On the characterization and information contents of dynamic default correlation under the doubly stochastic assumption: the case of iTraxx CDO tranches

Abstract

In this paper we present a tractable model that dynamically characterizes the correlation structure for a portfolio of credit entities. Heterogeneous credit events are modeled as the arrivals of mixed Poisson jump processes and magnitudes of jumps represent the associated impacts of credit events on the joint survival probabilities of the credit portfolio. We assume that the default intensities of both sources of risk are driven by a combination of two parameter Gamma and Pareto distributions as to capture the clustering effects of default events under different market scenarios. We conduct calibration of the model to iTraxx Europe as an example and verify the goodness of fit between the arket spreads and the model spreads of ours. We extract the implied jump-sizes that reveal information on the correlation structure, and further explore their impacts on the risk characteristics of CDO tranches.