"Innovative Approaches to Bank Default Prediction: The Integration of ESG Factors and Machine Learning" by Alberto Citterio

Topic

The predictive power of Environmental, Social, and Governance (ESG) indicators for forecasting financial distress in banks is analyzed using a sample of 362 commercial banks based in the United States and EU-28 member states over the period from 2012 to 2019. The results indicate that incorporating ESG factors enhances the model’s predictive capability in accurately identifying distressed banks. Notably, ESG significantly reduces the likelihood of misclassifying distressed or defaulted banks as financially healthy. The model, estimated through six alternative approaches—including traditional statistical methods, machine learning techniques, and ensemble methods—offers insights relevant both for practical application by banking sector supervisors and for the academic literature on default prediction.