考虑跳跃和知情交易概率的高频波动率研究 ----基于CSI300股指期货实证分析

his paper selects 5-minute high-frequency data of CSI300 stock index futures, builds five HAR class models to explore the fluctuation law of CSI300 stock index futures based on indicators such as up and down volatility, sign jump and other indicators as well as the probability of informed trading. The data of May 19, 2010 solstice and February 12, 2017 are selected as the in-sample data. The one-step rolling method was used for the out-of-sample prediction of each volatility model to analyze whether the model with the probability index of informed trading could enhance the fitting ability and prediction effect of the model compared with the benchmark model, and the MCS test was carried out on the model. The results show that the volatility of stock index futures has the characteristics of long memory and sharp peak and thick tail, the good and bad volatility of rising and falling has asymmetric impact on the future volatility, and the sign jump part has the opposite effect on the volatility of a period in the future. The decomposition of the rising and falling semivariances is more explanatory than the decomposition of the continuous and jumping parts. The probability of informed trading can explain the prediction of future volatility and has a good risk warning ability in the case of abnormal fluctuation. MedRTQ, which is used to adjust the coefficient to correct the noise bias, can significantly improve the prediction accuracy of the full sample interval of HARQ models. MCS test showed that HARQ-RS-VPIN model performed best in the medium term.