Does financial reporting quality affect the quality of non-GAAP disclosures?

(1)

Does financial reporting quality affect the quality of

non-GAAP disclosures?

Name: Heleanne Hazewindus Student number: 11120274

Thesis supervisor: prof. dr. D. Veenman Date: 24 June 2017

Word count: 16,842

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

(2)

Statement of Originality

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

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

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

(3)

Abstract

The increasing usage of non-GAAP disclosures suggests that Generally Accepted Accounting Principles (GAAP) are not as efficient as standard setters are thinking. The different ways to determine the unaudited non-GAAP earnings are a heavily debated topic because the incentives of managers deciding to disclose non-GAAP earnings differ among companies. Using archival data gathered from Compustat and IBES over the time period 2006-2015, I focus on companies in the United States. First, I examine if a relation exists between the likelihood of non-GAAP disclosures and high financial reporting quality. Second, I examine if the financial reporting quality affects the quality of non-GAAP disclosures. The financial reporting quality is operationalized by measuring the accrual quality using the modified Dechow & Dichev model (2002). Furthermore, the quality of non-GAAP disclosures is first measured by the magnitude of the total exclusions. Hereafter, I extend this relation by splitting total exclusions in ‘special item’ exclusions and ‘other item’ exclusions. Overall, I find a positive relation between the likelihood of non-GAAP earnings and high accrual quality of companies, indicating that companies disclosing non-GAAP earnings have a higher accrual quality. Besides, my regression analysis reveals a positive relation between the accrual quality measure and total exclusions, meaning that companies with lower quality financial reporting make higher total exclusions. This suggests that companies with lower financial reporting quality have lower quality non-GAAP earnings. Moreover, I find that there is a negative relation between the accrual quality measure and ‘special item’ exclusions, meaning that companies with lower financial reporting quality exclude lower amounts of ‘special items’. Lastly, I find a positive relation between the accrual quality measure and ‘other item’ exclusions, suggesting that companies with lower financial reporting quality make higher ‘other item’ exclusions.

Keywords: non-GAAP disclosures; special item exclusions; other item exclusions; financial reporting quality; accrual quality.

(4)

1 Introduction

Nowadays non-GAAP disclosures are more common than ever and the motivation of these disclosures is a heavily debated topic. By the early 2000s around 300 firms in the S&P 500 were excluding GAAP expenses from their non-GAAP earnings stated in press releases (Baumker, Biggs, McVay, & Pierce, 2014). The main reasons why managers present non-GAAP earnings are (1) the perceived performance of a firm can be improved by excluding expenses in non-GAAP disclosures, and (2) by disclosing non-GAAP earnings firms try to be more effective in communicating the permanent earnings of a firm (Heflin & Hsu, 2008). Because of the increasing number of non-GAAP disclosures, the U.S. Security and Exchange Commission (SEC) implemented, in March 2003, new non-GAAP earnings disclosure rules (Heflin & Hsu, 2008).

The reaction of investors to non-GAAP disclosures differs according to prior research. Johnson & Swartz (2005) find that investors are not misled by the non-GAAP disclosures. However, other studies find that only nonprofessional investors are misled by non-GAAP disclosures (e.g., Bhattacharya, Black, Christensen, & Larson, 2003; Elliot, 2006; Allee, Bhattacharya, Black, & Christensen, 2007). Furthermore, research is done about the reasons why non-GAAP disclosures are used to meet or beat analysts’ forecasts. Again different findings explain those reasons like being a credible company, being able to increase the stock price, improving the capability of the managers to the outside, and making sure that potential future growth is noticed by outsiders (Graham, Harvey, & Rajgopal, 2005). More recently, Isidro & Marques (2015) find that companies use GAAP earnings to meet or beat analysts’ forecasts when they are operating in countries with good working legal systems, where investors are protected, capital markets are well developed, and there are good communication channels.

Scepticism about the usage of non-GAAP disclosures comes also from the fact that the earnings within this disclosure are not audited (Black, Christensen, Joo, & Schmardebeck, 2016). On the other hand, financial reports of companies are audited and managers need to make sure that their reports are in accordance with GAAP. Because of all different outcomes in prior literature with relation to the usage of non-GAAP disclosures and due to the still developing nature of non-GAAP disclosures, it is interesting to investigate if the extent to which GAAP is followed by managers is influencing the way non-GAAP earnings are used. Specifically, my research question is: does financial reporting quality affect the quality of non-GAAP disclosures? The motivations for this research question are firstly that, in general, non-GAAP disclosures present a more positive picture of the firm value compared to the GAAP disclosures and, secondly, non-GAAP disclosures are often seen as substitutes for earnings management to meet or beat analyst

(7)

forecasts (e.g., Bhattacharya, Black, Christensen, & Mergenthaler, 2004; Black & Christensen, 2009; Chen, Krishnan, & Pevzner, 2012; Doyle, Jennings, & Soliman, 2013; Isidro & Marques, 2015). Interestingly, non-GAAP disclosures can also be seen as complements of financial reports. This all depends on the main reason why managers present non-GAAP disclosures and if these motives are driven by providing useful information for investors, or providing information to meet or beat analysts’ forecasts. So, when managers prefer high quality financial reports they can also strive for more space using non-GAAP disclosures to circumvent the strict GAAP rules and make high total exclusions, which results in lower quality non-GAAP disclosures. However, it is also possible that managers strive for high quality non-GAAP earnings as well and exclude only some costs to be more informative.

By first examining if there is a relation between the usage of non-GAAP earnings and the quality of the financial reporting of those companies and, second, examining if these companies make high or low total exclusions, I aim to provide evidence on whether companies disclosing non-GAAP earning have high or low quality financial reporting, followed by if those companies’ financial reporting quality is affecting their non-GAAP disclosures quality. I specifically focus on companies from the United States during the period 2006-2015. I follow prior research and use the modified Dechow & Dichev model (2002) for measuring the accrual quality. Just like prior research suggests, the better the link between accruals and cash flows of a company, the higher the quality of the accruals and thus, the less of the model is explained by the residual of the model (Dechow & Dichev, 2002). So, the quality of the accruals will increase (decrease) when the magnitude of the residual decreases (increases) (Dechow & Dichev, 2002). I retrieved data from Compustat and after cleaning the data I ended up with 39,738 observations in 468 specific industry groups per fiscal year. The cash flows and accruals are expected to correlate negatively because of the volatility of cash flows and because the accruals are used to make sure earnings are less volatile (Dechow & Dichev, 2002). However, I find a positive relation, which is in line with the findings of Bushman, Lerman, & Zhang (2016), suggesting a changing landscape of accrual accounting because the timing role of the accruals has declined in recent years. Because I was only interested in the residual value of the conducted regression analysis I continued with the modified Dechow & Dichev model (2002) as an operational variable for financial reporting quality. Like prior research, the calculated standard deviation of the residual contains at least three years of residual data with a total of five years (Ashbaugh-Skaife, Collins, Kinney Jr., & LaFond, 2008; Veenman, 2012). Due to the requirements of at least three years and a total of five years, the accrual quality is available for 25,239 observations in 7,749 specific industry groups per fiscal year of the 39,738 observations.

(8)

Next, I test whether non-GAAP disclosing companies have a high or low financial reporting quality. Retrieving data from IBES and merging all data into one master file result in a sample of 12,444 observations. This total sample can be split in 7,850 observations using non-GAAP disclosures and 4,594 observations using only non-GAAP disclosures. My sample reveals that companies disclosing non-GAAP earnings have higher accrual quality than companies disclosing only GAAP earnings. Black, Christensen, Ciesielski, & Whipple (2017) find in their research that companies reporting non-GAAP disclosures have lower GAAP earnings per share than companies that do not disclose non-GAAP earnings. However, I find that non-GAAP disclosing companies have higher amounts of GAAP earnings per share compared to GAAP-only companies. Results of the OLS regression reveal that companies disclosing non-GAAP earnings have a higher accrual quality. Control variables are added to the regression analysis based on prior research (e.g., Whipple, 2015; Bentley, Christensen, Gee, & Whipple, 2016; Leung & Veenman, 2016).

My main test focuses on the financial reporting quality of the group of non-GAAP disclosing companies and if their financial reporting quality affects the quality of their non-GAAP disclosures. The non-GAAP disclosure quality is measured by the total amount of total exclusions made by managers. Within my sample, the average total difference between non-GAAP and GAAP earnings is positive, indicating that by the creation of non-GAAP disclosures income-decreasing expenses of companies are excluded from the companies’ GAAP disclosures (Kolev, Marquardt, & McVay, 2008). My sample for this regression analysis is measured with 7,850 observations in 60 specific industry groups per fiscal year. The regression reveals a positive relation between the accrual quality measure and total exclusions, meaning that companies with a higher standard deviation of the residual make higher total exclusions. This suggests that companies with lower financial reporting quality have lower quality non-GAAP disclosures. Besides the ordinary regression, I corrected the regression analysis for changes per industry and per fiscal year. Correcting for these fixed effects makes sure the relation between the dependent and independent variable is not driven by side effects like industry changes and changes in non-GAAP reporting and accrual quality across years.

To extend the main test, the total exclusions are split in ‘special item’ exclusions and ‘other item’ exclusions. Both types of exclusions are used in a regression analysis. The results reveal that the accrual quality measure is significantly correlated with both the ‘special item’ exclusions and the ‘other item’ exclusions, suggesting that non-GAAP disclosing companies affect the type of exclusions made in their non-GAAP earnings. Moreover, there is a negative relation between the accrual quality measure and ‘special item’ exclusions, meaning that companies with a higher standard deviation of the residual are excluding less ‘special items’. This is in line with the findings

(9)

in prior research. Because Whipple (2015) finds that ‘special items’ are used to create a better understanding of the core performance of a company and Black & Christensen (2009) find that the exclusion of ‘special items’ improves the quality of the earnings number for predicting future cash flows and valuing the firms. So, companies with already low financial reporting quality are also not striving for high quality and improving non-GAAP disclosures. On the other hand, a positive relation exists between the accrual quality measure and ‘other item’ exclusions, suggesting that companies with lower financial reporting quality exclude more ‘other items’. This finding is in line with the finding of Whipple (2015) that despite the fact that ‘other item’ exclusions are recurring expenses and less easy to explain by managers, the exclusions of these ‘other items’ result in more informative earnings because most of these recurring items refer to non-cash items. So, low quality financial reporting companies try to be more informative. However, in this case it seems more plausible that lower financial reporting companies make higher ‘other item’ exclusions to be more positive about performances and misleading or beat analysts’ forecasts like other prior research states (e.g., Landsman, Miller, & Yeh, 2007; Black & Christensen, 2009; Doyle et al., 2013).

My study makes several contributions to prior literature. First, it contributes to extend the literature on non-GAAP disclosures because this will be the first paper that examines the relation between financial reporting quality and non-GAAP disclosure quality. Prior research about the topic non-GAAP disclosures finds mixed results why these disclosures are used, how investors react to those disclosures, and the incentives of managers why they meet or beat analysts’ forecasts, therefore, my relation is adding new insight in the usage of non-GAAP earnings. Second, my thesis contributes to the accrual accounting literature. My findings confirm the finding of Bushman et al. (2016), who find a changing landscape of accrual accounting because the timing role of the accruals has declined in recent years. I made an extra check to shift the timeline of the downloaded data to earlier years and conducted the same test over again. In this situation the expected negative relation between accruals and cash flows was visible again. Because I was interested in the residual value to measure the accrual quality, this was no problem for my thesis. This thesis has an important implication for the heavily debated topic, and increasing number of non-GAAP disclosures. Overall, questions arise if the GAAP are still effective and best way to report. According to chairman Russel Golden of the Financial Accounting Standards Board (FASB), they can learn from the need for non-GAAP disclosures and apply suited changes in the GAAP that results in reducing the amount of non-GAAP disclosures (Golden, 2017). My results suggest that financial reporting quality is associated with the quality of non-GAAP disclosures. In practice regulation G has a positive impact on the quality of non-GAAP disclosures (e.g., Heflin & Hsu, 2008; Kolev et

(10)

al., 2008; Black, Black, Christensen, & Heninger, 2012). Because of interference of regulators, standard setters, auditors, and other capital market intermediaries it is possible to increase the quality of voluntary disclosures (Healy & Palepu, 2001).

The next section describes my literature review. This section first describes non-GAAP disclosures and financial reporting in general and, second, it explains suitable theories for this thesis. My third section explains the relation between the variables, which results in the hypotheses. In the fourth section I explain my research methodology. Within this section I show all the empirical models I use in this thesis. Then, in section five I describe the results of the regression analyses. In addition, section six describes the conclusions of the results. Lastly, I add my references and three appendices.

(11)

2 Literature

2.1 Non-GAAP disclosures

2.1.1 Creation of non-GAAP disclosures

Non-GAAP disclosures arise from exclusions made from GAAP earnings of various nonrecurring or noncash items that managers believe are irrelevant for determining the financial performances of the firm (Entwistle, Feltham, & Mbagwu, 2006). These exclusions are done in two different ways (1) by increasing the amount of the excluded costs, and (2) by artificially creating new types of exclusions (Entwistle et al., 2006). So, in both cases the non-GAAP earnings increase compared to the GAAP earnings. According to prior research, the difference between GAAP and non-GAAP disclosures can be divided in exclusion costs called ‘special items’ and ‘other items’ (Whipple, 2015). One-time items are categorized in ‘special items’, which are exclusions made by managers and analysts in calculating non-GAAP earnings, like for example a litigation charge (Whipple, 2015). The other part of the exclusion costs that remains, is categorized as ‘other items’ (Whipple, 2015). According to Whipple (2015), ‘special item’ exclusions are done to create a better understanding of the core performance of a company. However, the explanation of managers for the exclusion of ‘other items’ is less clear because these ‘other items’ are recurring expenses (Whipple, 2015; Leung & Veenman, 2016). These exclusions are more susceptible to be misleading exclusions made by managers and are seen as more opportunistic exclusions compared to one-time or ‘special items’ (Whipple, 2015; Leung & Veenman, 2016).

The two main reasons why managers present non-GAAP disclosures are firstly, because managers try to be more effective in communicating permanent earnings of their firm and, secondly, the performance of a firm can be improved by excluding more expenses in the non-GAAP disclosure (Heflin & Hsu, 2008). Moreover, there are two different explanations for the two reasons why managers disclose non-GAAP earnings. On the one hand, managers want to reduce the information asymmetry and disclose more company information and, on the other hand, there is a possibility managers want to mislead investors by disclosing more positive earnings (Marques, 2006; Curtis, McVay, & Whipple, 2014). Another incentive for managers to create non-GAAP disclosures is, according to Isidro & Marques (2015), the fact that investors react to disclosures with non-GAAP earnings. Furthermore, they state that the most common expenses that are excluded by managers in GAAP numbers in developed countries are the recurring expenses like research and development, depreciation, and stock-based compensation expenses (Isidro & Marques, 2015).

(12)

Then the question arises who is responsible for the creation of those non-GAAP disclosures? In principle the management of the company is responsible because disclosures are shown in press releases and not shown in audited financial statements (Chen et al., 2012). So, scepticism about non-GAAP disclosures comes from the fact that the earnings within these disclosures are not audited (Black et al., 2016). On the other hand, according to the International Standard on Auditing (ISA) 200 “auditors obtain reasonable assurance about whether the financial statements as a whole are free from material misstatements, whether due to fraud or error, thereby enabling the auditor to express an opinion on whether the financial statements are prepared, in all material respects, in accordance with an applicable financial reporting framework; and to report on the financial statements, and communicate as required by the ISAs, in accordance with the auditor’s findings” (ISA, 2016).However, non-GAAP earnings are combined with GAAP earnings in one press release, therefore, auditors need to pay attention to what is stated in those non-GAAP earnings. Chen et al. (2012) state that an auditor is a protector of public trust and because of that they indirectly carry the responsibility that there are no material mistakes made or biased information is transferred in voluntary disclosures to the public. This all because interested people have an expectation that all numbers that are disclosed of a company give a true and fair view of the companies’ value.

2.1.2 Example of a non-GAAP disclosure

For a better insight of what a non-GAAP disclosure looks like an example is shown of Twitter Inc. in appendix 1. This example shows that Twitter Inc. reports under GAAP a net loss of $457 million at the end of 2016 and a net loss of $521 million at the end of 2015. However, their non-GAAP net income shows a net income of $119 million and $115 million, respectively. According to the press release Twitter defines their non-GAAP net income as “net loss adjusted to exclude stock-based compensation expense, amortization of acquired intangible assets, non-cash interest expense related to convertible notes, non-cash expense related to acquisitions, the income tax effects related to acquisitions and restructuring charges” (Twitter Inc., 2017). Twitter defines non-GAAP expenses as “total costs and expenses adjusted to exclude stock-based compensation expense, amortization of acquired intangible assets, non-cash expense related to acquisitions, and restructuring charges” (Twitter Inc., 2017). So, Twitter Inc. reported under GAAP earnings per share (EPS) at the end of the years 2016 and 2015 respectively, $(0.65) and $(0.79). But their non-GAAP EPS at the end of the years 2016 and 2015 are respectively, $0.57 and $0.40 (Twitter Inc., 2017).

(13)

Twitter’s reasons for disclosing those non-GAAP earnings are, firstly, because Twitter believes those earnings are assisting their investors in making clear what the management recognizes as operating results and, secondly, those earnings help investors to make comparisons with different companies in the same industry (Twitter Inc., 2017). Besides this, Twitter states that non-GAAP earnings are also helping by the identification of undetected trends and, moreover, they help to understand their performance from the past and for the future (Twitter Inc., 2017). Despite the given arguments that the non-GAAP earnings are helpful for investors, there is a difference in reporting a net loss (per share) or reporting a net income (per share) for a company. Because of this change it is possible that their non-GAAP earnings are seen as a more positive reflection of company performances. An overview of the differences between the quarterly disclosed non-GAAP and GAAP EPS of Twitter during the years 2015 and 2016 is shown in appendix 2. Overall, prior research states that firms exclude expenses to create their non-GAAP disclosures, but different firms do not make the same exclusions (Bhattacharya et al., 2004).

2.1.3 Reaction of investors to non-GAAP disclosures

A lot of research is done on whether the created non-GAAP disclosures are affecting the reactions of investors. For example, Frederickson & Miller (2004) find in their experimental research that unsophisticated investors are reacting to the non-GAAP earnings of firms. However, they state that sophisticated investors stay unaffected by non-GAAP earnings (Frederickson & Miller, 2004). Besides this, Brown, Christensen, Elliot, & Mergenthaler (2012) find in their research that managers are more willing to disclose non-GAAP earnings when the sentiment of the investor increases for those disclosures. Moreover, the prominence of these earnings in press releases increases with the level of investor sentiment as well (Brown et al., 2012). They also state that the relation between investor sentiment and non-GAAP earnings can be partially attributed to opportunistic motives of managers (Brown et al., 2012).

Additionally, Johnson & Schwartz (2005) find in their research, despite concerns regarding non-GAAP disclosures of regulators, that investors are not misled by the disclosure of non-GAAP earnings. Moreover, Elliot (2006) confirms the previous finding in another experimental research and states that the presence of non-GAAP disclosures is not influencing the judgement and decision of nonprofessional investors. However, Elliot (2006) also states that the decisions and the judgements of nonprofessional investors are affected by the focus management is placing on the non-GAAP disclosures compared to their GAAP disclosures. Results of both the experimental research of Frederickson & Miller (2004) as well as the experimental research of Elliot (2006) are confirmed in the archival data research of Allee et al. (2007). They conclude that less sophisticated

(14)

investors are affected by non-GAAP disclosures (Allee et al., 2007). Furthermore, Bhattacharya et al. (2003) state that investors find non-GAAP disclosures more informative and give a better representation of the main goals of a firm compared to the GAAP disclosures. Another study contradicts as well that non-GAAP disclosures are misleading because this study finds that companies keep developing their non-GAAP earnings’ calculations to create the most informative disclosure for investors (Black et al., 2017).

2.1.4 Timing of non-GAAP disclosures

Besides the investor’s interpretation of non-GAAP disclosures, there is also research done about the timing of those non-GAAP disclosures. Brown, Christensen, & Elliot (2012) investigated (1) if managers are timing the non-GAAP disclosures in the press releases and (2) if these disclosures are reflecting a more opportunistic view or a more informative view. In their research they conclude that the timing of non-GAAP disclosures are accelerated by management in quarters that companies disclose and adjust earnings metric, compared to quarters companies do not disclose and adjust at all (Brown et al., 2012). Furthermore, a second finding about the accelerated timing of the non-GAAP disclosures is that this acceleration increases when the exclusion of expenses in non-GAAP disclosures increase as well (Brown et al., 2012). Besides, they find that managers of companies accelerate the timing of the non-GAAP disclosures when information that is disclosed is less transparent (Brown et al., 2012). Another finding of Brown et al. (2012) relating to timing is that exclusion of expenses in the more accelerated non-GAAP disclosures are of relatively lower quality than exclusions in less accelerated disclosures.

Moreover, according to Gennotte & Trueman (1996) the impact of disclosures will be better reflected in market prices of companies when managers time the disclosure during the trading hours of a company compared to their disclosure timed after trading hours. Besides this, the management is at the same time more willing to disclose positive earnings during trading hours rather than disclosing less positive information (Gennotte & Trueman, 1996). This less positive information is more often disclosed after the trading hours of a company because the management is convinced that the market price reaction will be less extreme (Gennotte & Trueman, 1996). In the opinion of the managers, investors are in that case less able to notice the impact of the disclosed earnings. In accordance with this research Kothari, Shu, & Wysocki (2009) find that, with regard to the market price reaction, managers are on average delaying disclosures with less positive earnings and spreading more positive news about the company earlier.

(15)

2.1.5 Non-GAAP disclosures used to meet or beat analysts’ forecasts

Managers can have different reasons for meeting or beating analysts’ forecasts. According to Graham et al. (2005), these reasons are (1) being a credible company for the capital market, (2) being able to increase the stock price of your company, (3) improving the capabilities of managers to externalities, and (4) making sure that potential future growth of the company is noticed by outsiders. Besides this, prior research is examining the usage of non-GAAP disclosures to meet or beat analysts’ forecasts. In prior research findings show that non-GAAP earnings tend to meet or beat analysts’ forecast more often than the same firms’ GAAP earnings (Bhattacharya et al., 2004; Black & Christensen, 2009; Chen et al., 2012; Doyle et al., 2013; Isidro & Marques, 2015). For meeting or beating analysts’ forecasts managers are excluding most of the times recurring expenses like research and development, depreciation, and stock-based compensation expenses (Black & Christensen, 2009). When managers make sure that they are excluding those expenses in a way that they just meet or beat analysts’ forecasts they are behaving opportunistically (Doyle et al., 2013).

Furthermore, Isidro & Marques (2015) investigated the factors that influence managements’ behaviour to meet or beat analysts’ forecasts. They conclude that managers are more willing to use non-GAAP disclosures to meet or beat analysts’ forecasts when they are operating in countries with good working legal systems, where investors are protected, capital markets are well developed, and there are good communication channels (Isidro & Marques, 2015). Besides, they find that managers in countries with what they call “institutionally strong and economically developed jurisdictions” experience more pressure to meet or beat the analysts’ forecasts (Isidro & Marques, 2015).

2.1.6 Regulation G

Because of the increasing number of non-GAAP disclosures the U.S. SEC implemented, in March 2003, new non-GAAP earnings disclosure rules (Heflin & Hsu, 2008). Before the implementation of this rule the U.S. SEC already sent out press releases including a “cautionary advice” for companies if they want to disclose non-GAAP earnings at all, together with an “investor alert” that investors need to be sceptical about the disclosed information (SEC, 2001). This implemented regulation by the U.S. SEC is called regulation G and mandates companies that disclose non-GAAP earnings (1) to show the directly related non-GAAP earning of the disclosed non-non-GAAP earning, (2) contain the quantitative adjustment between the non-GAAP earning and the directly related GAAP earning, and (3) present these non-GAAP earnings in a way they are not misleading for investors (Heflin & Hsu, 2008). Overall, prior research provide a positive U.S. SEC’s opinion

(16)

towards the usefulness of interpretations or summaries made of GAAP disclosures (Katz, 2001). However, these interpretations or summaries can be misleading for investors when the real earnings are concealed (Katz, 2001). Thus, firms need to present specific information about the nature and magnitude of their non-GAAP disclosures compared to their GAAP disclosures (Black & Christensen, 2009).

This difference in the way of disclosing non-GAAP earnings after the implementation of regulation G has an impact on the quality of the exclusions. According to Kolev et al. (2008), exclusions of costs (on average) in non-GAAP disclosures are of a higher quality after the implementation of regulation G by the U.S. SEC. Besides this, they find firms that are not disclosing non-GAAP earnings anymore after March 2003 made exclusions of lower quality when they were still reporting non-GAAP earnings (Kolev et al., 2008). In the end, they conclude that this confirms the U.S. SEC’s objective to improve the quality of the non-GAAP disclosures (Kolev et al., 2008). Both Heflin & Hsu (2008) and Kolev et al. (2008) find in their studies a reduction in the amount of non-GAAP disclosures in the period after regulation G. Contradictory, Black et al. (2012) and Bentley et al. (2016) find only a temporary decline right after the implementation but further a general increase in the amount of non-GAAP disclosures since 1998. After the implementation of regulation G, investors are more interested in non-GAAP disclosures than GAAP earnings of a company (Black et al., 2012).

2.2 Financial reporting

2.2.1 Reasons for financial reporting

Within a capital market economy capital is flowing from household savings towards business firms and information is flowing the other way around from business firms towards households (Healy & Palepu, 2001). Because an information difference

exists between households (investors) and business firms, financial reporting is necessary to mitigate this problem. Another reason for financial reporting to exist is because there are agency conflicts between managers of a business firm and investors (Healy & Palepu, 2001).

The flow of capital and information is shown in figure 1 together with the financial and information intermediaries. The financial

Figure 1 The flow of capital and information (Healy & Palepu, 2001, p. 408)

(17)

information can be disclosed directly in press releases or indirectly in a financial report after the information is being audited by an auditor. If there is a perfect working market, disclosures with financial information would not be necessary to make and there will be no need for regulation of those disclosures with financial information (Fields, Lys, & Vincent, 2001). But, in real-life this is not the case and, therefore, we need auditors and accounting regulators to make sure the disclosures of managers give a true and faithful view of reality (Healy & Palepu, 2001).

2.2.2 Regulation of financial reports

Within the United States accounting standards are created by the FASB for the above mentioned participants in the capital market (SEC, 2000). These U.S. accounting standards, known as GAAP, help companies with the creation of the financial reports by making sure that these reports contain “transparent, consistent, comparable, relevant and reliable financial information” (SEC, 2000). The FASB states that “the purpose of financial reporting is to provide an objective look at a company’s financial situation” (FASB, 2017b). Moreover, this regulation of financial reporting is necessary for regulation of capital markets. With help of the regulation of financial reports it is possible for participants, like the investors, business firms, and auditors, to compare financial reports. This can result in a costs reduction for both preparers and users of the financial reports (FASB, 2017a). Furthermore, quality of the regulation is maintained by continually developing high quality GAAP standards that also results in influencing the future direction of financial standards (FASB, 2017a). Therefore, the FASB decided to participate in the development of the International Financial Reporting Standards (IFRS) and assists the International Accounting Standers Board (IASB) (FASB, 2017a).

According to Bradshaw & Miller (2008), this creation of a single set of accounting standards makes it possible to increase the comparability of financial reports across countries with different economic, political and cultural environments. However, Leuz (2010) finds that there are differences and will be differences in the future between countries’ institutions. Furthermore, he states that enforcements of those institutions differ between countries despite of applying the same regulation (Leuz, 2010). In the end, he concludes that it will be unfeasible to apply the same accounting rules globally.

2.3 Agency theory

2.3.1 Defining the agency theory

According to the previous sections, the agency theory is a suitable theory in this study. Jensen & Meckling (1976, p. 308) define the agency relationship as “a contract under which one or more

(18)

persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision making authority to the agent”. In this case, the relationship would be between shareholders (principals) and managers (the agents).

Because of this delegation of work problems will arise because both the principal and the agent have different interests. Overall, the agency theory is trying to resolve two problems that are possible to occur with regard to the relationship between the principal and agents (Eisenhardt, 1989). The first problem is the agency problem that will occur when there are differences between objectives and goals of principals and agents (Eisenhardt, 1989). Moreover, in this case it is difficult for the principal to follow the exact activities of the agents. So, it is not possible for the principal to determine if the agent act in a manner that is best for the company. Secondly, there is a problem of risk sharing (Eisenhardt, 1989). This problem occurs when there is a difference in the acceptance of risk between the principal and agent. Therefore, there can be a difference in the preference of actions that should be taken by the firm. These different actions can be based on for example, a principal who wants to receive constant dividend for the long-term and, therefore, prefers less risky actions. On the other hand, for an agent this could be reaching their target and receiving their bonus that leads to short-term thinking and more risky actions.

According to Eisenhardt (1989), there are two types of contracting problems with regard to agency risk sharing. This distinction will help to understand the relationship between the principal and agent. The first agency risk sharing problem is adverse selection or hidden-information. “Adverse selection problems arise when the agent has more information than the principal” (Darrough & Stoughton, 1986, p. 501). This arises before an establishment of a contract. The second agency risk sharing problem is moral hazard or hidden-action. “Moral hazard arises when the action undertaken by the agent is unobservable and has a differential value to the agent as compared to the principle” (Darrough & Stoughton, 1986, p. 501).

2.3.2 Solutions to the agency problem

Potential solutions for mitigating those problems are first of all to provide optimal contracts between the business firms and investors (Healy & Palepu, 2001). These contracts need to make sure that firms are disclosing information that gives investors the opportunity to check if managers are taking decisions that are best for the firm. A second solution given for the agency problem is the existence of the board of directors (Healy & Palepu, 2001). This because the task of the board of directors is to monitor the management of a firm on behalf of the shareholders (owners). Lastly, the third solution to the agency problem given by Healy & Palepu (2001) is that information intermediaries are there to produce their own information about their interpretation of the

(19)

performance of the firm. This is another way to be sure that the management is performing well and investors are receiving the right information. However, Healy & Palepu (2001) also state that the effectiveness of previous mentioned solutions are depending on several economic and institutional factors.

2.4 Disclosure theory

2.4.1 Role of disclosure

There are four main reasons given in prior research for the justification of regulating a company’s financial report and their disclosure activities (Leuz, 2010). These four main reasons for the justification are (1) that there will be a creation of externalities after disclosing financial reports, (2) there will be cost savings for companies because of the mandatory disclosures that makes it easier for interested people to find the information they need, (3) firms are willing to disclose voluntarily because they need for example funding, and (4) disclosure can mitigate the agency conflicts by providing more private information to outsiders (Leuz, 2010). Thus, the firms reporting under GAAP have a commitment to greater disclosure that should lower their costs of capital that arise from information asymmetries (Leuz & Verrecchia, 2000). These information asymmetries arise either between the firm and their shareholders, or investors buying or selling shares of the firm (Leuz & Verrecchia, 2000).

Specifically, the FASB defines voluntary disclosures as “disclosures, primarily outside the financial statements, that are not explicitly required by GAAP or a SEC rule” (FASB, 2001). The FASB (2001) explains in a report that the importance of voluntary disclosures will increase because of the changing business environment. Also, they state that the reason of voluntary disclosures for companies is to provide information that investors find informative (FASB, 2001). Furthermore, disclosures with opinions of managers about the company’s performances are seen as important disclosures. Another thing the FASB (2001) states is that the voluntary disclosures should not only be used to disclose positive information, but also bad information. In the end, the FASB (2001) states that they are stimulating companies to keep improving their reporting and they stimulate to use different types of information.

2.4.2 Managers’ motives for voluntary disclosures

According to prior research there are six forces that affect the decisions managers make with regard to disclosures for capital market reasons (Healy & Palepu, 2001). These forces are “capital market transactions, corporate control contests, stock compensation, litigation, proprietary costs, and management talent signalling” (Healy & Palepu, 2001, p. 420). Firstly, the influence of the capital

(20)

market transactions. This force affect the decision of managers’ disclosure because managers can anticipate that they are going to make transactions on the capital market and, therefore, are willing to provide voluntary disclosures for mitigating the information asymmetry (Healy & Palepu, 2001). By providing more information managers can reduce the cost of external financing. Secondly, there is the influence of corporate control contests. This force states that managers are responsible for the stock performance of the company. Because of that, managers are willing to disclose information voluntarily with explanations when the firm is performing badly and to make sure they reduce the possibility of being undervalued by externalities (Healy & Palepu, 2001). Thirdly, there is the influence of stock compensation. When managers are rewarded by receiving stock, the managers are willing to disclose inside information voluntarily to increase the liquidity of the companies’ stock (Healy & Palepu, 2001). Besides, managers disclose also information to make sure that the stock is representing the precise value of the company, so managers do not need to ask for extra compensation because of possible faults in the valuation (Healy & Palepu, 2001).

Fourthly, there is the influence of litigation costs. These costs influence managers’ motives on the one hand, by an increase in voluntary disclosures to prevent disclosures are too late or incomplete but, on the other hand, by a decrease in voluntary disclosures because especially future earnings are volatile information that can deviate easily from expectations. Fifthly, the talent for signalling is another force that affects the decision of a manager. Because the manager is responsible for the performance of the company, he or she is being evaluated on their manner of responding, and how he or she is being aware of environmental changes, investors are more positive about the manager if he or she is showing this in voluntary disclosures and as a result value the company higher. This is a reason for managers to disclose voluntarily. Lastly, the sixth force that affect the decisions of managers’ disclosure are the proprietary costs. This because prior research finds that companies are not willing to disclose information that will reduce their competitive situation on the market (Healy & Palepu, 2001).

Overall, Healy & Palepu (2001) state that managers can decide by themselves to disclose information, therefore, it is not clear whether the voluntary disclosures are credible. However, the credibility of the voluntary disclosures can increase because (1) intermediaries give a guarantee over managements’ disclosures and (2) the quality of the voluntary disclosures can be compared with the audited financial reports (Healy & Palepu, 2001). They conclude, with interference of regulators, standard setters, auditors, and other capital market intermediaries it is possible to increase the quality of those voluntary disclosures (Healy & Palepu, 2001).

(21)

3 Hypothesis development

Based on the mentioned references of prior literature and the theories used in the previous sections, it becomes clear the information availability between managers and investors differ, but also managers’ incentives differ between what information is disclosed. Because of the information asymmetry and agency conflicts between principles and agents, financial reporting exists (Healy & Palepu, 2001). Investors receive information about companies by reading financial reports and by following the companies’ disclosures in press releases. These users are especially interested in the EPS of a company but some users are even more interested in the non-GAAP disclosures (Linsmeier, 2016). There is a difference between ways the information is disclosed because the information in financial reports is audited, however, the information in voluntary disclosures placed in press releases is not. As a result, disclosures can be based on different manager incentives. Information in non-GAAP disclosures can be described using the formula:

𝐺𝐴𝐴𝑃 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠 = 𝑁𝑜𝑛‐ 𝐺𝐴𝐴𝑃 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠 + 𝐸𝑥𝑐𝑙𝑢𝑠𝑖𝑜𝑛𝑠

On the one hand, managers disclose voluntarily more information to reduce the information asymmetry. On the other hand, it is also possible that the managers want to mislead investors by disclosing more positive earnings (Marques, 2006; Curtis et al., 2014).

Because of the increasing number of non-GAAP disclosures it is also in recent years important to think about the quality of the GAAP in general. Of course, this raises questions if the GAAP are still the most efficient and best way of reporting? Even the chairman of the FASB Russel Golden is thinking about the increasing non-GAAP disclosures. He states the FASB can learn from the need for non-GAAP disclosures and apply suited changes in the GAAP that results in reducing the amount of non-GAAP disclosures (Golden, 2017). This is one of the reasons the FASB is continuously monitoring the non-GAAP disclosures. The influence of the implementation of regulation G on the quality of non-GAAP disclosures is researched by for example Heflin & Hsu (2008), Kolev et al. (2008), and Black et al. (2012). Their findings include that after the implementation of regulation G the frequency of ‘special- and other-item’ exclusions reduces (Heflin & Hsu, 2008). Besides this, Heflin & Hsu (2008) find a reduction in the excluded costs and the probability that non-GAAP disclosures are used to meet or beat analysts’ forecasts. Overall, they all conclude that regulation G has a positive impact on the quality of the non-GAAP disclosures (e.g., Heflin & Hsu, 2008; Kolev et al., 2008; Black et al., 2012).

There are two ways of thinking about the usage of non-GAAP disclosures besides GAAP disclosures. Because the usage of non-GAAP disclosures can both be seen as complements and substitutes with regard to financial reports that are in accordance with U.S. GAAP. For example,

(22)

when companies strive for high quality financial reports, thus follow GAAP strictly, it is possible that firms are seeking for more space by using non-GAAP disclosures for their explanation of their firm performance. In this case the non-GAAP disclosures will show too positive earnings that are misleading for investors and thus are of a lower quality (Katz, 2001). Specifically, in this case ‘other item’ exclusions are also excluded in the disclosed non-GAAP earnings. However, these ‘other item’ exclusions made by managers are less clear to explain to investors and, therefore, they can be seen as potential misleading exclusions (Whipple, 2015). On the other hand, it is possible that firms with high quality financial reports are also seeking for high quality non-GAAP disclosures. Now, companies find it important to explain in a more effective way their core earnings to investors (e.g., Heflin & Hsu, 2008). In this case the non-GAAP disclosures can be seen as a substitute next to the financial reports of a company. So, the exclusions made in this situation are called ‘special items’ and these exclusions exist of one-time items (Whipple, 2015). Positive values of the total difference between non-GAAP and GAAP earnings, existing of ‘special items’ and ‘other items’, indicate that by the creation of non-GAAP disclosures the income-decreasing expenses of companies are excluded from the companies’ GAAP disclosures (Kolev et al., 2008). However, investors do not know what incentive managers have had when they decided to disclose more information about the company in non-GAAP disclosures.

Before testing if financial reporting quality affects the quality of non-GAAP disclosures, I want to provide more insight whether companies that are disclosing non-GAAP earnings have high or low financial reporting quality. Therefore, a sub hypothesis is included measuring the likelihood of non-GAAP disclosures and their financial reporting quality. Do companies with high financial reporting quality use more or less non-GAAP disclosures compared to companies with low financial reporting quality? This will be tested with the help of dummy variables. The outcome can be used in the main hypothesis by testing if those non-GAAP disclosing companies are excluding a high or low amount of exclusions by calculating their non-GAAP earnings. The following sub hypothesis is formulated to test if there is a relation between using non-GAAP disclosures and financial reporting quality.

Hypothesis 1sub_{: The usage of non-GAAP disclosures is not associated with financial reporting quality.}

Because of the two different ways of using non-GAAP disclosures, as complements or substitutes, I will formulate a null hypothesis. Up to now, no relation is established between the financial reporting quality and the amount of exclusions made by managers to create their non-GAAP earnings. Despite this, the FASB points out that the amount of non-non-GAAP disclosures can reflect that the quality of GAAP is not effective, moreover, there exists some relation between

(23)

those types of disclosures. Also, the positive impact of regulation G on the amount of total exclusions, to calculate non-GAAP earnings, indicates that when companies are following regulation this has an effect on their quality of non-GAAP disclosures (Heflin & Hsu, 2008; Kolev et al., 2008; Black et al., 2012). All of the above mentioned uncertainties lead to the formulation of the following null hypothesis.

Hypothesis 1: The financial reporting quality of a company is not associated with the quality of non-GAAP

(24)

4 Research methodology

4.1 Overview operationalization of variables

To answer my research question, “Does financial reporting quality affect the quality of non-GAAP disclosures?” this thesis uses archival data from public databases. The data from the databases focus on the United States and covers the period from the year 2006 until the year 2015. This time period is chosen because it is interesting to include most recent data available in databases to determine if there exist a relation. As shown in the Libby boxes in figure 2 the conceptual variables are financial reporting quality and quality of non-GAAP disclosures. The conceptual variable financial reporting quality is operationalized by measuring the accrual quality. This is done with the help of the Dechow & Dichev model (2002) that is designed to measure accrual quality of firms. The independent variable, quality of non-GAAP disclosures, is operationalized by the magnitude of total exclusions made in GAAP earnings of a company. In addition, the total exclusions are split in ‘special item’ exclusions and ‘other item’ exclusions to explain the relation in more detail. Both operational variables are explained in more detail in the next sections. Furthermore, control variables are added to filter out noise in the regression analysis that is conducted.

Figure 2 Libby boxes of research

4.2 Financial reporting quality

The proxy for financial reporting quality is the accruals quality based on the Dechow & Dichev model (2002) (hereafter, DD model), as modified by McNichols (2002) and Francis, LaFond, Olsson, & Schipper (2005). This is a frequently used model in prior research. Biddle, Hilary, & Verdi (2009) describe that this measure is derived from the idea that accruals are the estimate of future cash flows. The DD model establishes a relation between the “current period of working

Explanatory variables

Conceptual Financial reporting quality Quality of non-GAAP disclosures

Dechow & Dichev model (2002) Accruals quality

Exclusion magnitude: Difference between non-GAAP and GAAP EPS

Controls Explained variables

(25)

capital accruals and operating cash flows in the prior, current and future periods” (Francis et al., 2005, p. 296). Because the DD model is limited to current accruals the modified version will be used. In this modified version, accruals quality is also determined by controlling for changes in revenues and level of property, plant and equipment (Francis et al., 2005). According to Dechow & Dichev (2002), the better the link between accruals and cash flows of a company, the higher the quality of the accruals and thus, the less of the model is explained by the residual value. To determine the quality of the financial reporting according to the DD model, data are retrieved from Compustat. Collins & Hribar (2002) provide evidence to use cash flow amounts instead of balance sheet amounts in calculating accruals because balance sheet amounts have a more frequent and higher magnitude of errors when calculating accruals. Therefore, data retrieved for the calculation of accruals are coming from cash flow statements of companies as well.

4.2.1 Operationalization using the Dechow and Dichev model

In particular, the following models will be applied for determining the financial reporting quality: ∆𝑊𝐶𝑖𝑡 = 𝑏0+ 𝑏1𝐶𝐹𝑂𝑖𝑡−1+ 𝑏2𝐶𝐹𝑂𝑖𝑡+ 𝑏3𝐶𝐹𝑂𝑖𝑡+1+ 𝜀𝑖𝑡 (1)

∆𝑊𝐶_𝑖𝑡 = 𝑏₀+ 𝑏₁∆𝑅𝑒𝑣_𝑖𝑡+ 𝑏₂𝑃𝑃𝐸_𝑖𝑡+ 𝜀_𝑖𝑡 (2)

∆𝑊𝐶𝑖𝑡 = 𝑏0+ 𝑏1𝐶𝐹𝑂𝑖𝑡−1+ 𝑏2𝐶𝐹𝑂𝑖𝑡+ 𝑏3𝐶𝐹𝑂𝑖𝑡+1+ 𝑏4∆𝑅𝑒𝑣𝑖𝑡+ 𝑏5𝑃𝑃𝐸𝑖𝑡 + 𝜀𝑖𝑡 (3)

The variable definitions:  ∆𝑊𝐶_𝑖𝑡 = 𝐶𝑎𝑠ℎ 𝑓𝑙𝑜𝑤

𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠=

(𝐼𝐵𝐶−𝑂𝐴𝑁𝐶𝐹+𝐷𝑃𝐶)

𝐴𝑉𝑇𝐴 = Change in working capital for firm i in year t;

o 𝐼𝐵𝐶 = Income before extraordinary items (from cash flow statement) o 𝑂𝐴𝑁𝐶𝐹 =Operating activities net cash flow

o 𝐷𝑃𝐶 = Depreciation taken from cash flow statement o 𝐴𝑉𝑇𝐴 =𝐴𝑇𝑡−1+𝐴𝑇𝑡

2 = Average total assets

o 𝐴𝑇 = Total assets  𝐶𝐹𝑂𝑖𝑡−1 =

𝐶𝑎𝑠ℎ 𝑓𝑟𝑜𝑚 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠_𝑡−1

𝐴𝑉𝑇𝐴 = Cash from operations in year t-1;

 𝐶𝐹𝑂𝑖𝑡 =

𝐶𝑎𝑠ℎ 𝑓𝑟𝑜𝑚 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠_𝑡

𝐴𝑉𝑇𝐴 = Cash from operations in year t;

 𝐶𝐹𝑂_𝑖𝑡+1 = 𝐶𝑎𝑠ℎ 𝑓𝑟𝑜𝑚 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠𝑡+1

𝐴𝑉𝑇𝐴 = Cash from operations in year t+1;

 ∆𝑅𝑒𝑣_𝑖𝑡 =𝑆𝐴𝐿𝐸𝑡−𝑆𝐴𝐿𝐸𝑡−1

𝐴𝑉𝑇𝐴 = Change in revenues deflated by beginning total assets;

 𝑃𝑃𝐸_𝑖𝑡 = 𝑃𝑃𝐸𝐺𝑇

𝐴𝑉𝑇𝐴 = Gross property, plant and equipment deflated by beginning total

(26)

Equation (1) is used to determine the quality of accruals following the original model of the DD model. The modified version of the DD model is shown in Equation (3) and this model will be used to determine the quality of accruals in this thesis. Within Equation (3) the absolute value of the estimation error (𝜀𝑖𝑡), or frequently called the residual, will determine the quality of

the accruals. Just like prior research is indicating, the better the link between accruals and cash flows of a company, the higher the quality of the accruals and thus, the less of the model is explained by the residual (𝜀𝑖𝑡) (Dechow & Dichev, 2002). So, the quality of the accruals will

increase (decrease) when the magnitude of the residual (𝜀𝑖𝑡) decreases (increases) (Dechow &

Dichev, 2002). The modified version of the DD model is chosen because McNichols (2002) finds in her research that the DD model has more explanatory power when the two variables change in revenue, and property, plant and equipment are included. Firstly, data are gathered to conduct the regression analysis in accordance to Equation (3) and to determine the residual (𝜀𝑖𝑡). Secondly,

the standard deviation of the residual (𝜀𝑖𝑡) will be included in the linear regression analysis to

determine if financial reporting quality affects non-GAAP disclosure quality.

4.2.2 Sample description for Dechow and Dichev model

Data are retrieved from Compustat for North American companies for the fiscal years 2006 to 2015. In the end, the sample of this research covers the years 2007 to 2015, but data need to be retrieved including 2006 because these data values are needed to calculate changes in the year 2007. The retrieved data from Compustat exist of the variables shown in table 1 panel A. After downloading the data into Stata, the process of cleaning the data starts. This process contains the next described steps.

Data are gathered for the calculation of the change in working capital. For this calculation the cash flows are divided by the average total assets. The average total assets are used in this calculation because the observations of the change in working capital should be comparable while controlling for the differences in the size of the companies. Because the calculated cash flow needs to be divided by the average total assets, companies need to have a value at all for their total assets and at the same time a value bigger than zero. So, the average total assets and thus the change in working capital will end up with a reasonable and comparable amount. Because of missing values in total assets 23,671 observations are dropped and because of negative total assets 400 observations are dropped. As shown in an overview of the sample selection in table 1 panel B.

(27)

Table 1 Sample selection Dechow and Dichev model (2002)

Second, 16,321 firms without an industry code are dropped, because we use this code called SICH to drop companies with less than ten observations per year within the same industry. This drop of observations can be explained as follows. The SICH code exist of four numbers per company. Within those four numbers the first two numbers explain the kind of industry a company is working in. Hence, after subtracting the last two digits, that represent the specific firm, only the two digits representing the industry code are left. So, within a year there are multiple industries available. Then, it is possible to create a new variable that counts the amount of companies within a specific industry per fiscal year. So, a minimum of ten observations is needed to conduct a precise regression analysis within a group of the same companies within an industry.

Third, all duplicates in the retrieved data are dropped. This duplicate drop is done by looking at the unique Global Company Key (hereafter, gvkey) that Compustat uniquely assigned to each company in the dataset and the fiscal year (hereafter, fyear). This results in a drop of 11,660 observations. After the drop of duplicates it is possible to teach Stata the structure of the data by

Sample selection used in Dechow and Dichev model (2002) Panel A: Input variables used in Dechow and Dichev model (2002)

# Variable name Description

1 AT Total assets

2 CUSIP Committee on Uniform Security Identification Procedures

A nine-digit code uniquely allocated to every North American company

3 DPC Depreciation taken from cash flow statement

4 IBC Income before extraordinary items (from cash flow statement)

5 OANCF Operating activities net cash flow

6 PPEGT Property, plant and equipment deflated by beginning total assets 7 REVT Change in revenues deflated by beginning total assets

8 SALE Sales revenue

9 SICH Standard Industrial Classification - Industry code

Panel B: Selection criteria used to obtain sample of Dechow and Dichev model (2002)

Description n

Compustat data of North American companies during 2006-2015 125,134

Drop if total assets have no value -23,671

Drop if total assets are less than zero -400

Drop if industry code SICH is missing -16,321

Duplicates drop gvkey fyear , force -11,660

Drop if change in working capital is missing -23,060

Drop if cash flow from operations t-1 is missing -253

Drop if cash flow from operations t+1 is missing -8,921

Drop if change in revenues deflated by total assets is missing -1

Drop if property, plant and equipment deflated by beginning total assets is missing -745 Drop if count<10 observations in a specific industry per fiscal year -364

(28)

conducting a tsset test of the gvkey and the fyear. This test results in an unbalanced result because not for every gvkey data are available for the time set of 2006 to 2015.

Fourth, all variables for the outcomes of the beta’s and residual are generated for conducting the regression analysis as shown in Equation (3). After the generations of the variables, all variables are dropped if there is no value available. This results in dropping the following observations because of missing values of 23,060 observations in the change of the working capital, 253 observations in the cash flow from operations in t-1, 8,921 observations in the cash flow from operations in t+1, one observation in the change of sales, and 745 observations in property, plant and equipment deflated by beginning total assets. Hereafter, the drop is done, of the earlier explained observations, with less than ten observations of companies in a specific industry per fiscal year end. The drop contained 364 observations. In the end, the regression of the DD model is conducted with 39,738 observations in 468 specific industry groups per fiscal year.

4.2.3 Sample description for measuring the accrual quality

After conducting the regression analysis for the DD model it is possible to continue with the outcomes and change the residual (𝜀𝑖𝑡) in an absolute value. These outcomes of the DD model

are shown in the results section. Because the proxy for financial reporting quality is measured by the standard deviation of the residual (𝜀𝑖𝑡) of the DD model, this absolute value of the residual

should represent a precise amount. Therefore, like prior research is applying, on a CUSIP-fyear basis a period of three to five years will be used to measure the standard deviation of the residual value of the DD model (Ashbaugh-Skaife et al., 2008; Veenman, 2012). Specifically, the calculated standard deviation of the residual contains at least three years of residual data with a maximum of five years. So, data for at least three years exist of data of the current year plus two years. Moreover, data for at least five years exist of data of the current year plus four years. The drops made of observations because of not meeting the requirement of 3-5 year residual data are shown in table 2. In the end, the accrual quality can be measured for 25,239 observations in 7,749 specific industry groups per fiscal year out of the total of 39,738 observations.

Table 2 Sample selection details accrual quality

Selection criteria used to obtain sample of accrual quality

Description n

Observations for regression analysis DD model 39,738

Observations containing the 3-5 year of residual data 25,239

-14,499 Dropped because standard deviation of residuals for every CUSIP-fyear

(29)

4.3 Non-GAAP disclosure quality

To measure non-GAAP disclosure quality this thesis obtains data from IBES. IBES represents one of the three major analyst-tracking services (Bradshaw & Sloan, 2002). The IBES-reported

Actual EPS (EPS) are used as non-GAAP measure, consistent with Brown (2001), Bradshaw &

Sloan (2002), Brown & Sivakumar (2003), Doyle, Lundholm, & Soliman (2003), Heflin & Hsu (2008), and Doyle et al. (2013). Besides, GAAP EPS (GPS) from IBES are used to measure the GAAP earnings per share. On the other hand, Compustat has also GAAP earnings data available and defines this variable as earnings per share before extraordinary items and discontinued operations, using either basic (Compustat data item epspx) or diluted (data item epsfx) EPS. But, for being consistent both the non-GAAP and GAAP EPS are used from IBES (Bradshaw & Sloan, 2002).

However, a disadvantage of using IBES according to Bhattacharya et al. (2003) and Whipple (2015) is that IBES is in only two-thirds of the time exactly the same as the non-GAAP disclosures. Another disadvantage when using the IBES database for both non-GAAP and GAAP EPS is the disappearing possibility to either categorize the exclusions made as ‘other items’ or ‘special items’ (Whipple, 2015), because IBES excludes all ‘special items’ from their non-GAAP EPS but do not exclude these ‘special items’ from their GPS (Bradshaw & Sloan, 2002). However, when using only IBES for both non-GAAP and GAAP EPS it is possible to determine the total exclusions made by managers. First, non-GAAP quality is measured by the total exclusions made using IBES data and, second, the non-GAAP quality is extended and split in the ‘special item’ exclusions and the ‘other item’ exclusions using a combination of IBES and Compustat data. Positive values of total exclusions indicate that income decreasing expenses were excluded from the GAAP earnings of a company (Kolev et al., 2008).

The difference between IBES and Compustat ‘actuals’ arises because only IBES makes adjustments in the reported earnings in a way they match with the forecast of analysts (Bradshaw & Sloan, 2002). Besides, it is possible to retrieve ‘special item’ exclusions data from Compustat. So, it is possible to split the total exclusions and determine the ‘other item’ exclusions. The advantage given of using databases in general compared to hand-collected data is the ability to end up with a larger sample of GAAP and non-GAAP reporting firms (e.g., Doyle et al., 2013). The focus of the data collection is based on the yearly earnings per share. According to Linsmeier (2016), the users of the financial reports are especially interested in the EPS of a company. Besides, he states that some users of the financial reports are even more interested in the non-GAAP disclosures (Linsmeier, 2016).

Does financial reporting quality affect the quality of non-GAAP disclosures?