Trust rhetoric and CEO gender
Abstract
EconStor is a publication server for scholarly economic literature, provided as a non-commercial public service by the ZBW.
Full text
Breuer, Wolfgang; Knetsch, Andreas; Salzmann, Astrid Juliane Article — Published Version Trust rhetoric and CEO gender Review of Financial Economics Provided in Cooperation with: John Wiley & Sons Suggested Citation: Breuer, Wolfgang; Knetsch, Andreas; Salzmann, Astrid Juliane (2023) : Trust rhetoric and CEO gender, Review of Financial Economics, ISSN 1873-5924, Wiley, Hoboken, NJ, Vol. 41, Iss. 3, pp. 322-344, https://doi.org/10.1002/rfe.1181 This Version is available at: https://hdl.handle.net/10419/288098 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc-nd/4.0/
322 | Rev Financ Econ. 2023;41:322–344.wileyonlinelibrary.com/journal/rfe 1 | INTRODUCTION Experimental evidence demonstrates that women are generally perceived to be more trustworthy than men (e.g., King et al.,1991). Even among researchers, beliefs about higher moral standards of women in the workplace and female executives in particular are widespread (e.g., Ford & Richardson,1994). In the realm of finance, Aggarwal et al.(2015) demonstrate that female microfinance lenders are considered to be more trustworthy. There is, however, little empirical evidence as to whether financial market participants perceive female executives and their communication as more trustworthy than that of their male counterparts. To the best of our knowledge, the study which comes closest to analyzing this issue is that of De Amicis et al.(2021), who do not observe that stock market reactions to the positivity in earnings conference call tone differ depending on the gender of the executive holding the call. However, tone positivity and trustrelated communication are distinct concepts, as only the latter relies on valuesrelated statements. Moreover, it is another matter as to whether female executives and their communication actually are more trustworthy. The experimental evidence on differences in the propensity of females and males to lie is mixed (e.g., Clot et al.,2014; Conrads et al.,2017; Ezquerra et al.,2018). The same is true for the empirical results on gender differences in terms of unethical behavior in the workplace (Kish- Gephart et al.,2010). Against the backdrop of this ambiguity, it is surprising Received: 1 February 2022 | Revised: 17 November 2022 | Accepted: 11 January 2023 DOI: 10.1002/rfe.1181 ORIGINAL ARTICLE Trust rhetoric and CEO gender WolfgangBreuer | AndreasKnetsch | Astrid JulianeSalzmann This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2023 The Authors. Review of Financial Economics published by Wiley Periodicals LLC on behalf of University of New Orleans. Department of Finance, RWTH Aachen University, Aachen, Germany Correspondence Andreas Knetsch, Department of Finance, RWTH Aachen University, Templergraben 64, 52056 Aachen, Germany. Email: [email protected] Funding information Deutsche Forschungsgemeinschaft, Grant/Award Number: 409307532 Abstract This study investigates the perceived and actual trustworthiness of female managers when using rhetoric to advertise their trustworthiness in public disclosure documents. We find that the stock market reacts more favorably to trust rhetoric if the document has been prepared under the responsibility of a female CEO rather than a male CEO. We rule out that this result could be driven by female and male CEOs talking about trust in different contexts. However, inconsistently with the notion that the trust rhetoric of women managers is more truthful than that of their male counterparts, trust rhetoric does not relate to less extensive earnings manipulation if such rhetoric stems from female CEOs compared to male CEOs. Our results thus do not confirm the popular view that higher female trustworthiness explains gender differences in accounting behavior. KEYWORDS earnings management, female CEOs, stock market reactions, textual analysis, trust JEL CLASSIFICATION D53, D91, G14, G41, G30, M41
| 323 BREUER et al. that research on financial reporting shows rather unequivocally that women executives behave in a more ethical way, when it comes to their accounting choices, namely earnings management (Barua et al.,2010; Ho et al.,2015; Peni & Vahamaa,2010), tax avoidance (Dyreng et al.,2010), or financial misreporting (Gupta et al.,2019). Given that managers can suffer legal consequences from aggressive accounting, differences in risk preferences between men and women rather than differences in trustworthiness are a likely explanation for female managers refraining from misrepresenting the financial situation of the firm to their own advantage (e.g., Francis et al.,2015; Zalata et al.,2019). This paper investigates the issue of perceived and actual female trustworthiness in the context of financial reporting by analyzing managerial rhetoric in 10- K filings. Recent literature has unveiled the potential of textual analysis of public disclosure documents for identifying managerial misconduct. Managers that represent themselves as trustworthy do so to conceal their opportunistic intentions and to mislead investors (e.g., Breuer et al.,2020; Loughran et al.,2009). This line of inquiry allows us to investigate the issue at hand in a realworld setting where lying seems to be (largely) free of risk, as the use of trust rhetoric in legal documents does not give rise to the threat of legal litigation (e.g., Huang,2005) and managers seem to be rather agnostic of the potential reputational concerns of deceptive trust rhetoric (Breuer et al.,2020). Investigating trust rhetoric as cheap talk enables us to contribute to the issues of perceived and actual trustworthiness of female executives by disentangling the effects of female ethics and female risk aversion on managerial behavior. Following prior literature, we measure the extent of trust rhetoric by counting the number of trustrelated words in the Management Discussion and Analysis (MD&A) section of 10- K filings. We relate this metric to abnormal shortterm returns around the date of the 10- K release to answer the question of whether investors perceive rhetoric that advertises the management's trustworthiness as more trustworthy when the firm is run by a female as compared to a male CEO. We employ discretionary accruals as a correlate of actual trustworthiness and test whether female CEOs manipulate reported earnings less extensively than their male counterparts when using trust rhetoric. Our analysis relies on different econometric approaches to ensure that CEO gender causally moderates the relationships between trust rhetoric and shortterm stock returns or discretionary accruals. We namely use panel regression models to control for unobserved firmlevel heterogeneity, matched samples capturing moderator effects between CEO gender and other observable factors, and a differencein- difference framework that compares male- to- female and female- to- male CEO turnovers to male- to- male CEO turnovers. Our results show that investors indeed react more positively to female rather than male CEOs' use of trust rhetoric. Even though additional analyses indicate that female CEOs aim their trust rhetoric more toward investors than male CEOs do, we rule out that this difference in female and male trust rhetoric is the reason for the more positive stock market reactions to female CEOs advertising their trustworthiness. Moreover, we also rule out that the differing reactions to female trust rhetoric are due to female CEOs discussing trustrelated issues more depending on the firm's CSR or operating performance or due to female CEOs talking more about trust in the context of other topics typically discussed in the MD&A section, which we extract using unsupervised machine learning. Overall, this finding contributes to the literature documenting that women are perceived to be more trustworthy in finance and accounting (e.g., Aggarwal et al.,2015; Shaub,1996). Our results do, however, not confirm that CEO gender moderates the relationship between trust rhetoric and earnings manipulation. We find no evidence that female CEOs can be trusted more than their male counterparts to disclose more reliable accounting information, when they advertise their trustworthiness in public disclosure documents. This is consistent with riskaversion rather than higher female morals being the reason for female managers refraining more from deceptive accounting behavior than their male counterparts (e.g., Francis et al.,2015; Gupta et al.,2019; Zalata et al.,2019). Overall, the issue of female managers' trustworthiness is important, since trust is a central concept in the economic literature. It affects economic (Knack & Keefer,1997) and financial (Guiso et al.,2008) development and, ultimately, the wealth of nations (Gur,2015). A key factor in the stock market reaction to a company's release of accounting information is the perceived credibility of that information (Pevzner et al.,2015). Even though women executives are still rare (e.g., Huang & Kisgen,2013), trust toward female managers in financial markets should not be based on stereotypes. Firms are finding themselves under growing pressure to promote the number of women in senior management (Adams & Ferreira,2009). Basing these demands on false assumptions about women's “virtues” might undermine female managers' reputation and ultimately efforts toward gender equality in management. This paper proceeds as follows: Section2 reviews the relevant literature on textual analysis in financial reporting as well as on gender differences in trustworthiness and develops our hypotheses. Section3 describes our dataset and measurement techniques. Section4 presents the empirical results. Section5 concludes.
324 | BREUER et al. 2 | LITERATURE REVIEW AND HYPOTHESES DEVELOPMENT 2.1 | Trust rhetoric in financial reporting A growing body of the finance and accounting literature has started to use textual analysis to identify deceptive behavior or intentions. Overall, this work suggests that managerial rhetoric which advertises a management's trustworthiness indicates opportunistic or unethical behavior: Loughran et al.(2009) find that the use of ethicsrelated words in 10- K filings is more common among firms that are subject to class action lawsuits, have poor corporate governance, or are socalled “sin stocks”. Larcker and Zakolyukina(2012) identify financial misreporting based on extremely positive emotion words, references to general knowledge, or shareholder value in conference calls. Hope and Wang(2018) demonstrate the robustness of these findings in the context of big bath accounting. Breuer et al.(2020) connect the use of trustrelated words to managerial opportunism. However, the cited research has not yet documented a significant stock market reaction to the use of trust- or ethicsrelated terms. This is consistent with investors being unaware of the potential for revealing managerial opportunism as well as investors being successfully lulled by managers into overlooking the latter's opportunistic intentions. Trust rhetoric therefore provides us with an opportunity to investigate female trustworthiness in a setting where lying appears to be (almost) riskless. Contrary to the use of positive tone (Rogers et al.,2011), trust rhetoric in public disclosure documents is extremely unlikely to be considered “material information”, and therefore it does not warrant legal litigation (e.g., Huang,2005). Another potential cost for managers could be a loss in reputation, which they suffer in the case that trust rhetoric misrepresents their true intentions. However, not only do investors fail to recognize trust rhetoric as an indicator of opportunistic intentions, but managers appear to ignore potential reputational costs involved with deceptive trust rhetoric (to a large degree) (Breuer et al.,2020). Based on these findings, it is safe to assume that even if there are any reputational costs from contradicting trust rhetoric by later actions, managers do not seem to care about these reputational concerns in an adequate way. Previous literature investigating gender differences in financial reporting finds that the earnings management of female executives is more conservative (Barua et al.,2010; Ho et al.,2015; Peni & Vahamaa,2010), and that female executives are less likely to engage in financial misreporting (Gupta et al.,2019). Given that managers can suffer legal consequences from aggressive accounting, differences in risk preferences between men and women are a likely explanation for female managers refraining from misrepresenting the financial situation of the firm to their own advantage (e.g., Francis et al.,2015; Zalata et al.,2019). Our analysis addresses the oftenconjectured alternative explanation that female managers behave more ethically by using trust rhetoric as a means of cheap talk in order to disentangle both potential explanations. 2.2 | Female trustworthiness The literature from management, economics, and psychology acknowledges significant genderbased differences in behavior and decisionmaking. One fundamental aspect of how the behavior of females and males is often suggested to differ is that of ethics and specifically trustworthiness. Trustworthiness is defined as an agent's preference to reciprocate another party's actions, which render that party vulnerable to opportunistic behavior by the agent (e.g., Alós- Ferrer & Farolfi,2019). One aspect of trustworthiness is truthfulness, which is an individual's preference to refrain from misrepresenting the truth to her or his own advantage, or in other words, the trustworthiness of an individual's communication. Experimental evidence supporting that women are perceived as more trustworthy than men at least dates back to Wright and Sharp(1979). King et al.(1991) show that defective behavior is typically attributed to men, while women are credited with cooperation. The views that women are more concerned with ethical behavior in the workplace than men (Ford & Richardson,1994) or that female executives have higher moral standards (Bernardi & Arnold,1997) are widely adopted. Empirical research in finance and accounting confirms the bias toward women being perceived as more trustworthy (e.g., Aggarwal et al.,2015; Shaub,1996). Of special importance to our study is evidence that demonstrates that women elicit more favorable reactions to valuesrelated statements (Wei & Ran,2019), like trust rhetoric. CEOs influence a firm's performance through their ability (e.g., Chang et al.,2010; Demerjian et al.,2012) as well as the extent of their opportunism at the cost of investors (e.g., Core et al.,1999; Gompers et al.,2003) and they often possess information about whether the firm's current valuation is justified (e.g., Chalmers et al.,2002; Kahle,2000; Lee,1997). By asking investors to trust them, CEOs claim that investing in the firm will turn out advantageous. Even though
| 325 BREUER et al. trustworthiness in a narrow sense only pertains to managers' honest intentions and truthfulness, i.e., the extent to which opportunism is absent, perceived competence, expertise, skills, or ability are also widely acknowledged to be antecedents of trust and thus aspects of trustworthiness in a wider sense (e.g., Caldwell & Clapham,2003; Mayer et al.,1995; McAllister,1995). Trust rhetoric thus is an attempt by managers to convince investors of their honest intentions, abilities, and positive private information, which all will ultimately increase shareholder wealth. Based on the evidence according to which women and their communication are perceived to be more trustworthy, we conjecture that market participants will also be more convinced when female instead of male CEOs ask investors to trust in them. In this case, trust rhetoric would be interpreted as a more (credible) positive signal of the management's intentions, abilities, and private information, if it comes from female CEOs as compared to male CEOs. We therefore expect stock market reactions to trust rhetoric to be conditional on CEO gender. Hypothesis 1. Trust rhetoric leads to more positive stock market reactions if the firm is run by a female compared to a male CEO. However, it is a different issue as to whether female CEOs also behave in a more trustworthy way than their male counterparts. Evolutionary pressure from greater intrasexual reproductive competition among men (Lee et al.,2017) or differences in socialization that promote kind, concerned attitudes in women and aggressive behavior in men (Heilman,2001) are potential reasons for women behaving more ethically. The higher ethical standards of women could explain why female managers are less likely to manipulate corporate disclosures (e.g., Gupta et al.,2019; MacLeod Heminway,2007). However, empirical results on gender differences in terms of unethical behavior in the workplace are mixed (Kish- Gephart et al.,2010). The experimental work that analyzes gender differences regarding trustworthiness largely confirms that women are more trustworthy (e.g., Croson & Buchan,1999). This is at least in part due to women feeling a stronger sense of obligation to reciprocate (e.g., Buchan et al.,2008). Recently, realworld evidence from lending relationships has corroborated these findings (e.g., Shahriar et al.,2020). Experimental studies on differences between genders regarding the special field of truthfulness, i.e., the trustworthiness of communication, are not conclusive. While some work observes that either women (Clot et al.,2014) or men (Conrads et al.,2017) are more prone to lying, many studies do not confirm any gender difference regarding trustworthiness of communication (e.g., Childs,2012; Dreber & Johannesson,2008; Ezquerra et al.,2018; Gylfason et al.,2013). Reasoning that trust rhetoric indicates opportunistic managers' intentions to mislead investors, Breuer et al.(2020) find that the use of trust words relates positively to the exploitation of latitudes in accounting standards. The use of discretionary accruals for selfish motives is widely regarded as unethical (Jha,2019). Even though CFOs have a greater influence on earnings management than CEOs (Jiang et al.,2010), CEOs also play an important role in this regard as highlighted by their career concerns, incentives, and even personal characteristics – like gender or marital status – determining the extent of discretionary accruals (e.g., Bergstresser & Philippon,2006; Cheng & Warfield,2005; Davidson III et al.,2007; Hilary et al.,2016; Ho et al.,2015). Given that the Sarbanes- Oxley Act requires the CEO as well as the CFO to certify financial reports, the relevance of both types of managers for discretionary accruals is not surprising. We are interested in whether the statements of female CEOs regarding their trustworthiness can be trusted more than those of male CEOs. By testing whether female CEOs engage to a lesser degree in unethical accounting decisions when advertising their trustworthiness, we hope to shed light on this issue. If female CEOs' communication is more trustworthy, their trust rhetoric should coincide to a lesser degree with earnings management than in the case of male CEOs. This argument is based on the assumption that trust rhetoric is not always an indicator of whether managers are trying to mislead investors. It may have a different, more truthful, meaning depending on the manager's gender. If female CEOs' trust rhetoric is truly more trustworthy than that of their male counterparts, the correlation between trust rhetoric and earnings management, with the latter being another means of deceiving market participants, should be weaker. Hypothesis 2. Trust rhetoric relates in a less positive way to the absolute value of discretionary accruals if the firm is run by a female CEO compared to a male CEO. Even though CEOs can gain from manipulating reported profits as well as from convincing investors of their trustworthiness, earnings management poses a risk to managers, whereas trust rhetoric is cheap talk. We can therefore answer the question of whether female CEOs are more trustworthy in a scenario where lying is riskless by testing Hypothesis2.
326 | BREUER et al. Following, e.g., Bergstresser and Philippon(2006), we use the absolute value of discretionary accruals for two reasons. First, CEOs do not only stand to gain from inflating but also from reducing reported earnings (e.g., Perry & Williams,1994). Second, the principle of accrual accounting requires every manipulation of reported earnings to be reversed in another period, which makes earnings manipulation an exercise of “earnings timing”. 3 | DATASET Our procedure to quantify trust rhetoric follows Audi et al.(2016): We download all 10- K reports from the Securities and Exchange Commission's EDGAR website for the years 1997– 2015. 10- K filings belong to the most important reports that managerial boards use to inform outsiders about a firm's operations. We concentrate our linguistic investigation on the MD&A section of the 10- K reports. Here, the management discusses past firm performance and outlines future plans and expectations. The MD&A section is subject to little regulation and not audited, giving the management the opportunity to describe its firm's operations, conduct, and prospects in its own words (e.g., Audi et al.,2016; Breuer et al.,2020). Hence, it serves as an ideal platform for communicating managerial trustworthiness to investors. The proxy for the amount of trust that managers try to convey is the number of words related to trust in a firm's MD&A section. The 21 trust words comprise “accountability”, “character”, “ethics”, “ethical”, “ethically”, “fairness”, “honest”, “honesty”, “integrity”, “respect”, “respected”, “respectful”, “responsible”, “responsibility”, “responsibilities”, “transparency”, “trust”, “trusted”, “truth”, “virtue”, and “virtues”. We delete all “respect to” phrases and capitalized trust words except those at sentence beginnings to ensure that the word does not relate to a financial asset or company name. We count the total number of occurrences for each trust word in the 10- K filing of firm i for year t and divide it by the total word count in the MD&A section. For our measure of trustworthiness ( Trust i,t) , we multiply this ratio by 1000 in order to obtain meaningful coefficient estimates. Consistent with prior work (Audi et al.,2016; Breuer et al.,2020), only 0.13‰ of all words in the MD&A sections of our sample are trust words. We obtain all other data for the firms in our dataset from Thomson Reuters. We match this data with the 10- K filings automatically based on CIK numbers and then manually verify the quality of this match. We exclude all firmyear observations with total assets below $10 million, as we expect that public disclosure documents from very small firms receive limited attention. We also exclude observations that have an MD&A section with 200 or fewer words, which mainly incorporate a reference to the annual report. Multivariate tests confirm that these observations are not significantly different from those that remain in our sample. Our full dataset includes 18,173 firmyear observations from 1869 firms over the years 1997 to 2015. However, many of our analyses also control for the firms' corporate social responsibility (CSR) performance, since CSR performance, for one, is correlated with top manager gender (e.g., Glass et al.,2015; Manner,2010) and, for another, could be an important determinant of how trustworthy managers are perceived and behave (e.g., Lins et al.,2017). When we exclude those observations for which information on CSR performance is missing, our sample reduces to 5583 observations from 935 firms over the years 2001 to 2015. Faced with the dilemma that we substantially reduce our sample size by controlling for CSR performance, but CSR performance being a potentially important correlate, we estimate all our regression models for our full sample as well as the reduced CSR sample, which requires that information on CSR performance is available. Note, however, that since we use CSR as an independent variable in all analyses, reducing our sample based on whether information on CSR performance is available is a case of exogenous sample selection, which does not introduce a selection bias or any other statistical problems than a smaller size (Wooldridge,2013). Table1 provides more detailed information on our sample construction. 10- K filings are prepared under the direct supervision of a firm's executive management. Public disclosure documents exhibit a significant CEO- fixed effect (Davis et al.,2015). CEOs are the most prominent management figures and shape outsiders' perception of a firm's management. We, therefore, focus our analysis on the CEO's gender using a dummy variable indicating whether a firm is run by a female CEO ( FemaleCEOi,t ). We obtain our data on executive gender from the ExecuComp database. In order to increase the number of observations for our reduced CSR sample, we manually obtain information on manager gender and turnover dates if the ExecuComp database does not cover a firmyear observation for which we have CSR performance data. We base our manual research on the Refinitiv database, which provides biographies for the most important managers of many firms. In order to ensure that the current CEO had an opportunity to assert her or his influence over the rhetoric in the 10- K filing, we require that she or he was not only in office on the day of the 10- K filing but also at least 30 days prior to that. Our final sample includes 526 observations with a female CEO. This corresponds to about 3% of our sample and is thus in line with prior studies on listed US firms (e.g., De Amicis et al.,2021; Liu,2021; McCarthy et al.,2017; Yuan et al.,2019).
| 327 BREUER et al. We measure investor reactions to the release of 10- K reports as the cumulative abnormal return over a threeday event window centered around the release date of the 10- K filing belonging to year t ( CARi,t ). Daily abnormal returns are calculated on the basis of the Fama– French threefactor model estimated over an event window of 252 trading days ending six trading days prior to the 10- K release date. We quantify the use of discretionary accruals ( DiscAcci,t ) following Capalbo et al.(2018), who utilize a modified version of the Dechow and Dichev(2002) model, which extends the Jones(1991) model to account for prior year, current year, and subsequent year cash flow from operating activities. We estimate the following regression model separately for each industryyear with at least 20 observations, based on the Fama– French 48 industry classification: TotalAccrualsi,t is the change in current assets minus the change in cash and shortterm investments, minus the change in current liabilities excluding changes of longterm debt in current liabilities, and minus depreciation and amortization over year t. AvgTAi,t is the average of total assets over years t and t- 1. ΔNetSalesi,t is the increase in net sales over year t. PropPlanEquipi,t represents property, plant, and equipment. CFOi,t is total cash flow from operating activities. We obtain the measure of how much reported earnings suffer from manipulation (| | DiscAcc i,t| |) , as the absolute value of the residuals ( 𝜀i,t ) from these regressions. Also following Capalbo et al.(2018) as well as others (e.g., Burgstahler et al.,2006; Stubben,2010), we exclude firms in the insurance and financial services industry from our analyses on discretionary accruals. Our multivariate regression models control for a variety of linguistic characteristics of the MD&A Section of the 10- K filing as well as for various firm characteristics. As one linguistic aspect, we consider the total MD&A length ( Wordsi,t ). It serves as a measure of disclosure readability (e.g., Li,2010), since Loughran and McDonald(2014) show that measures of document size outperform other common measures of disclosure readability or opacity. We also control for the use of positive words ( FinPosi,t ) as well as words that refer to uncertainty ( FinUnci,t ) in the MD&A section based on the word lists from Loughran and McDonald(2011). Even though the use of positive terms, especially in the MD&A section, only conveys very limited information (Loughran & McDonald,2011), we want to rule out that our results are driven by potential stock market reactions to rhetoric based on positive language rather than trust rhetoric. Examples of positive words are “efficient”, “profitable”, or “upturn”. We include words relating to uncertainty, like “depend”, “indefinite”, or “uncertain”, since female managers might be prone to describe the firm's prospects with another extent of certainty, as they are generally more risk averse when it comes to financial reporting (Barua et al.,2010; Gupta et al.,2019), and investors might react to this language, because it is an effective proxy for the certainty of the firm's prospects (e.g., Loughran & McDonald,2013). As for Trusti,t , we base our measures for FinPosi,t and FinUnci,t on the MD&A section, since we expect the use of rhetoric to be more pronounced in this part of the 10- K filing due to the MD&A section being subject to little (1) TotalAccruals i,t AvgTAi,t =𝛽1⋅ 1 AvgTAi,t +𝛽2⋅ ΔNetSales i,t AvgTAi,t +𝛽3⋅ PropPlanEquip i,t AvgTAi,t +𝛽4⋅ CFOi,t−1 AvgTA i,t +𝛽5⋅ CFOi,t AvgTA i,t +𝛽6⋅ CFOi,t+1 AvgTA i,t +𝜀i,t. TABLE 1 Sample construction Dropped Firmyear observations remaining Number of 10- K files downloaded for years 1997– 2015 166,679 Automated match of CIK number with Thomson Reuters database 101,195 65,484 Drop observations with total assets < $10 million 14,077 51,407 Drop if MD&A section words <201 6919 44,488 Manual control of CIK and Thomson Reuters database match 468 44,020 Drop if information on CEO gender missing 16,116 27,904 Drop if regression variables other than CSRi,t missing: Full sample 9731 18,173 Drop if CSRi,t missing: Reduced CSR sample 12,590 5583 Note: This table describes the procedure of our sample construction beginning with downloading 10- K reports from the Securities and Exchange Commission's EDGAR website.
328 | BREUER et al. regulation and it arguably being the part where managers most likely discuss potentially uncertain prospects (Loughran & McDonald,2016). We also consider the frequency with which the MD&A section refers to shareholder value. Larcker and Zakolyukina(2012) show that fewer such references indicate poor accounting quality and negative abnormal stock performance (Larcker & Zakolyukina,2012). We thus use a wordlist derived by Larcker and Zakolyukina(2012) to count the number of references to the firm's shareholder value. This list contains 15 phrases such as “shareholder value”, “value for our stockholders”, or “investor welfare”, which we detail in Table2. The resulting variable SHValRefi,t is the number of these phrases in the firm's MD&A section divided by its total wordcount. Moreover, numerous other topics could be discussed in the MD&A section, to which stock market participants might react or which could be related to accounting quality. We thus apply an unsupervised machine learning approach to determine relevant topics in 10- K filings called “latent Dirichlet allocation” (LDA), which has recently gained popularity in analyzing financial documents such as 10- K filings (e.g., Bao & Datta,2014; Bybee et al.,2020; Hanley & Hoberg,2019; Israelsen,2014). Following prior literature, we preprocess the MD&A section by lowercasing all words and removing punctuation, other symbols, and numbers. As is customary, we remove all words that are common in the English language but do not have meaning on their own, like articles or conjunctions, and words that are very commonly used in the MD&A section, but do not add to the underlying content of interest of the topics, like “millions”, “billions”, or “percent”. Moreover, we delete the “trust” words as well as the words from the shareholder value word list from the document, since we already captured both topics using the bag of words approach. By our choice, the LDA algorithm does not consider words that are contained in more than 80% of all documents. This ensures that our topics are not dominated by common words, which are relevant for most topics and thus do not add to a meaningful interpretation, like “costs”, “assets”, or “sales”. We estimate the LDA in the “quanteda” package of the opensource software “R”, where we employ the Gibbs sampler. The algorithm yields the most frequent words for a specified number of topics that are most commonly discussed in the M&DA sections of our sample as well as the extent to which these topics are referred to within each MD&A section, called “topic weights”. Again following prior finance literature using LDA to analyze 10- K filings, we specify that the algorithm reduces the documents to 25 topics (Bao & Datta,2014; Hanley & Hoberg,2019; Israelsen,2014), which should be enough to make the topics distinguishable from one another. Table S1 of the Online Appendix displays the 25 most relevant words for each topic extracted from the LDA sorted according to the weights that the words have within the respective topic. LDA does not provide labels for the extracted topics. Even though the presented word lists allow for an easy interpretation of most topics, we refrain from labeling the topics, as we only want to control for their potential effect on stock markets or correlation to discretionary accruals and our analysis does therefore not rely on an accurate interpretation of the topics' semantical meaning. We achieve this by adding the topic weights as control variables and present corresponding summary statistics in TableS2 of the Online Appendix. The set of firm control variables includes the firm's investments in fixed assets ( Investmenti,t ), measures of the firm's growth opportunities ( SalesGrowthi,t and MTBi,t ), metrics that capture the financial situation of the firm ( CashHoldingsi,t , ROAi,t , CashFlowi,t , Leveragei,t , and IndAdjROAi,t ), firm size ( Assetsi,t ), and institutional ownership ( IOi,t ) as well as analyst coverage ( ACi,t ) in order to reflect firm governance. As already mentioned, we estimate a specification of this model where we control for CSR performance ( CSRi,t ) and one where we do not control for CSRi,t . In the models where CARi.t is the dependent variable, we also control for the firm's abnormal stock performance ( PreReleaseAlphai,t ) as well as unexplained stock volatility ( PreReleaseRMSEi,t ) prior to the 10- K release, which we obtain from the Fama– French threefactor model estimated to calculate the daily abnormal returns (e.g., Audi et al.,2016; Loughran & McDonald,2011). If | |DiscAcci,t| | is the dependent variable, we add the absolute value of the change in ROAi,t from year t– 1 to t ( | | Δ ROAi,t| | ) as well as a dummy variable indicating whether ROAi,t is negative in both year t and year t– 1 ( negROAi,t ) (e.g., Klein,2002). Table2 provides definitions and Table3 descriptive statistics for all our variables. Using a ttest that controls for unequal sample variances, we do not observe significant differences in the use of trust words ( Trusti,t ), document length ( Wordsi,t ), 10- K release return ( CARi,t ), or the use of discretionary accruals ( | |DiscAcci,t| | ) between the female and male sample means. Female CEOs are on average working for firms with better CSR performance ( CSRi,t ). The language used by female CEOs differs from that of their male counterparts in that the former employ a less uncertain ( FinUnci,t ) tone. This study does, however, not investigate the linguistic differences between female and male CEOs per se. We analyze whether stock market participants react differently to trust rhetoric depending on CEO gender and whether these differing reactions would be justified.
| 329 BREUER et al. TABLE 2 Variable definitions Variable Definition FemaleCEOi,t A dummy variable indicating whether the individual holding the CEO position on the date of the 10- K release and in the 30 days prior to its release is a woman PostTransi,t A dummy variable which is zero during the tenure of the outgoing CEO and 1 during the tenure of the incoming CEO FemaleTransiA dummy variable indicating the incoming CEO is female Trusti,t Total number of trust words used in the MD&A section belonging to year t × 1000/total number of words used in the same MD&A section CARi,t The cumulative abnormal return over a threeday event window centered around the 10- K release day in percentage points. See Section3 for details |DiscAcci,t| Absolute value of discretionary accruals measured following Capalbo et al.(2018). See Section3 for details Wordsi,t Natural logarithm of the total number of words used in the MD&A section belonging to year t FinPosi,t Number of positive words in the MD&A section belonging to year t/total number of words used in the same MD&A section. The list of positive words is taken from Bill McDonald's webpage FinUnci,t Number of uncertainty words in the MD&A section belonging to year t/total number of words used in the same MD&A section. The list of positive words is taken from Bill McDonald's webpage SHValRefi,t The total number of occurrences of the phrases “shareholder value”, “shareholder welfare”, “shareholder wellbeing”, “value for our shareholders”, “value for shareholders”, “stockholder value”, “stockholder welfare”, “stockholder wellbeing”, “value for our stockholders”, “value for stockholder”, “investor value”, “investor welfare”, “investor wellbeing”, “value for our investors” or “value for investors” in the MD&A section belonging to year t/total number of words used in the same MD&A section Investmenti,t Purchase of fixed assets in year t/total assets in year t−1 SalesGrowthi,t (Net sales in year t/net sales in year t−1) − 1 MTBi,t (Market capitalization end of fiscal year t + total debt in year t)/total assets in year t−1 CashHoldingsi,t Cash and shortterm investments in year t/total assets in year t. ROAi,t (Operating Income in year t + Depreciation and Amortization in year t)/total assets in year t- 1. CashFlowi,t Cash flow from operating activities in year t/total assets in year t−1 Assetsi,t Natural logarithm of total assets in year t Leveragei,t Book value of total debt in year t/Book value of total assets in year t IndAdjROAi,t ROAi,t – median of ROAi,t in firm i's Fama– French 48- industry in year t CSRi,t Simple average of the social and environmental corporate social responsibility scores for year t AnalystCoveragei,t The yearly average of the number of earnings forecasts given for firm i per month. If the number of earnings forecasts is not provided, we set it to zero InstitutionalOwnershipi,t The percentage of shares held by 13- F investors in the firm at the end of year t PreReleaseAlphai,t Measure of abnormal stock performance prior to the 10- K release. Calculated as the intercept from the Fama– French regression prior to the respective 10- K release used to determine abnormal stock performance as described in Section3 PreReleaseRMSEi,t Measure of abnormal stock volatility prior to the 10- K release. Calculated as the root mean square error from the Fama– French regression prior to the respective 10- K release used to determine abnormal stock performance as described in Section3 |∆ROAi,t| Absolute value of the change in ROAi,t from year t−1 negROAi,t Dummy variable which is equal to one if ROAi,t and ROAi,t- 1 are both negative FemaleCFOi,t A dummy variable indicating whether the individual holding the CFO position on the date of the 10- K release and in the 30 days prior to its release is a woman Note: This table provides definitions for all variables used in the empirical analysis. We treat fiscal years that end between January 1 and May 31 as ending in the prior year. We deflate all financial data to 1990 using the Consumer Price Index from the Bureau of Labor Statistics.
336 | BREUER et al. TABLE 6 Alternative reasons for stock market reactions to trust rhetoric (1) (2) (3) (4) (5) (6) (7) (8) CARi,t CARi,t CARi,t CARi,t CARi,t CARi,t CARi,t CARi,t FemaleCEOi,t ∙ Trusti,t 2.160* 2.702* 2.043* 1.710* 2.074* 2.912* 2.154* 2.698* (1.194) (1.455) (1.144) (0.916) (1.227) (1.590) (1.236) (1.470) SHValRefi,t ∙ Trusti,t −1.836 −3.291 −1.073 −1.523 1.163 −0.686 (3.152) (4.284) (3.057) (5.599) (4.059) (5.364) IndAdjROAi,t ∙ Trusti,t 0.856 1.737 (1.669) (2.300) CSRi,t ∙ Trusti,t 0.012* 0.009 (0.007) (0.010) Trusti,t 0.017 0.047 −0.780** −0.751 −1.449 1.856 0.001 0.017 (0.191) (0.287) (0.383) (0.604) (3.244) (4.482) (0.181) (0.276) SHValRefi,t −2.132** −3.708** −0.497 −1.717 −2.441** −3.967*** −2.328** −4.085*** (1.026) (1.495) (1.192) (2.060) (1.049) (1.538) (0.909) (1.394) IndAdjROAi,t 0.114 0.148 0.530 0.527 0.114 0.140 0.115 0.148 (0.108) (0.116) (2.031) (2.640) (0.108) (0.118) (0.109) (0.116) CSRi,t −0.005* −0.001 (0.003) (0.005) FemaleCEOi,t −0.767*** −0.752* −0.526** −0.695 −0.777*** −0.813** 1.145 0.052 (0.264) (0.391) (0.247) (0.456) (0.261) (0.395) (1.416) (1.758) FemaleCEOi,t ∙ FinPosi,t −0.425 −0.050 (1.105) (1.569) FemaleCEOi,t ∙ FinUnci,t −1.031* −0.509 (0.571) (0.707) FinPosi,t −0.043 −0.172 −0.116 0.334 −0.034 −0.159 −0.031 −0.173 (0.176) (0.246) (0.218) (0.349) (0.176) (0.248) (0.176) (0.246) FinUnci,t −0.265*** −0.277* −0.243* −0.372 −0.261*** −0.296* −0.244** −0.267 (0.098) (0.164) (0.129) (0.249) (0.097) (0.164) (0.098) (0.165) Topic weights Yes Yes Yes Yes Yes Yes Yes Yes Topic weights ∙ Trusti,t No No No No Yes Yes No No Control variables Yes Yes Yes Yes Yes Yes Yes Yes Industryfixed effects Yes No Yes No Yes No Yes No Firmfixed effects No Yes No Yes No Yes No Yes Yearfixed effects Yes Yes Yes Yes Yes Yes Yes Yes Observations 17,211 17,211 5546 5546 17,211 17,211 17,211 17,211 Within R- squared 0.021 0.029 0.023 0.026 0.021 R- squared 0.026 0.147 0.035 0.205 0.027 0.148 0.147 Number of obs. with female CEO 508 508 163 163 508 508 508 508 Number of firms 1769 1769 929 929 1769 1769 1769 1769 Note: This table presents regression results with CARi,t as dependent variable. All models include our full set of control variables. We indicate standard errors, which are clustered at the firmlevel in Columns 1, 3, 5, and 7 and heteroskedasticityrobust in Columns 2, 4, 6, and 8, in parentheses. All variables are winsorized 1% in each tail. Variable definitions are provided in Table2. ***p < .01, **p < .05, *p < 0.1.
| 337 BREUER et al. 4.1.4 | Differencein- difference framework Inspired by Huang and Kisgen(2013) as well as Li and Zeng(2019), we also employ a differencein- difference framework to compare the changes in market reactions to trust rhetoric around the appointment of a new CEO for male- to- female or female- to- male transitions with a control group of male- to- male transitions. To this end, we solely consider the latest CEO transition recorded for each firm in our data and only include transitions where both the outgoing and the incoming CEO have tenures of at least 3 years. These criteria exclude the few female- to- female transitions we have in our sample. PostTransi,t is used as a dummy variable being zero during the tenure of the outgoing CEO and one during the tenure of the incoming CEO. GenderTransi is set to zero for male- to- male transitions, it is 1 for male- to- female transitions, and it is −1 for female- to- male transitions. This way, male- to- male transitions serve us as a control group to isolate whether reactions to trust rhetoric change around CEO turnovers where the incoming CEO differs in gender from his predecessor. By female- to- male transitions having the opposite sign as male- to- female transitions, we assume that markets perceptions of trust rhetoric change to a similar degree, but that they improve for male- to- female transitions and deteriorate for female- to- male transitions. To rule out that these changes are due to changes in observable firm and 10- K characteristics, the regression includes all our control variables. GenderTransi is not part of the regression, since it is absorbed by the firmfixed effects (Huang & Kisgen,2013, or Li & Zeng,2019): PostTransi,t ⋅ GenderTransi ⋅ Trusti,t indicates whether the change in sensitivity to trust rhetoric is different for transitions that result in a change of CEO gender. Higher values indicate that trust rhetoric is perceived more favorably after male- to- female transitions or more unfavorably after female- to- male transitions. Table8 presents the results for regression Model (4), with heteroskedasticityrobust standard errors in parentheses for our full sample, covering 31 male- to- female transitions, 17 female- to- male, and 537 male- to- male transitions, as well as the reduced CSR sample in columns 1 and 2, respectively. The triple interaction term PostTransi,t ⋅ GenderTransi ⋅ Trusti,t is positive and significant in both estimations. Thus, our results confirm that stock markets change their attitude toward trust rhetoric differently if a change (4) CAR i,t =𝛼+𝛽 1⋅ PostTrans i,t +𝛽 2⋅ Trust i,t +𝛽 3⋅ Trust i,t⋅ PostTrans i,t +𝛽4⋅Trusti,t⋅GenderTransi+𝛽5⋅PostTransi,t⋅GenderTransi + 𝛽6⋅PostTransi,t⋅GenderTransi⋅Trusti,t+∑ j Controlsj,i,t+𝜏t+𝛾i+𝜀i,t . TABLE 7 CEO gender and stock market reaction to trust rhetoric – Propensity Score Matching (1) (2) (3) (4) CARi,t CARi,t Female Male Female Male Trusti,t 2.884* −3.192** 1.195 −3.312** (1.521) (1.275) (1.171) (1.551) CSRi,t −0.033** −0.034*** (0.013) (0.012) Topic weights Yes Yes No No Control variables Yes Yes Yes Yes Industryfixed effects Yes Yes Yes Yes Yearfixed effects Yes Yes Yes Yes Chisquare 9.37 5.07 pvalue .002 .024 Observations 498 498 156 156 R- squared 0.229 0.170 0.375 0.424 Note: This table presents the results of a seemingly unrelated estimation including industry- and yearfixed effects with standard errors clustered at the firmlevel over a female and matched male sample, where CARi,t is the dependent variable. Columns 1 and 2 present the results for our full sample and columns 3 and 4 for our reduced CSR sample. The models presented in columns 3 and 4 do not include the 25 topic weights as control variables. The matching procedure is described in Section4.1.2. All models include industry- and yearfixed effects. All variables are winsorized 1% in each tail. Variable definitions are provided in Table2. ***p < .01, **p < .05, *p < .1.
338 | BREUER et al. in CEO genders occurs than when a male CEO succeeds a male CEO. Other than market participants reacting differently to the trust rhetoric of female than that of male CEOs, the sole alternative explanation for this result is that of unobserved firm characteristics changing around the CEO transition for male- to- female transitions only. These results also provide strong evidence in favor of the causal relationship postulated by Hypothesis1. 4.2 | Female trust rhetoric as an indicator of trustworthiness We test Hypothesis2 with the same econometrical approaches that address the endogeneity of CEO gender outlined in Sections 4.1.1 to 4.1.3. Table9 presents the results of estimating the panel regression model from Section 4.1.1 using | |DiscAcci,t| | as the dependent variable for our full sample as well as our reduced CSR sample. The interaction term FemaleCEOi,t ⋅ Trusti,t is insignificant in both models. Untabulated results confirm a positive effect of Trusti,t on | |DiscAcci,t| | in regressions without the interaction term FemaleCEOi,t ⋅ Trusti,t (Breuer et al.,2020). We also estimate untabulated regression models considering either firm- or industryfixed effects, where we exclude all textbased control variables, to examine whether CEO and CFO gender relate to the extent of earnings management, as established by prior work. We identify a negative and significant correlation between FemaleCEOi,t and | |DiscAcci,t| | in the model with firmfixed effects and the same for the relationship between FemaleCFOi,t and | |DiscAcci,t| | in the model with industryfixed effects. These results provide weak confirmation of previous findings and further substantiate that not only CFO (e.g., Barua et al.,2010) but also CEO (e.g., Ho et al.,2015) gender may affect earnings management. TABLE 8 CEO gender and stock market reaction to trust rhetoric – Differencein- difference estimation (1) (2) CARi,t CARi,t Trusti,t ∙ PostTransi,t ∙ GenderTransi4.557** 5.251** (1.796) (2.542) Trusti,t ∙ GenderTransi2.197 −0.881 (1.829) (2.387) Trusti,t ∙ PostTransi,ti −0.228 −0.045 (0.563) (0.639) PostTransi,t ∙ GenderTransi−0.232 −0.487 (0.503) (0.722) PostTransi,t −0.090 0.025 (0.218) (0.253) Trusti,t −0.735 −0.122 (0.462) (0.613) CSRi,t 0.000 (0.006) Topic weights Yes Yes Control variables Yes Yes Firmfixed effects Yes Yes Yearfixed effects Yes Yes Observations 6813 2795 Within R- squared 0.030 0.055 R- squared 0.140 0.205 Number of obs. with female CEO 219 103 Number of firms 643 405 Note: This table presents results for a differencein- difference regression where the treatment groups are male- to- female and female- to- male CEO transitions and the control group is male- to- male CEO transitions. Column 1 presents the results for our full sample and column 2 for our reduced CSR sample. The models include firm- and yearfixed effects and we indicate heteroskedasticityrobust standard errors in parentheses. All variables are winsorized 1% in each tail. Variable definitions are provided in Table2. ***p < .01, **p < .05, *p < .1.
| 339 BREUER et al. TABLE 9 CEO gender, trust rhetoric, and earnings manipulation (1) (2) (3) (4) (5) (6) |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| FemaleCEOi,t ∙ Trusti,t −0.031 −0.009 −0.032 −0.045 (0.019) (0.016) (0.020) (0.041) FemaleCEOi,t −0.003 −0.007 0.001 −0.002 (0.004) (0.006) (0.007) (0.007) FemaleCFOi,t ∙ Trusti,t 0.020 −0.012 0.020 −0.012 (0.020) (0.015) (0.020) (0.015) FemaleCFOi,t −0.005 0.006 −0.004 0.006 (0.005) (0.006) (0.005) (0.006) Trusti,t 0.006 0.013** −0.001 0.018* −0.000 0.019* (0.005) (0.006) (0.006) (0.010) (0.006) (0.010) Wordsi,t −0.001 −0.002 −0.004 −0.003 −0.004 −0.003 (0.002) (0.003) (0.003) (0.004) (0.003) (0.004) FinPosi,t 0.003 0.008* 0.000 0.004 0.000 0.003 (0.003) (0.004) (0.004) (0.005) (0.004) (0.005) FinUnci,t −0.001 −0.000 −0.005* −0.005 −0.005* −0.005 (0.002) (0.004) (0.003) (0.004) (0.003) (0.004) SHValRefi,t −0.017 −0.029 −0.025 −0.014 −0.025 −0.014 (0.020) (0.034) (0.026) (0.026) (0.026) (0.027) Investmenti,t 0.018 0.018 0.031 −0.001 0.031 −0.002 (0.012) (0.027) (0.020) (0.035) (0.020) (0.035) SalesGrowthi,t 0.003*** 0.002 0.004** 0.001 0.004** 0.001 (0.001) (0.003) (0.002) (0.003) (0.002) (0.003) MTBi,t 0.000 −0.001 0.001* −0.001 0.001* −0.001 (0.001) (0.001) (0.001) (0.002) (0.001) (0.002) CashHoldingsi,t 0.006 0.018 0.001 0.030 0.001 0.031* (0.008) (0.014) (0.011) (0.019) (0.011) (0.019) ROAi,t 0.013 0.046** 0.010 0.041 0.010 0.041 (0.009) (0.020) (0.012) (0.028) (0.012) (0.029) CashFlowi,t 0.002 −0.016 0.001 0.001 0.001 0.001 (0.009) (0.020) (0.012) (0.021) (0.012) (0.021) Assetsi,t −0.001 0.002 0.002 0.007* 0.002 0.007* (0.002) (0.003) (0.002) (0.004) (0.002) (0.004) Leveragei,t 0.001 0.002 0.002 0.005 0.002 0.005 (0.006) (0.014) (0.008) (0.017) (0.008) (0.018) IndAdjROAi,t −0.024*** −0.056*** −0.023*** −0.051** −0.023*** −0.051** (0.004) (0.016) (0.004) (0.022) (0.004) (0.022) CSRi,t −0.000 −0.000 −0.000 (0.000) (0.000) (0.000) AnalystCoveragei,t 0.000 0.000 −0.000 0.000 −0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) InstitutionalOwnershipi,t 0.001 0.004 −0.006 −0.007 −0.006 −0.007 (0.009) (0.015) (0.012) (0.017) (0.012) (0.018) (Continues)
340 | BREUER et al. Given the aforementioned relevance of CFOs on earnings management (Jiang et al.,2010), we also test whether CFO gender – rather than CEO gender – moderates the relationship between | |DiscAcci,t| | and Trusti,t . Like for CEO gender, we obtain information on CFO gender from ExecuComp and through manual research using Refintiv and construct the variable FemaleCFOi,t analogously. Since information on CFO gender is rarer than on CEO gender in both databases, we only manage to identify the gender of 4762 CFOs for the sample considered in our analyses on discretionary accruals. 8.8% of these CFOs are female, which is higher than in the case of CEOs and in line with prior work on listed US firms (e.g., Barua et al.,2010; Gupta et al.,2019). We estimate additional models which include FemaleCFOi,t as well as FemaleCFOi,t and its interaction with Trusti,t alternatively to FemaleCEOi,t and FemaleCEOi,t ⋅ Trusti,t for our full sample as well as our CSR sample in columns 3 and 4 of Table9. Like for CEO gender, we find no evidence of CFO gender moderating the relationship between Trusti,t and | |DiscAcci.t| | . When considering CEO and CFO gender simultaneously, the moderator effects of both CEO as well as CFO gender remain insignificant in columns 5 and 6 of Table9. We also find no significant difference in how CEO gender moderates the relationship between trust rhetoric and earnings manipulation over a female and a matched male sample or when adopting the outlined differencein- difference approach. We present the respective regression results for our full sample as well as our reduced CSR sample in TablesS3– S4 of our Online Appendix. In none of these cases, there is evidence in support of Hypothesis2. Overall, our findings cannot be interpreted as female CEOs being more trustworthy than their male counterparts in a scenario where deception is largely free of risk. 5 | CONCLUSION We present evidence that CEO gender causally moderates the effect of trust rhetoric in the MD&A section of 10- K filings on the respective 10- K release returns. Firms with female CEOs experience more positive announcement returns to the use of trust rhetoric. This outcome is consistent with rhetoric that advertises the management's trustworthiness being perceived as more convincing if the financial statement is prepared under the responsibility of a female CEO. We corroborate the results from prior literature which support the notion that females are perceived as more trustworthy (e.g., Aggarwal et al.,2015; King et al.,1991). Moreover, we show that the beliefs about female trustworthiness have practical implications for financial reporting. We also rule out that the more positive market reactions to female trust rhetoric are due to the fact that female CEOs talk about trust in different contexts than their male counterparts do. (1) (2) (3) (4) (5) (6) |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| |DiscAcci,t| |∆ROAi,t| 0.025*** 0.032*** 0.023*** 0.035** 0.023*** 0.035** (0.006) (0.012) (0.008) (0.015) (0.008) (0.015) negROAi,t 0.013*** 0.009* 0.010* 0.003 0.010* 0.003 (0.004) (0.005) (0.005) (0.005) (0.005) (0.005) Topic weights YES YES YES YES YES YES Firmfixed effects YES YES YES YES YES YES Yearfixed effects YES YES YES YES YES YES Observations 7945 2789 4762 1765 4762 1765 Within R- squared 0.039 0.052 0.049 0.069 0.049 0.070 R- squared 0.348 0.321 0.384 0.372 0.384 0.373 Number of obs. with female CEO 183 79 89 30 89 30 Number of firms 1191 517 775 380 775 380 Note: This table presents regression results with |DiscAcci,t| as dependent variable. Column 1 describes the results for our full sample and column 2 for our reduced CSR sample. Both models include firm- and yearfixed effects. Standard errors are heteroskedasticityrobust. All variables are winsorized 1% in each tail. Variable definitions are provided in Table2. ***p < .01, **p < .05, *p < .1. TABLE 9 (Continued)
| 341 BREUER et al. However, we cannot confirm that trust rhetoric relates to less exploitation of latitude in accounting standards if it stems from female CEOs rather than male CEOs. Hence, our results provide no reason for investors to place greater trust in the financial information presented by female managers who affirm their trustworthiness in the same disclosure document. Overall, we cannot support the widespread notion that female executives are more trustworthy than their male counterparts. Since the use of trust rhetoric is mostly free of risk, these results contribute to disentangling the effect of female trustworthiness from the alternative explanation for more conservative accounting choices of female managers: higher risk aversion of women (e.g., Gupta et al.,2019). We would have expected that female trust rhetoric is a better indicator of actual trustworthiness, if female CEOs actually behaved in a more trustworthy manner. Our results do, however, not lend support to this notion. This implies that the observed differences in accounting choices by gender, where managers take risks by behaving deceptively, can primarily be explained by higher female risk aversion rather than higher ethical standards (e.g., Francis et al.,2015; Zalata et al.,2019). It is worth mentioning that our analysis, like other work which investigates gender differences among CEOs, naturally relies on only relatively few sample observations with female CEOs. With the caveat of this limitation, our findings indicate that female managers seem to use trust rhetoric to lull investors into complacency, just as their male counterparts do. However, investors are more willing to buy into the rhetoric if it stems from women. This finding on perceived female trustworthiness contrasts with studies which confirm discrimination against female managers based on negative stereotypes about their competence (e.g., Bigelow et al.,2014; Niessen- Ruenzi & Ruenzi,2018). ACKNOWLEDGMENT Open Access funding enabled and organized by Projekt DEAL. FUNDING INFORMATION This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – 409307532. We would like to thank two anonymous journal referees, whose reviews have greatly improved the paper. ORCID Andreas Knetsch https://orcid.org/0000-0002-7656-0755 REFERENCES Adams, R., & Ferreira, D. (2009). Women in the boardroom and their impact on governance and performance. Journal of Financial Economics, 94, 291– 309. Aggarwal, R., Goodell, J. W., & Selleck, L. J. (2015). Lending to women in microfinance: Role of social trust. International Business Review, 24, 55– 65. Alós- Ferrer, C., & Farolfi, F. (2019). Trust games and beyond. Frontiers in Neuroscience, 13, 887. Audi, R., Loughran, T., & McDonald, B. (2016). Trust, but verify: MD&a language and the role of trust in corporate culture. Journal of Business Ethics, 139, 551– 561. Bao, Y., & Datta, A. (2014). Simultaneously discovering and quantifying risk types from textual risk disclosures. Management Science, 60, 1371– 1391. Barua, A., Davidson, L. F., Rama, D. V., & Thiruvadi, S. (2010). CFO gender and accruals quality. Accounting Horizons, 24, 25– 39. Bergstresser, D., & Philippon, T. (2006). CEO incentives and earnings management. Journal of Financial Economics, 80, 511– 529. Bernardi, R. A., & Arnold, D. F. (1997). An examination of moral development within public accounting by gender, staff level, and firm. Contemporary Accounting Research, 14, 653– 668. Bigelow, L., Lundmark, L., Parks, J. M., & Wuebker, R. (2014). Skirting the issues: Experimental evidence of gender bias in IPO prospectus evaluations. Journal of Management, 40, 1732– 1759. Breuer, W., Knetsch, A., & Salzmann, A. J. (2020). What does it mean when managers talk about trust? Journal of Business Ethics, 166, 473– 488. Buchan, N. R., Croson, R., & Solnick, S. (2008). Trust and gender: An examination of behavior and beliefs in the investment game. Journal of Economic Behavior & Organization, 68, 466– 476. Burgstahler, D. C., Hail, L., & Leuz, C. (2006). The importance of reporting incentives: Earnings Management in European Private and Public Firms. The Accounting Review, 81, 983– 1016. Bybee, L., Kelly, B. T., Manela, A., & Xiu, D. (2020). The structure of economic news. Working paper. Caldwell, C., & Clapham, S. E. (2003). Organizational trustworthiness: An international perspective. Journal of Business Ethics, 47, 349– 364.
342 | BREUER et al. Capalbo, F., Frino, A., Lim, M. Y., Mollica, V., & Palumbo, R. (2018). The impact of CEO narcissism on earnings management. Abacus, 54, 210– 226. Chalmers, J. M. R., Dann, L. Y., & Harford, J. (2002). Managerial opportunism? Evidence from directors' and officers' insurance purchases. The Journal of Finance, 57, 609– 636. Chang, Y. K., Dasgupta, S., & Hilary, G. (2010). CEO ability, pay, and firm performance. Management Science, 56, 1633– 1652. Chen, X., Harford, J., & Li, K. (2007). Monitoring: Which institutions matter? Journal of Financial Economics, 86, 279– 305. Cheng, Q., & Warfield, T. (2005). Equity incentives and earnings management. The Accounting Review, 80, 441– 476. Childs, J. (2012). Gender differences in lying. Economics Letters, 114, 147– 149. Clifford, C. P., & Lindsey, L. (2016). Blockholder heterogeneity, CEO compensation, and firm performance. Journal of Financial and Quantitative Analysis, 51, 1491– 1520. Clot, S., Grolleau, G., & Ibanez, L. (2014). Smug alert! Exploring selflicensing behavior in a cheating game. Economics Letters, 123, 191– 194. Conrads, J., Ellenberger, M., Irlenbusch, B., Ohms, E. N., Rilke, R. M., & Walkowitz, G. (2017). Team goal incentives and individual lying behavior. Working Paper. Core, J. E., Holthausen, R. W., & Larcker, D. F. (1999). Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics, 51, 371– 406. Cornett, M. M., Marcus, A. J., & Tehranian, H. (2008). Corporate governance and payfor- performance: The impact of earnings management. Journal of Financial Economics, 87, 357– 373. Croson, R., & Buchan, N. (1999). Gender and culture: International experimental evidence from trust games. American Economic Review, 89, 386– 391. Davidson, W. N., III, Xie, B., & Ning, Y. (2007). The influence of executive age, career horizon and incentives on preturnover earnings management. Journal of Management and Governance, 11, 45– 60. Davis, A. K., Ge, W., Matsumoto, D., & Zhang, J. L. (2015). The effect of managerspecific optimism on the tone of earnings conference calls. Review of Accounting Studies, 20, 639– 673. De Amicis, C., Falconieri, S., & Tastan, M. (2021). Sentiment analysis and gender differences in earnings conference calls. Journal of Corporate Finance, 71, 101809. Dechow, P. M., & Dichev, I. D. (2002). The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77, 35– 59. Demerjian, P., Lev, B., & McVay, S. (2012). Quantifying managerial ability: A new measure and validity tests. Management Science, 58, 1229– 1248. Demers, E., & Vega, C. (2014). Understanding the role of managerial optimism and uncertainty in the price formation process: Evidence from the textual content of earnings announcements. Working Paper. Dreber, A., & Johannesson, M. (2008). Gender differences in deception. Economics Letters, 99, 197– 199. Dyreng, S. D., Hanlon, M., & Maydew, E. L. (2010). The effects of executives on corporate tax avoidance. The Accounting Review, 85, 1163– 1189. Ezquerra, L., Kolev, G. I., & Rodriguez- Lara, I. (2018). Gender differences in cheating: Loss vs. gain framing. Economics Letters, 163, 46– 49. Ford, R. C., & Richardson, W. D. (1994). Ethical decision making – A review of the empirical literature. Journal of Business Ethics, 13, 205– 221. Francis, B., Hasan, I., Park, J. C., & Wu, Q. (2015). Gender differences in financial reporting decision making: Evidence from accounting conservatism. Contemporary Accounting Research, 32, 1285– 1318. Glass, C., Cook, A., & Ingersoll, A. R. (2015). Do women leaders promote sustainability? Analyzing the effect of corporate governance composition on environmental performance. Business Strategy and the Environment, 25, 495– 511. Gompers, P., Ishii, J., & Metrick, A. (2003). Corporate governance and equity prices. The Quarterly Journal of Economics, 118, 107– 156. Guiso, L., Sapienza, P., & Zingales, L. (2008). Trusting the stock market. The Journal of Finance, 63, 2557– 2600. Gupta, V. K., Mortal, S., Chakrabarty, B., Guo, X., & Turban, D. B. (2019). CFO gender and financial statement irregularities. Academy of Management Journal, 63, 802– 831. Gur, N. (2015). Trust and the wealth of nations. Progress in Development Studies, 15, 107– 124. Gylfason, H. F., Arnardottir, A. A., & Kristinsson, K. (2013). More on gender differences in lying. Economics Letters, 119, 94– 96. Hanley, K. W., & Hoberg, G. (2019). Dynamic interpretation of emerging risks in the financial sector. Review of Financial Studies, 32, 4543– 4603. Heilman, M. E. (2001). Description and prescription: How gender stereotypes prevent women's ascent up the organizational ladder. Journal of Social Issues, 57, 657– 674. Hilary, G., Huang, S., & Xu, Y. (2016). Marital status and earnings management. European Economic Review, 26, 153– 158. Ho, S. S. M., Li, A. Y., Tam, K., & Zhang, F. F. (2015). CEO gender, ethical leadership, and accounting conservatism. Journal of Business Ethics, 127, 351– 370. Hope, O.- K., & Wang, J. (2018). Management deception, bigbath accounting, and information asymmetry: Evidence from linguistic analysis. Accounting, Organizations and Society, 70, 33– 51. Huang, J., & Kisgen, D. J. (2013). Gender and corporate finance: Are male executives overconfident relative to female executives? Journal of Financial Economics, 108, 822– 839. Huang, P. H. (2005). Moody investing and the supreme court: Rethinking the materiality of information and the reasonableness of investors. Supreme Court Economic Review, 13, 99– 131.
| 343 BREUER et al. Israelsen, R. D. (2014). Tell it like it is: Disclosed risks and factor portfolios. Working paper. Jha, A. (2019). Financial reports and social capital. Journal of Business Ethics, 155, 567– 596. Jiang, J. X., Petroni, K. R., & Wang, I. Y. (2010). CFOs and CEOs: Who have the most influence on earnings management? Journal of Financial Economics, 96, 513– 526. Jones, J. J. (1991). Earnings management during import relief investigations. Journal of Accounting Research, 29, 193– 228. Kahle, K. M. (2000). Insider trading and the longrun performance of new security issues. Journal of Corporate Finance, 6, 25– 53. King, W. C., Miles, E. W., & Kniska, J. (1991). Boys will be boys (and girls will be girls): The attribution of gender role stereotypes in a gaming situation. Sex Roles, 25, 607– 623. Kish- Gephart, J. J., Harrison, D. A., & Trevino, L. K. (2010). Bad apples, bad cases, and bad barrels: Metaanalytic evidence about sources of unethical decisions at work. Journal of Applied Psychology, 95, 1– 31. Klein, A. (2002). Audit committee, board of director characteristics, and earnings management. Journal of Accounting and Economics, 33, 375– 400. Knack, S., & Keefer, P. (1997). Does social capital have an economic payoff? A crosscountry investigation. Quarterly Journal of Economics, 112, 1251– 1288. Larcker, D. F., & Zakolyukina, A. A. (2012). Detecting deceptive discussions in conference calls. Journal of Accounting Research, 50, 495– 540. Lee, I. (1997). Do firms knowingly sell overvalued equity? The Journal of Finance, 52, 1439– 1466. Lee, M., Pitesa, M., Pillutla, M., & Thau, S. (2017). Male immorality: An evolutionary account of sex differences in unethical negotiation behavior. Academy of Management Journal, 60, 2014– 2044. Li, F. (2010). The information content of forwardlooking statements in corporate filings – A naïve Bayesian machine learning approach. Journal of Accounting Research, 48, 1049– 1102. Li, Y., & Zeng, Y. (2019). The impact of top executive gender on asset prices: Evidence from stock price crash risk. Journal of Corporate Finance, 58, 528– 550. Lins, K. V., Servaes, H., & Tamayo, A. (2017). Social capital, trust, and firm performance: The value of corporate social responsibility during the financial crisis. The Journal of Finance, 72, 1785– 1842. Liu, C. (2021). CEO gender and employee relations: Evidence from labor lawsuit. Journal of Banking & Finance, 128, 106136. Loughran, T., & McDonald, B. (2011). When is a liability not a liability? Textual analysis, dictionaries, and 10Ks. Journal of Finance, 66, 35– 65. Loughran, T., & McDonald, B. (2013). IPO firstday returns, offer price revision, volatility, and form S- 1 language. Journal of Financial Economics, 109, 307– 326. Loughran, T., & McDonald, B. (2014). Measuring readability in financial disclosures. The Journal of Finance, 69, 1643– 1671. Loughran, T., & McDonald, B. (2016). Textual analysis in accounting and finance: A survey. Journal of Accounting Research, 54, 1187– 1230. Loughran, T., McDonald, B., & Yun, H. (2009). A wolf in sheep's clothing: The use of ethicsrelated terms in 10- k reports. Journal of Business Ethics, 89, 39– 49. MacLeod Heminway, J. (2007). Sex, trust, and corporate boards. Hasting's Women's Law Journal, 18, 173– 199. Manner, M. H. (2010). The impact of CEO characteristics on corporate social responsibility. Journal of Business Ethics, 93, 53– 72. Mayer, R. C., Davis, J. H., & Schoorman, F. D. (1995). An integration model of organizational trust. The Academy of Management Review, 20, 709– 729. McAllister, D. J. (1995). Affect- and cognitionbased trust as foundations for interpersonal cooperation in organizations. Academy of Management Journal, 38, 24– 59. McCarthy, S., Oliver, B., & Song, S. (2017). Corporate social responsibility and CEO confidence. Journal of Banking & Finance, 75, 280– 291. Niessen- Ruenzi, A., & Ruenzi, S. (2018). Sex matters: Gender bias in the mutual fund industry. Management Science, 65, 3001– 3025. Peni, E., & Vahamaa, S. (2010). Female executives and earnings management. Managerial Finance, 36, 629– 645. Perry, S., & Williams, T. (1994). Earnings management preceding management buyout offers. Journal of Accounting and Economics, 18, 157– 179. Pevzner, M., Xie, F., & Xin, X. (2015). When firms talk, do investors listen? The role of trust in stock market reactions to corporate earnings announcements. Journal of Financial Economics, 117, 190– 223. Rogers, J. L., Van Buskirk, A., & Zechman, S. L. (2011). Disclosure tone and shareholder litigation. The Accounting Review, 86, 2155– 2183. Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 41– 55. Shahriar, A. Z. M., Unda, L. A., & Alam, Q. (2020). Gender differences in the repayment of microcredit: The mediating role of trustworthiness. Journal of Banking & Finance, 110, 105685. Shaub, M. K. (1996). Trust and suspicion: The effects of situational and dispositional factors of auditors' trust of clients. Behavioural Research in Accounting, 8, 155– 174. Stubben, S. (2010). Discretionary revenues as a measure of earnings management. The Accounting Review, 85, 695– 717. Wei, H., & Ran, Y. (2019). Male versus female: How the gender of apologizers influences consumer forgiveness. Journal of Business Ethics, 154, 371– 387. Wooldridge, J. M. (2013). Introductory econometrics: A modern approach (5th ed.). South- Western Cengage Learning. Wright, T. L., & Sharp, E. G. (1979). Content and grammatical sex bias on the interpersonal trust scale and differential trust toward women and men. Journal of Consulting and Clinical Psychology, 47, 72– 85.
344 | BREUER et al. Yuan, Y., Tian, G., Lu, L. Y., & Yu, Y. (2019). CEO ability and corporate social responsibility. Journal of Business Ethics, 157, 391– 411. Zalata, A. M., Ntim, C., Aboud, A., & Gyapong, E. (2019). Female CEOs and core earnings quality: New evidence on the ethics versus riskaversion puzzle. Journal of Business Ethics, 160, 515– 534. SUPPORTING INFORMATION Additional supporting information can be found online in the Supporting Information section at the end of this article. How to cite this article: Breuer, W., Knetsch, A., & Salzmann, A. J. (2023). Trust rhetoric and CEO gender. Review of Financial Economics, 41, 322–344. https://doi.org/10.1002/rfe.1181