This thesis aims at investigating a dynamic portfolio default model. Unlike most intensity based models where the default event is driven by a Poisson process, an obligor defaults in this model if a Lévy subordinator exceeds a critical barrier. Under the assumption of a large and homogeneous portfolio and, if the Lévy process satisfies the jump constraint, the computation of the expected loss and the pricing of any tranches are possible. The dependence among the default times in the portfolio is explained by the Cuadras-Augé family of copulas. We consider different subordinators, for example: the compound Poisson process (with exponentially, chi square, Poisson ... distributed jumps), the gamma and the inverse Gaussian process. For each process we implement the model and calibrate it to the 5-year CDO prices. Moreover, we improve the calibration results by changing from decreasing to increasing default intensities. Thereafter, we simultaneously calibrate five, seven and ten year CDO market quotes.     «

This thesis aims at investigating a dynamic portfolio default model. Unlike most intensity based models where the default event is driven by a Poisson process, an obligor defaults in this model if a Lévy subordinator exceeds a critical barrier. Under the assumption of a large and homogeneous portfolio and, if the Lévy process satisfies the jump constraint, the computation of the expected loss and the pricing of any tranches are possible. The dependence among the default times in the portfolio is...     »