Decoding the Financial Determinants of ESG Outcomes: Evidence from Machine Learning Analysis

This study investigates the determinants of Environmental, Social, and Governance (ESG) outcomes associated with financial performance, utilizing data on publicly listed firms in Taiwan from 2015 to 2023. To address challenges of high dimensionality, multicollinearity, and variable selection bias, two machine learning methods, PCA and LASSO, are integrated into an integrated empirical framework to improve model robustness and interpretability. The PCA results indicate that a limited number of latent factors, primarily associated with profitability, operational efficiency, and human resource structure, account for most of the variance in ESG performance across industries. The LASSO analysis further identifies industry-specific determinants, suggesting that financial robustness, managerial capability, and workforce stability play crucial roles in shaping ESG outcomes. However, the magnitude and direction of these effects differ across industries, indicating that sustainability strategies should be tailored to specific industry characteristics rather than implemented uniformly. Overall, the findings underscore that firms with strong financial foundations are better positioned to implement sustainable practices effectively; in contrast, workforce instability and a short-term focus on profitability may hinder ESG performance.