Comment on “A New Capital Adequacy Framework”
Description
Comment on “A New CapitalAdequacy Framework”KMV CorporationCopyrightCOPYRIGHT 2000, KMV LLC, SAN FRANCISCO, CALIFORNIA, USA. All rights reserved.No part of this publication may be reproduced, stored in or introduced into a retrieval system, ortransmitted, in any form or by any means now known or hereafter developed (electronic,mechanical, photocopying, recording or otherwise) without prior written permission of thecopyright owner.KMV, LLC retains all trade secret, copyright and other proprietary rights in this document.Except for individual use, this document should not be copied without the express writtenpermission of the owner.Trademark Information®KMV and the KMV Logo are registered trademarks of KMV LLC. EDFCalc and Private Firm®Model are registered trademarks of KMV LLC. Portfolio Manager™, Credit Monitor™, GlobalCorrelation Model™, GCorr™, EDF Calculator, Expected Default Frequency™ and EDF™ aretrademarks of KMV LLC.All other trademarks are the property of their respective owners.Document Number: 999-0000-094. Revision 1.0.0.Published by: Author(s):KMV LLC Stephen Kealhofer1620 Montgomery Street, Suite 140San Francisco, CA 94111 U.S.A.Phone: +1 415-296-9669FAX: +1 415-296-9458Web: http: // www.kmv.comPage ii Release Date: 30-March-2000Comment on “A New Capital Adequacy Framework”Introductory CommentsThe Basel Committee on Banking Supervision has proposed a significant revision (Revision) tothe 1988 Accord. There are many large ...
Informations
| Publié par | Thioj |
| Nombre de lectures | 24 |
| Langue | English |
Extrait
Comment on A New Capital Adequacy Framework
KMV Corporation
Copyright COPYRIGHT 2000, KMV LLC, SAN FRANCISCO, CALIFORNIA, USA. All rights reserved. No part of this publication may be reproduced, stored in or introduced into a retrieval system, or transmitted, in any form or by any means now known or hereafter developed (electronic, mechanical, photocopying, recording or otherwise) without prior written permission of the copyright owner.
KMV, LLC retains all trade secret, copyright and other proprietary rights in this document. Except for individual use, this document should not be copied without the express written permission of the owner.
Trademark Information KMV and the KMV Logo are registered trademarks of KMV LLC. EDFCalc ® and Private Firm Model ® are registered trademarks of KMV LLC. Portfolio Manager, Credit Monitor, Global Correlation Model, GCorr, EDF Calculator, Expected Default Frequency and EDF are trademarks of KMV LLC.
All other trademarks are the property of their respective owners.
Document Number: 999-0000-094. Revision 1.0.0.
Published by: KMV LLC 1620 Montgomery Street, Suite 140 San Francisco, CA 94111 U.S.A. Phone: +1 415-296-9669 FAX: +1 415-296-9458 Web: http: // www.kmv.com
Page ii
Author(s): Stephen Kealhofer
Release Date: 30-March-2000
Introductory Comments
Comment on A New Capital Adequacy Framework
The Basel Committee on Banking Supervision has proposed a significant revision ( Revision ) to the 1988 Accord. There are many large questions surrounding the Accord and the Revision. The purpose of this comment is, however, directed to a particular aspect of the Revision: the use of external ratings such as S&P or Moodys debt ratings as a standard for determining credit quality. The proposed use of external ratings is never specifically justified in the Revision. Based upon a close reading of the document, the justification would appear to rest upon the following perspectives: • A major shortcoming of the current Accord [is] that inadequate recognition is given to the differing credit quality of claims 1 ; and • The lack of homogeneity among the rating systems at different banks, together with the central role of subjective risk factors and business judgments in assigning internal grades 2 . Taken together, these suggest the desirability of having a single standard that could be relied upon to distinguish credit quality, and the implicit conclusion that agency debt ratings are as close to being such a standard as practically and feasibly exists today. The viewpoint of this comment is the following. • It is a regulatory wrong turn to move towards agency debt ratings and away from a market-based approach to assessing default risk. Market prices aggregate the fundamental and necessarily subjective assessments of individual market participants to produce a consensus view that cannot be readily arbitraged. Importantly, market prices represent the terms on which risk can be bought, sold and hedged; i.e. they give the appropriate tradeoffs. Theoretically and practically, credit risk is no different than the so-called market risks in this regard. • Agency debt ratings of the type envisioned to be acceptable (e.g. Fitch/IBCA, Standard & Poors, Moodys) are inherently subjective and do not reflect market prices. Agency debt ratings, because they are subjective, cannot be used as a meaningful benchmark to evaluate alternative measures, such as a banks internal ratings. • Although credit markets are not as complete or as liquid as other markets, there is broader and better quality information coming from markets today than is available from ratings. A direct implication is that approaches that rely upon agency ratings to determine risk can be readily arbitraged in existing markets.
1 Annex 2.B.5.18, p. 30. 2 Annex 2.C.1.45, p. 39.
Page 1
Release Date: 30-March-2000
KMV Corporation
In summary, while the motives behind using agency ratings as a basis for assessing default risk are laudable, the reality of a standard based upon such ratings would be to move away from a more market disciplined approach to credit risk management. There is already better and broader information available from market prices than that available from agency debt ratings. These points are presented in greater depth below.
Overview
Default Probabilities and Capital Determination The process for determining economic capital has three steps. The first is to measure the loss risk associated with a single given facility. The second is to use correlation measures together with the facility loss measures to determine an overall portfolio distribution of possible losses and associated probabilities. The third is to use the distribution to find the amount of capital that would be sufficient to cover possible losses to a desired level of confidence. Although there are many technical details in making these measurements, one common element is that the loss risks for individual facilities, and for the portfolio as a whole, depend critically on the default probabilities associated with individual facilities. There are three general methods of obtaining default probabilities: judgmental approaches, statistical models, and market-implied probabilities. Conventional credit analysis, such as underlies agency ratings, is judgmental. It cannot be explicitly defined; it combines judgment, experience and theory in some undefined manner. Statistical methods, also called scoring models 3 , are based upon exploiting the correlation between ex ante measured firm characteristics and subsequent default. The advantage of these models is that they are objective and quantifiable; however, their disadvantage is that they do not measure causation but rather association. Despite some of the risks attendant upon a non-causative approach, mainly susceptibility to adverse selection, these methods can be very useful and are the only practical approaches available for certain asset classes. Market-implicit default probability models determine the default probability that is implied by the price of a specific liability of the borrower. For instance, at KMV we determine EDFs as the default probability of a firm implied by the value and volatility of its stock price. Default probabilities can also be determined from the market yield on a firms traded debt by isolating the component of the yield due to credit risk, and adjusting for the risk premium and expected loss given default. Because market values represent the current price of hedging risk, these approaches are accurate and timely, and generally not subject to adverse selection.
3 See, for example, Altman, Edward I. [1968]:"Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy ," Journal of Finance, September 1968.
Page 2 Release Date: 30-March-2000
Comment on A New Capital Adequacy Framework
Capital Regulation and Capital Arbitrage In determining the adequacy of capital, it is key to determine the adequacy of the default probabilities used in the banks risk measurement model. The Revision suggests that this can be done by using agency debt ratings such as Moodys, implicitly calibrating them to default rates. As will be argued below, the main difficulty is that the use of ratings in this fashion will lead to extensive capital arbitrage, because of the differences between default probabilities implied by ratings and default probabilities implicit in market prices. These arbitrages exist today and can be documented; they are not theoretical and they are not incidental. The world of fixed income management and structured finance vehicles provide plentiful examples. It is not the purpose of this paper to argue against banks being able to use their internal ratings to determine capital adequacy. In our view, banks should be free to determine the methods that they use. The issue is how the regulator will decide if the ratings used are satisfactory. Implicitly or explicitly, the Revision appears to suggest that agency debt ratings are satisfactory for that purpose. Based upon our empirical research over the last decade, agency debt ratings as well as banks internal ratings convey significant amounts of information about default risk. Statistical models also can perform quite satisfactorily in predicting future events of default. However, both approaches are subject to a common failing: where market price information exists, both approaches by construction will differ significantly from the assessment of the market. The result is that these methods will produce capital arbitrage. Only information based upon market prices can provide a non-arbitrageable benchmark for measuring risk. The Basel Accord is intended to be a work in progress. It is designed to encourage banks to move steadily in the direction of managing their own risks in an appropriate and ultimately transparent way. The danger is that instead of leading toward this goal, the regulation creates artificial constraints and costs on banks. In order to compete with non-banks, portfolio management gets distorted towards reducing regulatory capital rather than actually managing risks, or moves to shift assets out of the banking sector altogether. Unfortunately, we see both of these trends at present. The way to avoid these dead-end side paths is to recognize that risk management machinery must be grounded in market prices. This point can be illustrated by a simple example. Suppose that in examining a bank, it is found that a bank is requiring a particular amount of capital against a loan because it has determined the rating to be A. The concern of the regulator should be, not that the banks internal rating match some given external agency rating, but rather that the loans risk should actually be consistent with the rating. An objective way to make this determination is to see if the borrower has a bond outstanding that is trading at a price consistent with the default risk of an A-rated borrower. An alternative is to see if the value and volatility of the borrowers stock price is consistent with the default risk of a A-rated borrower. If not, the regulator should be able to require the bank to change its ratings appropriately. If the regulator cannot do this, then the bank will have an opening to arbitrage the rules.
Page 3
Release Date: 30-March-2000
KMV Corporation
Although these market-implicit default probabilities do not exist for all borrowers, they exist for the vast majority of the borrowers that have agency ratings because those are the firms with traded debt. Moreover, they are available for an order of magnitude more borrowers that do not have agency ratings, but that do have traded equity. Much of the total capital usage in the banking system is due to such borrowers. The rapid expansion of whole loan, credit derivative and corporate bond markets, coupled with better price disclosure, should lead to greater coverage in the near future. Two final observations: market-implicit default probabilities are consonant with the existence of real opportunities to hedge risk. One does not have to believe that the market price is right for the market-implicit probability to be the correct measure of risk. All that is necessary is that one can hedge the risk at that price. Might agency debt ratings simply be surrogates for market-implicit prices? Empirical research shows that there are major differences between default rates implied from ratings, and default rates implied from market prices.
1. Market Prices and Risk Assessment Market prices aggregate the fundamental and necessarily subjective assessments of individual market participants to produce an objective and consensus view that cannot be readily arbitraged. Importantly, market prices represent the terms on which risk can be bought, sold and hedged; i.e. they give the appropriate tradeoffs. Theoretically and practically, credit risk is no different than the so-called market risks in this regard. There are two important connections between risk and market values. First, realizations of risk are variations in market values. If we are talking about risky outcomes, we are talking about variations in future market values. All other outcome measures income, default, and so forth are simply imperfect short hands for market value. If there were default, but no variation in the market value of the instrument, then there would not be risk. The converse does not hold. The reason for this simple identity is that market prices give the terms on which we can trade. If we cannot trade on those terms, then they are not market values. If a bond defaulted but still traded at par, we could sell it and get out without loss. If a firms income went down, but its market value did not, we could sell it and get out without loss. The second connection between risk and market value is more complex. In trying to understand the range and likelihood of future market values, we can often use current market values. This connection requires some development. In the absence of a market price, the issue in evaluating an opportunity is the cost of creating the opportunity versus the value of the payoff. Investors in the risky outcome need to reach an assessment of the risk of the payoff. They do this via what one might call judgmental analysis. When there is a market where participations in the opportunity can be traded, the price will reflect the assessment of the participants in the market in some way that balances supplies and demands. This price will always reflect a probabilistic assessment of the outcomes in some general sense.
Page 4
Release Date: 30-March-2000
Comment on A New Capital Adequacy Framework
A further important point is that actual prices in many markets evolve in such a way as to approximately reflect the true risk of the opportunity per unit of time. In other words, if we look at the variation in a market value over a certain interval of time, we can project that variation as a guide to variation in future intervals of time. There is no necessary reason for this to occur, but it often does. Both the probabilistic nature of prices and the information in price variability can be exploited when assessing future risk. Rather than having to perform a judgmental, and thus inherently subjective, assessment of future risks, we can deduce the consensus information of market participants from prices. For instance, in studying interest rate risk we can look at the history of interest rates to draw objective inferences. This contrasts with doing macro-economic analysis and guessing what the central bank will do as regards monetary policy, i.e. judgmental analysis. In fact, there is a term of art, market risk , which carries the implication of being able to assess risk directly from prices. Index funds in stock markets exemplify the information trade-off that prices create between judgmental and market risk assessment. In looking to get information about risk, we must either turn to judgmental assessments, or to statistical approaches, or to market risk assessment. There are no other sources of information. Market prices, when they are available, provide the consensus information of market participants. Any single set of judgments, such as an agencys debt ratings, is generally inferior to the information in market prices. If it were not, then there would exist an opportunity to make an abnormal profit in the market using those judgments, and abnormal profit opportunities are not persistent. Capital rules based upon agency debt ratings can always be arbitraged because of inconsistencies between ratings and prices. The information contained in prices cannot be arbitraged because prices, by definition, give the terms on which the risk can be traded.
2. Agency Debt Ratings Are Not Market Assessments Agency debt ratings of the type envisioned to be acceptable (e.g. Fitch/IBCA, Standard & Poors, Moodys) are inherently subjective and do not reflect market prices. Agency debt ratings, because they are subjective, cannot be used as a meaningful benchmark to evaluate alternative measures, such as a banks internal ratings. Standard & Poors describes their debt ratings as follows: As part of the media, [S&P] simply has a right to express its opinions in the form of letter symbols. Recognition as a rating agency relies on investors willingness to accept its judgment. 4 In a more detailed description of its analysis, S&P states that within a given rating category, expected levels of financial ratios vary with the business or operating risk of the company. The business risk assessment is based on S&Ps qualitative evaluation. Evaluation of a companys industry environment and competitive position is, by nature, subjective. 5
4 S&P Corporate Finance Criteria, p. 3. 5 Ibid, p. 63.
Page 5
Release Date: 30-March-2000
KMV Corporation
Further, the rating agencies rely upon making projections that are also necessarily subjective. Quoting again from S&P, S&Ps ratings are not based on the issuers financial projections of what the future may hold. Rather, ratings are based on S&Ps own assessment of the firms outlook. 6 Ratio standards relate to projected financial condition, as ratings are an assessment of a companys ability to meet future obligations. S&Ps confidence regarding a companys ability to achieve its plans has a direct bearing on the expected level of ratios. 7 S&P has been quoted here because it has historically made the best efforts to articulate its rating process. Similar statements could be found, however, in the descriptive material of any of the rating agencies. In general, the logic of their business franchises rests upon pursuing an analysis that is in some measure independent of the information contained in prices. As will be documented below, the differences between measures based upon market prices and agency debt ratings are highly significant.
3. Current Market Risk Assessments of Credit Risk Dominate Agency Ratings Although credit markets are not as complete or as liquid as other markets, there is broader and better quality information coming from markets today than is available from ratings. A direct implication is that approaches that rely upon agency ratings to determine risk can be readily arbitraged in existing markets. KMV provides market-implicit default probabilities called EDFs that derive their information from equity prices. These data are used by a significant number of the large, international banks targeted by the Accord. It is our best estimate that there are approximately 35,000 firms globally that have traded equity, some 21,000 of which have commercially available EDFs. Of these 21,000 firms, only approximately 2,700 have agency debt ratings by our count. In addition to equity information that has much broader coverage than debt ratings, it is almost exactly those firms that have debt ratings that also have publicly traded debt instruments. Their prices provide additional sources of information for determining default probabilities. In sum, the set of corporate borrowers with available market-implicit default probabilities includes almost all firms with agency debt ratings and seven times that number that do not have agency debt ratings. It includes corporate borrowers whose borrowings represent probably 50% of total corporate exposure in the banking system, and a very high fraction of the larger exposures. Regardless of the logic and theory, can it be established that market-implicit default probabilities are good empirical predictors of default? For this purpose, we will use the KMV EDFs as an example of market-implicit default probabilities. The purpose is not to tout one particular commercial product, but rather to illustrate that agency debt ratings are substantially
6 Ibid, p. 9. 7 Ibid, p. 9.
Page 6
Release Date: 30-March-2000
Comment on A New Capital Adequacy Framework out of sync with market prices in ways that significantly undermine the usefulness of agency debt ratings as a basis for capital rules. Chart 1 shows power curves from a study based upon all identified defaults of non-financial companies with public debt ratings between 1978 and 1990. Power curves are a tool for comparing the error rates of default predictions on a comparable basis. 8 The EDFs significantly outperform agency debt ratings. For instance, using cutoff values for each measure that result in equal type 2 errors of 20%, EDFs successfully identify 72% of the defaults versus 61% for agency ratings. This result is particularly evident in the lower quality firms, because most of the defaults occur amongst those firms. Chart 1: S&P Rating vs. EDF (1978-1990) [EDF represented in red; S&P in green]
100
90
80
70
60 50 40
30
20
10
0 10 20 30 40 50 60 70 80 90 100 Percentage of Companies Excluded
EDF S&P
Chart 2 shows power curves for agency ratings and EDFs based upon a recent study that uses all publicly rated defaults from 1991 to 1999. Unlike the previous study, it includes both
8 See Kealhofer, Stephen, The Quantification of Credit Risk, January, 2000, KMV Corporation, unpublished, for a more detailed description of power curves.
Page 7 Release Date: 30-March-2000
KMV Corporation financial and non-financial firms. Again, the EDFs do a significantly better job than agency ratings in predicting default 9 . Chart 2: KMV EDF vs. Debt Ratings (1990-1999) Six Months Before Default 100
80
60
40
20
0
5 10 15 20 25 30 35 40 45 50 55 60 100 Percentage of Companies Excluded
EDF S&P
9 The performance of EDFs and agency ratings is insignificantly different for very high quality firms. The crossovers in the power curves are caused by small differences in ordering for a few default events. Default power tests, because of their inherent focus on default, do not provide much evidence about differences in ordering for firms with very low average default rates.
Page 8 Release Date: 30-March-2000
Comment on A New Capital Adequacy Framework
To understand these results pragmatically, it is useful to look at a specific example. Chart 3 shows the EDF and the agency debt rating for a recently defaulted firm, Harnischfeger Industries. Harnischfeger defaulted in June 1999, but it had been rated investment grade as late as March 1999. The rating is shown as the stepped line, with time on the horizontal axis, and default probability on the vertical axis 10 . However, the EDF for Harnischfeger indicated that it was no longer investment grade as of October 1997, and in fact, by October 1998, there was an enormous disparity between the agency rating and the EDF. The rating was BBB-; the EDF was more typical of companies then rated B- by the agencies.
Chart 3: EDF and Debt Rating Harnischfeger Industries Inc
If one looks at companies with a given letter rating, one can almost invariably find a significant group of companies with EDFs that are both significantly higher and lower than what is typical for that rating. These are usually companies whose credit quality has changed significantly over the previous year to two years, but whose rating has not changed. For instance, returning to Harnischfeger, its EDF was approximately consistent with its rating as late as mid 1997.
10 For purposes of graphing, the agency debt ratings are depicted at the level of default risk consistent with the median EDF of all firms with the given rating over the last five years.
Page 9 Release Date: 30-March-2000
© 2010-2024 YouScribe