Audit committees and climate risk disclosure: a transformer-based textual analysis approach

Purpose This study aims to examine whether audit committee (AC) attributes are associated with the extent of climate risk disclosure in mandated US annual reports, distinguishing between physical- and transition-risk discussion in 10-K Item 1A (Risk Factors). Design/methodology/approach Using S&P 500 firms, the study extracts Item 1A from 10-K filings and applies a transformer-based NLP classifier (ClimateBERT) to score firm-year physical- and transition-risk disclosure. Multiple regression models and alternative proxies are used to test the robustness of results. Findings The results show that AC independence, gender diversity and expertise are associated with higher physical- and transition-risk disclosure, consistent with stronger monitoring and closer review of climate-related narratives in 10-K filings. Longer committee tenure is linked to lower disclosure, suggesting that familiarity with management may soften scrutiny over time. The presence of an AC is also associated with higher disclosure. These results are stronger in carbon-intensive sectors and remain stable across multiple robustness tests. Research limitations/implications The study is limited to large US non-financial firms (S&P 500), which may restrict generalizability. The dependent variables are disclosure-based measures derived from textual classification of Item 1A and reflect how risks are communicated rather than underlying exposure. Practical implications Boards, AC chairs and investors can expect AC composition to shape the depth and clarity of climate-risk discussion in 10-K risk factors. Independence, gender mix and expertise are linked to fuller narratives that help users assess physical hazards and transition pressures. Regulators may also view AC features as part of the reporting system that helps explain cross-firm differences in climate-risk communication under a common US regime. Originality/value The study connects AC design to climate-risk disclosure in Form 10-K filings by using transformer-based text analytics that separate physical from transition risk within Item 1A. This provides fresh evidence on how reporting oversight relates to emerging-risk narratives and offers a scalable method for governance–disclosure research.