ESG Confusion and Stock Returns: Tackling the Problem of Noise
How does ESG (environmental, social, and governance) performance affect stock returns? Answering this question is difficult because existing measures of ESG perfor- mance — ESG ratings — are noisy and, therefore, standard regression estimates suffer from attenuation bias. To address the bias, we propose two noise-correction procedures, in which we instrument ESG ratings with ratings of other ESG rating agencies, as in the classical errors-in-variables problem. The corrected estimates demonstrate that the effect of ESG performance on stock returns is stronger than previously estimated: after correcting for attenuation bias, the coefficients increase on average by a factor of 2.6, implying an average noise-to-signal ratio of 61.7%. The attenuation bias is stable across horizons at which stock returns are measured. In simulations, our noise-correction pro- cedures outperform the standard approaches followed by practitioners such as averages or principal component analysis.
Florian Berg, Julian Kölbel, and Roberto Rigobon are grateful to Massachusetts Pension Reserves Investment Management Board, AQR Capital Management (Applied Quantitative Research), Qontigo, Asset One, and MFS Investment Management — members of the Aggregate Confusion Project Council — for their generous support of this research. We are also extremely grateful to the ESG rating agencies that provided their data to this project. We thank Ing-Haw Cheng, Aaron Pancost, and Jonathan Parker and seminar participants at the 2022 American Finance Association meetings, ARCS 2022, the 2022 ASU Sonoran Winter Finance Conference, George Washington University, London Business School, MIT GCFP, MIT Sloan, the 2021 PCOB colloquium on ESG Disclosure, the University of Amsterdam, the University of Mannheim, and the University of Michigan for excellent feedback. Julian serves as a member of the RepRisk Academic Advisory Council. All remaining errors are ours. Correspondence to: Roberto Rigobon, Sloan School of Management, MIT, 50 Memorial Drive, E62-520, Cambridge, MA 02142-1347, email@example.com, tel: (617) 258 8374. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.