Investor Shelf

Machine Learning Based ESG Disclosure Sentiment and Financial Performance: Corporate Sustainability and Behavioural Finance Implications from Nifty 50 Firms (International Journal For Multidisciplinary Research)

Save
Research paperInternational Journal For Multidisciplinary ResearchNovember 202630 pages45 min readA Binita Singha, Moonmoon Kalita, Biju Mani DasOriginal source
Key number
56.3%
ESG statements with negative tone p.1
Largest segment
61.1%
Highest negative share by ESG dimension (Social) p.1

Summary

  • Most ESG disclosures from fifteen Nifty 50 firms have a negative tone (56.3%), indicating a culture of risk disclosure.p.1
  • Negative sentiment is highest in social disclosures (61.1%), followed by governance (57.8%) and environmental (50.0%).p.1
  • The study draws on 810 ESG statements from fifteen Nifty 50 firms for FY2023–FY2025, classified with machine learning models.p.3
  • The support vector machine was the best classifier, with an accuracy of 70%.p.19
  • ESG sentiment correlates positively with financial performance: r = 0.337 with ROA and r = 0.439 with ROE.p.21
  • Sun Pharma records the most positive ESG sentiment, with 50.0% positive statements.p.12

Key insights

  1. BRSR reporting became mandatory for the 1,000 largest companies by market cap from financial year 2022-23, including all Nifty 50 firms.p.2
  2. Apollo and Wipro show the highest share of negative statements at 72.2%.p.12
  3. NTPC and ONGC also show high negative sentiment at 68.5%.p.13
  4. Sun Pharma has the lowest governance negative sentiment among the firms studied, at 16.7%.p.15
  5. In the OLS model, ESG sentiment has a significant positive effect on ROA (β = 0.123) and ROE (β = 0.193).p.22
  6. The explained variance is 33.1% for ROA and 20.5% for ROE in the overall ESG sentiment models.p.22
  7. Sub-dimension coefficients are unstable because E, S and G scores are nearly perfectly multicollinear, so those hypotheses are inconclusive.p.25

Pages

Funders mentioned

Topics and entities

Summary and figures generated from the document and checked automatically: every number shown appears on the cited page. Publishers can request removal.