基于不同记忆性异方差模型的中国股票市场波动率预测  

Comparison of Forecasting Ability of Heteroskedastic Volatility Models with Different Memory

作　　者：王鹏[1] 吕永健[2] 

机构地区：[1]西南财经大学中国金融研究中心,四川成都610074 [2]西南财经大学金融学院,四川成都611130

出　　处：《中国管理科学》2013年第S1期270-275,共6页Chinese Journal of Management Science

摘　　要：以中国股票市场最具代表性指数——上证综指的高频数据样本为例,通过构建具有不同记忆性的异方差模型并采用样本外的滚动时间窗(rolling time windows)预测法,计算了不同记忆性异方差模型对未来指数波动率的预测值,然后进一步运用具有bootstrap特性的SPA(superior predictive ability)检验法,实证研究了不同记忆性异方差模型对中国股票市场真实波动率的预测精度差异。实证结果显示:就中国股市而言,短记忆异方差模型的预测精度较长记忆模型更为优异,且SKST分布可以对中国股市收益分布所展现出的有偏(skewed)和尖峰胖尾(leptokurtic and fat tailed)等典型特征提供全面综合刻画。Take the most important stock index in Chinese stock market as sample,volatility predicting results are computed based on volatility models with different memory and out- of- sample rolling time windows method.Using bootstrapping SPA test,the predicting performance of different volatility models are compared.The empirical results show that,considering the skewed,leptokurtic,and fat tailed distribution in Chinese stock market,there is no evidence that volatility models with long memory outperform the models with short memory when volatility prediction accuracy is concerned.

关 键 词：波动预测 实现波动率 滚动时间窗 SPA检验 

分 类 号：F832.51[经济管理—金融学] F224.0