Forest Through the Trees: Building Cross-Sections of Stock Returns

Abstract

We build cross-sections of asset returns for a given set of characteristics, that is, managed portfolios that serve as test assets for asset pricing models and building blocks for new risk factors. We use decision trees to endogenously group similar stocks together by selecting optimal portfolio splits to span the Stochastic Discount Factor. Our portfolios are interpretable, and reflect many characteristics and their interactions. Compared to combinations of traditional sorts and machine learning prediction-based portfolios, our cross-sections have up to three times higher out-of-sample Sharpe ratios and pricing errors, and do not suffer from excessive repackaging/duplication of the original stocks.

Keywords: Asset Pricing, Sorting, Portfolios, Cross-Section of Expected Returns, Decision Trees, Elastic Net, Stock Characteristics, Machine Learning

JEL Classification: G11, G12, C55, C58