Forecasting gains of robust realized variance estimators: evidence from European stock markets

The classical realized variance (RV) estimator is biased due to microstructure effects and asset price jumps. Robust realized variance (RRV) estimators adjust for these biases, and make more efficient of use of the intraday data. This article examines the benefits of using RRV estimators instead of the RV estimator, in the context of volatility forecasting. The recently proposed Realized GARCH framework is used to generate daily forecasts of the conditional variance for eight European stock indices. The out-of-sample comparisons indicate that the RRV estimators improve upon the RV estimator on efficiency and bias criteria.