Modeling the Dependence Structure of the WIG20 Portfolio Using a Pair-copula Construction

Abstract

Elliptical distributions commonly applied to modeling the returns of stocks in highdimensional  portfolio are not capable of adequate describing the dependence between the components  when their statistical properties are very diverse. The MGARCH and standard dynamic  copula models are often of little usefulness in such cases. In this paper, we apply a methodology  called the pair-copula decomposition to model the joint conditional distribution of the returns on  stocks constituting the WIG20 index, and show some advantage of this construction over the  approach using the t Student DCC model.