The Study On The Selection Of Trading Time In Pair Trading

As a market neutral strategy,pair trading has a brighter prospect in Chinese financial market with the development and promotion of mechanism of short sale.Therefore,the analysis of the application of pair trading in various industries is meaningful,which can help investors to identify the arbitrage opportunities in current market condition or quickly catch opportunities when the short sale systems are further improved.The success of pair trading depends on the optimization of two aspects: the choice paired stocks and the choice of trading time.The study in choosing paired stock is relatively more mature.Therefore,this essay paid more attention on the identification of trading time.There are two elements for choice of paired stock: the determination of price spread and threshold.So,this essay analyzed the problem of choosing trading time from these two perspectives to exploit the optimized parameters and models so as to help investors more accurately identify trading signals and achieve the profit maximization.First,the determination of price spread relies on the hedge ratio between paired securities,which can be acquired via error correction model and Kalman filter model.They both base on the least square regression.However,the former one is static and the later one is dynamic.Compare these two methods,Kalman filter was worse than basic cointegration method because it daily-varied spread series causes the daily change of state between two stocks and the scope of identification of trading signals also changes.Namely,the spread which lie in closing scope may belong to down-upper scope,which influences the profitability.So,cointegration relationship and error modification can provide better mean-reverting property and profitability.In the perspective of threshold of opening and loss-stopping position,it compared method of Common Parameters and Parameterization.Although different pairs of stock had different optimal opening and loss-stopping threshold,the non-parameter method was better than O-U process in general,which may because thresholds under the O-U process always lower than non-parameter.So,it limited the profitability.So,basic cointegration relationship with non-parameter method was more suitable to be used in pair trading.However,given that the data has different features in different industries and periods,the general applicability of these findings cannot be asserted,which need the further experience in newer and wider market and continuously improve and track their performance.