Predicting ESG controversies

Abstract

In this thesis, I examine the correlations between firm characteristics and future Environmental, Social, and Governance (ESG) controversies using Refinitiv ESG data. Specifically, I investigate how various levels of ESG scores, along with fundamental and stock market characteristics, predict future controversies one or two years forward. Employing logistic regression models, I regress dummy variables of future controversies on these explanatory variables.

The findings reveal significant correlations between ESG controversies and ESG scores, pillar scores, and most sub-pillar scores. Notably, higher ESG scores are associated with a higher likelihood of future controversies, potentially due to increased visibility or managerial opportunism, where companies inflate their ESG scores to gain the benefits associated with high ESG ratings. Additionally, firm size and past returns are significant predictors of controversies. The analysis indicates that English-speaking countries are overrepresented in ESG controversies, and industries with high public visibility or scrutiny, particularly those involving consumer products, experience more controversies.

To develop a more comprehensive predictive model, I employ an elastic net algorithm, for essentially a combination LASSO and Ridge logistic regressions for model selection and regularisation. This method yields more intuitive results, highlighting the relative significance of the different predictors like ESG scores, firm size, past returns, and country-specific factors in predicting future controversies.

Overall, this thesis contributes to the understanding of ESG controversies and their predictors, offering practical insights for investors and other stake-holders. The results suggest that while ESG scores alone may not effectively mitigate risks, a nuanced approach considering firm characteristics and industry context can enhance the prediction of ESG-related issues.