The Return's Predictability Of China's Stock Market

The predictability of the financial assets' return is one of the earliest and most lasting theses in financial econometrics.The return's predictability of China's stock market has been researched with many kinds of methods by domestic economists. According to their different methods and sample of study, they have drawn different conclusions. But almost all these researches either chose the market index or some listed companies randomly as their objects instead of the submarkets grouped according to some ratios. In other words, these researches had some weakness. The existed studies carried on by foreign economists have shown that the companies' size, turnover rate, trading volume, P/E ratio and other variables could effect the predictability of return and the fluctuation of stock's price.For example the three factors asset pricing model of Fama and French (1993) tells us undoubtedly that some factors have distinct explanation to the change of stock's average income besides the factors reflected in traditional asset pricing model. These "new" factors include the company' size - the small company's stock displays better than the big company's; P/E ratio - the stock with high P/E doesn't displays as well as the stock with low P/E; leverage - the stock with high leverage displays better than the stock with low leverage; B/M - the stock with high B/M displays better than the stock with low B/M.In view of this, we group all the companies listed in Shanghai stock market during 1994 to 2000 according to their turnover rate, circulatable stock market value proportion to total market value, circulatable stock market value and B/M every year, and then test the return's predictability on the submarkets using daily return from January 9, 1995 to December 31, 2001.There are two kinds of methods that we adopt in this paper. The first is variance ratio of Lo Mackinlay (1988) which is under the traditional financial theory that is conditional on normal distribution and linear analysis. The second is Hurst exponent conditional on non-linear analysis whose prevailed could be attributed to lineranalysis' failure in explaining our complex world. It is the thing of recent not more than 20 years to introduce the non-linear analytical method into the capital market analysis. According to the materials that the author is exposed to, this method is used in home just from the beginning of this century.Under the first method, except the fifth group grouped by the proportion of circulatable stocks and the first group grouped by B/M, show distinct predictability of return during the first-half period (January 9, 1995 - June 19, 1998), the other groups can't refuse the hypothesis of return's random walk no matter during the whole or the first-half or the last-half period. In other words, we have no ability to predict the change of return in these submarkets during that time. However, Under the second method, all Hurst exponent don't equal to 0.5, that is to say that at all the submarkets we are greatly likely to predict the change of return to gain profit. We also find that the return has long term persistence and memory about 160 days, namely today's price or news can exert an influence on the price within 160 days in the future, or if today's income increased (reduce) relative to yesterday's, then tomorrow's income will increase (reduces ) compare today's most probably. After 160th day, the income demonstrates against persistence, namely if today's income increases (reduce) relative to yesterday, then tomorrow's income will reduction (increases ) compare today's most probably.We attribute the different results of the two methods to the different prerequisite. Under the first method, the investors are all " rational" and they will response to information linearly. We notice that we use variance-the standard measurement of risk put forward by Markowitz (1952) with high frequency. We must noticed that this standard is set up under linear frame, which implying the securities that this standard measures should comply with effective market assumption,...