中国股市的长期记忆性实证研究.pdfVIP

中文摘要 本文首先阐述了有效市场假说和分形市场假说的理论背景、主要观点、假 设前提等。然后结合我国股票市场的现实情况进行实证分析。首先对沪市的五 个行业分类指数的日对数收益率AR (1) 的残差进行正态性检验，发现我国股票市 场日收益率AR (1) 的残差不符合正态分布，与有效市场假说所假定的前提条件矛 盾；在此基础上，进一步用分形统计学的相关方法(经典R / S 法)为手段，对我 国沪市的有效性进行实证检验，结果表明：沪市的这五个行业都具有分形结构， 这五个行业指数日收益的赫斯特指数H 分别为：0.6685、0.6332、0.5983、0.6277 和 0.6541 均大于 0.5，长期记忆周期分别为 131、152、170、232 和 131 天。之 后又采用 Lo 修正R / S 分析对这五个行业指数的日对数收益率进行分析，得出 的结论是中国股市并不具有长期记忆性。分析产生这种分歧的原因可能是短期 记忆的干扰导致R / S 分析得出较高的 Hurst 指数，也可能是修正R / S 分析法中 滞后阶数q 的选取不当使得对长记忆不敏感。 关键词：有效市场假说；分形市场假说；R / S 分析；修正R / S 分析；赫斯 特指数；长期记忆性 Abstract This article first demonstrates the theory background, main ideas and theory prerequisites of the Efficient Market Hypothesis and Fractal Market Hypothesis, and then make empirical analysis on our country’ stock market. We test the normal testing of AR (1 ) residual of the five daily indexes. We find that these residual is not in agreement with the normal distribution, which is contradict with the hypothesis precondition of the EMH. Based on this base, we make farther test on the efficiency of Shanghai stock with the classical method: R/S analysis. The results show that the five industries stock market take on the distributing structure, and the Hurst indexes of the five industries are 0.6685, 0.6332, 0.5983, 0.6277 and 0.6541 respectively, which are bigger than 0.5; and the cycle of the long-run memory are 31, 152, 170, 232 and 131 days respectively. However, applying Lo modified R / S to test these five industries daily indexes of interest rate, the results show that five industries stock is not accor