HAR-RV-EMD-J模型及其对金融资产波动率的预测研究

The HAR-RV-EMD-J Model and Its Application to Forecasting the Volatility of Financial Assets

本文在HAR-RV模型的基础上,运用EMD等方法将模型中的已实现波动率分解成高频已实现波动率、低频已实现波动率和趋势已实现波动率,并加入跳跃波动率成分,构建HARRV-EMD-J模型;接着,以沪深300股指和沪深300股指期货的5分钟高频交易数据为实证样本,对HAR-RV-EMD-J模型以及常见的四个HAR类波动率模型进行样本内分析和样本外分析,并对其分析结果进行稳健性检验。研究发现:在HAR-RV-EMD-J模型中,高频已实现波动率和低频已实现波动率包含对未来1日、1周、2周和1月波动率的预测信息较多,而趋势已实现波动率和跳跃波动率包含的预测信息较少;HAR-RV-EMD-J模型对未来1日、1周、2周和1月波动率的样本内和样本外预测能力都明显强于其他四个HAR类波动率模型。

On the basis of the HAR-RV model,we develop a new HAR-type volatility model（ i.e.,HAR-RV-EMD-J model）. Then we use 5-min high-frequency transaction data of the CSI 300 stock index and CSI 300 stock index futures as the study sample,and respectively analyzes the HAR-RV-EMD-J and the other four HAR-type volatility models. The results indicate that the high-frequency realized volatility,low-frequency realized volatility can be used to predict the future 1 day,1 week,2 week and 1 month volatilities,while the trend realized volatility and jump volatility are poor at its prediction accuracy. Besides,the HAR-RV-EMD-J model shows obviously better forecasting performance than the other HAR-type volatility models on forecasting the future volatilities of the CSI 300 stock index and CSI 300 stock index futures.