Abstract
This study blends the simplicity and empirical success of univariate GARCH processes with an easy to estimate and interpret dynamic correlation estimator. A two step estimator and a simple test are employed to verify the null of constant correlation against an alternative of dynamic conditional correlation. The real strength of the DCC estimation process is its flexibility of univariate GARCH but not the complexity of conventional multivariate GARCH, therefore large correlation matrices can be estimated. One of the primary motivations for this study is that the correlations between assets are not constant through time. The focus of the study is hence to explore the empirical applicability of the multivariate DCC-GARCH model when estimating large conditional covariance matrices. Among the adopted models, DCC-GARCH(1,1)-t can be considered as the best model in measuring VaR, and DCC-GARCH(1,1) can be considered as the second best, while SMA is in the last. The results have suggested that a more complete model which carries more time series characteristics may outperform the others in the sample.