How can artificial intelligence-driven language analysis models enhance the detection of greenwashing tactics in corporate ESG disclosures and what influence does this detection have on the stock market?
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- In recent years, Environmental, Social and Governance disclosures have become increasingly important for companies and investors worldwide. However, the credibility of these reports is often undermined by a phenomenon called “greenwashing”. This occurs when firms exaggerate or misrepresent their sustainability efforts. This thesis explores how artificial intelligence, particularly language analysis models, can enhance the detection of greenwashing in corporate ESG disclosures, and how such detection influences investor behaviour. A new Greenwashing Score is developed based on five key linguistic indicators: buzzword density, lack of specificity, sentiment tone, hedging language, and candor. This composite score is computed through rule-based natural language processing (NLP) techniques in R and challenged through an AI co-analyst using GPT-4. This gives a hybrid methodology that combines interpretability with advanced language understanding. The second part of the study examines the financial impact of greenwashing by analyzing cumulative abnormal returns (CARs) following ESG report disclosures. Event study regressions test the relationship between greenwashing risk and stock market reactions across short- and medium-term event windows (CAR2, CAR5, and CAR60). By linking AI-driven textual analysis to market outcomes, this research contributes to the growing field of sustainable finance and highlights the potential of AI tools in promoting transparency and accountability in ESG communication. It also provides investors and regulators with a data-driven method to identify potentially misleading disclosures and encourage more authentic sustainability reporting.