Certain macroeconomic and political events often have a similar and simultaneous impact on companies belonging to the same industry sector. Prices of portfolio credit derivatives are driven by the obligors' dependence structure and their marginal default probabilities. Thus, a common approach to evaluate default models is to assess their ability to replicate market prices of so-called collateralized debt obligations (CDOs). [21] show that a sectordependent credit default model with default triggers having a nested Archimedean copula, considerably reduces pricing errors for these portfolio credit derivatives compared to an exchangeable Archimedean copula. Multivariate default is often modeled by a conditional independence approach due to computational simplicity. Particularly, given a stochastic market intensity, the lifetimes of the respective entities are independent. This structure might be extended by another hierarchy when introducing sector intensities. The present master thesis analyses a specific model exhibiting the extended structure - the hierarchical Lévy-frailty default model. Thereby, the survival copula of the joint vector of default times is given by a hierarchical Lévy-frailty copula. The particular structure of the hierarchical copula is determined by the linkage of the intensity processes. We and that the hierarchical Lévy-frailty model, based on two distinct constructions, reproduces market quotes of CDOs more accurately than the basic Lévy-frailty model and the nested Archimedean model for a given set of historical market quotes.
«
Certain macroeconomic and political events often have a similar and simultaneous impact on companies belonging to the same industry sector. Prices of portfolio credit derivatives are driven by the obligors' dependence structure and their marginal default probabilities. Thus, a common approach to evaluate default models is to assess their ability to replicate market prices of so-called collateralized debt obligations (CDOs). [21] show that a sectordependent credit default model with default trig...
»