Zorunlu ESG açıklamalarıyla birlikte kazanç duyuru toplantılarında ton değişimi

This thesis investigates whether mandatory Environmental, Social, and Governance (ESG) disclosure regulations influence the tone of corporate communication during earnings conference calls (ECCs). Using a large global panel of transcripts from 16,327 firms across 95 countries (2005–2024), the study employs advanced natural language processing tools to quantify tone. Specifically, FinBERT-Tone is used to detect sentence-level sentiment (positive, neutral, negative), while FinBERT-ESG and ESG-BERT models classify ESG-related content across environmental, social, and governance dimensions. For each ECC, the most optimistic ESG-related sentence is extracted to construct ESG_Tone_max, E_Tone_max, S_Tone_max, and G_Tone_max indicators. A Difference-in-Differences (DiD) approach compares tone changes in firms from countries with ESG mandates (treatment group) to those without (control group), while controlling for firm, year, and sector-year fixed effects. Results indicate a statistically significant increase in social tone (S_Tone_max) following ESG mandates, whereas environmental and governance tones show no robust change. Overall ESG tone improves slightly but is sensitive to model specification. Firms with stronger ESG performance consistently use more positive ESG language. This study bridges NLP and causal inference to show that regulation affects not only what firms disclose but also how they speak about ESG, enriching the literature on corporate communication and sustainability.