Samuel Malone, Abel Rodriguez and Enrique ter Horst
04/09/2009 09:00 AM
Applied Mathematics & Statistics
We develop a structural credit risk model in which the asset volatility of the firm follows a GARCH process, as in Heston and Nandi (2000). We benchmark the out-of-sample model prediction accuracy against the calibrated Merton (1974) model and the Duan (1994) ML estimation of the Merton model, both using simulated data and an empirical application to the bank CDS market in the US during the crisis period 2007-2008. The GARCH model outperforms the competitors in out-of-sample spread prediction in both cases. We document a high incidence of empirical bank CDS spread term structure inversion, and analyze its relationship with model performance.