CFE 2019: Start Registration
View Submission - CMStatistics
Title: Dynamic regular vine copulas with an application to exchange rates dependence Authors:  Alexander Kreuzer - Technische Universität München (Germany) [presenting]
Abstract: Modeling dependence among financial assets is an important research topic, since the dependence structure has high influence on the risk associated with a corresponding portfolio. Regular vine copulas have proven as a useful tool in this context. They allow for characteristics like asymmetric tail dependence, which cannot be modeled with a multivariate Gaussian or Student-$t$ copula. Usually it is assumed that the dependence parameters of the regular vine copula remain constant as time evolves. We get rid of this assumption and propose dynamic regular vine copulas. In this dynamic model dependence parameters are described through latent AR(1) processes. Since maximum likelihood estimation is infeasible for these latent AR(1) processes, we employ Markov Chain Monte Carlo within a sequential estimation procedure. The approach is illustrated with 25-dimensional exchange rates data, where we find clear evidence for dynamic dependence.