基于上证综合指数和深证成份指数, 文章将广义已实现测度引入 ARFIMA-Realized GARCH 模型, 同时考虑已实现方差、已实现极差、已实现双幂次变差和已实现极差双幂次变差, 比较不同已实现测度下模型的波动率预测能力和 VaR 度量效果. 实证结果表明:\ ARFIMA-Realized GARCH 模型能够充分捕获波动率的非对称性、长记忆性和持续性等特征; 采用已实现方差的\ ARFIMA-Realized GARCH 模型具有最优的波动率预测能力; 已实现平均绝对离差能够改进模型的拟合效果, 并且引入已实现风险值显著提高了\ ARFIMA-Realized GARCH 模型的\ VaR 预测精度.

Based on Shanghai Composite Index and Shenzhen Component Index, the paper introduces generalized realized measures to the ARFIMA-Realized GARCH model and considers realized variance, realized range-based volatility, realized bipower variance and realized range-based bipower variance at the same time, then volatility forecasting ability and VaR measurement effect of models under different realized measures are compared. The results show that the ARFIMA-Realized GARCH model can fully capture the asymmetry,\ long term memory and persistence of volatility. With realized variance, the ARFIMA-Realized GARCH model has the best volatility forecasting ability. The realized mean absolute deviation can improve the fitting effect of the model, and the introduction of realized value-at-risk significantly improves the VaR forecasting accuracy of ARFIMA-Realized GARCH model.