Assessing Alternative Assumptions on Default Risk Capital in the Trading Book

1. Assessing Alternative Assumptions on Default Risk Capital in the Trading Book Gary Dunn: UK Financial Services AuthorityMichael Gibson: Federal Reserve BoardGloria Ikosi: Federal Deposit Insurance CorporationJonathan Jones: Office of Thrift SupervisionCharles Monet: US Securities and Exchange CommissionMichael Sullivan: Office of the Comptroller of the Currency

2. Disclaimer • The views presented here are solely those of the authors and do not necessarily represent those of the institutions with which they are affiliated. • The model presented is for discussion purposes only to illustrate certain elements of the issue and is neither endorsed nor prescribed by any agency.

3. Motivation for the study • Basel 2 requires firms to model incremental default risk in the trading book • AIG Trading Book Working Group is discussing guidelines for models of default risk in the trading book • People are interested in knowing the effects of different modeling choices

4. Outline of the talk • A model of default risk • How to quantify the benefit of liquidity? • Three test portfolios • Results • Key findings

5. A model of default risk • Single-period Gaussian copula • Similar to A-IRB, reflects concentration • 99.9 percentile VaR • Correlation parameter = 10%, 20%, 30% • Fixed recovery = 40% • “Constant level of risk” incorporated by scaling up short-horizon PD

6. Liquidity and capital horizons • Liquidity horizon represents the frequency at which the portfolio is rebalanced to a target level of risk (or rating). • Capital horizon represents the time period over which default events are measured. • We consider 1 month, 3 months, 12 months for both LH and CH.

7. How to quantify the benefit of liquidity? • What PD to use in the model? • Two sources of data • Moody’s default database • Directly compute default rates at various liquidity horizons • MKMV July 2004 study • Compute “surprise default” ratio at 1-month liquidity horizon using MKMV EDFs

8. Constant position vs. constant risk for a Ba credit

9. Constant position vs. constant risk for a Caa-C credit

10. Constant position vs. constant risk for a Ba credit A = 1.30% PD scaling factor = 0.34 = B/A B = 0.44%

11. PD scaling factors (shown for a 1-month liquidity horizon and a 1-year capital horizon)

12. Three test portfolios Note: All portfolios have the same A-IRB capital requirement.

13. Results (Moody’s, corr.=0.2) (See Appendix for full results).

14. Key findings • Model suggests lower capital than A-IRB • Short positions reduce default risk • Liquidity alone reduces default risk by 20-40 percent with monthly rebalancing • Reducing capital horizon from 1 year to 3 months reduces default risk by 30-50 percent • Could not compare with Specific Risk Add-ons without a more realistic portfolio

15. Appendix – Details of all scenarios • In the tables below, the reported values indicate the 99.9% downside loss over the capital horizon, net of recoveries. The liquidity horizon is denoted by LH.

19. Appendix – Details of portfolios

20. Long only

21. Long bias

22. Long bias with lumps