Application Of High Frequency Data In Estimation Of Risk Minimization Hedging Ratio

This paper proposes a new class of time-varying Copula-Realized GARCH models,which is combined of the Realized GARCH model and the time-varying Copula model.This paper use the new model to predict the hedging ratio which can minimize the risk of hedging portfolio.Based on the CSI 300 index futures and spots high frequency data,then the effectiveness of hedging ratio obtained by different models was compared and analyzed.In the construction of RG structure,this paper not only considers nine different high frequency realized estimators and five different Copula functions with different tail sensitivity.Furthermore,the effect of the introduction of high frequency information in the Copula function is discussed and analyzed.In the empirical analysis,this paper carries out in-sample and out-of-sample prediction on the entire sample range of the CSI 300 stock index futures and spots,and adopts the variance reduction ratio HE,the realized variance reduction ratio HE^RV and the tracking error volatility TEV as the evaluation indexes of the model hedging ratio prediction efficiency.DM test and MCS test were used to compare their scores with those of other common models.In order to test the robustness of the model,the samples were divided into high-volatility sub-samples and low-volatility sub-samples by the NPCPM method in this paper,so the performance of the model under different market fluctuation conditions was further discussed.At the same time,it analyzes the empirical results of the S&P 500 index futures and spots data in the US market and discusses the performance of the model in different market structures.Our empirical examination on Chinese market and American market shows that,1)models with high-frequency information can significantly improve the hedging effectiveness over the traditional Copula-GJR models;2)Incorporating the realized measures through the RG structure achieves higher risk reduction than simply including them as exogenous variable in the marginal distribution（TVM Copula-GJRX）;3)Including the realized correlation measures in the dynamic copulas can improve the performance of mos T-Copula-RG type models.In high fluctuation range,the hedging performance of TVM-Copula-RG class model can be significantily improved by selecting tail-sensitive Copula;4)In terms of the selection of realized estimators,the RK with five-minute sampling which is robust to noise is obviously superior to other realized estimators;5)In the setting and selection of the Copula model.Clayton-Copula,which is sensitive to the left tail,significantly improves the effectiveness of hedging in the Chinese market and reduces the volatility of tracking errors.The SJC-copula,which is sensitive to both the top and bottom,is more prominent in the US market;6)The TVM Clayton（RC）-RG（RK）model that includes the realized kernel proves to be the superior hedging method in China’s market,especially during market fluctuations.