Selected Working Papers
"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.
"Cross-Industry Information Sharing among Colleagues and Analyst Research" (with An-Ping Lin and Amy Zang).
We identify a specific organizational resource in brokerage houses—information sharing among analyst colleagues who cover economically related industries along a supply chain. We hypothesize that this resource constitutes a knowledge asset that benefits analyst research and allows analysts to have a higher level of industry specialization. The impact of this resource depends on colleagues’ information complementarity, which we measure using the level of economic connection between an analyst’s industry and her colleagues’ industries. Controlling for economic determinants and brokerage selection effects, we find that colleagues’ information sharing improves the analyst’s earnings forecast accuracy, stock recommendation profitability, and investor recognition. We show that the colleagues’ coverage of downstream (upstream) industries improves the analyst’s revenue (expense) forecasts. We find that information sharing among colleagues allows analysts, especially less capable ones, to cover a more specialized research portfolio. Overall, our findings align with the theory of the firm that a firm’s organizational structures contribute to employee performance and specialization.
We identify a specific organizational resource in brokerage houses—information sharing among analyst colleagues who cover economically related industries along a supply chain. We hypothesize that this resource constitutes a knowledge asset that benefits analyst research and allows analysts to have a higher level of industry specialization. The impact of this resource depends on colleagues’ information complementarity, which we measure using the level of economic connection between an analyst’s industry and her colleagues’ industries. Controlling for economic determinants and brokerage selection effects, we find that colleagues’ information sharing improves the analyst’s earnings forecast accuracy, stock recommendation profitability, and investor recognition. We show that the colleagues’ coverage of downstream (upstream) industries improves the analyst’s revenue (expense) forecasts. We find that information sharing among colleagues allows analysts, especially less capable ones, to cover a more specialized research portfolio. Overall, our findings align with the theory of the firm that a firm’s organizational structures contribute to employee performance and specialization.
"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 committing accounting fraud before the fraud is publicly revealed as well as the economic impacts of such rating actions. We find that Standard & Poor’s (S&P), an issuer-paid rating agency, downgrades fraud firms and puts them on negative credit watch as early as four quarters prior to public fraud revelation, controlling for firms’ economic fundamentals and performance. Furthermore, Egan-Jones, an investor-paid rating agency without access to management and private information, is much less timely than S&P in downgrading fraud firms, especially when information uncertainty is high. In addition, S&P distinguishes fraud firms from similar-looking non-fraud firms even when the former appears financially healthy, whereas Egan-Jones does not. This further supports that S&P’s private information advantage facilitates fraud detection. Finally, S&P’s negative rating actions against fraud firms inform the market and are associated with management turnover and shorter fraud duration.
This study examines whether and when credit rating agencies take negative rating actions against issuers committing accounting fraud before the fraud is publicly revealed as well as the economic impacts of such rating actions. We find that Standard & Poor’s (S&P), an issuer-paid rating agency, downgrades fraud firms and puts them on negative credit watch as early as four quarters prior to public fraud revelation, controlling for firms’ economic fundamentals and performance. Furthermore, Egan-Jones, an investor-paid rating agency without access to management and private information, is much less timely than S&P in downgrading fraud firms, especially when information uncertainty is high. In addition, S&P distinguishes fraud firms from similar-looking non-fraud firms even when the former appears financially healthy, whereas Egan-Jones does not. This further supports that S&P’s private information advantage facilitates fraud detection. Finally, S&P’s negative rating actions against fraud firms inform the market and are associated with management turnover and shorter fraud duration.
"The Long-Term Consequences of Short-Term Incentives" (with Alex Edmans and Vivian Fang).
This paper studies the long-term consequences of actions induced by vesting equity, a measure of short-term concerns. Vesting equity is positively associated with the probability of a firm repurchasing shares, the amount of shares repurchased, and the probability of the firm announcing a merger or acquisition (M&A). However, it is also associated with more negative long-term returns over the 2-3 years following repurchases and 4 years following M&A, as well as future M&A goodwill impairment. These results are inconsistent with CEOs buying underpriced stock or companies to maximize long-run shareholder value, but consistent with these actions being used to boost the short-term stock price and thus equity sale proceeds. CEOs sell their own stock shortly after using company money to buy the firm’s stock, also inconsistent with repurchases being motivated by undervaluation.
This paper studies the long-term consequences of actions induced by vesting equity, a measure of short-term concerns. Vesting equity is positively associated with the probability of a firm repurchasing shares, the amount of shares repurchased, and the probability of the firm announcing a merger or acquisition (M&A). However, it is also associated with more negative long-term returns over the 2-3 years following repurchases and 4 years following M&A, as well as future M&A goodwill impairment. These results are inconsistent with CEOs buying underpriced stock or companies to maximize long-run shareholder value, but consistent with these actions being used to boost the short-term stock price and thus equity sale proceeds. CEOs sell their own stock shortly after using company money to buy the firm’s stock, also inconsistent with repurchases being motivated by undervaluation.
"FinBERT: A Pretrained Language Model for Financial Communications" (with Yi Yang and Mark Christopher Siy Uy).
Contextual pretrained language models, such as BERT (Devlin et al., 2019), have made significant breakthrough in various NLP tasks by training on large scale of unlabeled text re-sources.Financial sector also accumulates large amount of financial communication text.However, there is no pretrained finance specific language models available. In this work,we address the need by pretraining a financial domain specific BERT models, FinBERT, using a large scale of financial communication corpora. Experiments on three financial sentiment classification tasks confirm the advantage of FinBERT over generic domain BERT model. The code and pretrained models are available at this https URL. We hope this will be useful for practitioners and researchers working on financial NLP tasks.
GitHub link for Pre-Trained FinBERT
Contextual pretrained language models, such as BERT (Devlin et al., 2019), have made significant breakthrough in various NLP tasks by training on large scale of unlabeled text re-sources.Financial sector also accumulates large amount of financial communication text.However, there is no pretrained finance specific language models available. In this work,we address the need by pretraining a financial domain specific BERT models, FinBERT, using a large scale of financial communication corpora. Experiments on three financial sentiment classification tasks confirm the advantage of FinBERT over generic domain BERT model. The code and pretrained models are available at this https URL. We hope this will be useful for practitioners and researchers working on financial NLP tasks.
GitHub link for Pre-Trained FinBERT
Publications
"An Unintended Benefit of the Risk Factor Mandate of 2005" (with Jianghua Shen and Amy Zang), Review of Accounting Studies, conditionally accepted.
In 2005, the SEC began mandating that firms disclose risk factors to provide useful information on firm risk. Although not intended by regulators, mandatory RF disclosure may constitute “meaningful cautionary language” as defined in the Private Securities Litigation Reform Act, providing legal protection for forward-looking statement (FLS). Using both a difference-in-differences design and a two-stage least squares approach, we find that firms that had not previously disclosed risk factors (late RF disclosers) became more willing to provide qualitative FLS, particularly positive ones, than other firms, consistent with the mandate reducing managers’ perceived litigation risk in late RF disclosers. Cross-sectional tests show that this effect is stronger for firms whose managers perceive a higher level of benefit from safe harbor protection availed by the provision of meaningful cautionary statements. Last, we find that late RF disclosers experience improvement in their information environment, illustrating an unintended benefit of the 2005 RF mandate.
In 2005, the SEC began mandating that firms disclose risk factors to provide useful information on firm risk. Although not intended by regulators, mandatory RF disclosure may constitute “meaningful cautionary language” as defined in the Private Securities Litigation Reform Act, providing legal protection for forward-looking statement (FLS). Using both a difference-in-differences design and a two-stage least squares approach, we find that firms that had not previously disclosed risk factors (late RF disclosers) became more willing to provide qualitative FLS, particularly positive ones, than other firms, consistent with the mandate reducing managers’ perceived litigation risk in late RF disclosers. Cross-sectional tests show that this effect is stronger for firms whose managers perceive a higher level of benefit from safe harbor protection availed by the provision of meaningful cautionary statements. Last, we find that late RF disclosers experience improvement in their information environment, illustrating an unintended benefit of the 2005 RF mandate.
"Federal Judge Ideology: A New Measure of Ex-Ante Litigation Risk" (with Kai Wai Hui and Reeyarn Li), Journal of Accounting Research, May 2019, Vol. 57 Iss. 2, pp. 431-489.
Drawing on the political theory of judicial decision making, our paper proposes a new and parsimonious ex ante litigation risk measure: federal judge ideology. We find that judge ideology complements existing measures of litigation risk based on industry membership and firm characteristics. Firms in liberal circuits (the third quartile in ideology) are 33.5% more likely to be sued in securities class action lawsuits than those in conservative circuits (the first quartile in ideology). This result is stronger after the U.S. Supreme Court’s ruling in the Tellabs case. We next show that the effect of judge ideology on litigation risk is greater for firms with more sophisticated shareholders and with higher expected litigation costs. Furthermore, judicial appointments affect litigation risk and the value of firms in the circuit, highlighting the economic consequences of political appointments of judges. Finally, using our new measure, we document that litigation risk deters managers from providing long-term earnings guidance, a result that existing measures of litigation risk cannot show.
Online Appendix to Huang, Hui and Li (2019)
Federal Judge Ideology Data
Drawing on the political theory of judicial decision making, our paper proposes a new and parsimonious ex ante litigation risk measure: federal judge ideology. We find that judge ideology complements existing measures of litigation risk based on industry membership and firm characteristics. Firms in liberal circuits (the third quartile in ideology) are 33.5% more likely to be sued in securities class action lawsuits than those in conservative circuits (the first quartile in ideology). This result is stronger after the U.S. Supreme Court’s ruling in the Tellabs case. We next show that the effect of judge ideology on litigation risk is greater for firms with more sophisticated shareholders and with higher expected litigation costs. Furthermore, judicial appointments affect litigation risk and the value of firms in the circuit, highlighting the economic consequences of political appointments of judges. Finally, using our new measure, we document that litigation risk deters managers from providing long-term earnings guidance, a result that existing measures of litigation risk cannot show.
Online Appendix to Huang, Hui and Li (2019)
Federal Judge Ideology Data
"Analyst Information Discovery and Interpretation Roles: A Topic Modeling Approach" (with Reuven Lehavy, Amy Zang, and Rong Zheng), Management Science, June 2018, Vol. 64, Iss. 6, pp. 2833-2855.
This study examines analyst information intermediary roles using a textual analysis of analyst reports and corporate disclosures. We employ a topic modeling methodology from computational linguistic research to compare the thematic content of a large sample of analyst reports issued promptly after earnings conference calls with the content of the calls themselves. We show that analysts discuss exclusive topics beyond those from conference calls and interpret topics from conference calls. In addition, we find that investors place a greater value on new information in analyst reports when managers face greater incentives to withhold value-relevant information. Analyst interpretation is particularly valuable when the processing costs of conference call information increase. Finally, we document that investors react to analyst report content that simply repeats managers’ discussion. Overall, our study shows that analysts play the information intermediary roles by discovering information beyond corporate disclosures and by clarifying and confirming corporate disclosures.
Internet Appendix to Huang, Lehavy, Zang and Zheng (2018)
Java package to implement Topic Modeling (LDA) algorithm used in the paper
Another Java based package (MALLET)
Python package to implement LDA
This study examines analyst information intermediary roles using a textual analysis of analyst reports and corporate disclosures. We employ a topic modeling methodology from computational linguistic research to compare the thematic content of a large sample of analyst reports issued promptly after earnings conference calls with the content of the calls themselves. We show that analysts discuss exclusive topics beyond those from conference calls and interpret topics from conference calls. In addition, we find that investors place a greater value on new information in analyst reports when managers face greater incentives to withhold value-relevant information. Analyst interpretation is particularly valuable when the processing costs of conference call information increase. Finally, we document that investors react to analyst report content that simply repeats managers’ discussion. Overall, our study shows that analysts play the information intermediary roles by discovering information beyond corporate disclosures and by clarifying and confirming corporate disclosures.
Internet Appendix to Huang, Lehavy, Zang and Zheng (2018)
Java package to implement Topic Modeling (LDA) algorithm used in the paper
Another Java based package (MALLET)
Python package to implement LDA
"Imperfect Accounting and Reporting Bias?" (with Vivian Fang and Wenyu Wang), Journal of Accounting Research, September 2017, Vol. 55, Iss. 4, pp. 919-962.
Errors and bias are both inherent features of accounting. In theory, while errors discourage bias by lowering the value relevance of accounting, they can also facilitate bias by providing camouflage. Consistent with theory, we find a hump-shaped relation between a firm’s propensity to engage in intentional misstatement and the prevalence of unintentional misstatements in the firm’s industry for the whole economy and a majority of the industries. The result is robust to using firms’ number of items in financial statements and exposure to complex accounting rules as alternative proxies for errors and to using the restatement amount in net income to quantify the magnitude of bias and errors. To directly test for the two effects of errors, we show that, when errors are more prevalent, the market reacts less to firms’ earnings surprises and bias is more difficult to detect. Our results highlight the imperfectness of accounting, advance understanding of firms’ reporting incentives, and shed light on accounting standard setting.
Errors and bias are both inherent features of accounting. In theory, while errors discourage bias by lowering the value relevance of accounting, they can also facilitate bias by providing camouflage. Consistent with theory, we find a hump-shaped relation between a firm’s propensity to engage in intentional misstatement and the prevalence of unintentional misstatements in the firm’s industry for the whole economy and a majority of the industries. The result is robust to using firms’ number of items in financial statements and exposure to complex accounting rules as alternative proxies for errors and to using the restatement amount in net income to quantify the magnitude of bias and errors. To directly test for the two effects of errors, we show that, when errors are more prevalent, the market reacts less to firms’ earnings surprises and bias is more difficult to detect. Our results highlight the imperfectness of accounting, advance understanding of firms’ reporting incentives, and shed light on accounting standard setting.
"Short Selling and Earnings Management: A Controlled Experiment" (with Vivian Fang and Jonathan Karpoff), Journal of Finance, June 2016, Vol. 71, Iss. 3, pp. 1251-1294.
During 2005 to 2007, the SEC ordered a pilot program in which one-third of the Russell 3000 index were arbitrarily chosen as pilot stocks and exempted from short-sale price tests. Pilot firms’ discretionary accruals and likelihood of marginally beating earnings targets decrease during this period, and revert to pre-experiment levels when the program ends. After the program starts, pilot firms are more likely to be caught for fraud initiated before the program, and their stock returns better incorporate earnings information. These results indicate that short selling, or its prospect, curbs earnings management, helps detect fraud, and improves price efficiency.
List of pilot stocks published SEC's release No. 50104
Data and Codes to replicate our principal findings (README)
During 2005 to 2007, the SEC ordered a pilot program in which one-third of the Russell 3000 index were arbitrarily chosen as pilot stocks and exempted from short-sale price tests. Pilot firms’ discretionary accruals and likelihood of marginally beating earnings targets decrease during this period, and revert to pre-experiment levels when the program ends. After the program starts, pilot firms are more likely to be caught for fraud initiated before the program, and their stock returns better incorporate earnings information. These results indicate that short selling, or its prospect, curbs earnings management, helps detect fraud, and improves price efficiency.
List of pilot stocks published SEC's release No. 50104
Data and Codes to replicate our principal findings (README)
"Evidence on the Information Content of Text in Analyst Reports" (with Amy Zang and Rong Zheng), The Accounting Review, November 2014, Vol. 89, No. 6, pp. 2151-2180.
We document that textual discussions in a sample of 363,952 analyst reports provide information to investors beyond that in the contemporaneously released earnings forecasts, stock recommendations, and target prices, and also assist investors in interpreting these signals. Cross-sectionally, we find that investors react more strongly to negative than to positive text, suggesting that analysts are especially important in propagating bad news. Additional evidence indicates that analyst report text is more useful when it places more emphasis on nonfinancial topics, is written more assertively and concisely, and when the perceived validity of other information signals in the same report is low. Finally, analyst report text is shown to have predictive value for future earnings growth in the subsequent five years.
Weka: Data mining software to implement the naive Bayes algorithm
Using FinBERT trained on the 10,000 sentence from analyst reports to classify sentiments (FinBERT paper)
We document that textual discussions in a sample of 363,952 analyst reports provide information to investors beyond that in the contemporaneously released earnings forecasts, stock recommendations, and target prices, and also assist investors in interpreting these signals. Cross-sectionally, we find that investors react more strongly to negative than to positive text, suggesting that analysts are especially important in propagating bad news. Additional evidence indicates that analyst report text is more useful when it places more emphasis on nonfinancial topics, is written more assertively and concisely, and when the perceived validity of other information signals in the same report is low. Finally, analyst report text is shown to have predictive value for future earnings growth in the subsequent five years.
Weka: Data mining software to implement the naive Bayes algorithm
Using FinBERT trained on the 10,000 sentence from analyst reports to classify sentiments (FinBERT paper)
"CEO Reputation and Earnings Quality" (with Jennifer Francis, Shivaram Rajgopal and Amy Zang), Contemporary Accounting Research, Spring 2008, Vol. 25, Iss. 1, pp. 109-147.
We examine the relation between CEO reputation and measures of the firm’s earnings quality. Using press coverage (media counts) to proxy for CEO reputation, we find that more reputed CEOs are associated with poorer earnings quality. This finding is inconsistent with an efficient contracting view, which predicts that reputed CEOs take actions that result in good earnings quality. This seemingly counterintuitive result is, however, consistent with two other theories: a rent extraction hypothesis (which predicts that reputed managers are more likely to use their discretion to manipulate earnings in order to manage labor and stock market perceptions) and a matching hypothesis (which predicts that selection on the part of firms gives rise to a demand for reputed CEOs for firms where earnings quality is inherently poor). Further analyses provide little support for the rent extraction explanation and some support for the matching explanation.
We examine the relation between CEO reputation and measures of the firm’s earnings quality. Using press coverage (media counts) to proxy for CEO reputation, we find that more reputed CEOs are associated with poorer earnings quality. This finding is inconsistent with an efficient contracting view, which predicts that reputed CEOs take actions that result in good earnings quality. This seemingly counterintuitive result is, however, consistent with two other theories: a rent extraction hypothesis (which predicts that reputed managers are more likely to use their discretion to manipulate earnings in order to manage labor and stock market perceptions) and a matching hypothesis (which predicts that selection on the part of firms gives rise to a demand for reputed CEOs for firms where earnings quality is inherently poor). Further analyses provide little support for the rent extraction explanation and some support for the matching explanation.
