Decoding ESG risk ratings

: insights into financial sector sustainability by using machine learning

Abstract

With an emphasis credit institutions, this thesis explores the factors that influence environmental, social, and governance (ESG) risk ratings in the financial industry. ESG ratings are now an important tool to assess how financial institutions handle risks associated with sustainability because of the growing social and regulatory demands for transparency and sustainable practices. In addition to bank-specific attributes like market capitalization, the variables of the dataset of financial institutions were organized along the three ESG pillars. In order to guarantee model suitability, data preparation required the methodical handling of missing values and the transformation of relevant variables. To evaluate the predictive power of various machine learning techniques in determining the most significant ESG drivers, the study used ordinary least squares (OLS), LASSO, decision trees, random forests, and gradient boosting. The findings show that Random Forest and Decision Tree models tended toward overfitting, whereas Gradient Boosting and OLS provide the best balanced performance across training and test datasets. Governance-related elements, particularly the existence of a CSR sustainability committee and board diversity, were found to be consistently influential across models by feature importance analysis. OLS specifically highlighted market capitalization, underscoring the significance of financial size. Environmental factors like waste reduction programs and renewable clean energy products also emerged as important drivers. The results validate the usefulness of ESG risk ratings as a tool for both financial risk management and sustainable development, and they are consistent with regulatory frameworks like the EU Taxonomy and international sustainability goals.

Date of Award	14 Oct 2025
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Pedro Afonso Fernandes (Supervisor)