Empirical Analysis On Pairs Trading In Chinese A-Share Market

With the development of economic social and financial market, the introduction of all kinds of new financial products has been rolling out and makes the investment strategy become increasingly rich. By the end of September 2014, the number of stocks for securities margin trading in China’s A share market have increased to 700, covering 28% of the A-stocks. The academic research on pairs trading based on statistical arbitrage strategy began to has its practical significance.The main purpose of this article is to simulate the pairs trading by using C program and debug the threshold parameter values in the process of simulation, comparing the trading strategies and trade results of stocks in pairs which belong to banking and real estate. The sample data of empirical analysis include 50 stocks’ closing price during 250 days, from January 15,2013 to January 28,2014. Screened by correlation analysis and cointegration test, we get two pairs of stocks-stocks of Bank of Nanjing and Bank of Beijing, stocks of Rise Sun Real Estate Development and Poly Real Estate, whose correlation coefficient is 0.901 and 0.848, respectively. The ratio of each pair of stocks is based on the Error Correction Model and the goal of threshold parameter value test is to maximize the annualized return. The data out of sample include 100 days’closing price from January 29,2014 to June 30,2014. In order to improve the robustness, we process it with rolling time window to test the availability of the parameters and the strategies, so the cointegration test and Error Correction Model should be made every 250 days and each section of the forecast period includes 20 days. The data out of sample is divided into 5 sections to simulate the pairs trading.The results of the parameter debugging of the banks and real estate stocks are (0.28,1.1,2.4) and (0.06,1,1.9). The difference of parameter combinations reflects the difference of the two pairs of stocks’ price spread movement from the side, suggesting that we should choose the appropriate parameter combinations while doing pairs trading. Simulated trading outside the samples shows that the two pairs of stocks may not co-integrated in a forecast period, thus the trading session was shortened to 80 day. The difference of the mean value, standard deviation and model coefficients of each forecast period is also significant, so we can assume that the robustness of price spread sequence have been improved after processing.Banking and real estate stocks in the whole sample period found 15 and 11 opportunities for pairs trading, respectively, got high annual returns and positive Sharp ratios. The annual return out of the sample were 13.02% and 19.90%, respectively, but the Sharp ratios of Banks stocks are higher than real estate stocks both within the sample and out of the sample, so the Banks stocks are more suitable for pairs trading. Simulation results show that the pairs trading system built with cointegration and parameter debugging runs well and investors could get high investment return in the bear market.