Expected Stock Returns and Firm Characteristics: E-LASSO, Assessment, and Implications
68 Pages Posted: 13 Jun 2018 Last revised: 11 Sep 2021
Date Written: September 10, 2021
We develop new methods for constructing and analyzing cross-sectional stock return forecasts. We propose an E-LASSO approach that uses the LASSO, forecast combination, and forecast encompassing to implement shrinkage in a flexible manner designed to handle a large number of firm characteristics. We provide a cross-sectional out-of-sample R-squared statistic for assessing the accuracy of cross-sectional forecasts. Empirically, with presently the largest set of 193 firm characteristics, we find that our E-LASSO forecast produces significant cross-sectional out-of-sample R-squared gains and generates substantial economic value consistently over time. We further find that many firm characteristics matter, instead of a dozen or so.
Keywords: Cross-sectional expected stock returns, Characteristic premia, Shrinkage, LASSO, Forecast combination, Forecast encompassing
JEL Classification: C53, C55, C58, G11, G12, G17
Suggested Citation: Suggested Citation