Estimating Pair-Copula Constructions Using Empirical Tail Dependence Functions: an Application to Russian Stock Market

Tail dependence plays important role in portfolio optimization. The higher tail dependence among assets, the higher risk of simultaneous high loss in their prices. In this paper the choice of pair-copulas in pair-copula construction model is done by minimizing the distances between theoretical and empirical tail dependence functions. This method is believed to provide better approximation of tails of joint distribution (compared to maximal spanning tree methods), yet hold all advantages of pair-copula constructions as models of multivariate dependence.