Modeling the Dependency Structure of Stock Index Returns using a Copula Function Approach

In the present study we assess the dependency structure between stock indexes by econometrically estimating the empirical copula function and the parameters of various parametric copula functions. The main finding is that the t-copula and the Gumbel-Clayton mixture copula are the most appropriate copula functions to capture the dependency structure of two financial return series. With the dependency structure given by the estimated copula functions we quantify the efficient portfolio frontier using as a risk measure CVaR (Conditional VaR) computed by Monte Carlo simulation. We find that in the case of using normal distributions for modeling individual returns the market risk is underestimated no mater what copula function is employed to capture the dependency structure. 

Keywords: copula functions, copula mixtures, the efficient portfolio frontier, Conditional VAR, Monte Carlo simulation
JEL Classification: C51, C52, G10, G11