Research On Statistical Arbitrage Model Based On Multi Dimension Cointegration

In 2010,Hu-shen 300 Index Futures and margin trading to the launch of China’s stock market has brought short-mechanism,for the development of statistical arbitrage provided favorable conditions.Statistical arbitrage is a strategy being widely used by domestic and foreign investors,traditional statistical arbitrage focuses on pairs trading strategy,that is by buying the undervalued stocks,and selling overvalued stocks to get return.While statistical arbitrage is a risk arbitrage,it’s risk comes from the failure of the statistical model and the failure of pairing stock’s cointegration.This is the first time to establish statistical arbitrage models by using multi-dimensional cointegration test,by establishing a VAR model,and using the least squares method to estimate the parameters of the model and using AIC criterion for model order.With multiple stocks as arbitrage objects,through spreads deviation and regression to establish arbitrage model.First of all,this paper gets the stock index: asset liability ratio、total assets turnover、rate of return on common stockholders’ equity after deducting non recurring gains and losses、net profit growth rate、earnings per share growth rate、main profit growth rate,use the method of K-means clustering and Johansen cointegration test to filter stock,and finally,according to industry prorated to allocate cash.Calculate the spread of stock in each class,buy stock whose rank is the last and short Hu-shen 300 Index Futures,when spread back to near zero then close position.This paper uses historical data to carry on the back test,by constantly modifying strategy: firstly, use stop-loss strategy,it is found that the effect of the strategy has not greatly improved;then change spread back to zero to spread moved up to the specified rank,the policy effect has been improved;last,change conduct a cointegration to daily cointegration test,achieved an annual yield of 38.32% and a maximum of 10.6%.The empirical results show that the strategy is effective, it can be applied to the actual trader strategy.