Study On Applying A Statistical Arbitrage Technique Of Pairs Trading To High-frequency Data

The introduction of margin trading and stock index futures provides a good platform for the implementation of statistical arbitrage strategy. In this background, this thesis makes a comprehensive and in-depth study on the arbitrage strategy based on pairs trading in China's stock market.The thesis first makes a correlation analysis of 20 constituent shares' prices of SSE Mega-cap Index with daily data and five different intraday high-frequency datas, and chooses a share combination of China Shenhua Energy Company Limited and China National Coal Group Corp as potential statistical arbitrage portfolio whose stock prices have strong correlation relation among all six frequency sample.With the rising use of high-frequency data and the arbitrage profits dropping significantly caused by the use of low-frequency data in the more and more effective market recently, how to use high-frequency data to arbitrage is a tendency in the future.Taking this phenomena into account,this thesis use traditional daily data and five different intraday high-frequency datas to study the massive arbitrage opportunities that possibly exist, the effectiveness of high-frequency data in the statistical arbitrage strategy is verified. The data frequency's influence on the strategy is also compared.As the market environment is always changing, the common statistical arbitrage strategy is not necessarily suitable for all products in all market environments.In view of the problems that exist in applying the common statistical arbitrage strategy, the thesis makes corresponding improvements, and gets a new strategy which is significantly better than common one as shown by empirical research, and is taken as a basic strategy for further study.The traditionaly basic statistical arbitrage model-Cointegration model assumes thatβis constant, this deviates from that financial time series model's parameters are often time-varying.The thesis innovatively introduces Kalman filter model to the statistical arbitrage strategy to describe the relationship between asset prices, compares the performance of the two models in the arbitrage strategy, and finds neither model is obviously better. Finally, the thought of combination forecasting is introduced into the statistical arbitrage strategy, and, combined with concentrate targets in statistical arbitrage strategy, a combination strategy is developed. The empirical tests show that the new combination strategy is significantly better than the arbitrage program that adopts single model.