Benchmarking Asset Correlations

Abstract

Among the most crucial input parameters for credit portfolio risk models are the co-movements of default risks. Due to limited empirical evidence about the magnitude of correlations the New Basel Capital Accord sets standard requirements for calculating regulatory capital requirements, e.g. in the Consultative Document as of April 2003 asset correlations for sovereigns, banks and corporates between 12% and 24% depending on default probabilities are assumed. The present contribution shows how correlations can be estimated within the framework of the Basel II model. Using Default data from the G7 countries it is shown that asset correlations are much lower than broadly assumed. This may have valuable consequences for backtesting of PD forecasts carried out by banks and supervisors. Furthermore it is shown how current credit risk models can be parameterized with our estimates. We find that the differences between the model outcomes may become even more negligible than found up to now, thus reducing model risk.