Has the Regulation of Non-GAAP Disclosures Influenced Managers’ Use of Aggressive Earnings Exclusions?

Congressional and SEC Action Relative to Non-GAAP Reporting

In the midst of the 2001-2002 financial crisis, the SEC issued a warning on December 4, 2001 about the presentation of earnings and operating results on a non-GAAP basis (SEC, 2001). This warning emphasizes that the antifraud provisions of the federal securities laws also apply to companies reporting non-GAAP metrics and that the principles underlying the presentation of non-GAAP-related measures should be clearly discussed (SEC, 2001). The following year, Congress enacted SOX (on July 30, 2002), and as required by Section 401 (b) of the Act, the SEC implemented a number of new rules addressing the use of non-GAAP measures. First, it prohibits companies disclosing non-GAAP earnings from omitting material information. Moreover, they are required to provide explanations for exclusions, along with a detailed reconciliation between the GAAP and non-GAAP numbers to ensure that investors are not misled. These rules governing non-GAAP disclosures were implemented via a new regulation, known as Reg. G, which was finalized on January 22, 2003, and became effective on March 28, 2003 (SEC, 2002). However, consistent with Black et al. (2012), we expect that most changes in non-GAAP reporting occurred when the Act was passed, rather than when Reg. G was officially implemented in January 2003. To examine the impact of regulation on non-GAAP reporting, we include quarters ending after July 2002 in the post-SOX period.

Reg. G includes a general disclosure requirement and a specific content requirement. The general disclosure requirement indicates that

a registrant or a person acting on its behalf, shall not make public a non-GAAP financial measure that, taken together with the information accompanying that measure, contains an untrue statement of a material fact or omits to state a material fact necessary in order to make the presentation of the non-GAAP financial measure, in light of the circumstances under which it is presented, not misleading. (SEC, 2002)

The specific content requirement involves a reconciliation of any reported non-GAAP financial measure to a measure computed under GAAP. In particular, it requires companies disclosing non-GAAP financial measures to identify the most directly comparable GAAP measure and to provide a reconciliation of the differences between the GAAP and non-GAAP numbers. Similarly, amended Item 10 of Regulation S-K and Item 10 of Regulation S-B require companies disclosing non-GAAP numbers in filings with the Commission to present (a) the most directly comparable GAAP measure with equal or greater prominence, (b) a reconciliation of the differences between the two measures, and (c) an explanation of the reasons justifying the usefulness of the non-GAAP disclosures. In addition to these requirements, Item 10 of Regulation S-K and Item 10 of Regulation S-B prohibit companies from disclosing a non-GAAP measure to “eliminate or smooth items identified as non-recurring, infrequent or unusual,” if there is a high likelihood that these items will recur within 2 years or when a similar item was reported in the previous 2 years. These amendments also prohibit companies from “presenting non-GAAP financial measures on the face of the registrant’s financial statements prepared in accordance with GAAP or in the accompanying notes” as well as the utilization of descriptions for non-GAAP items that are similar to those used for GAAP measures (SEC, 2002).

Finally, Item 12 (“Disclosure of Results of Operations and Financial Condition”) of Form 8-K requires firms to furnish a Form 8-K to the SEC within 5 business days of the public release of annual or quarterly results. It also requires them to disclose the reasons they believe non-GAAP measures will be useful to investors and any additional reasons regarding the disclosure of non-GAAP measures that were not already mentioned (SEC, 2002). In January 2010, the SEC renewed its emphasis on non-GAAP financial measures by issuing a Compliance and Disclosure Interpretation (which was subsequently updated in July 2011). The SEC also began to scrutinize the consistency and accuracy of non-GAAP disclosures to ensure that they are not misleading (Deloitte, 2011). Recently, the SEC’s former chief accountant of the Enforcement Division, Howard Scheck, has emphasized that non-GAAP metrics are a “fraud risk factor” (Leone, 2010). Finally, the SEC formed a taskforce in July 2013 to scrutinize companies’ non-GAAP earnings metrics that could potentially be misleading “with an eye toward possible enforcement cases” (Rapoport, 2013).

Accounting Rule Changes

Two important accounting rule changes occurred during our sample period, which are relevant to our study. In particular, the Financial Accounting Standards Board (FASB) issued two new standards related to the accounting for (a) goodwill and other intangible assets (SFAS 142) and (b) stock-based compensation (SFAS 123R). We briefly summarize these accounting rule changes to clarify how they influence our analyses.

Background and Hypothesis Development

Prior research indicates that manager-adjusted non-GAAP earnings have generally become more common over time and that there is still potential for investors to be misled by these disclosures (e.g., Black et al., 2012; Brown, Christensen, Elliott, & Mergenthaler, 2012; Curtis et al., 2014). Black and Christensen (2009) explore the motivations for these non-GAAP reports using a hand-collected sample of non-GAAP earnings disclosures and find that firms may be motivated to disclose non-GAAP numbers based on concerns about meeting analyst expectations and earnings volatility. Bhattacharya et al. (2003) and Lougee and Marquardt (2004) also find that incentives to meet or beat analyst forecasts can influence the disclosure of non-GAAP earnings measures. Isidro and Marques (2015) explore how institutional and economic factors are associated with the disclosure of non-GAAP measures that meet strategic benchmarks, while the associated GAAP earnings numbers fall short. Their results suggest that the use of non-GAAP numbers to meet or beat strategic performance targets is higher in countries with strong investor protection, efficient law enforcement, developed financial markets, and better communication and information dissemination. In addition, there is some evidence that investors and analysts appear to pay more attention to non-GAAP earnings than standard GAAP operating earnings, or even I/B/E/S actual earnings (Bhattacharya et al., 2003; Bradshaw & Sloan, 2002; Lougee & Marquardt, 2004).

These studies do not provide direct evidence that investors actually use manager-adjusted earnings numbers in their individual valuation models. Market participants could calculate their own core earnings metrics to use in their valuation analyses. To the extent, however, that investors use all information at their disposal to develop their unobservable core earnings metrics used for valuation purposes, we would expect that they make use of management-provided non-GAAP earnings measures, along with their corresponding reconciliations, in assessing firm performance. Stated differently, market participants obviously come to their own conclusions about the appropriate core earnings numbers to use in their individual valuations, but to the extent that managers provide information about what they determine to be appropriate adjustments to the GAAP earnings number, investors are likely to consider this information in assessing core performance. If investors are able to determine that some of managers’ recommended adjustments are overly aggressive or inappropriate, they may choose to discount or ignore these disclosures (Bhattacharya et al., 2003; Black et al., 2012; Frederickson & Miller, 2004; Marques, 2006). ⁵

These results illustrate the need to better understand manager-disclosed non-GAAP earnings. Even though Choi et al. (2007) find evidence suggesting that manager-disclosed non-GAAP metrics are generally intended to better reflect sustainable operating performance in the United Kingdom, Marques (2010); Brown, Christensen, and Elliott (2012); and Entwistle, Feltham, and Mbagwu (2005) find that some managers may disclose non-GAAP performance metrics for opportunistic reasons. Baumker et al. (2014) and Curtis et al. (2014) find that after Reg. G, some managers may opportunistically refrain from reporting non-GAAP earnings in the presence of transitory gains. Elliott (2006) provides experimental evidence that managers’ emphasis on non-GAAP numbers can influence non-professional investors’ decisions. Bhattacharya, Black, Christensen, & Mergenthaler (2007) and Allee, Bhattacharya, Black, & Christensen (2007) find consistent archival evidence, suggesting that less-sophisticated investors are likely to be misled by aggressive non-GAAP disclosures. Furthermore, Christensen, Drake, & Thornock (2014) find that short sellers target companies that issue non-GAAP earnings disclosures. Their evidence suggests that if less-sophisticated investors inappropriately rely on adjusted earnings numbers in deciding to invest in companies that issue these disclosures, their wealth could be transferred to short sellers. Overall, the evidence from prior research suggests that (a) non-GAAP earnings are more highly associated with market returns, (b) some firms may report non-GAAP earnings opportunistically, and (c) less-sophisticated market participants may be misled by aggressive non-GAAP measures.

The difficulty in analyzing non-GAAP earnings disclosures lies in distinguishing “truthful” disclosures in which managers attempt to focus investors’ attention on sustainable core earnings from “opportunistic” ones that may be motivated by managers’ desire to overstate their operating performance. While manager intent is not directly observable, Black and Christensen (2009) shed light on the issue by examining the intersection of two settings where opportunism in non-GAAP reporting is most likely: (a) when managers exclude recurring items and (b) when they use non-GAAP adjustments to achieve strategic earnings targets. In addition to the exclusion of one-time items like restructuring charges, they find that managers often exclude recurring expenses such as depreciation, research and development, and stock-based compensation to meet these strategic targets. They argue that the exclusion of recurring items is especially indicative of aggressive non-GAAP reporting. ⁶

One of the stated reasons for active interventions in non-GAAP reporting by the U.S. Congress and the SEC (SOX and Reg. G, respectively) was to curb potentially misleading non-GAAP disclosures. ⁷ However, Black and Christensen’s (2009) sample is mostly comprised of pre-SOX observations. We extend their evidence by directly examining the impact of SOX and Reg. G on the likelihood that firms engage in aggressive non-GAAP reporting. We focus on SOX and Reg. G rather than the warning issued by the SEC in 2001 because, as discussed by Heflin and Hsu (2008), the warning may not have discouraged firms from reporting non-GAAP measures. In addition, non-compliance with SOX and Reg. G can result in SEC enforcement actions, whereas before the aforementioned legal and regulatory changes, an enforcement action was predicated on the SEC’s proving an intent to mislead investors.

As managers have a better understanding and an insider view of their businesses, analysts and market participants are likely influenced by managers’ non-GAAP exclusions. ⁸ Consistent with this notion, prior research finds evidence that managers are able to influence analysts’ earnings estimates during the accounting period (Cotter, Tuna, & Wysocki, 2006; Matsumoto, 2002; Richardson, Teoh, & Wysocki, 2004) and also their street earnings reported in I/B/E/S at the end of the period (Christensen, Merkley, Tucker, & Venkataraman 2011; Black, Christensen, Kiosse, & Steffen, 2014). Therefore, based on prior evidence that managers influence analysts’ behavior, we argue that it is important to examine managers’ non-GAAP reporting choices.

On average, companies are likely to report non-GAAP earnings metrics to portray firm performance more accurately. However, it is also possible that some managers exercise their discretion in reporting non-GAAP numbers to overstate their firms’ performance. Ex ante, it is unclear how the strict non-GAAP reporting requirements imposed by SOX and Reg. G may have affected these two types of behavior. On one hand, Reg. G may dissuade managers from using non-GAAP disclosures to opportunistically paint a more favorable picture of company performance. On the other hand, some managers may view Reg. G as a “safe harbor” for disclosing adjusted earnings metrics. That is, as long as they are careful to follow all of the rules prescribed by Reg. G, managers may feel safe in providing these additional voluntary earnings disclosures in the post-SOX environment. There is some evidence (based on I/B/E/S actual earnings as a proxy for manager-disclosed non-GAAP earnings) that, in the post-SOX period, managers are less likely to exclude recurring items, that exclusions are of higher quality, and that managers are less likely to use non-GAAP exclusions to meet analysts’ forecasts (Heflin & Hsu, 2008; Kolev et al., 2008). However, Curtis et al. (2014) and Baumker et al. (2014) find evidence that some managers opportunistically avoid non-GAAP disclosures when earnings are increased by transitory gains.

While it is unclear which effect dominates, we expect that managers are generally less likely to disclose non-GAAP earnings aggressively after the enactment of SOX. To shed further light on this question, we also examine how the post-SOX regulatory environment has affected the use of specific aggressive non-GAAP exclusions. Because SOX mandated the implementation of Reg. G’s specific reconciliation requirement, we expect the frequency of some non-GAAP exclusions to change in the post-SOX period. Specifically, because managers are required to disclose each line item exclusion in the non-GAAP-to-GAAP earnings reconciliation, they may be more likely to exclude items that are truly one time in nature. They may also be less likely to exclude recurring items because they do not want to appear to be opportunistic or aggressive. An alternative non-mutually exclusive explanation is that changes in the exclusions endorsed by managers may be explained by the accounting rule changes that occurred during the sample period. Our baseline analyses therefore compare all pre-SOX non-GAAP reporters (firm quarters ending through July 2002) with all post-SOX non-GAAP reporters (firm quarters ending after July 2002). We expect that firms report non-GAAP earnings less aggressively in the post-SOX period, leading to our first hypothesis (stated in the alternative form):

Kolev et al. (2008) find evidence suggesting that firms that stopped reporting non-GAAP earnings immediately after SEC intervention had lower quality analyst earnings exclusions in the pre-intervention period. However, they also find evidence consistent with a decrease in the quality of special item exclusions in the post-intervention period. Curtis et al. (2014) and Baumker et al. (2014) also document opportunistic non-GAAP behavior after Reg. G. Thus, there is mixed evidence about the impact of regulation on non-GAAP reporting. Based on the results of prior studies and also on our own untabulated univariate results suggesting that firms reporting non-GAAP earnings in both periods are systematically different from firms reporting non-GAAP earnings in either the pre- or the post-SOX period, we split the full sample into three subgroups: (a) continuer firms that report non-GAAP earnings in both the pre- and post-SOX periods, (b) stopper firms that report non-GAAP numbers only in the pre-SOX period, and (c) starter firms that report non-GAAP numbers only in the post-SOX period. We compare the behavior of the continuer firms pre- and post-SOX, and we also compare stopper firms with starter firms. ⁹

The continuer sample includes firms that report non-GAAP earnings numbers in both the pre- and the post-SOX periods, which enables us to hold firm-specific characteristics constant and focus on the effects of SOX and Reg. G on aggressive non-GAAP disclosures. Because these firms disclose non-GAAP earnings in both periods, we assume that they do not view the change in regulation as a threat to their choice to voluntarily report adjusted earnings metrics. While it is possible that some continuer firms may have become more conservative in their exclusion choices after the regulatory change, the fact that these firms disclose non-GAAP earnings figures in both periods suggests that the fear of regulatory costs did not cause them to cease disclosing adjusted earnings metrics. We posit that continuer firms are less likely to engage in behavior that might be viewed as aggressive in the post-SOX environment because it could result in regulatory costs and a loss of reputation. Thus, our next hypothesis (stated in the alternative form) predicts a decrease in the likelihood of aggressive non-GAAP reporting by continuer firms in the post-SOX period relative to the pre-SOX period:

It is not entirely clear whether firms that begin disclosing non-GAAP earnings in the post-SOX period are more or less likely to disclose adjusted earnings numbers for aggressive reasons than firms that stopped reporting non-GAAP earnings after SOX. As stated previously, we believe that post-SOX “starter” firms may provide non-GAAP disclosures in the post-SOX period because they perceive that the provisions of SOX and Reg. G create a type of safe harbor. In other words, as long as firms follow the guidelines outlined in Reg. G, they can avoid the threat of litigation and SEC sanctions related to non-GAAP disclosures. ¹⁰ We also separately explore firms that provided non-GAAP disclosures in the pre-SOX period, but ceased doing so after SOX (“stoppers”). The fact that stopper firms cease reporting non-GAAP numbers in the post-SOX period possibly provides ex post evidence that these managers felt their disclosures were potentially aggressive or that continuing to disclose such numbers could lead to regulatory (or criminal) action. However, we acknowledge that an alternative explanation may be that some firms appear in the stopper (starter) group because they only reported non-GAAP earnings during the pre-SOX (post-SOX) period as a result of large, one-time charges that they chose to exclude. Firms that only report adjusted earnings metrics sporadically because of the occurrence of a legitimate “one-time” event may be in these groups for logical reasons (other than reacting to the regulation of non-GAAP reporting). Nevertheless, Kolev et al. (2008), using I/B/E/S-adjusted earnings metrics as a proxy for manager-disclosed non-GAAP earnings, infer that firms that stopped reporting non-GAAP numbers in the post-SOX period had lower quality exclusions pre-SOX. Therefore, we expect that on average, stopper firms are more likely to exhibit aggressiveness in non-GAAP reporting than starter firms:

Sample Selection

We collect press releases containing non-GAAP earnings disclosures by searching the PR Newswire and Business Wire services on LexisNexis for the years 1998-2006. Our sample selection starts with all non-GAAP announcements in which the company discloses a non-GAAP diluted EPS figure that differs from the GAAP EPS number. Our original search uses the keywords “pro forma,”“pro-forma,” and “proforma.” However, companies often use other terms to describe their non-GAAP earnings figures. Wallace (2002) performs a detailed categorization of non-GAAP earnings nomenclatures used by companies. Based on Wallace’s (2002) categorization of adjusted earnings nomenclatures commonly used by firms, we use an expanded search string and retrieved 106,638 potential press releases. ¹¹ After carefully reading each press release, we find 24,018 releases containing an adjusted quarterly pro forma earnings number in addition to the GAAP number. Our analyses require financial statement information from Compustat, analyst information from I/B/E/S, and returns from Center for Research in Security Prices (CRSP). Moreover, we require complete data related to adjustment types. These requirements reduce our sample to 9,326 quarterly observations. To enhance comparability across our subsamples of continuers, starters, and stoppers, we subsequently delete firms that reported non-GAAP earnings only once in either the pre- or post-SOX period. ¹² The final sample includes 8,136 firm-quarter observations (2,797 observations in the pre-SOX period, which includes all observations through the second quarter of 2002 and 5,339 post-SOX observations from the third quarter of 2002 through December 2006) for which the firm discloses non-GAAP earnings numbers. We then split the full sample into continuers (4,467 observations), stoppers (947 observations), and starters (2,722 observations).

Variable Descriptions

Exclusion measures

As illustrated in Figure 1 (adapted from Christensen, 2007, and Brown, Christensen, Elliott, & Mergenthaler, 2012), we calculate total manager exclusions (MGREXCL_TOTAL), including special items and analysts’ exclusions, as the difference between EPS_Non-GAAP and GAAP EPS before extraordinary items (EPS_BefExt). We then decompose MGREXCL_TOTAL into the following components: (1) analysts’ total exclusions, calculated as EPS_I/B/E/S minus EPS_BefExt (ANALYSTEXCL_TOTAL); and (2) managers’ incremental exclusions of recurring items (MGREXCL_INCREMENTAL), calculated as EPS_Non-GAAP minus EPS_I/B/E/S.

OPEN IN VIEWER

Figure 1.

Illustration of the calculation of exclusion variables.

Note that positive values of the earnings exclusion variables indicate that income-decreasing items (such as expenses and losses) are excluded from the respective GAAP earnings figures, whereas negative values indicate that income-increasing items (such as one-time revenues or gains) are excluded. For example, positive (negative) values of MGREXCL_INCREMENTAL indicate that managers excluded more (less) recurring expenses than analysts. We classify both below-the-line items and special items as one-time or non-recurring in nature. Therefore, any additional exclusions by analysts and/or managers are generally recurring operating items such as depreciation, amortization, and R&D expenses (i.e., recurring items). ¹⁴

Descriptive Statistics

Table 1 provides descriptive statistics for the variables of interest in our analyses. In Panel A, we compare all pre-SOX non-GAAP reporters with all post-SOX non-GAAP reporters. We find that EPS numbers are higher, on average, in the post-SOX period, reflecting general trends in the economy. In particular, there was a major economic downturn in 2000-2002 but economic growth increased from 2003-2006, our post-SOX period. We find similar results when comparing EPS numbers of pre-SOX continuers with those of post-SOX continuers (Panel B) and in comparing pre-SOX stoppers with post-SOX starters (Panel C). When we examine differences in exclusions for the full sample, we find that, on average, managers’ and analysts’ exclusions decrease during the post-SOX period (variables MGREXCL_TOTAL, ANALYSTEXCL_TOTAL, and MGREXCL_INCREMENTAL in Panel A). ¹⁸ These three exclusion measures also decrease on average when we compare (a) continuers in the pre- and post-SOX periods (Panel B) and (b) pre-SOX stoppers with post-SOX starters (Panel C). This evidence suggests that pre-SOX non-GAAP reporters (both continuers and stoppers) were more likely to exclude more earnings components (or earnings components of greater magnitude) than post-SOX non-GAAP reporters (both continuers and starters). These results are consistent with a reduction in overall exclusions, by both managers and analysts, due to the enactment of SOX.

OPEN IN VIEWER

Table 1.

Descriptive Statistics.

Panel A: Full Sample.
Variable	Pre-SOX (n = 2,797)				Post-SOX (n = 5,339)
	Q1	M	Median	Q3	Q1	M	Median	Q3	Difference	t statistic
EPSNon-GAAP	0.010	0.216	0.150	0.390	0.100	0.355	0.280	0.520	−0.139	−14.375***
EPSI/B/E/S	0.000	0.203	0.150	0.370	0.100	0.349	0.270	0.510	−0.146	−15.272***
EPSGAAP-OP	−0.110	0.077	0.100	0.360	0.060	0.314	0.240	0.490	−0.237	−12.039***
EPSBefExt	−0.190	−0.033	0.050	0.320	0.020	0.233	0.200	0.460	−0.266	−11.569***
EPSNet	−0.200	−0.052	0.050	0.310	0.020	0.234	0.200	0.460	−0.286	−11.750***
MGREXCLTOTAL	0.030	0.249	0.080	0.210	0.010	0.122	0.040	0.120	0.127	6.185***
ANALYSTEXCLTOTAL	0.000	0.235	0.060	0.200	0.000	0.116	0.030	0.110	0.119	5.819***
MGREXCLINCREMENTAL	0.000	0.013	0.000	0.000	0.000	0.006	0.000	0.000	0.007	8.171***
TOTASSET	228.676	6735.794	756.913	2676.753	417.319	10849.566	1341.199	4504.645	−4113.772	−3.363***

Panel B: Continuer Subsample.
Variable	Pre-SOX				Post-SOX
	Continuers (n = 1,850)				Continuers (n = 2,617)
	Q1	M	Median	Q3	Q1	M	Median	Q3	Difference	t statistic
EPSNon-GAAP	0.020	0.217	0.150	0.350	0.070	0.323	0.240	0.470	−0.106	−8.589***
EPSI/B/E/S	0.020	0.205	0.150	0.340	0.080	0.320	0.240	0.470	−0.115	−9.374***
EPSGAAP-OP	−0.090	0.097	0.100	0.320	0.040	0.278	0.200	0.450	−0.181	−9.500***
EPSBefExt	−0.160	−0.046	0.050	0.280	0.010	0.196	0.170	0.410	−0.242	−6.954***
EPSNet	−0.170	−0.055	0.050	0.270	0.000	0.197	0.170	0.420	−0.252	−7.165***
MGREXCLTOTAL	0.030	0.263	0.080	0.220	0.010	0.127	0.040	0.120	0.136	4.205***
ANALYSTEXCLTOTAL	0.010	0.251	0.060	0.210	0.000	0.124	0.040	0.110	0.127	3.947***
MGREXCLINCREMENTAL	0.000	0.012	0.000	0.000	0.000	0.004	0.000	0.000	0.008	7.128***
TOTASSET	282.226	7040.632	854.695	2537.724	464.289	9649.634	1465.927	5315.867	−2609.002	−2.301*

Panel C: Stopper/Starter Subsample.
Variable	Pre-SOX				Post-SOX
	Stoppers (n = 947)				Starters (n = 2,722)
	Q1	M	Median	Q3	Q1	M	Median	Q3	Difference	t statistic
EPSNon-GAAP	−0.050	0.214	0.160	0.450	0.130	0.385	0.310	0.560	−0.171	−10.812***
EPSI/B/E/S	−0.050	0.198	0.140	0.430	0.130	0.377	0.300	0.560	−0.179	−11.354***
EPSGAAP-OP	−0.160	0.037	0.110	0.420	0.080	0.350	0.280	0.540	−0.313	−7.865***
EPSBefExt	−0.250	−0.006	0.050	0.400	0.050	0.268	0.240	0.500	−0.274	−9.914***
EPSNet	−0.270	−0.044	0.050	0.400	0.040	0.270	0.240	0.500	−0.314	−9.495***
MGREXCLTOTAL	0.020	0.220	0.080	0.190	0.010	0.117	0.040	0.120	0.103	4.789***
ANALYSTEXCLTOTAL	0.000	0.204	0.040	0.190	0.000	0.109	0.030	0.110	0.095	4.413***
MGREXCLINCREMENTAL	0.000	0.016	0.000	0.010	0.000	0.008	0.000	0.000	0.008	5.292***
TOTASSET	165.169	6140.281	541.549	2964.915	388.667	12003.211	1264.560	4036.534	−5862.930	−2.345*

Note. SOX = Sarbanes–Oxley Act; EPS_Non-GAAP = manager-adjusted GAAP earnings per share from the press release; EPS_I/B/E/S = I/B/E/S unadjusted actual earnings per share; EPS_GAAP-OP = Compustat diluted operating earnings per share; EPS_BefExt = Compustat diluted EPS before extraordinary items; EPS_Net = Compustat diluted EPS; MGREXCL_TOTAL = managers’ total above-the-line exclusions (EPS_Non-GAAP−EPS_BefExt); ANALYSTEXCL_TOTAL = analysts’ total above-the-line exclusions (EPS_I/B/E/S−EPS_BefExt); MGREXCL_INCREMENTAL = managers’ incremental above-the-line exclusions beyond those made by analysts (EPS_Non-GAAP−EPS_I/B/E/S); TOTASSET = total assets in US$ millions at the end of the quarter; I/B/E/S = Institutional Brokers’ Estimate System.

*

Significant at the .10 level. **Significant at the .05 level. ***Significant at the .01 level.

Furthermore, in untabulated analyses, we examine the industry breakdown of pre/post-SOX firms by one-digit Standard Industrial Classification (SIC) code by restricting the sample to one quarter per firm in each period (pre- and post-SOX) to remove the potential effects of some firms influencing the industry analyses by reporting non-GAAP figures many times in the sample. In other words, if some firms report non-GAAP earnings every quarter, this could result in overrepresentation of those firms’ industries. The single-firm analysis indicates a significant difference in industry breakdown when comparing (a) all pre-SOX firms with all post-SOX firms and (b) pre-SOX stoppers with post-SOX starters. However, as expected, there is no difference in the industry breakdown when comparing pre-SOX continuers with post-SOX continuers because these firms are the same in both periods. ¹⁹

Research Design

Multivariate regression analyses

Our main tests of H1, H2, and H3 examine the association between our two aggressive behavior proxies and the particular adjustments firms use to achieve these objectives. Specifically, for each subsample, we follow Black and Christensen (2009) and regress each of our measures of aggressive reporting (INCREXCL and STREETBEAT) on indicator variables representing recurring non-GAAP exclusions. This approach allows us to examine the extent to which specific exclusions are used in an attempt to portray firm performance in a positive light. ²⁰

INCREXCL = γ 0 + γ 1 R & D + γ 2 AMORT _ DEP + γ 3 STOCK _ COMP + γ 4 INTEREST + γ 5 TAX + γ 6 LNSIZE + γ 7 CONTROLS + ε ,

S T R E E T B E A T = τ 0 + τ 1 R & D + τ 2 A M O R T _ D E P + τ 3 S T O C K _ C O M P + τ 4 I N T E R E S T + τ 5 T A X + τ 6 L N S I Z E + τ 7 C O N T R O L S + ε ,

where INCREXCL is an indicator variable coded 1 if a firm’s incremental manager exclusions are greater than 0, STREETBEAT is an indicator variable coded 1 if EPS_I/B/E/S falls short of analysts’ expectations, while EPS_Non-GAAP meets or beats expectations, R&D is an indicator variable coded 1 if managers exclude research and development (R&D) costs, AMORT_DEP is an indicator variable coded 1 if managers exclude amortization or depreciation expense, STOCK_COMP is an indicator variable coded 1 if managers exclude stock-based compensation costs (excluding amortization of stock-based compensation), INTEREST is an indicator variable coded 1 if managers exclude interest-related costs, TAX is an indicator variable coded 1 if managers exclude tax-related charges, and LNSIZE is the natural log of total assets.

As with the univariate tests, our regression analyses compare (a) all pre-SOX with all post-SOX observations, (b) pre-SOX continuers with post-SOX continuers, and (c) pre-SOX stoppers with post-SOX starters. We first perform the regression analyses separately for each time period and subsample. That is, we estimate each of the regression models on six separate subsamples: all pre-SOX observations, all post-SOX observations, pre-SOX continuers, post-SOX continuers, pre-SOX stoppers, and post-SOX starters. Then, to make statistical comparisons, we add a POSTSOX indicator variable (equal to 1 if the observation occurs in the post-SOX period) to each regression model and interact POSTSOX with each non-GAAP exclusion variable. For each aggressive non-GAAP exclusion measure, we estimate this expanded model on the full sample and two subsamples: only continuers and only stoppers/starters.

Our purpose in these analyses is to determine whether managers are less likely to make aggressive recurring item adjustments in calculating their non-GAAP earnings metrics in the post-SOX period. The sign and magnitude of the coefficients provide information about the degree to which each adjustment type increases or decreases the likelihood that managers demonstrate one of the two types of aggressive behavior we examine. Our focus is on whether managers exclude recurring items to achieve these aggressive objectives. Following prior studies, all regression models control for the number of analysts following the firm and forecast dispersion (Chevis, Das, & Sivaramakrishnan, 2001; Heflin & Hsu, 2008). Firms followed by a larger number of analysts may have greater incentives to meet or beat analyst forecasts. We measure analyst following as the number of analysts following the firm during the quarter from I/B/E/S. Chevis et al. (2001) suggest that a meet or beat strategy may be easier to implement when analyst forecast dispersion is lower. Therefore, we also include a control for forecast dispersion, measured as the standard deviation of analyst forecasts over the quarter as reported in I/B/E/S.

Furthermore, we include earnings quality as an additional control following Heflin and Hsu (2008). We measure earnings quality using the quarterly e-loading measure used by Ecker, Francis, Kim, Olsson, and Schipper (2006) using CRSP and the Accruals Quality (AQ) factor data from Frank Ecker’s website. We also control for earnings guidance (which we collect from First Call), an indicator variable equal to 1 if managers issue earnings guidance during the quarter and 0 otherwise, following Brown and Higgins (2005) and Christensen et al. (2011). Other control variables include leverage, defined as total liabilities divided by common equity over the quarter (Chevis et al., 2001; Heflin & Hsu, 2008) and book-to-market computed as common equity divided by market value of equity over the quarter, using Compustat data (Heflin & Hsu, 2008). We follow Heflin and Hsu (2008) and include an indicator variable equal to 1 if the fiscal quarter is the fourth quarter. Finally, we include industry controls.

Univariate Results

We first test for changes in the frequencies of our aggressive non-GAAP exclusion measures (INCREXCL and STREETBEAT) after SOX. Table 2 presents direct univariate tests of H1, H2, and H3. Overall, the results provide evidence consistent with the notion that the frequency of aggressive non-GAAP reporting decreased in the post-SOX period. Panel A compares all pre-SOX non-GAAP reporters with all post-SOX non-GAAP reporters for our measures of aggressiveness: (a) the existence of incremental recurring exclusions beyond what analysts exclude (INCREXCL) and (b) using additional exclusions to move from missing analysts’ expectations based on I/B/E/S earnings to meeting or beating expectations based on a non-GAAP basis (STREETBEAT). In addition, it compares the exclusion of specific recurring items in the pre- and post-SOX periods. We find a significantly lower frequency of non-GAAP reporters in the post-SOX period that exclude recurring items, which are incremental to what analysts exclude: 17.1% versus 21.3%. Similarly, the percentage of firms that are able to meet or beat analyst forecasts based on non-GAAP earnings but not based on the I/B/E/S actual EPS number decreases from 4.9% to 4.3%, but the difference is not statistically significant. We find similar results in Panel B for the continuer sample and in Panel C comparing pre-SOX stoppers with post-SOX starters, except that the decrease in STREETBEAT is marginally significant in Panel C.

OPEN IN VIEWER

Table 2.

Pre/Post-SOX Comparisons of Strategic Benchmark Measures and Non-GAAP Exclusion Frequencies.

Panel A: Full Sample.
	Pre-SOX (n = 2,797)		Post-SOX (n = 5,339)
	Frequency	%	Frequency	%	Difference (%)	z score
Aggressive reporting measures
INCREXCL	595	21.3	912	17.1	4.2	4.62***
STREETBEAT	138	4.9	231	4.3	0.6	1.25
Recurring item exclusions
R&D	209	7.5	224	4.2	3.3	6.25***
AMORT_DEP	1,406	50.3	1,751	32.8	17.5	15.36***
STOCK_COMP	759	27.1	1,719	32.2	−5.1	−4.71***
INTEREST	25	0.9	144	2.7	−1.8	−5.42***
TAX	354	12.7	940	17.6	−4.9	−5.80***

Panel B: Continuer Subsample.
	Pre-SOX (n = 1,850)		Post-SOX (n = 2,617)
	Frequency	%	Frequency	%	Difference (%)	z score
Aggressive reporting measures
INCREXCL	339	18.3	365	13.9	4.4	3.95***
STREETBEAT	79	4.3	101	3.9	0.4	0.69
Recurring item exclusions
R&D	171	9.2	153	5.8	3.4	4.31***
AMORT_DEP	962	52.0	1,108	42.3	9.7	6.38***
STOCK_COMP	531	28.7	866	33.1	−4.4	−3.12***
INTEREST	21	1.1	70	2.7	−1.6	−3.59***
TAX	278	15.0	484	18.5	−3.5	−3.03***

Panel C: Stopper/Starter Subsample.
	Pre-SOX (n = 947)		Post-SOX (n = 2,722)
	Frequency	%	Frequency	%	Difference (%)	z score
Aggressive reporting measures
INCREXCL	256	27.0	547	20.1	6.9	4.45***
STREETBEAT	59	6.2	130	4.8	1.4	1.74*
Recurring item exclusions
R&D	38	4.0	71	2.6	1.4	2.19**
AMORT_DEP	444	46.9	643	23.6	23.3	13.50***
STOCK_COMP	228	24.1	853	31.3	−7.2	−4.22***
INTEREST	4	0.4	74	2.7	−2.3	−4.22***
TAX	76	8.0	456	16.8	−8.8	−6.57***

Note. SOX = Sarbanes–Oxley Act; INCREXCL = 1 if MGREXCL_INCREMENTAL is greater than 0; STREETBEAT = 1 if EPS_Non-GAAP is greater than or equal to the mean analyst forecast, while EPS_I/B/E/S is less than the mean analyst forecast; R&D = research and development (R&D) costs; AMORT_DEP = depreciation and amortization costs (excluding amortization of stock-based compensation); STOCK_COMP = stock-based compensation costs; INTEREST = interest-related items; TAX = tax-related items; I/B/E/S = Institutional Brokers’ Estimate System.

*

Significant at the .10 level. **Significant at the .05 level. ***Significant at the .01 level.

In addition, we examine the exclusion of specific recurring items in the pre- and post-SOX periods. While the exclusion of one-time items can be a way for a firm to disclose a measure of permanent earnings that is likely to be sustainable in future periods, the exclusion of recurring expenses is less justifiable and is generally regarded to be more aggressive (and thus potentially misleading to investors seeking to forecast future operating performance). Hence, our analyses focus on the following recurring items: R&D expense, amortization and depreciation, stock compensation, interest, and tax charges.

We first provide visual evidence. Figure 2 compares the percentage of observations excluding each type of recurring item. Panel A makes comparisons for the full sample, whereas Panel B makes comparisons for the continuer sample, and Panel C compares pre-SOX stoppers with post-SOX starters. In each case, we find that both R&D and amortization and depreciation decrease from the pre- to the post-SOX periods. However, stock-based compensation, interest, and tax expense exclusions increase from pre- to post-SOX. Figure 3 plots the frequency of each recurring item exclusion for the full sample over time. By superimposing all five exclusion types in the same line chart, it is easy to see how each type of exclusion has changed over our sample period. This figure indicates, for the full sample, exactly when the relative frequencies of each recurring exclusion type changed over time. Interestingly, while the exclusion of depreciation and amortization expenses peaked at 62% in 2001, it declined dramatically after the enactment of SOX in 2002. It also illustrates how the exclusion of stock-based compensation expense increased dramatically after the implementation of SFAS 123R in 2005.

OPEN IN VIEWER

Figure 2.

Non-GAAP exclusion frequencies pre/post-SOX.

OPEN IN VIEWER

Figure 3.

Recurring exclusion types as a percentage of total exclusions.

Table 2 tests whether these shifts in the relative frequencies of recurring exclusions in each category from the pre- to the post-SOX period are statistically significant. The z-score statistics indicate that all changes are statistically significant. Moreover, the results are virtually identical across comparisons in all three panels. Specifically, we find that non-GAAP earnings in the post-SOX period exclude two recurring items less frequently: (a) R&D costs and (b) amortization and depreciation costs, consistent with H1. It should be noted that the decrease in the frequency in which amortization and depreciation costs are excluded could be due to the changes imposed by SFAS 142 (which occurred around the same period as SOX). SFAS 142 eliminates the amortization requirement for goodwill and other indefinite-lived intangibles. Contrary to H1, non-GAAP disclosures in the post-SOX period have higher frequencies of the following recurring exclusions: (a) stock-based compensation expense, (b) interest expense, and (c) tax-related expenses. However, SFAS 123R, which became effective for reporting periods after June 15 2005, requires firms to recognize stock compensation in the income statement. Hence, it is likely that the increased exclusion of stock-based compensation is attributable to this change in standards. Consistent with this result, Barth, Gow, and Taylor (2012) find that stock-compensation expense is a common exclusion in the post-SOX period.

Overall, we find mixed evidence for H1 based on changes in the exclusion of specific recurring exclusions in the post-SOX period. However, we note that the frequency of tax and interest expense exclusions are relatively low, and the increase in stock-based compensation does not appear to be related to increased aggressiveness, but simply the change in accounting standards. Hence, our results are generally consistent with a decrease in aggressiveness in the post-SOX period.

Multivariate Regression Results

We further test H1, H2, and H3 using multivariate analyses. Specifically, we examine whether managers engage in aggressive non-GAAP reporting by excluding recurring items above and beyond what analysts exclude in the post-SOX relative to the pre-SOX period. We use our base Models 1 and 2 and their expanded versions with POSTSOX interactions to investigate how the regulatory changes imposed by SOX and Reg. G affect the extent to which specific types of exclusions are associated with each of our indicators of aggressive non-GAAP reporting. Table 3 reports the multivariate analyses pertaining to H1. Panel A (Panel B) presents the results of estimating Models 1 and 2 on all pre-SOX (post-SOX) observations. We present these period-specific estimations for completeness, but we focus on the results in Panel C to statistically test H1. Panel C provides results from estimating Models 1 and 2 on the entire sample, but augmenting each model with a POSTSOX indicator variable and its interaction with each recurring exclusion indicator variable. The estimated coefficients on these interaction terms constitute our formal tests of H1.

OPEN IN VIEWER

Table 3.

Full Sample Analysis: Exclusions Used for Strategic Non-GAAP Earnings Benchmarks.

Panel A: All Pre-SOX Observations (n = 2,797).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.168	2.155**	0.072	2.478**
R&D	−0.042	−1.673*	0.000	0.001
AMORT_DEP	0.214	10.024***	0.032	3.425***
STOCK_COMP	−0.043	−2.242**	−0.023	−2.197**
INTEREST	0.024	0.370	−0.037	−4.526***
TAX	−0.042	−2.360**	−0.008	−0.747
LNSIZE	0.018	2.017**	−0.003	−0.748
CONTROLS	Yes		Yes
Adjusted R2	.298		.020

Panel B: All Post-SOX Observations (n = 5,339).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.179	3.010***	0.060	2.382**
R&D	−0.069	−3.342***	−0.011	−0.914
AMORT_DEP	−0.003	−0.162	0.006	0.670
STOCK_COMP	0.229	11.507***	0.038	4.560***
INTEREST	−0.005	−0.113	0.012	0.637
TAX	−0.017	−1.126	0.005	0.645
LNSIZE	−0.001	−0.114	−0.002	−0.616
CONTROLS	Yes		Yes
Adjusted R2	.116		.020

Panel C: All Observations (N = 8,136).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.171	3.330***	0.066	3.158***
R&D	−0.066	−2.581***	0.001	0.080
AMORT_DEP	0.235	11.672***	0.034	3.780***
STOCK_COMP	−0.055	−2.689***	−0.020	−1.973**
INTEREST	0.006	0.091	−0.037	−5.168***
TAX	−0.060	−3.389***	−0.007	−0.614
R&D×POSTSOX	0.012	0.352	−0.013	−0.730
AMORT_DEP×POSTSOX	−0.240	−9.275***	−0.029	−2.360**
STOCK_COMP×POSTSOX	0.299	11.036***	0.058	4.677***
INTEREST×POSTSOX	−0.007	−0.091	0.049	2.439**
TAX×POSTSOX	0.051	2.213**	0.012	0.959
POSTSOX	−0.026	−1.671*	−0.006	−0.911
LNSIZE	0.008	1.375	−0.002	−0.984
CONTROLS	Yes		Yes
Adjusted R2	.173		.019

Note. The full sample includes 8,136 observations: 2,797 pre-SOX and 5,339 post-SOX. p values are based on heteroskedasticity robust standard errors clustered by firm; SOX = Sarbanes–Oxley Act; POSTSOX = 1 if the observation occurs in the post-SOX period (all quarters from the third quarter of 2002 through December 2006); LNSIZE = natural log of total assets from the respective quarter. All other variables defined in Tables 1 and 2.

*

Significant at the .10 level. **Significant at the .05 level. ***Significant at the .01 level.

When we estimate Models 1 and 2 using the full sample with interaction terms (Table 3, Panel C), we focus on changes in the association between excluding recurring items and indicators of aggressive non-GAAP reporting. We first summarize the results of Model 1. Specifically, we investigate the significance of interactions between recurring item indicator variables and the POSTSOX indicator variable in these models. The significantly negative coefficient on the AMORT_DEP×POSTSOX interaction term indicates that managers are less likely in the post-SOX period to exclude amortization and depreciation when analysts do not, consistent with H1. This result may also be influenced by the implementation of SFAS 142 in June 2001. The significantly positive coefficient on the STOCK_COMP×POSTSOX interaction term suggests that managers are more likely to exclude stock-based compensation when analysts do not. ²¹ Finally, the coefficient on the TAX×POSTSOX interaction term is also significantly positive at the 5% level. When we perform separate analyses for the pre- and the post-SOX periods, we find that in the pre-SOX period (Panel A) managers were less likely to exclude tax-related items when analysts did not. However, this relation does not hold as it is not significant in the post-SOX period (Panel B).

Model 2 explores our second measure of aggressive non-GAAP reporting (STREETBEAT): the use of non-GAAP exclusions to meet or beat analysts’ consensus forecast when the I/B/E/S actual EPS number falls short. The significantly negative coefficient on the AMORT_DEP×POSTSOX interaction term indicates that managers are less likely in the post-SOX period to exclude amortization and depreciation when analysts do not to meet/beat analysts’ expectations, consistent with H1. However, the significantly positive coefficient on the STOCK_COMP×POSTSOX interaction term suggests that in the post-SOX period managers are more likely to exclude stock-based compensation when analysts do not to meet expectations. We also find a significantly positive coefficient on the INTEREST×POSTSOX interaction term. When we perform separate analyses for the pre- and post-SOX periods, we find that while the coefficient on INTEREST is significantly negative in the pre-SOX period (Panel A), it is not statistically significant in the post-SOX period (Panel B).

Table 4 follows the same format as Table 3, except that it presents results for our tests of H2 for continuer firms. As with Table 3, we focus on Panel C of Table 4, which presents the results for Models 1 and 2 in which the exclusion variables are interacted with the POSTSOX indicator variable. The results for Model 1 indicate that, consistent with H2, managers are less likely to exclude amortization and depreciation in the post-SOX period when analysts are not willing to do so. However, they are slightly more likely after SEC intervention to exclude stock-based compensation. The results from estimating Model 2 investigate managers’ propensity in the post-SOX period to exclude incremental recurring earnings components to meet or beat analysts’ expectations when I/B/E/S actual EPS would otherwise fall short. The results suggest that managers are more likely in the post-SOX period to exclude stock-based compensation-related costs when analysts do not to meet/beat analysts’ expectations. We do not find evidence that any of the other recurring items are more or less likely to be excluded by managers for continuers in the post-SOX period.

OPEN IN VIEWER

Table 4.

Continuer Subsample Analysis: Exclusions Used for Strategic Non-GAAP Earnings Benchmarks.

Panel A: Pre-SOX Continuers (n = 1,850).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.099	1.133	0.042	1.342
R&D	−0.010	−0.357	0.010	0.613
AMORT_DEP	0.181	7.293***	0.028	2.721***
STOCK_COMP	−0.041	−1.891*	−0.019	−1.599
INTEREST	0.007	0.105	−0.036	−3.314***
TAX	−0.047	−2.488**	−0.007	−0.559
LNSIZE	0.018	1.932*	−0.000	−0.040
CONTROLS	Yes		Yes
Adjusted R2	.282		.024

Panel B: Post-SOX Continuers (n = 2,617).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	−0.030	−0.418	0.003	0.085
R&D	−0.035	−1.620	−0.011	−0.882
AMORT_DEP	0.032	1.243	0.016	1.367
STOCK_COMP	0.143	5.801***	0.018	1.629
INTEREST	−0.036	−0.585	−0.010	−0.476
TAX	−0.028	−1.417	−0.004	−0.353
LNSIZE	0.016	1.889*	0.004	0.968
CONTROLS	Yes		Yes
Adjusted R2	.148		.020

Panel C: All Continuers (n = 4,467).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.018	0.292	0.018	0.684
R&D	−0.021	−0.741	0.010	0.606
AMORT_DEP	0.188	8.041***	0.028	2.848***
STOCK_COMP	−0.041	−1.902*	−0.016	−1.437
INTEREST	0.004	0.062	−0.033	−4.144***
TAX	−0.053	−2.771***	−0.005	−0.433
R&D×POSTSOX	−0.006	−0.170	−0.021	−1.204
AMORT_DEP×POSTSOX	−0.159	−5.297***	−0.015	−1.005
STOCK_COMP×POSTSOX	0.191	6.202***	0.034	2.387**
INTEREST×POSTSOX	−0.034	−0.402	0.025	1.120
TAX×POSTSOX	0.028	1.039	0.002	0.153
POSTSOX	−0.022	−1.069	−0.004	−0.449
LNSIZE	0.019	2.572**	0.002	0.726
CONTROLS	Yes		Yes
Adjusted R2	.207		.022

Note. The continuer sample includes 4,467 observations: 1,850 pre-SOX and 2,617 post-SOX. p values are based on heteroskedasticity robust standard errors clustered by firm; SOX = Sarbanes–Oxley Act; POSTSOX = 1 if the observation occurs in the post-SOX period (all quarters from the third quarter of 2002 through December 2006); LNSIZE = natural log of total assets from the respective quarter. All other variables defined in Tables 1 and 2.

*

Significant at the .10 level. **Significant at the .05 level. ***Significant at the .01 level.

Taken together, the results presented in Table 4 indicate that managers are less likely in the post-SOX period to utilize certain recurring exclusions such as amortization and depreciation to meet their opportunistic motivations, while there is evidence that managers are more likely to exclude stock-based compensation in the post-SOX period to achieve their aggressive targets. Overall, these results are consistent with the notion that firms that reported non-GAAP numbers in both the pre- and post-SOX periods perceive net benefits from non-GAAP reporting, and they are therefore less likely to engage in overly aggressive non-GAAP reporting in the post-SOX period.

Similar to the previous two tables, Panel C of Table 5 presents our tests of H3 on the stopper/starter subsample of firms. The results of Model 1 indicate that, consistent with H3, managers are less likely to exclude amortization and depreciation in the post-SOX period when analysts are not willing to do so. However, they are slightly more likely after SEC intervention to exclude R&D expense (which is marginally significant). In addition, they are more likely to exclude stock-based compensation post-SOX. Model 2 investigates managers’ propensity to exclude earnings components that analysts do not exclude to meet or beat analysts’ expectations. Consistent with H3, the results indicate that managers are less likely in the post-SOX period to exclude amortization and depreciation (marginally significant). Contrary to H3, the results indicate that managers are more likely in the post-SOX period to exclude both interest and stock-based compensation to meet analysts’ expectations. ²²

OPEN IN VIEWER

Table 5.

Stopper/Starter Subsample Analysis: Exclusions Used for Strategic Non-GAAP Earnings Benchmarks.

Panel A: Pre-SOX Stoppers (n = 947).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.366	2.690***	0.141	2.150**
R&D	−0.148	−3.117***	−0.035	−1.160
AMORT_DEP	0.278	7.123***	0.042	2.180**
STOCK_COMP	−0.034	−0.841	−0.031	−1.527
INTEREST	0.171	0.948	−0.043	−2.315**
TAX	−0.023	−0.524	−0.018	−0.684
LNSIZE	0.011	0.587	−0.008	−1.004
CONTROLS	Yes		Yes
Adjusted R2	.325		.004

Panel B: Post-SOX Starters (n = 2,722).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.306	3.669***	0.105	3.141***
R&D	−0.099	−2.377**	−0.006	−0.245
AMORT_DEP	−0.028	−0.986	−0.001	−0.074
STOCK_COMP	0.306	10.629***	0.057	4.608***
INTEREST	0.027	0.498	0.034	1.128
TAX	−0.015	−0.648	0.012	0.901
LNSIZE	−0.007	−0.781	−0.006	−1.246
CONTROLS	Yes		Yes
Adjusted R2	.139		.026

Panel C: All Stoppers/Starters (n = 3,669).
Variable	Model (1) INCREXCL		Model (2) STREETBEAT
	Coefficient	t statistic	Coefficient	t statistic
INTERCEPT	0.336	4.539***	0.125	4.034***
R&D	−0.205	−4.335***	−0.030	−1.037
AMORT_DEP	0.318	8.461***	0.041	2.315**
STOCK_COMP	−0.068	−1.609	−0.023	−1.182
INTEREST	0.142	0.793	−0.049	−2.870***
TAX	−0.067	−1.575	−0.015	−0.592
R&D×POSTSOX	0.122	1.917*	0.023	0.584
AMORT_DEP×POSTSOX	−0.347	−7.527***	−0.043	−1.919*
STOCK_COMP×POSTSOX	0.395	7.865***	0.077	3.402***
INTEREST×POSTSOX	−0.112	−0.600	0.084	2.437**
TAX×POSTSOX	0.061	1.263	0.026	0.944
POSTSOX	−0.053	−2.085**	−0.013	−0.991
LNSIZE	0.002	0.189	−0.007	−1.802*
CONTROLS	Yes		Yes
Adjusted R2	.182		.022

Note. The stopper/starter sample includes 3,669 observations: 947 pre-SOX and 2,722 post-SOX. p values are based on heteroskedasticity robust standard errors clustered by firm; SOX = Sarbanes–Oxley Act; POSTSOX = 1 if the observation occurs in the post-SOX period (all quarters from the third quarter of 2002 through December 2006); LNSIZE = natural log of total assets from the respective quarter. All other variables defined in Tables 1 and 2.

*

Significant at the .10 level. **Significant at the .05 level. ***Significant at the .01 level.

Taken together, the results presented in Tables 3 to 5 indicate that SOX and Reg. G reduced the likelihood of opportunistic non-GAAP exclusions, but there is also some evidence that managers of firms that started reporting non-GAAP numbers in the post-SOX period are more likely to utilize certain recurring exclusions such as interest, tax, and stock-based compensation exclusions to meet their opportunistic motivations. Hence, our results illustrate the influence of SOX and Reg. G in non-GAAP exclusions as well as the changes in accounting rules. The following section specifically examines the impact of the introduction of SFAS 123R on our results.

Additional Analyses

In our main analyses discussed in the Multivariate Regression Results section, we find evidence that managers are more likely to exclude stock-based compensation (a) when analysts do not and (b) to meet analyst expectations. As the accounting standard governing share-based payments SFAS 123 was superseded by SFAS 123R (which requires the recognition of stock-compensation expense) during our post-SOX period, our results relative to STOCK_COMP may be influenced by the implementation of SFAS No.123R. To address this concern and shed more light on whether our results are attributable primarily to SOX or the implementation of SFAS 123R, we conduct additional analyses examining the impact of SFAS 123R, which became effective for reporting periods beginning after June 15, 2005. In particular, we repeat the multivariate analyses reported in Tables 3, 4, and 5 after excluding post-SFAS-123R observations. Finding similar results about stock compensation after excluding the post-SFAS-123R observations would suggest that managers exclude stock-based compensation opportunistically. However, finding that STOCK_COMP is no longer statistically significant in our regressions after excluding the post-SFAS-123R observations would suggest that our results regarding stock-based compensation are primarily attributable to the implementation of SFAS 123R and not the implementation of SOX.

We find that the results related to STOCK_COMP in Table 3 (using the full sample) are robust to the exclusion of post-SFAS-123R observations for both Models 1 and 2 (even though they are less statistically or economically significant). Hence, we conclude that stock-based compensation exclusions are used opportunistically by managers to appear to meet or beat analyst forecasts in the post-SOX period when analysts choose not to exclude this expense. This result is consistent with Barth et al.’s (2012) evidence, though it is clear that the implementation of SFAS 123R in the latter part of our sample period also influences our results (i.e., managers are also more likely to exclude stock-compensation expense in the post-SFAS-123R period).

When we repeat Model 1 reported in Tables 4 and 5 for continuer and stopper/starter firms respectively, the results are robust to the exclusion of post-SFAS-123R observations, suggesting that managers of continuer firms in the post-SOX period and of starter firms in the post-SOX period are more likely to exclude stock-based compensation when analysts do not. These results are consistent with those reported previously for the full sample. However, the coefficient on STOCK_COMP is no longer statistically significant when we exclude post-SFAS-123R observations and repeat the Model 2 regressions reported in Tables 4 and 5. This result, as well as the sharp rise in stock-compensation exclusions shown in Figure 3, suggests that the evidence for our second aggressiveness measure (STREETBEAT) for these specific subsamples is strongly influenced by the implementation of SFAS 123R. Overall, the additional analyses suggest that the results regarding stock-based compensation are partially attributable to the implementation of SFAS 123R during the latter part of our post-SOX period, requiring firms to recognize compensation-related expenses at fair value. However, we also find evidence consistent with the notion that managers exclude stock compensation for opportunistic reasons.