an State
rsity

I’ve

LIBRARY
n

Michi
U

 

 

 

This is to certify that the
dissertation entitled

HAS REGULATION G IMPROVED THE INFORMATION
QUALITY OF NON-GAAP EARNINGS DISCLOSURES?

presented by

HAN SANG Yl

has been accepted towards fulfillment
of the requirements for the

PHD degree in BUSINESS ADMINISTRATION

 

 

%/

Major rofessor’s Signature
(01m

Date

MSU is an Affirmative Action/Equal Opportunity Institution

..-.-.--.--.-.-.-.-.-.-.—-s—~-»—.- ——

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DateDue.indd-p.1

 

HAS REGULATION G IMPROVED THE INFORMATION QUALITY OF
NON-GAAP EARNINGS DISCLOSURES?

By

Han Sang Yi

A DISSERTATION
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Accounting and Information Systems

2006

ABSTRACT

HAS REGULATION G IMPROVED THE INFORMATION QUALITY OF
NON-GAAP EARNINGS DISCLOSURES?

By
Han Sang Yi

I examine whether the documented decreased frequency of non-GAAP earnings
disclosures post-Regulation G (e.g., Heflin and Hsu 2005, Marques 2005, National
Investors Relations Institute 2003) is evidence that the Sarbanes-Oxley Act of 2002 has
achieved one of its key disclosure objectives relating to non-GAAP earnings disclosures.
The intent of Congress and the SEC in issuing Regulation G was to improve the
transparency of non-GAAP earnings, facilitating management’s ability to communicate
the economic prospects of the firm and diminishing the prospect that management could
issue non-GAAP earnings to opportunistically mislead investors about firm performance.
Based on 10,896 sample firrn-quarters from 2001 to mid 2004, I find that managers of
sample firms that have communication motives, proxied by low historical retums-
eamings relations, are more likely to disclose non-GAAP earnings in the post-Reg G
period than in the pre-Reg G period. In contrast, I find that managers of sample firms that
have opportunistic motives, proxied by the firms’ GAAP earnings falling short of various
earnings’ benchmarks, are less likely to do so in the post-Reg G period than in the pre-
Reg G period. I also provide evidence that the incremental information content of sample
firms’ non-GAAP earnings surprises over GAAP earnings surprises is statistically
significant only in the post Reg G period, suggesting that investors’ improved perception
of the transparency of non-GAAP earnings also is consistent with a decrease in

misleading non-GAAP earnings post-Reg G. Overall, the findings of this paper appear

consistent with Congress’ and the SEC ’s intervention in pro-forma reporting practices
resulting in improvements in the quality of information provided in non-GAAP earnings

disclosures.

ACKNOWLEDGEMENTS

I believe that the best part of my Ph.D. program at Michigan State University was
the opportunity to meet many great people. I am deeply indebted to the members of my
dissertation committee for their insight and guidance: Thomas J. Linsmeier (Chair),
Kathy R. Petroni, K. Ramesh, and Jennifer Marietta-Westberg. I especially thank Tom
for his mentoring. Frequently, a single conversation with him proved more valuable than
many days of diligent study. My thanks also go to Bruce Bettinghaus, Marilyn Johnson,
Joan Lufi and Mike Shields for their thoughtful advice throughout my Ph.D. program.

I would like to thank former and current doctoral students in the Department of
Accounting and Information Systems at Michigan State University who made my journey
an intellectually stimulating and enjoyable one. I also thank Deb North and the other staff
for their generous help.

Finally, I sincerely appreciate the unconditional support from my family:
my wife Kyoung-hwa, my daughter Bin, my parents Jeonhee and Insoon, and my brother
Hankook. Their encouragement and understanding makes this dissertation as much theirs

as mine.

iv

TABLE OF CONTENTS

LIST OF FIGURES ................................................................................................... vii
LIST OF TABLES ..................................................................................................... viii
CHAPTER 1: INTRODUCTION .............................................................................. 1
CHAPTER 2: BACKGROUND AND REGULATION G ........................................ 7
2.1. Pro-Forma Earnings Debate ....................................................................... 7
2.2. Regulatory Development and Regulation G .............................................. 8
CHAPTER 3: LITERATURE REVIEW AND ANALYSIS ..................................... 12
3.1. Pre-Regulation G Studies ........................................................................... 12
3.2. Post-Regulation G Studies ......................................................................... 16
3.3. Distinctive Features of the Dissertation .................................................... 17
CHAPTER 4: HYPOTHESES DEVELOPMENT .................................................... 19
4.1. Comparison of Disclosure Motives Across Time ...................................... 19
4.2. Comparison of the Market’s Perception of the Transparency of
Non-GAAP Earnings ................................................................................. 22
CHAPTER 5. RESEARCH DESIGN 24
5.1. Comparison of Disclosure Motives Across Time ...................................... 24
5.2. Comparison of the Market’s Perception of the Transparency of
Non-GAAP Earnings ................................................................................. 28
CHAPTER 6: SAMPLE SELECTION AND DESCRIPTIVE STATISTICS ........... 31
6.1. Sample Selection ........................................................................................ 31
6.2. Descriptive Statistics .................................................................................. 32
CHAPTER 7: EMPIRICAL RESULTS .................................................................... 36
7.1. Characteristics of Disclosed Non-GAAP Earnings ................................... 36
7.2. Changes in Disclosure Motives Across Time ............................................ 38
7.3. Changes in the Market’s Perception of the Transparency of
Non-GAAP Earnings ................................................................................. 41
CHAPTER 8: CONCLUSION ................................................................................... 45
APPENDIX A: SUMMARY OF THE RULES ......................................................... 48
APPENDIX B: COMPARISON OF SAMPLE SELECTION APPROACHES ........ 52
APPENDIX C: MODEL SPECIFICATION ISSUE FOR EQUATION (1) ............ 54

APPENDIX D: PROBLEMS OF THE LONG-RUN RETURNS TEST ................... 56

BIBLIOGRAPHY ...................................................................................................... 80

vi

LIST OF FIGURES

Figure 1. Non-GAAP Earnings and Regulation G Timeline ..................................... 57

Figure 2. Frequency of Non-GAAP Earnings Disclosures by Calendar Quarters ..... 58

vii

LIST OF TABLES

Table 1. Sample Selection ..........................................................................................

Table 2. The Frequency of Disclosure of Non-GAAP Earnings Across

Regulatory Time Periods .............................................................................

Table 3. The Frequency of Disclosure of Non-GAAP Earnings Across Industries...

Table 4. Comparison of Non-GAAP EPS, I/B/E/S Actual EPS, GAAP EPS, and

S&P Core EPS Across Two Regulatory Periods .........................................

Table 5. The Frequency of the Nomenclatures Firms Used for Non-GAAP

Earnings Presentation Across Regulatory Time Periods .............................

Table 6. The Frequency of the Adjustment Items Used for Non-GAAP Earnings

Presentation ..................................................................................................
Table 7. Emphasis of Non—GAAP Earnings and Reconciliation Information ...........

Table 8. Non-GAAP Adjustments and Earnings Benchmarks ..................................

Table 9. Pearson and Spearman Correlations of Characteristics of Disclosed

Non-GAAP Earnings ..................................................................................

Table 10. Descriptive Statistics of Firm Characteristics by Non-GAAP Earnings

and Regulation G Regimes ..........................................................................

Table 11. Correlation Table of Firm Characteristics by Non-GAAP Earnings

and Regulation G Regimes ..........................................................................

Table 12. Probit Regression for H1 & H2 .................................................................

Table 13. Random Effect Probit Regression for H1 & H2: Alternative

Specification with the Post Time Dummy .................................................

Table 14. OLS Regression for H3 ..............................................................................

Table 15. OLS Regression for H3 with the Sub—Sample of Continuous

Disclosing Firms ........................................................................................

Table 16. OLS Regression with the Sub-Sample of Firms that Discontinued

Disclosing Non-GAAP Earnings in the Post-Reg G Period ......................

viii

59

60

61

62

63

64

65

66

67

68

72

75

76

77

78

79

CHAPTER 1: INTRODUCTION

In response to the highly publicized alleged misuse of pro-forma earnings
disclosures, the US. Congress ordered the Securities and Exchange Commission (SEC)
to issue new rules governing the presentation of non-GAAP financial metrics to improve
the quality and transparency of financial accounting information (Section 401(b) of the
Sarbanes-Oxley Act of 2002). The SEC released Regulation G (SEC 2003a, hereafier
“Reg G”) in January 2003, requiring firms that disclose non-GAAP earnings in
preliminary earnings announcements to clearly reconcile non-GAAP earnings to GAAP
earnings with equal emphasis on both numbers.l Recent studies by Heflin and Hsu
(2005) and Marques (2005) and a survey by the National Investor Relations Institute
(2003) document a significant decline in non-GAAP earnings disclosures after the SEC
intervention. This paper examines whether the decreased frequency of non-GAAP
earnings disclosures post-Reg G reflects intended or unintended consequences of Reg G
(and the Sarbanes-Oxley Act of 2002).

Business journalists, citing conflicting anecdotes, continue to make divergent
claims as to whether or not Reg G has effectively eliminated disclosures of misleading
non-GAAP earnings in preliminary earnings announcements. Some View Reg G to be a
lax enforcement mechanism, arguing that firms can still characterize routine expenses as
special charges to justify reversing these items in arriving at non-GAAP earnings;
misleading investors’ perceptions about firm performance (e. g., Bymes and Henry 2001,

Henry 2002, Stuart 2004, Taub 2003). Alternatively, others believe that Reg G helps

 

' Following Regulation G, I often use the term “non-GAAP earnings” to capture the concept of
management-defined earnings, which prior literature often refers to as pro-forma earnings. Throughout this
paper, I use pro-forma earnings and non-GAAP earnings interchangeably.

investors unravel the firm’s strategic motives underlying the disclosure of non-GAAP
earnings, discouraging firms with opportunistic disclosure motives from misleading
investors and increasing the frequency with which firms disclose non-GAAP earnings
post-Reg G to help inform investors about the economic prospects of the firm (e. g.,
Henry 2003; Thompson 2003).

Extant academic evidence examining the disclosure of non-GAAP earnings prior
to Reg G also provides mixed evidence about the transparency of non-GAAP earnings
disclosures. One line of literature shows that non-GAAP earnings function as a substitute
for less relevant GAAP earnings, and thus have incremental information content over
GAAP earnings and better map into future performance (e.g., Bhattacharya et al. 2003;
Bradshaw and Sloan 2002; Brown and Sivakumar 2003; Doyle et al. 2003; Vincent
1999). Another line of literature documents that managers may disclose non-GAAP
earnings opportunistically, altering investors’ perceptions about the finn’s ability to meet
or beat various earnings benchmarks (e. g., Bowen et a1. 2005; Doyle et al. 2004; Elliott
2006; Frederickson and Miller 2004; Hirshleifer and Tech 2003; Lougee and Marquardt
2004; Schrand and Walther 2000).

Thus, the decreased disclosure frequency of non-GAAP earnings after Reg G
reported in Heflin and Hsu (2005), Marques (2005) and NIRI (2003) could be attributed
either to the success or failure of the efforts of Congress and the SEC. If Reg G resulted
in fewer firms disclosing non-GAAP earnings to mislead investors, then Reg G has
achieved its objective. Alternatively, if Reg G resulted in fewer firms reporting non-
GAAP earnings to better inform investors, then Reg G has had unintended consequences.

The current paper attempts to distinguish between these two alternative consequences by

investigating the post-Reg G changes in: (1) managers’ underlying strategic motives for
disclosing non-GAAP earnings (i.e., to better communicate with investors when GAAP
earnings are less relevant or to opportunistically mislead investors ), and (2) investors’
perceptions of the relevance of non-GAAP earnings (i.e., by measuring the changes in the
incremental information content of non-GAAP earnings over GAAP earnings).

I examine 10,896 press releases of 792 randomly selected firms from the
intersection of CRSP and COMPUSTAT spanning from 2001 to mid 2004, and identify
3,228 non-GAAP earnings disclosures. Descriptive statistics suggest that sample firms
are less likely to emphasize non-GAAP earnings over GAAP earnings in the post-Reg G
period. Similarly, sample firms are more likely to provide detailed reconciliation
information of non-GAAP earnings to GAAP earnings in the post-Reg G period. These
results suggest that reporting firms have complied with Reg G’s requirements.

Based on prior research, I assume managers have two conflicting motives for
disclosing non-GAAP earnings (communication and opportunism). I examine changes in
the association between disclosure of non-GAAP earnings and these disclosure motives
over the pre- and post-Reg G periods. I assume that firms have communication motives
to disclose non—GAAP earnings when GAAP earnings have low relevance, as proxied by
a low GAAP eamings-security return relationship (Bradshaw and Sloan 2002, Lougee
and Marquardt 2004). I assume that firms have opportunistic motives for disclosing non-
GAAP earnings when GAAP earnings fall short of various earnings benchmarks (i.e.,
GAAP losses, negative seasonally differenced GAAP earnings, or GAAP earnings falling

short of consensus earnings forecasts).

I find that managers of sample firms are more likely to disclose non-GAAP
earnings in the post-Reg G period than in the pre-Reg G period when they have a lower
historical GAAP earnings-security return relationship (i.e., high communication motives).
In contrast, I find that managers of sample firms are less likely to disclose non-GAAP
earnings in the post-Reg G period than in the pre-Reg G period when they have losses or
negative GAAP earnings changes (opportunistic motives). These results imply that the
decreased disclosure of non-GAAP earnings post Reg G is more pronounced for firms
with opportunistic motives than for firms with communication motives.

To examine whether changes in investors’ perception of the transparency of non-
GAAP earnings also are consistent with a decrease in misleading non-GAAP earnings
disclosures post Reg G, I regress three-day size adjusted cumulative stock returns on both
seasonally differenced GAAP earnings and non-GAAP earnings surprises in the pre- and
post-Reg G periods. In contrast to the results from prior studies that report incremental
information content of non-GAAP earnings over GAAP earnings in the years preceding
my sample period, I do not find the incremental information content of non-GAAP
earnings over GAAP earnings in the pre-Reg G period (2001:01-2003:03). However, I
find that non-GAAP earnings are incrementally informative over GAAP earnings in the
post-Reg G period (2003:04-2004:08). This implies that investors, on average, perceive
non-GAAP earnings as more transparent in the post-Reg G period than in the pre-Reg G
period. Overall, the findings of this paper appear consistent with Congress’ and the
SEC ’5 interventions in pro-forma reporting practices resulting in improvements in the

quality of non-GAAP earnings disclosures.

This dissertation makes the following contributions to the literature. First, this
study provides timely and interesting evidence that addresses the Dechow and Schrand
(2004; 116) and Stuart (2004) calls for research on the effects of Reg G (and the
Sarbanes-Oxley Act of 2002) on firms’ (mis)use of pro-forma earnings disclosures. The
current study examines both temporal changes in managers’ disclosure motives and
temporal changes in investors’ perceptions of the transparency of non-GAAP earnings to
draw inferences about the consequences of Reg G on non-GAAP earnings disclosures.
Thus, this dissertation differs from Heflin and Hsu (2005), Marques (2005) and Zheng
and Zhang (2005) in that it focuses on (1) why, as opposed to whether or not, firms
discontinue presenting non-GAAP earnings disclosure following Reg G (i.e., by
examining temporal changes in non-GAAP earnings disclosure motives) and (2) whether
the capital market’s reaction to non-GAAP earnings adjustment is consistent with the
observed change in disclosure motives post Reg G.

Second, this dissertation adds to the debate on the transparency of pro-forma
earnings (e. g., Bradshaw 2003) by providing evidence that pro-forma earnings were used
as a means to obfuscate investors’ perception about firm performance in the pre-Reg G
period. The dual findings of this study that managers report non-GAAP earnings in a
more transparent way in the post-Reg G period and that the market values this
improvements of non-GAAP earnings transparency in the post-Reg G period suggest that
many firms disclosed non-GAAP earnings opportunistically in the less regulated
environment. In addition, investors perceived such disclosures as potentially
opportunistic, biasing downward the pricing of non-GAAP earnings in the pre-Reg G

period.

The remainder of the dissertation continues as follows. Chapter 2 provides the
background. Chapter 3 reviews relevant literature relating to non-GAAP earnings and
Chapter 4 develops the research hypotheses. Chapter 5 describes the research design.
Chapter 6 outlines my sample selection approach and provides descriptive statistics.
Chapter 7 presents the results of my analyses. Finally, Chapter 8 summarizes and

concludes.

CHAPTER 2: BACKGROUND AND REGULATION G

2.1. Pro-forma Earnings Debate

Although the SEC adopted the term “non-GAAP financial measures” in Reg G,
what triggered the legislative action was the proliferation in the late 1990’s of so-called
“pro-forma earnings” disclosures by US. corporations.2 Throughout time, pro-forma
earnings was the term used to describe hypothetical financial results reported by firms for
comparison purposes under financial reporting requirements (e.g., M&A, accounting
changes etc.). However, since the mid 90’s, firms in the technology sector started using
the term to describe earnings that eliminate the effect of non-cash, non-recurring charges
from core operating activities (Thurrn and Weil 2001). The practice has become
widespread in other industries and the list of excluded items has broadened over time.
Today, pro-forma earnings has become the general term describing a firm’s voluntary
disclosure of modified earnings measures that deviate from GAAP earnings (Bradshaw
2003), an important communication vehicle for management (e. g., Bowen et al. 2005,
Graham et al. 2004).

Advocates of pro-forma earnings reporting argue that pro—forma earnings reveal
management’s view about core operating results, and thus help investors by reducing
information asymmetry between management and investors. Consistent with this notion,

anecdotal evidence suggests that not only do financial analysts often find pro-forma

 

2 Some countries explicitly allow firms to present alternative EPS metrics in income statements. Among
them is the UK. where firms are allowed to present adjusted EPS on the face of their income statements
under Financial Reporting Standard No.3. Choi, Lin, and Young (2005) report that UK. firms’ adjusted
EPS is more informative about current and future performance than GAAP EPS.

earnings usefirl for equity valuation, but many oppose the exclusion of pro-forma
earnings from press releases (Taub 2001).3

However, as a result of a spate of financial frauds at the end of the tech bubble,
the financial press started touting the potential problematic use of pro-forma earnings,
igniting a debate about the transparency of non-GAAP earnings (e.g., Bymes and Henry
2001; Henry 2002; Teach and Reason 2002; Weil 2001). Critics often indicate three risks
relating to pro-fonna earnings reporting that may misinforrn investors: (I) insufficient
information being provided in press releases to reconcile GAAP and non-GAAP earnings
(e. g., Wallace 2002), (2) differential presentation in press releases focusing greater
attention on non-GAAP earnings (e.g., Bowen et al. 2005), and (3) strategic and
inconsistent selection of items reversed in determining non-GAAP earnings creating the
ability to upward bias non-GAAP earnings by desired amounts (e.g., Weil 2001).
Anecdotal evidence further suggests that some managers explicitly recognized that they

presented pro-forma earnings to simply “put the best spin on the results” (Goff 2001 ).4

2.2. Regulatory Development and Regulation G (See Figure 1 for the timeline)

Prior to Reg G, both private-sector organizations and the SEC alerted firms to the
three risks mentioned above associated with disclosing misleading pro-forma earnings
(hereafter referred to as “non-GAAP earnings”). Relating to the first risk identified above

(i.e., unclear reconciliation of GAAP and non-GAAP earnings), two influential private-

 

3 A survey conducted by a New York based research firm, Broadgate Consultants Inc., found that 67% of
223 portfolio manager respondents (buy side analysts) opposed exclusion of pro-forma earnings from press
releases although they wanted some consistency in pro-forma reporting (Taub 2001).

4 The CFO magazine and KPMG joint survey of 196 financial executives at an FBI conference reveals the
conflicting motives concerning the presentation of pro-forma earnings, viz. ( 1) to convey true performance
(45%), (2) to meet the demands of analysts (27%), and (3) to help ‘put the best spin on the results’ (25%).

sector organizations (Financial Executives International and the National Investor
Relations Institute, hereafter ‘FEI’ and ‘NIRI’ respectively) jointly proposed industry
guidelines for reporting pro-forma earnings. These guidelines advised firms to make clear
reconciliations between GAAP earnings and non-GAAP earnings and to explicitly
present GAAP earnings in earnings announcements (NIRI 2001). The SEC supported the
guidelines suggested by FBI and NIRI, and it reminded firms of the importance of
antifraud provisions and materiality in the pro-forma earnings context (the SEC’s
cautionary advice, see SEC 2001).5 Relating to the second risk mentioned above (i.e.
emphasis on positively spun pro-forma earnings over GAAP earnings), NIRI issued
further guidelines encouraging firms to present GAAP earnings with equal emphasis in
earnings releases (N IRI 2002). However, with respect to the third risk (i.e. inconsistency
in adjustment items), there was no attempt to issue any guidelines prior to Reg G,
because the choice to either include or exclude items, in itself, reflects the management’s
view about the item’s relevance to core operating activities.

The Sarbanes-Oxley Act of 2002 was signed into law by President Bush on July
30, 2002. The SEC’s implementation of Section 401(b) of the Act (“Conditions for Using
non-GAAP Financial Measures,” hereafter “the Rules”, January 22, 2003, see SEC
2003a), followed a two-step approach to deal with the three risks identified by critics of

pro-forma earnings (see SEC 2003b).

 

5 The SEC ’5 first administrative action, against Trump Hotel and Casino, was related to the first issue (SEC
2002) and was the first SEC administrative action against the opportunistic presentation of non-GAAP
earnings (January 16, 2002). In its earnings announcement, the company reported a profit of $14 million in
the third quarter of 1999, excluding a one-time charge of $81 million. Although the firm touted its EPS
exceeded the analysts’ forecast, what actually exceeded First Call estimates of 0.54 cents per share was
pro-forma EPS of 0.63 cents per share. It later turned out that the company included a one-time accounting
gain of around $17.2 million from the earlier termination of a lease agreement with the All Star Cafe’. What
is most striking in this case was that in their original press releases the firm provided a single revenue line
to hide their one-time gain.

The first step is to regulate the presentation of non-GAAP earnings and its
reconciliation with GAAP earnings in public communications other than SEC filings
(e. g., earnings releases and conference calls). In the first step, the Rules address the first
and the second problems identified above (i.e. no clear definition/reconciliation and the
emphasis placed on positively spun pro-forma earnings). Specifically, Reg G requires
firms to present GAAP earnings and provide reconciliations of non-GAAP earnings to
GAAP earnings. In addition, within 4 business days of releasing an earnings release
containing non-GAAP earnings, firms should disclose in item 12 of Form 8-K both
GAAP and non-GAAP amounts and the reason why management believes the non-
GAAP measure is useful to investors.6

The second step is geared towards dealing with the third problem
(i.e. inconsistency in adjustment items). If firms additionally disclose non-GAAP
earnings in SEC filings, the Rules give a set of specific guidelines about what should not
be excluded or included in the calculation of non-GAAP earnings (see Appendix A for
the details of the Rules). Because this requirement is highly likely to discourage firms
from presenting non-GAAP earnings, the SEC restricts this requirement only to non-
GAAP presentations in SEC filings.

Thus, when it comes to the reporting of non-GAAP earnings in preliminary
earnings announcements (one form of non-SEC filing), one could view Reg G simply as
a lax enforcement system, since firms still have the flexibility to select adjustment items

in a discretionary manner as long as they reconcile non-GAAP earnings with GAAP

 

6 I do not find any meaningful variation in the sample firms’ statements regarding why firms believe non-
GAAP earnings are useful to investors. Two boilerplate reasons I find are: (1) such earnings better measure
the firms’ operating performance, and (2) firms provide such earnings for consistency (i.e., the importance
of a disclosure precedent).

10

earnings and provide equal emphasis on both numbers. This is especially so because the
expected preparation costs concerning Reg G are minimal according to the SEC (SEC
2003a). Thus, one can hold the view that absent other factors, complying with the
mandated disclosure format requirement for the voluntary disclosure would have little
impact on the disclosure decision.7 However, recent studies by Heflin and Hsu (2005)
and Marques (2005) report that non-GAAP earnings disclosures outside SEC filings have
significantly decreased after Reg G for I/B/E/S and S&P 500 firms respectively. A recent
survey by NIRI also reports that 47% of survey respondents who previously provided
non-GAAP earnings in their earnings announcements and now provide GAAP earnings
only (26.4% of the total respondents) attribute their change in disclosure strategy to Reg
G (NIRI 2003).

There could be two extreme interpretations of the decreased frequency of non-
GAAP earnings disclosure after Reg G (Stuart 2004). One view suggests that Reg G has
limited the frequency of non-GAAP earnings disclosures that were previously used to
mislead investors. However because prior research also supports the value-relevance of
pro-forma earnings over GAAP earnings (see next section), it is also possible that Reg G
discouraged firms from providing value-relevant information. In this dissertation, I
provide evidence to distinguish between these divergent claims to infer whether or not
Reg G had intended (reducing misleading disclosures) and/or unintended (reducing

transparent disclosures) consequences.

 

7 This is the official position of the SEC. Three out of 81 comment letters explicitly echoed this notion.
Other comment letters were silent about whether Reg G would limit the use of non-GAAP earnings
disclosures.

ll

CHAPTER 3: LITERATURE REVIEW AND ANALYSIS

This section provides a summary of two strands of non-GAAP earnings research
germane to the current paper, discusses why the current paper differs from concurrent

studies on Regulation G, and identifies how the current paper extends prior research.

3.1. Pre-Regulation G Studies
3.1.1. Determinants of Non-GAAP (pro-forma) Earnings Disclosures

Prior literature assumes the following two conflicting strategic motives for
disclosing non-GAAP (pro-forma) earnings.
Substitute for Less Relevant GAAP Earnings

Despite the issues relating to using R2 and ERC’s from returns-earnings
regressions as a basis for inference about the value-relevance of reported earnings (e.g.,
Brown, Lo and Lys 1999), a number of studies provide evidence of a decline in GAAP
earnings’ relevance over time based on a temporal decline in the association between
GAAP EPS and stock prices (e. g., Collins, Maydew and Weiss 1997; Francis and
Schipper 1999; Lev and Zarowin 1999). In contrast, Bradshaw and Sloan (2002) report
an increasing association between street earnings and stock prices over time. They
provide preliminary evidence that earnings adjusted by managers in preliminary earnings
announcements are likely to contribute to the formation of street earnings, conjecturing
that managers may disclose non-GAAP earnings when their GAAP earnings are less
relevant. Consistent with the conjecture, Lougee and Marquardt (2004) find that pro-

forma earnings disclosure firms are more likely than their matched sample firms to have

12

less relevant GAAP earnings (i.e., lower Rz’s and ERC’s from returns-earnings
regressions).
Strategic Means to Alter Investors ’ Perception about Firm Performance

A number of papers show that managers may selectively report items in
preliminary earnings announcements as a strategic tool either to alter investors’
perception about the firm’s performance or to change perceptions about its ability to meet
or beat various earnings benchmarks. For example, Schrand and Walther (2000) find that
managers are more likely to remind investors of extenuating circumstances (i.e., one-time
gains) for strong performance in previous years than to remind investors of circumstances
for poor performance in previous years, potentially making current performance look
better (systematic spinning/tilting of information). Also, Bowen et al. (2005) show that
firms strategically emphasize in their preliminary earnings announcement a specific
earnings metric among multiple earnings metrics depending upon whether such an
earnings metric meets or beats past EPS and/or analysts’ EPS consensus. Relatedly,
McVay (2005) shows that firms are likely to use income classification (e.g., reporting
operating expenses as special charges) to make their core earnings look better.

Consistent with this notion, several studies show that firms are more likely to
provide non-GAAP earnings disclosure when they have losses (Bhattacharya et a1. 2003),
negative seasonally differenced GAAP earnings (Lougee and Marquardt 2004), and
GAAP earnings falling short of analysts’ BPS consensus (Doyle et al. 2004) in order to
change perceptions about whether they meet or beat typical past earnings benchmarks.

Also, Frankel et al. (2005) show that ill-govemed firms (proxied by fewer independent

13

board members) are more likely to disclose non-GAAP earnings that just meet or slightly

beat analysts’ consensus EPS.

3.1.2. Pricing of non-GAAP (Pro-forma) Earnings Disclosures

One line of research literature examines whether alternative earnings metrics have
higher associations with contemporaneous returns and future operating performance than
GAAP earnings. Vincent (1999) provides some evidence that funds from operations
disclosed by real estate investment trust firms have higher association with
contemporaneous returns than GAAP earnings measures. Similarly, a number of studies
compare “street earnings” with GAAP earnings, documenting that street earnings are less
conservative than GAAP earnings and have higher explanatory power for
contemporaneous returns, price, and future earnings and cash flows than do GAAP
earnings (Bradshaw and Sloan 2002; Brown and Sivakumar 2003; Frankel and
Roychowdhury 2004; Doyle et al. 2003). Using “pro-forma” earnings and “street
earnings” interchangeably, these studies conclude that management defined earnings (i.e.,
pro forma earnings) convey incremental information. However, these studies use earnings
extracted from analysts’ forecast data providers (i.e., First Call, I/B/E/S, and Zachs) as
proxies for pro-forma earnings. Thus, it is not clear whether it is actually managers or
analysts that provide the incremental information in earnings reported in these databases
(see Abarbanell and Lehavy 2002, 2005; Gu and Chen 2004). In contrast, recent research

explicitly collects earnings adjusted by managers from press releases and performs

l4

similar tests, documenting relative and incremental information content of pro-forma
earnings over GAAP earnings (Bhattacharya et al. 2003; Lougee and Marquardt 2004).8

Two studies examine whether pro-forma earnings are misleading investors. Using
I/B/E/S actual earnings as a proxy for pro-forma earnings, Dolye et al. (2003) show that
the difference between pro-forma earnings and GAAP earnings is a strong negative
predictor of future performance, suggesting that pro-forma earnings are misleading
investors. Using a P/E ratio approach with pro-forma earnings collected from actual press
releases, Johnson and Schwartz (2005) do not find evidence that the magnitude of the
pro-forma earnings adjustment explains market multiples.

Taken together, evidence on the transparency of non-GAAP earnings in the pre-
Reg G period is mixed. Studies investigating causes (i.e., managers’ motives) of non-
GAAP earnings disclosures show that non-GAAP earnings disclosures are associated
with both communication motives and opportunistic motives. Studies investigating
consequences (i.e., pricing) of non-GAAP earnings disclosures show that while non-
GAAP earnings are, on average, informative in short windows, the non-GAAP adjusted
amount may lead to negative abnormal returns in a long window. This mixed evidence
warrants further investigation into the transparency of non-GAAP earnings. Reg G
provides a setting where researchers can address this issue by examining how the
disclosure motives has changed across two regimes. Two caveats must be noted here on
pre-Reg G non-GAAP earnings studies. First, inferences from studies that use actual

earnings from analysts’ forecast data providers as a proxy for non-GAAP earnings may

 

8 Following Biddle et al. (1995), these authors describe the “relative” information content of A being
greater than that of B when the R2 from a regression of 3-day cumulative returns on A is greater than that
on B. Also, these authors describe A as incrementally informative over B when the coefficient on A is
positive and significant from a regression of 3-day cumulative returns on A and B.

15

not be appropriate for drawing inferences about the transparency of non-GAAP earnings
because the earnings construct provided by forecast data providers is not identical to that
of non-GAAP earnings.9 Second, capital market studies address the joint hypotheses of
information content of non-GAAP earnings and market efficiency. The evidence of long-
terrn mispricing of pro-forma earnings and a higher association of non-GAAP earnings
and various market metrics reported in capital market studies could be a mere
manifestation of investors’ functional fixation on reported metrics rather than indicating
these metrics are more value-relevant (Bowen et al. 2005; Dyck and Zingales 2003).
Thus, studies examining the effect of Reg G must address both determinants of and
pricing of non-GAAP earnings across two regulatory regimes to draw a conclusion about

the effect of Reg G on the transparency of non-GAAP earnings.

3.2. Post-Reg G Period Studies on non-GAAP Earnings

I am aware of three papers that attempt to draw inferences about the consequences
of Reg G (Heflin and Hsu 2005; Marques 2005; Zheng and Zhang 2005). These studies
provide mixed evidence on the consequences of Reg G.

Heflin and Hsu (2005) focus on the changes in disclosure frequency of various
types of non-GAAP earnings. They report that the frequency of non-GAAP earnings
disclosures has declined significantly in the post-Reg G period. They also report that
firms with both “other exclusions (i.e., their proxy for opportunistic disclosure motives
relating to non-GAAP earnings)” and “special items (i.e., their proxy for benign
communication motives)” are less likely to disclose non-GAAP earnings in the post-Reg

G period, perhaps indicating that Reg G dampens both the number of opportunistic and

 

9 See Appendix B for details.

16

non-opportunistic non-GAAP earnings. However, Heflin and Hsu (2005) do not provide
direct tests of this hypothesis because they fail to examine attributes of actual non-GAAP
earnings disclosures instead using I/B/E/S actual earnings as a proxy.

Marques (2005) focuses only on the capital market’s temporal assessment of S&P
500 firms’ non-GAAP adjustments from 2001 to 2003 based on information from actual
press releases. She documents a decline in non-GAAP earnings disclosures in the post-
Reg G period and shows that unique non-GAAP adjustments made by management
beyond the adjustments made by analysts are negatively associated with prices and
returns in the post-Reg G period. The results perhaps suggest a failure in Sarbanes-Oxley
Act of 2002 achieving its objectives relating to improving the transparency of non-GAAP
earnings disclosures. Marques (2005) does not examine strategic motives for firms
providing non-GAAP earnings disclosures.

Zheng and Zhang (2005) also focus on the mispricing of pro-forma earnings only.
Based on pro-forma earnings collected from actual press releases, they document that the
negative association between pro-forma earnings and one-year size-adjusted returns in
the pre-Reg G period is less pronounced when more reconciliation information is
provided. They further report that the negative association has disappeared in their 2002

sample, suggesting that Reg G curtails mispricing of non-GAAP earnings.

3.3. Distinctive features of this study
I examine whether the decreased frequency of non-GAAP earnings disclosures
post-Regulation G is consistent with the Sarbanes-Oxley Act of 2002 (1) increasing the

propensity for firms to issue non-GAAP earnings to communicate economic prospects of

17

the firm and (2) decreasing the propensity for firms to issue non-GAAP earnings to
opportunistically mislead investors’ perceptions about firm performance. No previous
research has explicitly tested these hypotheses. Prior research either has made inferences
on these issues without examining actual non-GAAP earnings disclosures (Heflin and
Hsu 2005) or has performed indirect security prices tests to make inferences about related
issues (Marques 2005 and Zheng and Zhang 2005).

Specifically, this dissertation provides an incremental contribution to the literature
by focusing on the quality of, as opposed to the quantity of, non-GAAP earnings
disclosure following Reg G by examining the post-Reg G changes in: (1) managers’
underlying strategic motives for disclosing non-GAAP earnings and (2) investors’
perceptions of the relevance of non-GAAP earnings. In addition, I follow a group of
predetermined firms from the intersection of CRSP and COMPUSTAT to avoid the
sample selection bias and measurement error issues in Heflin and Hsu (2005), Marques

(2005) and Zheng and Zhang (2005) (see Appendix B for details).

18

CHAPTER 4: HYPOTHESES DEVELOPMENT

4.1. Comparison of Disclosure Motives across Time

Among various costs and benefits associated with the disclosure of non-GAAP
earnings (e.g., proprietary costs, litigation costs, cost of capital consideration etc.), I posit
that management’s incentives relating to the cost of capital and firm valuation are likely
to play the first order role in the non-GAAP disclosure decision.

With respect to the proprietary costs, both Schrand and Walther (2000) and
Hirshleifer and Tech (2003) posit that the primary target audience of earnings
announcements is likely to be equity investors, suggesting that the costs of additional
disclosure that may trigger the entrance of a competitor may be, at most, the second order
effect. Litigation costs certainly could affect the non-GAAP earnings disclosure decision
in the post-Reg G period because it provides an additional context for the SEC to enforce
disclosure rules. However, it is not certain whether the additional disclosures required by
Reg G uniformly will induce declines in the disclosure of non-GAAP earnings for two
reasons. First, what often triggers class action litigation filings is an ex post event such as
sudden price drop, which can be caused by both appropriate and inappropriate
representations of current firm performance (i.e., both disclosure and/or no disclosure of
management defined earnings can trigger litigations to the extent that there is an ex post
price drop). Second, plaintiffs’ decision to file class action litigations may be more
related to the expected settlement amount (e. g., Directors’ and Officers’ insurance
amount, see Kim (2004)) than to the accounting issues per se. Thus, it is not only hard to

make a directional prediction about the relation between the firm specific litigation risks

19

and the non-GAAP earnings disclosure decision before and after Reg G, but also difficult
to suggest that litigation risks play a first order role in disclosure decision making.

The two strategic disclosure motives I examine are linked closely to (1) the
avoidance of mispricing by voluntarily revealing management’s beliefs about firm
performance, and (2) the avoidance of the market’s negative reaction to missing various
earnings benchmarks. Although the SEC contended it did not expect the rules to hinder
the flow of information to investors,‘0 the decreased frequency of non-GAAP earnings
disclosures post-Reg G raises several possible consequences depending on the motives
for providing non-GAAP earnings disclosures.

One possibility is that Reg G has effectively mitigated the incentive to disclose
non-GAAP earnings to mislead investors, while having no impact on firms disclosing
non-GAAP earnings for improved communication motives. An implicit assumption
behind this scenario is that Reg G’s disclosure requirements for non-GAAP earnings
would serve as truth-telling mechanisms under which investors can clearly see through
the strategic motives of non-GAAP earnings disclosures. Prior literature suggests that an
improved disclosure format can help investors better assess the value relevance of an
information cue by (1) reducing cognitive processing costs and (2) helping investors
avoid functionally fixating on the saliently presented information cue (e. g., Dietrich et al.
2001, Maines and McDaniel 2000, Hirshleifer and Tech 2003). Elliott (2006) confirms
this notion experimentally in a pro-forma earnings setting. If managers rationally factor
in the effect of enhanced disclosure format on the investors’ ability to evaluate firm

performance, managers with better communication motives would continue to present

 

'0 The SEC spokesman was reported to say “We don’t expect less disclosure, we expect more meaningful
disclosure. There’s nothing in the rule that precludes people from providing information that they want to
provide to investors, just as long as it's reconciled.” (Babington 2003).

20

non-GAAP earnings disclosure while managers with opportunistic motives would not,
because any attempt to mislead investors with non-GAAP earnings will be screened and
penalized by investors under the Reg G regime.

Another possibility is that Reg G has discouraged firms from providing value-
relevant non-GAAP earnings disclosures by making managers overly sensitized to
investors’ skepticism about non-GAAP earnings in the post-Reg G period. For example,
Robert Willens, an accounting analyst at Lehman Brothers, suggests that “a lot of people
are gun shy about unwittingly providing information that could be seen as misleading,”
perhaps causing firms with communication motives to reconsider providing non-GAAP
earnings information (see Babington 2003). In addition, business journalists contend that
given the nature of Reg G (i.e., a reporting requirement), firms with opportunistic
motives may still have incentive and ability to reverse routine expenses as special charges
in deriving non-GAAP earnings insofar as they are technically permitted under the Rules
(e.g., Countryman 2003, Stuart 2004, Taub 2003).H Thus, it also is possible that the
decreased frequency of non-GAAP earnings disclosures post Reg G could occur more in
firms with communication motives than in firms with opportunistic motives.

Finally, an alternative scenario is that Reg G dampens the propensity to disclose
non-GAAP earnings regardless of disclosure motives. In order to distinguish among these

possibilities, I test a set of hypotheses (stated in null form) relating to the pre- and post-

 

” Reg G explicitly states that “neither the requirements of Reg G nor a person’s compliance or non-
compliance with the requirements of Reg G shall in itself affect any person’s liability under Exchange Act
Section 10(b) or Rule 10b-5.” Chuck Hill, former director of research at Thomson Financial First Call, was
reported to say “while few are blatantly breaking the law by omitting GAAP equivalents or reconciliation
tables, some companies are taking advantage of the relatively lax enforcement of regulations on press
releases to spin their numbers in ways that would be illegal in official filings” (see Stuart 2004).

21

Reg G changes in the relative importance of managers’ underlying strategic motives for

disclosing non-GAAP earnings (i.e., communication vs. opportunism).

HI: (Communication motives and Reg G) The incremental propensity of firms to
disclose non-GAAP earnings to communicate the economic prospects of
firms rather than to opportunistically mislead investors has not changed
over the pre- and the post-Reg G periods.

H2: (Opportunistic motives and Reg G) The incremental propensity of firms to
issue non-GAAP earnings to opportunistically mislead investors '
perceptions about firm performance rather than to communicate economic
prospects has not changed over the pre- and the post-Reg G periods.

If Reg G had the consequences intended by Congress and the SEC, either the
association between non—GAAP earnings disclosures and communication motives will have
increased in the post-Reg G period and/or the association between non-GAAP earnings
disclosures and opportunistic disclosure motives will have declined in the post-Reg G period.
In contrast, Reg G will have had consequences unintended by Congress and the SEC, if
either the association between non-GAAP earnings disclosures and communication motives

declined in the post-Reg G period and/or the association between non-GAAP earnings

disclosures and opportunistic motives increased in the post-Reg G period.

4.2. Comparison of Market’s Perception of the Transparency of non-GAAP Earnings
across Time

The final hypothesis relates to the incremental information content of non-GAAP
earnings surprises in addition to GAAP eamings surprises across the pre- and post-Reg G

periods. My motivation for this hypothesis is to examine whether inferences from H1 and

22

H2 are consistent with the market’s perception of the transparency of non-GAAP
earnings in the post-Reg G period. Assuming that with the help of increased disclosure
requirements investors can determine the transparency of non-GAAP earnings more
efficiently in the post-Reg G period than in the pre-Reg G period, I posit two plausible
scenarios. On one hand, if Reg G is effective such that firms with communication
(opportunistic) motives are more (less) likely to disclose non-GAAP earnings in the post-
Reg G period than in the pre-Reg G period, it is likely that the incremental information
content of non-GAAP earnings is greater in the post-Reg G period than in the pre-Reg G
period. On the other hand, if Reg G is not effective, it is likely that the incremental
information content of non-GAAP earnings is either unchanged or smaller in the post-
Reg G period than in the pre-Reg G period. This discussion leads to the following

hypothesis in null form.

H3: (Incremental Information Content of non-GAAP earnings and Reg G)
The incremental information content of non-GAAP earnings over GAAP
earnings has not changed over the pre- and the post-Reg G regimes.

23

CHAPTER 5. RESEARCH DESIGN

5.1. Comparison of Disclosure Motives across Time

To jointly test H1 and H2, 1 separately run the following probit regression (l)
for pro-Reg G and post-Reg G sample firm-quarters. I compare the coefficients and the
marginal effects of two different disclosure motives across the time periods to draw
inferences about the temporal change in the association between non-GAAP earnings
disclosures and two strategic disclosure motives.12
Probability (NG=1|x)= G680 + ,BIEQRANK + ,BZLOSS + ,B3NES + [34C0NSENSUS

+ fl-CONTROLS)
(l), firm/quarter index omitted

The dependent variable is non-GAAP earnings disclosure, NG. The dichotomous
variable, NO, is coded as 1 if managers voluntarily disclose non-GAAP earnings in
preliminary earnings announcement press releases for a quarter, 0 otherwise.

EQRANK is an empirical proxy for the firm’s communication motives to disclose
non-GAAP earnings. EQRANK is a rank variable based on EQ, which is defined as the t-
statistic on the coefficient of seasonally differenced GAAP earnings from the following
firm-specific retums-GAAP earnings regression similar to Bradshaw and Sloan (2002)

and Lougee and Marquardt (2004).

MKTADJRE T = a0 + aIAGAAPEARN + a (2),‘3 firm/quarter index omitted

 

'2 I prefer testing for differences in coefficients in equation (1) over time rather than using an alternative
model specification which regresses non-GAAP earnings disclosures on disclosures motives and disclosure
motives interacted with a post-Reg G time dummy variable. See Appendix C for an explanation behind this
choice. See Table 13 for the results from the alternative specification.

'3 For each fimi/quarter, I estimated equation (2) using its time-series observations of at least eight quarters.
Following Lougee and Marquardt (2004), I do not include the current quarter’s earnings for the estimation
of E Q to make sure that the current quarters earnings do not spuriously correlate with HQ

24

Where

MK T ADJRET is cumulative market adjusted returns (i.e., RET - VWRETD) from two
days after the last quarterly earnings announcement to the day after the current quarter
earnings announcement date, and

AGAAPEARN is seasonally differenced GAAP earnings before extraordinary items and

discontinued operations deflated by the market capitalization at the beginning of the
current quarter (i.e., [(COMPUSTAT Data25t — (Data25t.4))/(Data61,* Data17t* Data14 ..

1)] )-

This empirical proxy follows from the previous discussion of non-GAAP earnings as a
substitute for less relevant GAAP earnings.” If managers disclose non-GAAP earnings
when their GAAP earnings are less relevant (i.e., communication motives), I expect a
negative association between NG and EQRANK, that is, [31 in equation (1), will be
negative and significant in equation (1). I use a rank variable, EQRANK, instead of EQ,
because coefficients on EQRANK estimated on a cross sectional basis can be used to
measure well communication motives over time.15 I rank firm specific t-statistics (E Q)
from equation (2) within each calendar quarter in which preliminary earnings
announcements were announced, assigning a number from 0 to 99 as the E QRANK
variable. Thus, EQRANK of 0 indicates the firm’s GAAP earnings relevance is lowest
among sample firms (i.e., suggesting these firms have higher incentives to communicate
better with investors by disclosing non-GAAP earnings) while E QRANK of 99 indicate
the firm’s GAAP earnings relevance is highest among sample firms (i.e., suggesting these
firms have lower incentives to disclose non-GAAP earnings). I use t-statistics for a] in

the equation rather than the coefficient (on) or adjusted R2 of the equation because t-

 

" Although “no single culprit is to blame” for the temporal decline of an association between earnings and
stock returns, Dechow and Schrand (2004, p.114) suggest any firm-specific attributes for the weak
association are likely to be captured by low eamings-retums relation in a parsimonious fashion.

'5 This is because EQ is time variant and, therefore, is not the appropriate specification when one wants to
calculate the difference in the marginal effect of a variable across two different periods. EQRANK provide
one with the ability to make an unequivocal interpretation of the changing marginal effect.

25

statistics parsimoniously combine the mean and variance of GAAP earnings
inforrnativeness in a single variable.16

LOSS, NES, and CONSENSUS are empirical proxies for the firm’s opportunistic
disclosure motives. LOSS, NES, and CONSENSUS are dichotomous variables,
respectively coded as 1 if a firm has a GAAP loss, negative seasonally differenced GAAP
EPS, or GAAP EPS falling short of the consensus EPS estimates, 0 otherwise. LOSS,
NES, and CONSENSUS represent circumstances where firms may disclose non-GAAP
earnings in order to mask poor GAAP earnings that fall short of various earnings
benchmarks.” These empirical proxies follow from the previous discussion of non-
GAAP earnings as a possible means to alter investors’ perception of firm performance
(Hirshleifer and Tech 2003, Lougee and Marquardt 2004; Doyle et al. 2004).18 Thus, if
managers disclose non-GAAP earnings to mask poor GAAP earnings performance falling
short of various earnings benchmarks (i.e., Opportunistic motives), I expect positive and
significant associations between NG and LOSS, NES, and CONSENSUS, respectively

(i.e., positive and significant [32, B3, and B4 in (1)).

 

'6 The results of this dissertation are almost identical when I replace EQRANK with EQ, (11, and the rank of

0.1. The results are qualitatively similar but statistically weaker when I replace EQRANK with R2 or the
rank of R2.

'7 When identifying LOSS and NES, I use diluted GAAP EPS excluding extraordinary items and
discontinued operations (Data 9) after adjusting for stock splits. When identifying CONSENSUS, I compare
GAAP EPS excluding extraordinary items and discontinued operations with the most recent mean
consensus EPS forecast before the current quarter’s earnings announcement date. In making this
comparison, I use either basic (Data 19) or diluted (Data 9) GAAP EPS depending upon whether the
corresponding EPS consensus forecast is basic or diluted. I extract the mean consensus EPS from the
I/B/E/S unadjusted summary file and consider the stock splits factor to address the concern of rounding
errors (Payne and Thomas 2003). For REIT (real estate investment trust) firms, I use FFO (funds from
operations) forecasts, if available, for the comparison.

'8 Out of the three earnings benchmarks described in Degeorge et al. (1999), academic evidence suggests
that firms appear to focus more on avoiding negative earnings surprises (i.e., missing the analysts’
consensus EPS) than losses or negative seasonally differenced earnings (e.g., Brown and Caylor 2005,
Matsumoto 2002). However, the survey of Graham et al. (2005) documents that US. CFOs themselves
believe that the prior year’s quarterly EPS is the most important yet hardest-to-beat number. Also, some
argue that the kinks around earnings and earnings changes distribution may not be evidence of managers’
efforts to meet or beat benchmarks (Durtschi and Easton 2005).

26

I also control for a vector of firm characteristics known to be associated with non-
GAAP earnings disclosures and other voluntary disclosures.19 I include special items (SI,
expected sign: — ) because there often is a mechanical association between non-GAAP
earnings and negative special items. I include the log of market capitalization (LNMK T,
+), the book to market ratio (BT M, — ), the debt to asset ratio (LE VERA GE, +), and the
amount of intangible assets (IN T ANGIBLE, +) to capture the firm’s overall voluntary
disclosure environment. I include the Herfindahl-Hirschman Index (HHI, —) to capture
the firm’s propensity, if any, to conceal information due to competition. I also include the
dichotomous variable indicating high litigation industries (LI T IGA T IONIND, ?) to
capture the firm’s propensity, if any, to avoid additional disclosure for litigation concerns

I infer the consequences of Reg G on managers’ strategic disclosure motives by
comparing coefficients on EQRANK, NES, LOSS, and CONSENSUS across two time
periods.20 If Reg G has decreased (increased) the propensity of firms with opportunistic
disclosure to provide non-GAAP earnings disclosures, I expect the assumed positive
associations between NG and LOSS, NES, and CONSENSUS (i.e., [32, B3, and B4) to be
weaker (stronger) in the post-Reg G period than in the pre-Reg G period respectively:
Bz_Posr— Burma < 0, I33_P0sr— B3_PRr-: < 0, I34_POST- I34_PRE < 0 (Bzyosr- Burma > 0, B3_Posr-
B3_PRE > O, B4_p()s’|' — PURE > 0). If Reg G has increased (decreased) the propensity of
firms with communication motives to non-GAAP earnings disclosures, I expect the

assumed negative association between NG and EQRANK to be statistically stronger

 

'9 This should mitigate concerns that non-GAAP earnings is a part of voluntary disclosure strategy that
merely depicts the firm’s overall voluntary disclosure environment.

20 These two disclosure motives may not be mutually exclusive. For example, LOSS, NES, and
CONSENSUS firms may have legitimate reasons to communicate better. By including EQRANK as another
variable, Equation (1) allows the possibility of the co-existence of these two disclosure motives for a given
firm-quarter.

27

(weaker) in the post-Reg G period than in the pro-Reg G period: B1_pos'r— BLPRE < O

(I31_Posr- Bryan, > 0).

5.2. Comparison of the Market’s Perception of the Transparency of non-GAAP
Earnings

Assessing whether investors’ perception of the transparency of non—GAAP
earnings disclosures are consistent with inferences drawn from the change in managers’
disclosure motives (H3), I run the OLS regression of three-day size adjusted cumulative
returns on both non-GAAP earnings surprises and GAAP earnings surprises after
controlling for known covariates (i.e., book-to-market ratio and losses) and the
interaction terms of these covariates with GAAP earnings surprises. I believe that this
approach is better than an alternative approach of regressing long-run returns on non-
GAAP adjustment and other risk factors (see Appendix D). See equation (3). Note that
the sample in this test is restricted to firms that provide both GAAP and non-GAAP
earnings surprises,2| limiting test observations to firms reporting GAAP and non-GAAP

numbers both in the current quarter and in the four quarters prior.

 

2’ Many firms that disclose non-GAAP earnings also disclose comparable four quarters ago non-GAAP
earnings numbers. In such cases, I extract information about non-GAAP earnings surprises from the current
quarter press release. However, in some cases, fimts simply provide the current period non-GAAP earnings
and there is no available four quarters ago non-GAAP earnings information from the current quarter press
release. In the latter cases, I try to locate the fours quarters ago press releases and extract non-GAAP
earnings, if any. If I still cannot find the four quarters ago non-GAAP earnings, the observation is treated as
missing. If there is a systematic sample selection bias such that only transparent reporters provide non-
GAAP earnings surprises, the result from my study is biased toward the effectiveness of Reg G. In order to
mitigate this bias, I use an alternative specification of earnings surprises (i.e., using I/B/E/S consensus EPS
as the market’s expectation) and find similar results. I do not report this alternative specification in the table
because of the known fundamental econometric problems with using I/B/E/S as proxy for actual earnings
(see Bradshaw 2003, p.331).

28

SADJRET = y, + 7,057 _ GAAP + yzUE _ NONGAAP + 73POST
+ y,UE _ GAAP . POST + 75w; _ NONGAAP . POST
+ moss + 7,BTM + moss . UE_ GAAP+ y,BTM ~UE_ GAAP + a

(3)23 (firm/quarter index omitted)
Where

SADJRE T is cumulative abnormal returns, defined as the sum of size adjusted daily
returns over the three-day window (-1, 0, +1), where 0 is the date of the preliminary
earnings announcement press release. Note that size factor for returns is implicitly
controlled by size-adjusted returns.

UE_GAAP (UE_NONGAAP) is seasonally differenced quarterly GAAP (non-GAAP)
earnings deflated by the market capitalization at the beginning of the current fiscal
quarter adjusted by stock splits factor, and

POST is an indicator variable coded as 1 if the earnings announcement date for a fiscal
quarter is after March 2003, 0 otherwise.

LOSS is an indicator variable coded as 1 if GAAP earnings is negative, otherwise 0.

BT M is the book to market ratio capturing growth factor.

If Reg G has discouraged firms with opportunistic disclosure motives from
providing non-GAAP earnings disclosures and reconciliation information helps investors
better appreciate the information content of non-GAAP earnings post-Reg G, I expect the
assumed positive associations between SADJRE T and UE_NONGAAP to be stronger in
the post-Reg G period than in the pre-Reg G period (i.e., 75>0). If Reg G has discouraged
firms with communication motives from presenting non-GAAP earnings disclosures and
reconciliation information helps investors better discern managers’ strategic motives

post-Reg G, I expect the assumed positive associations between SA DJRE T and

 

2" Note that size is implicitly controlled by the dependent variable. I choose this implicit SIZE control due
to a high variance inflation factor score (i.e., if I include SIZE in the regression, the mean VIF is greater
than 10).

29

UE_NONGAAP to be weaker in the post-Reg G period than in the pre-Reg G period (i.e.,
75<O). Alternatively, if Reg G has discouraged firms with both opportunistic and
communication motives relating to non-GAAP earnings disclosures, I may not expect the
assumed positive association between SADJRE T and UE_NONGAAP to be statistically
different across the pre- and the post-Reg G periods (i.e., 75:0).

I also run a test as to whether POST coefficients on UE_GAAP and
UE_NONGAAP are different from each other (i.e., (75 —— 72) = (74 — yl) ) to mitigate a
concern that 75 may merely capture a factor associated with cotemporaneous changes in
both GAAP EPS and non-GAAP EPS.

The test described in this chapter provides confirmatory evidence as to whether
investors’ perceptions of the information content of non-GAAP earnings disclosures are
consistent with evidence provided by equation (1). I do not draw direct inferences about
the effectiveness of Reg G from this test without parallel evidence from disclosure
equation (1), because my market test expectations are based on the joint hypothesis of

market efficiency and managers’ disclosure motives changes.

30

CHAPTE 6: SAMPLE SELECTION AND DESCRIPTIVE STATISTICS

6.1. Sample Selection

I first determine sample firms of interest and then manually collect press releases
for those firms over the sample period, constructing sample firm-quarters similar to a
panel dataset. Table 1 summarizes my sample selection approach. I randomly draw 1,000
firms from the intersection of CRSP and COMPUSTAT as of 2004. Sample firms in the
initial random sampling meet the following criteria: (1) the fiscal quarter ends (i.e.,
FQENDDT in COMPUSTAT) in 2001:01-2004:06 range, (2) membership on
NYSE/AMEX/NASDAQ, and (3) availability of EPS data from 1999 to 2004. The first
requirement makes sure that inferences are not confounded by including the pre-
Regulation Fair Disclosure (Reg FD) periods (e.g., my sample encompasses only post
Reg FD periods).24 The second requirement is added to facilitate the acquisition of press
releases of sample firms from public databases. The final requirement provides sufficient
time-series of observations to permit estimation of an empirical proxy for communication
motives (t—statistics from firm-specific regressions of returns on GAAP EPS). This
process yields 13,913 initial firm-quarters.

I search preliminary earnings announcement press releases of those firms,

collecting 12,238 press releases from 944 firms.25 I include sample firms that have less

 

2’ In the pre-Reg FD period, the nature and form of communication between firms and investors in
conference calls and other voluntary disclosures may have differed because there was no requirement to
inform all investors of disclosures made to a few. Reg FD imposed such a requirement.

25 I first search PR News Wire and Business News Wire through F ACTIVA to locate preliminary earnings
announcement press releases of sample firms. If I cannot locate press releases of a sample firm through
FACT IVA, 1 further search the sample firm’s website (typically, the investor relations section) to locate
press releases. If I fail to locate press releases even at the company’s website, I further search business
journal articles that seem to “relay” press releases. The number of “relay” cases was less than 1% of total
observations ( 107 out of 12,238).

31

than or equal to one missing observation in their time-series to make my sample firm-
quarters similar to a panel dataset (792 firms’ 10,896 firm-quarters).26 Out of 10,896
press releases, I identify 3,228 non-GAAP earnings disclosures. I exclude 34 more firms
(281 firm-quarters) for my probit regression due to unavailability of financial statement
variables. Finally, out of 3,228 non-GAAP earnings disclosures, I use in my market test
2,458 firm-quarters where non-GAAP earnings surprises for both the current quarter and

four quarters ago are available.27

6.2. Descriptive Statistics

Table 2 provides descriptive statistics about the frequency of non-GAAP earnings
over time. I first divided the sample firm-quarters into three regulatory regimes based on
their earnings announcement dates: (1) 2001 101—2001 :12 (pre Reg G (1) period), (2)
2002:01—2003203 (pre Reg G (2) period), and (3) after 2003:03 (post Reg G period) to
check whether other preceding events in the pre-Reg G (2) period (e.g., the SEC ’3
cautionary advice as of December 2001, a series of corporate scandals, or the passage of
Sarbanes-Oxley Act of 2002 etc.) affected the disclosure frequency of non-GAAP
earnings prior to issuance of Reg G. For the three time periods, 34.6%, 33.96%, and
23.27% of firm-quarters disclosed non-GAAP earnings in their preliminary earnings

announcement press releases, respectively.

 

2” The proportion of firms issuing preliminary earnings releases is qualitatively similar to what Arnir and

Livnat (2005) found: “80% of companies consistently issue preliminary earnings announcements to the
market through a press release prior to their SEC filings (Prelims), 8% consistently file lOQ/Ks without
first issuing a preliminary earnings press release (Filers). The remaining 12% use a mixed strategy.”

27 I assume seasonal random walks of both GAAP and non-GAAP earnings in calculating earnings
surprises. I do not use consensus EPS estimate for the market’s expectation because such an approach
introduces an errors in variables problem in favor of the informativeness of non-GAAP earnings (see
Bradshaw 2003, p.331).

32

This result is consistent with Heflin and Hsu (2005) and Marques (2005) in terms
of the overall declining reporting frequency (see Figure 2). However, there is a subtle
difference between my result and their results. First, disclosures are most prevalent in the
highly capitalized firms (S&P 500 firms) followed by bigger firms (I/B/E/S-
COMPUSTAT firms), then by my sample firms. This implies that inferences drawn from
highly capitalized firms (S&P 500 or I/B/E/S firms) may not depict the changing
landscape of non-GAAP disclosure in the population when disclosures are correlated
with firm characteristics (e.g., here, SIZE). Second, a )6 test of independence shows that
there is no significant frequency difference between pre Reg G (1) period and pre Reg G
(2) period (p-value=0.60). The decline in frequency of disclosures occurs only after Reg
G is effective (p-value < 0.001). This indicates that the decreased frequency of non-
GAAP earnings in my sample firm-quarters is not likely to be triggered by other
preceding events. Thus, for the remainder of this paper, I provide various analyses based
on two regulatory time periods; the pre- and post-Reg G periods (i.e., 2001:01—2003z03
vs. 2003:04-2004:08).28

Table 3 provides industry breakdowns (Barth et al. 1998) for the disclosure of
non-GAAP earnings. Reflecting my random selection procedure, the industry
composition of sample firm-quarters (row (2)) closely parallels that of

COMPUSTAT/CRSP firms as of 2004 (row (1)). The industries where non-GAAP

 

2” Fiscal end quarter periods of my sample firms span from 2001201 to 2004206. Correspondingly, periods
of preliminary earnings announcement releases span from 2001 :01 to 2004208.

33

earnings are more pronounced than their sample composition (row (6)) include insurance
and real estate, chemicals, computer, transportation, and service and others.”

Table 4 compares non-GAAP EPS, I/B/E/S actual EPS, GAAP EPS excluding
extraordinary items and discontinued operations, and S&P Core EPS (see Blitzer and
Friedman (2002) for the definition) across two regulatory time periods conditional upon
non-GAAP earnings disclosures. Three points are noteworthy. First, all EPSs but non-
GAAP EPS have increased over time ((a) through (d), p-values less than 0.05 except the
mean difference of non-GAAP EPS), indicating that earnings performance in the post-
Reg G period is better than in the pre-Reg G period. Second, non-GAAP EPS always is
greater than other EPS metrics ((e) through (g), p—values less than 0.01), but the
magnitude of difference between non-GAAP EPS and other EPSs vary across EPS
definitions. This implies earnings adjusted by managers are not identical to earnings
adjusted by other constituents (i.e., equity or credit analysts). Third, upward biases in
non-GAAP earnings have decreased somewhat over time. Compared to GAAP EPS, the
upward bias in non-GAAP earnings seems to have decreased over time (see the row (0
and columns (h) and (i), p-values less than 0.01). However, I do not find a similar
decrease regarding I/B/E/S Actual EPS and S&P Core EPS (except the mean statistics in
(e) and (h)). This raises the possibility that researchers can make an erroneous inference
regarding the effect of Reg G when they restrict their sample firms to samples followed
by specific constituents or screened by specific firm characteristics (e.g., highly

capitalized firms).

 

29 Some may view that the results of this dissertation are confounded by a 9-11 effect because of the high
frequency of non-GAAP earnings disclosures in the transportation industry in Table 3. However, the results
are unaffected after dropping this industry in my tests for H1 through H3.

34

I collect exact labels used for non-GAAP earnings in each press release to provide
descriptive statistics about various nomenclatures which firms employ for their non-
GAAP earnings disclosures.30 Surprisingly, Table 5 reports that “pro-forma” earnings and
its variations constitute only 13% of the nomenclatures, preceded by “earnings excluding
xxx” and “(adjusted) EBITDA and its variation.” The implication is that key-word search
using a specific string (e.g., “pro-forma”) may generate under-represented sample firm-
quarters if such firms’ labeling choices are endogenously determined by firm’s
opportunistic disclosure motives. I

Finally, Table 6 provides descriptive statistics about adjustments firms make
across the two time periods. Two points are noteworthy. First, while the restructuring
charge adjustment has increased, amortizations adjustments have decreased over time.
This implies that F AS 142, but not F AS 146, may contribute slightly to the decreased
frequency of non-GAAP earnings in the post-Reg G period.3 1 Second, non-
recurring/special items not clearly specified in press releases has slightly decreased but
still remains in the post-Reg G period, implying that the consequence of Reg G on the

transparency of non-GAAP earnings is still an empirical question.

 

’0 Untabulated descriptive statistics indicate that about 14.3% of sample firms that disclose non-GAAP
earnings provide more than one non-GAAP earnings metric, and that the frequency of multiple non-GAAP
earnings conditional upon non-GAAP earnings disclosures has decreased over time (19.43% in the pre-Reg
G period vs. 9.6% in the post-Reg G period, )8 test independence p-value <0.01).

3' Beginning after December 15, 2001 FAS 142 requires that intangible assets with indefinite lives no
longer be amortized over their usefirl life (i.e., previously up to forty years), but rather be subject to annual
tests of validity of the previous assigned value. To the extent that the frequency of both amortization and
related non-GAAP adjustments decreases, the frequency of non-GAAP earnings disclosures is expected to
decrease.

35

CHAPTER 7: EMPIRICAL RESULTS

7.1. Characteristics of Disclosed non-GAAP earnings

In Table 7, 8, and 9, I provide descriptive statistics of various properties regarding
“disclosed” non—GAAP earnings. In Table 7, I provide descriptive statistics of the
presentation quality of non-GAAP earnings in press releases. EMPSCORE measures the
emphasis placed on non-GAAP earnings (e.g., Bowen et al. 2005), coded from 3, where
non-GAAP earnings was highlighted in the headline or in the lead paragraph of press
releases over GAAP earnings, to 0, where non-GAAP earnings are not emphasized at all.
DISCQUAL measures the amount of reconciliation information provided in press releases
(e.g., Zheng and Zhang 2005; Wallace 2002), coded from 3, where the firm provides a
detailed pro-forma income statement or a tabular/columnar reconciliation table, to 0,
where there is no clear definition of non-GAAP earnings. Panel A shows that firms are
less likely to emphasize non-GAAP earnings in the post-Reg G period than in the pre-
Reg G period (Z-value<0.001). Also, firms are more likely to provide detailed
reconciliation information in the post-Reg G period than in the pre-Reg G period (Z-
value<0.001). These results suggest that reporting firms have complied with Reg G
requirements, improving the presentation quality of non-GAAP earnings in preliminary
earnings announcements.

Table 8 presents descriptive statistics relating to whether disclosed non-GAAP
earnings make the firm’s ability to meet or beat earnings benchmarks look better. The
frequency of non-negative non—GAAP earnings masking GAAP losses (H YPE_LOSS)

has not decreased over time (row (1) and (2)). The frequency of positive seasonally

36

differenced non-GAAP earnings masking negative seasonally differenced GAAP
earnings (H YPE_NES) has declined (from 13.43% to 10.26%, p-value 0.01, row (3)).
Once non-GAAP earnings are disclosed for negative seasonally differenced GAAP
earnings, however, the frequency of positive seasonally differenced non-GAAP earnings
does not change over time (p-value=0.961, row (4)). Rows (5) and (6) suggest that the
firm’s tendency to disclose non-GAAP earnings masking GAAP earnings falling short of
analysts’ consensus earnings forecasts has slightly decreased over time both with and
without the condition of GAAP earnings falling short of consensus earnings. Finally, the
percentage of non-GAAP earnings that were adjusted downward from GAAP earnings
(NEGADJ) has increased over time by 4.62% (p-value < 0.01), suggesting firms in the
post-Reg G period are more likely to clearly communicate by disclosing non-GAAP
earnings, because these firms apparently had an option not to disclose income decreasing
non-GAAP earnings.32 Overall, the descriptive statistics of “disclosed” non-GAAP
earnings in Table 8 provide mixed preliminary (univariate) evidence on whether Reg G
curtails the firm’s general tendency to present non-GAAP earnings opportunistically.
One thing to note is that EMPSCORE and DISCQUAL are positively correlated
(p-values <0.001), suggesting that firms emphasizing non-GAAP earnings are more
likely to provide detailed reconciliation information (Table 9). This is consistent with
Hutton et al. (2003), finding that firms are more likely to provide additional quantitative
disclosures when they need to bolster the credibility of their favorable voluntary

disclosures.

 

32 For example, Hirshleifer and Teoh’s model (2003) characterizing the pro-forma earnings disclosure
decision views GAAP earnings as a lower-bound of management defined earnings. Many financial press
articles also indicate that pro—forma earnings are always income increasing. Thus, firms that disclosed non-
GAAP earnings that are lower than GAAP earnings are likely to have a motivation to better communicate
with investors by disclosing non-GAAP earnings.

37

7.2. Changes in Disclosure Motives across Time (Hl & H2, Equation (1))

H1 and H2 suggest two alternative explanations for the decline in non-GAAP
earnings reports post Reg G. Table 10 compares the characteristics of firms that disclose
and do not disclose non-GAAP eamings across the pre- and post-Reg G time periods. In
the pre-Reg G period, firms that disclose non-GAAP earnings have lower EQ and
E QRANK and higher LOSS, NES, and CONSENSUS than firms that do not. This is
consistent with the notion that, in the pre-Reg G period, both firms with low earnings
relevance (i.e., firms with communication motives) and firms with poor GAAP earnings
performance (i.e., firms with possible opportunism motives) disclose non-GAAP earnings.
Although these differences also are similarly pronounced in the post-Reg G period, the
difference in E QRANK seems to be bigger in the post-Reg G period than in the pre-Reg G
period while the differences in LOSS, NES, CONSENSUS seem to be either smaller or
insignificant in the post-Reg G period than in the pre-Reg G period.

Table 11 shows both Pearson (upper) and Spearman (lower) correlations of the
dependent variable (NG), independent variables (E QRANK, NES, LOSS and
CONSENSUS) and control variables. Consistent with the descriptive statistics in Table 10,
the disclosure of non—GAAP earnings (NG) are significantly correlated with independent
variables (EQRANK, NES, LOSS and CONSENSUS) in predicted directions (negative for
E QRANK and positive for LOSS, NES, and CONSENSUS) in both the pre- and the post-
Reg G period. While the coefficients for NO and EQRANK are higher in the post-Reg G

period than in the pre-Reg G period, the coefficients for NG and

LOSS/NES/ CONSENSUS are smaller in the post-Reg G period than in the pre-Reg G
period.

Table 12 provides the results from the probit regression examining non-GAAP
earnings disclosure motives both pre- and post-Reg G. Within each regression (column
(A) and (B)), all control variables have the expected signs and are statistically significant
except the high litigation industry dummy variable (LIT GA T IONDUM), which had no
predicted sign. Firms with negative special items (SI) are more likely to provide non-
GAAP earnings. Firms with higher market value (LNMK T), more leverage (LE VERA GE)
and greater intangible assets (IN T ANGIBLE) are more likely to provide non-GAAP
earnings. Firms with higher grth potential (BT M) and lower competition (HHI) are
more likely to conceal information (e. g., Bamber and Cheon 1998).

Column (A) of Table 12 shows the association between NG and non-GAAP
earnings disclosure motives in the pre-Reg G period. I do not find a significant negative
association between NG and EQRANK in the pre-Reg G period (column (a), p-value
0.197). In contrast, I find evidence consistent with firms with opportunistic motives
(LOSS, NES, and CONSENSUS) disclosing non-GAAP earnings in the pre-Reg G period
(column (a), p-values 0.000 respectively). The results in column (A) imply that firms
primarily disclosed non-GAAP earnings for opportunistic but not communication
motives in the less regulated environment. Column (B) of Table 12 shows the association
between NG and disclosure motives in the post-Reg G period. I find a significant
negative association between NG and E QRANK, suggesting that firms with

communication motives disclose non-GAAP earnings in the post-Reg G period. While I

39

also find significant positive associations between NG and LOSS/ CONSENSUS, I do not
find such an association between NG and NES in the post-Reg G period.

In order to assess whether Reg G affects the magnitude of these associations over
time (H1 and H2), I compare the coefficients of disclosure motive variables across the
two time periods (x2 statistic from the Chow test in column (C)).33 The coefficient
difference on EQRANK across time is negative and significant (p-value=0.029),
suggesting that the prOpensity of firms to disclose non-GAAP earnings for better
communication has increased over time.34 Specifically, compared to firms at the top of
EQRANK, firms at the bottom of EQRANK (i.e., higher communication motives) in the
post-Reg G period are about six percent more likely to disclose non-GAAP earnings than
in the pre-Reg G period. In parallel, the positive association between NG and LOSS is
lessened in the post-Reg G period (p-value=0.059), suggesting that compared to profit
firms, LOSS firms in the post-Reg G period are about seven percent less likely to disclose
non-GAAP earnings than LOSS firms in the pre-Reg G period. Similarly, while NES
firms in the pre-Reg G period are 5.7% more likely to disclose non-GAAP earnings, this
tendency no longer exists in the post-Reg G period (see column (B), p-value=0.135). The
results regarding LOSS and NES suggest that the propensity of firms to issue non-GAAP
earnings to opportunistically mislead investors’ perceptions about GAAP earnings
performance has decreased over time.

All of these results (E QRANK, LOSS, and NES) are consistent with Reg G having

the consequences intended by Congress and the SEC. However, I do not find similar

 

3’ Consistent with my discussion, it appears that FAS 142 contribute to the decreased frequency of non-
GAAP earnings. See the change of coefficients on [N TA NGIBLE in column (C) (P-value<0.01).

3" In order to measure the economic effect, I use the coefficient estimates in the model to calculate the
marginal effect of EQRANK, LOSS, NES and CONSENSUS on the probability of presenting non-GAAP
earnings (Wooldridge 2002).

40

evidence on the association between NG and CONSENSUS. That is, firms with GAAP
earnings falling short of consensus EPS estimates in the post-Reg G period are 2.5%
more likely to disclose non-GAAP earnings than the similar firms in the pre-Reg G
period (p-value=0.032). This finding suggests that regulation of non-GAAP earnings may
not have alleviated the concern that firms use non-GAAP earnings’ to meet or beat
analysts’ consensus EPS forecasts. We need to recall, however, the result in Table 8 that
the frequency of “disclosed” non-GAAP earnings that actually meet or slightly beat the
consensus EPS by up to 2 cents has decreased over time (4.53% difference, p-
value=0.046). Taken together, the evidence from this section is likely to be more

consistent with Reg G having consequences consistent with the objectives of Congress

and the SEC.

7.3. Changes in the Market’s Perception of the Transparency of non-GAAP
Earnings

The previous section suggests that managers with benign communication motives
(opportunistic motives) are more (less) likely to provide non-GAAP earnings in the post-
Reg G period than in the pre-Reg G period. This section provides confirmatory evidence
on whether investors’ perceptions of the transparency of non-GAAP earnings are
consistent with these changes in managers’ disclosure motives by examining whether the
incremental information content of non-GAAP earnings surprises over GAAP earnings
surprises changes pre- and post-Reg G.

Table 14 presents the results of OLS regression of equation (3). The sample in

this test is restricted to a sub-sample of firms where sufficient time series data exist to

41

calculate seasonal difference in non-GAAP earnings (2,458 observations out of 3,228
non-GAAP disclosures observations). Alter running the model using the initial 2,458
observations, I eliminated 141 influential observations.35 The coefficient on GAAP
earnings surprises (UE_GAAP) is positive in the pre-Reg G period, but only marginally
(P-value=0.096). This is not surprising because firms that disclose non-GAAP earnings
are more likely to exhibit low GAAP earnings quality. The coefficient on non-GAAP
earnings surprise (UE_NONGAAP) in the pre-Reg G period is positive but statistically
insignificant. The incremental information content of non-GAAP earnings in the pre-Reg
G period may not be significant for two possible reasons: ( 1) investors perceive that firms
were disclosing non-GAAP earnings previously for opportunistic motives and/or (2)
investors cannot discern the disclosure motives but are so skeptical about the information
content of all non-GAAP earnings that they undervalue non-GAAP earnings disclosures
provided for either reason. The sample period of prior studies reporting the incremental
information content of actual non-GAAP earnings covers years preceding 2001. Thus, the
results may reflect skeptical investors in 2001 and 2002 years due to recent corporate

scandals.36

 

35 Following Belsley, Kuh and Welsch (1980), I identified 141 observations that have the absolute value of
RSTUDENT greater than 2 or the absolute value of DFITTS greater than 2*sqrt(2,458 obs/9 variables).
The estimation provided in Table 14 is the result excluding those influential observations. The sign and the
significance of two interaction terms of interest do not change when including those influential
observations. Because 1 controlled for the size effect implicitly by using size adjusted returns,
multicollinearity does not seem to affect the estimation (the mean variance inflation factor score is just
1.12). Possible heteroskedasticity is controlled for by providing heteroskedasticity-robust estimation.

3° Untabulated results show that the incremental information content of non-GAAP earnings is more
significant when GAAP earnings are less relevant (i.e., EQRANK>=50) than when GAAP earnings are
more relevant in the pre-Reg G period (i.e., EQRANK<50). This suggests that the market distinguishes the
transparency of non-GAAP earnings based on historical GAAP eamings’ relevance. However, the
incremental information content of non-GAAP earnings in the pre-Reg G period does not vary based upon
NES, suggesting that the market’s ability to distinguish non-GAAP eamings’ transparency might have been
limited.

42

In contrast, the coefficients on both UE_GAAP (y. + 14, p-value=0.0054) and
UE_NONGAAP (y; + 75, p-value=0.0198) in the post-Reg G period are positive and
significant, suggesting both GAAP earnings and non-GAAP earnings are priced in the
post Reg G period. The interaction term of POST and UE_GAAP (74) is positive and
marginally significant at the 0.07 level, indicating that the information content of
earnings surprises in the post-Reg G period increases (consistent with Table 10 of Heflin
and Hsu (2005)). The variable of interest in this study is the incremental information
content of non-GAAP earnings surprises in the post-Reg G period
(POST’ UE_NONGAAP). The coefficient of the interaction’term (75) is positive and
significant (p-value=0.028), suggesting that the information content of non-GAAP
earnings surprises in the post-Reg G period also increases. LOSS, BT M and the
interactions of LOSS and BT M with UE_GAAP all have predicted signs but BT M related
variables are not statistically significant.

Some may view, however, that the increases in the coefficients of both GAAP and
non-GAAP earnings in the post Reg G period imply other contemporaneous changes
rather than the Reg G effect. In order to mitigate this concern, I first provide the test
result that the POST coefficient UE_NONGAAP is greater than the POST coefficient on
UE_GAAP. Table 14 suggests that while the combined coefficients of the pre- and post-
UE_GAAP (i.e., y] + 74 ) and UE_NONGAAP (i.e., y; + y5) are significantly different
from zero, the information content change of non-GAAP earnings over the two periods
(i.e., (75 — yz) ) is greater than that of GAAP earnings (i.e., (Y4 — 71)) (p-value of 0.0575),
suggesting that the significant and positive 75 is not merely the manifestation of a

contemporaneous factor associated with both GAAP and non-GAAP EPS.

43

In order to further investigate the source of non-GAAP earnings’ incremental
informativeness in the post Reg G period, I run the same regression for two sub-sets of
firms that discontinued and continued disclosing non-GAAP earnings in the post Reg G
period. First, Table 15 suggests that firms that disclosed non-GAAP earnings in both the
pre- and post-Reg G periods show the incremental informativeness of non-GAAP
earnings even in the pre-Reg G period. For this type of firms, the incremental
informativeness in the post Reg G is not observed. When I regress the size adjusted
returns on both GAAP and non-GAAP earnings for the firms that discontinued disclosing
non-GAAP earnings in the post-Reg G period (Table 16), however, I find that both
GAAP and non-GAAP earnings are not informative. Thus, Table 15 and 16 suggest that
results in Table 14 are likely to come from the fact that firms that had noisy non-GAAP
earnings had dropped their non-GAAP earnings disclosures in the post-Reg G period.

Overall, the results presented in Table 14 through Table 16 are consistent with the
idea that investors in the post-Reg G period are more likely to value the information
content of non-GAAP earnings. Assuming market efficiency, I interpret the results as
evidence consistent with the story that Reg G affects managers’ disclosure motives of
non-GAAP earnings in a more transparent way, and investors concurrently view non-

GAAP earnings as more transparent in the post-Reg G period.

44

CHAPTER 8: CONCLUSION

In this dissertation, I examine whether the decreased frequency of reporting non-

GAAP earnings in the post-Regulation G period reflects either intended or unintended
consequences of Congressional intervention. Based on 792 firms’ quarterly earnings
announcement press releases spanning from 2001 to mid 2004, I first document that firms
in the post-Regulation G period provide more information reconciling non-GAAP
earnings to GAAP earnings and put less emphasis on non-GAAP earnings over GAAP
earnings than in the pre-Reg G period. Thus, firms appear to be complying with Reg G’s
requirements.

Borrowing from prior literature that posits firms may have two strategic disclosure
motives (communication vs. opportunism) for the disclosure of non-GAAP earnings, I
examine how managers’ underlying strategic motives change over the pre- and the post-
Reg G periods. Specifically, I regress the disclosure of non-GAAP earnings on variables
(1) measuring GAAP earnings’ historical relevance (proxied by t-statistics from a firm-
specific returns-eamings regression) and (2) identifying situations where firms have the
incentives to disclose non-GAAP earnings to mask the fact that GAAP has fallen short of
various earnings benchmarks (zero, last year same quarter earnings, and consensus EPS
from analysts). The results from the probit regression suggest that compared to the pre-Reg
G period, the same sample firms in the post-Reg G period are more likely to provide non-
GAAP earnings when they have low GAAP earnings relevance (i.e., high communication
motives). In contrast, compared to the pre-Reg G period, the same sample firms in the post-

Reg G period are less likely to provide non-GAAP earnings when they have GAAP losses

45

or negative seasonally differenced GAAP earnings. Although I do not find the same
declining propensity for firms with GAAP earnings falling short of consensus EPS
forecasts, descriptive statistics suggest that the frequency of non-GAAP earnings that meet
or slightly beat the consensus EPS up to 2 cents decreased over time. Taken together, it
appears that Reg G encourages firms with communication motives to continue disclosing
non-GAAP earnings while discouraging firms with opportunistic motives from disclosing
non-GAAP earnings.

To investigate how investors perceive the transparency of non-GAAP earnings
across two regulatory time periods, I regress three-day size adjusted cumulative returns
on both seasonally differenced GAAP and non-GAAP earnings. In contrast to the results
from most prior studies that used I/B/E/S actual earnings as a proxy for non-GAAP
earnings in sample periods preceding my period, I do not find the incremental
information content of non-GAAP earnings over GAAP earnings in the pre-Reg G period
(2001:01-2003z03). However, I find non-GAAP earnings incrementally informative over
GAAP earnings in the post-Reg G period (2003:04-2004:08). The results are consistent
with a joint statement that, compared to the pre-Reg G period, firms in the post-Reg G
period disclose their non-GAAP earnings in a more transparent way, and investors
perceive the increased transparency by assigning significant and positive returns to
seasonally differenced non-GAAP earnings.

The current dissertation makes two contributions to the literature. First, my study
provides timely and interesting evidence that responds to calls for research on the effects
of Reg G (and the Sarbanes-Oxley Act of 2002) on firms’ (mis)use of pro-forma earnings

disclosures (e.g., Dechow and Schrand 2004, 116; Stuart 2004). This dissertation

46

examines both temporal change in managers’ disclosure motives and temporal change in
investors’ perceptions of the transparency of non-GAAP earnings to draw inferences
about the consequences of Reg G on motives for disclosing non-GAAP earnings. Thus,
this dissertation differs from Heflin and Hsu (2005), Marques (2005) and Zheng and
Zhang (2005) in that it focuses on (1) why, as opposed to whether or not, firms
discontinued presenting non-GAAP earnings disclosure following Reg G and (2) whether
the capital market’s reaction to non-GAAP earnings adjustment is consistent with the
implied disclosure motive change post Reg G.

Second, this dissertation adds to the debate on the transparency of pro-forma
earnings (e.g., Bradshaw 2003) by providing evidence that pro-forma earnings were used as
a means to obfuscate investors’ perception about firm performance in the pre-Reg G
period. The dual findings of this study that managers report non-GAAP earnings in a more
transparent way in the post-Reg G period and that the market values this improvements in
non-GAAP earnings transparency in the post-Reg G period suggest that many firms
disclosed non-GAAP earnings opportunistically in the unregulated environment. In
addition, investors perceived such disclosures as potentially opportunistic, biasing

downward the pricing on non-GAAP earnings in the pre-Reg G period.

47

APPENDIX A: SUMMARY OF THE RULES: “CONDITIONS FOR USING N ON-
GAAP FINANCIAL MEASURES”

The Rules consist of three parts: (1) Regulation G; (2) amendments to the SEC
filings; and (3) amendment to Form 8—K. The role of Regulation G is to provide general
guidelines in releasing pro-forma earnings either in SEC filings or other disclosure
venues. Amendments to the SEC filings contain specific rules about which items to
exclude or include for non-GAAP earnings.

Under the Rules, if firms want to use pro-forma earnings in their SEC filings (e. g. their
lO-K), they have to strictly abide by some restrictive rules discussed in (2) below.
However, if firms simply want to disclose pro-forma earnings in public communications
other than the SEC filings (e. g. press releases or conference calls), they need only abide
by Regulation G.

The Rules require firms that are releasing public non-GAAP earnings disclosures
other than SEC filings to provide pro-forma earnings in their Form 8-K. Firms who so
utilize their 8-K are treated as “furnishing” this information as opposed to “filing” it.
Thus, firms can avoid the calculation restrictions in the SEC filings for pro-forma

earnings in the 8-K. I will explain the role of the three rules below.

(1) Regulation G: This regulation requires firms to present: (1) the most directly
comparable financial measure according to GAAP; and (2) a reconciliation of the
differences between GAAP and non-GAAP measures in the form of schedules or any
other clearly understandable method, when firms release any non-GAAP measures

publicly. There is no definition of “most directly comparable GAAP measures” in the

48

Rules. However, footnote 26 of the Rules provides general guidance. For example, if a
firm uses pro-forma earnings (e. g. self-defined cash flows), the firm may present
equivalent GAAP earnings (cash flows.)

Although the regulatory focus of Regulation G is pro-forma earnings, Regulation
G also applies to non-GAAP financial measures to circumscribe all measures that are
different from a performance measurement in GAAP or a measure of liquidity in
Statement of Cash Flow under GAAP. Thus, not only alternative performance measures
(e. g. pro-forma earnings, cash earnings, adjusted earnings, etc.) but also alternative
liquidity measures (e. g. Tyco Intemational’s self-defined “free cash flow from
operations”) and other operational measures are also covered by Regulation G. This
regulation G applies to all entities except for registered investment companies if the
entity publicly releases or discloses any material information that includes non-GAAP

financial measures.

(2) Amendments to Item 10 of Regulation S—K (the SEC filings) The purpose of the
amendments is to govern a case where a registrant presents non-GAAP financial
measures in its SEC filings. In such cases, in addition to Regulation G’s two presentation
requirements, a firm must present the most directly comparable GAAP financial
measures “with equal or greater prominence” and add two additional statements: (1)
disclosure as to why management believes non-GAAP financial measures are useful; and
(2) disclosure of any additional purposes for which management uses non-GAAP

measures paragraphs (e)(1)(ii) of Item 10 (considered “additional disclosures”)

49

In addition, the Rules impose additional direct restrictions on the presentation of
non-GAAP information in SEC filings (paragraphs (e)(1)(ii) of Item 10, considered
“calculation constraints”). That is, non-GAAP financial measures in the SEC filings
should: (1) not exclude cash settlement charges or liabilities; (2) not use non-GAAP
earnings to remove or smooth items described as non-recurring if such items are likely to
recur within 2 years or occurred within the previous 2 years; and (3) not present non-
GAAP earnings on the face of the GAAP financial statements. Commentators seem to
agree these constraints would effectively stamp out pro-forma earnings in SEC filings
(KPMG 2003).

Thus, it is generally expected many firms would use pro-forma earnings in
alternative disclosure venues other than lO-K. However, even in that case, the Rules still

require firms to provide pro-forma earnings information in 8-K as follows.

(3) The amendment to Form 8-K: This change adds a new Item 12 to Form 8-K, which
requires that whenever an entity discloses material non-public information regarding
results of operations or financial conditions for a completed quarterly or annual period,
the firm must provide such information dissemination in the Form 8-K within 5 business
days. Earnings releases are highly likely to be simultaneously subject to Regulation FD
(Item 9 in Form 8-K), which also requires firms to furnish earnings releases to 8-K. Thus,
Form 8-K could meet both regulations if disclosure of non-GAAP were furnished to the
SEC within the common timeframe of both requirements.

Although it seems the pro-forma earnings provided for Form 8-K are subject to

the aforementioned calculation constraints for the SEC filings, the Rules explicitly treat

50

this information provision as “furnishing” rather than as filing under item 5 of Form 8-K.
This special treatment in the Rules (as opposed to filing with the SEC) allows firms to
enjoy two important benefits: (1) furnished (as opposed to filed) information is not
directly subject to Section 18 (Liability for Misleading Statements) of the Exchange Act;
and (2) furnished (as opposed to filed) information is not subject to the aforementioned
calculation constraints for the SEC filings.

Form 8-K information provision applies to any releases or announcements
irrespective of the inclusion of non-GAAP financial measures. However, if such releases
include non-GAAP financial measures, they trigger both Regulation G and additional
disclosure requirements for the SEC filings; i.e. firms that present pro-forma earnings
other than in SEC filings still have to present: (1) the most directly comparable GAAP
financial measures “with equal or greater prominence”; (2) the reconciliation information
between pro-forma earnings and GAAP earnings; and (3) two additional explanations for
the reasons management believes non-GAAP financial measures are useful, as well as

any additional purposes for which management uses non-GAAP measures, in 8-K.

5]

APPENDIX B: COMPARISON OF SAMPLE SELECTION APPROACHES

Many studies use the difference between I/B/E/ S actual earnings and GAAP
earnings as an empirical proxy for the disclosure of non-GAAP earnings (e.g., Doyle et
al. 2004, Heflin and Hsu 2005). This empirical proxy has three issues. First, conceptually,
earnings adjusted by managers are not identical to earnings adjusted by analysts because
of different incentives relating to managers and analysts. Second, the disclosure,
magnitude, and adjustment of non-GAAP earnings are often significantly different from
those of earnings adjusted by either analysts or I/B/E/S. For example, Bhattacharya et al.
(2005) shows that “a mechanical screen of the COMPUSTAT-UB/E/S population only
results in a 64% hit rate in identifying actual manager-reported adjusted GAAP earnings
figures.” Finally, the intersection of I/B/E/S and COMPUSTAT does not translate well
into the demographic profile of non-GAAP earnings disclosures (e.g., firm size and
industries), limiting inferences about the impact of Reg G on the population of firms (see
Bhattacharya et al. 2005).

In order to avoid these issues, later studies use key-word search programs through
press releases databases (e. g., Zheng and Zhang 2005), and identify non-GAAP earnings
disclosure firms. This approach is presumably better than the former approach in terms of
the measurement error of non-GAAP earnings disclosures. Because language seems to
play an important role in pro-forma earnings disclosures (Bradshaw 2003; Johnson and
Schwartz 2005; Wallace 2002), however, it is likely that the sample firms in the post-Reg
G period from the keyword search approach may suffer from a self-selection issue. For

example, if firms tend to avoid the nomenclature of “pro-forma” earnings in the post-Reg

52

G period to avoid scrutiny from investors, a study that extracts sample firms based on the
“pro-forma” key-word may generate under-represented sample firms, leading to an
erroneous conclusion.

The sample selection strategy adopted in this dissertation is to follow a group of
pre-determined firms for a time period of interest to construct a panel dataset. This
approach is likely to effectively mitigate the issues raised in the previous two approaches.
Although the sample selection approach in Marques (2005) resembles my sample
selection, her sample firms of S&P 500 membership are biased to bigger firms. If bigger
firms had been subject to severe public scrutiny even in the pre-Reg G period, studies
with bigger firms might conclude that there was no evidence of the misuse of pro-forma
earnings and, therefore, no justification of the SEC’s intervention. My sample firms of
randomly drawn firms from COMPUSTAT are designed to help me draw inferences

about the population regarding managers’ changing behaviors.

53

APPENDIX C: MODEL SPECIFICATION ISSUE FOR EQUATION (1)

I separately run equation (1) for both the pre— and post- Reg G period and
compare the coefficients from two regression results to draw inferences about the effect
of Reg G on the firm’s disclosure motives relating to non-GAAP earnings disclosures. I
prefer equation (1) over an alternative model specification such as a regression of non-
GAAP earnings disclosures on disclosures motives and the interaction terms of disclosure
motives and the post-Reg G time dummy variable as in (Al)

Probability (NG=1) = (I) (E [Communication + ,BzOpportunism + E 3POS T +
,84Communication *POST + fl5Opportunism *POS T + Xfl) (Al)

While the coefficient of an interaction term in a linear model (i.e., OLS) is
straightforward to interpret because the coefficient could translate into the marginal effect
of the interaction term, it is not always straightforward to translate the coefficient on such
interaction terms in a non-linear model (e. g., probit in this case) into the marginal effect
of variables across the partitioning variable (i.e., regulatory time periods in this case).

For example, suppose a simplified probit model as in (A2).

Probability (NG=1|EQ, POST, X) = <1) ([3,15Q + fl2P()ST + [33EQ~POST+ X/i) (A2)

where
(I) is the normal cumulative distribution function and NG, EQ, POST stand for the

disclosure of non-GAAP earnings, GAAP earnings quality, and the Post-Reg G dummy
respectively.

54

The coefficient of EQ-POST from the probit model is
6<D()/ 6(EQ- POST) = fl3<l>'(-), and “most applied researchers interpret this as the
interaction effect.” (Ai and Norton 2003, Norton 2004). However, the full interaction
effect in this case must be the discrete difference (with respect to POST) of the single
derivative (with respect to EQ). That is,
A<a<1>(-)/ aEQ)/ APOST = (a + a )<I>'((fl. + it. )EQ + ,3, + Xfl) — flr<l>'(/31EQ + Xfl).
Because we have three variables of interest, it is practically non-trivial to compute
the exact full interaction effect. Thus, my approach gives a reliable yet conservative

methodology to address H1 and H2. I provide the results of the alternative specification

with a Post Reg G dummy variable in Table 13. Results are qualitatively identical.

55

APPENDIX D: PROBLEMS OF THE LONG-RUN RETURNS TEST

Many studies regress long run abnormal returns on the difference between non-
GAAP earnings and GAAP earnings deflated by assets or sales to draw inferences about

whether non-GAAP earnings are misleading or not, as in (A3) (e. g. Doyle et al. 2003;

Zheng and Zhang 2005).
One to three year = f (the diflerence between non-GAAP earnings and
long-run returns GAAP earnings deflated by assets or sales

| beta, size, book-to-market, momentum, other risk factors)
(A3)
These studies assume that the negative (no) association between long-run returns
and the difference is evidence that non-GAAP earnings are (not) misleading investors.
The test like (Al) has two issues. First, as well discussed by Easton (2003), it is hard to
conclude that the negative association is evidence of misleading non-GAAP earnings
because often the nature of adjustment changes the firm’s risk characteristics. For
example, firms that went through restructuring and adjusted the amount would experience
low returns in the future (e.g., the negative association in (A3)) not because non-GAAP
earnings are misleading but because the firm lowered risks by restructuring. Second, the
negative association in (A3) is often driven by some extreme observations and, therefore,
depends heavily on the power of the test. Thus, given the decreased number of non-
GAAP earnings observations in the post Reg G, it is not obvious that no association in
the post Reg G period can translate into evidence of non—GAAP earnings being

transparent in the post Reg G (e. g., Zheng and Zhang 2005).

56

 

oE—oEfi. tan U news—=3“ A PEEK

57

Figure 2. Frequency of Non-GAAP Earnings Disclosures by Calendar Quarters

—0—Yi (2005): 792 randomly selected firms from the COMPUSTAT/CRSP .

rsectl
—I - Heflin and

Inte o

Hsu (2005): 1,403 I/B/E/S Coverage firms

- A - Marques (2005): 314 S&P 500 Firms

Frequen

0.700

 

cy of Non-GAAP Earnings Disclosures

 

0.600 ~

0.500

0.400 ‘

0.300 -,

0.200 ,

0.100 ~~

 

 

 

. A, W i “ “fi—T
0‘ ' ‘ s E
_2" A~A 4". .A E __u_ _._
T V ‘7 s I
.\ s :
/ 9 o.\ E.‘ i
,- L) V ‘ : . ‘--
r y o I. T I/ ‘ :K “ ’-
‘ : I
I -:-l’

      

Post- Reg G Period

 

 

0.000

 

f ---------------------- >
I

2001 2001 2001 2002 2002 2002 2002 2003 52003 2003 2003 2004 2004

Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 :Q2 Q3 Q4 Q1 Q2

Note: This figure describes the frequency of non-GAAP earnings disclosures over regulatory time periods
depicted in Appendix A. Proposed Rules were circulated in Q4 2003. Finalized Rules were announced in
01 2003. Reg G became effective from March 28, 2003. Thus, the effect of Reg G on the disclosure

frequency

of non-GAAP earnings disclosures could be captured from 02 2003. The current study covers

792 randomly selected firms from the intersection of COMPUSTAT and CRSP. Heflin and Hsu (2005) and
Marques (2005) follow 1,403 l/B/E/S firms and 314 S&P 500 firms respectively. The current paper and
Marques (2005) manually collect non-GAAP earnings. Heflin and Hsu (2005) use the difference between
GAAP EPS and IBES Actual EPS as a proxy for non-GAAP earnings disclosures.

58

000800 ESE a 88800 2 m0t0m 08: convoomlo“ Som 8 8088030 @888 ~ :05 0.08 020: 38 80803030 288:0 ~ .0

H<HmDA=ZOO 80¢ 3:50:80 8308? 8080230. 30826 8 020 convoom “a 8:0 Etc 038%

.8 .80 80830 Home 0:... 088—005 80m noun—$0M 3 00900»: ~08 03 008000.88 35 08m 088 9 8080 8 53 80% 80830 80¢ 0.830. 0883 0:... .9
.80803030 0 wBEmoa 05 8 0880.20 08 moom .mm £802 00% 030 808008088 88830 b88208 .082. .030 808008583

$8880 b88208 980500.85 :80 80 0003 80880030 0 M0”. Smog 0:0 -08 05 88 80803030 0:80.85 0036 _ £08208 $880.0 fl 8088
$5 .8 0:00.“ 05 0880m .88 .3 £032 Ban 0008 .OMm 08 53> wwEE 88 880E $8800 $8208 £83286 0288 :m 8 82%? O coca—03.x .a

 

 

 

 

 

 

«flaw wa one; meonm G 030.9 0388 “m0“ 00:02

CB 8 Ammo Q Cmm v“ Ammflommv 008.83
$88.80 30-80: @688 85 80505m8m 0004 v

20.2 33. 88.0 ”we a .4 seat 205% 2208 23280

3me ~00: Amp: 903 838? 0:080:03
3088a w8088 028 85 8080:3085 0.004 m

03.3 $30 Nmmd «2. Am .m 030C moumcfim 0>cdtom0Q 00w 000: 2ng

CS mu Awom 3 Noon 3 wow 0050-08: 05 8
0 000208 808 w80m8 H :08 0008 022 88 .085 0004 N

$0.2 mnwfi God coo; voom 2 000— 80c 800 mam $888-80: ..

o<om<z\xm§<\mm>zao 88.8808 -

s 338-5” 58 a as 0:20 some as -

”0:080 w830:8 05 8008 85 000.0. («0 mm

883 8 832 080 05 8300200 3088
.8 80300838 05 80¢ 085 “0080—00 38088 coo; _

 

 

30H a O w0m 800 a O w0m 08 0008003 80:00 80:00—0m 038%
8088088& . 080E

 

80500.5 088% A mam—ad.

59

A 083.2 000 .0838: E03808 08 .8 0:820 80..— .Amoom £033: 0 coca—80¢ .8
828808298 05 28 Scam .00nd @8880 30.8: 880228 88 00:60 882800 0.0mm 08 ”0808 030 :0 0008 2 83202008 02% .8288 088
b08830 0:288 808 0080—08 0008 808008880 $88.80 b88208 8 00802020 38800 “385.808 .8 20808008 08 080008 033 25‘ ”082

 

 

 

 

520v 22.3 $8.: 329 0095228 232000
So: 81m ANV $050286 088.80 $20802
mama 02:0 8500.020 02
$30 . mmmd A: $000200 30.8 .8 800832

0030?; 80cm 0:0 wonmoom MonmoONAOu Bow

8 - E 0 00¢ 088 8 o 000 20 2

0080—8030:— .8 000,—. "x 82.88.03 03. am
Sodv Sodv wood $5.3 $00.3 , $8.3 2:8 .0w080000m 080.285
moo; om: . 0% A8 £088.86 mrnw88.00 m<<0-:0z
mofim $m.m 0mm; 00302020 02
830 nwwfi 22mm A: £00020. $08 8 800802

 

838$ 0023i m0:_0>-m 00am 0:0 wonmoom Monmoo~-_onmoom N _ “ :5onH Sea
30 - £0 3 - 3 £0 - E 0 00¢ 080 3 0 030 20 30 0 mom 05 3
00800800005 .8 608. «x 2.2.—090:0 00.. 5 mm

82.3.— 08; E088?“ 000830. mMEEam m<<080z .8 0802005 8 580:3; 2:. .N mam—(H

6O

A32. . .N can ooooA 82.8 d2??? .388 daoméocn magnum 530.83 888 .8. E... 8:83....

.. $0.008 838.58. 3235". $3983 :83. $3.183 82...: .oamYooov cougoaamfi. .onoménom v.8 dumménmm 528-025 882.800
.ancméhom “.5 .3265? .388 $3253 8058.358 038:9 .3.” 782 98 532.03. 8.586... 2,5...be .emwmémwm 28....8gnm .32
63.2 8.. «82.88 28.826 88.88 8.23.... 2:. 88.... 8.509 .. . .382 .8... 82-82 .828 $8.68. 8.32.88 Ba 8.52 .88-..
aha—8&3 082.8 8335320 883 m...“ .0 53m no 63.3 888338.... 2958 9... 5.3.32. 2.. 8.. 3:82.88 .986... a... 3:88.. 038 m3... .202

 

 

 

 

 

 

 

$8. $82 $9.. $:. $8 $5 $8. $8 $0.. $3 $8. $2. $8. $2. $8. .2. 8......355 92...... .
G. 53259....— =a..u>O 3.
$2.. $2. $3 $2 $2 $3 $3. $3.. $2 $3 $2 $3 $2 $2 $...o .853... 2.. 80.23.0518...
$55” $2.0 $2.. $...~ $.02 $.63 $.33 3&2 $3.3 fiend. $5.3 $5.2 $.22 $32 $6.? 3.8.6... o... 55.3. 3:83.“.
En «we v2 32 8 _ mom 2% 8m E mm. mm 2. R 2. 3 858.8... $5.58 n.<<O-:o.. .0 .oz
3... 2.. 2%.. .3... 8m 8.. 8.... $2 .8 . .m .52 8.. $2 82 mm $8.2»... .255 a.
$8.... $3. $3. $3 $02 $3 $2. $3.. $o.m $2 $...~ $3 $2 $3 $3 $282.... 2.. 8.28. .0582.
$...n $2.. $2. $3: .82 $.32 $32 $8.: $12 $8.: $.....~ $3.. $2.8 $8.2 $22 3.32... 2.. 55.3. 8:88..
2. 2.. .o. .e an on 3. E 2 .m 2 mm .2 m. m 358.8... 85.5.. $20-8: .8 .oz
2.... «R 8.. m9. on. 2.2 to O8 02. 82 .2. .52 8 8 . . was.3. 82.. .o .2
.qonvuz. 2...... o a... .2... .3
$. .2 $6: $3 $2... $2 $3 $8.... $3. $8.. $. .m $2 $8..” $2 $3 $3 $252.... 2.. .85... 88:8:
$.99. $52 $.m.2 {8.8" ..\o. .3 gm: $.09. $.32 $.92 $.32 .5 .3 $3.2 $0.3 £1 .3 $8.3 3.3.3.... o... 55.5: 3:252”.
8. N .2 m: 8. an S. .3 9..” NV 5 mm n. on em . . m238.8... 85.58 85.0-8: .0 .2
2m 8m 3.. 8c 8. SN 2.” Km. 8. 32 ON. 82 2.. ... _ t 3.8.3. 82A. .o .2
.83"... 2.5. o a... 2.. 5
$. ... $3 $2.... $2.. $2 $3. $2. $22 $2 $5. $8.. $3. $2 $8.. $2. 38.2%.
fiata=O\E..E 9.956 .8
$2. $8,: $2.... $3. $2 $. .0 $3.. $2. $2 $8... $3 $X $.... $2 $2. :9...» al. 2...” .c 2. 25....
ammoifimzmzou . ..
lm m Ilm h s n r k e k k. .5 d n m s
m m m m m m m .. m m ... m ..
m m m m u n m u m m m m. m .m m
d a n U m. 0 D m c e g t r d
m e H s C E x mn. .m m. g n
R n m m o A I
a d m m .n C
.m m T .n P .m
m m p .m a
s m m m.
m m .m
m w M
T.

8.52:...— nmeu< aunt—ham A32.0.5.7. .3 9.3285 .3 55:3; 2:. .m mam—<8

61

8:888 0:00 0.80: :8 :8::Sm .8 £800: 08 :8 388 88:08: :8 885 00m 88380 88:0: 808 8002880: :08808: :8

.3008 8:808 03:80:08: :0 0308080: .8 0:30:88? 88:08: 90:8 .0080 8.80830 :8 8:88 >8 :0 0:88:08 808:: .8 88088: .8880:
.088» :880 88.: 00.3880 8 :088: 880:8 0:208 .8830: 8:888 080 mfim 888:8: 0: :088: 08080800 :8 88808: 88:68 :8 0:88
805.88: .8980 53:00: .8 8088898 .8080 .8 080 08 808 8::0>0: :0: .8880: 88:0: 0: :088: 88: 80 8:208 mam 0:00 0.80.: 0% :8::Sm A3
.08 :88: a m: mmm n1:898: 80:0 88 08:08 _ .0:0 88:: :0 0:80 0 0: mam A:80680: 8808.» 80:0 8: m: 8 :0; .888 8.8:: Mmmm n1.20:0:
9880:8800 :08 88:08: 0:88:08 8:80:88: :8 0800: 888888 88208 mmm 3089 88:: :0 8889 0:8: 888 08088 mam 30 CV

, :90: 000000 05. :92: £0. 0:2... B 028.00... .20 :22 023.80: 8.0:: 20 ES: 8: 0.33.. 80:: 30

Avfl :89 ..0.c mam :88: 088080 0: :00: 88:0 .8 808:: 08 :80: 8:8 mam :0 008 880800 08

88:80 A340.8: 80:0 .382: 8 88:80 n$2.080: rod 8:8 mam :0 8 :00888 8: :0; 85:08 0.8888: 08 :8 88:0 0: 000.: 008 08 8 mam
n3.10.0.8: .0808: 08:: a 8 :88 08 mmm 3080: 082:8 :05? .8808: 08:: 80800888: 88:80 8:88:08 0:: 808 mam 8:00:00 38:82 A3
”3028098:

86:8 8 0.8 mam 8:090. @mfib :8 .mmm 30 .95 A8.010.080: .8 0:08:88: .:008::-88 :08 :8 8:08 88:: a 8 :03 8 8:08 :80 0.0000 8

08:00 8 80:80:00 08 mam =8 88 0.80 0088 0: 80:0 080: H .mmm 0:00 mam :8 mam m<<o .mmm 880< mEEb .mmm n2.06:0: 80:00 80:88.2: :088080
H8; A8888 a: 8080-806 30080088: 0_0>0_ Ed :8 mod 8 0:0: 808 80:08:: 30808880 08 80:08:88: 88088008 88 08:0: 1.... :8 1. ”082

 

 

 

 

 

 

 

 

08.0 8:0 3.00 30 0 we: :8:
30.0 0000 0:0: 5:0 3.0.0 08: o 00: 2.: A00 - 3 000208: :0
80.0 00:.0 08.0 N00,: 0 we: :8:
03.0 80 8.8 03.0 08.0 003 0 00: a: 3 - :0 .8808: 8
80.0 000.0 020 N00 0 00: :8:
000.0 80.0 80.0 000.0 0:0 3: 0 we: a: 30 - 00 .8028: 8
000.0 08.0 80.0 30 0 00: so:
000.0- 800. 20 000.0 0:00- 08.: 0 00: 2.: 8: 200 .80 :0
000.0 02.0 0: .0 80.: 0 00: :8:
000.0- 03.0. 008 000.0 000.0 002 0 we: 2.: 00 08 8 05023.0 8: 820 3
000.0 00: .0 08.0 N00 0 we: 08.:
.. 08.0- : 80.0. 80.0 0:0 080 a: o 0...: a: 0050.3: 800... 088: 30
80.0 08.0 08.0 80.: 0 mo: :8:
: 08.0- 080.0 088 08.0 03.0 008 0 00: 2.: .0580: 30-000 00
88:08V A808u $088 8802 802 Z 08:85 8808:: 0:08:81”: 888>

:8: I a: e 08.: .. 2.: 00
00080.: 880.80: 03:. 08:00 8: 0:00 .50 00.. .8: :36 .8: .053. 08:: .8: 880-002 :0 00:89:00 .0. :80»:

62

:Amrofio: Eu: 05 moms 8E 05 5:3 ANV can Fflofio ES 8930 msg#92585. swat acogmspum
mo tan a mm 95: 5380 Mo 250:8 “5:5:va 05 @6on 8: ow acne can? A: ”8.0.3 wows—o5 bowofio :3050: BE. ”Bo Z

 

 

53 23V 35 gm

$an MS $12 El £050 .2
$2 R $2 Nm BBB mam 8:32 .o
$w.~ mm $§ 2: so: 25 2: 325 .w
$2 8 $3 3 mowsfiEofimaaeoo 823 £85 s
$§ mm $3 5 $8-25» “82 .o
$§ 9: $3 $2 £8 8325532 .m
$3 2: $0: an 90%52 ~on 83% Eooamueo: chaeeoz .v
$3 a: $3: $8 SEEOE<E2§8§Q .m
$3; 02 $2: a? 3228 .N
$NS «on $3; 5 83% wagoémom g
3:0:on 2:50 Ecoswoum E500 25: E086: €<

a 3%

555585 wage—«m m<<0éez .5. tomb 8.8— 28.53.84. 2: .3 mono—:55 2E. .m HAM—«Q

63

.2385? 80: 05 moms EE 2: :2?» AS 28 quofio new mowumno wgogmor. rwdv 53533
mo tan a mm. as: £8.60 mo 250:5 “songs?“ 05 xmmooam 8c ow Etc :23 CV M38 89:2: @0330 :flofio: oak 68 Z

 

 

:3 , as} 38. ea
$3M 80 $3M E: 82:0 .2
$3 2 $3 mm 328 a? 8:32 .m
$3 mm $3 2: a»: 25 2: 32$ .w
$3 8 $2 cm: 8365:3538 was :85 .5
$3 mm $3 5 598-25, :32 .0
$3 2: $6. 8N was BEBEOEQGE .m
$2 2: $0: 8m sags: :on swat _2oam\mea_ mcEaeéoz .v
$3 8: $2: $3 EEEEEEOEQEEQ .m
$3: 08 $2: a: 3228 .N
$3: . 3m $2: :6 mowsfi $533.58: .:
Qua—wen 3:00 55:35 5300 was: Hangman}:
Mg g

55358.5 museum m<<Uécz 3.. 62.5 2:8— 2—25:22 2: .3 55:3; 2: 6 ”mu—a;

64

688.2 $558 5 05858
5055505005 050-5-0553550QBBDS 50 8825 $558 .50 38585,.” 50525 $558588 05 5 50550.55 5055:5005 8:38 5 505 (5 um
.5055an 30-50: $0 55050 05 58 .50 5o: 05— 05 $55553 58$.“an a 5 205 p: ”N
.$558 3050:
855002 8 5055585 505085 508cc .«0 55085 5.850 508: 0:0 :5 8.0.85 $558 5 $258 m<<0502 M0 :5568 820 a 5 085 h ”—
8m85 $558 5 $558 30502 («0 5055.80 .820 0: 5 505 .5 no
”85.3 $5323.“ 05 80—5 4<300m5 558E890
Roomy 0.00533 50 889 3cm $0 80885 $558 5 80305 50.55585 5055:5005 .50 5:050 05 50.5 5385 guinea >5 5 5<Dooma

.808 968 .50 0:0: 5

.$558 30 8>0 8822 3.05 .50 $005 05 5 5:8 585% :5 558.53 08— 05 50 25 05 5 8555qu “0: 2m $558 3050: (5 ”—
.$558 $0.50: w530=£ 38585.5 $558 m<<0 .53 :5 58$53 v8— .05 50 25 05 5 85555: 08 $558 $0.50: .5 ”N

595858 82 05 50 £5 05 5 85$in 08 $558 $050-50: ‘5 ”m

858 $535.8 05 858 mMOUmmEm .$558 m<<o 8>0 $558 $0.50: .50 58558 05 58 5385 585505 55 5 mMOUmmEm

2.ch
058$

 

mmod

 

 

a: 585
m m WE
m m 582
N N mg
33 RS. :82
a8. T70 62.870 85.9%
0 $0M «mom 0 mod 05
$83 as $3.8 $3 m
$3.20 as $2di 30 N
$8.0 E $8.5 02 .5
$3.3 mm. $2.8 w: o
$5;qu 62.820 85955
0 $0M “mom 0 mod 05
122085

Sodv
028$

 

 

 

.802

~33 5: $me
o 5 m5
o o 5622
o o 5
£3 32 :82
£8.30 62.820 885%
O mom “mom O mum 5m
$5.3 s $2.5 5 m
$.30 we $2.5; a N
$38 8 $8.” a v8 _
$830 9% $8.8 33 o
83qu 65.820 8:395
0 $0M “mom 0 mod 05

500525

503550.5— 50=5=0=08¢ 55 $553— miwéoz 50 58555 K mam—5H

65

$3553:an m<<0éoa “US$858 rod mam 30 323 98 mam n33.0.:o2 05 20:3 88

05 mm 220% .8258 338:8 05 26—3 33 mam 30 :23 5:3 3:8 N 8 a: 83850 $888 mam 383:8 Q88 05 Sup .8 BE mmm m<<0
.qu FEE 05 98:3 ammo 05 mm m Dm 2mm 200192.»: .Omtfizm meEmo A33.0.52 63:0st bficogom 2660a a 9 8:93 $5an 30 3055206
23:88"... gamma a @255 n$30:qu 985 05 20:3 83 05 mm mmZImgm .850 N 8 a: mam A$90.32 2568 8 Son 9 832 m<<0 BEE
A3~<0éoz FEE 05 30:3 83 05 a mmOAImE .332. 8.“ m x65aq< oomv buifioamfi 5888 mam 383:8 2: 98 .95 8.226 088 :3» “mm—
ofi .83 333 0.8 mam 30 ME 05 8023 3.0.3 05 08 mDmmeZOU .mmz .mmOA 9323.6 $5an A33.0.52 $ 205 223 88 05 a 02 .802

 

 

 

wood $8.2 8o. :a: $3.: 8333 92:282 S

2.3 $2.8 Geog $2.2 oz. :2 m Amam2mmzou_oé\m:m2mmzouumm>m §

Sod $3.2 $352 $2.3 832 m meDmmezouimmE 6

$3 $3.8 m 2.5: $2.8 E 1 Sam Ammzagmmzdgm €

Bod $3.2 So. 3: $2.2 833m oémmzlmgm 5

v86 $w§m 83M: $£~m 3S: $362038-me 5

83 $3.: $952 $_ _.2 832v ©7584:me E

mos—Syd X. 553on c\o >053on @553. A$3.0.coa 38285 HO 8553820
“mom. «x 85% 0 mom “mom cocoa 0 mom 05

9.35.6.3 $555 23 3552i? $4052 .w mini.

66

.bguoommo. £96. 36 98 mod .25 05 8 oucaoc_=w_m .8...ng 828.. .22.. can ...... .... .mcouaaov .8. m can u 2an .8 86: com .082

 

 

 

. 1.... .3... ......... 8...- ...... m8... 8.... ...... so... o 3. .2...
. ......... ma... 1.... N8..- 1.... 8o... ...._. 23... 8o... 0 may. 2.. 4.3085 .c
1.... .2... . ......... o8..- ...... 8o... 2...... 3.....- 0 may. .8.
......... N2... . ........ m8..- .. m8... .8... 3.... 8o... o 3. 8.. $88.2”. .m
......... E3- .. .3..- . ......... ES- .2... ME..- 23. o 358..
..:..... wood- ......... mood- fl ......... wc—d- ......... N~ 5o- ......... wm_.o.. O mom 0.5 HQ<OWZ .v
2.... 8o... .. ms... .1... ES- . ...... 8o... :9..- 0 m3. .8.
......... 8.... ........ So... ......... $2.- . .. 3o... .. as... 0 may. 8.. mammezoolmEE .M
8o... 2...... ........ £3- ...... $3. . ........ £2. 0 m3. .8. I
......... hhcd ... omod ......... NH Hd- ...... awed H ...... owed 0 mom 0.5 mmz mgr: .N
.. a3... 2.....- 3.... SN..- :3- ........ £3 . o 3.. .2... I
8o... .. .3... .1... .33- .. So... .1. o8... . 0 ms. 2.. $0.. $5.. ..
o n v m N . 50%... 85...;

«weir—am 30-32 come—uma .3 85.36230 .3 mucus—9.30 €2.64. 5.8.8on 9:. C258 ugh-«om .a Hand;

67

TABLE 10. Descriptive Statistics of Firm Characteristics by Non-GAAP Earnings

 

 

and Regulation G Regimes
Pre Reg G period (N=6,229)

(1) Non-GAAP EPS (2) No disclosures
Predictions Disclosures Firms Firms P-VALUES

Variables Statistics flvs. (2) (N=2,166) (N=4,063)
EQ Mean < 0.21 1 0.359 <.0001
Median < 0.201 0.304 <.0001

Stddev 0.028 0.021
EQRANK Mean < 47.481 50.404 0.0001
Median < 47.000 51.000 0.0001

Stddev 0.621 0.449
LOSS Mean > 0.365 0.223 <.0001
Median > 0.000 0.000 <.0001

Stddev 0.010 0.006
NES ‘ Mean > 0.516 0.397 <.0001
Median > 1.000 0.000 <.0001

Stddev 0.01 1 0.008
CONSENSUS Mean > 0.522 0.343 <.0001
Median > 1.000 0.000 <.0001

Stddev 0.011 0.007
LNMKT Mean > 6.479 5.544 <.0001
Median > 6.459 5.614 <.0001

Stddev 0.043 0.032
LEVERAGE Mean > 0.573 0.548 0.0006
Median > 0.580 0.534 0.0010

Stddev 0.006 0.004
BTM Mean > 285.220 261.230 0.0658
Median > 131.606 121.673 0.0181

Stddev 10.372 7.896
SI Mean < -0.012 -0.001 <.0001
Median < 0.000 0.000 <.0001

Stddev 0.001 0.000
HHI Mean < 0.210 0.239 <.0001
Median < 0.151 0.191 <.0001

Stddev 0.004 0.003
INTANGIBLE Mean > 0.094 0.056 <.0001
Median > 0.000 0.000 <.0001

Stddev 0.004 0.002
LITIGATIONIND Mean ? 0.236 0.178 <.0001
Median ? 0.000 0.000 <.0001

Stddev 0.009 0.01 1

 

68

TABLE 1o. (Continued)

Post Reg G period (N=4,477)

 

( 1) Non-GAAP EPS

(2) No disclosures

 

Predictions Disclosures Firms Fimis P—VALUES

Variables Statistics (1) vs. (2) (N=1,062) (N=3,415)
EQ Mean < 0.142 0.455 <.0001
Median < 0.121 0.368 <.0001

Stddev 0.037 0.023
EQRANK Mean < 44.391 51.241 <.0001
Median < 42.000 51.000 <.0001

Stddev 0.867 0.492
LOSS Mean > 0.288 0.233 0.0005
Median > 0.000 0.000 0.0003

Stddev 0.014 0.007
NES Mean > 0.409 0.347 0.0003
Median > 0.000 0.000 0.0002

Stddev 0.015 0.008
CONSENSUS Mean > 0.528 0.332 <.0001
Median > 1 .000 0.000 <.0001

Stddev 0.01 5 0.008
LNMKT Mean > 6.755 5.811 <.0001
Median > 6.608 5.852 <.0001

Stddev 0.058 0.035
LEVERAGE Mean > 0.591 0.550 <.0001
Median > 0.597 0.531 <.0001

Stddev 0.007 0.005
BTM Mean > 325 .490 307.100 0.2814
Median > 161.744 148.084 <.0001

Stddev 13.236 10.789
SI Mean < -0.006 -0.002 <.0001
Median < 0.000 0.000 <.0001

Stddev 0.001 0.001
HHI Mean < 0.229 0.256 0.0005
Median < 0.166 0.199 <.0001

Stddev 0.007 0.004
INTANGIBLE Mean > 0.151 0.1 15 <.0001
Median > 0.055 0.036 <.0001

Stddev 0.006 0.003
LITIGATIONIND Mean 0.205 0.194 0.4309
Median 0.000 0.000 0.4256

Stddev 0.012 0.007

 

69

Note: Variables are as follows.

 

EQ An empirical proxy for the firm’s communication motives to disclose non-
GAAP earnings due to low GAAP earnings relevance. Defined as

the t—statistics of the coefficient on seasonally differenced GAAP earnings
from the following firm-specific returns-earnings regression:

MKTADJRET = a0 + a,AGAAPEARN + e
(firm/quarter index omitted)
where

MK T ADJRE T is cumulative market adjusted returns (i.e., RET - VWRETD)
from two days afier the last quarterly earnings announcement to the day afier
the current quarter earnings announcement date, and

AGAAPEARN is seasonally differenced GAAP earnings before extraordinary
items and discontinued operations deflated by the market capitalization at the
beginning of the current quarter (i.e., [Data25. — Data25H))/(Data61.* Data17.*
Data14 ...)]). I require at least 8 quarters for the estimation of t-statistics. The
estimation does not include the current quarter EPS observation.

 

EQRANK EQRANK is a rank transformed variable from EQ. I rank EQ by the calendar
quarters in which preliminary earnings announcements were announced,
assigning from 0 (lowest EQ) to 99 (highest EQ). EQRANK is designed to
facilitate the interpretation of the coefficient and marginal effect of HO
variable in probit regression model across different regulatory regimes.

 

LOSS An indicator coded as 1 if the GAAP EPS excluding extraordinary items and
discontinued operations for the quarter is negative, 0 otherwise.

 

NES An indicator coded as 1 if the seasonally differenced GAAP EPS is declining
(i.e. negative GAAP earnings changes), 0 otherwise.

 

CONSENSUS An indicator variable, coded as 1 if the GAAP EPS excluding extraordinary
items and discontinued operations (dilution factor considered) is less than the
mean EPS estimate from the I/B/E/S summary file. For REIT (real estate
investment trust) firms, I use FFO (funds from operations) forecast for the
comparison, if available.

 

BTM Book-to-market ratio

 

HHI Hirschmann-Herfindahl’s index of market concentration (an indicator of
competition among firms).
2

n sales,
Defined for firm i as 2):] n
21—: sales,

where n denotes the number of firms in each industry broken down by 2 digit
SIC code and i denotes a firm in the industry. Higher HHI values in a given
industry can translate into less competition in the market.

 

IN T A NGIBLE The total intangible assets divided by the total assets (from annual data). 1
assign 0 if this item is missing in COMPUSTAT. The same non-missing value
was assigned for other quarters in the same year.

 

 

7O

 

 

 

 

LE VERA GE Total liabilities divided by total assets.

L] T [GA T IONIND An indicator coded 1 for the high litigation industries (i.e., SIC codes
2833~2836, 3570~3577, 7370~7374, 3600~3674) following Francis et a1.
(1994).

LNMKT Log of market value at the end of quarter.

S] The amount of special items deflated by the total assets. 0 was assigned for
missing values.

 

 

71

TABLE 11. Correlation Table of Firm Characteristics
by Non-GAAP Earnings and Regulation G Regimes

 

Variables Pre/Post 0 l 2 3 4
0.NG Pre RegG 1 -0.052 "‘ -0.048 ”* 0.152 ““ 0.113 ”"
Post RegG 1 -0.101 ‘” -0.100 ‘” 0.054 "‘" 0.055 ""
1.EQ Pre RegG -0.050 “‘ 1 0.942 ‘" 0.064 “* 0.001
Post RegG -0101 ‘“ 1 0.932 ‘" 0.056 "'” 0.032 "
2. EQRANK Pre Reg G 0048 ""‘ 0.998 *" 1 0.064 ”‘ 0.010
Post Reg G 0100 ‘” 0.999 "" 1 0.063 ‘" 0.042 "‘
3. LOSS Pre Reg G 0.152 "“ 0.063 ”" 0.064 "‘ 1 0.293 ‘"
Post RegG 0.054 ‘” 0.063 ”’" 0.063 ”" 1 0.300 ”‘
4. NES Pre Reg O 0.113 *“ 0.007 0.010 0.293 “‘ 1
Post Reg G 0.055 ”" 0.042 ‘" 0.042 "‘ 0.300 ‘“ l
5. CONSENSUS Pre RegG 0.173 "‘ -0.053 "" -0.054 "“" 0.135 *“ 0.232 "”
Post RegG 0.171 ”‘ -0.058 “* -0.058 ‘” 0.077 ”" 0.222 ‘"
6. LNMKT Pre RegG 0.205 ‘" -0.172 ”’ -0.l73 ““ -0.l92 “" 0001
Post RegG 0.185 ‘" -0.231 “" -0.232 ‘” -0.307 ‘" -0.134 "‘
7. LEVERAGE Pre RegG 0.041 "* -0.086 "* -0.084 ”“' -0.l3l ‘" -0.075 “"‘
Post Reg O 0.079 *" -0.023 -0.023 -0.077 ‘“ 0.028 "
8. BTM Pre Reg O 0.030 " -0.153 “"" —0.152 "‘ -0.359 "‘ -0.046 ”*
Post RegG 0.059 "“ -0.166 ”* -0.166 "* -0.464 “" -().113 ""
9. 81 Pre RegG -0.253 *" 0.017 0.018 -0.321 "‘ -0.224 ""'
Post RegG -0.205 ”* 0.038 " 0.038 *" -0.238 *“ -0.l96 "‘
10. 11111 Pre Reg G -().091 ‘” 0.004 0.001 0.000 -().002
Post Reg G -0.089 "* 0.002 0.002 0.021 0.004
11. INTANGIBLE Pre Reg O 0.069 *" -0.011 -0.016 -0.016 0131 ““
Post RegG 0.059 ““ -0.(')97 “" -0.097 ‘”" -0.042 ”" -0055 ”"
12. 1-1'1‘1GATION1ND Pre Reg O 0.069 “" 0.101 "’ 0.100 "“ 0.284 ‘" 0.057 "‘
Post Reg C 0.012 0.079 "* 0.079 ‘” 0.200 ‘” -0.()37 ”

72

TABLE 11. (Continued)

 

 

Variables Pre/Post 5 6
0.NG Pre Reg G 0.173 """‘ 0.213 "" 0.043 ‘" 0.023 *
Post Reg O 0.171 "" 0.195 "" 0.064 ”* 0.013
1.EQ PreReg G -0.065 ‘” -0.l73 "‘ -0.073 ““ -0.074 *"
Post Reg G -0.055 ‘" -0.225 "‘ -0.019 -0.060 “"
2. EQRANK Pre Reg G -0.054 ‘" -0.170 "‘ -0.081 ”‘ -0.069 ‘"
Post Reg G -0.058 "" -0.222 ”* -0.018 -0.064 ”"
3. LOSS Pre RegG 0.135 ”" -0.190 "‘ -0.109 ”“ -0.188 ‘”
Post Reg G 0.077 *" -0.313 ""' -0.063 *" -0.219 ""
4. NES Pre Reg G 0.232 ‘”" 0.000 -0.077 "* -0.032 "
Post Reg G 0.222 "* -0.135 "" 0.021 -0.038 "
5. CONSENSUS Pre Reg G 1 0.282 “" -0.049 *” 0.009
Post Reg G 1 0.224 ‘“ -0.007 -0.024
6. LNMKT Pre Reg G 0.274 "" 1 0.050 "" 0.355 ‘"
Post Reg G 0.217 ‘" 1 0.091 ““" 0.336 ""
7. LEVERAGE Pre Reg G -0.041 "* 0.066 "" 1 0.107 *"
Post Reg G -0.002 0.108 “" 1 0.104 *"
8. BTM Pre Reg G 0.104 ”“ 0.625 "" 0.129 *" 1
Post Reg C 0.068 "" 0.613 "* 0.139 "“ l
9. SI Pre Reg G -0.277 "“ -0.082 ‘" 0.060 ”" 0.061 ‘"
Post Reg G -0.235 "" -0.062 ”“ -0.043 *" 0.083 “‘
10.11111 Pre Reg G 0.041 ‘" -0.036 "“' -0.189 ‘”" -0.087 *"
Post Reg G -0.028 “ -0.065 ‘“ -0.212 "* -0.088 ”‘
ll. INTANGIBLE Pre Reg G 0.070 "“‘ 0.106 "” -0.003 0.016
Post Reg G 0.120 ”’" 0.186 "* -0.112 "* -0.013
12. LlTlGATlONIND Pre Reg G 0.091 *” -0.030 " -0.335 "* -0.248 ‘”
Post Reg G 0.029 "' -0.055 “‘ -0.290 "* -0.295 ‘"

73

TABLE 11. (Continued)

 

Variables Pre/Post 1 1 12
0.NG Pre Reg G -0.1 19 ‘” -0.070 ”‘ 0.127 “" 0.069 ”"
Post Reg G -0061 "‘ -0.052 "‘ 0.086 "" 0.012
1.EQ Pre Reg G -0.040 ” 0.018 -0.039 ”‘ 0.081 ”‘
Post Reg G 0.018 0.035 ” 0089 cu 0.067 n.
2. EQRANK Pre Reg G -0.031 ” 0.009 -0.047 ‘" 0.100 "W
Post Reg G 0.026 ‘ 0.011 -0.094 ““ 0.079 ”"
3. LOSS Pre Reg G 0216 ‘” -0.022 " 0.011 0.284 *"
Post Reg G -0. 192 '” -0.005 -0.021 0.200 “W
4. NES Pre Reg G -0.122 "‘ -0.008 -0.085 ‘“” 0.057 ‘"
Post Reg G 0136 "‘ 0.007 -0.029 ” -0.037 ”
5. CONSENSUS Pre Reg G -0.114 “‘ 0.032 ” 0.083 "‘ 0.091 ‘”"
Post Reg G -0.098 ”‘ -0.024 0.105 ‘“ 0.029 "
6. LNMKT Pre Reg G 0022 " -0.040 "" 0.107 “* -0.027 *‘
Post Reg O 0.019 -0.044 ”" 0.148 "‘ -0.043 "W
7. LEVl-ZRAGE Pre Reg O 0.033 “‘ -0.125 ”" -0.07l ”‘ -0.299 ""
Post Reg G 0.010 0133 ”" -0.115 ’" -0.257 ‘””
8. BTM Pre Reg G 0.039 "‘ -0.027 " -0.007 -0.143 ‘“
Post Reg G 0034 " -0.040 ‘“ -0.004 -0149 u:
9. 81 Pre Reg G 1 0.007 -0.055 "’ -0.101 ‘”
Post Reg G 1 -0.003 0,036 n -0045 an
10. HHI Pre Reg G 0050 ”’ 1 0.119 ‘" -0.164 ”‘
Post Reg G -0.034 " 1 0.148 ‘" -0.138 n.
1 1. INTANGIBLE Pre Reg G -0.082 *" 0.118 "‘ 1 0.046 ”‘
Post Reg G 0108 "" 0.282 ‘" 1 0.081 ‘”
12. LITIGA'I‘IONIND Pre Reg G -0.130 "" -0.145 "‘ 0.025 ‘ 1
Post Reg G -0.066 '" -0.104 ‘” 0.089 "" 1

Note: This table presents Pearson (upper) and Spearman (lower) correlations for the variables for two

regulatory time periods. *, ** and *** denote significance at <0.10, <0.05, and <0.01 levels, respectively,
for two-tailed tests. Pre-Reg G observations are obtained from preliminary earnings announcement press

releases from 2001:01-2003:03. Post Reg G observations are obtained from preliminary earnings
announcement press releases after 2003:03. See Appendix B for the definition of variables. NO is an
indicator variable of disclosures of non-GAAP earnings in preliminary earnings announcement press

releases. NG is coded as 1 if firms disclose non-GAAP earnings, 0 otherwise. The underlying construct of

NO is voluntary disclosure of management defined earnings. Thus, NO is not coded as 1 when, for

example, “pro-forma” earnings are required by GAAP (e.g., Reg S-X, accounting changes etc.). For other

variables, see the note for Table 10.

74

.3268qu _ 9 o Eco 832:3 >883 mo muwgo 8286 no.“ 8m mDmmeZOU 28 £94 .meq no Saute EEwEE 65 253 .96— :8:

05 8 3%: one Boa 653."? 49c 05 .«o owns? 828% 8.“ mm MZéOm mo Soho 353$: 05 Joint? .8 553588 2: mm .Amoom owotEooBv mozntg
mew—EBB 05 no.“ mos—g 56E wqawwmmn an 33:28 Ba $258 A<<Déo= $8885 .8 b59305 05 :o mDmmeZOU 28 mm Z .mqu ICE/Dam

.«o 06ka “coho fifiwafi RE. .383 9353 38636.85.“ no woman 98 muoumficmm .88“ 3:863» not. .boéooamon £32 moov one .moov .o_ .ov

“a coaumafi 88% .3... Ba .1. ._._ dogma :82:me gum—:83 RES: 05 no.“ mozfim .0 .2 98 2 3an 8o 86: o3. .moEmtg .3 85.6% 65 8m ”BoZ

 

$3.5 n «m onozmm

o\a\.o.o u «x 083mm

 

 

 

 

 

 

 

 

Rtvuz ommdnz
Nomo wnoo- moooo- mono owmo- NNoo- vomoo Nvmo on _ ._ omoo ... QZ_ZO_._.<O:.:
mono V0 _ .o onwoo- wooo owed- homo- mom: .o- ooo.o oté- _ov.o- I .II
34.0 ooo.o oooo.o nooo o _ 5N- ooo.o- ooooo mwoo oow. _- ooo.o- I 2km
mwoo ooo.o oNvoo ooo.o owe; _ mw _ .o w _ moo ooo.o 25.3 9 _ .o + EMEZA
Foo wood- wmooo ooo.o om_.v Sumo 59 _.o ooo.o omoé _mm.o + m0<mm>mq
t. ooo.o omoo- oom; .o ooo.o oom.m goo onwmo ooo.o onoo mno; + m4m_OZ<hZ_
ono mimo mmmoo- ooo.o omod- mo _ .N- m _ mo. _- _oo.o oovd- wood- I _m
t Nmoo + .m> I ”N: 02o vowoo ooo.o ohvo homo Nmooo ooo.o omoé :. _ .o + mDmmeZOU
mo_.o + .m> I ”N: mwoo- flmoo mflo oOm._ poo :moo ooo.o oowé hm _ .o + mmZ
. onoo + .m> I "N: om_.o- _Noo.o ooo.o ow _ .m womo one—o ooo.o omoo «mvo + mmOA
: oNoo + .m> I LI Nooo- ooooo- ooo.o coma- mooo- moooo- homo ooN. _- _oo.o- I ¥Z<m0m
ooo.o omoo _ - mam. _- ooo.o om _ .o _ - one. _ - E082:—
829»... o>oootm 82 O wax 2580 x03." 629$ .Em.N E ooo.o €36 029i ..Sm-N A: ooou 5:83.: 58 55.3 mozmtm>
an; 326 .2, 3:85 o as. I 3:80 8562..
88H 8.850%: ADV octan— O mom “mom EV coca 0 wow 95 $3
ovum—co 59: “285385

«gages + mbmzmmzoufi + mmzfi + $6an + M2§om£ + $5 I E 79: SEBEm H382

mocha Egon-3m cabana—0-5..E no .523

633.8% $5:an mini—2. 52 n 293...; Eco-835 NE a. 2* .8.“ 5:8..qu «EPS .NH HAM—AC.

75

TABLE 13. Random Effect Probit Regression for H1 & H2:
Alternative Specification with the Post Time Dummy
(Dependent variable = NG, non-GAAP earnings disclosure)

Probability (NG=I Ix)
= G(flo + fl/EQRANK + flzLOSS + ,63NES + fl4C0NSENSUS + fl'CONTROLS + fl5POST
+fl6EQRANK*P0ST + 37L OSS *POST + 3811/55 *POST + flgCONSENSUSWOST

 

 

+ fi-CONTROLS’WOSD

(firm/quarter index omitted)
Varaibles Coefficient Standard Errors Z Stat. P-Values
INTERCEPT -2.507 0.237 -10.590 0.000
EQRANK -0.002 0.001 -1 .400 0.162
LOSS 0.587 0.070 8.410 0.000 ***
NES 0.100 0.053 1.890 0.059 *
CONSENSUS 0.225 0.054 4.150 0.000 ***
SI -3.352 0.857 -3.910 0.000 ***
INTANGIBLE 0.405 0.195 2.070 0.038 **
LEVERAGE 0.176 0.173 1.010 0.311
LNMKT 0.246 0.028 8.760 0.000 ***
BTM 0.000 0.000 -3.360 0.001 ***
HHI —0.381 0.251 -1.520 0.129
LITIGATIONRISK 0.263 0.162 1 .620 0.105
POST -0.395 0.206 -1.920 0.055 *
EQRANK*POST -0.004 0.001 -2.790 0.005 ***
LOSS*POST -0.213 0.103 -2.070 0.039 **
NES*POST -0.081 0.088 -0.930 0.353
CONSENSUS*POST 0.144 0.083 1.740 0.083 *
SI*POST 0.424 1.321 0.320 0.748
INTANGIBLE*POST -0.967 0.234 -4.130 0.000 ***
LEVERAGE* POST 0.074 0.157 0.470 0.636
LNMKT*POST 0.020 0.024 0.840 0.403
BTM*POST 0.000 0.000 -1.510 0.132
HHI*POST 0.172 0.197 0.870 0.382
LITIGATIONRISK*POST -0.215 0.109 -1.970 0.049 **

 

Number of Observations: 10,615
Psuedo R Square: 9.3 %

Note: The results are based on random effect probit model (Gaussian assumption) and firm clustered
standard errors. Psuedo R2 is calculated by sealing the log-likelihood value of equation (1) with the log
likelihood of the constant-only model. For the definition of variables, see notes for Table 10 and 11. G
stands for the normal cumulative distribution function. *, ** and *** denote significance at <0.10, <0.05,
and <0.01 levels, respectively, for two-tailed tests.

76

TABLE 14. OLS Regression for H3

SADJRE T = 70 + 7,UE _ GAAP+ yzUE _ NONGAAP + ysPOST
+ 74UE _ GAAP - POST + ysUE _ NONGAAP- POST
+ yéLOSS + 77BTM + ysLOSS -UE _ GAAP+ 798TM -UE __GAAP + a
(Firm-quarter index omitted)

 

Expected Robust T

 

 

Variables . Coef. Estimates P>|t|
Sign Stat.
Intercept 70 0.0067 3.310 0.001
UE_GAAP + 7, 0.0187 1.670 0.096
UE_NONGAAP + 72 0.0064 0.990 0.322
POST 73 0.0023 0.930 0.351
UE__GAAP*POST ? 74 0.0185 1.810 0.070
UE_NONGAAP*POST ? 75 0.1225 2.200 0.028
LOSS - 76 -0.0105 -3.450 0.001
BTM + 77 0.0000 0.040 0.965
UE_GAAP*LOSS - 73 -0.0217 -1.940 0.053
UE_GAAP*BTM + 79 0.0000 1 .000 - 0.319
No. of Obs. 2,317
Adjusted R2 1.70%
[cits gm P-values (Two-sided)
y; + 74 = 0 Different 0.0054
72 + 75 = 0 Different 0.0198
(75 — 72) = ("Y4 — 71) Different 0.0575

Note: Sample consists of 2,317 press releases where seasonally differenced non-GAAP EPS surprises are
defined. In order to avoid a spurious inference from influential observations, I exclude 141 observations, as
suggested by Belsley, Kuh, Welsch (1980) (e. g., abs(RSTUDENT) greater than 2 or abs(DFITTS) greater
than 2*sqrt(2,458/9) from original 2,458 observations). If I include those influential observations, 1 get a
stronger result for the UE_NONGAAP*POST interaction term. Alternatively, if I use a rank-transformed
regression to avoid a spurious inference from outliers, I also get a stronger result for the interaction term.
All t-statistics are heteroskedasticity-robust (White 1980) and the mean VIF (variance inflation factor)
score is 1.12. SADJRE T is cumulative abnormal returns, defined as the sum of size adjusted daily returns
over the three-day window (-1 , 0, +1), where 0 is the date of the preliminary earnings announcement press
release. UE_GAAP (UE_NONGAAP) is seasonally differenced quarterly GAAP (non-GAAP) earnings
deflated by the market capitalization at the end of prior fiscal quarter. Stock-split factor has been
considered for both surprises. POST is an indicator variable coded as 1 if the earnings announcement date
is after March 2003, 0 otherwise. Size factor for returns is implicitly controlled by size-adjusted returns.
LOSS is an indicator variable coded as 1 if GAAP earnings is negative, otherwise 0. 3 TM is the book to
market ratio capturing growth factor.

77

TABLE 15. OLS Regression for H3
with the Sub-Sample of Continuous Disclosing Firms

SADJRE T = yo + y,UE _ GAAP + yzUE _ NONGAAP + 73POST
+ 74UE _ GAAP - POST + 75UE _ NONGAAP. POST
+ 76LOSS + 7,BTM + ysLOSS -UE _ GAAP + 798TM ~UE _ GAAP + s
(firm-quarter index omitted)

 

 

 

 

Variables Expected Coef. Estimates RObUSt T P>|t|
Slgn Stat.
Intercept . 70 0.0072 1.840 0.065
UE_GAAP ” ‘ ' ' '+ i V y. " 3 0.0217 . 1.440 0.150
UE_NONGAAP + ' 72 :7: 0.0517 ~ 1.670 0.096
POST . 73 0.0046 1.190 0.233
UE_GAAP*POS’T ? ‘ y. ‘} 0.0168 ' 1.130 " 0.257
UE_NONGAAP*POST ? 75 0.0904 1.310 0.189
LOSS - 76 -0.0137 -2.750 0.006
BTM + 77 0.0000 -1.420 0.156
UE_GAAP*LOSS - 78 -0.0228 -1.490 0.137
UE_GAAP*BT M + 79 0.0002 1.240 0.214
No. of Obs. 1,866
Adjusted R2 2.00%
m Rgilts P-values (Two-sided)
y. + Y4 = 0 Different 0.0402
72 + 75 = 0 Different 0.0204
(75 — 72) = (Y4 — 71) Not Different 0.6423

Note: This table presents the results of regression for a sub-sample of 225 firms that disclose at least one
non-GAAP earnings disclosure both in the pre- and the post Reg G periods. Sample consists of 1,866 press
releases where seasonally differenced non-GAAP EPS surprises are defined. All t-statistics are
heteroskedasticity-robust (White 1980). SADJRE T is cumulative abnormal returns, defined as the sum of
size adjusted daily returns over the three-day window (-1, 0, +1), where 0 is the date of the preliminary
earnings announcement press release. UE_GAAP (UE_NONGAAP) is seasonally differenced quarterly
GAAP (non-GAAP) earnings deflated by the market capitalization at the end of prior fiscal quarter. Stock-
split factor has been considered for both surprises. POST is an indicator variable coded as 1 if the earnings
announcement date is after March 2003, 0 otherwise. Size factor for retums is implicitly controlled by size-
adjusted returns. LOSS is an indicator variable coded as 1 if GAAP earnings is negative, otherwise 0. BT M
is the book to market ratio capturing growth factor.

78

TABLE 16. OLS Regression with the Sub-Sample of Firms
that Discontinued Disclosing Non-GAAP Earnings in the Post-Reg G Period

SADJRET = 7, + y,UE _ GAAP + yzUE _ NONGAAP + a

(firm-quarter index omitted)

 

Expected Robust T

 

Variables . Coef. Estimates P>|t|
Slgn Stat.
Intercept 70 0.0051 0.004 0.163
UE_GAAP + y, 7’ 0.0014 0.003 0.6701
UE_NONGAAP ~ + 72 -0.0168 V 0.019.j } 0.401
No. of Obs. 590
Adjusted R2 0.32%

 

Note: This table presents the results of regression for a sub-sample of 154 firms that discontinued
disclosing non-GAAP earnings in the post-Reg G period. Thus, sample consists of 590 pre-Reg G period
press releases where seasonally differenced non-GAAP EPS surprises are defined. All t-statistics are
heteroskedasticity-robust (White 1980). SA DJRE T is cumulative abnormal returns, defined as the sum of
size adjusted daily returns over the three-day window (-1, 0, +1), where 0 is the date of the preliminary
earnings announcement press release. UE_GAAP (UE_NONGAAP) is seasonally differenced quarterly
GAAP (non-GAAP) earnings deflated by the market capitalization at the end of prior fiscal quarter. Stock-
split factor has been considered for both surprises.

79

BIBLIOGRAPHY

Abarbanell, J., and R. Lehavy. 2002. Differences in Commercial Database Reported
Earnings: Implications for Empirical Research. Working Paper, the University of

 

Michigan.

, and ------------ . 2005. Letting the ‘Tail Wag the Dog’: The Debate over
GAAP versus Street Earnings Revisited. Working Paper, the University of
Michigan.

Amir, E., and J. Livnat. 2005. The Economic Consequences of (not) Issuing Preliminary
Earnings Announcement. Working Paper, London Business School.

Ai, C., and E. Norton. 2003. Interaction terms in logit and probit models. Economics
Letters 80 (1)2123—129.

Babington, D. 2003. New earnings rules bring new investor concerns. Reuters News.
(April 25, 2003)

Bamber, L., and Y. Cheon. 1998. Discretionary management earnings forecast
disclosures: antecedents and outcomes associated with forecast venue and forecast
specificity. Journal of Accounting Research 36: 167-190.

Barth, M., W. Beaver, and W. Landsman. 1998. Relative valuation roles of equity book
value and net income as a function of financial health. Journal of Accounting and
Economics

25(1): 1-34

Belsley, D. A., Kuh, E. and Welsch, R. E. 1980. Regression Diagnostics: Identifying
Influential Data and Sources of Collinearity. New York, NY: John Wiley and Sons,
Inc.

Bhattacharya, N., E. Black, T. Christensen, and C. Larson. 2003. Assessing the relative
informativeness and permanence of pro-forma earnings and GAAP operating
earnings. Journal of Accounting and Economics 36: 285-319.

 

, ,and R. Mergenthaler. 2005. Who trades on
pro- -forma earnings information? Working Paper, the University of Utah.

Biddle, G., G. Seow, and A. Siegel. 1995. Relative Versus Incremental Information
Content. Contemporary Accounting Research 12: 1-23.

Blitzer, D., and R. Friedman. 2002. Measures of core earnings. Standard & Poor’s. (May
14,2002)

80

Bowen, R., A. Davis, and D. Matsumoto. 2005. Emphasis on pro-forma versus gaap
earnings in quarterly press releases: determinants, sec intervention, and market
reactions. The Accounting Review 80(4): 1011-1038.

Bradshaw, M. 2003. A discussion of ‘assessing the relative informativeness and
permanence of pro-forma earnings and gaap operating earnings.’ Journal of
Accounting and Economics 36: 321-335.

------------ , and R. Sloan. 2002. GAAP versus the Street: An empirical assessment of two
alternative definitions of earnings. Journal of Accounting Research 40: 41-65.

Brown, L., and K. Sivakumar. 2003. Comparing the value relevance of two operating
income measures. Review of Accounting Studies 8: 561-572

--------- , and M. Caylor. 2005. A temporal analysis of thresholds: propensities and
valuation consequences. The Accounting Review 80 (2): 423-440.

Bruce, B. and M. Bradshaw. 2003. Analysts, Lies and Statistics: Cutting through the

Hype in Corporate Earnings Announcements. Institutional Investor Books
(Euromoney Institutional Investor PLC) NYC, NY.

Bymes, N., and D. Henry, 2001. Confused about earnings? Business Week. (November
26, 2001)

Choi, Y.S., S. Lin, and S. Young. 2005. Voluntary disclosure of Adjusted Earnings per
Share Numbers. Working Paper, Lancaster University (UK).

Collins, D., E. Maydew., and I. Weiss. 1997. Changes in the Value-Relevance of
Earnings and Book Values over the Past Forty Years. Journal of Accounting and
Economics 24: 39-67

Countryman, A. 2003. Some corporations continue to record routine expenses as special
charges. Chicago Tribune. (October 26, 2003)

Dechow, P. and C. Schrand. 2004. Earnings Quality. Research Foundation of CFA
Institute. Charlottesville, VA.

DeGeorge, F ., J. Patel, and R. Zeckhauser. 1999. Earnings management to exceed
thresholds. The Journal of Business 72: 1-33.

Dietrich, J ., S. Kachelmeier, D. Kleinmuntz, and T. Linsmeier. 2001. Market efficiency,

bounded rationality, and supplemental business reporting disclosures. Journal of
Accounting Research 39: 243-268.

Doyle, J ., J. Lundholm, and M. Soliman. 2003. The predictive value of expenses
excluded from ‘pro-forma’ earnings. Review of Accounting Studies 8:145-174.

81

------- , J ., M. McNichols, and M. Soliman. 2004. Do managers use ‘street’ exclusions to
exceed analyst forecasts? Working Paper, Stanford University.

Durtschi, C., and P. Easton. 2005. Earnings management? Alternative explanations for
observed discontinuities in the frequency distribution of earnings, earnings changes,
and analysts’ forecast errors. Journal of Accounting Research 43: 557-592

Elliott, W. 2006. Are Investors Influenced by Pro Forma Emphasis and Reconciliations in
Earnings Announcements? The Accounting Review 81 (1): 113-133

Frankel, R. and S. Roychowdhury. 2004. Testing explanations for the difference between
I/B/E/S and GAAP Earnings. Working Paper, MIT.

--------- , R., S. McVay, M. Soliman. 2005. Street Earnings and Board Independence.
Working Paper, New York University.

Francis, J ., D. Philbrick., and K. Schipper. 1994. Shareholder Litigation and Corporate
Disclosures. Journal of Accounting Research 32 (2): 137-164.

--------- , and K. Schipper. 1999. Have Financial Statements Lost Their Relevance.
Journal of Accounting Research 37: 319-352

Frederickson, J. and J. Miller. 2004. Pro-forma earnings disclosures: Do analysts and
nonprofessional investors react differently? The Accounting Review 79: 667-686.

Goff, J. 2001. Finance executives pro pro-forma, says CFO.com poll. CFO Magazine.
(November 16, 2001)

Graham J ., C. Harvey, and S. Raj gopal. 2005. The economic implications of corporate
financial reporting. Journal of Accounting and Economics 40: 3-73.

Gu, Z., and T. Chen. 2004. Analysts’ treatment of nonrecurring items in street earnings.
Journal of Accounting and Economics 39: 129-170

Heflin, F., and C. Hsu. 2005. The impact of the Sarbanes-Oxley act of 2002 on the use of
non-GAAP earnings measures. Working Paper, Purdue University.

Henry, D. 2002. The numbers game. Business Week. (May 14, 2002)
-------- , D. 2003. Ouch! real numbers. Business Week. (March 24, 2003)

Hirshleifer, D., and S. Teoh. 2003. Limited attention, financial reporting and disclosure.
Journal of Accounting and Economics 36: 337-386.

Hutton, A., S. Miller, and D. Skinner. 2003. The Role of supplementary statements with
management earnings forecasts. Journal of Accounting Research 41 :867-890.

Johnson, W., and W. Schwartz. 2005. Are investors misled by pro-forma earnings?
Contemporary Accounting Research 22(4): 915 - 976.

82

Kim, I. 2004. The role of directors’ and officers’ insurance in financial reporting and
class action securities litigation. Working Paper, Duke University.

Lev, B. and P. Zarowin. 1999. The Boundaries of Financial Reporting and How to
Extend Them. Journal of Accounting Research: 353-3 85.

Lougee, B., and C. Marquardt. 2004. Earnings quality and strategic disclosure: An
empirical examination of pro-forma earnings. The Accounting Review 79: 769-795.

Maines, L., and L. McDaniel. 2000. Effects of comprehensive-income characteristics on
nonprofessional investors' judgments: the role of financial-statement presentation
format. The Accounting Review 75: 1 79-207.

Marques, A. 2005. SEC intervention and the frequency and usefulness of non-GAAP
financial measures. Working Paper, The University of Texas at Austin

Matsumoto, D. 2002. Management’s incentives to avoid negative earnings surprises. The
Accounting Review 77: 483-514.

McVay, S. 2006. The use of special items to inflate core earnings. Forthcoming. The
Accounting Review.

National Investor Relations Institute (NIRI). 2001. Earnings press release guidance.
Vienna, VA: NIRI. (April 26, 2001)

 

. 2002. NIRI Guidelines to Improve Earnings
Releases. Vienna, VA: NIRI. (October 8, 2002)

 

. 2003. Information Dissemination under
Regulation G. Vienna, VA: NIRI. (April 15, 2003)

Norton, E., H. Wang, and C. A1. 2004. Computing interaction effects and standard errors
in logit and probit models. The STA TA Journal 4 (2)2103-116.

Payne, J ., and W. Thomas. 2003. The Implications of Using Stock-Split Adjusted I/B/E/S
Data in Empirical Research. The Accounting Review. 78(4): 1049-1067.

Schipper, K. and L. Vincent. 2003. Earnings Quality. Accounting Horizons. Vol. 17
Supplement: 97-110.

Schrand, C., and B. Walther. 2000. Strategic benchmarks in earnings announcements:

The selective disclosure of prior-period earnings components. The Accounting
Review 75: 151-177.

Securities and Exchange Commission (SEC). 2001. Cautionary advice regarding the use

of “pro-forma” financial information in earnings releases. Release No. 33-8039.
(December 4, 2001)

83

 

. 2002. Proposed Rule: Conditions for the
use of non-GAAP financial measures. Release No. 33-8145. (November 5, 2002)

 

. 2003a. Conditions for Using non-GAAP
Financial Measures. Release No. 33-8176. (March 28, 2003)

 

. 2003b. Testimony Concerning
Implementation of the Sarbanes-Oxley Act of 2002. available at www.sec. gov
(September 9, 2003)

 

Stuart, A. N. 2004. A matter of emphasis: Regulation G was supposed to end the abuses
of pro-forma reporting. Has it succeeded? CFO Magazine. (July 7, 2004)

Taub, S. 2001. Finance executives pro pro-forma, says CFO.com poll. CFO Magazine.
(November 8, 2001)

------ , S. 2003. A hard line on financial reporting didn’t deter a small group from their
pro-forma ways. CFO Magazine. (August 13, 2003)

Teach, E., and T. Reason. 2002. Lies, damn lies, and pro-forma. CFO Magazine. (March
1,2002)

Thompson, L. 2003. IR: Poised for Change and New Opportunities. available at
http://wwwniriorg (November 4, 2003)

Thurm, S., and J. Weil. 2001. Tech Companies Charge Now, May Profit Later. The Wall
Street Journal. (April 27, 2001)

Vincent, L. 1999. The information content of funds from operations (FF 0) for real estate
investment trusts (REITs). Journal of Accounting and Economics 26: 69—104.

Wallace, W. 2002. Pro-forma before and after the sec's warning: a quantification of
reporting variances from GAAP. F Iorham Park, NJ: Financial Executives Research
Foundafion.

Weil, J. 2001. Ignore the items behind the curtain: EPS means different things to different
earnings. The Wall Street Journal. (16 October 2001, C2)

White, H. 1980. A Heteroskedasticity-Consistent Covariance Matrix and a Direct Test for
Heteroskedasticity. Econometrica 48: 817-838

Wooldridge, J. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge,
MA: The MIT Press.

Zheng, L. and H. Zhang. 2005. Determinants and Valuation Impact of Reconciling Pro
Forrna Earnings to GAAP Earnings. Working Paper, The University of Hong Kong.

84

 

1111111111111111'