Abstract
This paper discusses the practical aspects of modeling the structure of dependence of national stock indices based on copula functions. An approach is proposed to identify groups of stock markets with homogeneous dynamics and to select typical representatives within these groups. It can be used as a preliminary step to build multivariate copula models as it rationalizes the criteria and principles for selection of national stock indices. Constructed model of the joint distribution of indices is used to estimate the probability of crisis events and to demonstrate the benefits of diversification.