This paper presents two applications of Extreme Value Theory (EVT) to financial markets: computation of value at risk and assets returns dependence under extreme events (i.e. tail dependence). We use a sample comprised of the United States, Europe, Asia, and Latin America. Our main findings are the following. First, on average, EVT gives the most accurate estimates of value at risk. Second, tail dependence decreases when filtering out heteroscedasticity and serial correlation by multivariate GARCH models. Both findings are in agreement with previous research in this area for other financial markets.
JEL classification: C22, G10.