Criteria Structured Finance ABS: Revised U.S. ABS Rating Methodology And Assumptions For Credit Card Securitizations

Transcription

1 January 28, 2010 Criteria Structured Finance ABS: Revised U.S. ABS Rating Methodology And Assumptions For Credit Card Global ABS Criteria Officer: Joseph F Sheridan, New York (1) ; joseph_sheridan@standardandpoors.com Primary Credit Analysts: Ildiko Szilank, New York (1) ; ildiko_szilank@standardandpoors.com Frank J Trick, New York (1) ; frank_trick@standardandpoors.com Analytical Manager, ABS Term New Issuance: Felix Herrera, CFA, New York (1) ; felix_herrera@standardandpoors.com Analytical Manager, ABS Term Surveillance: Eric Hedman, CFA, New York (1) ; eric_hedman@standardandpoors.com Structured Finance Research: Erkan Erturk, Ph.D., New York (1) ; erkan_erturk@standardandpoors.com Table Of Contents SCOPE OF THE CRITERIA SUMMARY OF CRITERIA UPDATE IMPACT ON OUTSTANDING RATINGS EFFECTIVE DATE AND TRANSITION METHODOLOGY Unemployment Rates And Credit Card Losses Benchmarking And Peer Comparison Credit Card ABS Rating Sensitivity To Economic Downturn And Relative Stability Of Different Rating Categories APPENDIX 1

2 Table Of Contents (cont.) Regression Analysis RELATED RESEARCH Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

3 Criteria Structured Finance ABS: Revised U.S. ABS Rating Methodology And Assumptions For Credit Card (Editor's Note: The criteria in this article supplement and partially supersede the criteria described in the "Base Case And Stress Assumptions" section for "loss rate" of the criteria article "General Methodology And Assumptions For Rating U.S. Credit Card," published Dec. 15, 2008.) 1. Standard & Poor's Ratings Services is refining and adapting its methodology and assumptions for rating U.S. credit card asset-backed securities (ABS). We are publishing this article to help market participants better understand our approach to reviewing securities backed by credit card loans. This article is related to our criteria article "Principles-Based Rating Methodology For Global Structured Finance Securities," which we published on May 29, SCOPE OF THE CRITERIA 2. Standard & Poor's revised its criteria for credit card ABS as part of its overarching objective to make ratings comparable across sectors, geographic regions, and over time. The changes reaffirm the long-standing definitions of our ratings. We intend for each rating symbol (e.g., 'AAA') to connote a comparable overall view of creditworthiness wherever and whenever it appears. 3. To help enhance the comparability of our ratings, we recently embraced the use of specific economic stress scenarios to calibrate our criteria. For our 'AAA' stress scenario, we use the Great Depression, a time of extreme economic stress. Accordingly, we intend for 'AAA' rated credit card ABS securities under the revised criteria to be able to withstand that level of economic stress without defaulting (although, they could suffer downgrades under that level of stress). 4. The principal updates to our criteria are: We will supplement our established criteria methodology by comparing all U.S. credit card master trust portfolios with our U.S. Bankcard Credit Card Quality Index (CCQI). The CCQI will serve as an industry average benchmark (the benchmark pool), and we will use it to compare and measure all rated pools when we forecast portfolio-specific stressed charge-off rates and when we evaluate the adequacy of provided credit support; We have established a 'AAA' peak charge-off rate of 33% for the benchmark pool that will be operative, provided the historical relationship between unemployment rates and industry charge-off rates continues to hold, when expected charge-off rates for the benchmark pool are within a range of about 5% to 10%. By holding the 'AAA' stressed peak charge-off rate constant during normal economic cycles, we effectively increase the loss coverage multiple when charge-off rates are lower during more benign economic environments; We will evaluate the credit risk of an actual pool relative to that of the benchmark pool as well as its peer group. During our evaluation, we will review originator-specific data on historical performance, pool-specific loan and borrower characteristics, and our forward-looking base case performance; and After we compare an actual pool with the benchmark pool and peer group, we will adjust the actual pool's specific 'AAA' stressed charge-off rates to the extent there are deviations in originators' historic and expected 3

4 performance relative to the benchmark pool and peer group or current loan and obligor characteristics relative to the peer group. SUMMARY OF CRITERIA UPDATE 5. The criteria outlined in this article supplement and partially supersede previously published criteria for credit card ABS described in the "Base Case And Stress Assumptions" section for "loss rate" of the criteria article "General Methodology And Assumptions For Rating U.S. Credit Card," published Dec. 15, IMPACT ON OUTSTANDING RATINGS 6. At this time, the changes outlined in this article will not in and of themselves have an impact on our outstanding U.S. credit card ABS ratings. 7. It is important to note that while this criteria article is focused on credit card losses (also referred to herein as charge-offs), other variables may have a more significant impact on a transaction's credit profile, ratings, and performance. For example, a one percentage point decrease in the base-case scenario payment rate could have a more significant impact on credit enhancement levels and on ratings than a one percentage point increase in the base-case charge-off rate. 8. This article is intended to provide increased transparency and establish a clear framework for benchmarking pools of credit card loans and provide insight into the stability of credit card ABS ratings. We believe that all rated securities meet or exceed the new criteria outlined in this article, and we expect that the consistent application of these criteria will enhance ratings quality and consistency. 9. As of December 2009, Standard & Poor's maintained surveillance on 770 rated U.S. credit card ABS securities. In 2009 we raised 183 ratings and lowered 96. However, the number of downgrades would have been greater and the number of upgrades would have been fewer if certain issuers did not provide additional credit enhancement during the past year. Four-hundred-and-thirteen rated securities benefited from this additional enhancement. Without the additional support, 236 of 413 ratings (approximately 57%) would have remained in the same rating category. We would have likely lowered the remaining 177 ratings (approximately 43%) by up to one rating category. For additional information on ABS transactions that benefited from additional support, see "ABS Credit Ratings Would Have Remained Relatively Stable Even Without Additional Support," published Jan. 25, 2010, and "U.S. Credit Card ABS Issuer Report: Additional Credit Support From Originators Results In 154 Upgrades Of Ratings Previously On Watch Negative," published Oct. 2, We expect that the criteria changes outlined in this article will enhance the stability of our 'AAA' ratings during normal economic cycles. EFFECTIVE DATE AND TRANSITION 11. These criteria are effective immediately for all new and outstanding U.S. credit card ABS transactions. Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

5 METHODOLOGY 12. We will explicitly use our U.S. CCQI as an industry benchmark to compare and measure outstanding pools against, based on collateral performance. The CCQI is a monthly performance index that aggregates performance information across Standard & Poor's rated bankcard transactions in several key risk areas. We are also emphasizing the use of peer group comparisons in refining our evaluation of a specific pool relative to other similar portfolios based on collateral characteristics. The 'AAA' peak charge-off rate for the benchmark pool is 33%. This benchmark 'AAA' peak charge-off rate reflects our opinion of expected performance under conditions of extreme economic stress. 13. In addition to stresses on performance variables such as payment rate, purchase rate, yield, and cost of funds, the credit enhancement supporting a 'AAA' rated security should also be sufficient to withstand a pool-specific 'AAA' stressed charge-off rate that peaks 12 months following the breach of an early amortization event trigger. We consider the 'AAA' peak charge-off rate of 33% associated with the benchmark pool as a fixed anchor point. However, pool-specific variations in historic actual and projected base-case performance relative to the benchmark pool, as well as loan or obligor characteristics relative to an issuer's peer group, may cause the 'AAA' stressed charge-off levels for actual pools to vary (higher or lower) from the benchmark pool's 33% peak charge-off level. In addition, we review pools on an ongoing basis and portfolio-specific base-case and stressed peak loss projections may change over time. For example, we may adjust our projections if performance deviates from industry trends, if performance is not in line with what we would expect based on current economic conditions, or if there has been a material change in underwriting standards, geographic diversification, or product mix. 14. We believe that issuer-specific historical pool performance for a comparable managed portfolio is a useful indicator of future performance. Therefore, issuer-specific portfolio performance will be a significantly weighted variable in the comparative analysis. We will consider the managed portfolio's historical performance relative to the benchmark pool as well as our expectations for future performance in the base-case scenario going forward. 15. When rating or surveilling credit card ABS transactions, we use fundamental credit analysis, which incorporates various qualitative and quantitative elements. Our ratings arise in large part from certain base-case assumptions for key variables, including losses, yield, payment rates, purchase rates, and coupon rate. The base-case scenario reflects our view of how a pool is likely to perform and reflects historical performance during various economic environments. The base-case scenario is also based on our opinion of the card originator's/servicer's overall experience, the quality and consistency of its origination activities and underwriting, and its account management, collection, and servicing practices. 16. We also look at factors such as the detail and depth of data on the collateral, the amount of performance history for the card originator/servicer through varying economic and business cycles, and examples of operational strategies the issuer has used to mitigate credit risk. For example, we may use more stressful assumptions for pools with less than three years of data available, or for pools from originators/servicers with, in our opinion, less experience managing and servicing credit card asset pools. Our assumptions for such pools generally will rely more heavily on comparisons with portfolios with similar loan characteristics, vintage data (for example, information on one year of originations), and worst-case scenario figures. 17. After we determine our base-case assumptions for each variable, we use them as starting points to simulate a deteriorating credit environment by stressing all variables simultaneously and applying the stresses to the base-case 5

6 assumptions progressively until we reach the maximum stress level (for charge-offs, the peak charge-off). We then maintain the maximum stress level until the notes' legal final maturity date. The stresses vary for each rating category. 18. Our base-case and stress assumptions for pool-specific charge-off rates include factors that will affect losses in a stress scenario. Some of the considerations that affect our base-case loss assumptions include vintage loss performance, delinquency roll rates, lagged loss rates based on receivables growth, account origination strategies, account seasoning, FICO credit tier distribution of new accounts and migration of FICO scores for existing accounts, geographic concentrations, and credit line management strategies. 19. When we compare a pool's historic actual and future expected loss rates with the benchmark pool's historical performance, we also make adjustments to the actual pool performance based on the quality and consistency of underwriting over time. For example, if FICO scores are drifting downward, it may be necessary to adjust pool specific charge-off rates upward. Likewise, for each account and obligor attribute, we will make qualitative adjustments for variations from the benchmark pool. 20. If a pool's expected charge-offs are higher than what we would expect for the benchmark pool in the base-case scenario, then the pool-specific 'AAA' stressed peak charge-off will likely be above 33%. 'AAA' stressed peak charge-off rates that are lower than 33% could also be applied in our cash flow analysis to the extent there are pools with higher credit quality assessments relative to the benchmark pool. 21. Although the stressed peak loss rate plays a central role in our rating analysis, it's not the only factor we consider. Our assessment of the interplay of many factors will affect our rating decisions. For example, higher payment rates by the obligors or higher yield from the underlying assets could offset an increase in losses when we're determining a new rating or considering a rating action during the surveillance of an existing rating. 22. For greater detail on our established criteria methodology and assumptions for rating U.S. credit card ABS, please refer to "General Methodology And Assumptions For Rating U.S. Credit Card," published Dec. 15, Supporting research and rationale for the 'AAA' peak stressed charge-off assumption for the benchmark pool 23. General-purpose bank credit cards were not around in the U.S. during the Great Depression. Bank of America launched its BankAmericard credit card program the predecessor of Visa in A group of banks joined together in 1966 and created the InterBank Card Association now known as MasterCard. In the absence of historic credit card performance data during periods of extreme economic stresses, we have formed an opinion of how credit card losses should behave in a 'AAA' stress scenario based on an examination of loss performance during recent economic downturns and an analysis of the relationship between charge-off rates and key economic indicators. 24. The rationale supporting our peak charge-off rate assumption for 'AAA' credit card ABS ratings draws from recent observed data and an analysis of the relationship between credit card charge-off rates and key economic indicators such as the unemployment rate. During times of recession, some studies have shown that there is a one-to-one relationship between the unemployment rate and credit card losses. Standard & Poor's published a research study investigating this relationship as well (see "The Recession Combined With Severe Home Price Drops Has Increased The Unemployment Rate s Correlation With U.S. Credit Card Losses," published Aug. 31, 2009). Based Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

7 on the findings of our regression analysis, we expect that the annual credit card loss rate could reach at least 27% in an extreme scenario where the unemployment rate reaches 25% (as we saw during the Great Depression). 25. Based on our analysis, we expect that the correlation between the unemployment rate and credit card losses will remain strong in periods of double-digit unemployment. The regression coefficient may increase beyond the current level, however, implying that credit card loss rates could rise at a faster pace than unemployment. In other words, the historical linear regression relationship could become an exponential one in double-digit unemployment rate environments. Overall, this finding provides the basis for our assumption that, in a 'AAA' scenario, we would see a rapid increase in charge-off rates that would correspond with rising unemployment rates. This expectation is in line with our revised 'AAA' stressed scenario industry benchmark of about 33% peak charge-off rates, which reflects our opinion of expected performance under conditions of extreme economic stress. For a more detailed review of our regression analysis, please refer to the Appendix at the end of this article. Unemployment Rates And Credit Card Losses 26. Generally speaking, the unemployment rate mirrors the economic environment. During the Great Depression in the 1930s, the unemployment rate reached 25%. In December 1982, the unemployment rate was 10.8%, the highest we've observed since the monthly unemployment data became available in Since then, the unemployment rate has hovered around 6%, and it remained under 8% through 2008 (see chart 1). It then passed 8% in February By June 2009, the unemployment rate had climbed to 9.5% and then surpassed 10% in October Unemployment was at 10% at the end of December Chart 1 7

8 27. The reason why there is a significant relationship between the unemployment rate and credit card losses is simple: Unemployment tends to increase during economic downturns and the loss of income leads to borrowers' inability to repay credit card loans (see chart 2). Unemployed consumers are also more likely to rely on their credit cards, and they may use their credit cards to compensate for their lost income, but only for a while. Many credit card borrowers who fall behind on their monthly bills and are unable to find jobs will eventually face credit card delinquencies and charge-offs. 28. In addition, other borrowers who are employed and have relatively strong credit profiles but are worried about job security during periods of economic stress may reduce their credit card purchases to increase their savings. This can lower the revolving debt and consequently increase charge-off rates. This is why market researchers use the unemployment rate as a key determinant of delinquencies and bankruptcies during economic crises. As observed in the current credit environment, this relationship becomes stronger in a stressed economy because traditional non-employment-related sources of credit card loan repayments decrease. That is, many U.S. households may experience a significant decline in their worth held in the stock market and a more limited ability to extract depleted equity from their homes. As a result, those households may struggle to meet their credit card obligations, which, in turn will generally lead to higher credit card losses. Chart 2 Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

9 Benchmarking And Peer Comparison 29. To help maintain ratings comparability across the U.S. credit card ABS sector, we will compare the credit risk of a master trust portfolio with that of the industry average CCQI benchmark pool as well as other master trust portfolios that we consider to be in the same peer group. The factors that we will review include originator-specific data on risk management and operational strategies, historical performance, pool-specific loan and borrower characteristics, and our forward-looking base case expected performance. The peer group analysis affords analysts the opportunity to make comparisons across issuers and can be useful in identifying trends and market developments that may be less apparent when looking exclusively at a single portfolio or originator. 30. We use our CCQI as a proxy for the average U.S. credit card portfolio. As of October 2009, the U.S. CCQI tracked approximately $401.6 billion bankcard receivables backing Standard & Poor's rated credit ABS. In aggregate, the U.S. bankcard CCQI tracks the performance of receivables which represent approximately 45% of the total U.S. revolving consumer debt balance. The benchmark pool is an aggregate of all bankcard pools tracked by the CCQI. 31. The CCQI aggregates performance information across Standard & Poor's rated bankcard transactions in the following key risk areas: receivables outstanding, yield, payment rate, charge-off rate, delinquencies, base rate, and excess spread rate. Each master trust's weighting is determined by the size of its outstanding eligible principal receivables divided by the total outstanding eligible principal receivables for all trusts included in the index. The resultant CCQI weighted average performance variables are then determined. 32. We also publish comparative portfolio-specific data in our U.S. Credit Card ABS Issuer Report (Issuer Report) to complement our CCQI. The Issuer Report highlights issuer-specific performance and provides the collateral data for all credit card receivables pools backing outstanding public/144a credit card ABS transactions that we rate. The report includes comparative data on pool performance as well as collateral characteristics such as average credit limit, average account balance, credit line utilization rate, geographic concentrations, and age of accounts. Similar statistics are used to compare pools within an issuer's peer group when we review a master trust as part of the rating process. Credit Card ABS Rating Sensitivity To Economic Downturn And Relative Stability Of Different Rating Categories 33. Standard & Poor's ratings generally reflect our opinion of the level of stress that a bond with a particular rating should be able to withstand without defaulting. The higher a bond's rating, the stronger, in our opinion, is the bond's capacity to meet its financial obligations during adverse economic scenarios. 34. The more junior the notes are in the capital structure, the more likely it is that their stress scenario is closer to that of the existing base-case scenario. In our benchmark pool, we define the 'B' stress scenario as one where the 'B' peak charge-off rate is also the expected loss given the existing base case economic scenario. The 'B' stress scenario is consistent with the base-case scenario and will change to reflect changing economic conditions or outlook. As such, the junior notes' credit enhancement and ratings will be more sensitive to changes in our base-case expectations than the more senior notes in the capital structure. For example, 'B' rated securities during a booming economy featuring rapid employment and income growth would be at greater risk of downgrade or default than higher rated securities as the economy weakened going forward. 9

10 35. Since the stress scenarios for the more senior tranches are further removed from the existing base-case scenario (more stressful), we would expect that 'AAA' rated senior tranches would have relatively stable enhancement levels, and ratings, until the downturn hits a point that is beyond a level considered consistent with a normal economic cycle. 36. Although we assign our initial ratings based on our outlook at that time, reality may play out differently. As our loss expectation for pools backing an outstanding credit card ABS changes because of changes in the market and economy, we review the ratings. 37. We would consider the relationship between projected losses and currently available credit enhancement before taking any rating action. Other factors that would likely influence a potential rating action include the seasoning of the portfolio, the structural features of the transaction, and the expected number of months before maturity. We typically would also consider the performance of the other primary performance variables including portfolio yield, payment rate, and purchase rate. Ratings can be adversely affected by an adverse change in any of theses variables, and the effects of deterioration in one variable could be offset by stability or improvement in another. Moreover, a rating may be more sensitive to a change in a performance projection that is not related to charge-offs. For example, a one-percentage point decrease in the base-case payment rate could have a more significant impact on credit enhancement levels and on the ratings than a one percentage point increase in the base-case charge-off rate. 38. Provided the historical relationship between the unemployment rate and industry charge-off rates continues to hold, the 'AAA' stressed peak charge-off rate for the benchmark pool of 33% should remain operative in periods of mild to moderate economic stress when unemployment rates are below 10%. When charge-offs are stressed to 33% and the expected industry average charge-off rate ranges from 5% to 10%, this corresponds with a loss coverage range of about 6.6x-3.3x the base case. In the past, the stress multiples we observed were typically in the 3x-5x range for 'AAA' rated credit card ABS. The increase at the upper end of the loss coverage band reflects the "floor" concept that we are instituting by holding the 'AAA' stressed peak charge-off rate at 33% even if expected charge-off rates of 5% are being realized in a mild economic environment. Although we expect that our 'AAA' stressed peak charge-off rate assumption for the benchmark pool to remain constant at 33% during normal economic cycles, we recognize that if economic and market conditions migrate significantly beyond the normal ranges for cyclical fluctuations, even the 'AAA' stressed peak charge-off rate would have to be higher. We expect that the peak charge-off rate assumption for the benchmark pool would range from about 3x to 3.5x the base-case scenario in an economic environment where the unemployment and base-case charge-off rates are between 10% and 20%. For the benchmark pool, the 'AA', 'A','BBB', and 'BB' peak loss scenarios are increasingly less stressful relative to 'AAA' stress case and are expected to be in the range of about 2.5x-5.0x, 2.0x-3.75x, 1.5x-2.5x, and 1.25x-1.5x the base-case when the base-case charge-off rates are in the 5%-20% range. The base-case and stressed peak charge-off rates will vary (higher or lower) for actual securitization trusts based on a portfolio- and servicer-specific analysis and stressed charge-off rates as a multiple of the base case will not be the same for all pools. For example, subprime pools with charge-offs that are materially higher than the industry average are likely to have higher stressed peak charge-off rates that are a lower range of multiples of the base case for that pool. In the 'B' rating scenario, peak losses are unstressed or equal to 1x the base case. Rating stresses and peak charge-off assumptions for actual pools will vary based on differences in pool characteristics and historic performance relative to the benchmark pool. 39. As our outlook for the economy, the credit card industry, and specific pools change, we reassess the potential impact of these changes on our outstanding ratings. This impact will be greatest for credit card ABS rated 'B' and 'BB' and least for securities rated 'AA' and 'AAA'. Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

11 40. Chart 3 illustrates the expected stressed charge-off rates at various rating categories based on the base-case charge-off rates (5% to 15%) in various economic scenarios for the benchmark pool. In the 'B' rating scenario, the peak charge-off rate would equal the base-case forecast. For the 'AAA' rating scenario, provided the relationship between charge-off rates and economic variables such as unemployment remains in line with historical patterns, the 'AAA' peak would be expected to remain constant at 33% for the benchmark pool until base-case charge-off percentage forecast moves into the double digits. 'AAA', 'A', and 'BBB' are the three most common ratings in the credit card ABS market. Chart 3 illustrates that the peak loss rate assumed when assigning a rating in the 'B' category moves with the base case forecast under all economic conditions and as ratings move up the scale, the stressed peak loss rate assumed is less vulnerable to change as the economy moves away from benign conditions to more stressed conditions. As a result, 'AAA' ratings are generally expected to be relatively stable during normal economic cycles and lower rated securities are generally expected to be more pro-cyclical as stressed charge-off rates are expected to change at earlier stages of economic declines. Further, the point in a normal economic down cycle with rising unemployment rates where a rating is vulnerable to change is expected to occur sooner as the ratings move down the scale. In other words, as the economy moves from benign conditions with low unemployment rates to more stressful conditions, the 'BBB' stressed loss assumptions are likely to move higher before the 'A' stress case assumptions, and they are both expected to change before the 'AAA' stress case would need to be adjusted. Although base-case and stressed charge-off rates will vary for actual securitization trusts based on a portfolio- and servicer-specific analysis and are subject to change as discussed herein, this same basic rating scenario framework will apply in evaluating the adequacy of provided credit enhancement. 41. We believe that benchmarking and the use of specific economic scenarios to calibrate our ratings criteria will enhance ratings comparability. However, the changes to our criteria outlined herein do not in any way diminish the 11

12 importance of issuer- and portfolio-specific analysis. Credit card ABS master trust pools are actively managed and some card issuers will employ account origination and management strategies that may be more successful at mitigating risk in challenging economic environments than others. In addition, other portfolio-specific credit characteristics, including performance variables such as portfolio yield, payment rate, and purchase rate, are important considerations in our analysis. APPENDIX Regression Analysis 42. For our analysis, we used information from our monthly CCQI (1992 to 2009) and other economic data, such as the unemployment rate, the S&P Case-Shiller Home Price Index, the S&P 500 equity index, personal bankruptcy filings, personal income, the Federal Reserve's outstanding consumer credit amount, and GDP growth rate from Bloomberg. 43. Based on the pair-wise correlations, which are consistent with our expectations, we observed that the changes in credit card losses are positively correlated with changes in the unemployment rate (0.60) and in personal bankruptcy filings (0.31), while credit card losses are negatively correlated with changes in home prices (-0.53), personal income (-0.44), the GDP growth rate (-0.29), and the S&P 500 (-0.23) (see appendix table 1). The unemployment rate is the most significant economic variable that affects credit card loss rates, followed by the home price index. Similarly, the changes in unemployment rates are negatively correlated with the changes in personal income, the equity market, home prices, and the GDP growth rate. When two or more independent variables are correlated, they basically measure the same relationship. Given the high correlation among independent variables, we believe it's reasonable to assume that unemployment conveys essentially the same information for personal income, the equity market, GDP, and to a certain extent, home price behaviors in our regression analysis. As a result, we could consider eliminating these variables from our key factors list and regression equations when necessary. Appendix Table 1 Correlation Coefficients 12-Month change in loss rate 12-month change in unemployment rate 12-month change in personal bankruptcy filings 12-month change in personal income 12-month change in consumer credit outstanding 12-month change in loss rate 12-month change in unemployment rate 12-month change in personal bankruptcy filings 12-month change in personal income 12-month change in consumer credit outstanding 12-month change in S&P month change in S&P Case-Shiller GDP growth rate (0.44) (0.01) (0.23) (0.53) (0.29) (0.80) (0.30) (0.72) (0.43) (0.57) (0.11) 0.03 (0.05) (0.25) (0.13) (0.44) (0.80) (0.11) (0.01) (0.30) Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

13 Appendix Table 1 Correlation Coefficients (cont.) 12-month change in S&P month change in S&P Case-Shiller (0.23) (0.72) (0.05) (0.53) (0.43) (0.25) GDP growth rate (0.29) (0.57) (0.13) While the actual levels of credit card losses and unemployment rates are somewhat correlated, the relation is more meaningful and stronger when we use the rolling 12-month percent changes to estimate the functional relationship between them (see appendix chart 1). 45. Beginning with a regression model, we considered most relevant variables regardless of any multi-collinearity problem in the regression model. Multi-collinearity results from high correlations among independent variables, such as unemployment and personal income or unemployment and home price changes, which makes it difficult to determine reliable estimates or regression coefficients. 46. The regression equation 1 in appendix table 2 shows that percent changes in unemployment, bankruptcy filings, outstanding consumer credit, and home prices are the key loss determinants with acceptable t-statistics. If we eliminate the home price variable a potential source of multi-collinearity from the equation (see regression equation 2 in appendix table 2), the model's ability (r-square) to explain the behavior of loss rates is around 38%, down from 62% in regression equation 1. This highlights how important the percent changes in home prices are in explaining the volatility of credit card losses. We then eliminated all independent variables and kept only the percent 13

14 changes in unemployment rates to explain the credit card loss behavior (see appendix chart 1). In appendix table 2, regression equation 3, with an r-square of about 36% and significant t-statistics for the unemployment rate, suggests that the contribution of other variables (other than home prices) to explain credit card loss behavior is very limited once the unemployment rate is present in the model. That's why researchers typically use unemployment as the sole variable to predict credit card losses. While the multi-collinearity between unemployment and home prices may, to a certain extent, be an issue in the regression analysis, the correlation and regression analysis highlight how important home prices and the unemployment rate are in explaining credit card loss behavior. Appendix Table 2 Regression Results Regression equation 1 R square: 61.7% Coefficients Standard error t-stat P-value (%) Intercept (0.06) 0.03 (1.74) month change in unemployment rate month change in personal bankruptcy filings month change in consumer credit outstanding month change in S&P Case-Shiller (1.36) 0.14 (9.83) 0.00 Regression equation 2 R square: 37.7% Coefficients Standard error t-stat P-value (%) Intercept (0.11) 0.04 (2.63) month change in unemployment rate month change in personal bankruptcy filings month change in consumer credit outstanding Regression equation 3 R square: 35.6% Coefficients Standard error t-stat P-value (%) Intercept month change in unemployment rate Regression equation 4 - reduced data R square: 70.00% Coefficients Standard error t-stat P-value (%) Intercept (0.03) 0.01 (2.57) month change in unemployment rate Finally, after attempting to control the effects of the expansion period in the early 1990s (due to rapid rise of consumer credit) and the bankruptcy period in 2005 (following a change in bankruptcy law) in the regression models, we removed these two time periods from the analysis to understand the undistorted functional relationship between the unemployment rate and credit card losses (see regression equation 4 in appendix table 2 and appendix chart 2). 48. During the expansion and bankruptcy periods, unemployment wasn't the obvious driver of credit card losses. In fact, the rolling correlation coefficients between the unemployment rate and credit card losses were negative during those times due to bankruptcy code changes and the expansion of the credit card industry, which eventually led to higher losses in We eliminated about 27% of the monthly data points where the monthly unemployment rate and credit card loss relationship was negative for regression equation 4 (see appendix table 2 Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

15 and appendix chart 2). As we expected, r-square reached the 70% level with significant t-statistics for the unemployment rate variable: The coefficient for the unemployment rate is almost one (it is 0.99), which means that a 100% increase in the unemployment rate roughly results in a 100% increase in credit card losses during times of high correlation and economic downturns. We also used the regression equation 4 in appendix table 2 to calculate the tightly correlated relationship between actual and predicted credit card losses. RELATED RESEARCH "Criteria Structured Finance ABS: General Methodology And Assumptions For Rating U.S. Credit Card," published Dec. 15, "The Recession Combined With Severe Home Price Drops Has Increased The Unemployment Rate s Correlation With U.S. Credit Card Losses," published Aug. 31, "U.S. and Canada Credit Card Quality Index Report: Losses Climb 60 Basis Points After Two Months Of Improvement," published Jan. 14, "U.S. Credit Card ABS Issuer Report: Additional Credit Support From Originators Results In 154 Upgrades Of Ratings Previously On Watch Negative," published Oct. 2, "ABS Credit Ratings Would Have Remained Relatively Stable Even Without Additional Support," published Jan. 25, "Criteria Structured Finance ABS: How Credit Card Bank Downgrades Could Affect Ratings On U.S. Credit Card ABS," published Oct. 7, "Criteria Structured Finance ABS: Methodology And Assumptions For Rating Short-Term U.S. Money Market-Eligible Notes Backed By Credit Card Receivables," published July 2,

16 "U.S. Economic Forecast: To A Prosperous New Year," published Jan. 11, "2010 Forecast: The Recovery Is Slowly Picking Up Steam, But Some Obstacles Could Still Derail It," published Dec. 23, These criteria represent the specific application of fundamental principles that define credit risk and ratings opinions. Their use is determined by issuer- or issue-specific attributes as well as Standard & Poor's Ratings Services' assessment of the credit and, if applicable, structural risks for a given issuer or issue rating. Methodology and assumptions may change from time to time as a result of market and economic conditions, issuer- or issue-specific factors, or new empirical evidence that would affect our credit judgment. Standard & Poor s RatingsDirect on the Global Credit Portal January 28,

17 Copyright ( c ) 2010 by Standard & Poor s Financial Services LLC (S&P), a subsidiary of The McGraw-Hill Companies, Inc. All rights reserved. No content (including ratings, credit-related analyses and data, model, software or other application or output therefrom) or any part thereof (Content) may be modified, reverse engineered, reproduced or distributed in any form by any means, or stored in a database or retrieval system, without the prior written permission of S&P. The Content shall not be used for any unlawful or unauthorized purposes. S&P, its affiliates, and any third-party providers, as well as their directors, officers, shareholders, employees or agents (collectively S&P Parties) do not guarantee the accuracy, completeness, timeliness or availability of the Content. S&P Parties are not responsible for any errors or omissions, regardless of the cause, for the results obtained from the use of the Content, or for the security or maintenance of any data input by the user. The Content is provided on an "as is" basis. S&P PARTIES DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, ANY WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE OR USE, FREEDOM FROM BUGS, SOFTWARE ERRORS OR DEFECTS, THAT THE CONTENT'S FUNCTIONING WILL BE UNINTERRUPTED OR THAT THE CONTENT WILL OPERATE WITH ANY SOFTWARE OR HARDWARE CONFIGURATION. In no event shall S&P Parties be liable to any party for any direct, indirect, incidental, exemplary, compensatory, punitive, special or consequential damages, costs, expenses, legal fees, or losses (including, without limitation, lost income or lost profits and opportunity costs) in connection with any use of the Content even if advised of the possibility of such damages. Credit-related analyses, including ratings, and statements in the Content are statements of opinion as of the date they are expressed and not statements of fact or recommendations to purchase, hold, or sell any securities or to make any investment decisions. S&P assumes no obligation to update the Content following publication in any form or format. The Content should not be relied on and is not a substitute for the skill, judgment and experience of the user, its management, employees, advisors and/or clients when making investment and other business decisions. S&P's opinions and analyses do not address the suitability of any security. S&P does not act as a fiduciary or an investment advisor. While S&P has obtained information from sources it believes to be reliable, S&P does not perform an audit and undertakes no duty of due diligence or independent verification of any information it receives. S&P keeps certain activities of its business units separate from each other in order to preserve the independence and objectivity of their respective activities. As a result, certain business units of S&P may have information that is not available to other S&P business units. S&P has established policies and procedures to maintain the confidentiality of certain non-public information received in connection with each analytical process. S&P may receive compensation for its ratings and certain credit-related analyses, normally from issuers or underwriters of securities or from obligors. S&P reserves the right to disseminate its opinions and analyses. S&P's public ratings and analyses are made available on its Web sites, (free of charge), and and (subscription), and may be distributed through other means, including via S&P publications and third-party redistributors. Additional information about our ratings fees is available at