Analyzing the Value and Bias in ESG Information
Innovative approach using NLP to evaluate ESG information from various media sources and assess relationship with stock price movements
Overview
ESG information doesn’t exist in a vacuum — different media sources report on the same company’s environmental, social, and governance practices with very different emphases and biases. This project used NLP to systematically measure those biases and explore whether they affect stock prices.
Research Questions
We asked two main questions. First, do different media outlets exhibit systematic biases when covering ESG topics? (For example, do some sources disproportionately emphasize environmental issues while downplaying governance concerns?) Second, is the way ESG information is framed in media coverage meaningfully related to stock price movements?
Approach
Using multi-source text analysis, we collected ESG-related articles from diverse Korean media outlets and built classification models to identify coverage patterns across E, S, and G dimensions. We then used statistical analysis to examine correlations between media ESG sentiment, coverage bias patterns, and contemporaneous stock price changes.
Findings
The research found measurable differences in how media outlets cover ESG topics, suggesting that investors who rely on a single source for ESG information may develop an incomplete picture of a company’s ESG profile. The relationship between ESG media coverage and stock movements also varied by the type of ESG factor being discussed.
Collaborators
- Aju Research Institute of Corporate Management: Research partner
- Cheju Halla University: Technical analysis and NLP development