Font Size: a A A

A Study On Statistical Arbitrage Based On The High-frequency Data Of Index Future

Posted on:2013-10-31Degree:MasterType:Thesis
Country:ChinaCandidate:H YanFull Text:PDF
GTID:2249330374490800Subject:Quantitative Economics
Abstract/Summary:PDF Full Text Request
The official implementation of stock index futures and securities margin trading isa symbol that china’s security market has stepped into the hedge era of long-shorttrading,which also means that hedging trading strategies including statistical arbitragestrategy maturely applied in the developed western capital market can be absorbed andapplied in china market.Statistical arbitrage is a kind of market-neutral strategy that makes profitexploiting the mean-reverting owned by spread between assets which have stableco-integration relationship, instead of depending on the judgment to market trends.The key of this strategy is how to depict the features of spread volatility, based onwhich a trading model is designed.Unlike most research literature which are limited to just one kind of model andsample data, this paper has designed three statistical arbitrage trading model based onOLS model implying constant variance,GARCH model implying time-varying varianceand Ornstein-Uhlenbeck stochastic process, has conducted emprical tests using threetypes of high-frequency sample data of HuShen300Index Future,which aredesprectively5minutes,15minutes and30minutes. In addtion, a more rigorous andcomparative method is conducted to evalute trading performance based on out-sampletesting results.Testing results suggest that, on the level of data frequencies, the returns obtainedfrom5minutes high-frequency data is the highest,followed by15minuteshigh-frequency data; on the level of trading models, the returns obtained from O-Utrading model is higher than OLS trading model, while the GARCH trading model getsthe worst performance.The performance evaluation results indicate that on the level oftrading models, O-U trading model always has a better performance than OLS model,no matter adopting K-Ratio which is more reasonable and comparative, or usingtraditonal annualized Sharp Ratio,or annualized cumulative returns. This conclusionis consistent with the the former emprical testing results. It also indicates thatcompared to the OLS and GARCH trading model, the O-U trading model has strongadvantages on trading performance ultimately because of its more accurate modelinganalysis and depiction of mean-reverting. However, on the level of data frequcncies,the traditional performance evaluation methods reveal that the most suitable data frequency is5minutes high-frequency, then15minutes high frequency. Whereas theevaluation result based on K-ratio reveals that the most suitable data frequency is15minutes high-frequency. Then compared to the returns of reference index, no matterhow the market behavior fluctuates, the curve of statistical arbitrage yield rate stayssteady upward, which identifies the characteristics of statistical arbitrage strategy:market-neutral, low risk and steady return.Basd on the results above, this paper suggests that for steady investor, the beststrategy of index future high-frequency arbitrage is using O-U trading model and15-minutes high-frequency; for aggressive investor, the best strategy is using O-Utrading model and5-minutes high-frequency. Lastly, statistical arbitrage has broadapplication prospects in China market.
Keywords/Search Tags:Statistical Arbitrage, Co-integration, GARCH, Ornstein-Uhlenbeck, High-Frequency Data, Sharpe Ratio
PDF Full Text Request
Related items