摘要

This paper considers two dynamic robust portfolio optimization models based on the framework of Kakouris and Rustem(2014). We use copula-GARCH and DCC copulas approaches to capture the dynamics of the distribution of the returns. We compare our proposed methods with the static robust and nonrobust portfolio optimization models based on the CSI300 data. The experimental study shows that the dynamic WCVaR models perform better in out-of-sample tests when considering the uncertainty in the estimated model. The static nonrobust method produces higher returns in the in-sample tests, since there is no room to capture model uncertainty.