From corporate earnings calls to social impact: Exploring ESG signals in S&P 500 ESG index companies through transformer-based models

Abstract

This study introduces a comprehensive framework for extracting and analyzing Environmental, Social, and Governance (ESG)-related content from earnings call transcripts of companies in the S&P 500 ESG index using natural language processing (NLP) technologies. By systematically identifying ESG discussions, the framework discovers that ESG themes are prevalent and evolve over time, reflecting corporate priorities and external pressures. ESG content was extracted using the Refinitiv-Sustainable Leadership Monitor's framework and summarized through Bidirectional Encoder Representations from Transformers (BERT)-based summarization and topic modeling employed for clustering and trend analysis. The results indicate that ESG-related topics are influenced by global events and are closely correlated with the S&P 500 ESG index. Additionally, a similarity matrix analysis highlights the strong interconnection between ESG elements and corporate performance, particularly in financial management and social responsibility. Causality analysis further demonstrates that social factors significantly impact the S&P 500 ESG index, suggesting that companies prioritizing social responsibility, transparency, and communication can improve both their ESG scores and financial performance. These findings underscore the strategic importance of integrating ESG initiatives for long-term corporate success.