The relative importance of ESG pillars: A two-step machine learning and analytical framework

This study aims to contribute to the ongoing and inconclusive debates regarding the relative significance of environmental, social, and governance (ESG) sustainability key performance indicators and their correlation with overall sustainability performance. We present a unique two-step analytical framework to leverage Thomson Reuters' ESG Score (ASSET4) database to determine the most impactful ESG pillars and their subcomponents at both the firm and industry levels. This framework integrates the Method based on the Removal Effects of Criteria (MEREC) with K-means cluster analysis. Through the MEREC-K-means framework, we determine the two most noteworthy ESG pillars within various industries, subsequently clustering them to form peer groups for comprehensive comparative analysis. We find that while the social and economic pillars are the two fundamental pillars of ESG performance in all industries in general, this prioritization sometimes differs from industry to industry. This research makes theoretical and practical contributions by introducing a novel dimensionality reduction technique in ESG pillars, offering valuable insights for strategic resource allocation in corporate social responsibility (CSR) and sustainability initiatives. The implications of our findings extend broadly to investors, policymakers, and practitioners, empowering them to make more informed decisions.