Anais
Resumo do trabalho
Finanças · Governança Corporativa, Risco e Compliance
Título
From Sustainability to Solvency: ESG’s Role in Predicting Financial Distress in Energy Companies
Palavras-chave
Financial Distress
ESG
Machine learning
Agradecimento:
We thank CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
Autores
-
Anderson Dias BritoFaculdade de Economia, Administração e Contabilidade da Universidade de São Paulo - FEA
-
Leandro dos Santos MacielFaculdade de Economia, Administração e Contabilidade da Universidade de São Paulo - FEA
-
José Roberto Ferreira SavoiaFaculdade de Economia, Administração e Contabilidade da Universidade de São Paulo - FEA
Resumo
Introdução
Predicting financial distress is critical for investors, creditors, and managers as it highlights firms’ ability to meet obligations and supports economic stability. Models began with qualitative assessments in the 1930s and advanced with Altman’s (1968) Discriminant Analysis, later evolving to logistic regression and machine learning. However, gaps remain in integrating ESG variables into corporate distress prediction.
Problema de Pesquisa e Objetivo
Despite growing interest in ESG, its influence on corporate financial stability—especially in South America’s energy sector—remains uncertain. This study examines how environmental, social, and governance practices affect creditworthiness of listed electric utilities. The objective is to quantify the individual and combined impacts of ESG pillars on distress prediction using machine learning techniques
Fundamentação Teórica
Drawing on Stakeholder Theory, ESG practices strengthen ties with investors, customers, and communities, mitigating operational and reputational risks. Environmental factors cover resource efficiency and access to green finance; Social includes employee retention and customer loyalty; Governance entails transparency and reduced agency conflicts. Prior research shows high ESG scores correlate with lower capital costs and reduced insolvency risk.
Metodologia
Sample: 30 electric utilities listed in South America (2016–2023), yielding 776 observations after data cleaning. Variables: firm size (SIZE), return on assets (ROA), leverage (LEV), liquidity (LIQ), asset turnover (TURNOVER), plus Refinitiv ESG scores (ENV, SOC, GOV). Models: logistic regression, SVM, XGBoost, neural networks, Random Forest, and a bagging ensemble. Evaluation: AUC, DeLong tests, Type I/II errors, sensitivity, specificity, MAE, and RMSE
Análise dos Resultados
Incorporating ESG variables improved AUC and accuracy across all models, with governance yielding the largest performance gains, followed by social and environmental pillars. The bagging ensemble achieved the best performance (test AUC = 0.9705) and significantly reduced Type II errors, enhancing sensitivity to distress cases. DeLong tests confirmed these improvements were statistically significant.
Conclusão
ESG factors add significant predictive value to solvency models, complementing traditional financial metrics. Governance stands out as the primary early-warning driver, followed by social and environmental dimensions. Integrating ESG into machine learning enhances early alert capabilities, aiding more informed decisions by investors, managers, and regulators.
Contribuição / Impacto
This study is the first to evaluate individual and combined effects of ESG pillars on distress prediction in the South American electric utility sector using robust statistical tests (DeLong) and advanced machine learning. It offers practical guidance for investors in portfolio adjustment, managers in policy enhancement, and policymakers in designing incentives to strengthen financial stability and sustainability.
Referências Bibliográficas
Altman, E. I. (1968). Financial Ratios, Discriminant Analysis And The Prediction Of Corporate Bankruptcy. The Journal of Finance, 23(4), 589–609.
Barboza, F., Kimura, H., & Altman, E. (2017). Machine learning models and bankruptcy prediction. Expert Systems with Applications, 83, 405–417.
Donaldson, T., & Preston, L. E. (1995). The Stakeholder Theory of the Corporation: Concepts, Evidence, and Implications. Academy of Management Review, 20(1), 65–91
Barboza, F., Kimura, H., & Altman, E. (2017). Machine learning models and bankruptcy prediction. Expert Systems with Applications, 83, 405–417.
Donaldson, T., & Preston, L. E. (1995). The Stakeholder Theory of the Corporation: Concepts, Evidence, and Implications. Academy of Management Review, 20(1), 65–91