金融时间序列记忆模式识别与实证基于方差标度指数-数量经济学专业论文.docxVIP

II II 摘 要 格变动会产生持久的正相关性的影响，美国股指序列会表现出对冲击的一定抵 抗力。因此，对中美股指序列的记忆模式识别与实证的结果，也可以反映出中 国股票市场的非有效性，美国股票市场的有效性较高。 关键词：金融时间序列；长记忆；记忆模式；方差标度指数 AB ABSTRACT III III ABSTRACT The long memory of financial time series has been the most active and most controversial issue in nearly three decades on financial academic research. The modern financial theory based on the efficient market hypothesis and the random walk theory gave the point that financial markets are efficient, the time series of financial price volatility is a white noise process, at most is only a short memory pro- cess. The new financial theory which bases on the fractal market hypothesis, hold the basic views that financial markets has long memory. The most significant feature is that autocovariance functions and autocorrelation coefficients of time series both show same trends of slowly decaying, the phenomenon known as long memory. Whether there is long memory in financial markets, kinds of long memory identification or test methods can not reach a unanimous conclusion. On the issue of memory pattern, the methods of R/S, DFA and GPH tend to conclude that there is long memory in financial market, Due to that these methods do not take into account the impact of short memory. In contrast, considering short memory effect, the test methods of MR/S and V/S statistics which weighting large-lag order autocorrelation coefficients may go to another extremeness and both tend to reject the assumption that there is no long memory in financial markets. The fundamental cause of the differences of the two research paths is that these methods have not been able to establish valid mathematical relationships and can not efficiently identify short memory and long memory, also result in the controversy. The innovation of this paper is that based on views of theoretical deduction, I advance a memory pattern hypothesis that there is white noise, short memory and long memory in financial markets, construct two statistics to make statistic