In this paper we compared multivariate GARCH models to estimate Value-at-Risk. We used a portfolio of weekly indexes including TEDPIX, KLSE, XU100 during ten years. To estimate Value-at-Risk, first we estimated CCC, DCC of Engle, DCC of Tse and Tsui, Dynamic Equi correlation models by OxMetrics. Then, optimum lags were estimated by minimizing the information criteria. To estimate VaR, the models accuracy was validated by using variance-covariance matrix. The results show that although CCC model estimates variance matrix better, Dynamic Equi correlation is preferable to estimate Value-at-Risk, employing more complete correlation matrix.
