Measuring What Matters: Constructing an ESG Controversy Index Using AI and Institutional ESG Frameworks

ABSTRACT

Environmental, Social, and Governance (ESG) controversies increasingly influence investment decisions, corporate accountability, and sustainability regulation. However, existing commercial controversy measures often rely on proprietary methodologies that limit transparency and make it difficult to understand how controversy assessments are generated. This paper aims to develop an ESG Controversy Index (ESGCI) that combines internationally recognised ESG principles with explainable artificial intelligence to provide a systematic approach to measuring firm‐level ESG controversy exposure. The ESGCI is constructed from more than 5000 annual and sustainability reports issued by S&P 500 firms between 2010 and 2023, together with ESG doctrines published by the United Nations, the Principles for Responsible Investment, the Organisation for Economic Co‐operation and Development and the International Labour Organisation. Natural language processing and supervised machine‐learning techniques are used to identify, classify and aggregate ESG controversies within a structured measurement framework. The proposed index is evaluated through a multi‐layer validation strategy incorporating conceptual assessment, document‐level out‐of‐sample testing, external comparison with Refinitiv ESG controversy measures, inter‐rater reliability, incremental validity and complementary robustness analyses. The findings show that the ESGCI captures controversy‐related information recognised by existing commercial measures while providing additional insights through a fully documented measurement process. The reported relationships remain consistent across alternative weighting schemes, lagged specifications and instrumental‐variable estimations, supporting the robustness of the empirical evidence. Rather than replacing existing ESG controversy ratings, the ESGCI offers a complementary approach that improves methodological transparency and facilitates reproducible ESG controversy measurement. By integrating institutional ESG principles with explainable AI and comprehensive validation procedures, this research seeks to contribute to the advancement of ESG measurement and provide a practical foundation for future research, investment analysis and sustainability reporting.