• •

### 基于门槛效应的Realized MAT-HAR GARCH模型波动预测

1. 浙江工商大学统计与数学学院, 杭州 310018
• 出版日期:2021-06-25 发布日期:2021-09-17

CAI Guanghui, WU Zhimin. Volatility Forecasting of the Realized MAT-HAR GARCH Model Based on Threshold Effects[J]. Journal of Systems Science and Mathematical Sciences, 2021, 41(6): 1548-1571.

### Volatility Forecasting of the Realized MAT-HAR GARCH Model Based on Threshold Effects

CAI Guanghui， WU Zhimin

1. School of Statistics and Mathematics, Zhejiang Gongshang University, Hangzhou 310018
• Online:2021-06-25 Published:2021-09-17

Based on the perspective of threshold effects, combining with the characteristic of the MAT-HAR model capturing the threshold effects and the advantage of Realized GARCH model in high-frequency volatility prediction, this paper establishes the Realized MAT-HAR GARCH model to study the threshold effects of high-frequency financial market. Different from the Realized HAR GARCH model, this model introduces MAT-HAR structure into the volatility equation, takes a moving average of the lagged realized volatility series as the threshold, and selects the optimal structure of volatility equation based on AIC information criterion. The numerical simulation results show that the Realized MAT-HAR GARCH model has the good effects on parameter estimation and volatility prediction. The empirical study based on the CSI 300 index shows that under the assumption that the residuals obey normal distribution, significant threshold effects exist at the weekly and monthly level, and the Realized MAT-HAR GARCH model based on AIC criterion performs better on in-sample fitting, volatility forecasting and VaR measurement than Realized HAR GARCH model. Finally, the rationality of the model structure selected based on AIC information criterion and the robustness of volatility prediction are studied under the assumption of different error distributions. The results of robustness test based on MCS test show that under the assumption of different heavy-tailed distributions, the out-of-sample performance of Realized MAT-HAR GARCH model performs better than that of Realized HAR GARCH model, which does't depend on the length of rolling windows.
()
 No related articles found!