notices - See details
Notices
building-capital-markets
THEME: CAPITAL MARKETS
27 February 2026 Financial Analysts Journal Volume 82, Issue 2

Rethinking Variable Importance in Machine Learning

An Economic Perspective on Empirical Asset Pricing

Yonghwan Jo and Yong Hwi Kim

Which firm characteristics truly add economic value in ML portfolios? Out-of-sample tests show microcaps distort results, some predictors hurt returns, and liquidity and risk signals matter most.

Are you a CFA Institute Member? Sign in to access the full article CFA Institute Member Content Not a CFA Institute Member? View or purchase on Taylor & Francis online

Abstract

We study which firm characteristics drive the economic value of machine learning portfolios. Three results stand out. First, in-sample variable importance overfits and provides little reliable guidance, highlighting the need for out-of-sample evaluation using economic criteria. Second, conventional models are dominated by microcaps, which inflate returns and concentrate gains in costly-to-trade stocks; excluding microcaps is essential for meaningful inference. Third, some predictors carry negative importance and consistently degrade performance; removing them improves risk-adjusted returns and clarifies which characteristics matter. These findings show that only with economic restrictions can machine learning deliver robust asset pricing insights.