Research On Future-spot Arbitrage Strategy Of Stock Index Futures Based On AR-GARCH Model

The research focus of this paper is to analyze the spread series of future and spot,and use AR-GARCH model to construct a complete future-spot arbitrage strategy of stock index futures,including modeling,parameter optimization and back-tested off-sample,count the arbitrage results within and outside the sample.Some suggestions for parameter selection are given and the arbitrage results are compared with the arbitrage strategy based on the residual constant volatility model before the improvement,aiming to provide investors with more effective arbitrage strategy.Firstly,based on the statistical arbitrage principle and the theoretical basis of the AR-GARCH model,the empirical estimates are made by using the HS300 stock index futures prices and the HS300 stock index prices in the sample.The correlation between futures and spot is analyzed.It is found that the price series of futures and spot are highly correlated and the price series have cointegration,which is in line with the basic assumption of statistical arbitrage.Then,the autocorrelation of the spread series is studied,result shows that the AR(5)process exists in the spread series and the perturbation term of the autoregressive equation has the ARCH effect.Hansen and Lunde(2005)indicate that the GARCH(1,1)model is no less than the complex GARCH model in predicting the volatility.And in the IBM Revenue Analysis,it performs better than the complex model,so the GARCH(1,1)model can be used to fit the spread series.Secondly,simulate arbitrage trading.The strategy is back-tested based on the residual constant volatility model.Then the AR-GARCH time-varying variance model is applied to the future-spot arbitrage strategy for improving the strategy,optimize the parameters in the sample and select three groups of transaction thresholds representing different risk preferences.Then,perform back-tested outside the sample under three sets of transaction thresholds.The strategy arbitrage results are compared with the pre-improvement strategy and the effectiveness of the strategy is analyzed,and some suggestions for parameter selection are given.The final results show that the arbitrage strategy based on the AR-GARCH time-varying variance model performs better than the arbitrage strategy based on the residual constant volatility model,indicating that GARCH model can better predict the fluctuation trend of the spread series in the short term.Regarding the selection of parameter,when considering the rate of return,the winning rate and the number of transactions,it is helpful to select the optimal parameter group.The yields of the future-spot statistical arbitrage strategy based on AR-GARCH model after deducting transaction costs under three different transaction thresholds are 0.23%,15.17%,12.04%,higher than the pre-improvement strategy yield-16.19% and the yield of the same period of the market-2.74%,proving the effectiveness of the strategy.