Predicting ESG Scores Using Firms' Financial Indicators: A Machine Learning Regression Approach

The oil and gas industry plays a pivotal role in the global energy and financial markets. With increasing concerns surrounding Environmental, Social, and Governance (ESG) scores, their impact on this sector has become a growing area of focus. This study aims to forecast ESG scores in the oil and gas sector using extensive datasets comprising publicly available financial and ESG indicators of firms. The research analyzed data from 497 companies within the industry over a 12-year period. A total of 10 machine learning algorithms were utilized to predict ESG scores, including decision trees, en-semble methods (boosting, bagging, and voting), XGBoost, LightGBM, random forest, extreme random trees, linear regression, robust linear regression, and elastic net. The analysis incorporated a one-year lag in ESG scores and employed panel data regression techniques in machine learning. The findings demonstrated a high predictive performance, with the best R² value reaching 0.922. These results provide a practical framework for investors and decision-makers to evaluate a firm's ESG performance, facilitating more informed and sustainable investment decisions within the oil and gas sector.