中国股市波动率的广义周内特征及其预测模型

A volatility model for Chinese stock market with generalized day-of-the-week effect

本文将Realized GARCH模型推广至基于周历日的时变参数情形以刻画杠杆和溢出效应的周内特征并避免传统GARCH类模型在拟合长记忆性与周内效应时两者相互干扰问题.将新模型应用于上海股票市场2001至2013数据的分析发现:我国股市波动率存在时变的杠杆效应和溢出效应.实证结果表明:新模型无论在样本外的预测能力还是在样本内的拟合度上都明显优于现有模型.

This paper extends the realized GARCH model to allow time varying parameters for leverage effect and spillover effects. The extended model not only fit daily volatility well, but also distinguish the day-of-the-week effect from the long memory volatility. Fitting proposed model to a sample of high- frequency data from Shanghai Composite Index （SSEC） from 2001 to 2013, we find the volatility in Chinese stock market has the time varying leverage and spillover effects. Finally, both the in sample fitness and out of sample predictability to the existing models, we find that the proposed model compares favorably.