Modelling credit default swap spreads by means of normal mixtures and copulas

This paper develops a multivariate statistical model for the analysis of credit default swap spreads. Given the large excess kurtosis of the univariate marginal distributions, it is proposed to model them by means of a mixture of distributions. However, the multivariate extension of this methodology is numerically difficult, so that copulas are used to capture the structure of dependence of the data. It is shown how to estimate the parameters of the marginal distributions via the EM algorithm; then the parameters of the copula are estimated and standard errors computed through the nonparametric bootstrap. An application to credit default swap spreads of some European reference entities and extensive simulation results confirm the effectiveness of the method.