摘要
我国股票市场存在高风险股票反而伴随较低收益的低风险定价异象,这有悖于传统资产定价理论.本文使用宏观经济和微观企业特征构建了六百多个变量的宏微观混合大数据集,并结合多种经典机器学习算法开发了基于大数据和机器学习的智能动态CAPM模型,检验了时变系统性风险对我国股市收益解释能力.实证结果表明:本文的智能动态CAPM定价模型能够显著解释我国股市低风险定价异象;随机森林等非线性机器学习算法表现最佳;影响股票时变系统风险的主要因素是市场类因子,基本面因子居次.本文对于我国股市系统性风险测度、动态资产定价模型构建和金融与大数据和人工智能融合创新有重要理论与实践指导意义.
This paper proposes a risk-based explanation for the low risk anomalies in Chinese stock markets,i.e.,stocks with high beta risks tend to generate surprisingly lower expected returns.A novel conditional CAPM model is constructed with big data and machine learning,in which more than 600 conditioning variables from macroeconomy and firm characteristics are employed,together with several popular machine learning methods in order to precisely modelling the time-varying beta systematic risk measures.Empirical Results show that our conditional CAPM model can fully explain away the low risk anomalies andthat market-related firms characteristics are the most important predictive variables.Further,random forest featuring nonlinearity is the most effective machine learning method in modelling time-varying beta.This study contributes to the time-varying risk modelling,conditional asset pricing,and application of machine learning in finance literature.
作者
姜富伟
马甜
张宏伟
JIANG Fu-wei;MA Tian;ZHANG Hong-wei(School of Finance,Central University of Finance and Economics,Beijing 100081,China)
出处
《管理科学学报》
CSSCI
CSCD
北大核心
2021年第1期109-126,共18页
Journal of Management Sciences in China
基金
国家自然科学基金资助项目(72072193,71872195)
国家社科基金资助重大项目(19ZDA098).
