Listed on April 16 in 2010, CSI300 futures have got the favor of many investors
and have become a mature future variety. Institutional investors in China are keen on
the low risk arbitrage trading in CSI300 future market, especially the stock index
futures and spot arbitrage with high robustness. But since 2011, the CSI300 future
market has been relatively mature, opportunity of future and spot arbitrage is not much,
which seriously restrict the arbitrage activity of arbitrage funds. Therefore, digging new
arbitrage strategies becomes the urgent topic.
The calendar spread arbitrage of CSI300 futures is a new direction of arbitrage
activity. Especially with the development of systematic trading mode, and statistical
arbitrage has become the main hedge strategy in quantitative investment field, the study
of statistical arbitrage in the CSI300 future market has very vital significance, whether
on the test of market efficiency, or for the supervision and administration of stock index
futures, or for investor’s program trading.
In this paper, 5 minutes high-frequency closing price data of CSI300 futures is
selected as study object. First, we study the relationship of the recent month consecutive
contract and the next month consecutive contract with co-integration theories, and
outcome shows that there is a long-term equilibrium relationship between them, so they
can be used to construct co-integration arbitrage. Based on the theory of cost of carry
model, we study factors that could cause changes of the spread between these two
contracts, outcome shows that risk-free interest and dividend yield do have significant
influence on spread fluctuations and they are the granger causality of the spread
changes. Altogether, risk-free interest and dividend yield can explain 20% of the spread.
For more in-depth analysis of the spread, then spread is decomposed into two parts
——mean central and random disturbance. Different from previous research, EWMA
method is used in this paper to estimate the mean central. Related parameters are
determined afterwards through sensitivity analysis and Monte Carlo simulation, so a
general statistical arbitrage model is constructed on the sample data. Empirical results of
the strategy show that we can get 16.71% accumulated earnings in a year within the
sample, while the market’s earning is minus 25.02% in the same period, which fully
embodies the advantages of statistical arbitrage regardless of the influence of market
overall trend.
In addition, considering the time-varying variance characteristics of spread
fluctuations, we base our statistical arbitrage model on sample data, and two different
strategies are built for data out of the sample period. One is based on constant historical
volatility; the other is based on time-varying volatility which is produced by using
AR(3)-GARCH(2,1) model. Results show that the statistical arbitrage strategy can still
earn more than 7% of the annual income outside the sample, and compared to the
constant historical volatility, strategy based on time-varying volatility has better
stability, yields more arbitrage opportunities and is expected to obtain a higher risk
adjusted returns, which provides a new idea for the construction of statistical arbitrage
strategy.
Key Words:Stock index futures, Co-integration, Cost of carry model,