A comparative analysis of correlation skew modeling techniques for CDO index tranches

In this work we present an analysis of CDO pricing models with a focus on “correlation skew models”. These models are extensions of the classic single factor Gaussian copula and may generate a skew. We consider examples with fat tailed distributions, stochastic and local correlation which generally provide a closer fit to market quotes. We present an additional variation of the stochastic correlation framework using normal inverse Gaussian distributions. The numerical analysis is carried out using a large homogeneous portfolio approximation.