Seeking Signals from ESG Data

More and more attention has been paid in recent years to Environmental, Social, and Governance (ESG) issues at companies and to the data those companies release that seek to quantify those issues. Along with that attention has come a debate about whether ESG data benefi ts investors in evaluating companies for their portfolios.

In this paper, we show that applying Gradient Boosting Trees to Bloomberg's own ESG dataset allows us to create an equity port-
folio with higher return and lower volatility than its benchmark Russell 3000 Index. We also investigate the interpretability of our model using SHapley Additive exPlanations and compare the results to a traditional Logistic Regression-based approach.