摘要：In this report we examine time-varying correlations of asset returns using the Dynamic Conditional Correlation (DCC) models, recently proposed by Engle (2002), that are estimated by a two-step procedure. First, we conclude that correlations vary considerably over time. Secondly, the conditional correlations exhibit significantly asymmetric effects for positive and negative asset return shocks. These asymmetric effects differ between stocks and bonds however. Thirdly, the loss of efficiency by using the two-step procedure is relatively large for the standard DCC model, but this procedure reduces the computational burden for the extended specifications. Finally, we compared the standard DCC model to other multi-variate GARCH models. The DCC model seems to outperform the alternatives.
