Empirical Asset Pricing via Machine Learning
Comparative analysis of machine learning methods for measuring equity risk premiums (expected stock returns), using 94 stock characteristics, their interactions with eight aggregate time-series variables, and 74 industry dummies.
Verdict
Comprehensive empirical study of ML in asset pricing. Trees and neural networks perform best by capturing nonlinear predictor interactions. Highlights the value and limitations of ML for measuring risk premia.
Practical Use
Essential reading for applying ML to equity factor investing and return prediction. The paper compares linear regression, penalized linear models (elastic net), principal components regression, partial least squares, random forests, gradient-boosted trees and neural networks on nearly 30,000 US stocks from 1957 to 2016. Key finding: trees and neural networks beat linear models by capturing nonlinear predictor interactions, and all methods agree on the same dominant signals (variations on momentum, liquidity and volatility). The authors' GitHub repo holds MATLAB simulation code, not the full empirical pipeline; Dacheng Xiu's website offers a separate dataset of 94 lagged firm characteristics (datashare.zip, about 1.6 GB, labelled "UPDATED June 2021"); the site says its information is "for research purposes only and should not be used for investment or other commercial purposes", and whether that covers the download is not stated. Stock returns in the paper come from CRSP; the dataset's readme says CRSP returns are not included. A research blueprint, not a production library.
Paper
https://doi.org/10.1093/rfs/hhaa009Shihao Gu, Bryan Kelly, Dacheng Xiu. Review of Financial Studies, 2020
Code Repository
https://github.com/xiubooth/ML_CodesLicense
Custom. Review of Financial Studies version: CC BY-NC-ND 4.0 ("Open Access"); NBER working paper: "All rights reserved". Code repo xiubooth/ML_Codes: no license file. Author site (datashare.zip readme: "@copyright Shihao Gu, Bryan Kelly and Dacheng Xiu"; site disclaimer: "for research purposes only and should not be used for investment or other commercial purposes").