A Study On Statistical Arbitrage Of Chinese Stock Index Futures

Considering fluctuate future markets, rational investors always bear the smallest risk for the highest benefits. Margin trading pilot was officially launched at March 31,2010, Shanghai and Shenzhen 300 Index futures contracts are also officially listed for trading at April 16, 2010, the first listing in Shanghai and Shenzhen 300 Index futures contract for May, June, September and December 2010 contracts.In foreign investment banks, hedge funds, institutional investors and other financial institutions, statistical arbitrage is in depth research, and constantly systematic trading procedures. But our margin trading and stock index futures started shortly, and the relevant aspects of the domestic limited. This article will apply our statistical arbitrage in stock index futures market, and test whether statistical arbitrage strategy is feasible or not in China. The main contents are listed as follows:(1) We introduce a traditional statistical arbitrage method which based on co-integration. Choose stock index futures as dependent variable, ETF50 as independent variables. Two variables of unit root test and found two variables in an order difference after are stationary sequence. In order to properly eliminate VAR model in the error term between autocorrelation, we determined the lag stage is three. The multivariate VAR model based on co-integration relationship have constant items, without the time trend function Johansen-Juselius co-integration test methods, stock futures price and ETF50 exist co-integration relation. The conclusion shows that this model can obtain 6.42% yield.(2) Owing to considering co-integration relationship of residuals of the more troublesome, and for residual item hypothesis can't be very good tally with the actual situation. Select our stock index futures and Hushen 300 constituent stock price as the research object. Don't consider residuals of the assumptions, use flexible least squares methods to estimate the time-varying regression coefficient, and spread stream. The empirical results showed that our model won 20.734% yields. The empirical result indicates that our model have an advantage over co-integration model.