Machine-Learning about ESG Preferences: Evidence from Fund Flows

Abstract

We construct Environmental, Social, and Governance (ESG) scores for U.S. active equity mutual funds based on over 500 underlying metrics covering a wide range of ESG issues and rating agencies. We use a revealed preference approach, combined with machine learning methods, to identify key issues driving fund flows, including waste and pollution (E), product responsibility (S), and business ethics (G). We also generate ESG-driven flows as proxies for a fund’s ESG performance. ESG-outperforming funds subsequently attract greater flows but yield lower benchmark-adjusted returns. Investors pay $17 million per year more for a top ESG fund.