Financial Materiality of ESG in Global Manufacturing: Econometric and Machine Learning Perspectives

Amidst significant digital and sustainability shifts in the global manufacturing industry, this research examines the relationship between Environmental, Social, and Governance (ESG) considerations and excess stock returns of manufacturing companies worldwide. A dual methodology integrating panel regression and machine learning techniques (Random Forest and XGBoost) is employed to assess overall ESG ratings, specific operational indicators, and ESG risk classifications. The panel regression analysis reveals no immediate, statistically significant linear relationship for most ESG measures; however, board size displays a positive correlation with returns. Notably, firms with high ESG risk exhibit a slight positive association with excess returns. In parallel, machine learning models achieve fair predictive accuracy, identifying market variables as key drivers while also ranking environmental scores and ESG risk categories among the leading predictors. These results suggest that, although direct linear effects are limited, ESG elements possess predictive capabilities, potentially through complex, non-linear interactions. The study offers nuanced perspectives for stakeholders navigating the evolving landscape of sustainable and technologically advanced manufacturing.