A Data-Driven Approach to Mitigating ESG Risks in Rare Earth Mineral Supply Chains

The global shift toward clean energy and digital technology has made rare earth minerals critical to modern supply chains. Despite their importance, existing ESG assessment tools remain inadequate: corporate sustainability reports systematically underreport negative incidents, and major ESG rating agencies exhibit an average inter-agency correlation of only 0.54. This paper develops and validates a data-driven analytical framework using machine learning (ML) and natural language processing (NLP) to automatically detect, classify, and score ESG risks in rare earth mineral supply chains. Applied to a corpus of approximately 840 documents from five major rare earth producers across Australia, the USA, and China (2015–2025), the fine-tuned BERT classifier achieves an F1-score of 0.84. The framework detects ESG controversies an average of 127 days earlier than rating agency updates, reveals a 23.5-point composite ESG risk score gap between Chinese and Western producers, and confirms financial materiality with −3.2% average abnormal returns following controversy disclosure. All four research hypotheses are statistically supported. The study contributes a validated, sector-specific framework for investors, procurement managers, and regulators seeking improved ESG transparency in critical mineral supply chains.