Аннотация:This study investigates market risk management methods for high dimensional portfolios composed of Russian stocks. We employ a general copula framework that allows for flexible marginal distributions, as well as different types of dependence represented by the copula function. We compare different multivariate distribution assumptions by measuring the Value at Risk for a portfolio composed of thirty Russian stocks and using various back-testing techniques. Our empirical results point out at the importance of marginal leptokurtosis while the type of copulas plays a minor role. Besides, we find that marginal misspecification may result in significant misspecified dependence structures with important consequences on VaR estimation