动量策略收益尾部风险的实证研究——基于两状态隐马尔科夫模型

The Empirical Study of Tail Risk in Momentum Strategy Returns: Based on Hidden Markov Model with Two States

本文通过构建两状态隐马尔科夫模型,利用EM算法系统地研究了动量策略收益的尾部风险。研究发现,不论是波动状态还是平稳状态下,动量收益对市场上行的反应程度都要低于其对市场下行的反应程度。本文的研究也发现,在动量策略收益的残差项为学生分布及市场超额收益的残差项为正态分布的情形下,动量策略收益和市场超额收益的四阶矩均位于其模拟分布的95%置信区间内。在对动量崩溃进行预测的过程中,本文研究还发现,当隐马尔科夫模型中包含期权特征时,动量崩溃的错误预测次数相对较小,这表明期权特征在隐马尔科夫模型中的重要性。利用市场超额收益波动预测方法,其产生的动量崩溃错误预测次数显著低于波动状态预测概率的相应值。

This paper constructs a hidden Markov model with two states,we study the tail risk of momentum strategy returns using EM algorithm.We find that whether it is turbulent state or calm state,the response of momentum strategy returns to up moves in the market is less than the response to down-moves in the market.When the distributions for the residuals of momentum strategy returns and market excess returns are Student-t and Normal distributions respectively,the first four moments of momentum strategy returns and market excess returns lies inside of the 95%confidence intervals of Monte Carlo simulation.Meanwhile,in the process of predicting momentum crashes,we find that when the constraint of option-like feature is included,the false positives of momentum crashes are relatively smaller;this shows that the necessity of the option features in our hidden Markov model.Using the volatility of market excess returns as tail risk measure is obviously better than the measure of predicted probability for turbulent state,because the former will lower the number of false positives in predicting momentum crashes remarkably.