Saad Mouti, UC Berkeley
There is a lack of consensus on whether environmental (E), social (S) and governance (G) attributes are associated with excess return or risk factors. From sin stocks to socially responsible companies, different research papers and professional reports lead to different conclusions. The lack of consensus can be attributed, in part, to disagreement in aggregate ESG scores provided by major vendors. Other sources of disagreement are experimental design, which includes the choice of which firm characteristics to control, as well as outcome metrics.
Using yearly KLD-MSCI data from January 1992 to December 2019, we explore the alpha and beta generated by long and long/short positions on the ``good’’ and ``bad’’ companies with respect to the CAPM model, the Fama-French three- and five-factor models, and the Carhart four-factor model. We make use of the Sustainability Accounting Standard Board (SASB) Materiality Map to associate industries with the ESG variables most likely to be financially relevant. The novelty of the study, however, comes from the application of causal inference framework and matching methods to control for firm characteristics such as market beta, size, profitability or momentum. We use a set of covariate balancing techniques, such as matching with a propensity score and Mahalanobis distance, to match companies with a positive score (treated) to otherwise similar companies with a neutral or negative score (control). This improves the solution to the imbalance problem between the ``good’’ and ``bad’’ set of companies and creates better grounds for comparison.