ESG and firm valuation: causal effects versus predictive relevance

This study examines the relationship between environmental, social, and governance (ESG) performance and firm valuation, with a particular emphasis on distinguishing between causal effects and predictive relevance. Using a panel dataset of publicly listed firms, the analysis combines panel econometric models, including pooled OLS, fixed effects, dynamic specifications, and System GMM estimators, with machine learning techniques to provide a comprehensive assessment of ESG valuation dynamics. The econometric results consistently show that ESG does not exhibit a statistically significant or robust linear effect on firm valuation across model specifications. This suggests that previously documented ESG–value relationships may be driven by cross-sectional differences rather than within-firm causal effects. In contrast, the machine learning analysis demonstrates that ESG variables improve out-of-sample predictive performance, particularly when included in lagged form, indicating that ESG contains incremental predictive information. Further evidence points to a nonlinear, threshold-like relationship between ESG and firm valuation. ESG appears to have limited relevance at lower levels but becomes positively associated with predicted valuation at higher levels of ESG performance. At the same time, firm valuation is primarily driven by persistence and traditional financial fundamentals, with ESG playing a secondary and context-dependent role. Overall, the findings highlight a fundamental distinction between explanatory and predictive modeling in ESG research. While ESG does not emerge as a robust determinant of firm value, it contributes to prediction through nonlinear and interaction-based mechanisms. These results provide a more nuanced perspective on the financial relevance of ESG and carry important implications for both research and practice.