Executive Gender and the Opportunistic Use of Non-GAAP Exclusions

(1)

Executive Gender and the Opportunistic Use of Non-GAAP Exclusions

Name: Wouter Beens Student number: 10438165 Thesis supervisor: Dr. A. Sikalidis Date: August 20, 2018

Word count: 17.209

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student Wouter Beens who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

ABSTRACT: This study investigates the association between executive gender and the opportunistic use of non-GAAP exclusions. For this purpose the impact of Chief Executive Officer (CEO) gender and Chief Financial Officer (CFO) gender is investigated, in the context of companies using income-increasing non-GAAP exclusions to meet or beat their analyst earnings forecasts. It is expected that female executives are associated with decreased opportunism in non-GAAP exclusions. This expectation is mainly based on several gender-related differences, with effects on financial decision-making. Multivariate logistic regressions indicate that female executives are associated with decreases in the opportunistic use of non-GAAP exclusions, in line with predictions. A decrease in opportunism is found both for total non-GAAP exclusions, as for the underlying special exclusions and other exclusions, being slightly stronger for special exclusions. Since early regressions with separate gender variables lack statistical significance, this paper’s main results flow from regressions using one combined gender variable. Therefore, while the associations with CEO and CFO gender seem to be behave in a similar manner, no definitive inferences are made with regard to the separate associations of CEO gender or CFO gender.

Keywords: non-GAAP disclosures; executive gender; earnings benchmarks; analyst earnings forecasts; earnings quality; earnings management; non-GAAP earnings management; accruals manipulation; real activities manipulation.

(4)

1 Introduction

Gender diversity under companies’ boards and executives currently is a widely discussed topic among regulators. In 2011, Viviane Reding, Vice-President of the European Commission and EU Commissioner for Justice, Fundamental Rights and Citizenship, published the ‘Women on the Board Pledge for Europe’, calling on publicly listed firms to sign a voluntary commitment to increase women’s presence on their boards. The debate that followed resulted in a call for national legislation among EU member-states to improve the representation of women on boards. Several European member states adopted such legislation, including gender quotas, in the years that followed (European Commission, 2012, p. 5). In April 2016, representation of women on publicly listed firms’ boards was measured to be 23,3%, against 13,7% in October 2011 (European Commission, 2016, p. 2). Despite the significant increase, the European Commission aims at improving this figure to 40% by 2020. While such regulation is already in effect in the EU, a similar discussion is ongoing in the US. However, this has not yet led to similar regulation being created in the US. All in all, these developments show the growing relevance of research on gender differences in a corporate context.

Extant literature indicates male and female executives differ in several key aspects, influencing their financial decision making. In a general sense, males are documented to have a larger risk-appetite, are more opportunistic and overconfident, and strive more for economic gain. Females are more risk-averse, more pessimistic, less confident, and strive more for security and harmony than economic gain. In a corporate setting this results in males employing more risky strategies, less conservative accounting, and being more likely to engage in earnings management. However, this association between male executives and increased earnings management is solely based on the more classical forms of earnings management: accruals earnings management and real activities manipulation. There is currently no research on whether this association with gender holds for a relatively new and controversial form of earnings management: non-GAAP earnings management.

Non-GAAP earnings are an alternative earnings figure, not subject to the General Accepted Accountings Principles that are used to determine GAAP earnings. In the creation of non-GAAP earnings incidental costs and revenues are excluded from GAAP earnings, resulting in a more permanent, recurring earnings figure. While managers argue non-GAAP earnings is primarily used to inform stakeholders of future performance, critics argue the high levels of discretion in creating non-GAAP earnings can be used opportunistically. Since the recent surge in firms disclosing non-non-GAAP earnings, researchers have increasingly researched the managerial motives behind disclosing it. Bhattachary, Black, Christensen, & Larson (2003) find non-GAAP earnings is more informative than GAAP earnings, and Johnson and Schwartz (2005) find non-GAAP earnings do not mislead investors. However, more recent studies find supporting evidence for management using non-GAAP earnings opportunistically to influence stakeholders (Barth, Gow & Taylor, 2011; Doyle, Jennings & Solliman, 2013). Empirical evidence on the opportunistic use of non-GAAP earnings predominately revolves

(6)

around firms using it to meet or beat earnings benchmarks, among which analyst earnings forecasts (Marques, 2017).

Researching whether executive gender has an impact on non-GAAP earnings management can add to the international discussion on gender diversity, possibly supplementing the wide range of effects executive gender has on firm characteristics. Therefore, this paper’s main research question is: ‘Is executive gender associated with the degree of opportunism in non-GAAP exclusions?’

This paper posits that the association between executive gender and non-GAAP opportunism, or non-GAAP earnings management, is dependent on the decision-making differences between male and female executives. Considering these decision making differences, it is hypothesized that male executives are more likely to engage in non-GAAP earnings management.

To answer the main research question, the model used by Doyle et al. (2013) is adapted with gender variables. This model’s main indicator of non-GAAP opportunism is the association between income-increasing non-GAAP exclusions and the propensity to meet or beat analyst earnings forecasts. Firm-quarter observations are collected from all available firms in the S&P1500, from 2007 to 2016. The US market is chosen because it offers more easily collectable data to test with, and since the regulatory discussion there is still ongoing, this research’s results could add to the ongoing discussion on whether to implement aforementioned corporate gender regulation. Prior literature often associates both CEO and CFO characteristics to the prevalence of earnings management. Therefore, both CEO gender and CFO gender characteristics are collected, resulting in the main executive gender variables.

First, the basic Doyle et al. (2013) model is used to confirm the presence of non-GAAP earnings management in the collected sample. Because without the occurrence of non-GAAP earnings management, it might be difficult to ascertain an association between non-GAAP earnings management and gender. Multivariate logistic regression indicates income-increasing non-GAAP exclusions are significantly associated with the propensity to meet or beat the analyst earnings forecast, thereby suggesting the sample is appropriate for investigating the association between executive gender and the degree of non-GAAP opportunism.

Next, the Doyle et al. (2013) model is adapted with executive gender variables. Several multivariate logistic regressions are run, indicating companies with female CEOs or CFOs result in a smaller association between income-increasing exclusions and the propensity to meet or beat earnings benchmarks. Thus, it seems that female executives engage less in non-GAAP earnings management. However, early regressions lack the significance to answer the research question. Female underrepresentation itself seems to be the main cause of these significance issues. In an effort to overcome these significance issues, the gender variables are combined into one, adding both groups of companies with female executives to each other. These final regressions provide an answer to this paper’s main research question. Companies with female executives are found to be significantly

(7)

associated with a decrease in the level of opportunism in non-GAAP exclusions, as was expected by the main hypothesis. Moreover, the gender differences in non-GAAP opportunism are stronger for the use of non-recurring special exclusions than for recurring other exclusions.

The results of this paper add to corporate gender literature, documenting an association with a controversial new form of earnings management. This paper also adds to existing non-GAAP literature, indicating a previously unknown determinant of the degree of opportunism in non-GAAP earnings, and suggesting which key executives might be associated with its content. Lastly, this paper adds to the literature on earnings benchmarks, substantiating notions of male and female executives resorting to different tools in order to meet or beat them. Considering the recent focus on gender diversification in corporate boards, this paper’s implications could be of interest to legislators and companies themselves.

The remainder of this paper is constructed as follows. Chapter 2 provides a theoretical background on all associated subjects: non-GAAP earnings, earnings benchmarks, and executive gender differences. In chapter 3, several hypotheses will be constructed to test the main research question, based on the theoretical background. The data-collection procedures and statistical methods are described in chapter 4. Next, chapter 5 includes the sample’s descriptive statistics, regression results, and some additional robustness testing of the results. Chapter 6 provides a discussion around the findings, ending with some concluding remarks and ideas for future research.

(8)

2 Background

2.1 Non-GAAP earnings

2.1.1 Reporting non-GAAP earnings

Non-GAAP earnings, also called ‘street earnings’ or ‘pro forma earnings’ are an alternative earnings figure, not subject to the General Accepted Accounting Principles which are used to define GAAP earnings. In deriving non-GAAP earnings, management excludes items from GAAP earnings that they deem incidental and non-recurring, arguing that the resulting non-GAAP figure gives a better picture of a firm’s recurring, permanent profitability (Bhattachary et al., 2003). Regularly excluded items are restructuring charges, write-downs and impairments, research and development expenditures, merger and acquisitions costs, mandatory stock compensation expense and goodwill amortization (Bradshaw & Sloan, 2002). Prior literature as well as the Compustat database divide these non-GAAP exclusions into “special items” and “other items” (Doyle et al., 2003; Heflin, & Hsu 2008). ‘Special items’ are unusual and non-recurring items, for example restructuring charges. ‘Other items’ are the remainder of the exclusions, that are not deemed unusual or non-recurring, such as goodwill amortizations.

Management generally motivates their decision to disclose GAAP earnings by arguing non-GAAP earnings contain useful information that supplements the non-GAAP earnings measure, enabling investors to make better decisions (Entwistle et al., 2006). However, because this non-GAAP earnings figure is not subject to the same regulation as GAAP earnings, critiques claim it can be used opportunistically to influence investor’s perception of a firm’s profitability (Bhattacharya et al, 2003; Johnson & Schwartz, 2002). Prior literature labels these two opposing motivations for disclosing non-GAAP earnings as being either to inform, or mislead investors, respectively.

Several researchers have investigated these motives, coming up with mixed results. Bradshaw and Sloan (2002) document a preference for disclosing non-GAAP earnings higher than GAAP earnings and that firms with negative GAAP earnings surprises are more likely to issue non-GAAP earnings information. Moreover, Brown et al. (2012) conclude this willingness to report non-GAAP earnings above GAAP earnings is positively related to investor sentiment, showing this association is at least partially explained by managerial optimism. Research by Bhattacharya et al. (2003) shows that firms that disclose non-GAAP information are more likely to report GAAP losses. Although these findings substantiate the existence of an opportunistic motive for disclosing non-GAAP earnings, they offer no clarification as to whether this is the predominant motivation for disclosing non-GAAP earnings, opposed to informing. By examining the disclosure of non-GAAP earnings in the presence of transitory gains, Curtis et al. (2013) find evidence primarily in support of the informing motive. The majority of their sample – 37,6% - discloses non-GAAP earnings consistently in an informative manner. However, a significant portion of their sample - 27,7% - discloses non-GAAP earnings in an

(9)

opportunistic manner, in the sense that they only disclose non-GAAP earnings if it increases the investor’s perception of core earnings.

2.1.2 Investor reaction to non-GAAP earnings

Besides management’s motivation for disclosing GAAP earnings, the investor reaction to non-GAAP earnings has been widely researched. Based on their investigation of short-term abnormal returns around earnings announcement dates, Bhattacharya et al. (2003) conclude investors view non-GAAP earnings as more informative than non-GAAP earnings, thus as more indicative of core earnings than GAAP earnings. However, Lougee and Marquardt (2004) show the informativeness of non-GAAP earnings is dependent on non-GAAP informativeness and the presence of strategic considerations, such as earnings benchmarks. Investors do not view non-GAAP earnings as more informative if GAAP earnings informativeness is high or in the presence of strategic considerations. Lastly, according to Doyle et al. (2003), relatively higher non-GAAP exclusions result in relatively lower future cash-flows and lower stock returns. This association is most powerful for ‘other exclusions’, which they find are often just as recurring as non-excluded items. Investors fail to assess this predictive power of non-GAAP exclusions.

Moreover, Frederickson and Miller (2004) examine the effects of non-GAAP earnings on sophisticated and unsophisticated investors’ equity valuations. They use an experimental setting, in which they find that equity valuations by less sophisticated investors are influenced by non-GAAP earnings, while valuations from more sophisticated investors such as analysts are not affected. Using historical instead of experimental data, Johnson and Schwartz (2005) find that firms using non-GAAP earnings figures are not priced differently than firms solely reporting GAAP earnings. This suggests investors are, on average, not influenced by non-GAAP earnings. However, as a limitation of their study, Johnson and Schwartz (2005) argue it is possible that sophisticated users arbitrage away any price premium resulting from unsophisticated users being misled by the non-GAAP earnings information.

2.1.3 Non-GAAP regulation

To counter the growing concern of misleading non-GAAP earnings disclosures, a part of the Sarbanes-Oxley Act of 2002 directed the SEC to draw up new regulation for these disclosures. This new non-GAAP disclosure regulation, named Regulation G, became effective in March 2003. It aimed at enabling non-GAAP earnings to aid investors in their investment decisions, and disabling management to mislead investors (Entwistle et al., 2006). Regulation G contained two new requirements for firms issuing non-GAAP information. Firstly, a general disclosure requirement prohibiting firms from disclosing misleading non-GAAP earnings (SEC, 2003, par. IIA3a). Secondly, it contained a reconciliation requirement, obligating firms to reconcile any presented non-GAAP

(10)

earnings to the most directly comparable GAAP financial measure, and provide an explanation for the differences (SEC, 2003, par. IIA3b).

Subsequent to the implementation of Regulation G several researchers have examined its effects on non-GAAP earnings reporting, with findings primarily in support of the SEC’s choices. Research by Entwistle et al. (2003) documents Regulation G led to a decline of firms issuing non-GAAP earnings from 77 to 54 percent. Moreover, they conclude that Regulation G resulted in managers reporting non-GAAP in a less opportunistic and misleading manner. These findings are confirmed by Heflin and Hsu (2008), who also find that although Regulation G reduced managers’ opportunistic use of non-GAAP earnings, it also reduced managers’ informative use of non-GAAP earnings. Furthermore, Zhang and Zheng (2011) examine the effects of the reconciliation requirement. They find evidence of firms releasing non-GAAP earnings with low reconciliation quality being mispriced prior to regulation G, whereas they find no such mispricing after Regulation G. After controlling for changes in the quality of items excluded from non-GAAP earnings, they conclude this difference is due to the reconciliation requirement in regulation G. Aubert and Grudnitski (2014) extend these findings by examining the reconciliation quality of 314 European firms. They empirically establish increased reconciliation quality to be associated with reduced mispricing, explaining the effectiveness of regulation G on securities’ mispricing.

2.2 Benchmark beating

2.2.1 Incentives for meeting earnings benchmarks

A firm’s earnings is perceived as the most important item included in a firm’s financial reports by investors, analysts, and other stakeholders (Degeorge et al., 1999). In order to assess these earnings and overall firm performance, stakeholders tend to use specific thresholds to benchmark earnings against. Research suggests three primary earnings benchmarks companies aim to meet: zero earnings, last year’s earnings, and analysts’ earnings forecast (Burgstahler & Dichev, 1997; Degeorge et al., 1999; Jiang, 2008). The importance of these earnings benchmarks is reiterated by CFOs themselves, naming last year’s earnings and analysts’ forecasts as the most important, based on a survey among 400 CFOs done by Graham et al. (2005). Extant literature provides several arguments for companies’ willingness to meeting these earnings benchmarks, primarily distinguishing equity incentives and debt incentives (Bartov et al., 2001; Jiang, 2008; Kasznik & Mcnichols, 2001; Lopez & Rees, 2000).

With regard to the equity incentives of meeting earnings benchmarks, results from Lopez and Rees (2000) show that the investor’s reaction to earnings is significantly stronger if a firm meets analysts’ forecasts, compared to if a firm does not. Furthermore, whether a firm meets analysts’ forecasts is found to be a stronger determinant of abnormal stock returns than the absolute value of earnings. Lastly, Lopez and Rees (2000) document the investor response to be incrementally stronger for firms that meet analysts’ forecasts in consecutive periods. These findings are consistent with the

(11)

research of Bartov et al. (2001), documenting that firms who meet or beat analysts’ forecasts are rewarded by investors, through their shares trading at a premium. Alternatively, firms who do not meet analysts’ forecasts are penalized, and trade at a discount. This reward or penalty is found not to be related to the absolute value of earnings, further substantiating the notion of investors rewarding or penalizing the firm for (not) meeting the earnings benchmark, instead of merely for higher or lower earnings (Bartov et al., 2001). In concurrent research, Kasznik and Mcnichols (2001) offer a rationalization of the shareholder’s reaction, by showing that firms meeting analysts’ forecasts have significantly higher future earnings than firms that do not. The investor’s reaction of rewarding forecast-meeting firms with a higher share price therefore might be the result of a rational expectation of higher future earnings for firms that meet their earnings forecasts.

In addition to the aforementioned equity incentives, Jiang (2008) documents incentives concerning a firm’s debt contracting for meeting the earnings benchmarks of zero earnings, last year’s earnings, and analysts’ earnings forecasts. Proxying for cost of debt with bond rating and yield spread, she finds firms that meet earnings benchmarks have a reduced cost of debt. Of the three examined earnings benchmarks, the effect of beating the zero earnings benchmark generally leads to the biggest reduction in cost of debt (Jiang, 2008).

While the existence of equity and debt incentives are substantiated by empirical research, the CFO’s in Graham, Harvey and Rajgopal’s (2005) survey propose a third incentive: personal welfare. They argue that meeting earnings benchmarks can be beneficial to their career, and their external reputation. Moreover, the results of Graham et al. (2005) suggest managers are primarily incentivized to meeting earnings benchmarks by the equity effects and benefits to their personal welfare, and less so by the effects on the firm’s cost of debt.

These three effects incentivize management to such an extent, that the CFOs’ in Graham et al. (2005) are willing to sacrifice long-term economic value in order to meet them, by engaging in real activities manipulation. They even feel this is the appropriate choice, arguing that the more short-term negative impact of not meeting the earnings benchmarks outweighs the possible long-term negative impact on economic value (Graham et al., 2005). There are several opportunistic methods under management’s discretion usable for benchmark beating, roughly divided into expectations management, GAAP earnings management, and non-GAAP earnings management. Research shows that the positive effects on equity and debt are only marginally smaller for firms that seem to have met their earnings benchmarks through earnings management or expectations management (Bartov, Givoly, & Hayn, 2001; Jiang, 2008). This offers some understanding for the CFOs’ willingness to use these opportunistic benchmark beating methods in order to meet their earnings benchmarks.

(12)

2.2.2 Expectations management

Management has some control over the information analysts receive, and therefore over the assessments these analysts make about a firm future economic performance. When management uses this control for deliberately downplaying analysts’ earnings forecasts, this is called expectations management (Bartov et al., 2001). The use of this first form of benchmark beating therefore is confined to lowering analysts’ earnings forecasts. It demands no further explanation that lower earnings forecasts will generally be easier to meet than higher earnings forecasts. Evidence of firms using expectations management to meet earnings forecasts is provided by prior research, on several occasions (Bartov et al., 2001; Matsumoto, 2002; Burgstahler & Eames, 2006).

2.2.3 Within-GAAP earnings management used for benchmark beating

Instead of trying to manage forecasts downwards, firms can also opportunistically manipulate their earnings upwards, called earnings management. For clarity, earnings management can be used for meeting any of the three aforementioned earnings benchmarks. Prior research describes two main earnings management methods for manipulating a firm’s GAAP earnings figure: accruals manipulation and real activities manipulation. The former refers to the changing of revenues or expenses, through the use of accounting accruals. Real activities manipulation refers to the altering of the timing or structure of economic decisions, thus also having an impact on actual cash-flows (Badertscher, 2011). Both of these methods are associated with multiple costs and other constraints. Accruals earnings management is constrained by the flexibility of the General Accepted Accounting Principles itself, and is more susceptible to outsider scrutiny and litigation, especially since the Enron scandal. Costs of real activities manipulation mainly pertain to deviations from optimal business strategy, because of its cash-flow effect. Both of these methods are therefore used as substitutes, subject to their respective costs (Zhang, 2011).

According to Burgstahler and Dichev (1997), companies seldom present small earnings decreases or small losses, and often present small earnings increases or small profits. They argue this skewed distribution is due to companies managing their earnings upwards, with both within-GAAP earnings management methods, in order to meet last year’s earnings and zero earnings. Research by Degeorge et al. (1999) shows the same methods are used for firms trying to exceed analysts’ forecasted earnings, just as for exceeding zero earnings and last year’s earnings. Besides finding evidence of firms managing their earnings upwards, in the case of nearly met earnings benchmarks, they find evidence of firms shifting current period’s earnings to the next period if the earnings benchmark is currently far from met. For firms only just meeting their earnings benchmarks, they document future performance is predictably worse than for firms whose earnings seem more genuine. Later research corroborates these results, adding that earnings management is used in conjunction with expectations management for meeting analysts’ earnings forecasts, instead of it being used as an alternative method (Burgstahler & Eames, 2006; Matsumoto, 2002).

(13)

2.2.4 Non-GAAP earnings management used for benchmark beating

Thus, GAAP reporting firms can downplay analysts’ forecasts, or use various within-GAAP methods of earnings management to meet earnings benchmarks. As an alternative to manipulating GAAP earnings, non-GAAP reporting firms also have the option of manipulating their non-GAAP earnings, as shown by recent studies. Anecdotal evidence by Bhattacharya et al. (2003) shows that, for non-GAAP reporting firms, 80 percent of issued non-GAAP earnings meet or beat analysts’ earnings forecast, while only 39 percent of their GAAP earnings manages to do so. They interpret this as being in support of the often exclaimed critique that non-GAAP earnings are used opportunistically to meet earnings benchmarks. More insight into the mechanics of non-GAAP earnings management is offered by Black and Christensen (2009), showing that the most frequently performed exclusions are recurring ‘other items’, instead of nonrecurring ‘special items’ that are supposed to be excluded from non-GAAP earnings. These other items are found to result in predictably lower future cash-flows, and are significantly associated with a firm’s ability to meet earnings benchmarks, whereas this association is considerably weaker for special items (Black & Christensen, 2009; Doyle et al., 2003; Doyle et al., 2013). This suggests management opportunistically excludes recurring items to manipulate non-GAAP earnings.

Barth et al. (2011) further substantiate the notion of opportunistic non-GAAP exclusions by examining stock-based compensation, which SFAS 123Rs mandates to be recognized as expenses. They find that firms are more likely to exclude these expenses from non-GAAP earnings if they have higher incentives for increasing earnings, smoothing earnings, and meeting analysts’ earnings forecasts. Heflin and Hsu (2008) also document opportunistic use of non-GAAP exclusions for meeting the zero-earnings and previous year’s earnings benchmarks, although this declined as a result of Regulation G. The study of Doyle et al. (2013) delivers critique on previous researchers, such as Bhattacharya (2003), arguing they often focus on non-GAAP reporting firms, without making a comparison between GAAP and non-GAAP reporting firms. By doing so, Doyle et al. (2013) find that non-GAAP exclusions are used opportunistically for meeting analysts’ earnings forecasts, and that solely non-GAAP reporting firms are more likely to engage in earnings management than solely GAAP reporting firms. Thus it seems the decision to include non-GAAP earnings is already driven by opportunistic motives. Collectively, these results suggest management is actively using non-GAAP exclusions in order to meet earnings benchmarks.

Researchers document a substitutive association between non-GAAP earnings management and the other benchmark beating tools, adding to the legitimacy of non-GAAP earnings management as being an alternative benchmark beating tool, used for the same goals. Doyle et al. (2013) find this substitutive association is dependent on the costs and benefits of each benchmark beating method, examining constraints for the other benchmark beating tools on firm-level. In their sample, firms that are more constrained in using GAAP earnings management or expectations management exercise

(14)

more non-GAAP earnings management for meeting analysts’ earnings forecasts. Whereas Doyle et al. (2013) investigate earnings and expectations management restrictions on firm level, Isidro and Marques (2015) investigate earnings management restrictions on a national level. They are the first to ascertain an association between institutional and economic factors, and non-GAAP earnings management for meeting earnings benchmarks. Specifically, if a firm operates in a country with more strict GAAP regulation, law enforcement or investor protection, thus being constrained in their use of GAAP earnings management, they seem to rely more on non-GAAP earnings management for benchmark beating (Isidro & Marques, 2015). An earlier study by Badertscher (2011) also finds this substitutive association, although with regard to implicit earnings expectations instead of explicit earnings benchmarks. He documents that overvalued firms struggling to meet their relatively high implicit earnings expectations as a result of their overvaluation, resort to all three different methods of earnings management, in succession. Non-GAAP earnings management is found to be management’s last resort, and is especially used if firms are highly overvalued, or for a longer duration (Badertscher, 2011). In conclusion, this substitutive association between non-GAAP earnings management and other forms of earnings management further adds to the legitimacy of opportunistic non-GAAP exclusions being an alternative earning management method, actively used by management for benchmark beating.

2.3 Executive gender differences

2.3.1 Risk-appetite

The notion that men and women do not always act in the same way, has been extensively researched in prior research. Since the recent regulatory discussion with regard to corporate gender diversity and female representation on corporate boards, researchers increasingly examined the implications of these gender differences in a corporate context. One gender difference with several implications for financial decision-making is males having a bigger risk-appetite, and engaging in more risk-taking.

Byrnes, Miller and Schafer (1999) examine the association between gender and risk-taking in a general context, on the basis of 150 earlier studies, and find men to engage in significantly more risk-taking than their female counterparts. Furthermore, men engage in more risk-risk-taking, even if it is apparent that this is a bad idea, whereas women engage in less risk-taking, even if it is apparent that they should (Byrnes et al., 1999). Earlier research by Powell and Ansic (1996) documents the same risk-taking differences with regard to gender in two experimental settings of financial decision making. The female participants in their study seem to be more aimed at creating security, whereas the male participants strive for the best possible gains, willing to engage in riskier decision-making to attain this (Powell & Ansic, 1996). Recent evidence by Khan and Vieito (2013) show the implications in a corporate setting, by comparing panel data of firms with a female CEO to firms with a male CEO.

(15)

They document that firms with a female CEO operate at a significantly lower risk level, with more constant returns, and achieve better overall performance. Similar results are found by Faccio, Marchica and Mura (2016), indicating that firms with a male CEO have less leverage, less volatile earnings, and are more likely to stay operating. Both these findings seem in line with Powell and Ansic (1996), with regard to females being more security-seeking. Although this strive for security might be perceived as solely beneficial, Faccio et al. (2016) also document less efficient capital allocation for firms with a female CEO than for firms with a male CEO. This suggests male executive’s increased risk-taking can, in some cases, be advantageous for the firm.

2.3.2 (Over)confidence and optimism

Extant research also documents males, in general and in a corporate context, are overconfident and more optimistic in their decision making. Early research by Powell and Ansic (1996) documents that male participants are systematically more optimistic in their view of the world, whereas the female participants are more pessimistic. Also, women more often ascribe their success to good luck, while men ascribe success to their own abilities and failures to bad luck. Ascribing success to your own abilities, and attributing failures to luck are named as defining characteristics of overconfidence by Habib & Hossain (2013). Subsequent literature often perceives optimism and overconfidence to overlap (Huang & Kisgen, 2013). Overconfidence is defined as “a behavioural bias of having unrealistic (positive) beliefs about any aspect of the distribution of an uncertain outcome (e.g., future cash flows) such that the mean is overstated” (Habib & Hossain, p. 92, 2013). Parallel to a bigger risk-appetite, overconfidence results in riskier decision making. However, the definition suggests overconfidence and optimism differ from risk-appetite in one important aspect. A bigger risk-appetite will still result in managers ascertaining the existence of risk in a decision, but deciding to accept this risk as a part of their bigger risk-appetite. However, overconfident managers will assess the same decision to be of a less risky nature, or it having a higher pay-off, and will accept the riskier decision based on the biased interpretation of the decision’s implications. This notion of overconfidence and risk-taking being two distinctly different concepts is empirically substantiated by Huang and Kisgen (2013).

The effects of overconfidence and optimism on financial decision making can be substantial, and are not all negative. Huang and Kisgen (2013) find that firms with male executives grow faster, perform more acquisitions, issue more debt, but have lower announcement returns on acquisition or debt announcements. Although these findings can also be partly described by the increased risk-appetite of males, they perform additional testing, proving overconfidence and optimism best explain their results. Habib and Hossain (2013) conclude from their literature review that while overconfident executives are in essence undesirable, they are also found to be better innovators.

(16)

2.3.3 Earnings quality, conservatism and earnings management

The wide use of the earnings figure in stakeholders’ decisions indicates the importance of high quality earnings, being the degree to which it captures a firm’s economic reality (Krishnan & Parsons, 2007). Examining factors that influence earnings quality has often been the subject of previous studies. In recent years, several researchers also found executive gender and diversity to be related with earnings quality, earnings management and accounting conservatism. One such study is that of Krishnan and Parsons (2007), indicating that companies with more gender diverse boards, thus having more women on the boards, have higher earnings and higher earnings quality. Moreover, these companies employ more conservative accounting practices, have more persistent earnings, and engage less in GAAP earnings management for beating earnings benchmarks. They conclude gender diversity in corporate boards seems to be beneficial for firm earnings, without this increase in earnings being attained through a lower earnings quality or increase in earnings management. Francis et al. (2015) substantiate the association between executive gender and accounting conservatism, documenting a significant increase in accounting conservatism after male CFOs are replaced by a female CFO. This association between executive gender and conservatism is found to be even stronger in riskier circumstances, such as with default risks or litigation risks. Logically, the smaller the risk itself, the smaller the impact of a different risk-appetite will be. Thus, this is also consistent with prior notions of females being more risk-averse compared to men (Byrnes et al., 1999; Powell & Ansic, 1996).

According to Barua et al. (2010), female CFOs are also associated with higher accruals quality, proxied for with the magnitude of discretionary accruals and accrual estimation errors. Discretionary accruals are often used as proxies for accruals earnings management in prior research. This implies, consistent with Krishnan and Parsons (2017), that female executives engage in less accruals based earnings management compared to male executives. Examining the effects of both CFO and CEO gender on accruals earnings management, Peni and Vähämaa (2010) find that female CFOs in the US use more conservative accounting, and less accruals-based earnings management. They do not find a significant association between CEO gender and earnings management. While previous studies have primarily sampled the developed markets of the US and the UK, a study by Liu, Wei and Xie (2016) examined the effects in the developing market of China. They find similar results, documenting female CFOs engaging in significantly less accruals and real activities manipulation.

The notion that male executives engage in more opportunistic and aggressive, and less conservative accounting policies can be explained by several aforementioned male characteristics, such as male risk-taking. However, they can also be explained from an ethical perspective. In their literature review, Habib and Hossain (2013) conclude that male CEO’s and CFO’s are more interested in economic gain than their female counterparts, and are more inclined to act unethical and break the rules in order to attain this economic benefit. Female executives are more interested in harmony and security, and are less likely to be unethical and break the rules.

(17)

3 Hypothesis development

In this section, the hypotheses will be developed that will be used to examine the main research question: ‘Is executive gender associated with the degree of opportunism in non-GAAP exclusions?’ But before this research question is operationalized and decomposed in hypotheses, a brief summary of the above literature review is given.

According to the literature, investors and other stakeholders tend to use normative thresholds for benchmarking a firm’s earnings against, aiding them in their decision-making. There are three primary earnings benchmarks companies aim to meet or beat: zero-earnings, previous year’s earnings, and analysts’ earnings forecasts (Burgstahler & Dichev, 1997; Degeorge et al., 1999; Jiang, 2008). Attaining these earnings benchmarks offers several benefits to management, including equity appreciation, more efficient debt contracting, and personal success and status improvements. Moreover, management is incentivized to such an extent, that they are willing to engage in earnings management in order to attain this. Earlier research on earnings management to meet or beat earnings benchmarks primarily focussed on the more classical within-GAAP earnings management methods. With regard to the within-GAAP earnings management methods, researchers consistently documented an association with several gender-related differences between executives and the use of opportunistic accounting policies and earnings management.

Several of these gender differences are described in the literature review. For example, male executives tend to take more risky and aggressive financial decisions, possibly due to a bigger risk-appetite or male overconfidence, and tend to be more focussed on economic gains, willing to break rules in order to attain this. Female executives are more risk-averse and pessimistic, and aim more at creating stability and harmony, being less likely to break rules and engage in earnings management. All together, these gender differences seem to result in male executives being more inclined to engage in within-GAAP earnings management in order to meet or beat their earnings benchmark.

Since the recent surge in popularity of disclosing non-GAAP earnings figures, often accompanying the GAAP earnings figure itself, researchers increasingly examined management’s opportunistic use of non-GAAP earnings. Researchers document non-GAAP earnings as a new earnings management method, and as a substitute for other opportunistic benchmark-beating methods such as expectations management, accruals manipulation, and real activities manipulation (Black & Christensen, 2009; Doyle et al., 2013; Isidro & Marques, 2015). Opportunities for opportunism primarily flow from the high level of discretion that is allowed in its creation. While companies are supposed to merely exclude incidental ‘special items’, they often exclude recurring expenses from their non-GAAP earnings. The latter type of exclusions is often named ‘other exclusions’, and is strongly associated with firms ability to meet or beat their earnings benchmark. However, whether the same association as between executive gender and within-GAAP earnings management also holds for

(18)

non-GAAP earnings management, is yet undiscovered. Therefore, this paper’s main research question is: “Is executive gender associated with the degree of opportunism in non-GAAP exclusions?”

A recent review by Marques (2017) notes that empirical evidence on non-GAAP opportunism is primarily focussed on companies using it to meet or beat earnings benchmarks. Numerous of these papers are discussed in the previous chapter. Although prior literature documents three different types of earnings benchmarks, the analyst’s earnings forecast is arguably the most frequently mentioned. CFO’s themselves named this as one of the most important benchmarks, and this benchmark was repeatedly examined in the context of non-GAAP earnings. Considering the proposed importance, and strong empirical basis in prior literature for companies using non-GAAP earnings for meeting this earnings benchmark, this research focusses on analyst’s earnings forecasts.

Before examining the impact executive gender has on opportunistic non-GAAP earnings exclusions, it is investigated whether non-GAAP earnings management even exists in the collected sample. This is the focus of the first hypothesis. Remember that both incidental costs and revenues should be excluded from non-GAAP earnings, in a similar manner. As prior literature explains, analysts generally also forecast on a non-GAAP basis, with monthly adjusting of the forecast itself in case of incidental items (Doyle et al., 2013; Heflin & Hsu, 2008). Thus, whether a company has more incidental costs than revenues in a period or vice versa should not be associated with its chance of meeting or beating its earnings forecast, because the analyst’s non-GAAP forecast will be adjusted to it. However, if companies use non-GAAP exclusions opportunistically in order to meet or beat analyst’s earnings forecasts, there would exist an association between the use of earnings-increasing non-GAAP exclusions, and the chance of meeting or beating their analyst earnings forecasts. This approach is in line with prior literature (Barth et al. 2012; Doyle et al., 2013; Heflin & Hsu, 2008), and results in the first hypothesis:

H1: Companies using exclusions to increase non-GAAP earnings are more likely to meet or beat analyst’s earnings forecasts.

Once it’s been established that opportunistic non-GAAP exclusions are prevalent in this paper’s sample, the impact of executive gender on the extent of opportunism can be investigated. As said, prior literature indicates male executives are associated with significantly more within-GAAP earnings management. Thus, if all earnings management methods are completely comparable, it might be expected that male executives are also associated with more non-GAAP earnings management. However, the literature review points out that these earnings management methods are not comparable, and all come with several constraints. Real activities earnings management has a cash-flow effect, leading to deviations from optimal business strategy and thus real costs. Besides constraints from GAAP itself, accruals manipulation bears significantly more scrutiny and litigation risks for the company and management itself (Badertscher, 2011).

(19)

If the more risk-averse female executives perceive non-GAAP earnings management to be of low risk, they might be more inclined to engage in this form of earnings management. Not meeting an earnings benchmark even comes with risks, in the form of a negative equity effects, debt contracting and reputational effect. The difference between men and women can be even smaller if male executives also use the within-GAAP substitutes that female executives might find too risky or costly. However, especially with the existence of Regulation G, non-GAAP earnings management is unlikely to be perceived as being of no risk. In that sense, non-GAAP earnings management is probably still similar to other forms of earnings management. Due to its unethical nature, women might also be less inclined to engage in non-GAAP earnings management. And lastly, male overconfidence and optimism could result in male executives more often perceiving costs as incidental and revenues as permanent, or underestimate any associated risks and costs of employing non-GAAP earnings management. Therefore, it is still expected that male executives use non-GAAP exclusions more opportunistically than female executives, leading to the second, and main hypothesis:

H2: The association between income increasing non-GAAP exclusions and meeting or beating earnings benchmarks is stronger for companies with male executives, than for companies with female executives.

A construct that is still ambiguous is ‘executives’. For executives to be of interest to this research, they must be associated with the decision making process leading to non-GAAP earnings management. The two executives that prior literature associates to earnings management are a firm’s Chief Financial Officer (CFO) and Chief Executive Officer (CEO). The CFO is primarily responsible for the integrity of the financial reporting system, while the CEO is responsible for overall company strategy and policy. For earnings management happening within the system of financial reporting, research generally finds that CFO characteristics and incentives have the strongest association with the prevalence of earnings management (Jiang, Petroni & Wang, 2010). Still, CEO characteristics and incentives also have an empirically established association with earnings management, mainly because CEOs put pressure on CFOs to alter reports and manage earnings (Friedman, 2014; Habib & Hossain, 2013).

However, in the case of non-GAAP earnings, this association is still quite unknown. Researchers have found that CEO characteristics and incentives are more associated with the chance that non-GAAP earnings is disclosed at all, but no research has examined which executive has the strongest impact on the contents of non-GAAP earnings itself, and thus on non-GAAP earnings management (Black, Black & Christensen, 2015; Bansal, Seetharaman & Wang, 2013). Marques (2017) therefore names this as an important research subject for future research in her literature review.

(20)

Because prior research offers no definitive answer as to which of the two executives will have the strongest impact, this research will focus both on CEO gender, as CFO gender, leading to the following two sub-hypotheses:

H2a: The association between income increasing non-GAAP exclusions and meeting or beating earnings benchmarks is stronger for companies with male CEO’s, than for companies with female CEO’s.

H2b: The association between income increasing non-GAAP exclusions and meeting or beating earnings benchmarks is stronger for companies with male CFO’s, than for companies with female CFO’s.

(21)

4 Data and method

4.1 Sample

This chapter includes the procedures that will be employed to answer the research question. First, the procedures that resulted in a usable sample are described, before describing the research design for empirically testing the hypotheses. The used sample is based on the US market. While corporate gender regulation is not yet existent in the US, or at least not in a similar extent as in the EU, sampling the US offers several advantages. The primary advantage is data-availability. Several used databases, especially with regard to non-GAAP information, are significantly more comprehensive for the US market. Besides, in the US, the discussion whether to create such regulation is still ongoing. Therefore, researching the US market could more directly add to the ongoing regulatory discussion. Considering the notion of underrepresentation of women in executive boards, a primary concern is not having enough data on female executives’ companies. If too few women are included, this could hinder statistical significance. Therefore, firm-quarter data from all companies in the S&P1500, instead of the S&P500, is collected over the period from 2007 to 2016. Note that this period is wholly after the adoption of Regulation G, assuring that found inferences are still actual.

As a starting point of the data collection process, identifiers for the S&P1500 are downloaded from the Compustat North-America database. The most important identifier is each company’s unique CUSIP code, which is used to merge the different databases. After obtaining these identifiers, this list is remodelled, now including all different years and quarters to be included in the sample. The remaining databases can now be merged on the basis of CUSIP code, data year, and data quarter.

Information on CEO and CFO gender is downloaded from the Execucomp database. The complete database is searched, resulting in a list of names for all executives in the Execucomp database, along with a denotation of their function being either CEO or CFO for a specific year, the respective company and CUSIP code, and their gender being either female or male. The gender variable is transformed, denoting a 0 in the case of a male executive, and a 1 in the case of female executives. Filtering the list solely on executives being either CEO’s or CFO’s results in two lists that can be linked to the obtained S&P1500 identifiers. This is done on the basis of company CUSIP-codes and year, considering this database only specifies the relevant CEO and CFO information on a yearly basis. Because the Execucomp database primarily includes 6-digit CUSIPS, the respective identifiers are first shortened into including only 6-digits, as advised in the Execucomp database itself. The accuracy of these new 6-digit CUSIP codes is manually verified after merging.

Information on analyst forecasts and non-GAAP information is downloaded from the Institutional Brokers' Estimate System (IBES). Note that the IBES forecasts are also on a non-GAAP basis, named ‘street earnings’ by analysts. The IBES Actuals figure is used as a proxy for disclosed non-GAAP earnings, as explained in the next chapter. The IBES Summary History database is used for deriving

(22)

the applicable analyst earnings forecasts and actuals. The list is composed of monthly consensus earnings forecast and respective actuals, for a certain period. This is filtered, so it only contains the last relevant reported earnings forecast prior to the companies’ quarterly earnings announcement. This last forecast is the forecast companies primarily aim to meet or beat, and is used by prior non-GAAP research (Barth et al., 2012; Doyle et al., 2013; Heflin & Hsu, 2008). These forecasts and actuals are merged to previous data on the basis of the unique 6-CUSIP codes, along with the data year and quarter. Several companies have a 5-digit CUSIP code in the IBES database, hence these companies are merged on the basis of this 5-digit CUSIP code, with manual verification.

After collecting data on executive gender, analyst forecasts and actuals, what remains is the more general company information including quarterly GAAP earnings and several control variables such as company size and sales. The Compustat North-America database is used for this matter, which is also merged using company CUSIP codes, data years and quarters.

The resulting sample composes of 89.080 firm-quarters, being quarterly observations on 2.227 unique companies over a period of ten years. Note, the sample includes more than 1.500 companies. This is due to changes in the companies included in the S&P1500 over the sample period. The resulting dataset includes several incomplete quarters. Excluding all observations with missing non-GAAP information results in the dropping of 28.973 firm-quarters, still keeping 60.107 quarters. The remaining omissions only pertain to the CFO and CEO gender variables. There are 1.299 quarters missing for CFO gender and 1.318 missing for CEO gender. Manual inspection reveals these missing quarters mostly overlap. Therefore, all missing observations for both CFO and CEO gender will be dropped, resulting in one sample that will be used for all regressions. This results in the dropping of 1.474 extra quarters, keeping 58.633 complete firm-quarters, about 1.951 firms.

For at least a part of our sample period, 116 companies have a female CEO, and 293 companies have a female CFO. This corresponds to data on 1.860 female CEO quarters and data on 4.943 female CFO quarters. A priori, it is assumed these figures are sufficient for this paper’s purposes. Table 1 gives an overview of the different steps taken, ending with the final sample.

Table 1. Sample selection

Procedure Firm-quarters Unique firms

Merged file total 89.080 2.227

Dropped missing non-GAAP actuals 62.087 2.223

Dropped missing Compustat variables 60.107 2.024

- minus incomplete CEO gender, or; 58.808

- minus incomplete CFO gender 58.789

(23)

4.2 Research design

4.2.1 Proxy for non-GAAP earnings

Companies usually release their non-GAAP earnings figure in quarterly earnings announcements, often together with their GAAP earnings. This non-GAAP earnings figure cannot be freely collected from a database. The only legitimate alternatives are hand-collecting all non-GAAP firm-quarters from press releases, or using a proxy. Considering a large sample size is deemed especially important to counter for female underrepresentation, hand-collecting is not deemed feasible. Earlier research often uses IBES actuals as a proxy for disclosed non-GAAP earnings (Doyle et al., 2013; Doyle & Soliman, 2005: in Heflin & Hsu, 2008; Heflin & Hsu, 2008). These IBES actuals are the non-GAAP earnings number as provided by analysts. Generally, after receiving management’s published non-GAAP earnings number, this figure is included as the IBES actuals figure, except for situations where analysts unwind detected managerial opportunism (Doyle et al., 2013).

But, prior research suggests IBES actuals mostly match disclosed non-GAAP earnings, and it still contains managerial opportunism. For example, Doyle and Soliman (2005) find that for 92 of 100 randomly selected 100, IBES actuals correspond perfectly with the non-GAAP earnings disclosed in press releases. Heflin and Hsu (2008) also examined this, by reviewing 1928 press releases with non-GAAP earnings figures and concluded that the IBES actual figure perfectly matches with the press release for 90% of these releases. Doyle et al. (2013) conclude that for 94,4 % of 1.000 randomly examined press releases, the disclosed non-GAAP earnings figure is the same as the figure included in IBES actual. Bansal, Seetharaman and Wang (2013) add that while IBES actuals undoubtedly include some noise as a proxy, this noise will more likely result in a bias against findings results, because analysts partly unwind managerial opportunism. Any association with opportunism found while using this proxy will likely only be stronger if the real non-GAAP earnings is used. Considering prior research finds a fit of more than 90%, and that any deviations will rather weaken the results than induce them, it is assumed IBES actuals are an appropriate proxy for non-GAAP earnings. Using IBES actuals as a proxy allows for a significantly larger sample size, because it requires no hand-collecting. Moreover, using this method makes any results more comparable to results found in earlier research.

4.2.2 H1: Investigating the prevalence of non-GAAP earnings management

This research primarily uses an adaptation of the model used in Doyle et al. (2013) for detecting opportunistic non-GAAP exclusions to meet or beat earnings benchmarks. In order to examine H1, the basic Doyle et al. (2013) model is used without any gender variables. The main dependent variable that Doyle et al. (2013) employ is whether a company meets or beats their analyst earnings forecast (MBEF), and is denoted as a dummy variable. It takes on the value of 1 if a company has a higher IBES actual (as a proxy for non-GAAP earnings) than the last median analyst forecast before the

(24)

earnings announcement date, and 0 otherwise. The main dependent variable is whether the company uses positive non-GAAP exclusions (PosExcl) or, differently phrased, discloses a non-GAAP figure that is higher than the GAAP reported earnings. This is also denoted as a dummy variable, being 1 in the case of positive total exclusions, thus a non-GAAP earnings figure that is higher than GAAP earnings, and 0 otherwise. Using a dummy variable for exclusions measures the decision to use income-increasing exclusions itself, instead of the absolute value of exclusions, and is consistent with prior literature (Barth, Gow & Taylo, 2012; Black & Christensen, 2009; Doyle et al., 2013).

The model also includes several control variables, which prior literature documents to have an association with the chance of meeting or beating earnings benchmarks. The first control variable is the company’s book-to-market ratio (Book-to-Market), calculated by dividing the book value of equity by the market value of equity. The company’s growth in sales (SalesGrowth) is calculated by dividing current quarter’s sales by sales in the same quarter of the prior year. Controlling for a company’s size is the natural logarithm of the market value of equity (LnSize). Whether a company attains a non-GAAP profit (Profitable) is included as a dummy variable with the value of 1 if IBES actual is positive, and 0 otherwise. The last control variable is return on assets (ROA), calculated by dividing a company’s IBES actual by its assets. Considering the dependent variable (MBEF) is a dummy variable, and thus not continuous, logistic regressions are used. This results in the following regression model:

(1) MBEFt = γ0 + γ1PosExcli,t + γ2Book-to-Marketi,t + γ3SalesGrowthi,t

+ γ4LnSizei,t + γ5Profitablei,t + γ6ROAi,t + υi,t

Multivariate logistic regression is used to investigate the association between the use of income-increasing exclusions (PosExcl) and the propensity to meet or beat analyst’s earnings forecast (MBEF). If firms using positive exclusions are significantly more likely to meet or beat their analyst earnings forecast, this indicates the opportunistic use of non-GAAP exclusions. This interpretation was explained in the hypothesis section, and is in line with prior literature (Barth, Gow & Taylo, 2012; Black & Christensen, 2009; Doyle et al., 2013). Thus, in line with the first hypothesis, it is expected that the coefficient for PosExcl is positive. While total exclusions provide insights into the overall extent of non-GAAP earnings management, prior literature often divides total exclusions into incidental ‘special exclusions’ and recurring ‘other exclusions’ (Doyle et al., 2003; Heflin, & Hsu 2008). These researchers suggest other exclusions are used primarily for opportunistic reasons, having a stronger association with meeting or beating earnings benchmarks than special exclusions. Also performing regressions with these underlying types of exclusions indicates which specific types of exclusions are being used differently, or more opportunistically, as a result of executive gender. All hypotheses in this research will therefore be tested twice: once with a regression model including

PosExcl, and once including PosOtherExcl and PosSpecialExcl. In line with prior literature, special

(25)

al., 2003). This indicates the part of exclusions that is non-recurring of nature. The remaining difference between GAAP EPS and non-GAAP EPS indicates other exclusions that are more likely to be recurring, non-incidental exclusions. The PosOtherExcl and PosSpecialExcl variables are dummy variables that take on the value of 1 in case of increasing other exclusions and income-increasing special exclusions, respectively. This results in the following secondary regression model for the first hypothesis.

(2) MBEFt = γ0 + γ1PosOtherExcli,t + γ2PosSpecialExcli,t + γ3Book-to-Marketi,t

+ γ4SalesGrowthi,t + γ5LnSizei,t + γ6Profitablei,t + γ7ROAi,t + υi,t

The expectation is that other exclusions are the main medium of opportunism, in line with prior literature. Therefore, the coefficient for PosOtherExcl is expected to be positive, and more positive than the coefficient for PosSpecialExcl.

4.2.3 H2: Investigating the association of executive gender with non-GAAP earnings management

After confirming the opportunistic use of non-GAAP exclusions in our sample, the association with executive gender can be investigated through the second hypothesis. For this purpose, the basic Doyle et al. (2013) model will be adapted with several gender-related variables. Both the gender variable itself and an interaction with the previous independent variables will be added to the model. Both executives of interest, CEO’s and CFO’s, will be investigated separately and therefore result in two different regression models. The model for investigating the association of CEO gender with non-GAAP earnings management is:

(3) MBEFt = γ0 + γ1PosExcli,t + γ2CEOgenderi,t + γ3CEOgender*PosExcli,t

+ γ4Book-to-Marketi,t + γ5SalesGrowthi,t + γ6LnSizei,t + γ7Profitablei,t + γ8ROAi,t + υi,t The model for investigating the association of CFO gender with non-GAAP earnings management is:

(4) MBEFt = γ0 + γ1PosExcli,t + γ2CFOgenderi,t + γ3CFOgender*PosExcli,t

+ γ4Book-to-Marketi,t + γ5SalesGrowthi,t + γ6LnSizei,t + γ7Profitablei,t + γ8ROAi,t + υi,t The variables CEOgender and CFOgender both are dummy variables, denoting a 1 if the executive is female, and a 0 if it is a male. The interaction variables CEOgender*PosExcl and

CFOgender*PosExcl will thus consist of an interaction between two dummy variables. The interaction

variables therefore take on the value of 1 in case of a female executive (=1) that uses positive exclusions (=1), and 0 otherwise. Note the effect of gender itself on MBEF will be included in the gender variable’s coefficient, while the effect of using positive exclusions on MBEF will be included in the coefficient of PosExcl. Therefore, by constructing the interaction variable in this manner, it denotes the incremental effect of gender on the association between the use of positive exclusions (PosExcl) and meeting or beating of analyst earnings forecasts (MBEF). The latter association is this paper’s main indicator of opportunistic non-GAAP exclusions. The interaction variables

(26)

between executive gender and the opportunistic use of non-GAAP earnings. If male executives are more inclined to engage in earnings management, the coefficient for the interaction variable in both models should be positive. Following the second hypothesis, it is expected that both

CEOgender*PosExcl and CFOgender*PosExcl will have a positive coefficient. Same as for H1, total

exclusions (PosExcl) will also be decomposed into the underlying other exclusions (PosOtherExcl) and special exclusions (PosSpecialExcl). This results in the following regression model for CEO’s:

(5) MBEFt = γ0 + γ1PosOtherExcli,t + γ2PosOtherExcli,t + γ3CEOgenderi,t

+ γ4CEOgender*PosOtherExcli,t +γ5CEOgender*PosSpecialExcli,t

+ γ6Book-to-Marketi,t + γ7SalesGrowthi,t + γ8LnSizei,t + γ9Profitablei,t + γ10ROAi,t + υi,t And lastly, the following regression model is used for CFO’s:

(6) MBEFt = γ0 + γ1PosOtherExcli,t + γ2PosOtherExcli,t + γ3CFOgenderi,t

+ γ4CFOgender*PosOtherExcli,t +γ5CFOgender*PosSpecialExcli,t

+ γ6Book-to-Marketi,t + γ7SalesGrowthi,t + γ8LnSizei,t + γ9Profitablei,t + γ10ROAi,t + υi,t Just as before, the interaction variables are the primary indicators in these models. However, in this case they provide insight into the origins of any association found in model (3) and (4). They indicate whether the association between executive gender and non-GAAP opportunism is primarily driven by one of either underlying exclusion types. If men use non-GAAP exclusions more opportunistically, it might be expected that they use both types of exclusions more opportunistically. However, prior literature offers slightly less information to make this expectation. For sake of completeness, it is expected that the coefficients for all interaction variables CEOgender*PosOtherExcl, CEOgender*PosSpecialExcl, CFOgender*PosOtherExcl, and CFOgender*PosSpecialExcl will be

Executive Gender and the Opportunistic Use of Non-GAAP Exclusions

Table of contents

1 Introduction

2 Background

3 Hypothesis development

4 Data and method