Detecting Greenwashing in Sustainability Disclosures: A Prediction Model for KOSPI 200 Enterprises using ESG-BERT

This research is centered on the development of a BERT-based metric for greenwashing, designed to address the existing limitations inherent in ESG assessment methodologies. Unlike standard assessments, ESG-BERT considers real-time policy details and reduces the risk of inaccurate evaluations and greenwashing. We employed ESG-BERT along with financial and environmental data to predict greenwashing among Korean KOSPI200 companies. By using advanced machine learning models like ANN, LR, RF, and XGB, the study found that XGB performs best in predicting greenwashing. Furthermore, the study compares greenwashing predictions between companies with top5 and bottom5 ESG ratings. The results showed better performance for the top 5 companies compared to the bottom 5 companies.