Optimal Simultaneous Validation Tests of Default Probabilities, Dependencies, and Credit Risk Models
by Uwe Wehrspohn of Heidelberg University
July 15, 2004
Abstract: We provide a general, model-independent approach to the construction of optimal simultaneous validation tests of credit default probabilities, dependencies between creditors, and credit risk models that maximize the power of test for any given portfolio-size and number of periods of data available. Results can be used to validate banks' estimates of rating default probabilities, correlations and choice of credit risk models in the Basel II supervisory review process. Example-analyses are given for the generalized asset value model.
Keywords: rating validation, backtesting, model validation, banking regulation.
Published in: Journal of Risk, Vol. 5, No. 4, (Summer 2003), pp. 1-23.