Importance sampling for integrated market and credit portfolio models

Abstract: Standard credit portfolio models do not model market risk factors, such as risk-free interest rates or credit spreads, as stochastic variables. Various studies have documented that a severe underestimation of economic capital can be the consequence. However, integrating market risk factors into credit portfolio models increases the computational burden of computing credit portfolio risk measures. In this paper, the application of various importance sampling techniques to an integrated market and credit portfolio model are presented and the effectiveness of these approaches is tested by numerical experiments. The main result is that importance sampling can reduce the standard error of the percentile estimators, but it is rather difficult to make statements about when the IS approach is especially effective. Besides, the combination of importance sampling techniques originally developed for pure market risk portfolio models with techniques originally developed for pure default mode credit risk portfolio models is less effective than simpler two step-IS approaches