Computation of the Loss Distribution Based on the Structural Model for Credit Portfolios

Speaker: Ken Jackson

Affiliation: University of Toronto

Credit risk analysis and management at the portfolio level is a challenging issue for financial institutions due to their portfolios’ large size, heterogeneity and complex correlation structure.  We have explored both asymptotic approximation methods and fast exact methods to compute the distribution of a loan portfolio’s loss in the CreditMatrics framework.

In this talk, we focus on the fast exact methods, which improve the efficiency by exploiting the sparsity that often arises in the obligors’ conditional losses. A sparse convolution method and a sparse FFT method are proposed, which enjoy significant speedups compared with the straightforward convolution method. We also construct truncated versions of the sparse convolution method and the sparse FFT method to further improve their efficiency. To balance the aliasing errors and round off errors incurred in the truncated sparse FFT method, an optimal exponential windowing approach is developed as well.