We model 1927-1997 U.S. business failure rates using a time series approach based on unobserved components. Clear evidence is found of cyclical behavior in default rates. The cycle has a period of around 10 years. We also detect longer term movements in default probabilities and default correlations. Our findings have important implications for portfolio credit risk analysis. First, a static analysis of portfolio credit risk can underestimate credit risk significantly by not accounting for the dynamic and cyclical behaviour of default probabilities. Second, estimating default correlations over long horizons without accounting for time variation may lead to misspecified risk management models. We highlight the main effects in an actual credit risk experiment, addressing the issue of pro-cyclicality in ratings and capital buffer formation. It turns out that dynamic models anticipate much better on required capital buffer increases than rating strategies based on recent historical data. In this way, dynamic credit risk models may help to alleviate part of the pro-cyclicality problem.
credit risk pro-cyclicality capital requirements dynamic models common factors credit cycles time varying parameters