Theory Study And Its Application On Wild Bootstrap Cointegration Test Among Several Assets Portfolio

In this paper, we study the statistical arbitrage among multiple assets portfolio and suggest the method for the selection of stocks on the basis of influencing parameters of the stock price rather than the traditional correlation coefficient. Moreover, considering the heavy-tailed properties in the economic and finance data, we suggest the more robust Wild bootstrap method rather than the commonly used Johansen method in cointegration test and construct the statistical arbitrary based on the corresponding cointegration vector. Simulation studies show that the three main parameters have significant influence on arbitrage profits, and the test on the base of wild bootstrap method is also significantly better than the Johansen cointegration test. Empirical results further show that the statistical arbitrary method suggested in this paper also has high ratio of success in both the inside and outside sample data, and paired (setted) trading by selecting larger differences on the shift parameter index has higher yield and sharpe ratio. Specific chapter is organized as follows:The first chapter mainly introduces the background and significance of the research, the related knowledge of the statistical arbitrage, literature review and innovation of this paper.The second chapeter named pairs trading theory. Introducing the existed theoretical approaches in the literature according to the three-step of pairs trading, including stock selection, trading parameter setting and trading signal construction.The third chapter specified the shortcomings of the correlation coefficient method, EG cointegration test and Johansen cointegration test which mentioned in Chapter two. And proposed the method which select stocks based on price-sensitive parameter to replace the correlation coefficient method. In addition, using Wild bootstrap method alternative to EG test and Johansen test.The simulation study in chapter fouth verified the improved method that proposed in chapter third. The results showed that:(1)Selecting drift parameter with larger gap, higher volatility parameter values and similar and smaller initial stock price can excluded stock portfolio which have smaller yield. Using this method can effectively relpace the correlation coefficient method which only suitable for two stocks pairing.(2)Wild bootstrap method is much better than Johansen cointegration test, and the transaction parameters based on this method also performance better than other methods of transaction parameter setting.(3)Extending the two stocks pairing to three stocks pairing not only provides more diverse stock portfolio,but also has a higher strategy income and the cointegration relationship outside sample also more stable.The empirical research in fifth chapter verified the imroved strategies that proposed in this paper using stock market data and futures market data. The empirical results and simulation results are basically the same in stock market, while in futures market, the empirical results are slightly worse than expected.Chapter sixth summarized the results of this study.