Evaluating climate-risk language in 10-K filings

: a ClimateBERT-driven study of firm valuation

Abstract

This thesis presents a transformer-based approach to quantify and price corporate climate-risk disclosures. By utilizing ClimateBERT, I classify Item 1A risk-factor text in S&P 500 10-K filings (200532024) to generate annual transition, physical, and general climate-risk scores for 8,001 firm-year observations. Fixed-effects regressions link these lagged scores to Tobin’s Q, revealing a significant negative valuation effect for transition-risk language, especially in high-exposure sectors (Utilities; Transportation & Warehousing), while physical-risk impacts primarily arise within the same industries. By combining advanced NLP with rigorous panel econometrics, this study provides detailed, sector-sensitive metrics that illuminate how investors value different aspects of corporate climate-risk disclosure.

Date of Award	8 Jul 2025
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Julien Fouquau (Supervisor)