Regime switches in the dependence structure of multidimensional financial data

Misperceptions about extreme dependencies between different financial assets have been an important element of the recent financial crisis, which is why regulating entities do now require financial institutions to account for different behavior under market stress. Such sudden switches in dependence structures are studied using Markov switching regular vine copulas. These copulas allow for asymmetric dependencies and tail dependencies in high dimensional data. Methods for fast maximum likelihood as well as Bayesian inference are developed. The algorithms are validated in simulations and applied to financial data. The results show that regime switches are present in the dependence structure and that regime switching models provide tools for the accurate description of inhomogeneity during times of crisis.