Tail approximations for portfolio credit risk

Any simulation procedure has difficulty achieving accuracy for rare events that lie in the tails of the probability distributions one is simulating from. But that is where the outcomes that produce defaults in a credit portfolio occur, making pricing and risk management for CDOs and similar instruments difficult and (computer) time-consuming. In this article, Glasserman introduces several approximation procedures for estimating the tails of the distribution of default risk exposure for a credit portfolio. The technology appears somewhat daunting, but the results are impressive in terms of accuracy in fitting even the remote tails for a broad range of portfolios.