基于跳跃、好坏波动率的混频已实现EGARCH模型的波动率预测与风险度量

为探究资产价格的跳跃行为和收益波动的非对称效应对波动率预测的影响，以高频数据建模为视角，基于跳跃、好坏波动率将Realized EGARCH-MIDAS模型进行拓展，以提升模型的波动率预测能力与风险度量效果。运用拓展后的模型，以沪深300指数价格高频数据为样本进行实证分析，探究中国股票市场的波动性规律，并采用似然函数、信息准则和基于损失函数的DM与MCS等检验方法，综合比较了改进前后的模型对波动率及风险值的预测效果。实证结果显示：(1)沪深300指数收益的长期波动主要来源于连续波动而非跳跃波动，且受正连续波动影响更大，而负跳跃对波动具有明显的负向冲击；(2)文章提出的拓展模型均能更好地捕捉波动率的长记忆性，在样本内估计和样本外预测上也都有更好的表现，其中同时考虑跳跃与非对称影响的Realized EGARCH-MIDAS-RSJ拓展模型拥有最优的估计及预测效果。

Realized EGARCH-MIDAS Model Based on Jumps and Good-badVolatility with Its Volatility and VaR Forecasting

To study the impact of jumps and asymmetric effects on volatility forecasting and further improve the risk measurement performance, the Realized EGARCH-MIDAS model is extended by introducing the decomposition of the realized volatility into the long-term volatility equation. Specifically, the decomposition includes the continuous-jump volatility, the good-bad volatility, and the positive-negative jump volatility with positive-negative continuous volatility. Then, high-frequency data from the CSI 300 is used for empirical analysis to explore the volatility pattern of the Chinese stock market. Different statistical and economic criteria, including the likelihood function, information criteria, and loss-functions-based tests (DM and MCS tests), are applied to assess the performance of proposed models in both volatility modeling and risk measurement. The empirical results show that: 1) The long-term volatility of the CSI 300 mainly comes from the continuous component rather than the jump component and is more affected by the positive continuous component, while the negative jump has a significant negative impact on the volatility. 2) All improved models can better capture the long-memory of volatility and thus perform better in both the in-sample fitting and out-of-sample forecasting of volatility and risk, with the REGARCH-MIDAS-RSJ considering both jumps and asymmetries being the optimal model.