Extreme values dependence of risk in Latin American markets

This paper aims to determine which extreme value copula is best suited to the bivariate relationships between shocks of U.S market with Brazilian, Argentine and Mexican markets. We used prices of S&P500, Ibovespa, Merval and IPC from January, 3, 2009 to December, 31, 2010, totaling 483 observations. We estimated Gumbel, Galambos, Husler Reiss and Student's t (extreme-value - TEV) copulas. Results allow concluding that there is a strong dependence on the tails of the joint probability distribution of these markets. Nevertheless, the Gumbel copula for Brazil, Husler-Reiss copula for Argentina and TEV copula for Mexico had the best fit. Thus, its important use an extreme value properly diversify the risk in a portfolio that consist of assets in these countries, especially in turbulence periods.