In this paper we use a semiparametric dynamic copula approach to estimate
Value at Risk (VaR) and Expected Shortfall (ES). First, VAR-GARCH models are used in
order to deal with temporal dependence. Second, in contrast to previous works which assume
parametric distribution functions, marginal distributions of the standardized residuals
of the first step models are estimated by a semiparametric approach using empirical
distributions and Extreme Value Theory. Third, copula distributions are estimated on the
basis of uniform random variables obtained in the second step. Finally, VaR and ES are
estimated through Monte Carlo simulations. This methodology could be useful for portfolios
with a large number of assets, because the identification process of the univariate
marginal distributions is easier, and additionally, this methodology reduces the model risk
derived from selecting an ex-ante specific distribution.
As an application of this methodology, VaR and ES are estimated for