股票市场的长记忆及其动态结构

Long-Term Memory in Stock Market and Its Dynamic Structure

本文提出了结合平均小波系数法和自回归原始自助法的稳健长记忆检验,蒙特卡罗模拟显示该方法对于短期记忆过程具有稳定性。基于该方法对2005年4月8日至2015年6月30日的中国、美国、香港和德国股市进行了实证分析。全局检验结果表明仅中国的股票市场存在显著的长记忆,并且风险因素无法对长记忆解释,而美国、德国和香港的股市不存在长记忆。基于递增窗口的动态Hurst指数分析显示,金融危机时期4个股市都存在显著的长记忆。2010年后,除中国股市外,其余三个股市几乎不存在长记忆现象。

In this paper we propose robust long-term memory test by using both Average Wavelet Co- efficient method and Autoregressive Wild Bootstrap. Monte Carlo simulation shows that our method is robust for short-term memory process. The robust test is used to empirical analyze daily returns of stock markets, including China, America, Germany, and Hong Kong, from April 8, 2005 to June 30, 2015. The global test shows that only Chinese stock market existed significant long-term memory, and risk factors can＇t explain it, whereas stock markets of America, Germany and Hong Kong didn＇t exist long-term memory. The dynamic Hurst analysis Based on increasing window gives information that in financial crisis period, the 4 stock markets existed significant long-term memory. After 2010, stock markets of America, Germany and Hong Kong almost did not exist long-term memory except Chinese stock market.