The Impact of Media Attention on the Use of Alternative Earnings Measures
We are grateful for the helpful comments from Graeme Dean (the then editor), an anonymous referee, Abe de Jong and the participants at the European Accounting Association Conference 2007.
Abstract
The practice of reporting earnings measures that deviate from generally accepted accounting principles (non-GAAP measures) has received negative attention in the media. In a period of increased regulatory concern for these reporting practices, we explore whether there has been a shift away from the use of non-GAAP metrics. This study focuses on the Dutch situation, where regulators responded conservatively (‘light’) to the accounting scandals. This contrasts with the U.S., where regulators intervened with a radical (‘heavy’) reform of regulation. We analyse a sample of earnings press releases published in the period 2000–05 from companies listed at Euronext Amsterdam. Our findings indicate that Dutch companies report non-GAAP measures frequently and prominently. However, companies' reporting behaviour changes after a peak in negative media attention for non-GAAP reporting. The magnitude of the adjustments to GAAP earnings becomes smaller and companies seem to have different reasons to report non-GAAP measures. The effect of the media attention is stronger when companies have been criticized for their non-GAAP reporting in the press. Investors seem to have become more hesitant towards the use of non-GAAP measures for their decision-making after negative media attention. Together, these findings suggest that the negative media attention for non-GAAP measures has influenced the decisions of investors and managers.
One of the controversial areas in financial reporting that has received considerable media attention is the disclosure of self-constructed earnings measures. Regulators such as financial market authorities and accounting standard setters repeatedly expressed their worries about the possibly misleading use of financial information that does not comply with generally accepted accounting principles (GAAP), that is, non-GAAP measures. In the U.S., the Sarbanes-Oxley Act (SOX) of 2002 and pursuant SEC regulation have addressed the practice of non-GAAP reporting, allowing it only under strict conditions. Several U.S. studies report that non-GAAP regulation seems to be effective, in the sense that the reporting practices became less opportunistic (e.g., Heflin and Hsu, 2008). Additionally, investors act as if non-GAAP information is more informative after the SEC regulation (Marques, 2006). Still, the reporting environment changed in various ways, regulatory changes being only one factor.
This study focuses on the influence of media attention on both firms' reporting behaviour and investors' response to the reported information. It is argued that media attention directed at allegedly misleading non-GAAP information has increased awareness of the use or potential abuse of these alternative earnings measures and has had a substantial influence on the behaviour of companies and investors. To study the impact of the media in the U.S. is difficult, since all listed companies are affected post-Enron by substantial new regulation. This would create the problem of disentangling the effect of regulation from the effect of media attention. We circumvent this problem by exploring non-GAAP reporting in the setting of the Dutch financial market. The practice of non-GAAP reporting also induced a sharp debate in the media in the Netherlands1; however, in this case regulators and policy makers did not respond with additional regulation similar to SOX in the U.S. Hence, the Netherlands allows study of reporting practices and, more specifically, the use of non-GAAP earnings measures of publicly listed companies in an environment of changing public opinions and negative media attention.
The contribution to the literature is threefold. First and most important is the contribution to the debate concerning the effectiveness of regulation in general and SOX more specifically. SOX has been criticized for being a hasty overreaction to corporate scandals, imposing substantial costs on companies without compelling evidence that this would create economic benefits (Romano, 2005). This assumes that financial regulation should be based on scientific evidence, an argument that was made more explicitly by Buijink (2006). Recent papers have tried to fill this gap and investigate the effect of SOX by comparing financial reporting before and after the Act became effective (Marques, 2006; Bartov and Cohen, 2007; Cohen, et al., 2008; Kolev, et al., 2008). However, these studies inevitably suffer from the problem that the effect of SOX cannot be isolated and that other factors may have caused the observed changes in financial reporting (Coates, 2007; Leuz, 2007). Our paper employs a different institutional setting—that is, the Netherlands—to provide evidence on how companies and investors change their behaviour in the absence of such a regulatory shock.
Second, the study adds to the growing literature on the effects of media on financial markets. While other studies try to infer the effect of company-specific media coverage on stock prices (Dyck and Zingales, 2003) or on corporate governance characteristics (Dyck, et al., 2008; Joe, et al., 2009), we focus on the effect of media coverage on financial reporting practices.
Third, we provide additional evidence on the use of alternative earnings measures. Recent studies have examined the use of non-GAAP measures as reported in earnings releases in the United States (Bhattacharya et al., 2003; Lougee and Marquardt, 2004; Bowen et al., 2005). A vast majority of the international accounting literature stresses the importance of institutional factors and market forces in shaping management's incentives to report informative earnings measures (Ball et al., 2003). To our knowledge, no research has been conducted on the use of alternative earnings measures outside the United States and Canada.
Evidence is reported here that the practice of reporting non-GAAP earnings measures in earnings press releases published during 2000–April 2005 is popular in the less regulated reporting environment of the Netherlands. First, analysis of the financial reporting environment during the sample period reveals a dynamic environment that leads to increased negative attention from the media for non-GAAP reporting. We then measure the popularity of non-GAAP reporting in terms of reporting frequency and prominence. The reporting behaviour is found to change after a peak in negative media attention. Although non-GAAP earnings measures are reported frequently after a peak in negative media attention, the motivation to report these measures seems to have changed. The majority of firms that reported non-GAAP earnings before 2003 stop thereafter. The companies that initiate or continue to report non-GAAP in 2003 and after are more strongly motivated by opportunistic motivations (loss avoidance or meeting analyst's forecasts). In other words, companies stop reporting non-GAAP measures after negative media attention unless the expected benefits of non-GAAP reporting are higher. On the other hand, the difference between reported non-GAAP and GAAP earnings is smaller in 2003 and after. From these results one can infer that companies' reporting behaviour is affected by media pressure, even without regulatory intervention. The effect is stronger for companies whose non-GAAP reporting practices were specifically criticized in the press.
Next the focus shifts to the investors' use of non-GAAP information as a basis for their decisions. The association between abnormal returns and non-GAAP earnings as well as GAAP earnings is investigated. We identify the development of the informativeness over time and find that the appreciation of non-GAAP information has changed. Prior to 2003 investors seemed to price non-GAAP earnings rather than operating earnings as defined under GAAP. Starting in 2003, the situation changed, as investors seemed to turn away from non-GAAP measures and start to price GAAP operating earnings instead. GAAP bottom-line earnings are informative throughout the entire period. Overall, in the period from January 2003 to June 2005 investors seemed to turn away from non-GAAP information after a peak in negative media attention.
The empirical results support the conjecture that the use of financial reporting information can change in the absence of any regulatory intervention. Investors' perception of specific reporting practices is shown to change after the public debate, and more specific, media attention. Arguably, investors that become more aware of the possibly negative aspects of certain non-GAAP earnings measures ignore these disclosures in their decision making. At the same time, media attention seems to affect firms' reporting behaviour, in particular those companies that were the subject of a critical newspaper article. Although the overall frequency and prominence with which non-GAAP measures are reported increases, companies seem to have different reasons to report non-GAAP earnings after a period of increased negative media attention. It seems investors are careful in the absence of regulation and ignore reported information that may not be reliable. Regulation may not be necessary to protect investors against confusing or misleading reporting practices when investors are aware. Nevertheless, regulation may restore the credibility, at least to some extent, of reporting practices that were publicly criticized, as some U.S. findings for non-GAAP reporting seem to suggest (e.g., Marques, 2006; Kolev et al., 2008). For companies, cost-benefit comparisons of reporting non-GAAP measures are different in a regulated or non-regulated situation. These findings are particularly relevant for standard setters, policy makers and financial market participants, since the debate on optimal level of regulation is on a new high.
LITERATURE REVIEW
The Effectiveness of Regulation
The effectiveness and desirability of regulation of financial markets has long been debated (e.g., Stigler, 1964; Benston, 1973). In the aftermath of recent corporate scandals, regulators imposed far-reaching regulation on financial markets to restore investors' trust. These new regulations, such as the Sarbanes-Oxley law of 2002, evoked the discussion on the effectiveness of regulation and led to several academic studies discussing the costs and benefits of SOX (e.g., Romano, 2005; Coates, 2007; Leuz, 2007).
The effects of SEC interventions concerning non-GAAP reporting have been investigated in a number of studies. Looking at firms' reporting behaviour, Marques (2006) and Heflin and Hsu (2008) find that non-GAAP reporting decreases significantly after SEC intervention. Entwistle et al. (2006) also find that the number of firms reporting non-GAAP information after the introduction of SEC regulation declines sharply. They also report that the reported non-GAAP information is less biased and presented less prominently after the SEC regulation. Kolev et al. (2008) find that after SEC intervention the predictive ability of the exclusions from GAAP earnings improves. Companies that stop reporting non-GAAP information had significantly lower quality exclusions in the period before SEC intervention, which suggests that SEC intervention caused the more opportunistic non-GAAP reporters to stop.
These results seem to suggest that SEC regulation was effective in the sense that non-GAAP reporting became less opportunistic. On the other hand, the decreasing frequency of non-GAAP reporting may also imply that some informative reporters stopped publishing non-GAAP measures (Marques, 2006; Heflin and Hsu, 2008). Similarly, Kolev et al. (2008) report evidence of unintended negative consequences. When exclusions are specified and split into special items and other exclusions, they find that the quality of reported special items deteriorates after SOX. They interpret their results as evidence that managers adapt to the stricter regulation by replacing opportunistic non-GAAP reporting with accounts manipulation2 through the use of special items.
Besides companies' reporting behaviour, investors' reaction to non-GAAP information has been analysed as well. Investors' response to non-GAAP measures has been compared to the response to GAAP earnings in several studies. Based on press releases, investors were revealed to find non-GAAP earnings more informative, more persistent and cause stronger revisions to analysts' beliefs than GAAP earnings (Bhattacharya et al., 2003; Lougee and Marquard, 2004). This evidence suggests that the ‘flexible’ non-GAAP reporting leads to information that investors find useful.
A number of studies have analysed investors' reactions to non-GAAP information before and after the implementation of SOX and Regulation G. For example, Heflin and Hsu (2008) find a decline in the reaction of investors to forecast errors after the implementation of Regulation G. On the other hand, Marques (2006) reports that investors react more positively to non-GAAP disclosures after the regulation became effective. She suggests the findings may be explained by SEC regulation increasing the credibility of non-GAAP disclosures.
Taken together, the U.S. evidence suggests that SEC regulation has influenced the use of non-GAAP disclosures, but that there may also be unintended consequences that effectively decrease the quality of reported financial information. Moreover, studies that examine the effectiveness of regulation face the problem of contemporaneous changes in the reporting environment. Our research design addresses precisely that issue, asking what would happen in the absence of additional regulation. First the reporting environment is analysed, identifying media attention for non-GAAP reporting as a changing characteristic that has likely influenced companies and investors.
Media Attention
Negative media attention for non-GAAP disclosures is assumed to have changed the reporting environment. This section of the paper deals with the theoretical underpinning of the effect of media attention on regulators, companies and investors.
Intuitively media attention is likely to influence financial market institutions and regulators. This notion is supported by the institutional crisis literature in political science. Research in political science strongly suggests that policy makers are more responsive to public pressure when the issue is more salient (e.g., Page and Shapiro, 1983; Monroe, 1998; Burstein, 2003). The financial scandals, the backdrop of the non-GAAP reporting discussion, attracted substantial media attention and yielded several salient stories. Public pressure sustained by mediaattention can ‘force’ institutions to respond in order to restore confidence (e.g., Boin and 't Hart, 2000; Lodge and Hood, 2002). According to the literature, regulators can respond to institutional crises by radical departures from the status quo or by a more conservative adaptive strategy. In the U.S., the SEC responded with a radical reform of the financial regulations. In the Netherlands, institutions tried to reinforce the existing regulations by pointing out the importance of the existing regulation and the definition of net income. Framed in this literature, the U.S. and the Netherlands represent two alternative responses to public pressure sustained by media attention.
Other studies show that companies adjust their behaviour in reaction to media attention. For example, Dyck et al. (2008) find that press coverage increases the probability of companies taking action to improve corporate governance. Joe et al. (2009) explore the impact of Business Week's publication of the worst corporate board officers and find that companies are inclined to enhance their corporate governance after negative media coverage and change their financial reporting strategy. On the other hand, Core et al. (2008) find no evidence that negative press coverage influences executive compensation.
Press coverage can also impact investors' trading decisions. An early study by Foster (1979) analyses the effect of the critical articles by the iconoclastic accounting critic Abe Briloff. Foster finds an average drop in price of 8% for companies whose accounting practices are criticized by Briloff. Brown et al. (2009) adopt a similar approach examining the impact of a leading Australian financial journalist, Trevor Sykes. Dyck and Zingales (2003) find that stock prices react most to the earnings metric that is emphasized by the press. In their study of the effect of Business Week's list of worst board members, Joe et al. find that individual investors overreact to this negative media coverage.
The empirical evidence from previous research on the effect of media coverage is based on press coverage of individual companies. This media attention is likely to change the reporting behaviour of managers. As Dyck et al. (2008) suggest, the influence of media attention on managers' behaviour works through reputation-based mechanisms. This relation is probably stronger in case of company-specific media attention.3 Investors' decisions are influenced by the media because they can provide credibility (Dyck and Zingales, 2003). Conversely, media attention can also destroy credibility. Mercer (2004) identifies the level of external assurance as one of the four factors influencing disclosure credibility. Although the evidence is largely anecdotal, the opinions expressed in the financial press can provide this external assurance to investors. Similarly, negative media attention for certain disclosure practices will harm the credibility of that information and make it less useful for economic decisions. Building on this literature, we argue that media attention on the misuse of non-GAAP information created more awareness and consequently influenced the behaviour of companies as well as investors.
THE CHANGING ENVIRONMENT OF NON-GAAP REPORTING
Changing regulation is only one aspect that may influence the disclosure choices and the use of financial information. This is illustrated by considering changes in the financial reporting environment that relate to the use of non-GAAP metrics.
Most notably, during the period of interest the first major accounting scandals in both the U.S. (Enron in 2001) and the Netherlands (Royal Ahold, early 2003) were revealed. Partly in response to these causes célèbres and alleged fraudulent practices, regulators issued several statements that directly implicated the practice of reporting non-GAAP information. Since the financial reporting environment is important for our analysis, developments are described in some more detail.
In the Netherlands, the Dutch Accounting Standards Board (the DASB4) noted that Dutch companies were reporting non-GAAP5 earnings measures in financial reports more frequently. In the 2002 edition of the DASB guidelines, the DASB addressed this issue in the introduction. The increased frequency with which companies reported EBITA and EBITDA in the income statement worried the DASB. In that same introduction, they stated explicitly that reporting EBITA or EBITDA within the income statement is incompatible with the law and that those measures can only be reported in the notes to the financial statements. This is because neither EBITA nor EBITDA fits into the prescribed formats of the income statement. Interestingly, the DASB does not include any guidelines as to how these measures should be reported or disclosed outside the income statement.6 It therefore seems that the DASB was primarily concerned with the application of the prescribed formats of the financial statements. So, contrary to SOX and Regulation G, the DASB communication was not induced by accounting scandals or alleged opportunistic use of non-GAAP measures. The DASB announcement was an affirmation of existing legislation rather than a change in regulation. Moreover, the affirmation left the publication of non-GAAP measures outside the income statement completely free, so it did not affect financial information in press releases or earnings announcements. Nevertheless, the DASB gave a clear signal that any earnings measure that does not fit in the prescribed format of the income statement is a deviation from GAAP.
The emergence of EBITA and EBITDA in the Netherlands coincided with changes in international standards on goodwill reporting. Following similar developments in the U.S., in 2001 a new accounting directive by the DASB that required goodwill to be capitalized and amortized became effective. Before 2001, it was allowed to set off acquired goodwill against retained earnings. This new accounting standard had been anticipated for some time and many companies had adopted this goodwill treatment before the standard became effective.7 On the other hand, even after 2001 a small number of companies ignored the accounting guideline and continued to record acquired goodwill against equity.8 Anecdotal evidence suggests that among companies that did capitalize goodwill, reporting earnings before interest, taxes, depreciation and amortization (EBITDA) became popular. This non-GAAP earnings measure allowed them to avoid the negative effect of goodwill amortization on net earnings. In the U.S., reporting EBITDA or similar earnings measures was already widespread by that time. The SEC issued a warning in 2001, with the intention of cautioning companies on their reporting of non-GAAP measures, and of calling the dangers of these measures to the attention of investors.
Following the Sarbanes-Oxley law of November 2002, the SEC established rules to regulate the disclosure of non-GAAP financial measures. Early in 2003, the SEC reduced the flexibility in non-GAAP reporting considerably with the passing of Regulation G. This rule required all publicly disclosed non-GAAP information to be reconciled with GAAP information. Furthermore, management had to explain why the non-GAAP information was relevant for investors. In addition, the GAAP information must be presented with the same prominence as the non-GAAP information. Besides the costly expanded disclosures that were required under Regulation G, management was also exposed to the risk of litigation if the requirements were not met.
In 2004, the Dutch professional accountants and auditors organization, Royal NIVRA,9 investigated the annual statements of listed companies for the years 2002 and 2003 and found that alternative measures such as EBITDA were reported frequently.10 In a research report by the Dutch Financial Market Authority (AFM), the various earnings measures reported in the annual reports over 2002 from fifty Dutch listed companies were criticized.11 Early 2004, the AFM issued a press release that urged companies to adhere to guidelines that were very similar to Regulation G.12 Royal NIVRA pressed external auditors and their clients that GAAP net income should be paramount in financial reports and that exotic alternative measures should be banned.13 Audit firms argued that financial disclosures in press releases should be regulated similarly to SOX.14 Despite this discussion, no specific regulation was issued to address the issue.
To summarize, the reporting of non-GAAP measures has led to discussions both in the U.S. and the Netherlands. Although faced with a similar challenge, the response of regulators in the Netherlands has been fundamentally different in comparison to the U.S. In contrast to the SEC, Dutch regulatory agencies such as the DASB and the AFM did not issue any specific rules for the disclosure of non-GAAP information in press releases. They did affirm the existing rules and legislation, stressing the importance of GAAP earnings. In the terminology of the institutional crisis literature, the Dutch regulators' response was ‘light’—conservative adaptive—while the U.S. regulators demonstrated a more ‘heavy’ radical approach. Overall, the Dutch environment of non-GAAP reporting is characterized by the absence of specific regulation directed at banning opportunistic non-GAAP earnings reporting practices in combination with negative attention from regulators and media. Nonetheless, there was substantial discussion about the practice of publishing non-GAAP earnings, warning against misleading and confusing use of alternative measures and advocating rehabilitation of GAAP net income. This provides a unique opportunity to explore whether financial reporting practices and investors' responses change without regulatory intervention.
Earnings Debate in the Dutch Press
Similar to the U.S. debate, the Dutch discussion on the use of alternative earnings measures took place in the media. It was not until late 2003 that financial market institutions (NIVRA, AFM and the DASB) started to participate in the discussion. Several other articles in Dutch newspapers had already warned against the use of alternative earnings measures.15 In order to get a more comprehensive picture of the media attention surrounding non-GAAP reporting, a search was performed in all Dutch newspapers in the years 1999–2005 as available in Lexis Nexis. As already explained, media attention was expected to influence behaviour of financial market participants (i.e., companies and investors). Newspaper coverage was used as a proxy for media attention. Articles that specifically covered the earnings-measures debate were searched. These articles typically pointed out that earnings are a hybrid concept that can be calculated in various ways and that this may lead to confusion. We used several text string16 searches in order to capture all the newspaper articles published on the subject. For each article, we determined if the central message dealt with the earnings debate. For example, articles that discuss the earnings announcement of a specific company and mention the use of non-GAAP earnings metrics in the context of that announcement were removed.
This yielded a total of 96 newspaper articles of which 42 were published in the Dutch financial newspaper Het Financieele Dagblad17 (Financial Daily). The articles published in the financial newspaper as well as the press in general all had a rather critical tone, warning against misleading reporting practices. As reproduced in The number of newspaper articles in Dutch newspapers from 1999–2005 that focused primarily on the arguments for and against different earnings definitions. Figure 1, the flow of articles seemed to build up to a peak in 2002, after which the number decreased again.
EARNINGS DEBATE IN DUTCH PRESS
Roughly one-third (33 of 96) of the newspapers articles concerning the discussion of earnings measures were published in 2002. In 2004, another spike in the interest in the earnings debate occurred. This time, the attention was motivated by concurrent statements of AFM and Royal NIVRA. Companies and their auditors were called upon to adhere to GAAP earnings as the most important earnings measure and to refrain from confusing non-GAAP measures. This led to an extensive debate in the newspapers in January and February of 2004 (17 of the 26 articles of 2004). The statements by the AFM and Royal NIVRA may be considered evidence of the influence of media attention on regulators. Arguably as a result of the negative press of non-GAAP measures, regulators had to respond and issued a statement. Regulators reaffirmed the existing rules and regulations, which qualifies as an adaptive conservative response in terms of the institutional crises literature (Boin and 't Hart, 2000).
The spread of the newspaper coverage supports the notion that 2002 is an important year in the earnings debate examined here. In the year where accounting was front-page news, substantial attention was paid to more technical topics such as alternative earnings measures. This set the spotlight on misleading reporting practices of non-GAAP earnings measures. During the first years of our sample (2000–02) the negative attention for the use of non-GAAP earnings measures increased. The most important year in this discussion is 2002, as the media were focused on accounting scandals after the Enron fraud and several scandals that were revealed in that year (e.g., Worldcom, Tyco, Qwest). Moreover, in 2002 the legislation concerning the use of non-GAAP information in the U.S. became effective. Two periods are distinguished in our sample based on the peak in negative media attention in 2002 (as depicted in Figure 1): the period before 2003, and the period 2003 and after. In other words, the sample is divided between press releases that were issued before and after 1 January 2003. This permits analysis of (a) the non-GAAP reporting behaviour by companies, and (b) investors' responses to GAAP and non-GAAP earnings measures before and after negative media attention.
SAMPLE SELECTION
The quarterly earnings announcements were retrieved from the companies' websites in order to obtain the earnings release in its original format. Under Dutch regulation, only (half-) year reports are compulsory, but the majority of companies voluntarily publish quarterly earnings releases.
Non-GAAP measures were collected from original earnings releases for the primary sample. This offers some advantages. For example, it leads to more accurate information on the reporting behaviour of companies. Adjustments to GAAP earnings made by analysts are not necessarily the same as those reported by firms in their press releases. When compared, non-GAAP earnings as reported in press releases on average differ significantly from the street earnings reported by analysts (Bhattacharya et al., 2003; Marques, 2006). A more practical reason for using press releases is that analysts' street earnings are not readily available given that analyst databases such IBES have only limited coverage of Dutch companies.
The Dutch capital market is relatively small, allowing us to collect all earnings press releases of the large and mid-cap listed companies, and determine whether they report non-GAAP earnings measures or not. This provides a more comprehensive picture of reporting practices in earnings announcements. Prior studies with U.S. data used text searches in order to collect a sample of non-GAAP reporting companies (Bhattacharya et al., 2003; Lougee and Marquardt, 2004). This way, only the earnings releases with the ex ante defined non-GAAP measures can be selected, which may lead to self-selection problems.
Quarterly, half-year and annual earnings release data are hand collected for (large- and mid-cap) companies that were listed at Euronext Amsterdam indexes issued between 2000 and April 2005.18 In line with prior research, all the collected earnings release data are referred to as firm-quarters, even for companies that report semi-annually.
The analyses in the following sections are based on different samples. The initial sample consisted of 766 earnings releases, reported in 21 different quarters for a total of 56 companies. The descriptive evidence of the reporting behaviour of companies was based on the full sample of 766 earnings releases. Analyst data is only available for a subset of 143 firm-quarters. For the analysis of the use of non-GAAP earnings by investors, the sample size is smaller due to data requirements. As set out in Table 1, data required to examine market reactions are only available for 545 press releases. Therefore our analyses of the response to the different earnings measures are based on 545 press releases.
| Companies listed AEX/AMX 1999–2004 | 70 |
| No press releases availablea | 14 |
| 56 | |
| Theoretical initial sampleb | 1,176 |
| Archive starts later than the 4th quarter 1999 | 188 |
| Not reported 1st and 3rd quarters for firms that report semi-annuallyc | 189 |
| 799 | |
| Missing release | 33 |
| Sample for descriptive analysis | 766 |
| No data for four quarters earlier (q-4) | 215 |
| Insufficient data to estimate market model | 6 |
| Final sample | 545 |
- a The following firms are removed from the list (necessary data not available, often merger-related): ASR Verzekeringen, Baan, Cap Gemini, CMG, Endemol, Gucci, KPN Qwest, Libertel, Pakhoed, PinkRoccade, Rodamco Asia, UPC, Vodafone Libertel and Volker Wessels.
- b 21 quarters of 56 companies would lead to a maximum of 1,176 earnings releases.
- c Listed companies at Euronext Amsterdam are not required to report quarterly but semi-annually. Therefore in the first and in the third quarter a number of companies do not report earnings.
For each of the firm-quarters, earnings releases were read and coded and the reported earnings measures listed. The non-GAAP earnings measure that is reported first in the earnings release was the focus. Our assumption was that this is the non-GAAP measure that management wanted to emphasize. An earnings metric was defined to be non-GAAP if the measure is not defined under Dutch regulation. Under Dutch law, the presentation format of the profit and loss account is prescribed.19 The Dutch standard setter DASB emphasized in 2001 that the prescribed formats should not be altered in order to present measures such as EBITA or EBITDA as a subtotal. Accordingly, measures that did not fit in the prescribed models of the profit and loss account were considered to be non-GAAP.
In contrast, GAAP earnings were defined as either bottom-line earnings or a recognized subtotal of the profit and loss account not in violation of the prescribed model, such as earnings before extraordinary items. Within these GAAP measures, we distinguished operating result from bottom-line earnings. Proponents of non-GAAP measures claim that these adjusted GAAP measures provide more insight into a company's permanent earnings by excluding transitory items. This can result in non-GAAP measures referred to as adjusted bottom-line measures or adjusted operating measures, depending on the items that management considers to be transitory. In order to compare the information content of the different measures we needed both operating GAAP and bottom-line GAAP.
THE NON-GAAP REPORTING BEHAVIOUR OF COMPANIES
The frequency and prominence of non-GAAP reporting by Dutch companies was examined first. From the descriptive evidence, we drew some tentative conclusions about companies' motivations. One motivation may be that underlying economics or accounting standards have changed. This possibility was explored in the second paragraph, where we analysed the items that were excluded from GAAP earnings to arrive at the reported non-GAAP measure. Finally, to get a more profound understanding of the factors that drive non-GAAP reporting, the likelihood of companies emphasizing non-GAAP earnings in their announcement was analysed by means of a logit model.
The Frequency and Prominence of Non-GAAP Reporting
To begin, an exploratory analysis was undertaken of the way Dutch companies reported non-GAAP measures in their earnings press releases. The popularity of earnings measures in both the frequency and the prominence with which they are reported was observed. Of the initial sample of 766 earnings releases, 523 (68%) contained at least one non-GAAP earnings measure. The companies that reported these self-constructed measures presented them prominently in their earnings press releases: in 341 firm-quarters, a non-GAAP earnings measure was emphasized by reporting it before GAAP earnings (45% of the total sample, 65% of the non-GAAP releases). Frequency (relative to the total number of press releases in our sample for the year) of press releases containing a non-GAAP measure and frequency of non-GAAP earnings metrics reported as the first earnings measure (primary measure) in the year of publication. Figure 2 graphically illustrates the development of non-GAAP reporting over time. The frequency of reporting non-GAAP measures increased steadily over the period 2000–05 (from 55% to 83%). Furthermore, non-GAAP earnings were reported more often as the first and therefore primary earnings measure. In 2005, 55% of the press releases published a non-GAAP measure first, as compared to 30% in 2000. Based on this we conclude that the popularity of non-GAAP reporting persisted in a period of negative media attention, even increasing after the turbulent year 2002.
FREQUENCY OF NON-GAAP REPORTING 2000–2005
In order to uncover the motivation for non-GAAP reporting, we analysed if the non-GAAP measures were used (more or less) opportunistically. For example, if companies reported non-GAAP to mislead investors, one would expect the non-GAAP measures to have been more positive than the GAAP measures. A simple way to measure this is to compare the frequency of non-GAAP profits to the frequency of GAAP profits in our sample. In the initial sample of 766 press releases, 93% of the reported non-GAAP earnings measures were a profit, compared to 77% of GAAP bottom-line earnings or 78% of GAAP operating earnings. On average, non-GAAP measures presented a more favourable view of a firm's financial performance. For companies in the U.S., similar results have been reported.20
Another reason why companies report non-GAAP measures is to meet analyst forecasts. We therefore collected median and average analyst forecasts from the IBES database. Not all firms were covered by IBES and therefore the sample size was reduced. In 54 cases firms did not meet median analyst forecasts with GAAP earnings. However, in 41 out of 54 cases (76%) firms did meet median analyst forecasts with non-GAAP earnings. Using average analyst forecasts this number equalled 40 (75%) out of 53 cases. In 121 quarters with loss according to GAAP earnings, only 32 (26.4%) reported losses under non-GAAP earnings. This is consistent with managers using non-GAAP earnings to avoid having to report a loss or to miss an analyst forecast under GAAP earnings.
At first glance, the growing popularity of non-GAAP reporting suggests that companies were not influenced by negative media attention and possibly increased investor scepticism. However, on closer inspection, one finds that only 9 (41%) out of the 22 firms that reported non-GAAP measures at least once before the 2003 continued to report a non-GAAP earnings measure at least once in 2003 or later years. The majority of 13 firms (59%) discontinued reporting non-GAAP earnings (10 firms stopped reporting non-GAAP earnings at one point in 2003 or after and 3 firms before 2003). This is consistent with managers changing their behaviour after negative media attention that previously went unnoticed in our analysis. At the same time, 13 firms reported non-GAAP earnings for the first time in 2003 or after. This explains why overall we did not observe a decrease in the number of non-GAAP reporters. Although the number of companies that report non-GAAP measures at least once remained stable in the period before 2003 as compared to the years 2003 and after, it was a different group of companies. Given the increased frequency with which non-GAAP measures were reported over time, it seems the companies that reported non-GAAP in 2003 or after did so in more quarters.
The subset of firms that reported non-GAAP earnings in 2003 or later therefore is a mix of 9 companies that already reported non-GAAP earnings before 2003 and 13 new non-GAAP reporters in 2003 and later years. This raises the question why firms started or continued to report non-GAAP earnings. Firms that newly adopted non-GAAP reporting in 2003 and thereafter must have seen offsetting benefits in the additional disclosure (especially after the negative media attention on non-GAAP reporting). The same argument applies to the firms that already reported before 2003 and continued to do so afterwards.
In sum, the descriptive evidence presented in this subsection reveals that non-GAAP measures as compared to GAAP measures were on average more positive and avoided reporting a loss or missing an analyst forecast more often. Although the proportion of earnings announcements containing non-GAAP earnings measures increased and these measures were reported more prominently, the composition of the group of non-GAAP reporters changed over time. In the next subsection, we explore non-GAAP reporting behaviour in more detail by examining the nature of the adjustments companies made.
Specific Exclusions From GAAP Earnings
In order to understand the underlying reasons that companies had to report a certain adjusted earnings measure, we examine the specific items that they excluded from GAAP earnings. For example, as Entwistle et al. (2006) suggest, the exclusion of certain items may be influenced by changes in accounting standards (e.g., goodwill amortization) or changes in the business environment (e.g., acquisition related charges). Moreover, the consistency with which a company chose to exclude specific items was examined to increase an understanding of companies' motivations.
For each earnings press release, we tabulated the items that are excluded from GAAP earnings to arrive at the reported non-GAAP earnings measure. Based on the descriptions of the non-GAAP measures in the earnings releases, 22 different categories were identified. A list of the exclusions is provided in Table 2.
| Exclusion | 2000–02n=225 | 2003–05n=298 | Totaln=523 | ||
|---|---|---|---|---|---|
| Non-operating items | 114 | 50.7% | 145 | 48.7% | 259 |
| Depreciation*** | 69 | 30.7% | 53 | 17.8% | 122 |
| Amortization*** | 175 | 77.8% | 196 | 65.8% | 371 |
| Impairment*** | 0 | 0.0% | 15 | 5.0% | 15 |
| Exceptional items** | 52 | 23.1% | 95 | 31.9% | 147 |
| Extraordinary items | 52 | 23.1% | 57 | 19.1% | 109 |
| Restructuring charges | 14 | 6.2% | 14 | 4.7% | 28 |
| Acquisition related charges | 3 | 1.3% | 1 | 0.3% | 4 |
| Sale of assets* | 0 | 0.0% | 5 | 1.7% | 5 |
| Share compensation expense*** | 12 | 5.3% | 2 | 0.7% | 14 |
| R&D*** | 7 | 3.1% | 0 | 0.0% | 7 |
| Revaluation (fixed/financial assets)** | 14 | 6.2% | 34 | 11.4% | 48 |
| Current cost valuation | 10 | 4.4% | 9 | 3.0% | 19 |
| Foreign currency | 10 | 4.4% | 13 | 4.4% | 23 |
| Provisions/accruals | 7 | 3.1% | 10 | 3.4% | 17 |
| Discontinued operations | 8 | 3.6% | 12 | 4.0% | 20 |
| Realized investment gains/losses | 11 | 4.9% | 13 | 4.4% | 24 |
| Rent | 5 | 2.2% | 6 | 2.0% | 11 |
| Penalties/claims* | 3 | 1.3% | 11 | 3.7% | 14 |
| Pension charges | 0 | 0.0% | 2 | 0.7% | 2 |
| Finance related charges | 0 | 0.0% | 3 | 1.0% | 3 |
| Excluded Segments** | 0 | 0.0% | 6 | 2.0% | 6 |
| Total exclusions | 566 | 702 | 1,268 | ||
| No specification | 4 | 1.8% | 9 | 3.0% | |
| Incomplete specification | 5 | 2.2% | 9 | 3.0% | |
| Switched primary measure | 67 | 29.8% | 86 | 28.9% | 153 |
| Switched definition of non-GAAP measure | 93 | 41.3% | 128 | 43.0% | 221 |
- Note: For each exclusion we test whether the percentages are equal across the two periods. For exclusions with superscript ***, ** or * equality is rejected at 1%, 5% or 10% significance, respectively.
In total, there were 1,268 exclusions from the 523 non-GAAP measures reported in the press releases. On average, a non-GAAP measure excluded 2.4 items from GAAP earnings. This number was stable throughout the sample period, suggesting that the average complexity of the non-GAAP measures remained fairly stable. On the other hand, the variation of non-GAAP measures grew, since the number of categories of exclusions increases from 17 to 21. When comparing the period before the negative media attention (before 2003) with the period after (2003 and after), the overall picture looked relatively stable with 10 types of exclusions increasing and 11 decreasing.
From the 523 non-GAAP measures, 371 exclude amortization charges, which is substantial. This may raise the question as to what extent non-GAAP reporting is driven by changes in goodwill accounting standards. To address this concern, we took a closer look at the goodwill accounting choices of the companies in our sample. For each company we determined the quarter when they first started capitalizing and amortizing goodwill. In the sample, 51% of the companies started to capitalize goodwill before the introduction of the new guidelines in 2001. In addition, 13 companies already voluntarily started to capitalize and amortize goodwill before the beginning of our sample period (fourth quarter, 1999). Another 13 firms only had quarters with amortized goodwill included in our analysis. This means that 26 companies (45%) had consistent goodwill accounting across sample quarters.
When comparing the two periods, a significant decrease was evident in the frequency with which amortization was excluded from GAAP earnings, together with an increase in the exclusion of annual impairments. This coincides with an alternative accounting treatment of goodwill that was gaining popularity in this period, according to which goodwill was not amortized but instead tested for impairment annually. Taken together, adjustments related to goodwill (amortization and impairment) have decreased from 77.8% to 70.8%. The declining frequency and the consistent goodwill reporting of 45% of the sample both support the notion that goodwill accounting changes did not drive the findings.
Another accounting guideline that was issued during the sample period related to the reporting of exceptional and extraordinary items.21 Exceptional and extraordinary charges were excluded 147 and 109 times, respectively. Although the new accounting guidelines effectively prohibited reporting extraordinary items in the income statement, this caused companies to report a non-GAAP measure excluding extraordinary items more often.
Finally, analysing these exclusions also helped to distinguish between opportunistic and informative use of these measures. Assuming informative incentives for non-GAAP reporting would lead to a consistent way of reporting, one would expect companies to report the same non-GAAP measure in its consecutive earnings releases. Moreover, companies would emphasize the same earnings measure in their press releases. Exploring the consistency with which the non-GAAP measures were reported, it is evident that a stable proportion of the sample switched the definition of the non-GAAP measure. Before 2003, 41% of the non-GAAP measures reported by a specific company were defined differently than the quarter before (43% of the non-GAAP measures in 2003 and later years). Thirty per cent of the non-GAAP earnings releases in the period before 2003 reported a different earnings measure more prominently in the press release than before (29% of the non-GAAP releases in 2003 and after). Furthermore, in a small number of earnings releases, the company omitted explaining the definition of the non-GAAP measure by either reporting an incomplete specification or no specification at all.
In sum, analysis of the excluded items shows that there were some shifts in the use of non-GAAP earnings measures that may have been induced by changes in accounting standards, but that these did not explain the overall trend. Moreover, a large proportion of our sample did not report the non-GAAP measures consistently. The next subsection will further examine what drives companies' choice to report non-GAAP measures.
Logit Regression
We now take a closer look at these descriptive results and estimate a logit model, trying to explain why firms used non-GAAP earnings as the primary earnings measure in some quarters but not in others. The logit model examines whether the likelihood that the firm used non-GAAP as its primary earnings measure was associated with quarterly GDP growth and the number of acquisitions undertaken by the firm. Poor economic conditions were expected to be associated with a larger amount of exceptional items, which made it more likely that firms adopt non-GAAP measures. Similarly, we expected companies to be more likely to report non-GAAP earnings because GAAP earnings are lowered by the goodwill amortization. A dummy variable NEGATIVE MEDIA takes one for quarters in 2003 and later years. This allows investigation of whether firms were less likely to report non-GAAP measures after the peak in negative media attention. Other determinants for non-GAAP reporting such as meeting analyst forecasts and avoiding reporting losses under GAAP earnings are also included.
The first column of Table 3 shows that firms were more likely to report non-GAAP earnings as their primary earnings measure in quarters with poor economic growth and when they reported a GAAP loss. Interestingly, there was no difference in the likelihood of reporting non-GAAP earnings when comparing the period before and after the negative media attention on non-GAAP earnings. This is consistent with the earlier finding that non-GAAP reporting did not decline over time.
| Full sample | Before media attention | After media attention | Full sample | Before media attention | After media attention | |
|---|---|---|---|---|---|---|
| GDP GROWTH | −3.111 | −5.131 | −2.435 | −19.791 | −32.639 | −8.947 |
| (−4.921)*** | (−1.723)** | (−1.233) | (−1.699)* | (−1.927)** | (−0.902) | |
| M&A ACTIVITY | −0.128 | −0.281 | 0.544 | 0.464 | 0.026 | 1.117 |
| (−0.713) | (−1.15) | (1.633)** | (0.891) | (0.034) | (1.678)** | |
| LOSS | 0.493 | 0.306 | 0.635 | 0.930 | 0.234 | 1.332 |
| (3.056)*** | (1.344)* | (2.733)*** | (2.350)*** | (0.393) | (2.293)** | |
| MISS FORECAST | 0.721 | 0.075 | 1.524 | |||
| (1.903)** | (0.135) | (2.798)*** | ||||
| NEGATIVE MEDIA | 0.191 | 0.934 | ||||
| (1.271) | (1.247) | |||||
| Pseudo R2 | 0.014 | 0.016 | 0.023 | 0.132 | 0.084 | 0.159 |
| Wald χ2 | 14.564*** | 8.557** | 10.563** | 21.291*** | 6.264* | 12.393*** |
| Number of firm-quarters | 766 | 419 | 347 | 143 | 82 | 61 |
- Note: Table shows the logit regression results using a dummy variable PRIM_NON as the dependent variable. PRIM_NON takes on the value one if the firm uses non-GAAP earnings as the primary earnings measure in its press release. GDP GROWTH is the quarterly growth in Gross Domestic Product from the Dutch Statistical Office (CBS). M&A ACTIVITY is the number of acquisitions the firm undertakes in that quarter and is downloaded from Thomson's SDC M&A database. LOSS is a dummy variable that indicates one if the firm reports a loss under GAAP earnings. MISS FORECAST is a dummy that indicates one if the firm fails to meet the median analyst forecast in that quarter. Analyst forecasts come from IBES. NEGATIVE MEDIA is a dummy variable that indicates one if the quarter is in 2003 or later years. We split the sample in observations before and after the negative general media attention. z-statistics are shown in parentheses. * statistically significant at the 10% level (one-tailed), ** statistically significant at the 5% level (one-tailed) and *** statistically significant at the 1% level (one-tailed).
If one compares the period before (second column) and after (third column) the negative media coverage the coefficient on the loss dummy was found to have doubled in size. This increase in the coefficient is statistically significant at the 5% level and suggests that companies that reported losses under GAAP were more likely to emphasize non-GAAP earnings in their press releases after the negative media attention in 2002. In the period after the negative media attention firms were also more likely to use non-GAAP earnings as their primary earnings measure in quarters in which they conduct more mergers and acquisitions.
Next, it was examined whether firms reported non-GAAP earnings to meet analyst forecasts that they would not meet with GAAP earnings. Column four supports this hypothesis. If firms missed the median analyst forecast they were more likely to highlight non-GAAP earnings in their press releases, but only after the negative media attention in 2002 (column six).
These results confirm the descriptive evidence here that turning GAAP losses into non-GAAP profits and meeting analyst forecasts are important motivations for firms to report non-GAAP earnings. Especially after the negative media attention, these two motivations increased in importance. In this period firms were also more likely to make use of non-GAAP earnings measures when they conducted more mergers and acquisitions. In the period before the negative media attention, poor economic conditions increased the likelihood of non-GAAP reporting.
To take a further look at this, the influence of these factors on the size of the adjustments was also investigated. Adjustments were measured as the difference between non-GAAP and GAAP earnings.22Table 4 shows the results. Adjustments are found to be larger when economic conditions are poor; the company reports a GAAP loss or when the company misses the median analyst forecast with its GAAP earnings. Interestingly, significantly lower adjustments were found after negative media coverage on non-GAAP earnings in general. This suggests that managers decided to deviate less from GAAP earnings when reporting non-GAAP earnings after the latter have been criticized in the media.
| Model (1) | Model (2) | |
|---|---|---|
| GDP GROWTH | −0.969 | −1.466 |
| (−3.041)*** | (−2.552)*** | |
| M&A ACTIVITY | −0.026 | −0.001 |
| (−1.124) | (−0.081) | |
| LOSS | 0.012 | 0.023 |
| (1.454)* | (1.563)* | |
| MISS FORECAST | 0.024 | |
| (1.456)* | ||
| NEGATIVE MEDIA | −0.036 | −0.052 |
| (−4.103)*** | (−2.935)*** | |
| R 2 | 0.093 | 0.184 |
| F-statistic | 8.866*** | 3.294*** |
| Number of firm-quarters | 379 | 90 |
- Note: Table shows the OLS regression results using ADJUSTMENTS as the dependent variable. ADJUSTMENTS is the difference between UE NONGAAP and UEGAAP as defined in footnote 22. GDP GROWTH is the quarterly growth in Gross Domestic Product from the Dutch Statistical Office (CBS). M&A ACTIVITY is the number of acquisitions the firm undertakes in that quarter and is downloaded from Thomson's SDC M&A database. LOSS is a dummy variable that indicates one if the firm reports a loss under GAAP earnings. MISS FORECAST is a dummy that indicates one if the firm fails to meet the median analyst forecast in that quarter. Analyst forecasts come from IBES. NEGATIVE MEDIA is a dummy variable that indicates one if the quarter is in 2003 or later years. t-statistics are shown in parentheses. * statistically significant at the 10% level (one-tailed), ** statistically significant at the 5% level (one-tailed) and *** statistically significant at the 1% level (one-tailed).
Finally, we investigated whether managers of firms that were criticized in the media for their use of non-GAAP earnings also changed their reporting behaviour; in particular, whether these managers were less likely to continue to use non-GAAP earnings as their primary measure in their press release one quarter after they were criticized in the media. As Dyck et al. (2008) suggest, managers respond to negative media attention to reduce reputation effects and therefore the effect of negative publications is probably stronger when a company is named specifically in a negative newspaper article. Managers are found less likely to emphasize non-GAAP earnings in their press releases after negative media attention and adjust their behaviour. Of the 12 firms that were criticized, 6 (50%) turned away from non-GAAP reporting the next quarter. However, 4 of these 6 firms (83%) took up non-GAAP reporting after some period. This is consistent with the press having at least a transitory impact on the behaviour of managers.
In conclusion, although companies continued to report non-GAAP measures and did so more frequently and more prominently, evidence was found of a disciplining effect of negative media attention on the reporting of non-GAAP metrics. About 60% of the firms that reported non-GAAP before 2003 stopped reporting non-GAAP earnings afterwards. The firms that continued to report non-GAAP earnings, or that were new non-GAAP reporters, deviated less from GAAP earnings. This suggests that even though the frequency of non-GAAP reporting did not decrease over time, the magnitude of the adjustments did decrease in 2003 and thereafter, possibly as a result of the negative media coverage of non-GAAP earnings in general. Furthermore, there is evidence of short-term disciplining effect on companies that were ‘named and shamed’ in the media for their non-GAAP reporting. On the other hand, several indicators of opportunistic non-GAAP reporting practices, such as avoiding losses or avoiding missing a benchmark, persisted or even increased after the negative attention.
INVESTORS' USE OF NON-GAAP MEASURES
In order to determine investors' use of non-GAAP earnings measures as compared to GAAP earnings, we examined the informativeness of the identified earnings metrics. First an event study was performed for the entire sample. Next, to delve deeper the sample was split between the period before 2003 and 2003 and after to analyse the effect of media attention in general. After that there was an examination of the effect of specific media attention by distinguishing between companies that were criticized in the media for their non-GAAP reporting and companies that were not. The section concludes with a subsection on robustness tests.
Event Study for the Entire Sample
A standard event study procedure was used to assess if stock prices changed in response to the different earnings measures disclosed in the press releases.23
Using a random-walk earnings expectations model we defined unexpected earnings as the three EPS figures (GAAP, operational GAAP and non-GAAP) minus GAAP earnings from four quarters earlier (q-4). Unexpected earnings were used instead of forecast errors, because analyst forecast data were lacking for most Dutch companies during our sample period. We calculated three measures of unexpected earnings or earnings surprise: UE GAAP, UE OGAAP and UE NONGAAP. On average, the non-GAAP measures resulted in unexpected earnings of 5.3%, compared to UE GAAP of minus 0.2%. The mean market capitalization MCAP (€10,397 million) was much higher than the median (€1,593 million), revealing that a few firms in our sample were much larger than most of the sample firms. This is in fact a characteristic feature of the Dutch financial market, which is dominated by a few large multinational companies.
We first examined which definition of earnings investors paid attention to non-GAAP earnings or GAAP earnings (either bottom-line or operating). To gain insight into the degree to which the market was processing each measure in prices, we investigated a short-window association between abnormal returns on each earnings surprise (unexpected earnings) measure separately. If the market found non-GAAP earnings to be a better summary measure of performance, returns would be more highly correlated with UE NONGAAP than with UE GAAP or UE OGAAP.
Table 6 presents the results of regressions of abnormal returns on unexpected earnings. The regression was estimated separately for UE GAAP, UE OGAAP and UE NONGAAP (Models 1–3, respectively). The regression was not estimated for the three unexpected earnings metrics together, because of high correlations between the earnings definitions (correlations above 0.95).
| Model (1) | Model (2) | Model (3) | |
|---|---|---|---|
| UE GAAP | 0.155 | ||
| (2.585)*** | |||
| UE OGAAP | 0.033 | ||
| (0.602) | |||
| UE NONGAAP | 0.107 | ||
| (2.432)*** | |||
| Intercept | −0.001 | −0.004 | −0.004 |
| (−0.177) | (−1.007) | (−0.979) | |
| R 2 | 0.012 | 0.001 | 0.015 |
| F-statistic | 6.684*** | 0.363 | 5.916*** |
| Number of observations | 538 | 363 | 381 |
| Comparison of earnings measures | ||
|---|---|---|
| Vuong's Z-statistic | Probability | |
| UE OGAAP vs UE GAAP | 7.48 | <0.0001 |
| UE NONGAAP vs UE GAAP | −1.62 | 0.1056 |
| UE OGAAP vs UE NONGAAP | 7.75 | <0.0001 |
- Note: Table shows the regression results using CAR as the dependent variable. We refer to Table 5 for variable definitions. t-statistics are shown in parentheses. * statistically significant at the 10% level (one-tailed), ** statistically significant at the 5% level (one-tailed) and *** statistically significant at the 1% level (one-tailed).
| Variable | Mean | Median | SD | N |
|---|---|---|---|---|
| EARN GAAP | 255.991 | 44.000 | 952.908 | 538 |
| EARN OGAAP | 433.674 | 69.000 | 1,329.333 | 363 |
| EARN NONGAAP | 474.054 | 106.000 | 870.280 | 381 |
| UE GAAP | −0.002 | 0.001 | 0.053 | 538 |
| UE OGAAP | 0.018 | 0.017 | 0.072 | 363 |
| UE NONGAAP | 0.052 | 0.022 | 0.090 | 381 |
| MCAP | 10,397.640 | 1,592.690 | 19,653.910 | 545 |
| CAR (%) | −0.073 | −0.430 | 7.376 | 545 |
- Note: Table shows summary statistics for quarterly earnings press releases issued by Dutch listed companies from January 2000 to April 2005. EARN GAAP, EARN OGAAP and EARN NONGAAP denote the GAAP earnings, operational GAAP earnings and non-GAAP earnings (in millions of euros), respectively. UE GAAP, UE OGAAP and UE NONGAAP denote the unexpected earnings (earnings surprise) for GAAP, operational GAAP and non-GAAP earnings, respectively. We use the random model to compute unexpected earnings and use the GAAP earnings four quarters earlier (q-4) as our proxy for expected earnings. The UE GAAP is computed as (EARN GAAP-EARN GAAP (q-4))/MCAP, UE OGAAP is computed as (EARN OGAAP-EARN GAAP (q-4))/MCAP and UE NONGAAP is computed as (EARN NONGAAP-EARN GAAP (q-4))/MCAP. The unexpected earnings (earnings surprise) is trimmed at the 5th and 95th percentile. MCAP denotes the market capitalization five trading days before the press release (in millions of euros). CAR denotes the cumulative abnormal return during the three trading day interval from one day before to one day after the press release.
In the separate unexpected earnings regression reported in Table 6, UE GAAP, UE OGAAP and UE NONGAAP were positively related to short window returns. The coefficients on both UE GAAP and UE NONGAAP were statistically significantly positive. These results indicated that the different definitions of unexpected earnings had different explanatory power with respect to short window abnormal stock returns. Remarkably, bottom-line earnings were informative whereas operating earnings were not. Normally it would be argued that operating earnings are closer to core earnings and are therefore more relevant to investors. Furthermore, non-GAAP earnings are informative, which is in line with prior research. Consistent with U.S. studies, non-GAAP earnings were found to be more informative than GAAP operating earnings (Bhattacharya et al., 2003), but this does not hold for GAAP earnings.
Before and After Negative Media Attention
If negative media attention had an effect on the way investors perceive non-GAAP information and if they adjusted their behaviour accordingly, the information content of the respective earnings measures should have changed from 2003 onwards. The results of the regressions in these two periods are reported in Table 7.
| Before negative media attention | After negative media attention | |||||
|---|---|---|---|---|---|---|
| Model (1) | Model (2) | Model (3) | Model (4) | Model (5) | Model (6) | |
| UE GAAP | 0.207 | 0.108 | ||||
| (2.140)** | (1.522)* | |||||
| UE OGAAP | −0.021 | 0.098 | ||||
| (−0.240) | (1.445)* | |||||
| UE NONGAAP | 0.130 | 0.061 | ||||
| (1.808)** | (1.148) | |||||
| Intercept | 0.004 | −0.005 | −0.003 | −0.004 | −0.004 | −0.005 |
| (0.754) | (−0.720) | (−0.331) | (−1.247) | (−0.890) | (−1.099) | |
| R 2 | 0.018 | 0.001 | 0.020 | 0.008 | 0.010 | 0.006 |
| F-statistic | 4.580** | 0.058 | 3.272** | 2.317* | 2.089* | 1.318 |
| Number of observations | 248 | 160 | 166 | 290 | 203 | 215 |
| Comparison of earnings measures | ||||
|---|---|---|---|---|
| Before negative media attention | After negative media attention | |||
| Vuong's Z-statistic | Probability | Vuong's Z-statistic | Probability | |
| UE OGAAP vs UE GAAP | 6.76 | <0.0001 | 0.42 | 0.6759 |
| UE NONGAAP vs UE GAAP | 0.55 | 0.5831 | −0.66 | 0.5110 |
| UE OGAAP vs UE NONGAAP | 4.29 | <0.0001 | 1.02 | 0.3103 |
- Note: Table shows the regression results using CAR as the dependent variable. We refer to Table 5 for variable definitions. We split the sample in observations before and after the peak in negative media attention. t-statistics are shown in parentheses. * statistically significant at the 10% level (one-tailed) and ** statistically significant at the 5% level (one-tailed).
Models 1–3 depict the results of the regression in the period before 2003, and Models 4–6 refer to the period 2003 until April 2005. Before 2003, both the bottom-line GAAP and the non-GAAP unexpected earnings measures were significantly positive. The coefficient on UE OGAAP was not significant during this period. This result was similar to the aggregated results for the entire period. So before 2003, investors seemed to price both GAAP bottom-line earnings and non-GAAP earnings.
In the second period (2003 and after), however, the results change. UE NONGAAP was no longer significant, while the unexpected earnings on bottom-line GAAP and operating GAAP earnings were significant (at the 10% level). According to our results, the decreased use by investors of non-GAAP information coincided with an increasing popularity of operating GAAP earnings. The coefficient on UE OGAAP switched to a positive sign, and was significant.
The coefficient on UE GAAP was statistically significant for the entire sample period as well as for the two sub-periods. This may suggest that bottom-line earnings were in fact informative to capital market participants. This contradicts the critics of bottom-line earnings, who claim that this earnings measure is not useful to investors because it includes items that are non-operating or transitory. The use of net income in financial reports is strongly encouraged by Dutch financial market authorities and regulators such as the DASB. Based on the results, it seems investors agreed with the regulators on the importance of bottom-line GAAP earnings.
Although the explanatory power (R2) of the Models was low, as reported in Table 6 and 7, this is in line with other research (Bhattacharya et al., 2003). Vuong's (1989) Z-statistic was calculated to compare the explanatory power of the Models and found that for the entire period UE GAAP and UE NONGAAP had significantly more explanatory power than UE OGAAP, suggesting that investors found operating GAAP the least informative earnings measure. This result held for the first period of our sample, before the negative media attention. After negative media attention, the explanatory power of the models did not differ significantly, suggesting that investors found the earnings measures equally (un)informative.
Effect on Companies That Were Specifically Criticized
As argued in earlier research (e.g., Dyck et al., 2008) the effect of media attention may have been stronger for companies that were specifically criticized for their use of non-GAAP earnings in newspaper articles. A regression was estimated using the stock market reaction surrounding the earnings press release as our dependent variable. In order to test whether there are any differences between firms that received negative media coverage and the ones that did not, a dummy FIRM IN MEDIA was included that indicated one if the company was criticized in the media for using non-GAAP earnings. This dummy (FIRM IN MEDIA) tells us whether there was a different intercept for these firms compared to firms that did not receive negative media attention. This dummy was then interacted with UE NONGAAP. This interaction term captured differences in slope coefficients between the two group of firms and tested whether investors perceived non-GAAP earnings differently between the two groups.
Regressions were run before and after the negative media attention in general, enabling one to compare the effect of general media and firm-specific media coverage. The results are reported in Table 8. Results show that investors assigned less value relevance to the non-GAAP earnings of firms that are targeted in the media. The interaction term turned statistically significant for 2003 and after. This suggests that the negative effect of negative firm-specific media attention on the use of non-GAAP measures was stronger when there was also more widespread critical discussion of the use of non-GAAP earnings.
| Before negative media attention | After negative media attention | |
|---|---|---|
| UE NONGAAP | 0.276 | 0.136 |
| (2.031)** | (1.701)** | |
| FIRM IN MEDIA | −0.021 | −0.012 |
| (−0.478) | (−1.141)* | |
| UE NONGAAP* | −0.237 | −0.128 |
| FIRM IN MEDIA | (−1.09) | (−1.34)* |
| Intercept | −0.004 | −0.031 |
| (−0.51) | (−0.600) | |
| R 2 | 0.059 | 0.032 |
| F-statistic | 1.622* | 2.271** |
| Number of firm-quarters | 166 | 215 |
- Note: Table shows the regression results using CAR as the dependent variable. FIRM IN MEDIA is a dummy variable that indicates one if the firm was mentioned in the press as an opportunistic non-GAAP reporter. We interact this dummy with UE NONGAAP. We refer to Table 5 for variable definitions. We split the sample in observations before and after the peak in negative media attention. t-statistics are shown in parentheses. * statistically significant at the 10% level (one-tailed) and ** statistically significant at the 5% level (one-tailed).
Investors continued to assign value relevance to non-GAAP earnings in 2003 and later years for firms without negative company-specific media attention but the regression coefficient has halved in size (from 0.27 to 0.13). This drop in regression coefficients was significant at 5% level. This shows that also for these firms the market awarded less information content to non-GAAP earnings compared to the period before 2003.
Overall, it seems that firm-specific media attention had a large impact on how investors perceive non-GAAP numbers, especially when it was part of a general discussion on the use of non-GAAP earnings. However, the negative media attention on non-GAAP earnings in general also had an impact on firms that were not targeted in the media. Investors continued to perceive their non-GAAP earnings as informative in 2003 and after, but less so compared to the period before 2003.
An additional analysis was performed on firms that were named specifically in the newspapers articles. Twelve firms that were targeted in the media for their (opportunistic) use of non-GAAP measures allowed us to examine the effect of the media more closely. Five firms were targeted before 2003, and seven in 2003 and later. Here one was able to identify the exact quarter in which the media reported negatively on the use of non-GAAP earnings measures by these firms. We then compared the quarters before and after the negative media attention and examined whether the stock market response was more negative after the firms were ‘named and shamed’. The results are reported in Table 9. The informativeness of non-GAAP earnings was found to decrease whereas the informativeness of GAAP earnings increased after the company was mentioned in the media as an (opportunistic) non-GAAP user.
| Before negative media attention on company | After negative media attention on company | |||||
|---|---|---|---|---|---|---|
| Model (1) | Model (2) | Model (3) | Model (4) | Model (5) | Model (6) | |
| UE GAAP | −0.122 | 0.157 | ||||
| (−0.394) | (1.625)* | |||||
| UE OGAAP | −0.245 | 0.036 | ||||
| (−1.211) | (0.405) | |||||
| UE NONGAAP | 0.215 | 0.016 | ||||
| (2.203)** | (0.233) | |||||
| Intercept | 0.003 | −0.004 | −0.003 | −0.018 | −0.019 | −0.018 |
| (0.335) | (−0.250) | (−0.263) | (−2.427)** | (−2.081)** | (−2.081)** | |
| R 2 | 0.002 | 0.025 | 0.032 | 0.048 | 0.003 | 0.001 |
| F-statistic | 0.152 | 1.461 | 4.844** | 2.245* | 0.163 | 0.053 |
| Number of firm-quarters | 53 | 29 | 51 | 67 | 45 | 67 |
- Note: Table shows the regression results using CAR as the dependent variable. We refer to Table 5 of the paper for variable definitions. We split the sample in observations before and after the negative media attention on a company. t-statistics are shown in parentheses. * statistically significant at the 10% level (one-tailed) and ** statistically significant at the 5% level (one-tailed).
Robustness Tests
We performed additional tests to determine if our results were robust for alternate model specifications. First, results were tested for sensitivity to the exact date that we chose to separate the observations before the negative media attention from the ones after (1 January 2003). While remaining within the boundaries of the period of intensified media attention, we shifted the date partitioning the sample one quarter backward (30 September 2002) and forward (30 March 2003) and ran the regressions again. Similar results occurred, suggesting our findings were insensitive to the exact date.
Next, a panel data analysis was performed, which added the time-series dimension to our cross-sectional analysis presented in Table 6. Panel data helped also to control for omitted variables that change over time but not across companies (i.e., accounting policies, economic conditions, etc.) and for unobserved differences in business practices across companies. Random effects were used to estimate the panel data regression (the Hausman test rejected the use of fixed effects). Unreported results showed that analysing the data using random effects did not yield materially different results than using OLS. Furthermore, regressions were repeated after excluding firms with a cross listing at a U.S. exchange. The results are consistent with the results for the entire sample. This alleviated concerns that the findings were driven by cross-listed firms that responded to the regulation in the U.S.
Moreover, in addition to the original specification of the regression presented in Table 6, we re-estimated the regression using UE GAAP and ADJUSTMENTS (defined as the difference between UE GAAP and UE NONGAAP). The (unreported) results showed that the magnitude of the adjustments mattered, as larger adjustments were more informative. This finding only held in the period before 2003. This corresponds with our finding that non-GAAP earnings were not informative in the period 2003 and after.
Finally, in addition to the tabulated results, the regressions were performed including the market capitalization five days before the press release was published. The results of this test were similar to the models excluding market capitalization.
CONCLUDING REMARKS
This paper investigated the use of different definitions of earnings: earnings calculated according to generally accepted accounting principles (GAAP earnings, both bottom-line and operating) and alternative versions of earnings that exclude various items recorded under GAAP (non-GAAP earnings). Our study is placed in the turbulent period where financial scandals were front-page news and investors' trust was on a historical low. During this period, influenced by the accounting scandals, the use of alternative earnings measures received negative media attention. Building on the growing literature on the influence of media attention and press coverage on the behaviour of managers and investors, it is argued that this negative media attention may have affected non-GAAP reporting. The Dutch setting offered us the possibility to study the effect of the negative attention while the rules and regulations remained the same as before the scandals.
In 2003 and later years, companies were found to have different motivations for non-GAAP reporting and that the difference between reported GAAP earnings and non-GAAP earnings was smaller. The change in reporting behaviour was stronger for companies that were named specifically in the media. This evidence suggests that companies' reporting choices may have been influenced by factors such as media attention, even without regulatory changes. Investors seemed to take the warnings in the media seriously and turned away from non-GAAP measures. From 2003 on, the evidence shows investors consider GAAP earnings to be informative, whereas they did not price non-GAAP earnings measures. This contrasts with the findings before 2003, where investors seemed to find non-GAAP earnings useful, as well as bottom-line GAAP earnings. Collectively, the findings here suggest that market participants perceived non-GAAP earnings measures to have been less informative after a peak in negative media attention.
This study is important to regulators and standard setters. Critical opinions in the media and warnings by regulators expressed through the media are shown to be effective means to create awareness among investors, and to some extent to alter companies' reporting behaviour. U.S. studies suggest that specific regulation has successfully restrained opportunistic non-GAAP reporting practices. On the other hand, there is evidence that the SEC regulation leads to suboptimal reporting decisions (Kolev et al., 2008; Heflin and Hsu, 2008). Our results cast doubt on the attribution of changes in investor and company behaviour to the effect of regulation and suggest that investor perceptions can change without regulation. This potentially has important implications for regulation effectiveness studies that evaluate the effect of new regulations on the behaviour of market participants.
The current study indicates the need to expand our understanding of the effect of regulation. In order to evaluate the effectiveness and necessity of the regulation of financial markets, we need to understand the effect of concurrent changes in the environment that may influence behaviour of financial markets participants. For example, media attention can induce reputation effects that discipline reporting behaviour or create awareness among the users of financial information. We report evidence that suggests that investors base their decisions on different earnings metrics after negative attention from media and regulators. Such effects may decrease the necessity of additional regulation. In order to disentangle the effects of regulation and reporting environment, more research in an international setting may be fruitful.
