CEO overconfidence and non-GAAP-earnings : opportunism or informative? : the influence of managerial overconfidence on non-GAAP reporting

CEO Overconfidence and Non-GAAP-earnings:

Opportunism or Informative?

The influence of managerial overconfidence on non-GAAP reporting

Name: Houda Benabbou

Student number: 11378484 Thesis supervisor: S.W.Bissessur

Date: 25-06-2018

Words: 19,508

MSc Accountancy & Control, specialization Accountancy Amsterdam Business School

Statement of Originality

This document is written by student Houda Benabbou 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.

Abstract

The aim of this research is to investigate whether CEO overconfidence has an influence on the use of non-GAAP reporting. The purpose of introducing non-GAAP reporting is to better depict core earnings by excluding items and to better inform investors. However, in the last decade, many regulators, academics and the financial press have expressed their concerns and scepticism regarding the motive of the use of non-GAAP reporting and started viewing this as a threat to the regular GAAP earnings. They claim it is unclear whether the motive of the use of non-GAAP reporting is to better inform investors or to mislead them. The use of non-GAAP earnings offers managers a considerable level of discretion on how to report firm performance to investors. Extant literature finds evidence that the use of non-GAAP reporting can lead to financial misreporting due to discretion offered to managers. Thus, non-GAAP earnings depends on the behaviour of the manager. Therefore, this research examines whether an overconfident manager is more likely to report non-GAAP earnings. This study uses the proxies based on the personal portfolio option exercising behaviour of CEO’s developed by Malmendier and Tate (2005a). This study provides empirical evidence that managerial overconfidence is positively associated with the use of non-GAAP reporting. In addition, this research finds evidence that the degree of managerial overconfidence increases or decreases the use of GAAP reporting. Overall, these results are consistent with the claims that non-GAAP reporting is used for opportunistic reasons. This study is the first to examine the association between managerial overconfidence and the use of non-GAAP reporting.

Keywords: Non-GAAP reporting, CEO overconfidence, Behavioural corporate finance, Financial misreporting, Opportunism

Contents

2.4 Relationship managerial overconfidence and the use of non-GAAP reporting ... 17

3 Hypothesis development ... 19

4 Research Design... 22

4.1 Data description Sample... 22

4.2 Non-GAAP earnings ... 26

4.3 Managerial Overconfidence ... 27

4.3.1 Measuring managerial overconfidence ... 27

4.3.2 Managerial overconfidence measures: Holder67 ... 28

4.3.3 Managerial overconfidence measure: NetBuyer ... 30

4.3.4 Managerial overconfidence measures: High Overconfidence and Low Overconfidence ... 31 4.4 Control variables ... 32 4.4.1 Firm size... 32 4.4.2 Financial leverage ... 32 4.4.3 Growth of firms... 33 4.4.4 Earnings Volatility ... 33 4.4.5 Loss indicator ... 34 4.4.6 Special Items ... 35 4.5 Regression specification ... 35

5 Results ... 38 5.1 Preliminary analysis ... 38 5.2 Descriptive statistics ... 38 5.3 Correlations ... 41 5.4 Empirical results ... 43 6 Conclusions ... 50

6.1 Discussion and conclusions ... 50

6.2 Future research ... 52

6.3 Limitations ... 52

7 Appendices ... 54

1 Introduction

“Companies Pollute Earnings Reports” (August 21, 2001), “Pro Forma (Non-GAAP) Profits Don’t Impress Pros” (November 25, 2001), these are several public remarks illustrated by Wall Street Journal from financial press that raised a number of red flags regarding the use of non-GAAP reporting. Another public remark by the SEC stated as follows: “We wish to caution public companies on their use of this ‘pro forma’ financial information and to alert investors to the potential dangers of such information” (SEC, 2001). In the last decade many regulators, academics and financial press have expressed their concerns regarding the motive of the use of GAAP reporting. These public remarks are consistent with the notion that the use of non-GAAP reporting can reflect the opportunistic motives by managers to overstate performance (Christensen et al., 2017). It is claimed to be unclear whether the use of non-GAAP reporting is to better inform financial statement users or if it is a way for CEOs to behave opportunistically for their own benefits (Doyle et al., 2013).

The purpose of this thesis is to examine the effects of the use of Non-GAAP reporting earnings on future performance measures. More specifically in this thesis, the consequences of managerial overconfidence on the use of non-GAAP earnings are researched. Therefore, the research question of this study is as follows:

RQ: Is managerial overconfidence positively associated with the use of non-GAAP reporting?

Providing an answer to this research question is important, because there are a lot of discussions and debates over the use of non-GAAP reporting. It is claimed to be beneficial to investors because it better depicts core operations on one hand, but on the other hand, non-GAAP reporting also affords discretion to managers, and firms can use non-GAAP reporting in more opportunistic ways. Prior studies show that there is an ongoing debate on whether managers use non-GAAP reporting to mislead or to inform investors (Bradshaw & Sloan, 2002; Bhattacharya et al., 2003; Doyle et al., 2003; Lougee & Marquardt, 2004; Johnson & Schwartz, 2005; Doyle, et al., 2013; Black & Christensen, 2009; Frankel et al., 2011). Therefore, this research is focused on investigating the influence of managerial overconfidence on the use of non-GAAP reporting. This is important because overconfidence can induce managers to make decisions make that may destroy the value of a firm (Roll, 1986).

The Securities and Exchange Commission (hereafter SEC) requires companies to disclose all the information, whether positive or negative, that can be relevant or useful for investors to make decisions, such as to buy, hold or sell a firms’ securities. A great part of this information falls under required disclosures mandated by the regulation, such as GAAP. Because the SEC recognizes the fact that not all relevant and useful information can be reflected in GAAP disclosure requirements, it allows companies to provide voluntary disclosures. One of the most common used type is non-GAAP earnings metrics, which are measures that deviate from GAAP disclosure requirements (Miller, 2008).

Non-GAAP reporting is a new phenomenon that started to become more popular during the mid-1990s, as managers, analysts and investors focused more on reporting forms other than GAAP based performance measures. This form of disclosing information to stakeholders has significantly increased over the last ten years (Black et al., 2017a), which also led to practitioners, regulators, investors and analysts becoming more involved in non-GAAP reporting. A prior study shows 88% of the S&P 500 firms use GAAP reporting, and non-GAAP adjustments increased the net income by 82% in 2015 (Coleman & Usvyatsky, 2015)

The motivation for the use of non-GAAP reporting for managers of firms is to better capture core operations and to better depict core performances, which is more beneficial to investors. With non-GAAP reporting customized earnings metrics are used, that can better depict core performance than GAAP-based earnings (Black et al., 2017a). However, the motive of public companies to report non-GAAP earnings is under scrutiny, because it is not clear whether managers’ motivation is to help investors to assess the core operating performance of a company or change perception of the operating performance that could possibly mislead investors (Curtis et al., 2014).

But there is also criticism on the use of non-GAAP reporting, because its usage can lead to exclusion of income-decreasing items and managers are being questioned on whether they exclude this just to depict better satisfactory earnings measures (Black et al., 2017a). Prior studies that investigate the incentives that motivates the use of non-GAAP reporting show that managers and analysts often exclude items that occur one time, like litigation charges (Bradshaw & Sloan, 2002). Prior studies also show that investors seem to respond more to non-GAAP earnings than they do to non-GAAP earnings. The use of non-non-GAAP reporting can give managers the opportunity to meet strategic earning targets by excluding recurring items, that

they would have missed if they would have reported based on GAAP-reporting (Doyle et al., 2003).

Besides the fact that non-GAAP practices gives managers the opportunity to exclude GAAP earnings components, it also does not require them to inform which items they chose to exclude from disclosing in the calculation of the non-GAAP earnings. This has led to regulators becoming more sceptical towards non-GAAP reporting and therefore, the SEC has implemented Regulation G (hereafter Reg G), which requires public companies that disclose non-GAAP earnings to include a presentation of the most directly similar GAAP performance measure and also a reconciliation of the disclosed non-GAAP performance measure to the most directly similar GAAP performance financial measure (SEC, Regulation D, 2003). The main goal of the implementation of Reg G is to improve the transparency and quality of the non-GAAP reporting. Even after the implementation of this regulation, evidence shows that aggressive non-GAAP reporting remains. It shows that managers are more willing to exclude one-time losses compared to one-time gains (Curtis et al., 2014). Despite the fact that the SEC has monitored the non-GAAP reporting practices since 2001, and the issue of the specific regulation (Regulation G), the concerns continued to exist in recent years. Moreover, the former chairperson, Mary Jo White, of the SEC has recently expressed that non-GAAP reporting practices have to be ‘reined in’ to keep potential- aggressive reporting practices in check through more regulations (Christensen et al., 2017).

This research is designed with an archival approach. In order to examine the association between CEO overconfidence and the use of non-GAAP reporting, the most largest and recent publicly available dataset of manager-disclosed non-GAAP metrics, which captures managers’ reporting choices collected by Bentley et al. (2018) is used. In this study, the stock-option measures Holder67 and NetBuyer developed by Malmendier and Tate (2005a), based on the personal portfolio decisions of a CEO are used to measure CEO overconfidence. In addition, the variables High Overconfidence and Low Overconfidence developed by Campbell et al. (2011) to measure the level of CEO overconfidence are also used. The data sample regarding CEO option behaviour and control variables is extracted from the databases ExecuComp and Compustat and ranges from the years of 2003 until 2016 and contains 2,191 U.S. firms. To enhance the robustness of this research several different measures of CEO overconfidence, a wide timespan and multiple control variables are added.

This study provides initial evidence that CEO overconfidence is positively associated with the use of non-GAAP reporting. This implies that an overconfident CEO is more likely to report non-GAAP earnings. All four CEO overconfidence measures support the hypotheses. However, the proxy NetBuyer indicates a positive but insignificant association between CEO overconfidence and non-GAAP reporting, this can be due to different datasets, signalling and private inside information.

The results of this research contributes to the broader voluntary disclosure literature and the behavioural corporate finance literature. As previously stated, both CEO overconfidence and non-GAAP reporting can lead to financial misreporting and opportunistic behaviour. However, there has not been a research on the association between CEO overconfidence and non-GAAP reporting yet. Therefore, this study examines the influence of CEO overconfidence on the use of non-GAAP reporting and offers initial evidence on this association. The results of this study confirm the concerns that non-GAAP reporting are used for opportunistic reasons, and that managerial overconfidence can lead to financial misreporting.

The remainder of this thesis is structured as follows. The next section reviews the literature on the agency theory, on what has already been established about non-GAAP reporting in the literature, CEO overconfidence and the relationship between non-GAAP reporting and CEO overconfidence. Section 3 states the hypothesis developed for this research. In section 4 the empirical methodology and sample construction are discussed. Thereafter the empirical results and their interpretations are discussed in section 5. Finally, the conclusions, limitations and directions for future research are discussed in the last section.

2 Literature review

This chapter contains the theoretical background of this study. It gives an extensive overview of the literature on the research subject. First, the agency theory and the agency relationship between the managers and stakeholders is discussed. Second, the new reporting practice non-GAAP reporting, is discussed and different insights on what is known about the reporting form in the current literature practice are contemplated. In the third paragraph the theoretical background on CEO overconfidence is discussed. At last, the relationship according the current theory between non-GAAP reporting and CEO overconfidence is discussed.

2.1 Agency Theory

It is important to research whether there is an association between CEO overconfidence and the use of non-GAAP reporting. The agency theory is important to understand this discussion. This is one of the most researched theories in the economic science, that describes the relationship between the agent and the principal. The agency relationship can be defined as one in which the principles engage the agent to carry out a service on their behalf that involves delegating decision-making authority to the agent.

The literature describes the risk sharing problem among groups or individuals that arises when cooperating parties have different views towards risks. The so-called agency problem occurs when the principal delegates work to the agent, that must perform the work. When the agency theory is used within a firm, the CEO is considered as the agent and the shareholders and stakeholders are considered as the principles of the firm, because they own the firm. The agency theory discusses two problems that occur in the relationship between the agent and the principal and how to resolve these problems (Eisenhardt, 1989).

The first problem is when the agent and the principals think in their own interest, the problem arises as the objectives or interests of the principal conflict with the objectives or interests of the agent and the agent has more inside information than the principal. This is also called information asymmetry in the literature. Because of the information asymmetry, it is difficult and/or costly for the principal (stakeholders/shareholders) to verify whether the behaviour of the agent is appropriate or not. Information asymmetry can be decreased by requiring companies to provide more information (Eisenhardt, 1989).

The second issue described by the agency theory is when the agent and the principal have different attitudes towards risk. When this occurs, the problem is that both parties may choose different approaches because of the different risk preferences. The agency theory tries to describe the relationship between the principal and the agent using the metaphor of a contract. The focus of the theory is to determine the most efficient contract governing the relationship given assumptions about organization, information and people.

Within a firm, the CEO is considered the agent and the shareholders are considered as the principles. The CEO is supposed to act in the best interest of the shareholders. But due to the conflicting interests, this is not always the case. The CEO wants to have a higher reputation and is more likely to make decisions that are beneficial for him and the shareholders who also want high returns on their investments. Because of information asymmetry it is difficult for the shareholders to monitor the actions of the CEO. Non-GAAP reporting provides managers with the opportunity to influence their earnings with discretion, based on their behaviour. The decision of the manager to use non-GAAP measures or not relies on the trade-off between the benefits and costs of non-GAAP reporting. To decrease information asymmetry, managers can use non-GAAP metrics to provide a better view of performance, when GAAP earnings are affected by noise (Black et al., 2017b).

2.2 Non-GAAP earnings

Over the last twenty years, the use of non-GAAP reporting has increased, and this growth has led analysts, investors and managers to accept non-standard performance measures to evaluate the performance of a firm (Black et al., 2017a). Non-GAAP reporting is the disclosure of adjusted earnings measures. What are non-GAAP earnings? Non-GAAP earnings represents the core earnings from a firm, excluding specific components of the GAAP-earnings (Black et al., 2017a). Managers have the possibility to voluntary disclose non-GAAP earnings and to determine the exclusion at their own discretion. Regulators, standard setters and academics became sceptic and started viewing non-GAAP reporting as a threat to the regular GAAP-reporting, because it is not clear whether the purpose of the use of non-GAAP earnings is informative for stakeholders or a tool for managers to mislead investors. The scepticism regarding the motives behind the use of non-GAAP reporting has led regulators (SEC) to caution the stakeholders about the possible misleading role of non-GAAP reporting. The chairman of IASB expressed his concerns about non-GAAP reporting, contending it as increasingly misleading and it less comparable and/or verifiable (Black et al., 2017b).

Opportunism can have an influence on managers’ choices in excluding components, since the non-GAAP earnings, that are disclosed in press releases are not audited (Black & Christensen, 2009; Doyle et al., 2003). However, prior studies have shown that companies are more likely to disclose non-GAAP earnings if GAAP earnings are less informative (Lougee & Marquardt, 2004) and also that non-GAAP earnings seem to be more persistent compared to GAAP earnings (Bhattacharya et al., 2003). According to Black et al (2017a) non-GAAP metrics are more valuable to investors than GAAP performance metrics

There are numerous studies that argue that managers can use GAAP-reporting in an opportunistically way to increase private benefits or equity valuations. Evidence was found that when the stock prices and earnings of a firm decreases, the use of non-GAAP reporting dramatically increases (Bhattacharya et al., 2004). Other research show that managers are more likely to emphasize metrics that reflect better firm performance (Bowen et al., 2005). Black and Christensen (2009) found evidence that managers outline their non-GAAP earnings in such a way that they could meet earnings targets. Earlier research has found evidence that there is association between non-GAAP reporting and opportunistic behaviour that influence firm performance. Literature identifies four non-GAAP reporting decisions: (1) whether a manager discloses non-GAAP earnings metrics in the press release of the earnings announcements

(Marques, 2006), (2) the importance given to the non-GAAP metrics in the press releases (Bowen et al., 2005), (3) the exclusion of recurring items from the standard GAAP earnings when calculating non-GAAP earnings (Black & Christensen, 2009), (4) whether the reconciliation between the non-GAAP numbers and GAAP numbers is disclosed or not (Marques, 2010).

The most recent literature of Black et al. (2017b) on non-GAAP reporting discusses all the insights and evidence on this form of reporting, non-GAAP reporting for over the last two decades. Black et al. (2017b) attempt to answer and summarize different questions of the increasing literature on non-GAAP reporting and provide a comprehensive overview.

GAAP vs non-GAAP earnings

Several studies compare the reactions of investors towards GAAP earnings versus non-GAAP earnings and what this shows is that in the recent years more investors seem to rely and focus on non-GAAP performance measures than on the standard GAAP measures (Black et al., 2017b). Prior research finds evidence that providing non-GAAP information can have an influence on the assessments of future cash flows of the investors in a more optimistic way. Furthermore, this research also shows that using non-GAAP reporting can ensure that stock prices reflect the fundamental value more accurately (Hirshleifer & Teoh, 2003). Other studies find that stakeholders view the non-GAAP earnings adjusted by managers as being more valuable and informative than the regular GAAP earnings (Bhattacharya et al., 2003).

Users of non-GAAP information

Since evidence found that investors rely more on non-GAAP earnings numbers than the regular GAAP earnings, researchers attempted to explore which types of financial statement users react to non-GAAP earnings. Formerly, the nonprofessional investors were more likely to rely on non-GAAP earnings in their decision making compared to the more professional investors (Elliot, 2006). However, in the recent years, evidence was found that the more professional stakeholders can also be susceptible to the non-GAAP earnings for their decision making. Analysts consider non-GAAP information as reliable, if firms reconcile the non-GAAP measures to the GAAP earnings (Elliot, 2006). In addition, other studies show that the non-GAAP earnings can have an influence on the earnings per share forecast of analysts (Andersson & Hellman, 2007). As it has been mentioned before, non-GAAP earnings seem to be

contemplated as more informative by nonprofessional stakeholders as well as professional stakeholders through the recent years.

Type of adjustments

The main distinction between non-GAAP and GAAP reporting is in the calculation by excluding nonrecurring items. The motivation for managers to exclude these items, is to draw financial statement user’s attention on sustainable earnings (Bhattacharya et al., 2003). Non-recurring items include profits and losses on disposals of assets, acquisition and merger costs and extraordinary items, transitory items. Curtis et al. (2014) find evidence that managers can be inconsistent in the choices of excluding items. Excluding these one-time losses or gains results in higher or lower earnings. Literature shows that not only non-recurring items, but also recurring items are being excluded in the calculation of non-GAAP earnings by managers in the last years increasingly. These items are claimed by managers to be cash and/or non-operating in their nature, like depreciation, stock compensations (Bradshaw & Sloan, 2002). Black and Christensen (2009) find evidence that the exclusion of recurring items, like stock-based compensation, depreciation and research and developments expenses, is associated with opportunistic motives. The extant literature shows that the exclusion of recurring items represents a lower quality of non-GAAP adjustments compared to excluding non-recurring items (Doyle et al., 2003; Kolev et al., 2008) The main reason the exclusion of recurring items has increased in the last decades is because of accounting standards mandating the inclusion of these items in GAAP earnings (Black et al., 2017c).

Regulation

The use of non-GAAP reporting has increased scepticism by regulators and literature due to unclear motive of use of non-GAAP reporting; misleading the financial statement users (opportunism) or informing the financial statement users. This scepticism had led to stricter regulation of non-GAAP reporting. Due to these concerns, SEC implemented Reg G in 2002 to caution financial statement users of possible financial misreporting. Reg G mandates that non-GAAP earnings numbers must comprise the most directly comparable GAAP number, that non-GAAP may not be presented in a way that could possibly mislead financial statement users and at last that the non-GAAP earnings number must contain quantitative reconciliation with the most comparable GAAP number (Heflin & Hsu, 2008). Consequently, Reg G the frequency of the use of non-GAAP reporting practice decreased directly after the implementation.

However, the use of non-GAAP reporting recovered shortly after regulation and remained increasing consistently (Bentley et al., 2018). Several recent studies find that Reg G increased the quality of non-GAAP reporting. Besides the increased quality of this reporting practice, prior evidence suggests that investors seem to react more to non-GAAP earnings compared to GAAP-earnings following the enactment of the regulation regarding non-GAAP reporting (Marques, 2006). Several studies conclude that due to stricter regulations, financial statement users are less likely to be misled by non-GAAP earnings and the quality of the new reporting practice have increased during the recent years (Jennings & Marques, 2011).

To provide a better insight of the effect and difference of non-GAAP earnings and GAAP earnings, a visual representation is included in appendix 7B. This contains a non-GAAP press release of Twitter Inc, it shows the reconciliation between GAAP and non-GAAP figures. The difference between the results are remarkable. The example shows that at the end of 2015 and 2016 Twitter report a net loss of respectively $457 million and $521 under GAAP-reporting. While the reports under non-GAAP are significantly different, Twitter Inc. shows a non-GAAP income of $115 million and $119 million, respectively. This indicates what a difference non-GAAP earnings can make and how this could possibly influence the decisions of financial statement users (Twitter Announces Fourth Quarter and Fiscal Year, 2017).

2.3

According to the research literature, overconfidence can be defined in three distinct ways, which is (1) over placement of someone’s performance compared to others, (2) overestimation of someone’s actual performance, (3) excessive precision in someone’s beliefs (Moore & Healy, 2008). Managerial overconfidence is the case when managers seem to overestimate their future return on investments. On the one hand these managers tend to systematically overestimate future returns from investments or systematically overestimate the impact and probability of events that are favourable on their firm’s cash flows. And on the other hand, they can underestimate the impact and the probability of adverse events on their firm’s cashflows.

In prior research, evidence was found that managerial overconfidence has an influence on managerial behaviour and economic decision making (Kidd & Morgan, 1969), (Lardwood & Whittaker, 1977). Prior literature distinguishes overconfidence in two main facets: over-optimism and miscalibration (Skala, 2008). Over-over-optimism refers to an overestimation of the mean, when managers are positive about outcomes that are not certain in an unrealistically

way. While on the other hand miscalibration refers to an underestimation of variance, where managers underestimate uncertainty. The psychology literature shows overconfidence as decision making under uncertainty as miscalibration and over-optimism (Kidd & Morgan, 1969). Humans tend to be more optimistic about the outcomes of their own decisions (Lardwood & Whittaker, 1977).

Theory suggest that the personal characteristics of a manager of CEO in large firms can result in distortions in investment decisions. Malmendier and Tate (2005c) find evidence that firms that are dependent on equity, overconfident CEOs tend to have a higher sensitivity of investment to the amount of cash within the firm. The two main causes for distortions in investments decisions are information asymmetry between principal and the agent and the misalignment of the stakeholders and CEOs interests (Jensen & Meckling, 1976). Overconfident CEOs underestimate the probability of failure and are more likely to select the capital construction that is more dependent on debt than on equity. This implicates that CEO overconfidence decreases the amount of equity of a firm (Shefrin, 2001). Prior social psychological literature shows that individuals are more likely to overstate their skills, when assessing them. So, in the case of overconfident managers, they expect that their actions and behaviour will create good results. (Lardwood & Whittaker, 1977). Theory describes three factors that trigger overconfidence, a higher degree of commitment towards good outcomes (1), the illusion of control (2) and abstract reference points that make it difficult to compare performance among individuals (3) (Weinstein, 1980). CEOs are more likely to be committed to high firm performance since their personal benefits and wealth is tightly related to the firm’s stock prices.

Malmendier and Tate (2005) have developed metrics to measure managerial overconfidence based on the stock option holding and exercising decisions and on the net stock purchases. The authors consider CEOs as overconfident if they hold stock options that are more than 67% in the money. That is, if the price of the stock exceeds the exercise price by more than 67%. The specific percentage of 67% originates the model of Hall and Murphy (2002) from earlier literature. The model uses a detailed dataset on CEOs stock options holding and exercise decisions. And the model recognizes that risk-averse CEOs characteristically hold undiversified portfolios and should exercise options in an earlier stage if they are expected to maximize their utilities (Campbell et al., 2011).

Prior literature documents that it is mainly the high-ranking managers that are susceptible to being too optimistic about the outcomes and overestimate outcomes to which they are highly committed (Weinstein, 1980). There are two reasons of why top executives can show this behaviour. First, CEO compensation depends for a great part on the performance of a firm and the human capital is associated with the returns of a firms. Secondly, a CEO usually has the final say in large and important decisions within a firm, this leads to a possible underestimation of the likelihood of a failure (Malmendier & Tate, 2005b).

2.4 Relationship managerial overconfidence and the use of non-GAAP reporting

The research on managerial overconfidence is related to the accounting literature and to managerial financial reporting, which is the main subject of this study. Prior research finds evidence that companies that miss their own forecast are more likely to have less flexibility in making accounting decisions and are less familiar and experienced with forecasting (Chen, 2004). Moreover, evidence was found that managerial overconfidence has an influence on forecasting decisions and other evidence shows that managerial overconfidence can result in aggressive accounting for earnings forecasting, by augmenting the use of discretionary accruals (Hribar & Yang, 2016). Furthermore, prior literature also found evidence that there is a positive association between managerial overconfidence and the chance of financial statement fraud (Schrand & Zechman, 2011). These results indicate that managerial overconfidence can result to a higher chance of optimistically biased misstatements in financial reporting. Other evidence was found that firms have a higher tendency to report financial information with a more positive tone for positive reactions of the market (Millo, Wisniew et al., 2014).

Numerous studies that investigate the motives for the use of non-GAAP reporting, focus on two contradictory incentives: opportunism and informativeness. The main motivation of the implementation of non-GAAP reporting was to better inform about the core operations of a firm and to provide value-relevant earnings information for financial statement users. Hence, numerous studies find evidence that supports this incentive to better inform financial statement users about the core earnings or to provide value-relevant earnings information (Black et al., 2017a). For example, Curtis et al. (2014) find that to show a more accurate view of the core performance in calculating the non-GAAP earnings, managers systematically include one-time profits and exclude one-time items. Other theory suggest that investors have greater reliance on non-GAAP earnings information than on the GAAP earnings numbers, and that investors

are more likely to respond to non-GAAP information than to GAAP metrics (Bhattacharya et al., 2003). In addition Johnson and Schwartz (2005) found evidence that suggests financial statements users are not misled by non-GAAP earning metrics, especially after the implementation Reg G. Andersson and Hellman (2007) conclude that the motivation for managers and financial analysts’ tendency toward the use of non-GAAP reporting is that they seek to show a more simple and comprehensible view of financial reporting.

However, despite the evidence found that the incentive for use of non-GAAP reporting is supposed to inform financial statement users, previous research shows the opposites finding examples of managers and CEOs misleading non-GAAP disclosures (Black D. E et al., 2017b). Bhattacharya et al. (2003) and Black and Christensen (2009) conclude that despite the fact the exclusion of the one-time items was to generate a performance measure that gives a better reflection of a sustainable performance, the exclusion of recurring items on the other hand seems to be less justifiable. Other academics find that the exclusion of recurring items is the type of non-GAAP adjustments of the lowest quality and that these can mislead financial statement users (Bentley et al., 2016; Black D. E. et al 2017a). In addition, numerous studies argue that the motive to exclude items is to mislead financial statement users by persuading them that an ‘artificial’ performance metrics meets a favourable outcome (Bhattacharya et al., 2004; Black & Christensen, 2009; Marques, 2006).

Agency theory indicates that CEO compensations that are based on performance is a manner to align the interest of the insider and outsider of a firm (CEOs and stakeholders) (Jensen & Meckling, 1976). So these performance-pay can create incentives to encourage CEOs to increase the wealth of shareholders, but can also create incentives to manipulate financial information. Previous literature finds evidence that changes in share prices fluctate around option awards, suggesting that CEOs influence the expectations of investors by pushing bad news and delaying good news (Aboody & Kasznik, 2000). The authors find evidence that managers make oppurtunistic voluntary disclosures that maximize their stock option compensation, which in this case is excluding items from non-GAAP earnings.

However, there is a perceived distinction between the motives of managers on the one hand and analyst on the other hand for providing non-GAAP earnings. It seems to be less likely for analysts to mislead financial statement users and focus on providing sustainable earnings numbers. In accordance with this, Brown et al. (2015) find that analyst commonly exclude one-time items from their earnings prognoses.

3 Hypothesis development

Regulators and academics have expressed their scepticism about the use of non-GAAP earnings. The use of non-GAAP earnings offers managers a considerable level of discretion on how to report financial firm performance to financial statement users (Isidro & Marques, 2013). Many previous studies on non-GAAP research whether CEOs choose to report non-GAAP earnings to (1) provide financial statement users with a better-quality measure to determine firm value to forecast future firm performance (i.e. informative motive) or (2) it is an attempt to increase reported performance (i.e. opportunistic motive). The reason behind the trade-off between the informative motive and the opportunistic motive is due to the discretionary nature of non-GAAP reporting (Bentley et al., 2018). With managers having this discretion, there is place for opportunistically behaviour by managers.

Prior studies show that there is an ongoing debate on whether managers use non-GAAP reporting is to mislead or to inform investors (Bradshaw & Sloan, 2002; Bhattacharya et al., 2003; Doyle et al., 2003; Lougee & Marquardt, 2004; Johnson & Schwartz, 2005; Doyle et al., 2013; Black & Christensen, 2009; Frankel et al., 2011). Bhattacharya et al. (2004) finds that the use of non-GAAP reporting increases dramatically when stock prices and earnings of a company decrease. This indicates that the use of non-GAAP earnings can mislead investors to portray a better firm performance. In this case non-GAAP earnings may not be so informative anymore. This is concerning, since one of the motivation of the introduction of non-GAAP reporting is to give investors more information that could be relevant or useful for them to make decisions.

Non-GAAP earnings are completely at the discretion of managers, they can choose for themselves which item to exclude from the traditional GAAP-earnings numbers. To give a better view of this exclusions, to better capture core operations and to better depict core performance, which is more beneficial to investors. With non-GAAP reporting you use customized earnings metrics, that can better depict core performance than GAAP-based earnings (Black et al., 2017a). Examples of exclusion of items that firms remove from the traditional GAAP earnings to calculate non-GAAP earnings can be: litigation charges, stock-based compensation, depreciation and amortization, foreign exchange effect on revenue and restructuring costs. There is evidence from prior research that managers use non-GAAP reporting to exclude recurring items to meet strategic earnings targets, that would have been missed if they have reported based on GAAP-reporting (Doyle at al., 2003). Other studies show

that managers can use non-GAAP reporting to opportunistically increase equity valuation or private benefits.

This implicates that personal managerial behaviour can affect financial reporting, which can again affect the decisions of investors. Therefore, this research focuses on examining the influence of managerial behaviour on the use of non-GAAP reporting. Since there is ongoing debate on whether the use of non-GAAP reporting aids investors in making better decision or whether it is to mislead investors and for managers to benefit themselves. In this research the focus is especially on the “misleading investors” aspect, since that can be achieved by managerial behaviour, such as managerial overconfidence.

Prior research has shown that managerial overconfidence can lead to financial misreporting. In the previous paragraph, literature reviews the two main facets (overoptimistic and miscalibration) (Skala, 2008), in this research the focus is on the situation in which managers are over-optimistic. Over-optimistic managers tend to overestimate the magnitude and probability of positive shows to future cash flows (Ahmed & Duellman, 2013). Prior research finds evidence that overconfidence increases the amount of optimism in management forecasts (Hribar & Yang, 2016). Therefore, this study focuses on the examination of the effects of overconfidence on the use of non-GAAP reporting. This is important because overconfidence can induce decisions managers make that can destroy the value of a firm (Roll, 1986).

The motivation for managers/CEOs to financial misreporting with the use of non-GAAP reporting can be explained through various ways. Prior evidence shows that non-non-GAAP disclosures can be used to enhance market valuation, and this increases CEO compensation that is based on performance (Isidro & Marques, 2013). Bhattacharya et al. (2007) find evidence that less sophisticated investors trade based on non-GAAP earnings, increasing market prices. Prior evidence shows that the market responds is more responsive to earnings announcements of firm when the non-GAAP earnings are higher than the comparable GAAP items (Isidro & Marques, 2013). Furthermore Black et al. (2011) find evidence that bonus compensation that is focused on the short-term performance is more likely to encourage opportunistic behaviour in non-GAAP reporting.

Non-GAAP reporting offers managers and CEOs the possibility of discretion. How will an overconfident manager react to this with the use of non-GAAP reporting? Managerial overconfidence can lead to financial misreporting, and managers can use non-GAAP reporting

to benefit themselves and mislead investors. Therefore, this study focuses on investigating the effects of managerial overconfidence on the use of non-GAAP reporting. This suggests the following hypothesis:

H1 Managerial overconfidence is positively associated with likelihood to opportunistically

report non-GAAP earnings.

So, this study predicts that when managers are overconfident they are more likely to opportunistically exclude items from GAAP earnings, resulting in non-GAAP earnings. Campbell et al (2011) extends the theory of Malmendier and Tate (2005) and creates two additional measures of managerial overconfidence by making a distinction between high and low managerial overconfidence. In addition to the research on association between managerial overconfidence and non-GAAP reporting, this study also researches whether the level of overconfidence (low versus high overconfidence) has an influence on non-GAAP reporting. Therefore, the following additional hypotheses are added to this research.

H2a. The propensity to disclose non-GAAP earnings is greater when the manager is highly

overconfident.

H2B. The propensity to disclose non-GAAP earnings is smaller when the manager/CEO is less

4 Research Design

This chapter of the thesis gives an extensive overview of the research methodology that is used to examine the association between CEO overconfidence and non-GAAP earnings metrics. In the first paragraph there will be a description of the data samples used for this research. Secondly, an overview will be given of the proxies to measure the independent variable CEO overconfidence and how to construct these. Thirdly the control variables used in this research are briefly discussed. At last the regression specification will be given.

4.1 Data description Sample

Formerly, academics had to either costly hand collect data about non-GAAP metrics or use proxies for managers’ disclosures from analyst forecast data providers, such as I/B/E/S. This is because there has not been an archived dataset available for managers’ non-GAAP metrics. But several academics have expressed their concerns about the appropriateness of using I/B/E/S, because it is not clear if analysts systematically differ from the managers’ disclosed metrics. Furthermore, theory argues that using the I/B/E/S to research the incentives of managers when using non-GAAP reporting underestimated the aggressiveness of the financial reporting choices and overstates the quality of non-GAAP reporting by managers (Bentley et al., 2018).

Because of this ambiguity regarding the examination of non-GAAP reporting incentives, Bentley et al. (2018) created and provided the first large-sample dataset regarding the non-GAAP earnings metrics disclosed by managers. This data set is an improvement compared to other data sets, because (1) it is more comprehensive and significantly larger than previous hand-collected data samples, (2) it captures the non-GAAP reporting metrics by managers in a more accurate way, (3) the data set is publicly available for everyone. Therefore, in this research, the publicly available dataset of manager-disclosed non-GAAP earnings metrics will be used.

The sample data of this study contains updated manager non-GAAP EPS disclosure data collected of firms included in the publicly available dataset1 of Bentley et al. (2018),

1_{The dataset of Bentley et al. (2018). https://sites.google.com/view/kurthgee/data. This link contains} updated manager non-GAAP EPS disclosure data collected using the method in Bentley et al. (Journal of Accounting Research, Forthcoming), with data from 2003 through part of 2016.

contains only North Americans firms. The data sample period is between 2003 and 2016 of manager non-GAAP EPS disclosures. The data sample period starts at the year 2003, because the identification of non-GAAP reporting uses only 8-K disclosed earnings announcements of firms on EDGAR (Electronic Data Gathering, Analysis, and Retrieval system). EDGAR is database that contains fillings and that performs validation, automated collection, indexing, acceptance and forwarding of submissions by firms that are required by the law prescribed by the SEC and are freely available to the public (SEC, 2010). The mandatory filings of 8-k for earnings announcement commenced in 2003 with the implementation of Reg G.SEC-fillings are financial statements or other documents that are required for public companies by the SEC for informative purposes for financial professionals, investors and other financial statement users that rely on these fillings about firms they evaluate for investment decisions. 8-K fillings notify financial statement users of current events or changes Between quarterly reports that may be important for decision-making.

From this data sample, the following items are excluded: real statements investments trusts, firm quarters for which 8-K earnings announcement cannot be identified, firm quarters that report discontinued operations and extraordinary items. The final sample data consists of 115.370 firm-quarter observation of 7,090 companies. The dataset contains a ‘GVKEY’ which is a unique six-digit key number that is assigned to each firm that are included in the capital IQ Compustat database. Further, the dataset contains two variables: Mgr_Exclude, and MGR_NG_EPS. The first variable ‘MgrExclude’ has two possible values (1,0) and informs whether managers report non-GAAP metrics or not. This variable is labelled as one if the earnings announcement contains a non-GAAP EPS sentence, and a zero if otherwise (Bentley et al., 2018).

Table 1. Non-GAAP dataset Bentley et al. (2018)

Description Number of

observations Firm-quarter observations from Compustat-CRSP-end date

from January 1, 2003 to December 31, 2016 I/B/E/S universe

with fiscal quarter earnings announcement date and CIK 115,370

It is important to recognize that the data is at quarterly level, this indicates that it can be possible that a firm discloses a non-GAAP measure in a one quarter, but not in another quarter. For this research, data at yearly level will be used. Therefore, it is important to determine when to classify as mgr_exclude=1. In this research a firm-year is classified as mgr_exclude=1, if the fourth quarter has a non-GAAP disclosure. Selecting the fourth quarter of every firm-year can be done by merging the data with data from Compustat Annual based on the GVKEY and DATADATE, because DATADATE is automatically the fourth quarter.

The data used in this research can be separated intro three different parts. The main dataset is the publicly available dataset of Bentley et al. (2017), the dataset is named as ‘Dataset 1 Non-GAAP”. Based on the firms (gvkeys) in this dataset, the remaining data for the other variables are collected in different databases. In the following table 2, the datasets are divided into three parts with each its own source. The second dataset contains the data that is collected to measure Holder67, NetBuyer, High_OC100 and Low_OC30. The last dataset contains data to measure the control variables.

To examine the association between CEO overconfidence and non-GAAP earnings quality, detailed information based on the CEOs net stock purchases and on the stock option’s holding and exercising decision of (exercise price, number of shares and residual etc.) each firm by year is required. Information regarding the executive compensation can be collected from Execucomp, which is a database available at Wharton Research Data Services

TABLE 2 Data Sample Selection

Dataset Variables Source Items in database

Dataset 1 Non-GAAP

Non-GAAP Publicly available dataset Bentley et al. (2017)2 Mgr_exclude Dataset 2 CEO Overconfidence Holder67 NetBuyer High_OC100 Low_OC30 Execucomp Compustat PRCC_F OPT_UNEX_EXER_EST_VAL OPT_UNEX_EXER_NUM SHROWN_EXCL_OPTS Dataset 3 Control Variables Firm size(LN) Leverage Loss indicator Growth of firms Earnings Volatility(LN) Special items

Compustat Total Assets(AT) Total Liabilities(LT) Net Income(NI)

Sales/Turnover(SALE) Income Before Extraordinary

Item(IB) Special Items (SPI)

that provides annual data on executive compensation and covers data from 1992 till present. To create the dummy proxies Holder67 and Holder100, the collected variables are:

‘Estimated value of in-the-money unexercised exercisable options’ and ‘Unexercised

Exercisable Options’ from Execucomp and ‘Stock price fiscal year-end’ from Compustat. For the proxy ‘NetBuyer’ the variable SHROWN_EXCL_OPTS was collected from Execucomp. The data is collected based on the gvkey from the master main file dataset 1, datadate, year and the chosen sample period was 2002-2016. The sample period is from 2003 till 2016, but because of the CEO overconfidence proxy ‘NetBuyer’, an extra prior year is added to compute the additional bought or sold stocks by a CEO compared to the previous year. In table 3 the variables used for the creation of the proxies for managerial overconfidence are shown. The data is coming from Execucomp and Compustat and are merged afterwards, resulting in the residual observations of 12,671.

TABLE 3. Dataset 2 Proxies CEO Overconfidence

Panel A: Input variables from Execucomp

Variable name Description

GVKEY Year Execid

Global Company key Year

Executive ID number

EXEC_FULLNAME Executive full name

Co_name Company Name

Pceo Current CEO

OPT_UNEX_EXER_EST_VAL Estimated Value of In-the-Money Unexercised OPT_UNEX_EXER_NUM Unexercised Exercisable Options

SHROWN_EXCL_OPTS Shares Owned - Options Excluded Panel B: Input variables from Compustat

PRCC_F Datadate

Stock price fiscal year-end Financial year in Compustat

TABLE 4 CEO Overconfidence Data (2)

Description n

Starting observations CEO Execucomp 160,296

Drop if Pceo has missing values -139,354

Starting observation PRCC_F Compustat 65,747

Merge based on gvkey and fyear

keep if _merge==3 24,692

Create dummy variables Holder67, Holder100 and NetBuyer)

The third part of the data used in this thesis is regarding the control variables that are included in this thesis. All data regarding the control variables is collected from WRDS Compustat. In the following table the used variables used are presented below. The sample period is from 2000-2016, this is because of the control variable earnings volatility, where data prior three year was needed. After creating and constructing all required control variables with the collected data, the three datasets are all merged in to one main datasets. The remaining amount of observation contains 11,107 observation and 2,191 firms.

TABLE 5. Dataset 3 Control Variables

Panel A: Input variables from Compustat

Variable name Description

GVKEY Datadate AT

Global Company key

Financial year in Compustat Total Assets

LT Total Liabilities

NI Net Income

IB Income before extraordinary items

SPI Special items

SALE Sales/Turnover

TABLE 6. Sample selection dataset 3

Description n

Starting observations Compustat 2000-2016 96,937

Drop if missing values -18,441

Observations 78,496

Creation of all control variables Merge dataset 1,2 and 3 keep if _merge==3

Sample 12,341

After trimming data (winsorization & log transformation) -1,234

Final sample 11,107

Firms 2,191

4.2 Non-GAAP earnings

Non-GAAP earnings are the adjusted type of the standard GAAP earnings that some CEOs report to their financial statement users. CEOs report non-GAAP figures and exclude specific components that are deemed to be less representative or informative of the core operations of a firms. Whether a manager reports non-GAAP earnings or not is the dependent

variable of this research. The dependent variable in this research is the decision whether a manager discloses a non-GAAP measure in the earnings announcement or not. This information is publicly available in dataset2 of Bentley et al. (2018). Where there are two possible values: 1 if the managers exclude items from the non-GAAP numbers and therewith providing GAAP earnings, 0 if otherwise. Like theory suggests, the disclosure of non-GAAP metrics may be associated with opportunism, because non-non-GAAP earnings typically depict better firm performance than GAAP earnings.

4.3 Managerial Overconfidence

4.3.1 Measuring managerial overconfidence

Malmendier and Tate (2005) argue that constructing a plausible measure for managerial overconfidence is a big challenge, because it cannot be measured directly. Malmendier and Tate (2005) construct measures for CEO overconfidence based on the stock option holding, exercising decisions and the net stock purchases of a CEO (Campbell et al., 2011). The authors construct three proxies of CEO overconfidence that are based on the personal portfolio decisions of CEO. The proxies of CEO overconfidence are based on shareholding status and Malmendier and Tate (2005) infer CEO beliefs on future performance of firm by measurement of their exposure to idiosyncratic risk. The three measures are as follows: Holder67,

Longholder and NetBuyer. To identify CEO overconfidence for the first two, the measures Holder67 and Longholder use the timing of option exercises. The last measure NetBuyer uses

the regular acquisition of firm’s stocks. These measures are based on the under diversification of CEOs because compensation contracts include great quantity of stocks and option stocks. And to maximize the incentives for CEOs the sale of these stocks is limited, and the trading of options are prohibited. Hence, the human capital of CEOs is invested in their own company, which implicates that a negative outcome within their firm will also negatively affect their personal wealth and affect their outside employment options (Malmendier & Tate, 2005a). Due to overconfidence CEOs tend to overestimate future returns on investment, therefore they expect that the stock prices will continuously increase under their leadership. The main characteristics of CEO overconfidence are self-attribution by the CEO and overestimation of

2_{The dataset of Bentley et al. (2018). https://sites.google.com/view/kurthgee/data. This link contains}

updated manager non-GAAP EPS disclosure data collected using the method in Bentley et al. (Journal of Accounting Research, Forthcoming), with data from 2003 through part of 2016.

skill (Malmendier & Tate, 2005a). For this research the variables Holder67 and NetBuyer will be used based on available data. The third measure Longholder of Malmendier and Tate (2005a) will not be included in this research, because it is difficult to measure. This is because it could be the case that the option has never become in the money, which means it has never been profitable to exercise the option. In addition, given the fact that CEO of firm changes over period, it is difficult to measure when options have duration of ten years or longer.

4.3.2 Managerial overconfidence measures: Holder67

The first measure, Holder67, uses the timing of a CEOs firm stock option exercising behaviour to assess overconfidence. The measure classifies a CEO as overconfident if the average value of this exercisable unexercised options is more than 67% of the average. This implicates that because of heavy involvement the CEO indicates that he is confident about the future of the firm. Theory states that the stock option value of a CEO is founded based on their initial wealth, the degree of under diversification of personal portfolio and risk aversion (Hall & Murphy, 2002). Based on these factors, the CEO should exercise his options directly after the vesting period if the stock option reaches a certain percentage of in-the-money, which is in this case 67%. This percentage is based on the assumption that a two-third of a manager’s wealth is tied to the firm’s stock. Overconfident CEOs overestimate future cashflows and overestimate their ability, therefore they are willing to expose themselves more to risk by not exercising their options beyond the exercise date (Malmendier & Tate, 2005a). When a CEO is holding an option that is 67% in the money, it indicates that the stock price is more than 67% higher than the exercise price indicating it is with to exercise the option. More specifically, CEO incentives are maximized by not allowing them to trade their options for a specific time, so CEOs that choose to not to exercise their options early to decrease their exposure to idiosyncratic risk. So, when a CEO choose to hold in the money options and expose themselves to idiosyncratic despite the incentives are classified as overconfident. Theory explains this behaviour by suggesting the CEOs are confident about the future firm performance and expect the stock prices to increase.

The required data regarding detailed data about option grant specific exercise prices for the construction of the measure Holder67 is not publicly available. Therefore, other academics made adjustment to the proxies of Malmendier and Tate (2005) that makes it possible to do research with data that is publicly available (Hirshleifer et al., 2012). To

determine if CEOs hold options of at least 67% in the money, Campbell et al. (2009) used the method to calculate the average moneyness of options. Data about CEOs behaviour of in-the-money is available in Execucomp, which makes it possible to construct the overconfidence measure. Following prior research, this research calculates the equations to construct overconfidence measure Holder67 (Campbell et al, 2011). To measure overconfidence, the confidence variable is based on the average percentage of the moneyness option. This is calculated as the stock price at fiscal year-end3 divided by the estimated average exercise price minus one.

𝑀𝑜𝑛𝑒𝑦𝑛𝑒𝑠𝑠 𝑜𝑓 𝑜𝑝𝑡𝑖𝑜𝑛 = 𝑆𝑡𝑜𝑐𝑘 𝑝𝑟𝑖𝑐𝑒 𝑎𝑡 𝑓𝑖𝑠𝑐𝑎𝑙 𝑦𝑒𝑎𝑟 𝑒𝑛𝑑³

𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑒𝑥𝑐𝑒𝑟𝑐𝑖𝑠𝑒 𝑝𝑟𝑖𝑐𝑒 -1

equation 1. Moneyness of options

To give a clearer view, the equation is decomposed and presented below step by step. 𝑅𝑒𝑎𝑙𝑖𝑧𝑒𝑎𝑏𝑙𝑒 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑎𝑛 𝑒𝑥𝑐𝑒𝑟𝑐𝑖𝑠𝑎𝑏𝑙𝑒 𝑜𝑝𝑡𝑖𝑜𝑛

=𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑢𝑛𝑒𝑥𝑒𝑟𝑐𝑖𝑠𝑒𝑑 𝑒𝑥𝑒𝑟𝑐𝑖𝑠𝑎𝑏𝑙𝑒 𝑜𝑝𝑡𝑖𝑜𝑛𝑠

4 _($)

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑒𝑥𝑐𝑒𝑟𝑐𝑖𝑠𝑎𝑏𝑙𝑒 𝑜𝑝𝑡𝑖𝑜𝑛𝑠5 _(#) equation 2. Realizable value of an exercisable option

𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑒𝑥𝑐𝑒𝑟𝑐𝑖𝑠𝑒 𝑝𝑟𝑖𝑐𝑒

= 𝑌𝑒𝑎𝑟 𝑒𝑛𝑑 𝑠𝑡𝑜𝑐𝑘 𝑝𝑟𝑖𝑐𝑒³ − 𝑅𝑒𝑎𝑙𝑖𝑧𝑒𝑎𝑏𝑙𝑒 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑎𝑛 𝑒𝑥𝑐𝑒𝑟𝑐𝑖𝑠𝑎𝑏𝑙𝑒 𝑜𝑝𝑡𝑖𝑜𝑛

equation 3. Estimated average exercise price

The average moneyness of option needs to be calculated to determine whether the CEO holds options of at least 67% in the money. The components of this equation are available in the database ExecuComp under the following data item description.

Data item description Execucomp data item Unit Source Table

Stock price at fiscal year-end PRCC_F Actual Compustat

Estimated value of unexercised exercisable

options ($) OPT_UNEX_EXER_EST_VAL Thousands Execuomp

Number of exercisable unexercised options (#) OPT_UNEX_EXER_NUM Thousands Execucomp

Table 7. Data description in Execucomp of components calculation moneyness options components (Holder67)

3 _{Compustat item Stock price at fiscal year-end – Variable PRCC_F}

4 _{Execucomp item Estimated Value of In-the-Money Unexercised Exercisable Options -}

OPT_UNEX_EXER_EST_VAL

As previously stated, CEOs are classified as overconfident if the average moneyness held in options is at minimum 67%. According the original developed proxy by Malmendier and Tate (2005), a CEO was classified as overconfident if he/she held options which were 67% or 100% in the money at least two times. However, other research documents that it does not make a statistical difference whether holding an option in the money once or twice

(Hirshleifer et al., 2012). Therefore, in this research, a CEO is classified as overconfident when he holds the option which is 67% or 100% in the money at least once only (Hirshleifer et al., 2012; Kim et al., 2016).

4.3.3 Managerial overconfidence measure: NetBuyer

The second proxy, ‘NetBuyer’ is to measure and determine whether a CEO is classified as overconfident or not. This proxy indicates the tendency of CEOs who are willing to buy additional company stock, even when their personal wealth is already highly exposed to idiosyncratic risk (Malmendier & Tate, 2005a). NetBuyer is determined as follows, first the quantity of shares the CEO has purchased or sold must be determined. This is available in the database Execucomp under the data key item ‘SHROWN_EXCL_OPTS’ with the description ‘Shares Owned - Options Excluded’ in table ‘AnnComp’. These are all the shares that are owned by CEOs, except for the options that are excisable or will become exercisable within 60 days. After the number of shares has been determined, the number of years a CEO has purchased more shared than he has sold must to be calculated. A CEO is classified as overconfident in a year if she or he purchases more of own firm stock than she or he sells in a year. More precisely, this is determined by calculating the difference between the shares owned in the current year and the previous year6. When this increases, the CEO is a NetBuyer and classified as overconfident. When this amount decreases, the CEO is not classified as overconfident. Finally, the variable ‘NetBuyer’ is constructed. The dummy variable gives a value of one if the CEO purchase addition shares of their own company and gives a value of zero if otherwise. This additional dummy variable is included in the research to enhance the robustness of this research. According to hypothesis 1, it is expected that this dummy variable ‘NetBuyer’ is positively associated with the use of non-GAAP reporting.

4.3.4 Managerial overconfidence measures: High Overconfidence and Low Overconfidence

Following prior literature regarding measurement of overconfidence, a dummy variable Holder67 was constructed to measure overconfidence. Campbell et al (2011) construct two additional proxies for overconfidence, which are ‘high overconfidence’ and ‘low overconfidence’. The authors construct these proxies based on the personal portfolio option exercising behaviour of CEOs (Campbell et al., 2011).

The proxy ‘high overconfidence’ follows the same theory as the measure Holder67 as discussed in the previous section of this thesis and is also derived from this measure. According to prior literature, a CEO is classified as high overconfident if he holds vested options that exceed 100% in the money at least one time (Campbell et al., 2011). The calculation of the ‘moneyness of option’, ‘estimated average exercise price’ and ‘the realizable value per option’ for this overconfidence is similar as in the equation 1,2 and 3 as discussed and defined in the previous section. However, there is a difference between the measures high overconfidence and Holder67 based on the threshold. Whereas for the proxy Holder67 a CEO is considered as overconfident from the instant she fails to exercise his options while these options are 67% or more in the money. The dummy variable High Overconfidence (High_OC100) is classifies a CEO as highly overconfident if he or she holds options that exceed 100% in the money from the first moment. The dummy variable High Overconfidence takes as value one if highly overconfident and zero otherwise. According to hypothesis 2A high overconfidence is positively associated with non-GAAP reporting. Following hypothesis 2B it is expected that low managerial overconfidence (Low_OC30) is negatively associated

The second option based managerial overconfidence measures constructed by Campbell et al. (2013) is ‘low managerial overconfidence’. The extant literature classifies a CEO as having low overconfidence if he or she holds options that do not exceed 30% or lower in the money. The calculation relies on the same method as measures Holder67 and High Overconfidence. Thus, the dummy variable Low_OC30 takes a value of one if the CEO is classified as low overconfident and if zero otherwise. Following hypothesis low overconfidence is expected to be less positively associated with non-GAAP reporting compared to high overconfidence. These additional dummy variables are included in the research to enhance the robustness of this research and examine whether the degree of managerial overconfidence is significant difference.

4.4 Control variables

For this study a number of control variables are included in the regressions, from which it is expected to be correlated with both non-GAAP reporting and CEO overconfidence. The control variables used in this research are firm size, financial leverage, growth of firms, earnings volatility, loss indicator and special items.

4.4.1 Firm size

The first control variable that is used for this thesis is firm size. Prior research finds evidence that larger companies provide more public information, which results to less information asymmetry (LaFond & Watts, 2008). In addition, theory suggests that larger companies are less likely to miss their forecast (Hribar & Yang, 2016). Francis et al. (1994) find that the costs of opportunistic behaviour can increase with firm size, because shareholders are more likely to sue larger companies. Firm size can be measured as total assets7_{, which is publicly available in}

Compustat following the method of LaFond and Watts (2008). For this research, the natural logarithm of total assets will be used, named LN_FirmSize. See appendix 7C for an overview of the log transformation.

4.4.2 Financial leverage

The second control variable which could influence non-GAAP reporting and managerial overconfidence association is the financial leverage. This variable measures to what extent the company is financed by debt and gives a view of the capital structure of a firm. Prior literature documents that when companies exceed a certain amount of their leverage, financial statements users and specifically investors perceive the positive earnings to be less informative. This is because of the higher likelihood of earnings manager and higher probability of firm failure. (Hodgson & Clarke, 2000). Therefore, firms with high leverage may benefit from non-GAAP reporting that result in highlighting effects of excluding nonrecurring items to depict higher earnings (Lougee & Marquardt, 2002). Thus, it is expected that companies with a high leverage are more likely to exclude non-recurring items and provide non-GAAP earnings than other companies (Lougee & Marquardt, 2004). In addition, literature documents evidence that financial leverage is also associated with managerial overconfidence. Because an overconfident manager is more likely to overestimate future cashflows and in making riskier investments