统计研究

• 论文 • 上一篇    下一篇

中国股票市场的长期记忆性与趋势预测研究

谭政勋 张欠   

  • 出版日期:2016-10-15 发布日期:2016-10-18

Research on the Long-term Memory of Chinese Stock Market and Its Trend Prediction

Tan Zhengxun & Zhang Qian   

  • Online:2016-10-15 Published:2016-10-18

摘要:

本文首次在国内利用较新的精准局部似然函数法(Exact Local Whittle),以上证指数为对象,估计了ARFIMA(p,d,q)模型的长期记忆参数d,并分析了上证指数的趋势性变化。估计结果和稳健性检验均表明,上证指数具有长期记忆性,以上证指数为代表的股票市场并非有效;模拟结果显示,当滚动窗口n=260,带宽m=[n0.65]时,长期记忆参数即估计量d既具备一致性,又具有渐进正态性。在2004年10月8日至2015年11月13日期间,模型给出了8次上涨或下跌的趋势转换信号,其中7次信号是正确的,仅有1次给出了错误信号;股票价格由下跌趋势转为上涨趋势、由上涨趋势转为下跌趋势两种情况相比,记忆参数d在前一种情况时下跌幅度更大,给出的信号更明显。

关键词: ARFIMA(p,d,q)模型, 长记忆性, ELW估计法, 趋势预测

Abstract:

By using of a new Exact Local Whittle estimation for the first time in the domestic, it estimates long memory parameter d of the ARFIMA (p, d, q) model, and analyzes the trend of Shanghai composite index. The estimate results and robustness tests indicate that the Shanghai composite index has a long-term memory and Chinese stock market is not effective. The long memory parameter estimator d is shown to be consistent and asymptotically normally distributed when rolling window n = 260, bandwidth m = [n0.65]. During the period from October 8, 2004 to November 13, 2015, the model gives eight times upward or downward trend conversion signal, seven times is correct, only one error signals is presented. Comparing two kinds of circumstances of stock price from the downward to upward trend and the uptrend into a downward trend, the memory parameter d fell even more sharply in the former case and signals are given more obvious.

Key words: ARFIMA(p,d,q) Model, Long-term Memory ;ELW Estimation ;Trend Prediction