A Causal AI Approach to Identifying the Financial Determinants of ESG Ratings

Jian-Fa Li

Keywords: ESG, Causal AI; DoWhy; Causal inference; Financial determinants

TL;DR: Using Causal AI (DoWhy), this study finds that key financial metrics—EPS, pre-tax net income per share, return on operating assets, quick ratio, debt ratio, and net asset value per share—causally drive ESG ratings.

Abstract: Sustainable development has become a priority for firms and investors, with environmental, social, and governance (ESG) ratings now widely used to assess non-financial performance and sustainability risk. Using Causal Artificial Intelligence (Causal AI) through the DoWhy framework, this study identifies financial variables that causally influence ESG ratings for listed and over-the-counter firms in Taiwan. Earnings per share (EPS) and pre-tax net income per share emerge as the primary causal drivers of overall ESG ratings and the environmental dimension. The social dimension is causally influenced by return on operating assets, quick ratio, and debt ratio, while the governance dimension is driven by return on operating assets and net asset value per share. These relationships remain robust across multiple refutation tests, demonstrating the effectiveness of Causal AI in providing causal—rather than correlational—evidence on the financial determinants of ESG performance.

Submission Number: 43