基于重现时间间隔分析的极端收益预测及交易策略研究  

Trading Strategies and Extreme Return Predictions based on the Recurrence Interval Analysis

作　　者：吴婧 蒋志强[2] 周炜星[1,2] WU Jing;JIANG Zhi-qiang;ZHOU Wei-xing(School of Science,East China University of Science and Technology,Shanghai 200237,China;School of Business,East China University of Science and Technology,Shanghai 200237,China)

机构地区：[1]华东理工大学理学院,上海200237 [2]华东理工大学商学院,上海200237

出　　处：《中国管理科学》2020年第5期39-51,共13页Chinese Journal of Management Science

基　　金：国家自然科学基金资助项目(U1811462,71532009,71671066);上海市哲学社会科学规划一般课题资助项目(2017BJB006);中央高校基本科研业务费资助项目。

摘　　要：极端收益的预测在金融风险管理中非常重要。本文系统研究了极端收益重现时间间隔的统计规律,提出了一种基于重现时间间隔分析的早期预警模型,并对极端收益的重现进行预测,检验了模型在样本内外的预测性能;最后分别针对极端正收益和极端负收益的样本外预测结果,设计了看涨和看跌的两种交易策略,并以中国上证指数、法国CAC40指数、英国富时指数、香港恒生指数和日本日经指数为例,对交易策略的日均收益率进行了统计显著性检验。研究结果表明,极端收益的重现时间间隔具有右偏、尖峰厚尾和强自相关等典型特征;极端收益预测模型在样本内和样本外检验中都具有良好的预测能力;看涨和看跌交易策略在卖出区间均能有效地避开下跌阶段,看涨策略有更显著的盈利水平。Predicting such extreme financial events as market crashes,bank failures,and currency crises is of great importance to investors and policy markers because they destabilize the financial system and can greatly shrink asset value.A number of different models have been developed to predict the occurrence of financial distresses.Here,an early warning model is built to predict the recurrence of financial extremes based on the distribution of recurrence intervals between consecutive historical extremes.The extreme returns are determined according to quantile thresholds which includes 95%,97.5%,and 99%.By taking into consideration the time in which extremes occur,our prediction of extreme returns is based on the hazard probability W(Δt|t),which measures the probability that following an extreme return occurring at t time in the past there is an additional waiting timeΔt before another extreme return occurs,where the hazard probability■.In this paper,three common functions are employed to fit the recurrence interval distributions,and it is found that the recurrence intervals follow q-exponential distribution.Using the hazard probability,an extreme-return-prediction model is developed for forecasting imminent financial extreme events.When the hazard probability is greater than the hazard threshold,this model can warn when an extreme event is about to occur.The hazard threshold is obtained by maximizing the usefulness of extreme forecasts.In order to test the validity of our extreme-return-prediction model,a recurrence interval analysis of financial extremes in the Shanghai Composite Index during the period from 1990 to 2016 is performed.The data before each turbulent period are used to calibrate the model and each turbulent period that follows for out-of-sample forecasting,which obtains three in-sample periods:2000–2002,2006–2009 and 2014–2016.It is found that the recurrence intervals exhibit the characteristics of positive skewness,fat-tailed distributions,and positive autocorrelations.Both in-sample and out-o

关 键 词：金融物理学 交易策略 极端收益 重现时间间隔分析 收益预测 

分 类 号：F831.5[经济管理—金融学]