摘要
股市风险溢价是金融学中的一个经典研究问题.常见的线性模型存在着模型误设和参数不稳定的问题,难以有效预测风险溢价.本研究从机器学习的视角重新检视了中国股票市场的可预测性.基于1996年1月—2019年12月的数据,构建提升回归树(boosting regression trees,BRT)模型对股市收益率与波动率进行样本外预测,并构建了最优风险资产配置模型.实证结果显示:1)提升回归树方法能够对收益率、波动率和最优风险资产权重做出准确预测;2)在收益率预测中最重要的三个变量分别是净权益增加值、换手率和股价方差;挖掘预测变量之间的非线性关系是BRT预测能力的来源;3)结合提升回归树预测构建的最优风险资产组合可以为投资者带来更高的收益和效用.本研究将机器学习方法引入股票市场风险溢价的研究,为此问题的研究提供了全新的视角.
Equity premium prediction is one crucial research problem in finance.The traditional linear model suffers from model misspecification and parameter instability,weakening its out-of-sample prediction performance.This paper re-examines equity premium predictability in the Chinese stock markets.Based on the data from January 1996 to December 2019,we adopt Boost Regression Trees(BRT)to predict market return and volatility.Empirical results show that BRT outperforms traditional methods in predicting return,volatility,and optimal portfolio allocation.The three most important variables in the prediction model include Net Issuance,Turnover,and Stock Variance.The BRT’s predictability is driven by its ability to capture nonlinearity among variables.Finally,the optimal portfolios constructed by BRT result in higher returns and utility to investors.Our work contributes to the literature by leveraging the nonlinear machine learning method for the equity premium study,thus providing new research perspective.
作者
李斌
龙真
LI Bin;LONG Zhen(School of Economics and Management,Wuhan University,Wuhan 430072,China;The Centre of Finance Research,Wuhan University,Wuhan 430072,China)
出处
《管理科学学报》
CSSCI
CSCD
北大核心
2023年第10期138-158,共21页
Journal of Management Sciences in China
基金
国家自然科学基金资助项目(71971164,72371191)
科技创新2030-“新一代人工智能”重大项目课题资助项目(2020AAA0108505)
国家社会科学基金资助重大项目(20&ZD105)
关键词
权益风险溢价
提升回归树
样本外预测
机器学习
equity risk premium
boosting regression tree
out-of-sample prediction
machine learning
