Commodity Futures Statistical Arbitrage Based on High Frequent Data
|School||Zhejiang Technology and Business University|
|Keywords||Statistical arbitrage Co-integration Moving window Time-varying coefficient Combined model|
The development of quantitative investment in foreign countries has a history of more than30years. During the period, it has gained steady profits and its market scale has been expanded. Up till now, quantitative investing has been a very important method of investment in America. In recent years,Chinese government implements the stock index futures, security margin trading and emulation trading of treasury bonds futures as well as options, which lay a solid foundation for quantitative investment. In November2012, some futures companies are licensed to do asset management business, which will bring about structural change in futures industry. Futures market is more risky than stock market due to its leverage. Therefore, to pursue steady profits, trading strategies based on programing and quantitative investing will be main products of asset management business. Statistical arbitrage is an important method in quantitative investing and it is market neutral.This thesis is mainly concerned about the application of statistical arbitrage in commodity futures market. Through liquidity test, correlation test and co-integration test, we choose RB1309and RB1310as our main empirical objects. First, we propose a statistical arbitrage framework based on co-integration, which mainly introduces the choice of arbitrage contract, calculation of arbitrage cost, design of arbitrage procedure and evaluation of strategies. Second, considering the time-varying feature of the financial market, we introduce moving window theory into the framework, which provides parameters with time-varying features. Statistical arbitrage strategy under moving window mainly includes the following aspects:First, optimize the opening and stop-loss threshold. Traditional opening threshold uses fixed value. As is known to all, if the opening threshold is larger, the opportunity for arbitrage becomes less; when the opening threshold is small, frequent opening will leads to severe loss. Therefore, based on dynamic optimization, we choose opening threshold that maximizes profits. The parameters will follow self-adapting process. As a result, profits will not be affected by the threshold. As to stop loss, we choose threshold based on VaR and estimate price volatility rate based on GARCH model.Second, estimate the impact cost. Impact cost is a very important part of trading cost. Many strategies cannot gain expected profits when run in real market because they ignore the impact cost. Although many literatures take impact cost into consideration, they usually set it directly instead of estimating it based on models. Based on mathematical models, this thesis uses the data of commodity futures to estimate the impact cost of arbitrage contract.Third, estimate the co-integration coefficient and back testing of strategies. Different co-integration leads to different spread,which will later leads to different arbitrage process. This thesis adopts OLS estimating and time-varying estimating method to acquire co-integration coefficient. We use the above two strategies to back testing the sample and get a satisfactory result. We also test the stability of the parameters through sensitivity test. Considering the relative independence of the two strategies, we combine OLS strategy and time-varying coefficient strategy and conduct a back testing of3months. By analyzing the profits and risk, we prove that the combined model is efficient.The shortcomings and prospect of the study are given in the end of this thesis.