| After years of efforts,the first stock index futures in China,CSI 300 stock index futures,was finally listed in April 2010,and then in 2015,SSE 50 stock index futures and CSI 500 stock index futures were listed at the same time.Although China's stock index futures have been listed for a period of time,the research results in China are still mainly focused on cross-species arbitrage of commodity futures or cross-period arbitrage of stock index futures.The research on cross-species arbitrage of stock index futures is relatively less,while the research on stock index futures abroad is earlier and relatively mature.On the basis of foreign research,domestic research on cross-species arbitrage of stock index mostly focuses on the idea of statistical arbitrage,that is,cointegration arbitrage,and then gradually constructs cointegration GARCH model on the basis of cointegration.On the one hand,in order to expand the research methods of stock index futures and provide investors with a more effective arbitrage scheme of stock index futures,this paper uses one minute closing price high-frequency data of IF1912 contract,IH1912 contract and IC1912 contract as the sample data.And after the data's cointegration analysis,establishes the cointegration GARCH model,and innovatively constructs the AR-GARCH model based on cointegration.Then,based on the AIC and SC rules,the optimal model is obtained,and the arbitrage strategy is constructed according to the optimal model.After using four transaction thresholds to carry out back test arbitrage on the sample data,analyze the arbitrage results to select a better transaction threshold.In order to verify the validity of the model and transaction threshold,we use two sets of data outside the sample to verify the validity.In this paper,we find that there are autocorrelation and heteroscedasticity in the residual sequence of cointegration model,and AR(5)-TGARCH(1,1)model based on cointegration can better fit the data in the sample than cointegration GARCH model,and can effectively eliminate the autocorrelation and heteroscedasticity of the residual sequence,at the same time,the model also reflects the asymmetry effect of the residual sequence.As for the setting of transaction threshold,most of the previous research literature used the classical threshold(1,2,0.1)or the threshold that can obtain the maximum return.However,this paper found that the classical threshold(1,2,0.1)can not make the transaction in the sample obtain the optimal return and winning rate,and(0.75,2,0.1)this group of thresholds can get better returns and winning rate,so this paper thinks that many factors should be considered when setting transaction thresholds.At the same time,it is found that the setting of opening threshold should be as accessible as possible,and the setting of stop loss threshold is necessary.However,through the back testing results of the two groups of data outside the sample,it is found that the effectiveness of this group of thresholds(0.75,2,0.1)is not stable,reflecting that the transaction thresholds should be continuously optimized according to the different transaction data.In this paper,AR(5)-TGARCH(1,1)model based on cointegration is used to build a cross-species arbitrage model of stock index futures,which proves that GARCH model can be used in the arbitrage research of stock index futures in China.However,it should be pointed out that although the theoretical rate of return obtained from arbitrage trading based on the model is relatively satisfactory,in consideration of the cost factor,this paper only considers the handling fee,commission and fixed margin interest,and does not consider the additional margin,impact cost and other relevant factors,so in the actual futures trading,this model needs to be improved to to some extent in practical operation.With the research and application of machine learning technology,it is believed that the research on arbitrage of stock index futures in China will further meet the needs of actual trading. |
PDF Full Text Request