A Review of Statistical and AI Methods for Predicting ESG Risks for Default

Abstract

In the context of financial stability, understanding the risk of default is crucial for both investors and institutions. This study examines the role of Environmental, Social, and Governance (ESG) factors in predicting the risk of corporate default, integrating statistical and artificial intelligence (AI) methods. Carrying out a review of empirical studies retrieved 221 papers, of which 31 are related to this topic. We identify how ESG risks affect the likelihood of default across sectors. Statistical methods like panel regression and EGARCH models offer interpretability for linear relationships, while AI techniques such as LSTM neural networks and natural language processing (NLP) excel in capturing non-linear patterns and dealing with unstructured data. A comparison reveals that environmental risks are highly correlated with systemic default in sensitive industries, social risks disrupt operational stability, and governance risks amplify agency costs. These findings underscore the need for integrated ESG-disclosure frameworks to enhance risk management for financial institutions and regulators.