Comparative Firm-Level Analysis of ESG and Sentiment Effects on Stock Market Performance: Evidence from 16 Companies Using Retrofitted Sentiment Embeddings

This study investigates how emotion-specific sentiment embedded in financial news headlines interacts with firm-level ESG (Environmental, Social, and Governance) ratings to influence stock return behavior. Using data from 16 firms and integrating retrofitted word embeddings with ESG metrics, the analysis evaluates three hypotheses: (1) emotion-specific sentiment independently predicts stock returns; (2) its moderating effect varies across ESG dimensions; and (3) positive (negative) sentiment amplifies (dampens) ESG performance effects. The empirical results support Hypotheses 1 and 2. Emotion-laden headlines -- especially those expressing trust, sadness, and anticipation -- significantly correlate with return variability, particularly when aligned with specific ESG dimensions such as Environmental and Social. However, results for Hypothesis 3 are mixed: while sentiment clearly moderates ESG-return relationships, contradictory interactions (e.g., negative sentiment enhancing returns) were more common and exhibited stronger effects than theory-consistent ones. Retrofitted word embeddings outperformed the traditional lexicon-based sentiment model (NRC Emotion Lexicon), offering higher explanatory power across interaction models. The study highlights the nuanced and context-dependent nature of ESG-sentiment dynamics, emphasizing the importance of emotional framing in investor behavior and ESG valuation frameworks.