Assessing the Predictive Power of ESG Scores on Company Default Probability

Abstract

For many investors risk avoidance is a feature they build their portfolios upon. Investing in a solvent company with prudent governance and fair practices over a longer term can make these investors feel secure about their future profits. A key principle of selecting a sustainable company is through ESG. ESG ratings evaluate Environmental Social and Governance practices of a business to be compared and recent research suggests that ESG ratings affect company performance. Treating the probability of default as a key measure of company performance, the research questions arise: Does ESG affect the probability of default, and can it be a significant predictor of insolvency?

To address these questions, the study first turns to the key theories. Legitimacy theory is examined to see how companies may use sustainable reporting to legitimize their own activities for more favorable investor sentiment. The modern Stakeholder theory is reviewed against the more classical Shareholder theory to view the arguments of both sides to understand whether a business should actively promote corporate social performance. These theories help to understand the background behind ESG frameworks and the push for business to do good. The research also shows that ESG influences corporate performance and firm value in particular.

This paper views the probability of default as the main metric of company performance and solvency. The probability of default is measured employing the methodology developed by Altman, a widely used Z score. The Z score focuses on several key financial ratios to evaluate the default risk of a company. The research approach is to treat Z scores as dependent variables and ESG scores as independent ones, employing the Ordinary Least Squares regression method. The second part is to take the same data and use machine learning, specifically gradient boosting, to evaluate the feature impact of ESG scores on default probabilities. Financial data and ESG scores come from the Refinitiv database. Using this data Z scores are calculated independently. The data focuses on the financial metrics of 346 companies from 12 countries from 2001 until 2022.

The results show that ESG scores affect the probability of default negatively and significantly, although the magnitude of this effect is close to zero. The second part of the research using the Xgboost functions shows that ESG can be an impactful factor in predicting insolvency. The results demonstrate that inclusion of ESG in the probability of default models can significantly improve the accuracy of these models. This research suggests that although ESG participation alone does not reduce insolvency risk, its inclusion improves predictive power, offering valuable insights for investors, companies, and policymakers.