Abstract
This paper offers a joint estimation approach for forecasting probabilities of default and loss rates given default in the presence of selection. The approach accommodates fixed and random risk factors. An empirical analysis identifies bond ratings, borrower characteristics and macroeconomic information as important risk factors. A portfolio-level analysis finds evidence that common risk measurement approaches may underestimate bank capital by up to 17% relative to the presented model.
Acknowledgements
The authors would like to thank two anonymous referees, the editor of this journal and the participants of the EFA Annual Meeting in Athens, the EFMA Annual Meeting in Milan, the International Conference on Price, Liquidity, and Credit Risks in Konstanz, the 11th Symposium on Finance, Banking, and Insurance, Karlsruhe, the Workshop Risk Management, Obergurgl as well as the financial seminars at the University of Bristol, Deutsche Bundesbank, University of Edinburgh, Global Association of Risk Professionals, Leibniz University Hannover, University of Ulm, Hong Kong Institute of Monetary Research, Melbourne Centre for Financial Studies and The University of Melbourne for valuable comments. The support of the Australian Centre for Financial Studies, the Australian Prudential Regulation Authority, the Centre for International Finance and Regulation and the Hong Kong Institute for Monetary Research is gratefully acknowledged.
Notes
1 We thank an anonymous referee for pointing this out.
2 A simulation study was conducted to ensure the consistency of the estimators.
3 Moody's collects general information on bond issuer and issues as well as default and market price information given default events. Note that the only information this paper uses which is uniquely generated by Moody's is the credit rating. This credit rating is very similar to the ones published by other rating agencies such as Standard & Poor's and Fitch: we have hand-collected from the Bloomberg database 63 151 ratings at origination which were rated by Moody's and Standard & Poor's, 38 346 bond ratings at origination which were rated by Moody's and Fitch and 34 578 bond ratings at origination which were rated by Standard & Poor's and Fitch. The Spearman correlation coefficient is 0.9819 for Moody's and Standard and Poor's, 0.9738 for Moody's and Fitch and 0.9702 for Standard and Poor's and Fitch. This analysis of these rating pairs suggests that credit ratings are similar for the three rating agencies as the correlation coefficients are very high. These findings are consistent with CitationGuettler and Wahrenburg (2007) as well as interviews with employees of the three agencies.