Research On New Quantitative Hedge Strategy Based On Multivariate Cointegration Model

Since 2010,the continuous improvement of domestic stock index futures and margin trading business provides a very favorable condition for statistical arbitrage.One of the typical representatives of the statistical arbitrage strategy is pairs trading,however,due to China's A-share market can not short-sell stocks,so we consider to use stock index futures to hedge the combination of indices for avoiding market risk.In order to improve the robustness of investment portfolio,and get excess returns with a lower risk,we use the k-means clustering firstly,then the Johansen cointegration test or the Gregory-Hansen cointegration test with structural breaks(GH test)to construct a stock selection model on the basis of industry neutral.Eventually,taking advantage of spread's deviation and regression principle,we construct a quantitative hedging model based on the multi-dimensional cointegration relationship.As statistics arbitrage of multiple stocks requires a good cointegration relationship,and belong to the same industry is the premise for stocks have cointegration relationship,so,firstly,the CSI 300 Index constituent stocks are classified according to SWS industry standard.Secondly,we use the k-means clustering to find out the stock portfolio from the same industry with similar fundamentals.first of all,for every stock,we can get variables from each report period are as follow:return on equity?return on total assets?net profit growth rate?net asset growth per share?primary business profit growth rate?net profit growth rate?total asset growth rate?debt ratio per share?asset liability ratio?current liabilities ratio?total asset turnover?current ratio and inventory turnover ratio.we use these variables to do complexity analysis.after do complexity analysis on all report periods,we would have a comprehensive matrix,then,we do the k-means clustering on it with the stock as dimension.Thirdly,according to whether to consider structure breaks,we use the Johansen cointegration test or GH test on all closing price sequences from the same class,if the test is adopted,we think that the combination has multiple cointegration relationship.Lastly,we do arbitrage operations on the portfolio with multiple cointegration relationship:calculate spread of each stock,buying in the stock with spread is max.when it's spread ranking is rose to a certain position or it's spreadexpanded for seven consecutive days,we sell it.In the whole process,we short the CSI 300 stock index futures to hedge.Inside,the industry in the CSI 300 index accounted for the weight of the allocation of funds,if there are more then one combination in the same industry,then divided.For three months as adjustment period,from January 1,2016 to March 31,2017,the two models were harvested 32.03%,37.27% annualized return,This shows that the models are effective,and the latter's annual return is significantly higher than the former,which shows that it is necessary to consider whether there will be a structural break in the construction of cointegration arbitrage model.