Work

An Empirical Framework for Measuring Corporate Social Responsibility (CSR) Disclosure Tone and Its Financial Implications

Year: 2026
Type: article
Abstract: This study introduces a model that measures the tone of Corporate Social Responsibility (CSR) disclosures in U.S. Securities and Exchange Commission (SEC) submissions and examines its association with corporate performance. We systematically gather Form 10-K and Form 20-F filings along with related eXtensible Business Reporting Language (XBRL) information from the Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system, identify CSR-related excerpts utilising a specialised sustainability lexicon, and evaluate sentiment of the excerpts using Financial Bidirectional Encoder Representations from Transformers (FinBERT) to develop firm-year CSR tone metrics. Financial data are standardized and supplemented with key metrics (e.g., Net Margin, Return on Equity (ROE), Return on Assets (ROA), free cash flow, market capitalization). We investigate the economic significance of disclosure tone through Ordinary Least Squares (OLS) with robust errors, regularised regression techniques (Ridge Regression, Least Absolute Shrinkage and Selection Operator (Lasso), Elastic Net) utilising cross validation, and panel fixed-effects models; additional predictive analyses (Elastic Net) evaluate out-of-sample validity. Across model specifications, a more positive CSR disclosure tone is positively associated with profitability and, to a lesser degree, with valuation indicators; the fits are modest yet durable out of sample and resilient to company and year effects. All artifacts—code, intermediate datasets, and rating outputs—are generated comprehensively for the purposes of replication and enhancement. The framework provides a clear and adaptable foundation for overseeing sustainability communication and its financial relationships as disclosure practices develop.
Language: English
Open Access status: diamond
OpenAlex ID: W7225830622
FWCI:
Cites:
Cited by:
Related to: