2018
Name: Isabel Eising
Student number: 10762469
Thesis supervisor: mw. dr. Réka Felleg Date: June 24, 2018
Word count: 11,023
MSc Accountancy & Control, specialization Accountancy
Faculty of Economics and Business, University of Amsterdam
The influence of corporate governance on non-GAAP
earnings quality after the 2010 CDI update of the SEC
1
Statement of Originality
This document is written by student Isabel Eising 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.
2
Abstract
Non-GAAP earnings are increasingly disclosed by companies while their quality is heavily debated: they can be informative or misleading. Corporate governance functions as a monitor of the company and improved non-GAAP earnings quality before the introduction of Regulation G. Since its introduction, Regulation G acted as a substitute governance mechanism to restrain managers from the disclosure of misleading non-GAAP earnings. However, in 2010, the SEC updated the Compliance and Disclosure Interpretations (CDIs) related to the use of non-GAAP financial measures which signaled increased flexibility for companies to disclose non-GAAP earnings. After this update, GAAP reporting increased substantially. This reflects that non-GAAP reporting became more accessible. Therefore, the monitoring role of corporate governance might influence non-GAAP quality after the CDI update. I use a recent dataset from US companies between 2007 and 2016 to examine the influence of corporate governance on non-GAAP earnings quality after the SEC’s 2010 CDI update. I find that board independence and board size, two out of three corporate governance variables, significantly influence non-GAAP earnings quality after the CDI update. Overall, my results show that corporate governance influences non-GAAP earnings quality after the CDI update.
Keywords: corporate governance, non-GAAP earnings, independent directors, board
3
Content
1. Introduction 4
2. Literature review and hypothesis development 8
2.1 Corporate governance 8
2.2 Non-GAAP earnings 8
2.2.1 Regulation of non-GAAP earnings 10
2.3 The influence of corporate governance on non-GAAP earnings quality 11
3. Methodology 14
3.1 Sample selection 14
3.1.1 Corporate governance 15
3.1.2 Non-GAAP earnings quality 17
3.2 Statistical model 19 3.3 Control variables 20 4. Results 21 4.1 Descriptive statistics 21 4.2 Main analysis 27 4.3 Robustness test 29
5. Discussion and conclusion 31
4
1. Introduction
Firms disclose non-GAAP earnings increasingly, while managers’ motivation for the disclosure of them is heavily debated (Black, Christensen, Ciesielski & Whipple, 2017a; Curtis, McVay & Whipple, 2015). This debate is about the quality of non-GAAP earnings, because they can be disclosed to inform or mislead. Marques (2017) states that the influence of corporate governance on non-GAAP earnings is especially important. Firstly, because non-GAAP earnings can both inform and mislead the users of financial statements and secondly, because these earnings are not audited and therefore might be unreliable. Werder (2011) states that corporate governance is about understanding and improving the way companies are managed and supervised. The corporate governance structure of a company functions as a monitoring device, which positively influences the financial reporting of a company (Ajinkya, Bhojraj & Sengupta, 2005). Therefore, it might influence the quality of non-GAAP earnings.
Non-GAAP earnings disclosures can be informative or misleading for users of financial statements. According to Jennings and Marques (2011), managers compute and disclose non-GAAP earnings for different reasons, of which one is to inform investors with their ‘core’ earnings that are likely to recur in the future. Non-GAAP earnings can be more informative than GAAP-earnings, because irrelevant income statement items are excluded (Jennings & Marques, 2011). However, non-GAAP earnings can mislead investors when they exclude recurring expenses that are included in GAAP earnings. This results in investors that treat these exclusions as transitory, when they may in fact be persistent. The potential that non-GAAP can mislead investors increasingly attracts attention from regulators, practitioners and the financial press, and motivates some regulators to establish more rules about non-GAAP disclosures (Marques, 2017).
Baber, Liang & Zhu (2012) state that weak corporate governance can facilitate managerial entrenchment, known as the process of transferring wealth to self-interested managers at the expense of shareholders’ interests. Non-GAAP measures can be used to determine executive compensation (Jennings & Marques, 2011). If the corporate governance of a company is weak, it is likely that managers who have incentives to provide financial statements users with opportunistic and thus lower quality non-GAAP disclosures in order to receive higher compensation, are not constrained. Jennings & Marques (2011) state that strong corporate governance promotes higher quality financial reporting. Baber et al. (2012) state that strong corporate governance disciplines managers and directors, therefore they are likely to provide higher quality non-GAAP disclosures.
Prior research of Frankel, McVay & Soliman (2011) about the influence of board independence, which is a component of internal corporate governance, indicated a positive association with non-GAAP earnings quality before the introduction of Regulation G. They find
5
that investors can rely on non-GAAP earnings more heavily in companies where independent boards are present, because they constrain opportunistic non-GAAP exclusions. In turn, they find that non-GAAP exclusions are of lower quality when board independence is low. However, this association declined after the introduction of Regulation G, which could be due to Regulation G acting as an alternative governance mechanism for firms with low board independence (Frankel et al., 2011).
Furthermore, other preliminary work from Jennings & Marques (2011) about the joint effects of regulation and corporate governance on non-GAAP earnings shows that prior to Regulation G investors were misled by disclosures of non-GAAP earnings, but only for disclosures made by firms with weaker corporate governance. Their research indicates as well that non-GAAP disclosures are no longer misleading after the introduction of Regulation G, because the regulation acts as an alternative governance mechanism. Jennings & Marques (2011) measured corporate governance with both internal and external governance mechanisms. Frankel et al. (2011) investigated observations between 1998 to 2005 and Jennings & Marques (2011) between 2001 and 2003. Both Frankel et al. (2011) and Jennings & Marques (2011) find that Regulation G substituted for corporate governance in the early 2000’s. However, according to Webber, Nichols, Street & Cereola (2013), non-GAAP reporting increased again after the updated Compliance and Disclosure Interpretations (CDIs) in January 2010. This update signaled substantially increased flexibility to include non-GAAP financial measures in financial statements. Black et al. (2017a) report that the amount of non-GAAP disclosures among S&P 500 firms increased substantially in the last years as well, from 53% in 2009 to 71% in 2014.
Moreover, the SEC’s 2010 update suggests that management should be able to provide investors with ‘appropriate’ non-GAAP financial performance measures, where they are classified as appropriate as long as the measures are accompanied by an adequate disclosure that explains the adjustments and the reasons for the adjustments (Webber et al., 2013). Their research finds that most of the S&P 500 firms that provide non-GAAP disclosures, does indicate why management believes that the disclosure of non-GAAP financial measures is useful for investors, but that their rationales are typically general and not informative. Therefore, corporate governance can step in firstly to deter managers from the disclosure of opportunistic non-GAAP earnings, or instead of other earnings management practices and secondly to ensure that managers disclose informative rationales to provide non-GAAP earnings, which the SEC’s 2010 update suggests.
In addition, non-GAAP earnings are suggested to be a substitute for earnings management (Black, Christensen, Taylor Joo & Schmardebeck, 2017b; Doyle, Jennings & Soliman, 2013). Since corporate governance constrains the tendency of management to manage their earnings (Bekiris &
6
Doukakis, 2011), corporate governance might deter managers from the disclosure of low quality non-GAAP earnings as well. Before the SEC’s update, Regulation G prevented managers from the disclosure of misleading non-GAAP earnings, acting as a substitute governance mechanism. The CDI update signaled more flexibility and non-GAAP reporting increased again. Because the regulation is more flexible since the SEC’s 2010 CDI update and the disclosure of misleading non-GAAP earnings is therefore more accessible, I expect that Regulation G does not substitute for corporate governance anymore.
My thesis provides new evidence about the influence of corporate governance on the quality of non-GAAP earnings after the SEC’s 2010 update. This is important, because the substantial increase in non-GAAP reporting illustrates that non-GAAP reporting has become more accessible for companies. My thesis is distinct from other studies, because it investigates the effect of the CDI update on the association of corporate governance and non-GAAP earnings. This is relevant, because the SEC’s 2010 update signaled increased flexibility that led to an increase in the appearance of non-GAAP earnings, after the decline that Regulation G initiated in the early 2000’s. Therefore, I expect that corporate governance influences non-GAAP earnings quality after the SEC’s 2010 CDI update, by deterring managers from disclosing misleading non-GAAP earnings. However, non-GAAP earnings attract attention from regulators, practitioners and financial press because of their potential to mislead investors (Marques, 2017). It is also possible that the change in regulation in 2010 attracted even more attention to non-GAAP earnings, which could deter managers from the misleading disclosure of them. Furthermore, the CDI update only made the regulation more flexible, but because there is still regulation present, the situation is not the same as prior to Regulation G.
I measure internal corporate governance with board independence, board size and average board age. Based on prior literature, I expect that corporate governance is stronger when board independence and board size are higher and when average board age is lower. I measure non-GAAP earnings quality with the ability of non-GAAP versus non-non-GAAP earnings to predict future performance. I consider non-GAAP earnings as informative if these are better at predicting future performance than GAAP earnings. I run a logistic regression on non-GAAP quality with each of my corporate governance variables on a sample of 15,063 observations of US companies between 2007 and 2016, to analyze the influence of corporate governance before and after the SEC’s 2010 update. My findings can be generalized to a greater extent than that of prior literature about the association between corporate governance and non-GAAP earnings quality, because my sample size is higher.
7
The results imply that non-GAAP earnings quality is positively influenced by board independence and board size, after the SEC’s 2010 CDI update. My findings contribute to prior literature that concludes that corporate governance only influences corporate governance when there is no regulation that acts as an alternative governance mechanism. According to my results, this only holds for board independence. Board size influences non-GAAP earnings quality before the SEC’s 2010 update as well, i.e. in the presence of Regulation G acting as a substitute governance mechanism. In addition, my results contribute to the corporate governance literature, because the influence of board size is different before and after the SEC’s 2010 CDI update. This implies that the influence of board size still remains an open question. In addition, I find no significant results with average board age in my sample, which might be due to the low variation in this measure. This indicates that average board age is not an appropriate measure for corporate governance.
A limitation of my study is that I could not investigate whether there is a difference between the influence of internal and external corporate governance on non-GAAP earnings quality after the 2010 CDI update, because I did not have access to institutional ownership data, which is a measure of external governance. However, prior research states that internal and external governance are substitutes (Baber et al., 2012), hence there might be no difference. Another limitation is that non-GAAP earnings data was limited to the value itself. Information about the types of exclusions, being recurring or non-recurring expenses, is valuable because excluding recurring expenses is misleading for investors and would classify non-GAAP earnings quality as low (Whipple, 2015).
My paper proceeds as follows: in section 2 I review prior literature about corporate governance and non-GAAP earnings, after which I present my hypothesis. I explain my research design, variables and methodology in section 3 and I present the results in section 4. I provide the discussion and conclusion in section 5.
8
2. Literature review and hypothesis development
2.1 Corporate governance
Werder (2011) states that strong corporate governance can mitigate the negative effects of the principal-agent theory by supervising the company. Citing Werder (2011, p. 1346): “Corporate governance designates the legal and factual framework for managing and supervising the corporation.” He states that corporate governance is about the determination of a companies’ overall goal and how this goal is achieved: which structures, processes and persons at the top are selected, how these are evaluated periodically and how adequate transparency of the activities of the company are protected via communication.
Baber et al. (2012) state that specific governance characteristics indicate an independent oversight of the financial reporting process of a company. They state that corporate governance mechanisms are classified as external or internal governance mechanisms. External governance is defined as the role of direct shareholder oversight and represents the corporate control of a company (Baber et al., 2012). External governance either facilitates or discourages active outside stakeholder participation and is known as an effective governance mechanism, because it leads to low or high participation costs for outside stakeholders. Low participation costs indicate strong external governance, because the threat of shareholder intervention disciplines managers and directors. High shareholder participation costs facilitate managerial entrenchment, which is known as the process of transferring wealth to self-interested managers at the expense of shareholders (Baber et al., 2012). The presence of statutory restrictions, imposed by regulation, is an example of an indicator of external governance according to Baber et al. (2012).
Baber et al. (2012) state that internal governance consists of the interactions between insiders of the company and refers to the extent of independence of the board from the management of a company. They describe that this is also known as board independence, which is the ability of the board to monitor, discipline and influence the management of the company. Strong internal governance, thus having an independent board, also discourages managerial entrenchment (Baber et al., 2012). Concluding, corporate governance consists of both internal and external governance and has a supervising and monitoring role that disciplines managers and directors.
2.2 Non-GAAP earnings
Earnings performance is traditionally measured with financial statements numbers produced in accordance with Generally Accepted Accounting Principles (GAAP) (Bradshaw & Sloan, 2002). In the last decades, there has been an increased focus on earnings announced by companies in their
9
press releases, which are not in accordance with GAAP. These alternative earnings are commonly referred to as non-GAAP earnings (Marques, 2017) or ‘Street’ earnings (Bradshaw & Sloan, 2002). Bradshaw & Sloan (2002) state that difference between GAAP and Street earnings increased dramatically in the early 2000’s.
Marques (2017) defines non-GAAP earnings as earnings measures that are not in accordance with GAAP. According to Jennings & Marques (2011), companies have multiple reasons for the disclosure of non-GAAP earnings. It is possible that firms have to meet requirements for loan agreements that are based on non-GAAP earnings. Other companies have CEO compensation agreements with performance measures based on non-GAAP earnings. They state that the most important reason for computing non-GAAP earnings is to inform investors about the recurring or ‘core’ earnings of the firm. Firms execute this by the disclosure of alternative earnings in which GAAP-items that the company regards as nonrecurring are excluded. Excluded items can be for example the recognition of intangible amortization or stock-based compensation expenses, because these GAAP-expenses are sometimes viewed as having no or ambiguous implications for future performance (Leung & Veenman, 2016).
Prior literature (e.g. Marques, 2017; Leung & Veenman, 2016; Frankel et al., 2011) describes that companies can disclose these non-GAAP earnings to inform capital markets or to mislead investors. Frankel et al. (2011) state that these two alternatives, to either inform or mislead, are not mutually exclusive. According to Leung & Veenman (2016), non-GAAP earnings can be informative because they provide investors with information about which items the managers regard as recurring. Whipple (2015) states that non-GAAP earnings can also mislead investors, because they often exclude expenses that are recurring. He shows that 22% of the exclusions of his sample are stock-based compensation and 21% is amortization, while both are recurring expenses.
According to Frankel et al. (2011), another issue of non-GAAP earnings is that the company computes them itself and that they are part of voluntary disclosures. Therefore, non-GAAP earnings numbers are not audited, which makes their appropriateness difficult to verify. Companies’ choices to exclude recurring items in non-GAAP earnings can be opportunistic, because of the fact that this earnings number is not audited (Frankel et al. 2011). They state that managerial entrenchment can be exercised by the disclosure of misleading non-GAAP earnings numbers. One of the reasons why companies disclose non-GAAP earnings is to determine executive compensation (Jennings & Marques, 2011). Therefore, a self-interested manager might want to influence the outcome of his compensation by the disclosure of misleading non-GAAP earnings, while this might be at cost of their financial reporting reliability. Stock-based
10
compensation of managers is another area where managerial entrenchment can be exercised through the disclosure of misleading non-GAAP earnings. This is because non-GAAP earnings became the primary determinant of stock-prices (Bradshaw & Sloan, 2002). Therefore, managers might disclose opportunistic and misleading non-GAAP earnings, in order to receive a higher compensation.
According to Black et al. (2017b), non-GAAP reporting can substitute for accrual or real earnings management. They state that non-GAAP reporting is a relatively costless form of earnings management, unlike accruals earnings management, where the accruals reverse in the next reporting period and hence lead to lower next period’s earnings, or real earnings management, which has real economic consequences. Moreover, non-GAAP reporting does not attract auditors’ attention, since non-GAAP earnings are voluntary disclosures and thus not subject to audit. More specifically, Black et al. (2017b) show that firms use non-GAAP reporting to meet expectations when other types of earnings management are not enough to do so. In addition, they find that companies are more likely to report non-GAAP earnings in an aggressive and misleading way, when they are unable to use other types of earnings management and when their operating performance is poor. Doyle et al. (2013) show that excluding expenses from non-GAAP earnings can substitute for accrual earnings management as well. They show that managers define non-GAAP earnings opportunistically in order to meet or beat analyst expectations and that managers exclude more expenses from non-GAAP earnings when accrual earnings management is costlier. Concluding, non-GAAP earnings became the primary determinant of stock prices in the early 2000’s (Bradshaw & Sloan, 2002), but these earnings can be informative or misleading. Companies disclose misleading non-GAAP earnings if they exclude expenses that are in fact recurring. There is also evidence that non-GAAP reporting is a relatively costless substitute for accrual and real earnings management, which companies use especially when other types of earnings management do not allow the company to meet expectations.
2.2.1 Regulation of non-GAAP earnings
Already in 2001, the Securities and Exchange Commission (SEC) released a cautionary advice regarding the use of earnings measures not in accordance with GAAP (SEC, 2001). They included a warning for investors about the dangers of this additional information. Jennings & Marques (2011) state that the fear emerged that some companies used non-GAAP disclosures to potentially mislead investors and that the United State Congress therefore directed the SEC to develop regulation in the Sarbanes-Oxley Act of 2002, related to the disclosure of manager-adjusted financial measures as GAAP. Their intention was to reduce opportunistic disclosures of
non-11
GAAP earnings. Consequently, the SEC issued Regulation G in January 2003, which requires firms to reconcile their disclosed non-GAAP measures to their most comparable GAAP measures (SEC, 2002). Regulation G was issued to provide more transparency about non-GAAP disclosures, because it requires to explicitly disclose the types of exclusions which enables investors to better determine which adjustments are transitory and which are persistent (Jennings & Marques, 2011).
Heflin & Hsu (2008) show that Regulation G initiated a decrease in both the volume of the non-GAAP disclosures and the opportunistic nature of them, which is consistent with the intentions of legislators and regulators. Frankel et al. (2011) and Jennings & Marques (2011) also show that after Regulation G was introduced, there is no evidence that non-GAAP earnings mislead investors. However, on January 11th 2010, the SEC updated the Compliance and Disclosure
Interpretations (CDIs) related to the use of non-GAAP financial measures (SEC, 2010). This signaled substantially increased flexibility to include non-GAAP financial measures because of the change in tone at the SEC, which caused an increase in non-GAAP disclosures (Webber et al., 2013). Black et al. (2017a) also report that the amount of non-GAAP disclosures among S&P 500 firms increased substantially in the last years, from 53% in 2009 to 71% in 2014, but their research did not discuss the cause.
The SEC’s 2010 CDI update signaled increased flexibility because it suggests that management should be able to provide investors with ‘appropriate’ non-GAAP financial performance measures. Webber et al. (2013) state that non-GAAP financial performance measures are classified as appropriate as long as the measures are accompanied by an adequate disclosure that explains the adjustments and the reason for the adjustments. They find that most of the S&P 500 firms that provide non-GAAP disclosures do indicate why management believes that the disclosure of non-GAAP financial measures is useful for investors, but that managers’ rationales are general and therefore not informative. In sum, after the implementation of Regulation G, the magnitude of non-GAAP earnings disclosures decreased. However, after the SEC’s CDI update in 2010 that signaled increased flexibility, non-GAAP disclosures increased again.
2.3 The influence of corporate governance on non-GAAP earnings quality
Corporate governance has a monitoring role on a company, thus can have impact on the quality of non-GAAP earnings. Ajinkya et al. (2005) state that strong corporate governance positively influences companies’ financial reporting. Corporate governance mechanisms deter managers’ opportunistic behavior, and this is especially important in the case of non-GAAP disclosures, because these are not audited (Baber et al., 2012; Marques, 2017).
12
Non-GAAP reporting can substitute for accruals and real earnings management (Doyle et al., 2013; Black et al., 2017b). Hazarika, Karpoff & Nahata (2012) find that internal governance works proactively to discipline earnings management problems, before they become severe enough to receive public attention. They state that internal governance acts as an important control on managers’ incentives to manipulate earnings. Bekiris & Doukakis (2011) also find that corporate governance constrains the tendency of management to manage their earnings. Therefore, I expect that strong corporate governance also deters the opportunistic and misleading use of non-GAAP earnings as a substitute for other types of earnings management.
Frankel et al. (2011) state that board independence, a proxy for internal corporate governance, positively influences non-GAAP earnings quality because it constrains opportunistic non-GAAP exclusions. However, after the introduction of Regulation G, this association diminished. Jennings & Marques (2011) show that investors were misled by disclosures of non-GAAP earnings prior to Regulation G, but only by firms with weak corporate governance. Their results also show that there is no evidence that non-GAAP earnings are misleading investors after the introduction of Regulation G. These findings illustrate that Regulation G deters managers from the disclosure of opportunistic non-GAAP earnings since its introduction and therefore acts as a substitute corporate governance mechanism. In 2010, the SEC’s CDI update signaled more flexibility for the disclosure of GAAP earnings measures. This resulted in increased non-GAAP reporting, what reflects that it became more accessible (Webber et al., 2013). I expect that the more flexible regulation does no longer substitute for corporate governance and that the disclosure of opportunistic and misleading non-GAAP earnings.
In addition, the SEC’s 2010 update on Regulation G states that management should be able to provide investors with ‘appropriate’ non-GAAP financial performance measures (Webber et al., 2013). Webber et al. (2013) concluded that most of the S&P 500 firms did provide rationales, but that these were not informative. I expect that strong corporate governance can step in to ensure that managers disclose informative rationales to provide non-GAAP earnings, because non-GAAP earnings can be opportunistic and misleading when a company does not disclose informative reasons for the disclosure of them.
In sum, there is extensive evidence that strong corporate governance deters opportunistic behavior of management, because it functions as a monitoring device. Therefore, I expect that strong corporate governance leads to informative GAAP earnings. The disclosure of non-GAAP earnings can be a relatively costless substitute for accruals and real earnings management and corporate governance has the ability to discipline earnings management problems. Hence, it is reasonable that corporate governance has an impact on the quality of non-GAAP earnings as well.
13
Prior research found an association between corporate governance and the quality of non-GAAP reporting, but only prior to the introduction of Regulation G. This suggests that Regulation G restrains opportunistic and misleading non-GAAP disclosures, acting as a substitute for corporate governance. Regulation G initially caused a decrease in non-GAAP reporting, but the SEC’s 2010 CDI update signaled increased flexibility in the reporting of non-GAAP disclosures. As a consequence, non-GAAP reporting increased substantially, which illustrates that it is more accessible for companies to disclose them. Accordingly, I expect that Regulation G no longer acts as a substitute governance mechanism to restrain managers from opportunistic and misleading GAAP reporting and that there is an association between corporate governance and non-GAAP earnings quality after the 2010 CDI update. My above arguments lead to the following hypothesis:
H1: Corporate governance is positively associated with non-GAAP earnings quality after the SEC’s 2010 update.
14
3. Methodology
3.1 Sample selection
My thesis uses archival research to investigate the influence of corporate governance on non-GAAP earnings quality after the SEC’s 2010 CDI update. I use the non-non-GAAP dataset from Bentley, Christensen, Gee & Whipple (2016), which is available from 2003 to 2016, but the data about board characteristics is only available from 2007. Therefore, my sample covers the period from 2007 until 2016. In this way, I can analyze the influence of corporate governance before and after the SEC’s 2010 update. The data is about United States companies.
To test my hypothesis, I require a sample of firms that disclose non-GAAP earnings. The original non-GAAP dataset from Bentley et al. (2016) contains 146,121 quarterly data points of 7,091 companies. First, I drop all the data from before 2007 and 96,781 observations remain. Second, I drop all the firm-quarters in which companies did not report non-GAAP earnings, because these are not relevant in my study. After that 38,961 observations remain. Then, following Leung & Veenman (2016), I retrieve GAAP data from the Compustat Daily Updates - Fundamentals Quarterly, available in WRDS. I merge this dataset with the non-GAAP dataset and remain with 38,375 observations. Following prior research, I drop data from Financial Services firms because the nature of these firms’ non-GAAP disclosures differs from those of nonfinancial firms and these firms are operating under different regulations and settings (Leung & Veenman, 2016; Jennings & Marques, 2011). I do this by dropping SIC codes 6000-6999 and 31,776 observations remain. Then I drop observations for which my control variables are not available, and 31,752 firm-quarter observations remains.
I retrieve data board affiliation data from the ISS director database, available in WRDS, only for firms that are in my combined non-GAAP and GAAP dataset. I calculate the percentage outside directors, board size and average board age from this data, which results in 14,810 annual observations. The ISS director database contains only annual data, but because a board remains broadly constant over the year, I merge these annual board characteristics data to the quarterly observations from my combined non-GAAP and GAAP dataset. Finally, 15,063 quarterly observations of 1,306 companies remain. The sample selection procedures are summarized in Table 1.
15
Table 1 Sample Selection
Description n
Starting observations 2003-2016 146,121
Drop data from before 2007 -49,340
Drop if no non-GAAP reporting -57,820
Drop if no GAAP earnings available -586
Drop financial firms -6,599
Drop if control variables not available -24
Drop if board data not available -16,689
Final sample 15,063
3.1.1 Corporate governance
The first variable that is necessary to conduct this research is data about firm characteristics that reflect their corporate governance quality, the independent variable of my study. As discussed in the literature review, corporate governance consists of both external and internal governance. I measure internal corporate governance quality with board characteristics. Based on prior literature, the board characteristics that I use to measure the corporate governance quality are the percentage outside directors to measure B_IND, the amount of board members to measure B_SIZE and the age of the board members to measure B_AGE (Shaukat & Trojanowski, 2018; Core et al., 1999; Paniagua, Rivelles & Sapena, 2018).
Frankel et al. (2011) state that the SEC and major stock exchanges impose requirements for board independence, which reflects that this is a major aspect of corporate governance. Therefore, this is the most important measure of corporate governance in my study. Internal board members are more loyal to management and therefore, if the CEO is part of the board, he can exert relatively more influence over inside than outside directors (Core et al., 1999). According to Shaukat & Trojanowski (2018), the primary responsibility for board oversight rests with the outside directors, which are called independent non-executive directors on corporate boards. They are expected to perform this oversight function effectively if they are represented adequately in the board, due to their reputational concern in the managerial and directional labor markets. Paniagua et al. (2018) state that outside directors play a crucial role in monitoring the firm’s activities. According to them, the most efficient boards have a large proportion of outside directors relative to inside directors.
16
Regarding board size, Core et al. (1999) state that this is associated with less effective board monitoring, based on the argument that larger boards are less effective and more susceptible to the influence of the CEO. According to Paniagua et al. (2018), an advantage of a smaller board is cohesiveness, because group cohesion has a positive effect on group performance. They state that another advantage of a smaller board is better strategic management, because large boards limit the directors’ ability to initiate strategic interactions. However, there are also studies that posit that a large board is better, especially for larger firms. Ali (2018) states that the complexities increase exponentially when a firm grows. These complexities require a larger board that consists of members that have expertise and knowledge in several operations. Coles, Daniel & Naveen (2008) state that the board’s monitoring role becomes more complex as firms grow, because of diversification, scale and scope of operations, and this requires a larger board. For this study I expect that corporate governance quality increases when board size increases, because my sample consists only of listed firms that provide earnings in accordance with GAAP, i.e. large firms.
The last board characteristic that I use to measure the corporate governance structure is the average age of the board. Core et al. (1999) state that outside directors may become less effective as they grow older and that reform advocates therefore suggest mandatory retirement ages. This is considered important in practice, since 73% of S&P 500 boards have a mandatory retirement age for directors (Hoang, Tygesson, & Ibok, 2016). However, board age is not certain to be an appropriate proxy of corporate governance. This is because 46% percent of the S&P 500 boards have an average age between 60 and 63 (Hoang et al., 2016), which shows it has little variance. Therefore, I consider this measure as the least important measure of corporate governance in my study.
Additionally, Jennings and Marques (2011) state that institutional ownership plays an important oversight role and that it can be used to measure the external governance structure of a company. Unfortunately, I have no access to a database with institutional ownership data and therefore, I cannot investigate the influence of external corporate governance.
To examine if corporate governance influences non-GAAP earnings quality after the CDI update, my three corporate governance variables are measured as a continuous variable. I expect that a company’s corporate governance quality is higher if the percentage outside directors is higher, if the board size is higher and if the average director age is lower. This means that the directions of my corporate governance measures are different. In order to interpret my results better, I multiply the board age variable with -1. In this way, the directions of the results of my corporate governance measures can be interpreted in the same way.
17
3.1.2 Non-GAAP earnings quality
The other variable that is necessary for my research is non-GAAP earnings quality, the dependent variable of my study. As discussed in the literature review, non-GAAP earnings can be informative or misleading. In this study, I consider non-GAAP earnings as informative when they are better at predicting future performance than GAAP earnings. Following prior research (Leung & Veenman, 2016; Black et al. 2017a; Whipple, 2015), I test non-GAAP earnings quality by comparing the ability of GAAP versus non-GAAP earnings to predict future performance (in my study the next period’s GAAP earnings), using the following OLS regression models:
GAAP_EPS2 = β0 + β1 GAAP_EPS + β2 CONTROLS + ε (1)
GAAP_EPS2 = β0 + β1 NG_EPS + β2 EXCL + β3 CONTROLS + ε (2)
Where:
GAAP_EPS2 is the dependent variable of models (1) and (2). It is the Earnings Per Share (EPS) number
(EPSFXQ in Compustat) of next period’s earnings, in accordance with GAAP.
GAAP_EPS are the company’s EPS in accordance with GAAP (EPSFXQ from Compustat).
NG_EPS are the company’s non-GAAP EPS, from Bentley et al. (2016).
EXCL are non-GAAP exclusions, i.e. NG_EPS subtracted with GAAP_EPS.
CONTROLS are the control variables of my study: firm size, firm performance, loss and diversity.
See Table 2 for the description of all the variables of my study. In the second equation, GAAP earnings (GAAP_EPS) are split into non-GAAP earnings (NG_EPS) and exclusions (EXCL). Non-GAAP earnings are higher than GAAP earnings if the company chooses to exclude expenses in their non-GAAP earnings and consequently the exclusions variable takes a negative value (Leung & Veenman, 2016). Before the analysis, the variables are winsorized at the bottom 1st percent and
the top 99th percent levels to mitigate the influence of outliers. The regressions used are robust, to
take care of heteroscedasticity.
I use the regressions (1) and (2) to linearly predict the next period’s GAAP earnings per observation. Hence, I have two variables with the linear predictions of models (1) and (2). The prediction that is closer to the actual next period’s earnings, i.e. deviates less from GAAP_EPS2, is better able to predict future earnings. I determine that current non-GAAP earnings are informative, if the prediction of non-GAAP earnings deviates less from GAAP_EPS2 than the prediction of GAAP-earnings. In order to examine the influence of corporate governance on GAAP earnings quality after the SEC’s 2010 CDI update, I create an indicator variable for non-GAAP quality for each observation. This indicator variable is equal to 0 if current non-GAAP earnings
18
better predict future GAAP earnings and equal to 1 if current non-GAAP earnings better predict future GAAP earnings, i.e. if non-GAAP earnings are informative.
Alternatively, I can subtract the non-GAAP deviation from the GAAP deviation to measure non-GAAP earnings quality. A higher value then indicates higher non-GAAP earnings quality, because the smaller the non-GAAP deviation, the closer the prediction is to the actual next period’s GAAP earnings. However, the magnitude of this value would not reliably measure the distribution of non-GAAP earnings quality. For example, when the GAAP deviation is high (1) and the non-GAAP deviation low (0.1), this means higher non-GAAP earnings quality (0.9) than when the GAAP deviation is less high (0.3) and the non-GAAP deviation as low (non-GAAP earnings quality value of (0.2). Therefore, I prefer to measure non-GAAP earnings quality as an indicator variable. Consequently, my dependent variable takes only two values and therefore I have to conduct a logistic regression to investigate my research question.
Table 2
Description of Variables
Variable Label Proxy
GAAP Earnings q GAAP_EPS Earnings Per Share (EPSFXQ), according to GAAP
Non-GAAP Earnings NG_EPS Non-GAAP Earnings Per Share
Non-GAAP Exclusions EXCL Amount excluded from GAAP earnings
GAAP Earnings q+1 GAAP_EPS2 Earnings Per Share (EPSFXQ) q+1, according to GAAP
Non-GAAP Quality
Corporate Governance variable NG_QUALITY CG Whether non-GAAP earnings better predict GAAP B_IND, B_SIZE and B_AGE q+1
% Independent Directors B_IND Ratio of independent directors to total board members
Board Size B_SIZE Total board members
Average Board Age B_AGE Average age of board members in years
CDI CDI Whether the observation is from after 2010
Firm Size SIZE Natural logarithm of total sales
Firm Performance ROA Return on Assets
Loss LOSS Whether the company made a loss
19
3.2 Statistical model
I examine the influence of corporate governance on non-GAAP earnings quality after the SEC’s 2010 update with the following logistic regression model:
NG_QUALITY = β0 + β1 CG + β2 CDI + β3 CG × CDI + β4 CONTROLS + ε (3)
Where:
NG_QUALITY is the dependent variable of model (3). It is an indicator value that is equal to 1 if
non-GAAP earnings are informative.
CG stands for each of the three corporate governance variables:
B_IND is the ratio of independent directors to total board members.
B_SIZE is the total amount of board members.
B_AGE is the average age of the board members in years.
CDI is an indicator variable that is equal to 1 if the observation is from after 2010.
CG × CDI is the interaction term of CG and CDI and measures CG when CDI is equal to 1.
CONTROLS are the control variables: firm size, firm performance, loss and diversity.
See Table 2 for the description of all the variables of my study. For the logistic regression all continuous variables are winsorized at the top 99th percent and bottom 1st percent levels and the
regressions are robust, to take care of heteroscedasticity. As discussed earlier, I have three different corporate governance measures: B_IND, B_SIZE and B_AGE. Therefore, I run logistic regression model (3) three times.
In order to test my hypothesis, I include an indicator variable for the CDI update. This variable equals 0 if the observation is from before 2010 and equals 1 if the observation is from 2010 or later, because January 11 2010 is the date of the CDI update. My hypothesis is supported if the corporate governance variables are significantly correlated with non-GAAP earnings quality after the SEC’s 2010 CDI update. I test this by interacting the CDI variable with the corporate governance variables. This means that the interaction term has to have a significant positive coefficient to support my hypothesis. However, when I interact a continuous variable (the corporate governance measure) with an indicator variable (the CDI variable), I measure the main effect of the indicator variable when the continuous variable is equal to zero. Therefore, I mean-center the corporate governance variables, after which I run the logistic regression model (3) with these variables.
20
3.3 Control variables
To make sure that the dependent variable ‘non-GAAP earnings quality’ is explained by the independent variable ‘corporate governance’, I control for firm size, firm performance, loss, board diversity and firm. I also control for the effects of industry and fiscal year-quarter. According to Frankel et al. (2011), board independence and firm size tend to be positively correlated. They also state that the costs of managers’ opportunistic behavior increase with firm size, as shareholders are more likely to sue larger firms, which is why firm size might explain the quality of non-GAAP earnings. I measure firm size as the natural logarithm of a company’s total assets.
Black et al. (2017b) state that companies are more likely to disclose misleading non-GAAP earnings when their operating performance is poor. Following Doyle et al. (2013) I control for firm performance, which I measure by the companies’ return on assets. They state that managers disclose opportunistic non-GAAP earnings when GAAP earnings are not sufficient to meet or beat analyst expectations. Following Whipple (2015) and Frankel et al. (2011), I control for loss by including an indicator variable for companies that suffer a loss (which is equal to 1 if GAAP quarterly earnings are negative, and 0 otherwise). This is because these companies may have less persistent earnings and are more likely to disclose non-GAAP earnings (Frankel et al., 2011). The proportion of female directors on audit committees, which consists of board members, improves the quality of financial information (Pucheta-Martínez, Bel-Oms & Olcina-Sempere, 2016). Therefore, it might improve the quality of non-GAAP earnings as well. Board diversity is measured by calculating the percentage female board members divided by the total board members. Based on Leung & Veenman (2016), I control for industry effects, because non-GAAP earnings quality might differ between industries. I follow their method by creating an indicator variable based on the 2-digit SIC code, which results in 58 industries. Following Leung & Veenman (2016), I also control for year-quarter effects and I cluster by firm, because my data has a panel structure.
21
4. Results
4.1 Descriptive statistics
The descriptive statistics for each variable of my sample are presented in Table 3. GAAP earnings have a mean of 0.368 and a standard deviation of 0.821, while non-GAAP earnings have a mean of 0.587 and a standard deviation of 0.535. Compared to Whipple (2015), who also investigate GAAP and non-GAAP earnings based on quarterly earnings per share, but between 2004 and 2012 and only for S&P 500 firms, these variables are on average higher in my study. His study shows a GAAP mean of 0.296 and a non-GAAP mean of 0.348. Consistent with prior literature (Frankel et al., 2011; Whipple, 2015; Black et al., 2017a), non-GAAP earnings are on average higher than GAAP earnings. The mean of exclusions also supports this, -0.226. The average exclusion in the sample of Whipple (2015) was only -0.056, which means that firms in my sample exclude more expenses. However, the average exclusion in the study of Frankel et al. (2011), that was conducted between 1998 and 2005, is -0.180, which is also lower, but more similar to my study. This suggests that firms’ exclusions are higher before regulation G (Frankel et al., 2011), lower in the period after Regulation G (Whipple, 2015) and higher in my sample which consists of years after the SEC’s 2010 CDI update. The standard deviation of the exclusions in my sample is 0.613, compared to a standard deviation of 0.350 of Frankel et al. (2011), which means that the exclusions in my sample have more variation.
The next period’s GAAP earnings have a mean of 0.394, which is higher than the current GAAP earnings, and a standard deviation of 0.774, which is lower than the current GAAP earnings standard deviation. Consistent with Frankel et al. (2011), future GAAP earnings are on average higher. This indicates that firms grow over time and that future GAAP earnings are less dispersed. The dependent variable of my study, non-GAAP earnings quality, has a mean of 0.588. Because this is an indicator variable that equals only 0 or 1, this means that non-GAAP earnings are better able to predict future performance than GAAP earnings, in 58.8% of the observations, i.e. non-GAAP earnings of 58.8% of my observations are informative. This is consistent with Leung and Veenman (2016). In their sample, they found that 58% of non-GAAP earnings have a better predictive ability. The indicator variable CDI has a mean of 0.758 which means that 75.8% of my observations are from the period after the CDI update.
The first corporate governance variable, the percentage independent directors, has a mean of 0.804 and a standard deviation of 0.100. Frankel et al. (2011) show a mean of 0.660 and a standard deviation of 0.172. The sample of Jennings & Marques (2011) show a mean of 0.680 between 2001 and 2003. This suggests that board independence increases over time. Board size, my second corporate governance variable, has a mean of 9.119 and a standard deviation of 2.111.
22
The average board age, the last corporate governance variable, has a mean of 62.278 and a standard deviation of 3.666. This standard deviation is small relative to my other corporate governance measures, which means that there is little variation in board age. I cannot compare these numbers with other studies, because prior research about non-GAAP earnings did investigate these corporate governance measures.
The control variable firm size has a mean of 7.997 and a standard deviation of 1.551, which is similar to 7.085, the average firm size of the sample of Whipple (2015). Firm performance has a mean of 0.010 and a standard deviation of 0.027, which is higher than the sample of Doyle et al. (2013), who show a mean of 0.005 and a standard deviation of 0.040. This means that firms in my sample perform better, but that the variation of their performance is higher. This might be due to his sample ending in 2012, which means that it consists of more years in which there were still consequences of the financial crisis. Loss has a mean of 0.170 and because this is an indicator value, this means that 17% of the firms in my sample suffers a loss. Frankel et al. (2011) show an average value for loss of 0.280, which means that my sample includes fewer loss firms. This can be due to the fact that their sample included only the first years after the financial crisis of the early 2000’s. The last control variable, diversity, has a mean of 0.132 and a standard deviation of 0.103. I cannot compare this number with similar studies, because prior research about non-GAAP earnings did not control for diversity.
Table 3
Descriptive Statistics
Variable N Mean SD Min Max
GAAP Earnings q 15,063 0.368 0.821 -3.890 2.860 Non-GAAP Earnings 15,063 0.587 0.535 -0.480 2.680 Non-GAAP Exclusions 15,063 -0.226 0.613 -4.290 0.950 GAAP Earnings q+1 15,063 0.394 0.774 -3.440 2.790 Non-GAAP Quality 15,063 0.588 0.492 0.000 1.000 CDI 15,063 0.758 0.428 0.000 1.000
Average Board Age 15,063 62.278 3.666 47.667 77.833
Board Size 15,063 9.119 2.111 4.000 20.000 % Independent Directors 15,063 0.804 0.100 0.143 1.000 Firm Size 15,063 7.997 1.551 4.050 13.511 Firm Performance 15,063 0.010 0.027 -0.134 0.076 Loss 15,063 0.170 0.376 0.000 1.000 Diversity 15,063 0.132 0.103 0.000 0.667
For the description of variables, refer to table 2.
Table 4 presents the descriptive statistics of the differences of the variables before and after the CDI update. The mean of GAAP earnings has increased significantly, from 0.192 to 0.424, possibly
23
explained by the financial crisis in the years before the CDI. Likewise, non-GAAP earnings and next period’s GAAP earnings are also significantly higher, the means increased from 0.445 to 0.632 and from 0.248 to 0.441. The mean of the exclusions decreased significantly after the CDI update, from -0.257 to -0.216, which means that firms on average exclude less expenses after the CDI update. The mean of non-GAAP earnings quality increased significantly from 0.577 to 0.597, what means that after the CDI update, there are more observations where non-GAAP earnings are better able to predict future performance than GAAP earnings. This suggests that non-GAAP earnings quality increased after the SEC’s 2010 CDI update, contrary to my expectation.
My corporate governance variables are all significantly higher after the CDI update. The percentage independent directors increased from 0.791 to 0.808. This suggests that board independence increases over time. Board size increased from 9.036 to 9.145. Based on these measures, corporate governance quality became stronger. Board age increased from 61.306 to 62.588, therefore corporate governance quality decreased after the CDI update based on this measure. However, it makes sense that board age increases over time, because a board remains broadly constant over the year and board members get older each year.
Table 4
Comparison Descriptive Statistics Before and After the CDI Update
N=3,643 N=11,420 Difference
of Mean
Before January 11, 2010 After January 11, 2010
Variable Mean SD Mean SD t-test
GAAP Earnings q 0.192 0.857 0.424 0.801 -14.96***
Non-GAAP Earnings 0.445 0.431 0.632 0.556 -18.99***
Non-GAAP Exclusions -0.257 0.711 -0.216 0.578 -3.52***
GAAP Earnings q+1 0.248 0.743 0.441 0.777 -13.19***
Non-GAAP Quality 0.577 0.494 0.597 0.491 -2.14**
Average Board Age 61.306 3.834 62.588 3.556 -18.59***
Board Size 9.036 2.149 9.145 2.100 -2.71*** % Independent Directors 0.791 0.103 0.808 0.098 -9.00*** Firm Size 7.820 1.525 8.054 1.555 -7.95*** Firm Performance 0.006 0.033 0.011 0.024 -9.93*** Loss 0.216 0.411 0.156 0.363 8.41*** Diversity 0.113 0.098 0.138 0.104 -12.81***
a Student t-test, significance is based on two-tailed p-values. *** and ** indicates statistical significance at the 1% and 5%
level respectively.
24
The increase in non-GAAP earnings quality can be interpreted in different ways. I expect that the CDI update made it easier for companies to disclose misleading non-GAAP earnings, because of the relaxation of the regulation. Therefore, I expect non-GAAP earnings quality to be lower after the CDI update. But if the hypothesized monitoring influence of corporate
governance is present, this could accordingly result in higher non-GAAP earnings quality, because corporate governance disciplines managers to disclose high quality non-GAAP earnings. Alternatively, the higher non-GAAP earnings quality after the CDI update could be explained by the fact that GAAP earnings are significantly higher after the CDI. This suggests that firms are more likely to disclose low quality non-GAAP earnings if their GAAP earnings are lower, which is consistent with Black et al. (2017a). Another interpretation could be that the change in
regulation attracted more attention to non-GAAP earnings and that firms are thus deterred from the disclosure of misleading non-GAAP earnings.
Table 5
Frequency of non-GAAP reporting per year
Year Frequency Percentage Increase (in %)
2007 845 5,83 2008 1,277 8,81 51,12 2009 1,521 10,49 19,11 2010 1,514 10,44 -0,46 2011 1,573 10,85 3,90 2012 1,759 12,13 11,82 2013 1,937 13,36 10,12 2014 1,988 13,71 2,63 2015 2,086 14,39 4,93 Total 14,500 100,00
2016 is not included because the dataset ends in July 2016.
The frequency of non-GAAP reporting per year is displayed in Table 5. Consistent with prior research (Black et al., 2017; Whipple, 2015), my sample shows an increase in non-GAAP reporting, from 845 observations in 2007 to 2,086 observations in 2015. Non-GAAP reporting increased in each year, except for 2010, the year in which the CDI update took place. This could be explained by the fact that the change in regulation had a deterring effect, possibly due to increased attention on non-GAAP earnings. However, in 2011, the number of firms that disclose non-GAAP earnings increased again. It could be that firms were deterred by the increased attention on non-GAAP earnings after the CDI update in the first place, but that they afterwards realized that the regulation became more flexible.
25
Table 6 displays the Pearson correlation matrix between the variables in my sample. Non-GAAP earnings (0.644) and exclusions (0.731) are positively correlated with current Non-GAAP earnings. This is not surprising, since non-GAAP earnings consist of current GAAP earnings subtracted with exclusions. Future GAAP earnings (0.521) are also positively correlated with current GAAP earnings. Firm performance is also positvely correlated with current GAAP earnings (0.772). It makes sense that these variables are highly correlated, since firm performance is measured with net income divided by total assets. Loss is negatively correlated with GAAP earnings and because loss equals 1 if a company’s GAAP earnings are negative, this makes sense.
Future GAAP earnings (0.540) are positively correlated with non-GAAP earnings, which means that the ability to predict future earnings is higher than that of current GAAP earnings. This suggests that non-GAAP earnings are more informative than GAAP earnings and confirms the descriptive statistics of non-GAAP earnings quality variable in Table 3. Firm size (0.399) and firm performance (0.398) are also positively correlated with non-GAAP earnings. This means that non-GAAP earnings increase when firm size increases and when performance increases. Loss is negatively correlated with non-GAAP earnings (-0.375), which means that firms that suffer a loss disclose higher non-GAAP earnings. The latter two correlations are contradictory, because non-GAAP earnings increase with both firm performance and loss, while these should be opposites.
Firm performance (0.666) and loss (-0.471) are significantly correlated with exclusions, which is also contradictory, because it means that exclusions increase when firm performance increases and when loss decreases. Firm performance (0.353) and loss (-0.334) are significantly correlated with future GAAP earnings. This relationship is logical, because it means that future GAAP earnings increase when firm performance increases, and that future GAAP earnings decrease when a firm suffers a loss.
Consistent with prior literature (Boone, Karpoff & Raheja, 2007), board size is positively correlated with firm size (0.637), which means that board size increases as firms grow. Diversity (0.374) is positively correlated with board size, which means that when the proportion of women on boards increases, board size also increases. Diversity is also positvely correlated with firm size (0.348). Board size increases with firm size, therefore it makes sense that firm size also increases with diversity. Loss is negatively correlated with firm performance (-0.650), which means that firm performance decreases when a suffers a loss. This is logical because firm performance is a formula of net income divided by total assets.
Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (1) GAAP Earnings q 1.000 (2) Non-GAAP Earnings 0.644*** 1.000 (3) Non-GAAP Exclusions 0.731*** -0.031*** 1.000 (4) GAAP Earnings q+1 0.521*** 0.540*** 0.211*** 1.000 (5) Non-GAAP Quality 0.001 0.012 -0.015* 0.029*** 1.000
(6) Average Board Age 0.080*** 0.140*** -0.020** 0.089*** -0.011 1.000
(7) Board Size 0.164*** 0.266*** -0.014* 0.181*** -0.014* 0.135*** 1.000 (8) % Independent Directors 0.072*** 0.125*** -0.011 0.077*** -0.008 0.038*** 0.207*** 1.000 (9) CDI 0.121*** 0.150*** 0.029*** 0.107*** 0.017** 0.150*** 0.022*** 0.075*** 1.000 (10) Firm Size 0.222*** 0.399*** -0.055*** 0.218*** -0.020* 0.172*** 0.637*** 0.215*** 0.064*** 1.000 (11) Firm Performance 0.772*** 0.398*** 0.666*** 0.353*** -0.001 0.014* 0.065*** 0.002 0.089*** 0.086*** 1.000 (12) Loss -0.607*** -0.375*** -0.471*** -0.334*** 0.003 -0.028*** -0.100*** -0.028*** -0.068*** -0.134*** -0.650*** 1.000 (13) Diversity 0.110*** 0.166*** 0.005 0.118*** -0.023*** -0.045*** 0.374*** 0.210*** 0.348*** 0.348*** 0.060*** -0.064*** 1.000
a ***, ** and * indicate a 1%, 5% and 10% level of statistical significance respectively.
b
Table 7 Logistic regression results based on the following model:
NG_QUALITY = β0 + β1 CG + β2 CDI + β3 CG × CDI + β4 CONTROLS + ε (3)
B_IND B_SIZE B_AGE
NG_QUALITY (1) (2) (3) (4) (5) (6)
Odds Ratio Coefficient Odds Ratio Coefficient Odds Ratio Coefficient
CDI 0.885 -0.122 0.882 -0.126 0.868 -0.142 (0.589) (0.589) (0.579) (0.579) (0.530) (0.530) B_IND 0.571 -0.560 (0.129) (0.129) B_IND × CDI 2.177* 0.778* (0.086) (0.086) B_SIZE 0.965* -0.036* (0.086) (0.086) B_SIZE × CDI 1.045** 0.044** (0.032) (0.032) B_AGE 0.988 -0.012 (0.257) (0.257) B_AGE × CDI 1.017 0.016 (0.187) (0.187) SIZE 1.033 0.033 1.034 0.034 1.032 0.032 (0.108) (0.108) (0.148) (0.148) (0.114) (0.114) ROA 0.207 -1.576 0.220 -1.514 0.212 -1.550 (0.189) (0.189) (0.207) (0.207) (0.199) (0.199) LOSS 1.062 0.060 1.064 0.062 1.065 0.063 (0.426) (0.426) (0.408) (0.408) (0.408) (0.408) DIV 0.986 -0.014 1.016 0.015 0.979 -0.021 (0.958) (0.958) (0.952) (0.952) (0.936) (0.936)
INDUSTRY YES YES YES YES YES YES
YEAR QTR YES YES YES YES YES YES
Observations 15,059 15,059 15,059 15,059 15,059 15,059
Pseudo R2 0.028 0.028 0.028 0.028 0.028 0.028
a For the description of variables, refer to table 2. b Board characteristics variables are mean centered.
c In order to better interpret the results, board age is multiplied by -1 before I run the regression. This is because
corporate governance quality decreases when board age increases.
d The bold values in the table are the interaction variables, these are the values that answer my research question. e ***, ** and * indicate a 1%, 5% and 10% level of statistical significance respectively.
Table 7 displays the results of the logistic regressions of the influence of corporate governance on non-GAAP earnings quality after the CDI update. The interaction terms measure the effect of corporate governance after the CDI, i.e. when the CDI variable is equal to 1. For my hypothesis to be supported, the interaction terms of the corporate governance variables with the CDI variable
28
have to be significant. The logistic regression outcomes are only available in probabilities. Therefore, it is not possible to discuss the economic significance of my results.
Columns 1 & 2 show that the interaction variable of CDI and board independence is significant at a 10% level (p = 0.086). The odds ratio is 2.177 and the coefficient is 0.778. To be precise, the odds ratio equals 2.177 to 1. This is interpreted as a 2.177 change in the odds ratio when there is a one-unit change in the interaction variable. This number is positive and higher than 1 and indicates a positive influence (odds ratios between 0 and 1 indicate a negative influence). In other words, after the CDI update in 2010, the ratio of the odds that non-GAAP earnings are informative increase. Regarding the coefficient, it means that for a one-unit increase in the interaction variable, thus an observation after 2010 and an increase board independence, an increase of 0.778 in the log-odds of the dependent variable non-GAAP earnings quality is expected, holding all other independent variables constant. This implies that after the SEC’s 2010 CDI update, non-GAAP earnings quality is higher when board independence is higher.
Column 3 & 4 show that the interaction term of CDI and board size is significant at a 5% level (p = 0.032). The odds ratio is 1.045 and the coefficient 0.044. This is interpreted as a 1.045 change in the odds ratio when there is a one-unit change in the interaction variable. This means that after the CDI update, the ratio of the odds that non-GAAP earnings are informative slightly increase. As for the coefficient, for a one-unit increase in the interaction term, thus an observation after 2010 and an increase in board size, a 0.044 increase in the log-odds of non-GAAP earnings quality is expected, holding all other independent variables constant. This result shows that after the SEC’s 2010 CDI update, higher board size has a positive effect on non-GAAP earnings quality.
However, board size is also significant before the CDI update, at a 10% level (p = 0.086). The odds ratio is 0.965 (to 1) and the coefficient is -0.036. The odds ratio is interpreted as a 0.965 change in the odds ratio when there is a one-unit change in the board size variable, therefore the ratio of the odds that non-GAAP earnings are informative decrease. Regarding the coefficient, it means that for a one-unit increase in board size, a decrease of 0.036 in the log-odds of the dependent variable non-GAAP earnings quality is expected, holding all other independent variables constant. This shows that before the CDI update, non-GAAP earnings quality decreases when board size is higher. However, this is not relevant to answer my research question, because this is the effect of board size before the CDI update. The effect of board size has changed during the years of my sample, because before the CDI update, higher board size negatively influences GAAP earnings quality, but after the CDI update, a higher board size positively influences non-GAAP earnings quality. This could be due to the significantly different means of all the variables of my sample, for which I refer to Table 3.
29
The interaction variable of CDI and board age appears to be not significant, shown in columns 5 & 6. This could be due to the fact that it is not an appropriate measurement construct for corporate governance quality. The relatively small standard deviation, as displayed in Table 3, can be the reason that there is no significant association. The standard deviation is small compared to the standard deviations of the other corporate governance measures, which confirms prior findings of Hoang et al. (2016). This means that there is low variation in board age, which might be the reason that there is no significant association. This finding could be the reason that board age is not an appropriate measure for corporate governance in the first place.
Not one of the control variables appears to be significant. This indicates that firm size, firm performance, loss and diversity do not influence non-GAAP earnings quality according to my results. The pseudo R2 of the logistic regression is 0.028 for each of the corporate governance
variables. However, this is not the proportion of variance explained by the predictors and therefore it is not possible to interpret this number in the same way as R2 of OLS regression.
According to my expectations, I find that board independence and board size influence non-GAAP earnings quality after the CDI according to my expectations. However, board age does not. Given these results, I can confirm my hypothesis that corporate governance influences non-GAAP earnings quality after the SEC’s 2010 update, because I considered board independence as the most important measure of corporate governance and board age as the least important.
4.3 Robustness test
As discussed in the literature review, firms can disclose misleading non-GAAP earnings instead of other types of earnings management (Black et al. 2017b; Doyle et al. 2013). According to prior research, the disclosure of non-GAAP earnings allows managers to meet expectations, when other types of earnings management are costlier or not an option.
Das, Shroff & Zhang (2009) investigate if quarterly earnings patterns indicate earnings management and find that reversals of year to date earnings patterns in the fourth quarter occur more frequently than expected when earnings reversals occurred purely by chance. Since the disclosure of misleading non-GAAP earnings can be a substitute for earnings management, firms may have incentives to disclose more misleading non-GAAP earnings in the fourth fiscal quarter. This is because the fourth fiscal quarter is the last quarter of the fiscal year and firms have to close their accounts. As a robustness test, I examine whether the effects of the fourth fiscal quarter are stronger than the effects of the other three quarters. More specifically, I expect that the effects of the interaction terms of the corporate governance variables with the CDI variable on non-GAAP quality are stronger in the fourth fiscal quarter than in the other quarters.
30
Untabulated results show that there are no significant differences of the interaction terms of the corporate governance variables and the CDI between the fourth quarter and the other quarters. This means that there is no stronger effect in the fourth fiscal quarter for the influence of corporate governance on non-GAAP earnings quality.
