Using Machine Learning to Predict Corporate Environmental Violation: A Stakeholder Pressure Perspective

ABSTRACT

Predicting corporate environmental violations remains a key challenge in practice and in environmental governance research. However, existing studies have largely focused on ex post associations between stakeholder pressures and realized environmental violations, offering limited insight into whether stakeholder pressures can be used ex ante to identify firms exposed to future violation risk. Using a sample of Chinese listed firms, this study develops a prediction framework that integrates multiple stakeholder pressure indicators and uses machine learning as an empirical tool for risk identification and substantive interpretation. The results show that creditor pressure is the most influential predictor, followed by regulatory and competitive pressures. Firms subject to greater creditor pressure or stronger competition exhibit a higher probability of environmental violation, whereas stronger regulatory pressure is associated with a lower probability. Analyses of violation severity yield consistent results: Creditor pressure remains highly predictive of more severe violations, whereas stronger regulatory and public pressures are linked to fewer severe violations. The prominent role of creditor pressure highlights the need to consider financial‐constraint distortions in corporate environmental compliance, and these findings show that stakeholder pressure does not operate as a uniform disciplinary force in environmental governance.