Importance Sampling for Portfolio Credit Risk in Factor Copula Models

This work considers the problem of the estimation of Value at Risk contributions in a portfolio of credits. Each risk contribution is the conditional expected loss of an obligor, given a large loss of the full portfolio. This rare-event framework makes it difficult to obtain accurate and stable estimations via standard Monte Carlo methods. The factor copula models employed to capture the dependence among obligors, poses an additional challenge to this problem. By conveniently modifying the algorithm introduced by Glasserman and Li (2005), this work develops importance sampling schemes which lead to signifivannt variance reduction, both in single and multi-factor models.