CAPITAL REQUIREMENTS Estimating Probability of Default via External Data Sources: A Step Toward Basel II Banks considering their strategies for compliance with the Basel II Capital Accord will likely use external data sources for the estimation of probability of default as part of the solution. Definitions and characteristics must be consistent or adjusted to be so if the external data is to be an effective and statistically useful supplement to internal data. Banks that do not have the most sophisticated by Nadeem A. Siddiqi and Shelley B. Klein External data will likely be necessary for most banks to achieve the foundation or advanced IRB approach. Although the decision to use the data may be relatively simple, effective use of it is more complicated and can be critical to success. Careful analysis of the definitions and characteristics implicit in the external data is essential, and banks that do this will likely be rewarded with more appropriate risk-rating results and an easier time defending their processes to the regulators. 68 The RMA Journal December 2002 - January 2003 methods of calculating risk capital as required by Basel II, can, and are, attempting to employ the foundation internal ratings-based approach proposed in Basel II. To calculate regulatory capital, banks must determine the probabilities of default associated with their portfolios, and then apply regulator-determined loss given default and exposure at default rates. One seemingly simple approach banks can take is to use default probability rates available from major external ratings agencies such as Standard and Poor s (S&P) and Moody s Financial Services. However, banks need to be careful when using this data as proxies for specific loan portfolios. 1. They need to understand the method of calculating default to ensure that it matches both the bank s philosophy and processes. 2002 by RMA. Siddiqi and Klein are senior consultant and director, respectively, with the Credit Analytics and Data Management Group of the Financial Services business unit of BearingPoint, Inc. The authors thank Charles Hill and Savitha Sagar for valuable research assistance. All information provided is of a general nature and is not intended to address the circumstances of any particular individual or entity. Although we endeavor to provide accurate and timely information, there can be no guarantee that such information is accurate as of the date it is received or that it will continue to be accurate in the future. No one should act upon such information without appropriate professional advice after a thorough examination of the particular situation. The views and opinions are those of the authors and do not necessarily represent the views and opinions of BearingPoint, Inc.
2. They must be careful in matching the characteristics and structures of the loans with those of the underlying bonds from the published default studies, especially for loans below investment grade, which may require specific adjustments. 3. Some loan grades have little published default data available, and banks may need to extrapolate their default rates from available data. Capitalizing on continuous advances in technology and analytical methods, Basel II is prompting banks to learn more about quantitative credit risk measurement. Many banks response will include using external data to calculate risk capital. Basel II offers three different ways of determining risk capital: 1. The standardized approach, which applies industry-based averages to different asset classes and is most suitable for very small banks that have a relatively homogenous asset base and do not have the resources to meet regulatory requirements to internally calculate risk. 2. The foundation internal-ratings-based (IRB) approach, whereby banks need to internally estimate part of the risk calculation formula, with the rest being determined by (conservative) regulator-determined rules. This approach may be suitable for small and medium-sized banks looking to get a foothold in the advanced approach, without having the resources to commit to it fully. 3. The advanced IRB approach allows banks to fully estimate the risk and regulatory capital requirements internally. The obvious advantage of using the advanced IRB approach is that if a well-managed bank calculates its capital requirements for itself, and its risk profile requires less capital than the conservative regulator-determined charges, more capital will be freed for other use, thus improving overall profitability. Recent Default Activity The poor global and domestic economic environment that prevailed in 2001 resulted in record defaults. During that year, S&P observed 216 defaults on approxi- Table 1 mately $116 billion of debt. These defaults represented a record 4.09% of all rated issuers at the beginning of the year. This default ratio surpassed the previous record of 4.01%, set in 1991. 1 In addition to a generally weak global economy, an abundance of recently issued speculative-grade debt, particularly in the leisure time/media and telecom sectors, contributed to the very high default rate. To reflect the changing economic environment and the associated increases in defaults, yield spreads on highly speculative debt widened to a point where costs associated with issuing debt at such levels were prohibitive to most prospective issuers. To lower their cost of capital, many issuers have been forced Corporate Defaults Rates All Ratings, 1982 through 2001 Year Default Rate (%) Number of Total Debt Defaulting Defaults ($ Billions) 1982 1.26 18 0.9 1983 0.68 10 0.4 1984 0.83 13 0.4 1985 1.09 18 0.3 1986 1.69 32 0.5 1987 0.93 19 1.6 1988 1.49 32 3.3 1989 1.75 39 7.3 1990 2.86 64 21.2 1991 4.01 88 23.6 1992 1.35 31 5.4 1993 0.86 22 2.4 1994 0.62 18 2.3 1995 0.97 32 9.0 1996 0.56 20 2.7 1997 0.59 23 4.9 1998 1.26 56 11.3 1999 2.19 108 37.8 2000 2.56 132 42.3 2001 4.09 216 116.1 Source: Standard and Poor's, Special Report - Record Defaults in 2001 the Result of Poor Credit Quality and a Weak Economy. 69
to use alternative debt structures. In some cases, issuers provide collateral when issuing debt, or issuing securitizations, to receive higher credit ratings. A number of issuers have also issued debt with structures, including interest reserves, which ensure interest payments to investors for a number of years after issuance. Table 1 provides historical corporate default rates for the past 20 years. 70 The Basic Process The risk calculation process outlined in Basel II for those institutions seeking to apply one of the IRB approaches is also composed of three parts. First, the probability of default (PD) must be obtained. Second, the loss given default (LGD) must be established. Finally, the exposure at default (EAD) must be estimated. The foundation IRB approach requires banks to estimate only the PD, while the LGD and EAD are determined by regulatory guidance. In this article, we focus on one methodology that can be used to estimate the probabilities of default without committing substantial internal resources, to utilize the foundation IRB approach to regulatory capital calculation, as well as to check any internal calculations under the advanced IRB approach. In risk management processes used by virtually all banks, commercial loans are assigned a credit rating. These credit ratings can be letter grades or numerical grades covering multiple states. A mapping of these bank credit ratings can then be made to published ratings of bonds by agencies such as S&P and Moody s. Once this mapping is complete, any loan can then be given an equivalent S&P rating, for example, based on the internal bank rating. Then, based on the applicant s credit rating, historical default performance observations from S&P may be used to estimate the probability and potential timing of default. Thus starting from existing internal bank ratings of commercial loans, probabilities of default can be attached. A Few Kinks While this process of obtaining default probabilities from external data vendors seems straightforward, several caveats are in order. First, most of the empirical work on corporate defaults thus far has concentrated on publicly traded bonds. Due to the private nature of the loan market, there is limited publicly available loan default data. Second, since loan portfolios vary from bank to bank, even if a reasonable default database were available, it would be inappropriate to generalize these results for the entire market. 2 What this implies is that external default data must be carefully scrutinized for suitability of application to any particular bank s portfolio. This scrutiny should center around at least two dimensions: default calculation methodology and characteristics of the instruments. Default Calculation Methodology At least two different methodologies have been used to estimate default probabilities in the commercial sector. S&P s standard default curves use static pools. As used by S&P, the static pool assesses the default rates of all bonds of a given bond rating, regardless of age. 3 The S&P default rates do not take into account the age of an issuance. Default curves obtained from S&P s CreditPro database do not reflect a new issue bias. New issue bias refers to the expected results of the hypothesis that an issuer that has just received a substantial cash inflow from a bond offering or loan is not likely to default in the near term, regardless of the issuer s credit rating. A new issue bias would be expected to reduce marginal default rates in the first three years subsequent to the issuance of new debt, with the effect getting more pronounced as the credit quality declines. To address this concern, Professor Ed Altman from New York University developed a series of default curves based upon this new issue bias hypothesis. As stated in Managing Credit Risk: The Next Great Financial Challenge, The aging effect is intuitively sound, since most companies have a great deal of cash just after they issue a bond. Even if their operating cash flow is negative, they are usually able to meet several periods of interest payments. 4 S&P also notes that relatively few issuers default early in their rated history. 5 Similar to S&P s default analysis, Altman grouped his analysis according to credit-rating cohorts. However, Altman s analysis differed in that his cohorts were organized by issuers with the same original rating at issuance, as opposed to issuers with the same ratings as of a ran-
Table 2 Cumulative Default Rates Year 1 2 3 4 5 6 7 S&P Credit Pro A 0.04 0.12 0.18.027 0.43 0.28.072 BBB 0.29.068.098 1.52 2.04 2.42 2.84 BB 1.07 2.97 5.27 7.26 8.94 10.73 11.82 B 9.29 18.21 24.22 27.71 30.23 32.47 33.99 CCC 24.72 33.06 38.40 42.60 46.87 48.48 49.62 Altman s MMR A 0.00 0.00 0.03 0.15 0.21 0.23.025 BBB 0.02 0.31 0.58 1.25 1.49 1.89 1.99 BB 0.38 1.13 3.78 5.26 7.56 8.49 10.50 B 1.16 4.15 9.75 15.30 19.21 21.62 23.82 CCC 2.06 15.6 28.51 34.53 36.52 42.71 44.91 dom observation date (that is, static pools). Dr. Altman developed his default probability curves by measuring defaults in a particular period relative to the base population in the same period. Therefore, depending on the number of observed issuers within each cohort at the beginning of the year that could possibly default during the year, the denominator used to calculate the probabilities of default may change every year. This methodology was based on the actuarial methodologies used by insurance companies and was named the Marginal Mortality Rate Methodology (MMR). This methodology also differs significantly from S&P s analysis in that Altman measured the magnitude of default as the dollar value of defaults as a percent of the total dollar value of debt rated at the beginning of the period. S&P differs in its methodology by measuring the number of issuers that have defaulted as a percentage of all the issuers rated at the beginning of the measurement period. Altman s data tends to be more volatile from year to year, especially when significant bonds default, such as the multi-billion dollar Texaco default in 1987 or the Enron default in 2001. The seven-year cumulative default rates implied by the two methodologies are shown in Table 2. The difference in marginal default rates over the first few years between the two methodologies is clearly visible, and the difference gets more pronounced as the credit quality declines. Table 3 Time to Default by Rating Category Table 3 indicates that, historically, issuers that received funds at speculative ratings tended to avoid default for longer periods compared with issuers who may have issued debt at higher ratings but subsequently slid down the rating scale because of financial difficulties. For example, Table 3 indicates that issuers who received an initial rating of CCC and eventually defaulted have taken, on average, 3.1 years to default (CCC; Average Years from Original Rating ). Conversely, issuers who had a last rating of CCC and eventually defaulted (but may have had a higher rating when they issued debt) typically defaulted within five months of receiving the CCC rating (CCC; Average Years from Last Rating ). Problems with both methodologies begin to emerge for the highly speculative issuances below CCC (CCC-, CC, and C). Whether using static pools or first rating cohorts, there is no data available for ratings below CCC. While cumulative and marginal default probability data are available from S&P for CCC-rated debt, no default rates are provided by S&P for any ratings below CCC. Default Average Years Average Years Original Rating from Original Rating Last Rating from Last Rating AAA 8.0 AAA N.A. AA 11.9 AA N.A. A 10.7 A N.A. BBB 7.4 BBB 1.3 BB 5.1 BB 1.5 B 3.8 B 1.3 CCC 3.1 CCC 0.4 N.R. N.A. N.R. 3.3 Total 4.7 Total 1.1 71
Table 4 Syndicated Bank Loans versus Corporate Bond At-Issuance Mortality Rates Year 1 Year 2 Year 3 Year 4 Year 5 Bank Bond Bank Bond Bank Bond Bank Bond Bank Bond Baa 0.04% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 054% 0.00% 0.00% Ba 0.17% 0.00% 0.60% 0.38% 0.60% 2.30% 0.97% 1.80% 4.89% 0.00% B 2.30% 0.81% 1.86% 1.97% 2.59% 4.99% 1.79% 1.76% 1.86% 0.00% Caa 15.24% 2.65% 7.44% 3.09% 13.03% 4.55% 0.00% 21.72% 0.00% 0.00% Source: Caouette, John B., Altman, Edward I., and Narayanan, Paul. Managing Credit Risk: The Next Great Financial Challenge. rates will therefore need to be extrapolated by banks for credit ratings below CCC. Hence, it should be noted that any comparisons made and conclusions drawn at the CCC level need to be tempered by a consideration of the lack of data points at this rating level. Instrument/structure characteristics. The second dimension that needs to be scrutinized before default probabilities can be applied to bank portfolios is the characteristics of the loans in the bank portfolio relative to those of the bonds in the published default studies. This is especially the case for speculative-rated debt, whose specific structures greatly affect the cash flow. This is probably one reason why, in Table 2, we note that the difference in default rates between the static pool and at-issuance methodologies increases as we move down the credit rating. In addition to the added liquidity provided by the proceeds of a bond issuance, recently issued bonds, especially speculative bonds, have had structural enhancements such as interest payment guarantees for the first years of the bond s life, or in some cases they have been structured with escrow accounts containing a portion of the bond s interest payments. These factors might contribute to lower initial default rates (and in some cases higher credit ratings). These structural characteristics need to be compared to the loan structures the bank is issuing. Corporate Bonds versus Syndicated Bank Loans As noted above, due to the lack of available data on the historical performance of syndicated bank loans, historical corporate bond performance data has often been used to predict the probability and timing of default that may occur. Table 4 indicates that while cumulative default rates for bonds and loans may even out after a number of years, syndicated loans tend to have much higher initial default rates, again especially at speculative ratings. These rates indicate that bond structures may have been substantially different from loan structures. These differences could be due to the use of more complex structures for bond issuances that are designed to protect investors from default during the years immediately following issuances well as the typically shorter average maturities for bank debt compared to corporate bonds. Unfortunately, very little data is available specifically on the historical performances of syndicated bank loans, making it very difficult to verify any of the findings presented in Table 4. Dr. Altman notes that, The marginal mortality rate results and its information content concerning the aging effect of corporate loan default rates is not conclusive. 6 As the liquidity of the secondary whole-loan market increases in the future, so will the quantity and quality of public information available. Until then, it will remain difficult to accurately draw meaningful conclusions, particularly at the speculative end of the credit-rating range. Siddiqi and Klein may be contacted at nsiddiqi@bearingpoint.net and sklein@bearingpoint.net Notes 1 Standard and Poor s, 2002, Special Report Ratings Performance 2001. 2 Altman, E., Suggitt, H., 2000, Default Rates in the Syndicated Bank Loan Market: A Mortality Analysis, Journal of Banking & Finance, 24, 229-253 3 Standard and Poor s, 2001, CreditPro 5.0 User Guide. According to S&P, A Static Pool consists of all of the rated obligors on the first day of the year or a quarter and these obligors are followed from that point on. Thus, a Static Pool is a grouping of obligors whose members remain constant. The results presented in tables with all Static Pools represent a weighted average based on the number of obligors in each pool and rating category, at the beginning of each period, over a specific time period. 4 Caouette, John B., Altman, Edward I., and Narayanan, Paul., 1998, Managing Credit Risk: The Next Great Financial Challenge. New York: John Wiley and Sons, Inc. 5 Standard and Poor s, 2001, CreditPro 5.0 User Guide. 6 Altman, E., Suggitt, H., 2000. 72 The RMA Journal December 2002 - January 2003