Parameterizing Credit Risk Models
36 Pages Posted: 13 May 2004 Last revised: 21 Feb 2009
Approaches for modeling and estimating individual credit risk have been considerably improved during the last years, and latterly practitioners and researchers in the banking industry increasingly focus on quantification of portfolio credit risk. The main problem of this task is the lack of adequate time series of default data. Therefore there is little empirical evidence on the relevant input parameters for the various credit risk modeling approaches. As a consequence, calculations of economic capital may yield very different results and internal models will not be envisaged for the determination of regulatory capital requirements. The present contribution firstly presents three popular portfolio credit risk models and shows how they can be comparably parameterized using a likelihood framework. Then the respective input parameters of all three models are estimated from a large database using a time-series of German corporate bankruptcies. Several restrictions on the available information set are introduced and compared. At last we analyze the forecasted loss distributions generated by each model. We find that the differences of the outcomes are very small when our empirical estimates are used. Hence, model risk may be considerably reduced.
Keywords: Credit Risk Models, Default Correlations, Basel II
JEL Classification: G20, G28, C51
Suggested Citation: Suggested Citation