讲座：Regression Tree in Factor Models
题 目：Regression Tree in Factor Models
嘉 宾：Guanhao Feng, Assistant Professor, City University of Hong Kong
掌管人：张然 助理传授 上海交通大学安泰经济与办理学院
This paper provides a regression tree factor model (RTFM) that provides a unified framework to generate the stochastic discount factor (SDF) for the linear factor model. The regression tree offers an alternative top-down solution to security sorting for splitting the cross-section of stocks based on their firm characteristics. In particular, we design a multi-period tree model for the imbalanced panel data structure of individual stocks. The tree split criterion follows the non-arbitrage condition by fitting the SDF model with time-varying factor loadings. The SDF is a mean-variance efficient portfolio within a bottom-up approach on those leaf-basis portfolios generated by the regression tree. Additional applications include multiple SDFs through boosting regression trees, OOB variable importance through the random forest, and time-series splitting with macro predictors. Using U.S. equity data, we find RTFM outperforms Fama-French factor models for different pricing and prediction measures.
Guanhao (Gavin) Feng is an assistant professor of business statistics at the City University of Hong Kong. He is also the program leader of MSc in Business Data Analytics (Quantitative Analysis for Business stream). Feng's research work has been published or accepted at the Journal of Finance and Journal of Econometrics. His research work has received major academic and practitioner awards, including the 2018 AQR Insight Award, 2nd Prize of the 2019 Crowell Prize, Unigestion Alternative Risk Premia Research Grant Award, INQUIRE Europe Research Grant Award, and PwC 3535 Finance Forum Annual Best Paper Award. Also, Feng has been invited to present at major academic conferences (AFA, CICF, Informs, and SoFie) and practitioner conferences (CQAsia, Wolfe Research, and INQUIRE Europe). Feng obtained his PhD and MBA degrees from the University of Chicago in 2017. His research interests include financial time series, empirical asset pricing, machine learning, and quantitative finance.