中国股市长期记忆效应的实证研究

Empirical Study on the Long-memory Effect in Chinese Stock Market

股票市场长期记忆效应问题一直是金融经济学家们倍感兴趣的一个研究热点。本文针对中国股票市场中价格指数与个股的日收益序列 ,在已有研究文献主要采用的经典R S分析方法基础上 ,引入修正R S分析与ARFIMA模型进行了实证研究。从研究结果来看 ,2 2个样本序列并不满足传统的正态分布假设 ,序列呈现出尖峰、肥尾、右偏等有偏特征以及独特的自相关与偏自相关结构 ,这些迹象预示着非线性动态系统的存在。而进一步的研究却表明 ,中国股市中代表市场总体的股价指数不存在长期记忆效应 ,而个股收益序列的分布特征存在着较大差异 ,仅少数个股存在长期记忆行为。这一结论明显地有别于以往那些由经典R S分析所得到的研究结果。

The long-memory effect in stock market is a hotspot for financial economists. Based On the literatures using classical R/S analysis, this paper has introduced the modified R/S analysis and ARFIMA model and puts forward to study the returns distribution in Chinese stock market. Twenty-two logarithm return rate series including stock price indices and corporate stocks are studied in this investigation. The empirical result makes clear that these series are non-normality with positive skewness, leptokurtosis, fat tails and special autocorrelation and partial autocorrelation structure. But further investigation shows that no evidence of fractal structure is found in the stock indices, so the aggregate market has no long-memory. The results highlight the differences in fractal structure across different companies' series, and show that only several have significantly long-memory effect. This conclusion is different from other studies using classical R/S analysis obviously.