Self-exciting corporate defaults: contagion or frailty?

Kay Giesecke (Stanford)

The ongoing credit crisis highlights the need for portfolio credit risk estimates that account for the feedback from credit events. We introduce and estimate from U.S. corporate default data spanning 1970 to 2006 a new model of self-exciting defaults that incorporates the feedback from events to arrival rates. Our filtered point process likelihood estimators indicate that a default has a significant contagious impact on the surviving firms, after controlling for firms' exposure to a common Feller diffusion risk factor that may be a frailty. Contagion is found to be an important source of the default clustering in the data. Goodness-of-fit tests indicate the statistical importance of incorporating the effects of contagion into a model of correlated default timing. These findings have significant implications for the risk management of corporate debt portfolios and the risk analysis of securities exposed to correlated default risk, such as collateralized debt obligations.