Relative Emphasis on Non-GAAP Earnings in Conference Calls: Determinants and Market Reaction

ABSTRACT

We use textual analysis to examine whether non-GAAP earnings receive greater emphasis than GAAP earnings in the conference calls that accompany earnings announcements. We measure relative emphasis, i.e. prominence, based on the first appearance or frequency of GAAP and non-GAAP earnings per share (EPS) dollar amounts in the transcripts of conference calls. To complement our analysis of relative emphasis on non-GAAP earnings, we measure general non-GAAP content using frequency counts of keywords. We find that firms place greater relative emphasis on non-GAAP earnings and include more general non-GAAP content when the non-GAAP results exceed the GAAP results, when the non-GAAP results achieve a benchmark that the GAAP results missed, and when the firm’s GAAP earnings are less value-relevant. We find somewhat weak evidence that impression-management motivation is the dominant explanation for greater relative emphasis on non-GAAP earnings but not for general non-GAAP content. Overall, the construct and measurement of relative emphasis on non-GAAP earnings and general non-GAAP content differ, but results indicate they are complements in explaining the market response to earnings conference calls.

Keywords:

• Non-GAAP
• Conference calls
• Disclosure emphasis
• Text mining

Acknowledgements

We thank the guest editors and anonymous referees for helpful comments and suggestions. Our keyword list benefited from a review by Jack Ciesielski and staff at The Analyst’s Accounting Observer.

Notes

1 ‘Regulation G [SEC 2002] applies to all public releases or disclosure of non-GAAP measures, even if such information is not part of a registrant’s SEC filing (e.g. conference calls, investor presentations, and webcasts) and regardless of whether the information is furnished or filed with the SEC’ (Deloitte & Touche, 2017, p. 19). However, while conference calls are subject to certain Regulation G disclosure requirements, they do not trigger a separate 8-K filing and are not subject to the restrictions on relative emphasis of non-GAAP financial measures (i.e. the requirement to present GAAP-compliant measures with equal or greater prominence). See Deloitte & Touche (2017, p. 19).

2 ‘Those interested in participating [in conference calls] are mostly the investment community’ (Westbrook, 2014, p. 140). Research confirms that conference calls are an important source of information, directly informing money managers’ trading decisions (Frankel, Johnson, & Skinner, 1999) and analysts’ earnings forecasts (Brown, Call, Clement, & Sharp, 2015; Mayew, Sharp, & Venkatachalam, 2013).

3 This measure is similar to that used in concurrent research by Black et al. (2019) who examine the relation between managements’ discussions of non-GAAP adjustments and analysts’ adjustments in deriving non-GAAP earnings. Given their focus on adjustments, their keywords also pertain to types of adjustments such as ‘restructuring’ and ‘impairment’ charges.

4 These findings are similar to those in Bowen et al. (2005) and Bradshaw and Sloan (2002) who examine firms’ choices about relative emphasis on non-GAAP earnings in earnings press releases prior to Regulation G’s restrictions in the United States. Beyond examining conference calls, which differ from earnings press releases in important ways previously enumerated, our paper differs from prior research in utilizing a unique computational linguistics approach for measuring emphasis, applied to a far larger sample size over a longer time frame, specifically, 19,521 firm-quarters, in contrast to 1168 firm-quarters in Bowen et al. (2005).

5 Shearman & Sterling (2016, p. 2) note: ‘By far the most common comment, appearing in roughly 40% of the letters, asked for compliance with the requirement to include a presentation of the most directly comparable GAAP financial measure ‘with equal or greater prominence [emphasis added].’ This applies whenever a non-GAAP measure is included in a document filed with the SEC or, for US domestic issuers, in an earnings release furnished to the SEC under Item 2.02 of Form 8-K (but not to earnings call transcripts [emphasis added] or slides not required to be furnished under Item 2.02)’.

6 For example, a manager’s over-emphasis of non-GAAP earnings could be perceived as intentionally misleading, which would likely incur reputational costs.

7 I/B/E/S has also been widely used as a source of non-GAAP earnings to proxy for managers’ non-GAAP disclosures (e.g. Brown & Sivakumar, 2003; Bradshaw & Sloan, 2002; Doyle et al., 2003). Bentley, Christensen, Gee, and Whipple (2018) note this source is not as accurate for managers’ disclosures as the hand collection of non-GAAP disclosures from earnings announcements as used, for example, in Bhattacharya et al. (2003), Johnson and Schwartz (2005), and Lougee and Marquardt (2004).

8 The algorithm to create the numbers-based measures involves converting each conference call into a string, searching the call for the corresponding EPS number (GAAP and non-GAAP) as a substring considering alternative format types, and inserting a mark at each point at which the amount is located. Emphasis.frequency uses a sum of the marks for each EPS number. For Emphasis.NGEarningsfirst and Emphasis.distance, the algorithm extracts the text preceding the first appearance of each EPS number and uses a comparison of the length to generate the variables.

9 Our keyword list benefited from a review by Jack Ciesielski and staff at The Analyst’s Accounting Observer.

10 Where neither the GAAP-compliant EPS nor the non-GAAP EPS amounts are included in the call (4685 observations), the value for the numbers-based measures of relative emphasis is zero, i.e. no evidence of relative emphasis.

11 Each measure of relative emphasis is positively correlated with each of the main test variables (NGEarnings_gt_GAAP, OnlyNonGAAPAchieved, OnlyNonGAAPIncreased, HighTech, HighNonrecurring, and Analysts). The pairwise correlations between the general emphasis measures and all main test variables are also positive, apart from Emphasis.KeyWordGeneric and HighNonrecurring which are positive but not statistically significant.

12 Our results are robust to estimating abnormal return based on CAPM, Fama-French two-factor model, and Fama-French three-factor model.

13 For robustness, we repeat the analysis substituting institutional ownership for analysts and get similar results.

14 We repeat the determinants analysis using the hand-collected data set on managers’ non-GAAP earnings from Bentley et al. (2018), which reduces the sample size to 8,361 because of missing EPS, lagged EPS values, or differences in conference calls included. The regression results yield similar inferences for NGEarnings_gt_GAAP, OnlyNonGAAPAchieved, and HighTech. Results are similar but somewhat weaker for HighNonrecurring. For OnlyNonGAAPIncreased and Analysts, the coefficients are similarly positive but not statistically significant at conventional levels.

15 An alternative interpretation is that a greater number of analysts increases the demand for emphasis on non-GAAP earnings. Further, a small pool of analysts could increase the impact of one credulous analyst on non-GAAP emphasis.

16 We repeat the determinants analysis using the keyword measure focused on non-GAAP adjustments used in concurrent research by Black et al. (2019) as the dependent variable and find some similarities: the coefficients are positive and significant for NGEarnings_gt_GAAP, OnlyNonGAAPIncreased, and HighNonrecurring, and negative for analysts following. We leave for future research a further exploration of differences.

17 This result should be viewed with some caution as results are sensitive to the choice of statistical methodology. Specifically, utilizing a same-model test (i.e. test for the equivalence of the sum of coefficients on impression-management proxies and the sum of coefficients on value-relevance proxies) gives weaker results. Moreover, in general, managers’ motivations are difficult if not impossible to assess with archival analysis.

18 Results are robust to inclusion of a control variable defined as equal to one when the call occurs on a different date than the earnings announcement. Results are also robust to inclusion of a control for beginning of period share price but are somewhat sensitive to variable specification. When UEPS is scaled by beginning of period share price, coefficients on the interactions are positive and significant for only one of the numerical measures and one of the keyword measures.

19 Differences between the two sections may also reflect the influence of investor relations or legal staff involved in preparing the scripted remarks.

20 Updates shown on the SEC website include references to the date on which the staff guidance was issued. See https://www.sec.gov/divisions/corpfin/guidance/nongaapinterp.htm.

21 Dictionary available at https://sraf.nd.edu/textualanalysis/resources/#Master%20Dictionary.

Additional information

Funding

We thank the following organizations for financial support from National Natural Science Foundation of China [71772150], Young Talent Recruiting Plans of Xi’an Jiaotong [GL1J009].