Abstract

Comparative Analysis of Alternative Credit Risk Models

Various models have been developed in recent years for quantifying banks’ default risks with the portfolio effects of such risks being taken into account. So far, no approach has been able to establish itself as the generally accepted standard. Since the models show fundamental conceptual differences for using different empirical input data, the credit-risk model choice may have a considerable impact on banks’ credit portfolio management. The purpose of this contribution therefore is to clarify whether models permit to calculate systematically deviating value-at-risk figures and, if so, what the origins of such deviations are. This contribution initially shows that the existing credit risk models may be divided into two categories: asset value-based models and default rate-based models. On the basis of a model portfolio of loans to German construction firms, the protagonists of the two model classes (CreditMetrics and CreditRisk*) are compared estimating the effects of the differing empirical input parameters on risk results. The analysis shows substantial differences between the models. However, an examination of the reasons explaining the deviations show that the wide value-at-risk variations primarily stem from the differences in the empirical input data leading to different assumptions for implied correlations. This contribution demonstrated how to choose model parameters to generate identical correlations. The results of both models are largely congruent where the correlation assumptions are consistent