关键词: 混合频率, 已实现波动率, SPA检验, 区块自助法

Abstract: This paper extends Hansen et al. (2012)’s Realized GARCH model to a mixed frequency Realized GARCH model (M-Realized GARCH model), including the data of intraday return, daily return and realized volatility. In our new model, we divide each trading day into two periods, and add mix frequency mean equations. We both build the conditional volatility equation of the residuals in mean equations and realized volatility equation. Mix frequency data of CSI 300 index's during 2013-2016 are analyzed in this paper. Different distributions including normal distribution, t distribution and generalized error distribution are discussed too. The results of Loss function, SPA Test, kupiec Test and Dynamic Quantile Test show that the volatility predict power measure of VaR of M-Realized GARCH model are better than the GARCH, Realized GARCH. Moreover, actual failure rate of VaR is consistent with the theoretical failure rate, and there is no correlation between the two failures occurred. In the end, we use Block bootstrap method to sample mix frequency data, and further prove that the M-Realized GARCH model has higher prediction accuracy than the Realized GARCH model.

Key words: Mix Frequency, Realized Volatility, SPA Test, Block Bootstrap