An Empirical Analysis On Statistical Arbitrage Based On GARCH And Ornstein-Uhlenbeck Models

In recent years, with continuously rapid development of the China’s financial reform process and the gradual improvement of the capital market, the domestic commodity futures market’s influence is increasing and then receives much attention from investors all over the world. Especially after the international financial crisis, as the domestic futures market system advantages gradually appeared, it has been further affirmed. However, even if we has a perfect market environment and a robust system of trading rules, it is impossible to completely avoid various unforeseen financial risks which will be brought by the financial globalization or liberalization and risks posed by the futures transaction itself. Meanwhile our country is relatively lack of mature, rational market participants.So under the dual influence of the competition and the risk, promoting the financial innovation and the development of the futures market and seeking a way to secure a stable income for derivatives investors have become the internal demand of China’s financial market development.High frequency statistical arbitrage is based the on mean reversion and the market neutral assumption of a high frequency low-risk arbitrage strategy, which can be obtained stable investment income out of the market trend, and the strategy does not require investors to invest futures depending on the the price charts or the ordinary technical analysis, so we only need selecting a higher consistency of the trading portfolio, and then according to the trend of the spread between securities, purchase the relatively undervalued securities while selling the relatively overvalued securities, and finally we can obtain the returns by perform the the reverse operation when the spread go back to be normal level. In addition, since spread’s special property of the convergence in probability, we must have a strict stop loss criteria so as to prevent significant losses causing by the unusual fluctuations which may not return to the equilibrium level in a short time. Programmed statistical arbitrage trading with high-frequency datas is mainly based on the computer automated trading, and the key is to establish transactions and stop signals based on the varying spread, and then follow the instructions to carry out arbitrage trading operations.In this paper, we choose the RU1401and RU1309rubber futures1minute high-frequency datas in2013as the research object. We respectively use the same in-sample mean method and the moving average method to determine the spread’s mean center of out-sample static model and dynamic model (model parameters and trading signals will keep updated as the sample datas change), at the same time we will choose the dynamic GARCH and Ornstein-Uhlenbeck arbitrage trading model as foundation of research, with expected revenue maximization method to determine the range of the optimal and close positions, and learning from VaR method’s thoughts to determine stop signal. In addition, considering the static model and the out-sample data may exist mismatch problem, we will compare the results of static and dynamic statistical arbitrage in the empirical part. And then by analyzing the profit ratio, net profit, yield, effective yield, average yield and other indicators, we expect to identify the strengths and weaknesses of each strategy and its suitable investment crowd. Final empirical results show that the dynamic arbitrage strategies based on standard deviation and the static arbitrage strategies based on O-U models are more conservative, they both pursuit the stability earnings with minimal risk which are more appropriate for risk aversion to invest; while the arbitrage strategies based on historical statistical laws, the static statistical arbitrage strategies based on standard deviation and the statistical arbitrage method based on time-varying O-U processes have fewer stops less and higher yield, which are more suitable for risk-neutral or risk preference investors to consider. Thus, investors can select the arbitrage method according to different profit objectives, their own risk preferences and the market conditions. In conclusion, the arbitrage strategies effects of our paper are better on the whole, and fully embodies the market advantages of statistical arbitrage out of the market trend, providing a new idea to build high-frequency statistical arbitrage strategy.