We develop FinBERT, a state‐of‐the‐art large language model that adapts to the finance domain. We show that FinBERT incorporates finance knowledge and can better summarize contextual information in financial texts. Using a sample of researcher‐labeled sentences from analyst reports, we document that FinBERT substantially outperforms the Loughran and McDonald dictionary and other machine learning algorithms, including naïve Bayes, support vector machine, random forest, convolutional neural network, and long short‐term memory, in sentiment classification. Our results show that FinBERT excels in identifying the positive or negative sentiment of sentences that other algorithms mislabel as neutral, likely because it uses contextual information in financial text. We find that FinBERT's advantage over other algorithms, and Google's original bidirectional encoder representations from transformers model, is especially salient when the training sample size is small and in texts containing financial words not frequently used in general texts. FinBERT also outperforms other models in identifying discussions related to environment, social, and governance issues. Last, we show that other approaches underestimate the textual informativeness of earnings conference calls by at least 18% compared to FinBERT. Our results have implications for academic researchers, investment professionals, and financial market regulators.

This study examines whether and when credit ratings are useful for accounting fraud prediction. We find that negative rating actions by Standard & Poor’s (S&P), an issuer-paid credit rating agency (CRA), have predictive ability for fraud incremental to fraud prediction models (e.g., F-score) and other market participants. In contrast, rating actions by Egan-Jones Rating Company (EJR), an investor-paid CRA relying on public information, have less predictive ability, which is subsumed by S&P and other market participants. Our results are robust to including firms not covered by EJR, using only rating downgrades, controlling for firm characteristics, and using alternative benchmarks. We also find that the ability of negative S&P rating actions to predict fraud becomes stronger after the 2008–2009 financial crisis. Last, compared with EJR, S&P is quicker to take negative rating actions against fraud firms. In sum, our results suggest that issuer-paid CRAs’ information advantage helps predict accounting fraud. Data Availability: Data are available from the public sources cited in the text.