Heterogeneous Archimedean copula and t-copula with application in credit portfolio modeling

Besides their advantage in modelling tail-dependency, the main drawback of standard non-Gaussian copula is the homogeneity in the tail dependency parameter. Several approaches to solve this are meanwhile developed, hierachical copula, the grouped t-copula and the heterogeneous t-copula as recently described by Luo and Shevchenko [1]. We will show results from a concrete implementation of a factor model using the later approach based on a two step estimation procedure. In particular the eects on capital allocation will be highlighted. We will also present how this can be extended to a wide class of Archimedean copula, in order to capture heterogeneous tail-dependencies and therefore tail-sensitive capital allocation in credit portfolio models. The rst part is joint work in progress with Carsten Binnenhei, Melanie Frick and Benedikt Mankel (Deka Bank)