基于状态空间HAR-RV-RS模型的中国股市波动率预测

Predicting the Volatility of Chinese Stock Markets Based on State Space HAR-RV-RS Model

文章以高频数据下异质自回归(HAR)模型为基础,将模型的部分参数进行时变化处理,利用卡尔曼滤波方法进行参数估计,并引入新的要素——已实现的偏度(RS)作为新解释变量,构造状态空间HAR-RV-RS模型,进行波动率预测研究。文章分别选取了上证综指和深证成指3410个交易日的五分钟高频数据,通过样本内参数估计和样本外滚动时间窗预测,结合损失评价函数与HAR-RV、状态空间HAR-RV进行模型比较。实证研究表明,参数时变化和加入RS后的状态空间HAR-RV-RS模型预测效果有显著提升。

Based on the heterogeneous autoregressive(HAR)model in high frequency data,this paper uses Kalman filter to estimate some parameters after time-varying processing,and introduces a new factor,the realized skewness(RS),as a new explanatory variable to construct the state space HAR-RV-RS model for volatility prediction.Five-minute high-frequency data of 3410 trading days of Shanghai Composite Index and Shenzhen Composite Index are selected in this paper.The model is compared with HAR-RV model and state space HAR-RV model by estimating the parameters in the sample and predicting the rolling time window outside the sample.Empirical research shows that the prediction effect of HAR-RV-RS model with time-varying parameters and RS is significantly improved.