Predicting drastic drop in Chinese stock market with local Hurst exponent

In this paper we employ detrended fluctuation analysis (DFA) technique to estimate the local Hurst exponent of Shanghai stock exchange composite (SSEC) index, which is used to predict the crash or other drastic decreases in Chinese stock market. We assume the local Hurst exponent can be used to as a measurement of actual excitation or nervous state of the market before the displaying of market index. The empirical result shows that a very clear decreasing trend from greater than 0.5 to less than 0.5 in local Hurst exponent is visible, preceding the market drastic drop point. In-sample vs. out-of-sample tests confirm above viewpoint, showing that local Hurst exponent below 0.5 can be used as an important signal of stock return when stock index is going to face drastic drop, and it has more significant predict power than historical average return forecasting method. So we find efficient market hypothesis (EMH) doesn´t receive support in Chinese stock market, and explore a new method for investment beyond the traditional technical analysis.