Characteristics of firms disclosing non-GAAP earnings

(1)

Characteristics of Firms Disclosing Non-GAAP Earnings

MSc Accountancy and Control Specialization Accountancy

Faculty of Economics and Business University of Amsterdam

Name: Myrthe E. Roet Student no.: 10373640 Word count: 15,290 Date: June 25, 2018 Supervisor: Prof. dr. D. Veenman

(2)

Statement of Originality

This document is written by student Myrthe Roet 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)

1

Abstract

This study examines the characteristics of firms reporting non-GAAP earnings in the U.S. during 2003 and 2016. Non-GAAP reporting is frequently used, despite its ambiguous use. On the one hand, firms can report non-GAAP earnings to provide investors with additional value relevant information. In doings so, firms want to show information that is more useful and informative than GAAP earnings. On the other hand, non-GAAP earnings can be used opportunistically to mislead investors, i.e. to meet or beat analysts’ expectations. While the usefulness of non-GAAP earnings is an extensively debated topic, similar to the determinants of voluntary disclosure, the characteristics of firms reporting non-GAAP earnings are not extensively examined. Therefore, this study uses a large dataset of Bentley et al. (2018) and analyses five possible determinants of disclosing non-GAAP earnings. The firm characteristics expected to show a relation with disclosing non-GAAP earnings are whether the company operates in a high-tech industry, the firm size, number of analysts following the firm, the intangible-intensity and whether the company reports a GAAP loss. By conducting a logit regression and analysing marginal effects, this thesis finds that all five firm characteristics have a positive and significant effect on the likelihood of non-GAAP reporting. To conclude, the results show that firms that operate in high technology industries, have a larger firm size, have higher analyst coverage, report a higher intangible-intensity ratio and report a GAAP loss, are more likely to disclose non-GAAP earnings than firm that do not meet these characteristics.

(4)

2

1 Introduction

The purpose of this thesis is to examine the relation between several firm characteristics and the likelihood of firms reporting non-GAAP earnings. The Financial Accounting Standards Board (FASB, 2001) states that voluntary disclosures contain information beyond the disclosed information required by Generally Accepted Accounting Principles (GAAP). Several studies suggest that firms are more likely to provide voluntary disclosures when current information is uninformative (Chen, DeFond, & Park, 2002; Lougee & Marquardt, 2004). One financial measure that firms disclose voluntarily is non-GAAP earnings. Regulation G of the Securities and Exchange Commission (SEC) provides a definition of non-GAAP financial measures (SEC, 2001). First, Regulation G states that non-GAAP earnings represent a measure of a firm’s financial performance, financial position or cash flow, either historical or future related. Subsequently, firms exclude amounts that are included in the most directly comparable measure that is calculated in accordance with GAAP, or firms include amounts that are excluded from GAAP measures. These measures, including a reconciliation from the GAAP measure to the non-GAAP measure are presented in the financial statements of the firm. By presenting Regulation G, the SEC wants to regulate the disclosure of non-GAAP financial measures, leading to more transparent disclosures.

However, for years a widespread debate exists regarding the purpose and usefulness of non-GAAP earnings. On the one hand, several prior studies suggest that managers disclose non-GAAP earnings to provide investors with additional value relevant information. In addition, prior literature argues that firms disclose non-GAAP earnings to reduce information asymmetry (e.g. Bhattacharya, Black, Christensen, & Larson, 2003; Entwistle, Feltham, & Mbagwu, 2010; Lougee & Marquardt, 2004). GAAP reporting requires items to be included that are transitory or nonrecurring. However, managers have the discretion to separate GAAP earnings into non-GAAP earnings and exclusions, where the exclusions consists of unusual and nonrecurring items. In this way, the GAAP earnings number contains more permanent items (e.g. Bhattacharya et al., 2003; Entwistle et al., 2010). Lougee and Marquardt (2004) argue that the separation of permanent items and exclusions results in less noise in earnings. On the other hand, critics provide evidence that suggests that managers report non-GAAP earnings opportunistically to mislead investors by showing a more favourable picture of the performance of a firm to meet or beat analysts’ expectations (e.g. Black, Black, Christensen, & Heninger, 2012; Doyle, Jennings, & Soliman, 2013; Entwistle et al., 2010). Despite the ongoing debate about non-GAAP earnings, evidence shows that non-GAAP financial measures are still frequently used (Bhattacharya et al., 2003; Zhang & Zheng, 2011). Further, Black et al. (2012) show that firms increasingly report non-GAAP earnings after the two-year dip caused by the Sarbanes-Oxley act and SEC’s Regulation G in 2002 and 2003, respectively.

(6)

4

In addition to the usefulness of non-GAAP earnings, are the determinants of voluntary disclosures in general another frequently debated topic. The industry a firm operates in has a positive relation with the likelihood to voluntarily disclose information. That is, high-tech firms make large investments in intangibles and this would distort the earnings of a firm, making the earnings uninformative (e.g. Broberg, Tagesson, & Collin, 2010; Chen et al., 2002). Second, larger firms tend to disclose more information than smaller firms, because larger firms have more resources and can better afford the costs associated with providing additional information (e.g. Broberg et al., 2010). By providing extra information, firms want to meet the demand of financial statement users. Third, prior studies argue that profitable firms are more likely to disclose voluntary information. This is because profitable firms want to distinguish themselves from firms that are not profitable (e.g. Zhang, 2015). Finally, prior literature suggests that analyst coverage has a positive effect on the likelihood of firms to disclose voluntary information (e.g. Chen et al., 2002; Shehata, 2014).

While the usefulness of non-GAAP earnings is an extensively debated topic, similar to the determinants of voluntary disclosure, the determinants of firms to report non-GAAP earnings are not extensively examined. Based on the foregoing debate, this research attempts to answer the following research question:

What are the characteristics of companies reporting non-GAAP earnings?

The five hypotheses of this study examine the determinants of a firm’s decision to report non-GAAP earnings. Based on prior literature, this study hypothesizes that firms that (1) operate in high technology industries, (2) have a greater firm size, (3) have more analysts following the firm, (4) have a higher intangible-intensity and (5) report a GAAP loss, are more likely to report non-GAAP earnings than firms that do not meet these characteristics.

This study uses a dataset of Bentley, Christensen, Gee and Whipple (2018) where for a large sample is indicated whether firms report non-GAAP earning or not. The sample period is 2003-2016. The sample is matched to datasets from Compustat and I/B/E/S, containing information regarding several firm characteristics. Then, this study conducts a logit regression and examines the marginal effects to see whether operating in a high-tech industry, firm size, analyst coverage, intangible-intensity and GAAP loss are significant characteristics of firms reporting non-GAAP earnings.

The results are consistent with the predictions. All firm characteristics this study examines have a positive and significant effect on a firm’s decision to disclose non-GAAP earnings. In other words, firms that operate in high-tech industries and have a larger firm size are more likely to report non-GAAP

(7)

5

earnings. Further, firms that have higher analyst coverage or a higher intangible-intensity are also more likely to disclose non-GAAP earnings. Finally, firms reporting a GAAP loss have an increased likelihood of disclosing non-GAAP earnings. All the results are significant.

After conducting the logit regression, this study examines the marginal effects. Marginal effects show the percentage point change in the likelihood of reporting non-GAAP earnings, when the independent variable increases by one unit. The results of this analysis show that if a firm changes its operations from low-tech industries or even non-tech industries, to operations in high technology industries, then the likelihood of reporting non-GAAP earnings increases by 13.4 percentage point. When controlling for all variables in this study, the likelihood of disclosing non-GAAP measures increases by 15.4 percent point. Also, an increase in the size, intangible-intensity and GAAP loss of a firm by one unit, increases the likelihood to report non-GAAP earnings by 6.1, 38.0 and 7.9 percentage point, respectively. These results are significant. Finally, when the number of analysts that follow a firm increases by one unit, i.e. by one person, then the likelihood of disclosing non-GAAP earnings in the financial statement increases by 0.3 percentage point. Though small, this result is significant. The results suggest that firms with the abovementioned characteristics find their GAAP earnings uninformative and therefore report non-GAAP earnings.

By answering the research question, this thesis contributes in the first place to the growing amount of literature of non-GAAP earnings as a voluntary disclosure measure. By examining certain firm characteristics to provide an answer as to why firms report non-GAAP earnings, existing literature concerning the usefulness of voluntary disclosures is extended. More specific, this study shows that the likelihood of non-GAAP reporting depends on several firm characteristics. By examining the characteristics of firms reporting non-GAAP earnings, this study provides insights into the determinants of non-GAAP reporting. Besides, evidence is added to Lougee and Marquardt (2004) and more recently Black, Black, Christensen and Gee (2018) who examine several characteristics of firms disclosing non-GAAP earnings. Second, this study contributes to the existing literature as it uses a recent publicly available dataset from Bentley et al. (2018). Researchers were previously forced to hand-collect non-GAAP earnings numbers or use proxies using earnings numbers available through analyst forecast data providers. However, as was the case with previously used datasets, the new dataset does not underestimate managers’ aggressive reporting decisions.

The examination of several characteristics of firms reporting non-GAAP earnings has several implications for both the market and policy makers. For instance, results show that firms that operate in high-tech industries are more likely to report non-GAAP earnings. This suggests that these firms find the

(8)

6

large investments they make in intangibles, e.g. R&D, more relevant and informative and therefore include this amount in the non-GAAP earnings number. Investors and other financial statements users can take the exclusion of large investments from the GAAP earnings number into account when assessing the firm’s performance. Investors can then decide whether to invest in high-tech firms, based on the increased likelihood these firms report non-GAAP earnings. The implications can also be interpreted the other way around. When investors know the firm reports non-GAAP earnings, they can take into account the characteristics the firm might have. These results can be applied for all firm characteristics examined. Overall, this suggest that the market has a better understanding of the firm and that investors can make deliberate decisions.

The second implication is as follows. Policy makers do not want a firm to use non-GAAP earnings opportunistically and mislead investors. However, firms operating in high technology industries make large investments in R&D, which are excluded from GAAP earnings. High-tech firms find the exclusions relevant and informative and hence, include these items in the non-GAAP earnings number. Though, firms still use the exclusion of items to meet or beat earnings benchmark and to mislead investors (e.g. Black et al. 2012). This interpretation applies to all firm characteristic this study examines. Since policy makers know that firms with certain characteristics exclude items, they may need to relax current regulation G and Compliance and Disclosure Interpretations on non-GAAP measures of the SEC. For instance, regulation on the exclusion on recurring items might need relaxation, in order to allow firms to show a more informative GAAP earnings measure.

The structure of the thesis is as follows. Chapter 2 presents the literature review and discusses voluntary disclosures in general, as well as background information on non-GAAP earnings, the use and usefulness of this earnings measure and the determinants of voluntary disclosures. Chapter three explains the relationship between non-GAAP reporting and several firm characteristics, which results in the hypotheses. Chapter four describes the sample this study uses, as well as the logit regression model and descriptive statistics. The fifth chapter discusses the regression results and the marginal effects. Finally, chapter six contains the conclusion, where also an answer to the research question is given.

(9)

7

2 Literature review

The first section of chapter two explains the term voluntary disclosure and the section thereafter discusses the information problem, the agency problem and ways how voluntary disclosure can resolve these problems. The next section discusses motives to use voluntary disclosures, as well as non-GAAP earnings as a form of voluntary disclosure. Third, the use and usefulness of non-GAAP earnings are discussed. Finally, the last section discusses several determinants of voluntary disclosure.

2.1 Voluntary disclosure

2.1.1 Definition of voluntary disclosure

Prior literature divides corporate information into two categories: mandatory and voluntary disclosure (Marston & Shrives, 1991). Required information needs to be in line with the applicable laws and regulations and is publicly available in the financial reports. These laws and regulations are set by regulatory bodies, such as the FASB, the SEC and domestic regulatory authorities. On the other hand, voluntarily disclosed information is information that firms report in excess of the required, mandatory information. Companies add this information to the reports if the perceived benefits of disclosure outweigh the costs.

The FASB (2001) provides a description of voluntary disclosure in its Business Reporting Research Project. It states that voluntary disclosures are beyond the disclosures required by GAAP or rules from the SEC. One can find voluntary disclosures primarily outside the financial statements. By providing a definition, the FASB wants to improve the quality of disclosures, leading to a more efficient allocation of resources. Dye and Sridhar (2002) provide a similar description and show evidence that the capital market reacts to disclosures, which in turn leads to managers allocating resources more efficiently.

2.1.2 Information and agency problem

Healy and Palepu (2001) discuss the role of both financial reporting and voluntary disclosure in the capital market. In this market, information and agency problems disrupt the efficient allocation of resources. The information problem or lemons problem arises as a result of information asymmetry between market participants. In other words, such problems arise when one party is in the possession of more or better knowledge than the other party. This is in line with the definition that Healy and Palepu (2001) give, as they argue that such problems arise between entrepreneurs and savers due to differences in information and conflicting incentives. Healy and Palepu (2001) provide several solutions to resolve this problem. One

(10)

8

way is by fully disclosing information. By doing so, all investors are better informed and thus, have the same knowledge that managers already had. This reduces the information asymmetry problem between two parties. Hence, investors can more accurately predict the value of the company and thus, managers cannot benefit by valuating the company differently.

The second problem that causes an inefficient allocation of resources is the agency problem. This problem arises because of misalignment of interest between the principal and the agent (Jensen & Meckling, 1976).Healy and Palepu (2001) refer to the two parties as on the one hand, the savers and on the other hand, the entrepreneur. An explanation for this situation is that although savers provide funds to a company, they are usually not actively involved in the operations of that business. The entrepreneur carries out these management tasks. Hence, the entrepreneur dispossesses the funds provided by the savers by acting in a self-interested manner. This may results in making decisions that harms the savers. One way of solving this problem is similar to solving the information asymmetry problem. The entrepreneur needs to disclose relevant information which enables investors to monitor and evaluate the actions of the entrepreneur and determine whether the actions made are in line with the interest of the investors.

2.1.3 Motivations for voluntary disclosure

This section briefly discusses six motivations for voluntary disclosure, provided by Healy and Palepu (2001). The authors argue that one motive affecting manager’s disclosing decision, is the capital markets transactions hypothesis. Of interest is the perception an investor has of the firm when the firm is in need of funding. By providing voluntary disclosure, which increase the information available to investors, the information asymmetry problem is reduced and this in turn leads to a reduction of the costs of external financing.

The second motive is the corporate control contest hypothesis, which states that managers are held accountable for a firm’s stock performance by investors and boards of directors. Voluntary disclosure can help explain potential poor performance of a firm. This reduces undervaluation as well as manager’s risk of losing their job due to bad performance.

Third, compensation plans based on stocks motivates firms to use voluntary disclosures. This is in the first place to meet restrictions set by insider trading rules. In addition, this leads to managers being able to correct any misvaluation of stocks prior to the date they expire. Further, stock-based compensation reduces the contracting costs associated with new employees.

(11)

9

Fourth, the risk of litigation has two effects on manager’s voluntary disclosing decision. On the one hand, litigation risk motivates managers to voluntary disclose information when the firm reports poor performance and faces litigation risks due to untimely and inadequate disclosures. Voluntary disclosures reduce litigation costs. On the other hand, potential penalization reduces managers’ motive to disclose voluntary information and in particular forward-looking information. This occurs because the legal system has trouble deciding whether errors in disclosures are made deliberately or unintentionally.

The fifth motivation describes the management talent signaling hypothesis. This hypothesis suggests that managers want to show their talent to investors by voluntarily disclosing earnings forecasts that are more accurate and by reporting the disclosures before other firms do. By doing this, investors will perceive the managers as more favorable because managers make quick response to changes. This positively affects the market value of the firm.

Finally, Healy and Palepu (2001) argue that prior studies suggest that the proprietary cost motive leads to firms not wanting to disclose additional disaggregated information, because this potentially attracts more competitors. Hence, this results in less competitive advantage. However, the authors also argue that firms operating in similar business segments are more likely to disclose disaggregated information (Healy & Palepu, 2001). In other words, operating in similar business segments motivates firms to disclose voluntary information.

2.2 Non-GAAP earnings

This section first discusses non-GAAP earnings as a form of voluntary disclosure by proving background information. Further, it explains the informative use and the opportunistic misuse that are associated with the disclosure of non-GAAP earnings.

2.2.1 Background information non-GAAP earnings

Companies are required to prepare their financial statements in according with the principles set out in GAAP. However, an increasing amount of companies disclose additional information in the form of non-GAAP earnings. While firms use this voluntary disclosure frequently, there is no uniform definition available. Nonetheless, prior literature states that non-GAAP earnings can be seen as GAAP earnings where a manager excludes items that are perceived as unusual and nonrecurring (Bhattacharya, Black, Christensen, & Mergenthaler, 2007; Gu & Chen, 2004). SEC’s Regulation G discusses that a non-GAAP earnings measure relates in the first place to a firm’s financial performance, financial position or cash flow. Second, the earnings measure excludes items under non-GAAP earnings, which are included under

(12)

10

GAAP earnings. Besides, the GAAP earnings number is the most directly comparable financial measure. The non-GAAP figure is subsequently reported in the financial statements (SEC, 2001). In addition to using the term non-GAAP earnings, other measures are also used to refer to the adjusted earnings. Two alternatives that other studies commonly use are pro forma earnings (e.g. Bhattacharya, Black, Christensen, & Larson, 2003; Lougee & Marquardt, 2004; Zhang & Zheng, 2011) and street earnings (e.g. Bradshaw & Sloan, 2002; Gu & Chen, 2004). The abovementioned terms refer to approximately the same concept. However, this study adopts the term non-GAAP earnings.

To show the differences that exists between ways to show non-GAAP earnings, Appendix A shows an example of the reconciliations two firms make to arrive at the non-GAAP earnings number. More specific, one firm that discloses non-GAAP earnings operates in a high technology industry and the other firm that reports non-GAAP earnings operates in a non-tech industry. By doing so, the example shows more clearly that differences that exists between reconciliations, depending on whether the firm operates in a high-tech industry or not. High-tech firms tend to disclose non-GAAP earnings, because they want to show the large investments in R&D they make. That is because these firms perceive the R&D expenses as relevant and informative (Lougee & Marquardt, 2004). Table A-1 Panel A in Appendix A shows that the high-tech firm makes substantial changes to arrive at the non-GAAP earnings number. One significant item is the purchase accounting adjustments. This item mainly consists of amortization and depreciation expenses and in particular amortization of intangible assets. Amortization and depreciation expenses represent 54.9 percent of the total adjustments made. By eliminating this items from the GAAP earnings number, the company wants to better show the underlying economics. More specific, the firm wants to provide a degree of consistency to intangibles that are internally generated with expensed R&D costs.

Panel of B of Table A-1 shows the reconciliations that a tech firm makes to arrive at the non-GAAP earnings number. This firm uses the earnings before interest, taxes, depreciation and amortization (EBITDA) as the non-GAAP financial measure. While the firm excludes a substantial amount to arrive at EBITDA, the only item it excludes is depreciation and amortization expenses. From the total adjustments, the amortization of intangibles expense represents 14.0 percent. The reconciliations of Panel A and Panel B show that both high-tech and non-tech firms make significant adjustments when computing non-GAAP financial measures. However, the high-tech firm shows that the adjustments mainly consist of amortization of intangible assets. This suggests that a high-tech firm find these costs relevant and informative and therefore adds the items to the non-GAAP earnings number.

Though the disclosure of non-GAAP earnings is legal, there are some requirements a company must comply with when disclosing this earnings number. These rules arose after the SEC (2001) warned

(13)

11

public companies that non-GAAP earnings could mislead investors by obscuring GAAP earnings. Brown, Christensen, Elliott and Mergenthaler (2012) and Zhang and Zheng (2011) state this is possible as managers have significant discretion to exclude GAAP items in computing the non-GAAP earnings number. Subsequently, the SEC issued in 2003 Regulation G to regulate the disclosure of non-GAAP earnings. This regulation leads to more transparent disclosures as companies disclosing non-GAAP are required to include the most directly comparable GAAP number as well as a reconciliation of the non-GAAP earnings number to the GAAP earnings number. Despite this regulation and monitoring by the SEC, managers still have discretion to show a more favorable image of the earnings. That is because non-GAAP earnings are not subjected to an audit (Black, Christensen, Taylor Joo, & Schmardebeck, 2017).

2.2.2 Use and usefulness of non-GAAP reporting

The FASB (2008) states that their primary objective is to focus on the use of accounting information in decision making. In order to evaluate the use, the usefulness of information must be determined. This usefulness of accounting information also explains the use of non-GAAP earnings. When reporting this earnings number, prior studies suggest there are two broad motives. On one hand, proponents see non-GAAP earnings as informative and on the other hand, opponents see them as opportunistic. The following two subsections describe these two motives.

2.2.2.1 Informative use of non-GAAP reporting

Proponents of non-GAAP earnings argue that the information that is provided in this measure, is more value relevant and thus more useful than information provided in the GAAP earnings number (e.g. Bhattacharya et al., 2003; Brown & Sivakumar, 2003; Entwistle et al., 2010). Several prior studies support the informative use of non-GAAP measures and argue that nonrecurring and unusual items are included in GAAP earnings, resulting in uninformative earnings (e.g. Bhattacharya, Black, Christensen, & Mergenthaler, 2004; Bradshaw & Sloan, 2002; Brown & Sivakumar, 2003; Entwistle et al., 2010; Lougee & Marquardt, 2004). Managers are required to prepare GAAP earnings that include nonrecurring and unusual items, such as gains and losses on asset sales. However, these items make comparison between two GAAP earnings numbers more difficult than comparison between two GAAP numbers. Eliminating the two distinct items, reduces noise in the earnings measure and provides additional value relevant information to investors and other users of the financial statement. In the end, this leads to non-GAAP earnings that are more informative and more useful than traditional GAAP earnings.

(14)

12

Bhattacharya et al. (2003) and Bhattacharya et al. (2004) support the abovementioned notice and argue that an ongoing discussion exists regarding the usefulness of non-GAAP earnings. Proponents state that non-GAAP reporting is a more accurate way of determining expectations of future cash flows and the value of a firm than GAAP reporting. Both Bhattacharya et al. (2003) and Bhattacharya et al. (2004) examine in their articles whether non-GAAP earnings are relatively more informative and permanent than GAAP earnings and further examine characteristics of firms voluntarily reporting non-GAAP measures. They also look into the type of adjustments firms make to arrive at the non-GAAP earnings number. They find that almost all firms of their sample operate in the service industry or high technology industry. Second, they find three types of adjustment that firms most commonly make when reporting non-GAAP earnings. Those adjustments are depreciation and amortization, stock-based compensation costs and changes in the number of used shares when computing the earnings per share. Firm make this last adjustment to decrease the loss of GAAP earnings per share. Isidro and Marques (2015) add to this results that firms also frequently adjust tax expenses. However, literature argues that investors perceive non-GAAP earnings to be more value relevant than non-GAAP earnings if the adjustments are not routine expenses, such as depreciation and amortization, but rather one-time exclusions (e.g. Bhattacharya et al., 2003).

Further, research shows non-GAAP earnings are in most cases higher than GAAP earnings and in general, this higher non-GAAP earnings number is reported before the GAAP earnings number (Bhattacharya et al., 2003). However, in supporting the informative use of non-GAAP, Bhattacharya et al. (2003) find that firms in a quarter of cases report the non-GAAP earnings number first, even though the GAAP earnings are higher. Firms do this because they believe non-GAAP figures are more value relevant than the higher GAAP figures. Overall, Bhattacharya et al. (2003) conclude that investors perceive non-GAAP earnings as more informative than non-GAAP earnings and analysts perceive them as more permanent.

Next, prior studies discuss the comparability of non-GAAP earnings. Critics argue that no single definition of non-GAAP earnings exists, nor is the same definition used in a single firm from one year to another (Bhattacharya et al., 2004). Besides, Bhattacharya et al. (2004) find in their study that, even though adjustments are split up into several categories, firms still classify the adjustments they make as other adjustments. The abovementioned points above do not enhance the comparability of non-GAAP earnings. On the other hand, Venter, Emanuel and Cahan (2014) examine what happens with the comparability of non-GAAP earnings when firms are obliged to report the earnings measure. They find that non-GAAP earnings are comparable over time, since a uniform definition of non-GAAP earnings is specified and all firms have to apply that definition. In addition, firms are required to use this definition over time, enhancing the comparability between firms and of one firm from year to year. Finally, Venter

(15)

13

et al. (2014) find that mandatory non-GAAP reporting is likely to be subjected to an audit, which increases comparability. However, Black et al. (2017) argue that voluntary non-GAAP reporting is not audited and hence, does not increase comparability.

Third, prior literature discusses the relevance of non-GAAP earnings. Brown and Sivakumar (2003) compare the value relevance of non-GAAP earnings to the value relevance of GAAP earnings. They determine the relevance of earnings by applying three methods. The first method is the extent to which earnings can be predicted. A second way to assess the relevance is by establishing the connection between earnings and stock prices. Third, the authors determine the interdependence between earnings surprises and abnormal stock returns. Under all techniques, research shows that the initially reported GAAP number is of less relevance than the additionally reported non-GAAP number. The authors explain this result by first referring to the transitory items. By excluding some of those items, more permanent items remain in non-GAAP figures and hence, this earnings number is more value relevant. Venter et al. (2014) support this result. A second explanation of the results is that managers try to provide relevant information to users of the financial statements by reporting non-GAAP earnings (Brown & Sivakumar, 2003). This suggest that firms report non-GAAP earnings as a motive to provide useful and informative information. Entwistle et al. (2010) build on the above discussed literature about the informative use of non-GAAP earnings, since they also analyze press releases of firms reporting non-non-GAAP earnings. The authors examine the value relevance of GAAP, non-GAAP and I/B/E/S earnings and state that I/B/E/S earnings represent GAAP earnings that are adjusted and subsequently disclosed by analyst forecasting tracking services. Furthermore, Entwistle et al. (2010) evaluate the relevance difference between the three earnings measures, as well as the effect of the Sarbanes-Oxley (SOx) act and Regulation G on the relevance of these three measures. The SOx act partially focuses on mitigating managers’ discretion to report non-GAAP opportunistically. Therefore, the authors argue that if SOx limits the ability of reporting non-non-GAAP earnings for strategic reasons, the relevance of this earnings number will be lower after implementation of the act. In providing support for this, Entwistle et al. (2010) examine a sample of press releases between 2000 and 2004, especially to see whether there is a change in relevance because SOx is implemented in 2002. However, the study shows hardly any evidence that SOx changes the relevance of GAAP or non-GAAP earnings. Though, research shows an increase in the relevance of I/B/E/S earnings after the implementation of SOx. The other findings are similar to the ones in the above discussed literature (Brown & Sivakumar, 2003) and suggest that non-GAAP earnings are higher than either GAAP or I/B/E/S earnings and that they are more value relevant during the entire period examined. Moreover, investors use non-GAAP earnings to predict future earnings more accurately than if the other two earnings metrics were

(16)

14

used (Entwistle et al., 2010). These findings support the view that managers exclude nonrecurring or unusual items to provide more information to the financial statement users.

Finally, Leung and Veenman (2018) also provide strong evidence that firms report non-GAAP earnings to show more informative numbers than when firms report GAAP earnings only. In particular, they show evidence that for firms that report both GAAP and non-GAAP earnings, the GAAP earnings number is half as informative than when GAAP earnings only would have been reported. The authors also show that firms that convert a GAAP loss into a non-GAAP profit have strong uninformative GAAP earnings (Leung & Veenman, 2018). The results suggest that firms use non-GAAP earnings to provide informative information to investors.

2.2.2.2 Opportunistic use of non-GAAP reporting

On the other hand, critics argue that non-GAAP earnings give managers a significant amount of discretion when excluding items from the GAAP earnings number. Graham, Harvey and Rajgopal (2005) examine whether managers consider earnings benchmarks to be important and what motivates managers to involve in earnings management. The results show that firms believe that outsiders see earnings as the most important earnings measure. Therefore, meeting or beating earnings benchmarks are important to managers. Especially, Graham et al. (2005) find that over 85 percent of the interviewed managers think they are perceived as credible, when they meet earnings benchmarks. Hence, managers are inclined to manage earnings by giving up actual economic value instead of manipulating accounting numbers, in order to meet or exceed those benchmarks. However, regulators prohibit the use of earnings management. Consequently, managers have found an alternative way to meet certain earnings numbers, i.e. by disclosing non-GAAP earnings (e.g. Bhattacharya et al., 2003; Bhattacharya et al., 2004). One of the findings Bhattacharya et al. (2004) show is that a reason to voluntarily disclose non-GAAP earnings, is to meet certain earnings benchmarks. They further show in their research that 80 percent of the press releases reporting non-GAAP earnings meet or exceed earnings benchmarks, while under half meet or beat those benchmarks when reporting GAAP earnings only. This suggests that managers considerably change GAAP earnings not meeting the benchmark, into non-GAAP earnings that do meet or beat the benchmarks. In other words, managers use non-GAAP earnings in an opportunistic way. This is why critics argue that investors should focus on GAAP earnings, instead of non-GAAP earnings disclosed in press releases (Bhattacharya et al., 2003).

One of the things Black et al. (2012) examine, is if SOx affects whether firms report non-GAAP earnings aggressively or not. They argue that managers have incentives to act opportunistically. Results

(17)

15

show that implementation of SOx leads to a small reduction in the number of firms that use non-GAAP earnings to meet or beat earnings benchmarks, while this benchmark would otherwise not be met. However, after SOx, still a third of firms use non-GAAP earnings to meet or exceed analysts’ expectations. Additional result show that managers are incentivized to exclude recurring items, such as depreciation, amortization and other usual operating expenses from the GAAP measure, in addition to nonrecurring and unusual items. Isidro and Marques (2015) and Whipple (2015) support this result. They argue that the main adjustments firms make, are recurring items. By excluding these items, firms have the opportunity to beat earning benchmarks. SOx has not led to a noteworthy change in the number of firms that exclude recurring items. Still three-quarters of firms exclude items after the implementation of SOx (Black et al., 2012).

Black, Christensen, Kiosse and Steffen (2017) extend the research of Black et al. (2012) by taking into account SEC’s Compliance and Disclosure Interpretations on non-GAAP measures, in addition to SOx and Regulation G. These interpretations are presented in 2010 and updated on April 2018 and emphasize non-GAAP earnings by taking into account the opportunistic use of it (SEC, 2010). Black et al. (2017) find in the first place that firms reporting non-GAAP earnings are less likely to exclude recurring items, such as depreciation and amortizations expenses. However, firms are still likely to exclude interest and stock-based compensation. This result is in line with Black et al. (2012) and Isidro and Marques (2015). Second, Black et al. (2017) find that firms are less likely to use non-GAAP measures to meet or beat analysts’ expectations. However, results show that some firms still use non-GAAP earnings opportunistically to mislead investors.

Also, Hsu and Kross (2011) demonstrate that managers act in an opportunistic manner. They find evidence that the exclusion of items when computing non-GAAP earnings is associated with three features. The first feature is whether the exclusion of items increases the non-GAAP measure. Second, the use of non-GAAP reporting must smoothen the earnings and last, it enables managers to meet or beat the earnings benchmark. By performing the abovementioned actions, managers fail to improve the informativeness of earnings, but behave opportunistically instead (Hsu & Kross, 2011).

In line with the previously discussed literature, Doyle et al. (2013) also examine the use of non-GAAP figures. In particular, they examine whether managers opportunistically report non-non-GAAP earnings. They provide evidence for this opportunistic behavior, since managers are more likely to meet or beat earnings benchmarks when they report non-GAAP earnings (Doyle et al., 2013). They further provide two ways by which managers change the non-GAAP earnings number to meet or beat analyst expectations. One way is by opportunistically increasing the amount of expenses excluded, which results in meeting or

(18)

16

beating the earnings benchmark. The second way is by artificially creating a different exclusion. Overall, results suggest that firms use non-GAAP earnings in an opportunistic way.

Despite the downside of non-GAAP earnings that prior literature discusses, research shows that investors pay attention to the adjusted earnings number. Both Bradshaw and Sloan (2002) and Brown and Sivakumar (2003) argue that investors put more emphasis on non-GAAP earnings than GAAP earnings when making decisions. Bhattacharya et al. (2003) support this notice and add that investors pay significantly more attention to non-GAAP earnings as they perceive this number to be more informative than GAAP earnings. Entwistle et al. (2010) argue that financial statement users pay more attention to non-GAAP earnings, because this number is pointed out in the press releases. As a result, the users give little attention to the GAAP earnings number. Further, Lougee and Marquardt (2004) provide evidence that non-GAAP earnings have incremental information over traditional GAAP earnings. At last, contrary to what prior literature discusses, Entwistle et al. (2010) find that the purpose of managers to disclose non-GAAP earnings is to provide informative earnings rather than to mislead investors.

Overall, the usefulness of GAAP is twofold. On the one hand, several studies argue that non-GAAP earnings are more permanent and informative than non-GAAP earnings (e.g. Bhattacharya et al., 2004; Entwistle et al., 2010) and that they are more comparable when they should be mandatory disclosed (Venter et al., 2014). In addition, prior studies argue that non-GAAP earnings are more relevant (Brown & Sivakumar, 2003; Entwistle et al., 2010). Finally, prior research shows that loss firms that convert a GAAP loss into a non-GAAP profit have strong uninformative GAAP earnings (Leung & Veenman, 2018). In other words, loss firms disclose non-GAAP earnings to provide more informative financial measures. On the other hand, other studies suggest that firms use non-GAAP earnings opportunistically to mislead investors about the firm’s performance. More specific, firms use non-GAAP earnings to meet or beat earnings benchmarks they would not meet under GAAP reporting (e.g. Bhattacharya et al., 2004; Black et al., 2017; Doyle et al., 2013; Hsu & Kross, 2011). In addition, research shows that the main adjustments firms make are recurring items, such as depreciation and amortization expenses (Black et al., 2012; Isidro & Marques, 2015; Whipple, 2015). The example in Table A-1 shows similar results.

2.3 Firm characteristics and voluntary disclosure

Non-GAAP reporting is an increasingly used earnings measure in financial statements of companies (e.g. Zhang & Zheng, 2011). However, determinants of non-GAAP reporting are not extensively examined. On the other hand, prior studies examine the determinants of voluntary information disclosure in general. This section discusses these determinants.

(19)

17

The first determinant of voluntary disclosure that prior studies frequently examine is the industry a firm operates in. Broberg et al. (2010) argue that industry is strongly associated with voluntary information disclosure. In addition, firms with high R&D expenses are more cautious in what information to disclose to competitors than firms that do not have these significant R&D expenses (Meek, Roberts, & Gray, 1995). Further, the relevance of R&D expenses differs from one firm to another, depending on the industry the firm operates in. Therefore, firms want to disclose extra information, in order to show a more relevant earnings measure. Meek et al. (1995) and Lougee and Marquardt (2004) argue that firms operating in high technology industries have significant R&D expenses, which distort earnings that are required by regulations. Hence, these firms provide extra information to show information that is more relevant. Chen et al. (2002) show a similar result and argue that high-tech firms have high R&D expenses and make large investments in other intangibles. In addition, high technology firms operate in an environment that rapidly changes, resulting in uncertain future operations and hence, uncertain future earnings. Valuing intangibles and future earnings based on required information only is difficult and may therefore lead to distorted earnings that are less informative. As a result, firms operating in high technology industries are more likely to voluntarily disclose information to make their earnings more informative (Chen et al., 2002). In conclusion, industry and in particular high technology industries, are associated with the decision to voluntarily report information.

The second characteristic is firm size. An often referred study is that from Meek et al. (1995). They argue that larger firms in general report more information than smaller firms. Prior literature supports this result and suggests that firm size is one of the main determinants of providing information on top of those required by regulations. That is because larger firms have more resources, and can better afford the costs associated with voluntary disclosure (Broberg et al., 2010; Depoers, 2000; Marston & Polei, 2004; Zhang, 2015). Further, larger firms can afford highly educated people as well as reporting systems that are better able to provide extra information. Finally, larger firms may also voluntarily provide information to meet the demand of analysts and other users of the financial statement. Overall, prior studies argue that firm size has a positive relation with the amount of voluntarily disclosed information.

Third, prior literature often discusses profitability as a firm characteristic to explain why firms voluntarily disclose information (e.g. Broberg et al., 2010; Marston & Polei, 2004; Meek et al., 1995; Zhang, 2015). The literature argues that firms that are profitable want to distinguish themselves from firms that are not, so they can raise capital at the best price. By voluntary information disclosure, firms can distinguish themselves. Hence, the profitability of a company is positively associated with disclosing extra information, on top of those that is required.

(20)

18

Last, analyst coverage positively influences a firm’s decision to voluntarily disclose information. Black et al. (2017) argue that analysts use alternative earnings measures, in addition to the standard earnings, to make more accurate estimations about the firm. However, obtaining extra information is associated with higher costs. By providing extra information, the costs for analysts to obtain information decrease. Besides, firms want to meet the demands from analyst for more information (Bowen, Davis, & Matsumoto, 2005). Overall, results suggest that a firm with higher analyst coverage is more likely to disclose voluntary information (Chen et al., 2002; Shehata, 2014).

(21)

19

3 Hypothesis development

This chapter discusses prior literature to develop hypotheses this study uses, in order to provide an answer to the research question. Evidence in prior research shows that companies frequently use non-GAAP measures in their earnings press releases. For example, Bhattacharya et al. (2003) show a significant increase of non-GAAP earnings disclosure during 1998 to 2000. Zhang and Zheng (2011) find a similar result in their research. Bhattacharya et al. (2007) and Entwistle et al. (2010) support this view as well and conclude that the disclosure of non-GAAP earnings increases until mid-2001. They add that disclosure of non-GAAP performance measures decreases after implementation of SOx and Regulation G by the SEC. However, Black et al. (2012) show that despite the two-year dip of reporting non-GAAP figures due to SOx and Regulation G, firms increasingly report non-GAAP earnings afterwards. While firms frequently report non-GAAP earnings, the disclosures tend to cluster in firms with certain firm characteristics. This thesis provides a comprehensive examination of these firm characteristics.

The first firm characteristic this study examines is the industry a firm operates in. Bhattacharya et al. (2003) find that managers are more likely to report non-GAAP numbers in the financial statement, when GAAP earnings are negative. Further, evidence shows that firms operating in high technology industries, often report GAAP losses and are more likely to disclose non-GAAP earnings than firms in other industries. Further, Bhattacharya et al. (2007) and Black et al. (2012) argue that non-GAAP reporting is highly concentrated in certain service industries and manufacturing industries. Bhattacharya et al. (2004) add that in the service industry, around two-thirds is included due to technology related services and they find that of the total non-GAAP earnings they examined, around half is reported by high technology firms. Additional research shows that firms operating in the high technology industry make large-scale investments in intangible assets, such as R&D (Lougee & Marquardt, 2004). These investments may lead to distorted GAAP earnings, which are less informative than GAAP figures reported in other industries. Further, results show that firms with less informative earnings are more likely to report non-GAAP earnings than other firms. The findings of Lougee and Marquardt (2004) therefore show that high technology industries make up nearly 40 percent of the sample reporting non-GAAP earnings. Zhang and Zheng (2011) show a similar result, by suggesting that high-tech firms make up over half of their sample reporting non-GAAP earnings. The authors follow the same reasoning as Lougee and Marquardt (2004) for these results. The abovementioned arguments lead to the following first hypothesis:

H1: Companies operating in high technology industries are more likely to report non-GAAP earnings.

(22)

20

The second characteristic that can influence the use of non-GAAP reporting is firm size. Prior literature argues that larger firms are more likely to make greater voluntary disclosures (e.g. Firth, 1979; Meek et al., 1995; Zhang, 2015). This would be the case because larger firms have more resources and can best afford the costs that are involved with voluntary disclosure. Baumker, Biggs, McVay and Pierce (2013) also argue that larger firms are more likely to voluntarily disclose additional informative information and in particular find evidence that larger firms tend to disclose non-GAAP earnings in contrast to small firms. However, more recently, Black et al. (2018) find no relationship between firm size and the likelihood to report non-GAAP earnings. Following the majority of prior studies, the second hypothesis is as follows:

H2: Larger firms are more likely to disclose non-GAAP earnings compared to smaller firms.

Third, several prior literature suggests that the number of analysts following a firm can also influence the use of GAAP reporting by companies. Analysts play a monitoring role, have influence on a firm’s non-GAAP disclosure and reduce information asymmetry (Black, Christensen, Ciesielski, & Whipple, 2018). This relationship is explained as follows. Prior literature suggests that besides using GAAP earnings to evaluate a firm’s performance, analysts and other stakeholders also use alternative earnings measures (Black et al., 2018). In general, they exclude items from GAAP earnings that are less relevant, such as unusual and nonrecurring items, to arrive at the non-GAAP earnings measure (Bradshaw & Sloan, 2002). In doing so, analysts can make more accurate estimates and analyses about the firm they follow. However, obtaining this extra information is associated with additional costs (Hamrouni, Benkraiem, & Karmani, 2017). As a result, Lang and Lundholm (1993) argue that one reason to voluntarily disclose information, is to lower the transaction costs that are involved. That is, more information provided by companies reduces the costs for analyst to acquire the information they need. Hence, these companies are more attractive to analysts. Following Bowen et al. (2005), firms that have higher analyst coverage tend to place more emphasis on non-GAAP earnings. An explanation for this is that companies want to meet analysts’ demand for more information (Bowen et al., 2005). Based on the foregoing, the third hypothesis is formed.

H3: Companies with higher analyst coverage are more likely to report non-GAAP earnings.

A fourth firm characteristic this thesis examines concerns the amount of intangibles a firm holds. Gelb (2002) examines the effect of intangible assets on firms’ disclosure decisions. He argues that firms may enjoy the benefits of these investments for multiple years. Yet, under GAAP reporting, intangibles are expensed, unless strict criteria are met. Hence, Gelb (2002) and Lougee and Marquardt (2004) argue that

(23)

21

if a company heavily invests in these intangibles, such as research and development (R&D), the GAAP earnings are distorted. As a result, this reduces the perceived usefulness of GAAP earnings to investors. Therefore, managers are more likely to emphasize voluntarily disclosed earnings and they use their discretion to exclude for example R&D expenses, which increases the non-GAAP earnings number. Besides, it provides investors with more useful information (Gelb, 2002; Lougee & Marquardt, 2004). Zhang and Zheng (2011) support this notion and provide evidence that for the non-GAAP earnings sample they examine, about two-thirds are intangible-intensive firms. Last, recent evidence shows that the intangible-intensity of a firm is positively associated with the use of non-GAAP earnings (Black et al., 2018). This theory leads to the following hypothesis:

H4: Companies with higher intangible-intensity are more likely to report non-GAAP earnings.

Finally, the GAAP earnings level of a company also influences a firms’ non-GAAP reporting decision. Prior research shows that managers aggressively use exclusions when preparing non-GAAP earnings. Bhattacharya et al. (2004) find evidence suggesting that firms disclosing non-GAAP earnings are in general less profitable than firms that do not report non-GAAP numbers. Specifically, evidence shows that over 10 percent of firms disclosing non-GAAP earnings, turn a GAAP loss into a non-GAAP profit (Bhattacharya et al., 2004). This suggests that firms use non-GAAP earnings opportunistically. However, Leung and Veenman (2018) provide evidence that firms that convert a GAAP loss into a non-GAAP profit have strong uninformative GAAP earnings. This suggest that loss firms use non-GAAP earnings to provide financial statement users with more informative earnings. Black et al. (2012) examine whether there is a change in the frequency of firms reporting non-GAAP measures opportunistically, after the implementation of SOx and Regulation G. They find that regulation has a positive effect, resulting in fewer firms using non-GAAP earnings to turn a GAAP loss into a non-GAAP profit. Yet, Black et al. (2012) support the notion of Bhattacharya et al. (2004), by stating that just under 10 percent of firms still disclose non-GAAP earnings in order to turn a GAAP earnings loss into a non-GAAP zero profit, or even a positive profit. Aubert (2010) finds a similar result. The author argues that of firms disclosing non-GAAP earnings, 95 percent reports a non-GAAP profit, while only 90 percent reports a GAAP profit. In other words, half of firms reporting a GAAP loss, report a non-GAAP profit. Despite the abovementioned results, Black et al. (2018) find that firms reporting a loss are less likely to disclose non-GAAP earnings. Based on the foregoing, this thesis follows the majority of prior results. Therefore, the fifth hypothesis is as follows:

(24)

22

The hypotheses discussed above are stated in alternative form. The corresponding null hypothesis is that companies operating in high-tech industries, that are larger, have higher analyst following, are more intangible-intensive and report a GAAP loss, are not more likely to report non-GAAP earnings in comparison to other companies that do not meet these firm characteristics.

(25)

23

4 Methodology

This chapter discusses the sample and sample selection details this study uses. The section thereafter shows the logit regression model and explains the operationalization of the variables used in the model. The final section discusses the descriptive statistics of the sample and the Spearman correlation matrix.

4.1 Sample selection procedures

This study examines the influence of firm characteristics on non-GAAP reporting decisions, described in the previous chapter. Therefore, this thesis uses a recently available sample obtained from Bentley et al. (2018). The authors provide a data set containing quarterly data about earnings announcements from U.S. firms. The period examined is from end 2003 through 2016 because regulation G, which requires firms reporting non-GAAP figures to disclose the most comparable GAAP figure as well as a reconciliation between the GAAP and non-GAAP earnings, became effective in 2003. The dataset contains 7,090 firms and 146,121 observations for which the authors identify whether these firms report non-GAAP earnings during the sample period, and if so, what the non-GAAP earnings number of the firm is. The sample indicates this as one if a firm reports non-GAAP earnings and zero otherwise.

Next, this study collects the data for the independent variables by searching for information about the same companies that are stated in the sample about non-GAAP disclosures. This matching is based on the global company key (hereafter gvkey), which is a uniquely identified code for every company in the Compustat database, as well as the fiscal year date of the firm. Data for the independent variables this study uses are available through databases within the Wharton Research Data Services (WRDS) system. The sample period is the same as the one available in the non-GAAP earnings disclosure dataset provided by Bentley et al. (2018). Data about the historical Standard Industry Classification (SICH) code, total assets, intangible assets and net income of the firm is retrieved through the Compustat database from WRDS. The I/B/E/S database provides the number of analysts that follow the firm and is also available through WRDS. While quarterly-level accounting data for U.S. companies reporting non-GAAP earnings is available, this thesis uses annual information, because not for all independent variables is quarterly-level information available. Further, observations are eliminated for which information regarding the dependent and independent variables is missing. Table 1 presents the abovementioned sample selection procedures.

Further, data with respect to the SICH code is an often used measure when determining the industry a firm operates in (e.g. Lougee & Marquardt, 2004; Zhang and Zheng, 2011). Firstly, as in prior

(26)

24

literature (e.g. Marques, 2006), this study eliminates financial firms with SICH codes 6000-6999, because these firms differ extensively from non-financial firms. As Table 1 shows, the abovementioned procedures lead to a sample of 58,273 observations with information regarding Compustat variables. Second, as prior literature often refers to Francis and Schipper (1999) to specify a high-tech industry, this study also follows these authors. The SIC code they refer to is based on three digits, while the SICH code retrieved from Compustat includes four digits. Where a three-digit SIC code defines the industry group a firm operates in, e.g. computer and office equipment, adds the four-digit SIC code a number to determine the specific industry of the firm, for example computer communications equipment. In order to determine whether a firm can be identified as high-tech, the fourth digit of the SICH code is dropped and the three digit SICH code is then compared with the industry group code as specified by Francis and Schipper (1999). The SICH codes are as follows: Drugs (283), Computer and Office Equipment (357), Electrical Machinery and Equipment, Excluding Computers (360), Electrical Transmissions and Distribution Equipment (361), Electrical Industrial Apparatus (362), Household Appliances (363), Electrical Lighting and Wiring Equipment (364), Household Audio, Video Equipment, Audio Receiving (365), Communication Equipment (366), Electronic Components, Semiconductors (367), Computer Hardware (368), Telephone Communications (481), Computer Programming, Software, Data Processing (737), and Research, Development, Testing Services (873). The independent variable SICH is a dummy variable equal to one if the firm operates in a high-tech industry and zero otherwise.

Table 1. Sample selection criteria to obtain sample

Description n

Unique firm-year observations of U.S. firms during 2003-2016 94,348

Drop if total assets have no value - 2,283

Drop if total assets are less than or equal to zero - 23

Drop if industry code SICH is missing - 3,810

Drop if industry code SICH is between 6000 and 6999 - 14,835

Drop if intangible assets have no value - 2,260

Drop if net income has no value - 12,864

Firm-year observations Compustat dataset 58,273

Drop if no match after merge - 31,902

Firm-year observations final sample 26,371

Notes: This table describes the sample selection based on firm-year observations of Compustat and I/B/E/S. SICH code is the historical industry code of the firm. SICH codes 6000-6999 are dropped because these financial firms are significantly different from non-financial firms. The table presents the sample after merging with the Bentley et al. (2018) and I/B/E/S dataset. Observations are dropped when there is no match between the Bentley et al. (2018) dataset and the Compustat dataset. In addition, observations are dropped when they are only present in the I/B/E/S dataset.

(27)

25

Second, data about the number of analysts that follow a firm cannot be obtained from Compustat, contrary to the other independent variables, but from the I/B/E/S database. In this database it is not possible to use the gvkey company code. Therefore, this study uses the eight-digit CUSIP, which is a uniquely identified code for every company, as is gvkey. However, the CUSIP available through Compustat contains nine digits, where the ninth digit checks the accuracy of the previous eight digits (CUSIP Global Services, n.d.). Therefore, the last digit is dropped, to retrieve information from the I/B/E/S database based on an eight-digit CUSIP. The obtained dataset contains information about the number of analysts that follow the firm, the date when the amount of earnings per share is estimated, and the forecast period this estimation has been made. In general, analysts make this estimation for the fiscal year date.

Using gvkey and quarterly reporting dates of the Bentley et al. (2018) dataset and gvkey and yearly reporting dates of the Compustat variables dataset, the two dataset are merged. Both dates belong to the variable date, and contain the reporting day, month and year, and can therefore be matched. As a result of matching quarterly and yearly dates, only the fiscal year reporting dates remain. Further, if there is no match between the two datasets based on gvkey and date, then the observation is dropped. After merging the Bentley et al. (2018) dataset and the dataset with Compustat variables, the sample consists of 26,371 observations in both datasets. Next, the first merged dataset (hereafter non-GAAP dataset) and the analyst coverage dataset are matched, using the eight-digit CUSIP and date. From the I/B/E/S dataset the date is when forecasts about the earnings per share of the firm are made. In order to match this date to the fiscal year date of the other dataset, this study uses only the month and year of both dates. Observations are dropped if they do not appear in both datasets, but only in the second dataset. However, if an observation only appears in the first dataset and no information is given regarding the number of analysts that follow a firm, then this missing value is replaced by zero. By doing this, no observations are dropped unnecessarily. Table 1 shows that the foregoing steps lead to a final sample of 26,371 observations.

4.2 Model

The model this thesis tests in order to provide an answer to the research question, takes the following form and is estimated using a logistic regression:

Non −GAAPit = α + β1Hi-Techit + β2Sizeit + β3Analystsit + β4BIit + β5Lossit + β6ROAit +

(28)

26

Where Non-GAAP is a dummy variable equal to one for company-year observations in which firm i reports non-GAAP earnings, and zero otherwise. Hi-Tech is also a dummy variable equal to one if the firm operates in the high technology industry and zero otherwise. The study of Francis and Schipper (1999) is often referred to by prior studies (e.g. Lougee & Marquardt, 2004; Bowen et al., 2005; Zhang & Zheng, 2011) when determining what is specified as a high-tech firm. Further, prior literature often refers to firms operating in high-tech industries with a specific SICH code. Therefore, this study adopts the SICH codes specified by Francis and Schipper (1999). Prior literature defines several measures as to how to specify the Size of a firm. Following Lougee and Marquardt (2004), this study defines firm size as the natural logarithm of the total assets of the firm. The third independent variable (Analysts) measures the number of analysts that follow the firm and is obtained through the I/B/E/S database of WRDS. Next, BI represents the intensity of the average book value of intangible assets on the firm’s balance sheet. The intangible-intensity represents intangibles assets divided by the total assets of the firm at the end of the fiscal year (Lougee and Marquardt, 2004). Finally, Loss is the fifth independent variable and again a dummy variable, meaning it is equal to one if firm i reports negative GAAP earnings at year end and zero if the GAAP earnings are positive.

This study adds two control variables to the model that could influence non-GAAP reporting. The first control variable is the profitability of the firm, or return on assets (ROA), which is measured as the net income divided by total assets. It is expected that a higher ROA, leads to an increase in the use of non-GAAP reporting. Besides, firm size, intangible-intensity and other independent variables are also also associated with the profitability of the firm and therefore, it is important to control for ROA. The second control variable (Trend) takes into account the time trend of reporting non-GAAP earnings. Table 2 shows the fraction of firms reporting non-GAAP earnings in a certain year. Results show that non-GAAP reporting significantly increases over time, from 25.2 percent in 2003 to 65.4 percent in 2016. Figure 1 graphically shows the increase in non-GAAP reporting during the sample period. Subsequently, this study adds the time trend variable to the model and controls for the increase in non-GAAP reporting, which is not explained by the independent variables. Further, the size of a firm may increase over time and may influence the use of non-GAAP reporting due to time effects. As a result, a trend variable (Time) is added to the model. Time is measured as the firms’ reporting year minus 2002, which is one year before the sample period starts.

(29)

27

4.3 Descriptive statistics

Panel A of Table 3 provides an overview of the descriptive statistics for the variables this study uses for the full sample. Table 3 Panel B shows the descriptive statistics for the variables, where a distinction is made between firms reporting non-GAAP earnings and firms reporting GAAP-only. Merging all data results in a sample of 26,371 observations. This sample is split into two subsamples, where the first subsample contains firms reporting non-GAAP earnings and consists of 10,659 firms. The second subsample consists of 15,712 firms that report GAAP earnings only. Minimizing the influence outliers have on the sample, the continuous variables are winsorized at the 1st_{and 99}th_{percentiles of the distribution. By doing so, the}

values in the smallest percentile are set equal to the smallest value of the 2nd_{percentile and the values in}

the largest percentile are set equal to the largest value of the 99th_{percentile. The descriptive statistics in}

Table 3 Panel B show that on average, one-third of the full sample operate in the high-tech industry. This is split into 39.1 percent of firms reporting non-GAAP earnings operating in the high technology industry, while 30.2 percent of firms only reporting GAAP earnings operate in this industry. The mean and median of firm size show that companies reporting non-GAAP are larger than firms that disclose GAAP earnings only. Further, the median number of analysts that follow a non-GAAP firm is 6, while the median number of GAAP-only firms is 3 analysts. The BI of non-GAAP versus GAAP-only firms

Table 2. Non-GAAP reporting during sample period (2003-2016)

Year Total non-GAAP

reporting (n) Total observations sample (n) % Non-GAAP reporting 2003 545 2,162 25.2 2004 541 2,306 23.5 2005 620 2,248 27.6 2006 693 2,214 31.3 2007 716 2,278 31.4 2008 903 2,258 40.0 2009 888 1,999 44.4 2010 858 1,895 45.3 2011 885 1,805 49.0 2012 941 1,781 52.8 2013 974 1,783 54.6 2014 978 1,763 55.5 2015 1,017 1,726 58.9 2016 100 153 65.4

Notes: Column one and column two show the number of firms that report non-GAAP earnings and the total number of firms in the specific year, respectively. Column 3 present the percentage of firms reporting non-GAAP earnings relative to the total amount of firms in the sample for that specific year. The numbers in the third column are rounded up to one decimal place.

Characteristics of firms disclosing non-GAAP earnings