Research About Statistical Arbitrage On Futures

After 2006 China's stock market and futures market have undergone rapid development, which not only give people the opportunity to increase the wealth, but also bring great risks, especially one-way speculative trading makes investors expose to great market risks. In order to increase trading opportunities and reduce risk, this paper uses statistical arbitrage approach to the Dalian Commodity Exchange listing of soybeans, soybean oil and soybean meal futures contract, using co-integration to building a portfolio, and mainly focus on portfolio building, unwinding and risk control mechanism, and thus obtain a relatively complete set of investment strategies.This article first undertake a review of relevant literature on statistical arbitrage at home and abroad, we find that these documents are primarily used to study the stock market (1) used to study the effectiveness of the stock market; (2) used to study the characteristics of the stock market momentum,seasonal characteristics, fat-tail and peak effect; (3) used to study the specific equity arbitrage strategies and methods. These studies basically cover the markets from both the developed and developing countries, and give the corresponding explanation for some phenomena. Of course, there are some studies the futures market literature, especially in the oil and chemical class futures contracts, which give quite a good empirical research.Then this paper makes a brief summary on the concept of statistical arbitrage, co-integration theory, extreme value theory, which paves the way for the empirical research. In this paper, co-integration theory is the fundamental basis for building the portfolio, while risk-control part of this article mainly relies on calculation of VaR and ES coming from extreme value theory.The most important part of this paper is empirical. First, we tested Stationary of three years'data of the soybean oil, soybeans, and soybean meal futures contracts. we found that they are a first-order single integration, then test co-integration on the basis of ADF test and found that co-integration relationship exists between them, and establish portfolio and identified each ratio of futures contract in the portfolio: soybean:soybean oil:meal=1:-0.1833:-0.8438. Then we establish the location of the position in 1.16σand -0.48σusing the expectation maximization function approach, but this calculation has not included the cost, so we get the transaction threshold which makes portfolio return largest using the method stepwise-style within the in-sample, the upper and lower thresholds are 0.8σ(there is another group threshold, but we forbid it because the probability is too small), when the residuals of portfolio reaches zero, then close position, one transaction is over; When a small probability event happens, residuals is more than VaR value, close the position, confess of the loss. We get the VaR values are 665 and -534, the small probability events happen when residuals are over VaR, arbitrage model fails.This paper compared portfolio returns and established market index. we found that the portfolio returns are far better than the market index in sample, showing a strong earning power and smaller volatility, showing the characteristics of market-neutral; portfolio earnings has a poor performance compared with the market index, but the volatility is still less than the market index, with market-neutral characteristics. In the aspect of excess revenue, the in-sample has great excess returns, the out-sample has negative excess returns, which have shown no market efficiency in China's futures market.