 of $0.43 per diluted shares and stressed that earnings would improve, forecasting “recurring earnings” of $2.15 per share for the following year. However, at the same time the company announced a non-recurring net loss of $628 mn or $0.84 per diluted share due to write-downs on various investments, casting doubt on the management’s recurring earnings estimate.42 On October 19, 2011, when it was revealed that Enron’s CFO, Andrew Fastow had earned over $7 mn by running the LJM partnerships, some of the SPEs Enron set up for hedging purposes, Wall Street started questioning Enron’s past activities even more.43 On October 22, 2001, Enron announced that the US Securities and Exchange Commission (SEC) had started an official inquiry into the firm’s accounting methods. Upon the announcement, Enron’ stock price plunged over 20% and the company’s 1-year PD increased sharply by over 300 bps to 842 bps. The increase in RMI PD was primarily induced by a 40% drop in the company’s stock price from October 15 to GLOBAL CREDIT REVIEW VOLUME 2

1250008.indd 185

185

8/30/2012 9:35:59 AM

b1423

Global Credit Review Volume 2

30 August 2012 9:08 AM

Table 4. Enron: Credit ratings and RMI 1-year PD during final months. Date

Fitch

Moody’s

S&P

Oct 22 Oct 25 Negative Watch, BBB Negative Outlook, BBB+ Oct 29 Downgrade, Baa2 Nov 1 Downgrade, BBB Nov 5 Downgrade, BBNov 9 Downgrade, Baa3 Downgrade, BBB− Nov 28 Downgrade, CC Downgrade, B2 Downgrade, B− Nov 29 Nov 30 Downgrade, CC Dec 3 Downgrade, D Downgrade, Ca Downgrade, D *Previous closes are used to avoid incorporating new rating changes into PD data.

October 23, negatively affecting various PD inputs such as DTD, idiosyncratic risk and relative size. On October 24, 2001, Enron’s Chief Financial Officer Andrew Fastow was fired. The firm’s stock price had dropped to $15.41 by October 24, 2001. Enron’s PD increased 391 bps overnight to 1305 bps on the same day, mainly driven by a large increase in idiosyncratic volatility. On October 31, 2001 the SEC began a formal investigation into Enron’s financial reporting practices. One day later, for the first time since 1995, Standard & Poor’s revised its rating downward to BBB from the BBB+ rating, while Moody’s downgraded Enron to Baa2 from Baa1 on October 29. Both the CRAs stated that the substantial loss of investor confidence and significant debt maturities in the short term were the primary reasons for the downgrades, which occurred just one month before Enron filed for a Chapter 11 bankruptcy protection. The firm’s PD had increased 1200 bps in the space of a week, to 2503 bps on November 1. Though uncertain if Enron was able to keep its investment grade rating, the CRAs stated that the firm’s wholesale trading business and its regulated pipelines were able to generate excess cash flow. On November 8, 2001, Enron announced that it had overstated profits by $586 mn over the past five years and disclosed that it was attempting to restructure a $690 mn debt obligation that matured on November 27. On the same day Enron’s rival Dynegy announced plans to buy Enron for about $9 billion in stock.44 The firm’s PD had again increased by over 1200 bps in one week, to 3754 bps on November 8. The following day, Moody’s and S&P each downgraded Enron one notch 186

1250008.indd 186

RMI 1-year PD (bps), previous close* 532 bps 1305 bps 1476 bps 1981 bps 2689 bps 3754 bps 5933 bps 7564 bps 7919 bps 9598 bps

to Baa3 and BBB− respectively. Rumors abounded that investment bankers had lobbied the CRAs to avoid downgrading Enron below investment grade, which would scuttle any potential acquisition by Dynegy. On November 20, 2011, Enron announced that its Q4 earnings would be negatively affected by the events of the past months. Enron’s PD had continued to increase throughout November, rising 1850 bps in a fortnight to 5623 bps on November 21. Up to that point, S&P had maintained that it was confident Enron would be able to successfully negotiate a restructuring and extend the maturity of current obligations. However, a slew of downgrades followed in the last week of November. Moody’s downgraded Enron to B2 from Baa3 on November 28, while Fitch downgraded the firm to CC from BBB− the same day. S&P also downgraded Enron to B− from BBB− on November 28. On November 29, 2001, it became clear that Enron had failed to salvage the deal with Dynegy. The same day, Enron’s new CFO Jeff McMahon said he preferred to fix Enron without a bankruptcy filing, but he did not rule out the possibility.45 S&P again downgraded the firm to CC from B− on November 30. The CRAs cited common reasons for the downgrades in Enron’s final week, including limited access to capital markets and a poor liquidity outlook. The RMI 1-year PD for Enron had increased by over 7000 bps in November, to 9598 bps on November 30 as the DTD variable fell significantly following a major sell-off of the firm’s stocks. Following the announcement on December 2, 2001 that Enron had filed for Chapter 11 bankruptcy, Moody’s revised its ratings to Ca. Both S&P and Fitch downgraded the company to D.

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:59 AM

b1423

Global Credit Review Volume 2

DISCLAIMER The graphs and tables included with this document are provided ‘as is,’ and are produced and provided strictly for comparative research purposes. No part of this document should be relied upon as legal advice or for legislative purposes. CRI acknowledges the fact that ratings include a high degree of qualitative opinion, making a quantitative analysis of credit ratings difficult, and that other comparative studies could produce diverging results.

NOTES 1

2

3

4

5

6

7

8

9

10

All CRA ratings data used in the application of this methodology were downloaded from ESMA’s Central Ratings Repository (CEREP), for the period from January 2000 to December 2010; for the CRI methodology we refer to the Technical Report published in this Volume of the GCR. It should be noted that the PDs included in this analysis are in-sample PDs and based on the calibration data of April 2012. The CRI PD is a prediction horizon specific number, for example, one year. As a point-in-time credit risk measure, the current market condition will have a greater impact on the shorter-horizon PD. When the prediction horizon is lengthened substantially, the impact of the current market condition will naturally diminish. The point-in-time PD in effect becomes a through-the-cycle risk measure when the horizon is reasonably long. Rating Action: Moody’s assigns long-term issuer rating of A3 to MF Global Ltd, Moody’s, July 24, 2007. The CRI 1-year PD for the Man Group remained below 70 bps, significantly more favourable than the PD assigned to MF Global. Rating Action: Moody’s downgrades MF Global; continues to review, Moody’s, February 28, 2008. Rating Action: Moody’s confirms MF Global’s Baa1 rating; assigns negative outlook, Moody’s, June 18, 2008. CRI’s probability of default model requires 1 year of trading and financial statement data to compute a firm’s PD. Rating Action: Moody’s downgrades MF Global to Baa2, changes outlook to stable, Moody’s, January 16, 2009. Rating Action: Moody’s changes MF Global’s (issuer rating at Baa2) outlook to negative, Moody’s, November 6, 2009.

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26 27

28

30 August 2012 9:08 AM

Announcement: Moody’s comments on MF Global’s CEO change, Moody’s, March 23, 2010. Fitch: No effect on rtgs of MF global on announced strategic initiatives; remains on Watch Negative, Fitch Ratings, May 20, 2010. Fitch affirms MF Global Holdings Ltd.’s ratings; Outlook Negative, Fitch Ratings, February 3, 2011. Corzine rebuffed internal warnings on risks, The Wall Street Journal, December 6, 2011. A bond offering with a twist, Bloomberg Businessweek, August 4, 2011. Congress presses rating firms, The Wall Street Journal, January 5, 2012. Rating Action: Moody’s downgrades MF Global to Baa3; reviews for further downgrade, Moody’s, October 24, 2011. Update 1 — S&P may cut MF Global rating to junk, Reuters, October 26, 2011. Rating Action: Moody’s downgrades MF Global to Ba2; reviews for further downgrade, Moody’s, October 27, 2011. Fitch downgrades MF Global’s LT IDR to ‘BB+’, ST IDR to ‘B’; Rtg Watch Negative, Fitch Ratings, October 27, 2011. Regulators Investigating MF Global for missing money, The New York Times, October 31, 2011. Duan, J.-C., J. Sun, and T. Wang (2012), Multiperiod Corporate Default Prediction — A Forward Intensity Approach. Journal of Econometrics, forthcoming (DOI: 10.1016/j.jeconom.2012.05.002). Between 2003 and 2004, Lehman acquired a number of mortgage lenders, including subprime lenders BNC Mortgage and Aurora Loan Services, which specialized in making Alt-A loans, or loans to borrowers without full documentation. These acquisitions allowed Lehman to dramatically expand its securitization business and vertically integrate the company’s mortgage securitization business. The New York Times, October 31, 2011. Lehman Brothers shuts down subprime unit, fires 1,200 (Update 4), Bloomberg, August 22, 2007. Lehman shares slide on fears over results, Financial Times, August 20, 2008. Lehman may split off weak holdings, The New York Times U.S. helps Lehman go up for sale, The Washington Post, September 12, 2008. In this section, Repo 105 refers to both Repo 105 and Repo 108 transactions. Lehman used the same accounting treatment for both transactions. GLOBAL CREDIT REVIEW VOLUME 2

1250008.indd 187

187

8/30/2012 9:35:59 AM

b1423

29

30

31

32 33

34

35

188

1250008.indd 188

Global Credit Review Volume 2

Lehman reported a net leverage ratio in its financial statements, as the bank believed it gave a more meaningful calculation of leverage than simple accounting leverage, by excluding low-risk, non-inventory assets. Lehman’s net leverage ratio was defined as net assets divided by tangible equity. Lehman defined net assets as total assets minus cash and securities segregated or deposited for regulatory or other purposes, securities received as collateral, securities purchased under agreements to resell, securities borrowed and identifiable intangible assets and goodwill. Tangible equity included stockholder’s equity and junior subordinated notes, but excluded identifiable assets and goodwill. Big banks mask risk levels, The Wall Street Journal, April 9, 2010. The RMI PDs are based on the originally filed financial statements. Enron: the rise and fall, Loren Fox, September, 2002. Enron’s 1st-Qrt Profit rises 18% on trading growth, Bloomberg, April 13, 2000. Enron president states rumors untrue, Bloomberg, November 24, 2000. Enron’s 4th-Qtr profit rises 34% on commodity sales, Bloomberg, January 22, 2001.

36

37

38

39 40

41

42

43

44

45

30 August 2012 9:08 AM

Dharan, B and W. Bufkins (2004). Red Flags in Enron’s Reporting of Revenues and Key Financial Measures, in Rapoport and Dharan (editors), Enron: Corporate Fiascos And Their Implications, Foundation Press. Press Release: Moody’s worldwide rating actions, moody’s, March 23, 2000. Rating Action: Moody’s places all Enron corp long term debt ratings on review for upgrade, Moody’s, February 3, 2000. Uniform System of Accounting, Oxbridge Writers Enron reports second quarter earnings of $0.45 per diluted, Bloomberg, July 12, 2001. Dunham, Lay, Raymond among few to met with Cheney on Energy, Bloomberg, May 16, 2001. Enron announces third quarter losses, The Washington Post, October 16, 2001. Enron CFO’s partnership had millions in profit, Wall Street Journal, October 19, 2001. Rival to buy Enron, top energy trader, after financial fall, New York Times, November 10, 2001. Enron faces bankruptcy as Dynegy abandons purchase, Bloomberg, November 28, 2001.

A LEAD-LAG INVESTIGATION OF RMI PD AND CRA RATINGS

8/30/2012 9:35:59 AM