The Research Of Bond Futures Intertemporal Arbitrage Based On Copula-SV Model

September6,2013, five-year bond futures pricing officially listed in China Financial Futures Exchange. As the country’s first interest rate derivatives, Treasury futures markets and investors will be launched with a positive sense. So after its launch, arbitrage strategies will become the market attention and research focus. However, unlike other stock index futures and commodity futures, the subject of the bond futures contract is the nominal virtual bond. So even if based on the risk-free arbitrage assumption in the same way, because of the cheapest deliverable bonds (CTD), the bond futures become uncertainty, thus make the pricing and subsequent strategy of the bond futures more difficult.In most of the papers, the bond futures arbitrage analysis methods can be divided into two:the Holding Cost Theory and the Spread Volatility Models. But first, these two methods ignore the reality of the impact of the spread of random factors, the second is the assumption of that two models are more stringent. However, the data in reality is often difficult to fit such perfect assumption. Therefore, this paper intends to another point, and try to avoid the stringent assumptions of the previous model, starting the work at the perspective of portfolio, treat the intertemporal arbitrage portfolio of bond futures as an asset portfolio, which conducted the analysis of intertemporal arbitrage bond futures. This paper base on the in-depth analysis and combining results of the previous studies of bond futures pricing and arbitrage strategies,and put forward a portfolio based on the idea of Copula-SV bond futures through intertemporal arbitrage model.In order to verify the validity of the model, this paper use data modeling methods to compare the Copula-SV model and the traditional spread volatility models and founded:In the steady spread arbitrage portfolio case, the traditional spread volatility model and Copula-SV model has their own advantages, the use of traditional spread volatility arbitrage model even more robust, lower risk, but the income is lower relatively; On the other hand, due to the Copula-SV model estimate the distribution of spreads directly, it can characteristics the characterization of wave more detailed. Therefore, the Copula-SV model under steady spread arbitrage portfolio situations can gains more profit than traditional models, but due to the increase of the number of transactions, the winning percent of the Copula-SV model become decline and the transaction costs therefore is higher than the traditional arbitrage model. When the spread of the arbitrage portfolio is a non-stationary sequence, the traditional model based on the assumption of steady spread volatility will fail, because the Copula-SV model estimate the distributed directly at each time point, so in theory the Copula-SV can be applied to non-stable spread sequence situation, but because this idea is based on statistical arbitrage,the result of using the Copula-SV model will not be as great as the arbitrage spreads stable situation, not only the income decreased, the winning percent also decreased. In summary,when the arbitrage portfolio spreads is non-stationary, even though the Copula-SV model can be apply, but the performance is still not good.In addition, after the analysis and test on model performance and applicability, this paper uses Copula-SV model, from the arbitrage point of view, to do the empirical analysis of5year bond futures arbitrage portfolio:first, through cointegration analysis found that the Treasury bonds futures in recent months and far month contract price sequence pass the cointegration test, and this two contract has a long-run equilibrium relationship, and through the cointegration regression equation we can obtained the arbitrage positions of the5year bond futures arbitrage portfolio. Secondly, considering the arbitrage portfolio in essence can be regarded as a portfolio, from the portfolio direction, at the view of the single futures contract, this paper select the stochastic volatility model of three kinds of thick tail to estimated and compared then found that SV-GED and SV-t model of the short term contracts and futures return’s volatility and marginal distribution are the highest degree of fitting. Then, according to the maximum likelihood method to estimate the parameters values and select the joint distribution of the elliptic functions include the normal Copula function, t-Copula function, and the Archimedes function include ClaytonCopula in group, FrankCopula, Gumbel Copula function. According to the maximum likelihood value, we found the best fitting joint distribution is the t-Copula function, and use it as the final joint distribution of the arbitrage portfolio. After getting the joint distribution, then use MCMC method to simulate the confidence interval of arbitrage portfolio, and the confidence interval is the confidence interval of price in fact, and then using the idea of statistical arbitrage with specific call and put point setting, and the stop surplus and stop loss point selection of arbitrage strategy to determined optimal arbitrage interval, then we can get the5year treasury bond futures arbitrage model.Based on the intertemporal arbitrage portfolio in the five-year bond futures contracts and forward contracts in recent correlation analysis we found that, under normal circumstances, the5-year bond futures arbitrage portfolio in recent contract with forward contracts have a strong correlation.Therefore intertemporal arbitrage bond futures under general market can guarantee higher returns and lower risk, but also has better operating accuracy. On the other hand, in extreme cases, especially in a bear market,the five-year bond futures arbitrage portfolio in recently its correlation between contracts and forward contracts is more intense compared to the performance of the bull market,so we can guarantee higher returns by doing arbitrage trading.In this paper, we test the estimated model from before and found the annualized revenue rate of the model is12.96percent, it’s annualized revenue rate is far beyond the general profitability of financial products on the market, it means the set up and the test result of the model are very satisfactory. On the other hand, we use the estimated model to predicted and test the arbitrage interval outside the period of the samples data, and make an examples by using the model to introduced the specific methods of operation in real bond futures intertemporal arbitrage transactions.