Research On The Statistics Arbitrage Strategy Based On Pair Trading

Pair trading is one kind of statistics arbitrage strategy, which is widely used in oversea securities markets. By creating both long and short position for two different stocks, this trading strategy tries to gain investment returns from the converging of the pair spread. The advantage of pair trading is that the strategy hedges the systematic risk. This advantage makes the strategy be able to get steady profit from the bear market. The studies on pair trading by the foreign researchers focus on two main aspects:the first aspect is finding a method to pick up stock pairs from the stock candidates; the second aspect is making an optimal trading strategy to maximize a certain utility function. The domestic studies on pair trading show that the pair trading strategy has the ability to gain profit from Chinese financial market since China opened margin transactions in year2010.However, few researchers pay attention to the influence between the time parameters and the return of the strategy.By reviewing the literature, the theory and models of pairs trading areintroduced in thethesis. An empirical research of the distance method strategy is made by using the HSI300stocks data between year2009andyear2013.Based on the implementation of the classical trading strategy, different length of the formation period and trading period aretested to find whether the time parameters make influence onthe returns. By setting both the formation period and trading period from1month to12month,144different combinations are tested in the thesis. The results show that the different time parameters do affect the return of the pair trading strategy.Finally, a methodto describe the length of the trading period is introduced. This methodis based on the Ornsteirn-Uhlenbeck process and the first passage time theory. In this method, one trading period is split into an empty period and a holding period. Both two periods are the statistics variants described by the first passage time theory. The sum distribution of these two variants is calculated as the distribution of one trading period. According to this distribution, the length of the trading period is set to support finishing trading once at a certain probability.