Conditional Distribution of the Colombian Exchange Rate Returns : An Empirical Exercise Using Multivariate GARCH Models

A set of multivariate GARCH models is estimated and its empirical validity is compared from the calculation of the Value at Risk. Data used are the daily returns of the nominal exchange rate of the Colombian peso vis-a-vis the American dollar, euro, sterling and Japanese yen for the period 1999-2005. The comparison of the estimations for the conditional covariance matrix and the results obtained for the proportion of failure and the dynamic quantile test of Engle and Manganelli (2004), show evidence in favor of the model of Conditional Constant Correlation