Selected Working Papers
"FinBERT--A Large Language Model Approach to Extracting Information from Financial Text" (with Hui Wang and Yi Yang).
In this paper, we develop FinBERT, a state-of-the-art large language model that adapts to the finance domain. Using a sample of researcher-labeled sentences from analyst reports, we document that FinBERT significantly 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 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 (BERT) model is especially salient when the training sample size is small. FinBERT also outperforms other models in identifying discussions related to environment, social, and governance issues and forward-looking statements. Last, we show that other approaches underestimate the textual informativeness of earnings conference calls by at least 18%, compared with FinBERT. Our results have implications for academic researchers, investment professionals, and financial market regulators who want to extract insights from financial texts.
FinBERT Sentiment Classification model with sample code
FinBERT ESG Classification model with sample code
FinBERT Forward Looking Statement Classification model with sample code
Pre-Trained FinBERT on GitHub
In this paper, we develop FinBERT, a state-of-the-art large language model that adapts to the finance domain. Using a sample of researcher-labeled sentences from analyst reports, we document that FinBERT significantly 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 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 (BERT) model is especially salient when the training sample size is small. FinBERT also outperforms other models in identifying discussions related to environment, social, and governance issues and forward-looking statements. Last, we show that other approaches underestimate the textual informativeness of earnings conference calls by at least 18%, compared with FinBERT. Our results have implications for academic researchers, investment professionals, and financial market regulators who want to extract insights from financial texts.
FinBERT Sentiment Classification model with sample code
FinBERT ESG Classification model with sample code
FinBERT Forward Looking Statement Classification model with sample code
Pre-Trained FinBERT on GitHub
"Securities Law Precedents, Litigation Risk, and Misreporting" (with Benedikt Franke, Reeyarn Li and Hui Wang).
We study securities law precedents and their impact on firms’ litigation risk and financial misreporting decisions. Using a sample of 438 circuit court precedents for alleged securities law violations from 1996 to 2018, we measure each circuit court’s leniency by its tendency to dismiss these lawsuits at each point in time and show that leniency affects future lawsuit rulings and firms’ litigation risk. Case-level analyses reveal that district courts heed the circuit court precedents and are more likely to dismiss pending cases when their home circuits dismiss more allegations. In firm-level analyses, we find that shareholders are less likely to sue misreporting firms that reside in more lenient circuits. The effect of precedents on misreporting firms’ litigation risk is more pronounced when managerial intent is difficult to judge and in the presence of sophisticated potential plaintiffs and higher expected lawsuit payoffs. Finally, consistent with lenient precedents lowering expected litigation costs, we show that investors react less negatively to the restatement announcements of firms residing in more lenient circuits and that such firms are more likely to misreport. We conclude that securities law precedents induce considerable heterogeneity in firms’ securities litigation risk, which affects their financial reporting quality.
We study securities law precedents and their impact on firms’ litigation risk and financial misreporting decisions. Using a sample of 438 circuit court precedents for alleged securities law violations from 1996 to 2018, we measure each circuit court’s leniency by its tendency to dismiss these lawsuits at each point in time and show that leniency affects future lawsuit rulings and firms’ litigation risk. Case-level analyses reveal that district courts heed the circuit court precedents and are more likely to dismiss pending cases when their home circuits dismiss more allegations. In firm-level analyses, we find that shareholders are less likely to sue misreporting firms that reside in more lenient circuits. The effect of precedents on misreporting firms’ litigation risk is more pronounced when managerial intent is difficult to judge and in the presence of sophisticated potential plaintiffs and higher expected lawsuit payoffs. Finally, consistent with lenient precedents lowering expected litigation costs, we show that investors react less negatively to the restatement announcements of firms residing in more lenient circuits and that such firms are more likely to misreport. We conclude that securities law precedents induce considerable heterogeneity in firms’ securities litigation risk, which affects their financial reporting quality.
"Credit Rating Agencies and Accounting Fraud Detection" (with Pepa Kraft and Shiheng Wang).
This study examines whether and when credit rating agencies take negative rating actions against issuers that commit accounting fraud before the fraud is publicly revealed and investigates the economic effects of such rating actions. After controlling for firms’ economic fundamentals and other sources of information, we find that Standard & Poor’s (S&P), an issuer-paid rating agency, takes negative rating actions against fraud firms, including downgrades and negative credit watches, as early as four quarters before the revelation of fraud. Egan-Jones, an investor-paid rating agency with no access to management and material nonpublic information, is slower than S&P to downgrade fraud firms, especially when information uncertainty is high and when fraud firms are subject to private SEC investigations. Last, we find that S&P’s negative rating actions against fraud firms are informative to the market and are associated with a shorter fraud duration. Taken together, our results suggest that credit rating agencies use their information advantage to detect accounting fraud.
This study examines whether and when credit rating agencies take negative rating actions against issuers that commit accounting fraud before the fraud is publicly revealed and investigates the economic effects of such rating actions. After controlling for firms’ economic fundamentals and other sources of information, we find that Standard & Poor’s (S&P), an issuer-paid rating agency, takes negative rating actions against fraud firms, including downgrades and negative credit watches, as early as four quarters before the revelation of fraud. Egan-Jones, an investor-paid rating agency with no access to management and material nonpublic information, is slower than S&P to downgrade fraud firms, especially when information uncertainty is high and when fraud firms are subject to private SEC investigations. Last, we find that S&P’s negative rating actions against fraud firms are informative to the market and are associated with a shorter fraud duration. Taken together, our results suggest that credit rating agencies use their information advantage to detect accounting fraud.
"Judge Ideology and Corporate Tax Planning" (with Travis Chow, Kai Wai Hui and Terry Shevlin).
We investigate whether and how the federal judiciary affects corporate tax planning. We find that firms engage in less aggressive tax planning when Circuit Court and Tax Court judges are more liberal. This effect is economically significant and robust across various measures of tax planning. We further detail specific tax planning tactics in response to liberal judge ideology, such as shifting less income overseas, conducting more foreign tax planning, and acquiring more auditor-provided tax services. Firms also avoid liberal judges through forum shopping. Finally, we show that IRS enforcement complements the judge ideology effect. Overall, we are the first to demonstrate the judicial branch as a key determinant of corporate tax planning, which contributes to a more complete understanding of tax enforcement.
We investigate whether and how the federal judiciary affects corporate tax planning. We find that firms engage in less aggressive tax planning when Circuit Court and Tax Court judges are more liberal. This effect is economically significant and robust across various measures of tax planning. We further detail specific tax planning tactics in response to liberal judge ideology, such as shifting less income overseas, conducting more foreign tax planning, and acquiring more auditor-provided tax services. Firms also avoid liberal judges through forum shopping. Finally, we show that IRS enforcement complements the judge ideology effect. Overall, we are the first to demonstrate the judicial branch as a key determinant of corporate tax planning, which contributes to a more complete understanding of tax enforcement.
"Judge Ideology and Opportunistic Insider Trading" (with Kai Wai Hui and Yue Zheng).
Extant evidence suggests that liberal judges prefer stricter securities enforcement to protect innocent investors. We find that firms located in circuits with more liberal judges perform fewer opportunistic insider sales, consistent with managers taking judges’ political ideology into consideration. This deterrent effect is stronger when insiders are more likely to be sued. The SEC also considers judge ideology when selecting litigation venues. Finally, we validate that liberal judges are associated with heavier penalties in insider trading cases. Overall, we provide the first evidence demonstrating the importance of judicial discretion and judge political ideology in deterring opportunistic insider trading.
Extant evidence suggests that liberal judges prefer stricter securities enforcement to protect innocent investors. We find that firms located in circuits with more liberal judges perform fewer opportunistic insider sales, consistent with managers taking judges’ political ideology into consideration. This deterrent effect is stronger when insiders are more likely to be sued. The SEC also considers judge ideology when selecting litigation venues. Finally, we validate that liberal judges are associated with heavier penalties in insider trading cases. Overall, we provide the first evidence demonstrating the importance of judicial discretion and judge political ideology in deterring opportunistic insider trading.