. .3 , 5:1...
. 1. l c!
cathf‘ "r
is“. ‘
.. "a

‘ 5%
.J.

. .c.

.I‘! 1 35!}
Inu'vaik

.. mil.

. . 3.33.1511
Ib‘lyfii...‘ ..
{.823}

 

 

 

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

THE CHANGES OF ACCOUNTING CONSERVATISM
AROUND EARNINGS RESTATEMENTS: AN ANALYSIS OF
THE IMPACT OF CEO COMPENSATION

presented by

Xin Xu

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Accounting

 

 

WW 5}. GM

Major Professor’s, Signature
'7/34)oq

Date

 

MSU is an Affirmative Action/Equal Opportunity Employer

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K.IProlecc&Pres/CIRC/DaleDueiindd

THE CHANGES OF ACCOUNTING CONSERVATISM AROUND
EARNINGS RESTATEMENTS: AN ANALYSIS OF THE IMPACT OF
CEO COMPENSATION

By

Xin Xu

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTORAL OF PHILOSOPHY
Accounting

2009

TH
EAR

consent
of 253m
deer-case
restatcm
tssociatr
“hilt th
Option 3‘
that alt-c
relations}
533111st
USCII as 3
OPP’Onun
demons”

finns‘ fin

ABSTRACT

THE CHANGES OF ACCOUNTING CONSERVATISM AROUND
EARNINGS RESTATEMENTS: AN ANALYSIS OF THE IMPACT OF
CEO COMPENSATION

By
Xin Xu

Using a sample of earnings restatements, I investigate how accounting
conservatism changes after restatement announcements. Using an accrual-based measure
of asymmetric timeliness of earnings, I observe an initial increase and a subsequent
decrease in the accounting conservatism of restating firms during the two years after
restatement announcements. I also find that the initial increase in conservatism is
associated with a high level of CEO option grants awarded before the announcements,
while the subsequent decrease of conservatism is associated with a reduction of CEO
option awards during the periods following the announcements. Further analysis reveals
that after controlling for the restatement magnitude and the accounting fraud, the
relationship between the initial increase in conservatism and the change in CEO option
grants becomes insignificant. These results suggest that accounting conservatism may be
used as a control mechanism to minimize the agency cost that results from management’s
opportunistic behavior induced by their compensation contracts. The results also
demonstrate the significant role of restatement characteristics in explaining restating

firms’ financial reporting behavior.

im 3

Kath

ilOl (

also

Marij-

ACKNOWLEDGEMENTS

I would like to thank the members of my dissertation committee for their
invaluable guidance and support: Marilyn Johnson (Chair), Ted Fee, Chris Hogan, and
Kathy Petroni. Without their support, I could not have finished this long journey. They
not only contributed their expertise to this project, but their guidance and expertise will
also be critical for my future growth as a researcher. I would especially like to thank
Marilyn Johnson for her guidance, insight, and consideration throughout the dissertation
process. One thing I learned through working with her is that you must believe in what
you are doing and work with great dedication to make things happen.

I am also indebted to workshop participants at Michigan State University, the
other faculty members in the Department of Accounting and Information System, and my

fellow doctoral students who have provided their comments and support.

-iii-

CHIPT

cum

{Titir

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. vi
LIST OF FIGURES ................................................................................. vii
CHAPTER 1: INTRODUCTION .................................................................. 1
CHAPTER 2: RELATED LITERATURE ......................................................... 8
2.1 Conservatism Research ..................................................................... 8
2.2 Restatement Research ..................................................................... 10
2.2.1 Causes of Restatement .............................................................. 10

2.2.2 Consequences of Restatements ..................................................... 12

2.2.3 Restatement and Financial Reporting Quality ................................. 13
CHAPTER 3: DEVELOPMENT OF HYPOTHESES ......................................... 15
3.1 Executive Compensation and Agency Problem ......................................... 15

3.2 Agency Problems and Accounting Conservatism ...................................... 18
CHAPTER 4: RESEARCH DESIGN ............................................................ 22
4.1 The Accrual-Based Measure of Conservatism .......................................... 22

4.2 The Market-Based Measure of Conservatism .......................................... 27
CHAPTER 5: SAMPLE AND DATA ............................................................ 29
5.1 Sample of Restating Firms ................................................................. 29
CHAPTER 6: EMPIRICAL RESULTS ......................................................... 32
6.1 Changes in Conservatism around Restatement Announcements ..................... 32
6.1.1 The Accrual-Based Measure ........................................................ 32

6.1.2 The Market-Based Measure ........................................................ 34

6.2 Changes in CEO Compensation ........................................................... 36
6.2.1 Changes in CEO Compensation ................................................... 36

6.2.2 The Determinants of the Changes in CEO Stock Option Grants .............. 38

6.3 Changes in Conservatism and CEO Compensation .................................... 40

6.3.1 Changes in CEO Compensation and the Level of Conservatism. . . . . . . . 40
6.3.2 Changes in CEO Compensation and the Changes of Conservatism. . . . .......43
6.3.3 The Impact of Restatement Characteristics on the

Change in Conservatism .................................................................. 44
CHAPTER 7: SENSITIVITY ANALYSES ..................................................... 50
7.1 Change in Conservatism and the “Big Bath” Theory ................................. 50

7.2 The Relationship between Change in Conservatism and Restating Firms’
Performance ..................................................................................... 53
7.3 CEO Compensation Incentives ........................................................... 54

-iv-

7.4 Analysis of Discretionary Accruals around Restatement Announcements. . . . . 55

CHAPTER 8: CONCLUSION .................................................................... 5 8
APPENDICES ..................................................................................... 113

Appendix A: Definition of Variables ...................................................... 1 13
BIBLIOGRAPHY ................................................................................. 117

Table I:

Table I:
Announ.

Table 3:
Amount

Table 4:

Table 5.:

for Rests
Tails 6: I

Table ”1 1

Option Cl

Table 8'. I
CEOS Op

Table 9; I
OI] IIIC Let

Table III;
on the Ch.

Table I l:
Announce

Table II;
Perfomur

Tible I3,
OPIIOII Jn‘

Table I4

 

LIST OF TABLES

Table 1: Sample Reconciliation and Descriptive Statistics of Restatement Firms ......... 64

Table 2: Testing of Conservatism for Restating Firms around Restatement

Announcements Using Accrual Model ........................................................... 70
Table 3: Testing of Conservatism for Restating Firms around Restatement

Announcements Using Basu Model ............................................................... 76
Table 4: Test of the Change in Conservatism within Restating Firms ....................... 82

Table 5: Distributional Statistics of Annual CEO Compensation and Its Components
for Restating Firms ................................................................................. 84

Table 6: Regression Analyses of Changes in CEO Option-Based Compensation ......... 87

Table 7: The Relationship between Level of Conservatism and CEOs' Bonus and
Option Changes ...................................................................................... 89

Table 8: The Relationship between Change in Conservatism and Change in
CEOs' Option Grants ................................................................................ 94

Table 9: The Impacts of the Restatement Magnitudes and the Accounting Irregularity
on the Level of Conservatism at Time t .......................................................... 96

Table 10: The Impacts of the Restatement Magnitudes and the Accounting Irregularity
on the Change in Conservatism .................................................................. 100

Table 11: Testing of Conservatism for Restating Firms around Restatement
Announcements Using Accrual Model after Deleting New-CEO F irm—Years ............ 102

Table 12: The Relationship between the Accrual Measure of Conservatism and Firm's
Performance ........................................................................................ 108

Table 13: The Relationship between Level of Conservatism and Changes of CEO
Option and Equity Sensitivities .................................................................. l 10

Table 14: Discretionary Accruals for Restating Firms around Restatement
Announcements .................................................................................... 1 12

-vi-

FIGL'R

LIST OF FIGURES

FIGURE 1: Event Year Illustration ............................................................... 63

-vii-

COHII
I I"? I .
FIOI‘II;

b3 3
L 5.

how 5.

EDIT OLII

executi

demanc‘

misaliw:

A
\—

IIICII Im

 

 

CHAPTER 1: INTRODUCTION

Conservative financial reporting has been hypothesized to facilitate efficient
contracting between managers and shareholders in the presence of agency problems (Ball
2001, Watts 2003). Jensen (2005) suggests that overvalued equity can lead to agency
problems and equity-based compensation of executives makes the problem worse. It has
been argued that equity-based incentives, especially option-based compensation, might
be at the root of the dramatic increase of earnings restatements in the last decade (Jensen
2005; Burns and Kedia 2006; Efendi et. a1. 2007)]. In this study, I examine whether and
how shareholders change their demand for accounting conservatism after restatement
announcements. Because a restatement announcement indicates a financial reporting
failure (GAO 2002) and severe agency problems that may have been induced by high
executives’ option-based compensation, I expect that shareholders will change their
demand for accounting conservatism in response to the change in the degree of the
misalignment of manager’s incentives with those of shareholders in order to ensure that
their interests are protected.

Under agency theory, moral hazard problems exist in financial reporting when the
reports’ accounting measures inform shareholders about managerial performance and
influence managers’ welfare. Managers have incentives to introduce bias and noise into
the accounting measures in order to transfer wealth to themselves from shareholders.
Because managers typically have limited liability and limited horizons, settling ex post
with managers once they have made excessive distributions to themselves is difficult,

generating deadweight costs for the firm and for shareholders. Watts (2003) argues that

 

’ Several recent empirical studies provide mixed results regarding the association between executives"
equity-based incentives and the likelihood of restatements. These studies are reviewed further in section
2.2.1.

eenser
By re.
conser
in firm

temper

sharehc
bCI‘IVL‘CI

sharehe

are amt
(Bums ;
incentit
tradition
Wealth b
film Val:
III? misa
annOUne:
IEXpeet

Compi‘ns.

I’I ,_

’ "“33 a I

can-ITO;
.N

firm? '

conservative accounting helps address the issues of limited horizons and limited liability.
By requiring a higher degree of verification of recognizing gains versus losses,
conservatism reduces the managers’ ability to overstate earnings and cumulative changes
in firm value at any point in time. Therefore, shareholders are protected from
compensation overpayments to managers. 2

Agency problems as described above arise when the interests of managers and
shareholders are not aligned. I expect that the greater the misalignment of interests
between managers and shareholders, the greater the demand for conservatism from
shareholders in order to minimize their potential losses in the future. Recent literature on
restatements shows that CEO compensation incentives, especially stock option holdings,
are among the key driving forces behind the aggressive financial reporting behavior
(Burns and Kedia 2006; Efendi et. a1. 2007). These studies demonstrate that very high
incentives from CEOs’ option holdings exacerbate the agency problem above a
traditional level whereby managers focus on the short-term maximization of their own
wealth by artificially inflating earnings beyond the scope allowed by GAAP and sacrifice
firm values in the long run. Because restatement announcements reveal to shareholders
the misaligned incentives induced by high CEO compensation, I expect that after the
announcements shareholders will demand more conservative reporting. More importantly,
I expect the increase in conservatism to be positively associated with the level of CEO

compensation incentives awarded during the years prior to the announcements.

 

2 In addition, Chen, Hemmer, and Zhang (2007) demonstrate that when accounting numbers are also used
to assess a firm’s expected future payoff, conservatism can dampen managers’ incentives to manage
earnings, especially in situations in which top management teams have significant equity stakes in their
firms.

cannot
incenti

HCCO 1111

contract
propose
and the
anagei
perform;
their (lit
announe
the C E0

thonghl

 

The above prediction is based on the assumption that restating firms, on average,
cannot address their CEOs’ compensation problem in a timely fashion. If the high CEO
incentives can be reduced immediately afier the announcements, the change in
accounting conservatism will only be a mechanistic response on the previous accounting
aggressiveness and we should not expect to see more conservative reporting by restating
firms when comparing them with similar non-restating firms. However, as documented in
Cheng and Farber (2008), it takes about two years on average for restating firms to re-
contraet with their CEOs and significantly reduce their option-based compensation. They
propose that restatements result from too high a level of CEO compensation incentives,
and the reduction in option grants alleviates the severe agency problem by better aligning
managerial incentives with those of the shareholders, resulting in improved firm
performance. Based on their finding, I expect that restating firms will, on average, reduce
their demand for the higher level of conservatism by the end of the second year after the
announcements. This decrease in conservatism should be associated with the reduction of
the CEO compensation incentives, which alleviates the severe agency problems conveyed
through the announcements.

Prior to testing my hypotheses, I provide descriptive evidence on how accounting
conservatism changes for restating firms after restatement announcements. I use two
measures of conservatism: an accrual-based measure developed by Ball and Shivakumar
(2006) and a market-based measure from Basu (1997). Using a sample of 354 restating
firms identified by the GAO (2002) and (2006) reports and with the accrual-based
measure, I find evidence consistent with my predictions. In particular, I compare the

levels of conservatism between restating firms and non-restating firms after the

announ
Compu
rename
amour
ahertl
eonser
pest- 2
eenser
pncrt
restate
eonse
inner

llilafli

C0119

at}?

COIN

C011

announcements. Two groups of non-restating firms are used: all other firms from the
Compustat database and a size-, industry-, and year-matched group of firms. 1 find that
restating firms’ earnings were more conservative in the year of the restatement
announcement (FY t). The trend continued in FY t+1. Interestingly, by the second year
after the announcement (FY t+2), there were no significant differences in accounting
conservatism between restating firms and non-restating firms. I also compare the pre- and
post- announcement periods for restating firms. I find that their earnings became more
conservative in the year of the restatement announcement (FY t) than in the two years
prior to the earliest year restated (FY tn-l and FY tn-2, where FY tn is the earliest year
restated and It refers to the number of restated periods)’. However, this high level of
conservatism declined during the two years (FY t+1 and FY t+2) following the
announcements. Overall, the evidence is consistent with restating firms changing their
financial reporting behavior within the two-year period afier the announcements.

The second objective of my paper is to examine whether the changes in
conservatism are consistent with my hypotheses. I analyze both the level of conservatism
at FY t and the change in conservatism from FY tn-l to FY t and their associations with
the change of CEO stock option grants from FY tn-l to FY t. Similar analysis is also
conducted for the period from FY t to FY t+2. Using the accrual-based measure and after
controlling for other factors affecting conservatism based on prior literature, I find a

positive relationship between the change in CEO option grants and the level of

 

’ Following Cheng and F arber (2008) and Badertscher (2007), I define FY tn-l as the first year prior to the
restating firm engaging in aggressive accounting activity. For example. in fiscal year 2001. Kroger
Corporation announced it will restate its fiscal year 1998, 1999, and 2000 financial statements. Therefore,
FY t is 2001. FY tn-l is 1997 when Kroger still had normal financial reporting behavior. Comparing FY tn-
1 (and FY tn-2) to FY t allows me to examine whether the firm’s financial reporting strategy changes after
the restatement announcements. Figure 1 illustrates the event years around a restatement announcement.

L<

I ‘u, I17,

I

conserr
consert
was pri
penod t
finns t
annount

lllCZiSUIL’

renatcni
COHSCI\E
accounn
danand
conduct
increase
magninu
films a]
Tesialcm.
hié’herle
(I
Witserta
AHHOUnQ

demand 1

Ismg the
EMT-um”.

accounl} n

1

(’3 rr

conservatism in both FY t and FY t+2. Moreover, results from the change of
conservatism specification reveal that the increase in conservatism from FY tn-l to FY t
was primarily driven by restating firms that increased CEO option grants during the
period tn—l to t, while the subsequent decrease in conservatism was driven by restating
firms that reduced their CEOs’ option awards in the two years following the
announcements. I fail to find results consistent with my hypotheses using the Basu
measure. This is explored further in section 6.1.2.

In addition to CEO compensation incentives, I also investigate the impact of the
restatement characteristics on the changes in conservatism, especially the increase in
conservatism in FY t. Specifically, a higher restatement magnitude or an indication of
accounting fraud might reveal a greater agency problem to shareholders, leading to a
demand for more conservative financial reporting immediately after the announcements. I
conduct further analysis that includes these two factors. Although I do not find that the
increase in conservatism from FY tn-l to FY t is driven by firms with high restatement
magnitudes or fraud firms, the significant association between the change in CEO option
grants and the level of conservatism in FY t disappears after I control for these two
restatement characteristics. More interestingly, I find fraud firms have a significantly
higher level of conservatism than non-fraud firms in FY t4.

Overall, the above evidence suggests that investors demand a higher level of
conservatism in response to the severe agency problem revealed by a restatement
announcement. In addition, subsequent to the announcement, they gradually reduce their

demand for conservatism as the agency problem is alleviated through re-contracting of

 

4 Using the change specification, 1 find both fraud firms and non-fraud firms significantly increase their
accounting conservatism from FY tn-l to FY t. The level and the change analyses on the impact of the
accounting fraud on conservatism present a troubling finding that I hope to explore in future research.

”.m1! ’4!

.—

C E0
charact

financi.

my W
has for
CI ltlt‘llt
contra:
Watts

negatix
authors

UCIWCC

aCCOun'

Underlt

CEO compensation. Furthermore, in addition to CEO option incentives, the
characteristics of restatements also play an important role in explaining restating firms’
financial reporting behavior immediately after the announcements.

I believe my study makes two contributions to financial accounting research. First,
my paper furthers our understanding of accounting conservatism. While recent literature
has focused primarily on the demand for conservatism arising from debt contracting,
evidence regarding the demand for conservatism from shareholders due to compensation
contracting has been lacking. My study not only provides empirical evidence to support
Watts (2003), but also complements LaFond and Roychowdhury (2008) who find a
negative association between managerial ownership and accounting conservatism. The
authors use managerial ownership as the proxy for the degree of alignment of interests
between managers and shareholders. They hypothesize and find that cross-sectional
variation in managerial ownership leads to cross-sectional variation in accounting
conservatism. In my study, I use the restatement announcement as an indicator of the
misalignment of manager’s incentives with those of shareholders, and examine how
accounting conservatism changes in response to the change in the degree of the
underlying agency problem.

Second, my study extends the restatement literature. Studies examining the
change in the financial reporting behavior of restating firms have provided mixed results.
Moore and Pfeiffer (2004) find that there are no differences in restating firms’ total
accruals before and after restatements. Wilson (2008) documents that the information
content of earnings announcements (ERC) declines in the period immediately following a

restatement, yet returns to the pre-restatement level in an average of four quarters.

Badensc
accrual

contrihut
financial
impohar
response
matlpr
ofmed

mnount

their re
L300Tt
"'IIIOUI‘
‘3‘» DIV c
Impun

financ

Badertscher (2007) provides evidence that restatements have reformative effects on both
accrual management and real activity management for restating firms. My paper
contributes to this literature by predicting and documenting a change of restating firms’
financial reporting behavior through examining accounting conservatism. More
importantly, existing studies do not examine whether their observed response (or lack of
response) is a function of the incentives that drive restatements. My study is different in
that I predict and directly examine the association between compensation incentive, one
of the determinants of restatements, and the observed change in conservatism after the
announcements.

Finally, this study is most closely related to the following two concurrent working
papers. Huang and Zhang (2009) hypothesize and find that overstating firms report more
conservatively after restatements than before restatements, and after the announcements,
their reports are also more conservative than the reports of their non-restating peer firms.
LaGore (2008) finds a similar result by comparing the three-year period before the
announcements to the three-year period after. Both studies, however, fail to examine the
evolvement in conservatism during the years after the announcements. And more
importantly, they do not investigate the relation between the incentives for aggressive
financial reporting and the change of conservatism.

The reminder of the paper is structured as follows. In chapter 2, I review the
literature on conservatism and restatements. Chapter 3 contains my hypothesis
development. Chapter 4 outlines my research design. In chapter 5, I introduce my sample
and data. Chapter 6 provides the empirical results. Chapter 7 presents several sensitivity

tests. I conclude my paper in chapter 8.

CHAPTER 2: RELATED LITERATURE

My paper is related to studies that examine the demand for conservatism due to
management compensation contracts. My paper is also related to studies that investigate
the causes and consequences of restatements and changes in the financial reporting
behavior of restating firms. Below I briefly review these literatures, their relations to my
study, and my contributions to these literatures.

2.1 Conservatism Research

Basu (1997) interprets conservatism as capturing accountants’ tendency to require
a higher degree of verification for recognizing good news than bad news in financial
statements. Thus, earnings is more timely or concurrently sensitive in reflecting publicly
available bad news than good news. Watts (2003) proposes that from the contracting
perspective, conservative accounting is a means of addressing moral hazard caused by
parties to the firm having asymmetric information, asymmetric payoffs, limited horizons,
and limited liability. The asymmetry in the application of conservatism leads to a greater
delay in the recognition of gains than losses. The probability of overstating earnings and
net assets at any point in time is lowered, thus reducing the likelihood of losses incurred
by investors that result from management’s manipulation of financial reports.

The role of conservatism as proposed in Watts (2003) has been investigated in
prior theoretical work in a principal-agent setting. Kwon, Newman, and Sub (2001) show
that under a limited liability construct, the optimal accounting system will always be
conservative in order to motivate and compensate the agent. Chen, Hemmer, and Zhang
(2007) extend Kwon et. al. (2001) by analyzing the dual roles of accounting numbers. In

their paper, the authors model accounting numbers as serving both a valuation role (for

Cons
OWI'IC
impli

HHCIS

 

sake

CUUBC
and F
and r
addre
Withir
FIObls
aSIFnt
Dunc]
DIEM

Zhang

HO”),

potential investors to assess firm’s value) and a stewardship role (for current investors to
monitor managers). They show that under conservative accounting standards, the
equilibrium level of earnings manipulation is reduced. Conservatism reduces the impact
of news on share prices and consequently reduces the benefit of earnings management.
Conservatism also increases the cost of earnings manipulation. Therefore, the current
owner will engage in less earnings management. As indicated by the authors, the model
implies that conservative principles are more likely to arise in situations where the self-
interested parties involved in the financial reporting process have a significant equity
stake in their firms.

However, there is little empirical evidence demonstrating the demand for
conservatism from shareholders due to management compensation contracts. In LaFond
and Roychowdhury (2008), the authors examine the association between conservatism
and managerial ownership. The authors argue that to the extent that conservatism
addresses the agency problem resulting from the separation of ownership and control
within a firm, increased managerial ownership should decrease the severity of the agency
problem and thus decrease the demand for conservatism. They document that the
asymmetric timeliness of earnings is negatively associated with the level of managerial
ownership. But interestingly, they find that conservatism does not decrease with the level
of managerial option holdings. This finding supports the theory of Chen, Hemmer, and
Zhang (2007) and implies that option holdings by management on average require a
higher degree of conservatism (or at least do not reduce the demand for conservatism

from shareholders).

Chen.
Roychr
ahgnnn
tanano
consert
nusahgr
account
undeflyi
EXCCUU\
equnyir
a Donna
manager
connolt
PIOI Ides
Change n:
It) IIIE'sc C

2.2 Restn

Inlernal

Indepcn.

I

LaFond
[EOPU0p.
“Vince a.

 

My paper provides empirical evidence supporting Watts (2003) and in particular
Chen, Hemmer, and Zhang (2007). Moreover my study complements LaFond and
Roychowdhury (2008). They use managerial ownership as the proxy for the degree of the
alignment of interests between managers and shareholders, and find that cross-sectional
variation of managerial ownership leads to cross-sectional variation in accounting
conservatism5 . My paper uses the restatement announcement as an indicator of the
misalignment of manager’s incentives with those of shareholders, and examines how
accounting conservatism changes in response to the change in the degree of the
underlying agency problem. Moreover, their study assumes equilibrium design in
executives’ compensation incentives, while my study adopts a setting in which CEOs’
equity incentives evolve from an off-equilibrium level to a new equilibrium level. Finally,
a normal cross-sectional test might have difficulty identifying a relationship between
management compensation designs and accounting choices without carefully specifying
control variables and the relations among the variables. Therefore, the restatement setting
provides the advantage of being able to exploit an event that highlights the need to
change management compensation incentives and how accounting conservatism responds
to these changes.
2.2 Restatement Research
2.2.1 Causes of Restatement

Plumlee and Yohn (2009) find that restatements are most often related to basic
internal company errors. Agrawal and Chadha (2005) find that the presence of

independent directors with finance expertise reduces the likelihood of a misstatement. On

 

5 LaFond and Roychowdhury (2008) use a measure of managerial ownership that excludes shares granted
in options. They find that conservatism does not vary significantly with the shares in outstanding options
owned by the CEOs. This is an equilibrium study while my study adopts an off-equilibrium setting.

10

term
restri
.hfnl:
equit
subj c
(300

and t

misr

SIIV

P‘_.
I):

annig;

COpf-

finih

N J

n'.
n 9,. -

the other hand, Baber et. al. (2007) fail to find a significant link between a comprehensive
set of internal governance mechanisms and the likelihood of restatements. Regarding
possible economic determinants of restatements, Richardson, Tuna and Wu (2002)
identify several capital market incentives (e.g., raising cash capital) and debt market
incentives (e.g., high leverage) that might explain accounting aggressiveness".

A stream of studies focuses on the role played by executive compensation
incentives. Larcker et a1. (2007) do not find an association between CEOs’ total long-
term compensation and restatements. Similarly, by adding the incentives of CEOs’ stock,
restricted stock, and stock options together and using propensity score matching,
Armstrong et al. (2009) also do not find evidence of a positive association between CEO
equity incentives and accounting irregularities. Using a sample of 87 firms that are the
subject of the SEC’s Accounting and Auditing Enforcement Releases, Johnson et al.
(2009) document a positive relation between executive incentives from unrestricted stock
and the likelihood of fraud.

Other studies consider the specific role of CEO stock options in providing
incentives for accounting misstatements7. Burns and Kedia (2006) demonstrate that the
CEO’s option sensitivity is positively associated with restatements. They further show
that higher incentives from stock options are also associated with higher magnitudes of
misreporting. Such a relation does not exist for equity and restricted stock. Efendi,

Srivastava, and Swanson (2007) extend Burns and Kedia (2006). They use CEOs’ in-the-

 

6 These incentives, other than CEO compensation. might also affect the change in conservatism after the
announcements, such as leverage. In chapter 4, I discuss other factors that might affect conservatism and
control for these factors in my subsequent testing.

7 Different from restricted stocks and stock ownerships, stock options possess distinctive features:
convexity and asymmetric payoff. These features induce strong incentives for executives to manipulate
financial reports. Burns and Kedia (2006) and Efendi, Srivastava, and Swanson (2007) contain more
detailed discussion on these issues.

11

IIIC

I,“
('4
F1

.0-‘

sew

their

Rich
anno
that r
mont:
resta:

Shun

bCN'Qg

money option holdings to capture both equity overvaluation and CEO option price
sensitivity. They find that large amounts of in-the-money options are much more likely to
be involved in restatements.
2.2.2 Consequences of Restatements

Empirical studies on restatements demonstrate that restatements typically result in
severe capital market consequences for restating firms and labor market punishments for
their management. From the perspective of the capital market effect, Palmrose,
Richardson and Scholz (2004) document a mean market reaction to restatement
announcements of —9.2% over a two-day event window. Hribar and Jenkins (2004) find
that relative percentage increases in the cost of capital average between 7 and 19% in the
month following a restatement. Gleason et. a1. (2008) find that the share prices of non-
restating firms in the same industry also decline. With respect to the labor market,
Srinivasan (2004) finds an abnormally high board turnover within three years of the
restatement announcement. Desai, Hogan, and Wilkins (2006) show that almost 60% of
restating firms experience top executive tumover within 24 months. Hennes, Leone, and
Miller (2007) report even higher (91%) CEO or CFO turnover by using a refined
methodology to distinguish intentional misstatements from unintentional ones. Finally,
Palmrose and Scholz (2004) demonstrate that 38% of restating firms have civil litigation
after restatements.

In addition to management turnover, Cheng and Farber (2008) examine CEOs’
compensation re-contracting following restatements. Building on the positive link
between stock-based compensation and aggressive accounting established in the prior

literature, their analysis shows that the ratio of the value of option grants to total

12

comp
nto-y

IIIIS

induct
financ
and bl

ofarc

renan
aRI'IOU
the r
301101}
accru,

EITOI'S

e“lib:

IUI I D at

'2

p.
U)

compensation for restating firms’ CEOs decreases significantly on average during the
two-year period after restatement announcements. More importantly, for those restating
firms with reductions in CEO option grants, operating performance improved during the
same period, implying a better alignment between management and shareholder interests
after CEO compensation re-contracting.

2.2.3 Restatement and Financial Reporting Quality

Conventional wisdom suggests that the severe consequences of restatements will
induce firms to reform their accounting policies and practices towards less aggressive
financial reporting behavior. In addition, heightened SEC scrutiny, more diligent auditors,
and board of director oversight could also lead to changes in financial reporting behavior
of a restating firm.

Yet empirical studies on the change in financial reporting quality after
restatements provide mixed results. Moore and Pfeiffer (2004) examine 72 restatements
announced between 1997 and 2000. They find that total accruals for the period preceding
the restatement do not differ significantly from those of the post-restatement
announcement period. Similar results hold for working capital accruals and other
accruals. The paper further documents that the patterns in earnings growth and forecast
errors are also not different across the two periods.

On the contrary, Badertscher (2007) documents that restating fimrs identified as
aggressive non-GAAP earnings management firms based on the GAO (2002) report
exhibit less real activity earnings management and less accrual earnings management
following a restatement. Wilson (2008) provides evidence that the ERCs of restating

firms fell significantly in the two quarters immediately following restatements when

13

compa

pret’lOI

armour

associz
(or lac
comp:
in fig;
doeurr

compared with the quarter prior to the restatements (the base quarter). But it returns to its
previous level in the 4th quarter. Furthermore, there is no loss in the information content
of earnings for firms that make changes in their governance structures following
restatements. Finally, both Huang and Zhang (2009) and LaGore (2008) find that
restating firms report more conservatively after the announcements than before the
announcements.

One limitation of the above papers is that they do not directly examine the
association between the incentives that induce misstatements and the observed changes
(or lack of changes) after restatements. If factors causing restatements, such as executive
compensation remain the same, it is plausible that management may continue to engage
in aggressive reporting behavior. My study is different in that I not only predict and
document the direction of the change in conservatism, but also directly test whether the

change is associated with compensation incentives.

14

them
that t
aecou:
manag
Healy'
COI‘por
partial}
PTOPric
based (
also dc
malfcag
earning
bem'een
“Ianlclt

Emma's

 

CHAPTER 3: DEVELOPMENT OF HYPOTHESES

In this chapter, I first use the earnings management literature to demonstrate the
agency problem between shareholders and managers induced by executive compensation.
I then focus on the role of conservatism in reducing this agency cost. Based on these
discussions, I develop my hypotheses.
3.1 Executive Compensation and Agency Problem

It has long been recognized that eamings-based compensation provides an
incentive for management to manipulate accounting income. Sloan (1993) demonstrates
that the level of bonus payments is usually linked to reported earnings and other
accounting based performance measures. Healy (1985) documents that executives
manage accruals to maximize their bonus payments. Guidry et. al. (1999) find support for
Healy’s conclusion using internal data from different business units within a single
corporation. Gaver, Gaver, and Austin (1995) and Holthausen, Larcker, and Sloan (1995)
partially confirm and also advance Healy’s finding by using refined methodology and a
proprietary dataset. In general, this literature demonstrates that incentives from eamings-
based compensation induce managers to manipulate earnings. The restatement literature
also demonstrates that for restating firms classified as having engaged in accounting
malfeasance, executive bonus payments provide an incentive to managers that lead to
earnings restatements (Efendi et. al. 2007).

There is also an emerging line of academic research that examines the link
between equity-based compensation and aggressive financial reporting. Cheng and
Warfield (2005) find that CEOs with high equity incentives are more likely to engage in

earnings management in order to meet or beat analyst forecasts. The magnitude of

15

abnonn
Bergstr:
abnonn

incentix

subsequ
demons
exercisi
discretit

abnonn;

enable 1
IIICCIIIIV
ttldene
Show rh
OPTIOH 5
Illlancja
Sl0CI§ OI

mm C

mafia ’r

,-

C

{I

abnormal accruals is positively related to CEO stock-based compensation. Similarly,
Bergstresser and Philippon (2004) find a positive correlation between the magnitude of
abnormal accruals and the proportion of CEO total compensation that is related to equity
incentives. Beneish and Vargus (2002) reveal that income-increasing accruals are
significantly less persistent for firms with abnormal insider selling and that these firms
subsequently suffer significantly lower stock returns. Bartov and Mohanram (2004)
demonstrate that top executives inflate earnings through accruals management prior to
exercising their stock options and selling the acquired shares. The abnormally high
discretionary accruals reverse in the post-exercise period, and stock price changes are
abnormally negative during that period.

In related research on restatements, Burns and Kedia (2006) argue that options
enable management to extract rents in the form of excessive compensation and create
incentives for management to adopt aggressive accounting practices. They find strong
evidence that option sensitivity is positively associated with misreporting. They further
show that larger magnitudes of restatements are associated with higher values of CEO
option sensitivity. Efendi et al. (2007) demonstrate that the likelihood of a misstated
financial statement increases greatly when the CEO has sizable holdings of in-the-money
stock options. These results indicate that substantially overvalued equity creates agency
costs by causing managers to take actions to support stock price.

The above studies demonstrate that there are agency costs induced by CEO stock-
based compensation, especially option holdings, and that these agency costs are
associated with earnings management. If option compensation is intended to provide

managers with incentives to act in the best interests of shareholders (Jensen and Meckling

l6

.4

1976; SI
examine
the agcn
Guay. 31
that tin
executix
to Chang
stock pr
det‘iatio:
a greate:

to earnir

 

 

 

1976; Smith and Stulz 1985), a natural question that arises is why the option incentives
examined in the above studies fail to curb aggressive reporting and, instead, exacerbate
the agency problem. Following Cheng and Farber (2008), I appeal to arguments in Core,
Guay, and Larcker (2003) and Core and Guay (1999) to help explain this. They propose
that firms choose an optimal level of equity incentives when they contract with
executives. But over time, managers’ equity incentives deviate from the optimal level due
to changes in firm and/or manager characteristics. For example, the incentives vary with
stock price, stock-return volatility, and the time remaining until the options expire. The
deviation exacerbates the sensitivity of CEOs’ wealth to short-term stock prices, inducing
a greater misalignment of interests between management and shareholders, which leads
to earnings management and eventual earnings restatement.

Earnings management and particularly earnings restatements are costly to a firm
and its shareholders. It is well documented that restatements are associated with
significant decreases in firms’ market value (Palmrose et al. 2004; GAO 2002; and GAO
2006). Moreover, as demonstrated in the earnings management literature, executives
receive high cash bonuses based on manipulated accounting numbers. Investors’ wealth
is damaged because the excess amount cannot be fiilly recovered in the subsequent period
when the misconduct is uncovered. This is caused by managers having limited liability
and short employment horizons. Similarly, executives engage in earnings management to
inflate short-term stock price in order to boost the value of their equity-based
compensation. Shareholder values are damaged when executives sell unusually large
numbers of shares during the manipulated period, which is normally followed by low

reported earnings and stock returns’ reversal during the subsequent years (Beneish and

17

 

Vargu:
values

3.2 Ag

cash \
in We
likely
limit:

is hat

COnt

Vargus 2002; Bartov and Mohanram 2004; Bergstresser and Philippon 2006). Firm
values are also reduced when executives focus on non-value added activities.
3.2 Agency Problems and Accounting Conservatism

The above discussion demonstrates that executive compensation provides
incentives for managers to overstate earnings in order to transfer wealth to themselves.
This excess compensation to managers, in the form of paid cash bonus and/or realized
cash value from exercising options and selling stocks, is difficult to recover. As discussed
in Watts (2003) and LaFond and Roychowdhury (2008), ex post settling with managers is
likely to be extremely costly and incomplete due to managers’ limited horizon and
limited liability. Moreover, the loss estimation is difficult to verify and the accused fraud
is hard to prove or to distinguish from bad outcomes that are due to chance. Even with a
legal finding of fraud and a damages award, the excess payments typically cannot be
firlly recovered (Watts 2003). As discussed earlier, shareholders lose not only in the form
of excess compensation payments, but also when stock returns plunge during the post-
manipulation period and when managers are distracted from their primary function of
efficiently managing the firm. These issues create a demand for more efficient
contracting ex ante.

Watts (2003) argues that conservatism emerges as one mechanism to facilitate
efficient contracting. By applying stricter verification standards for recognizing good
news as gains than for recognizing bad news as losses, the probability of overstating
earnings and net assets at any point in time is lowered, thus reducing the likelihood of
losses incurred by investors that result from management’s manipulation of financial

reports. Furthermore, as demonstrated in the analytical model of Chen, Hemmer, and

18

 

Zhang
pnces

ninrec

sharcht
prohler
are les
nusahg
executi
optunal
resunen
allllOLlnt
executr

IIIIS in

II'hich l

ahernat

aISO IQ]

 

Zhang (2007), conservative accounting standards reduce the impact of earnings on share
prices, thereby reducing the benefit of earnings management. Consequently, executives
with equity holdings of the firm will engage in less earnings management.

In summary, conservative financial reporting is hypothesized to facilitate
shareholder-management contracting in the presence of agency problems. Agency
problems are likely to be more severe when the interests of managers and shareholders
are less aligned. In chapter 3.1, I argue that for restating firms, the degree of
misalignment of interests between shareholders and managers is exacerbated when
executive compensation, especially in the form of stock options, deviates from the
optimal level. Thus, I predict that the demand for conservatism will increase after
restatement announcements in response to the agency problem revealed through the
announcements. This increase in conservatism will restrain the opportunities of
executives to further manipulate earnings and limit additional loss of shareholders’ value.
This increase in conservatism will also result in higher levels of conservatism for
restating firms after the announcements when comparing them with non-restating firms,
which face less severe agency problems. Accordingly, I hypothesize that (stated in the
alternative form):

Hla: immediately after restatement announcements, restating firms will, on
average, have higher levels of accounting conservatism relative to non-restating firms and
also relative to pre-announcement period;

Hlb: the increase in conservatism is associated with the CEO compensation

incentives awarded during the years leading up to the announcements.

l9

Hypothesis 1 is based on the assumption that restating firms, on average, cannot
address their CEOs’ compensation problem in a timely fashion. More specifically,
because restatements result from a high level of incentive compensation, restating firms
should re-contract with their CEOs, reducing their compensation and mitigating their
agency problems. If the high CEO incentives can be reduced immediately after the
announcements, the change in accounting conservatism will only be a mechanistic
response on the previous accounting aggressiveness, and we should not expect to see
more conservative reporting by restating firms when comparing them with similar non—
restating firms. But as proposed in Core and Guay (1999) and Core et al. (2003), the
compensation re-contracting process will likely take some time. Specifically, the authors
point out that information gathering and processing costs prevent firms from adjusting
incentives to optimal levels on a timely basis. Furthermore, if transaction costs associated
with re-contracting exceed the benefits, some restating firms may not start to reduce
compensation incentives when the issue emerges. Finally, the action of reducing bonus or
option grants may have a negative impact on managers’ effort levels, which might also
result in some restating firms slowing down the re-contracting processg.

In fact, Cheng and Farber (2008) find that, on average, it takes about two years for
restating firms to re-contract with their CEOs and reduce their option-based
compensation significantly. In particular, Cheng and Farber (2008) find that restating

firms become more aggressive with option usage during the misstated periods, and their

CEOs’ option grants are significantly reduced by the end of the second year (i.e., FY t+2)

 

8 In Core and Guay (1999) and Core et al. (2003), the authors do not address the compensation re-
contracting process when firms experience top executive turnover. It is possible that for restating firms who
experience CEO turnover, the hiring of a new CEO and the compensation re-contracting might occur at the
same time. I address the turnover issue in my subsequent analysis.

20

after the announcements. Their findings support the notion that high CEO stock options
are the main driving force behind the misstatements, and restating firms start to take
actions to address their greater agency problem after the announcements. They further
document that for those firms who reduce option grants by FY t+2, their operating
performance improved. But more importantly, their observation that there were no
significant changes in option grant usage until FY t+2 supports the notion that restating
firms will re—contract with their CEOs but the re-contracting process will take some time.

If restating firms reduce their CEO stock option grants by the end of FY t+2,
which results in a better alignment of interests between shareholders and managers, I then
expect that the demand for greater conservatism after the announcements will fall
subsequent to the two-year period following the announcements. Accordingly, I
hypothesize that (stated in the alternative form):

H2a: after initially increasing conservatism, restating firms will, on average,
reduce their high level of conservatism during the years subsequent to the announcements;

H2b: this decrease in conservatism is associated with the reduction of the CEO

compensation incentives during the same period.

21

CHAPTER 4: RESEARCH DESIGN

4.1 The Accrual-Based Measure of Conservatism

I employ two measures of conservatism. One is accrual-based and the other one is
market-based. The accrual-based method is developed in Ball and Shivakumar (2006),
who investigate two roles of accruals. The first is consistent with Dechow (1994), who
proposes that cash flow “noise” arises from timing and matching problems that reduce
the ability of cash flow to reflect firm performance. Accrual accounting shields
accounting income from this transitory noise, making it a more efficient performance
measure. Thus accruals and cash flows from operations are contemporaneously
negatively correlated, and accruals offset transitory increases and decreases in cash flows.

Ball and Shivakumar propose a second role for accruals: asymmetrically timely
loss recognition. The authors argue that economic gains and losses can be thought of as
the current-period cash flow plus any revision in the present value of expected future cash
flows. Because accruals are based in part on revisions in future cash flow expectations,
economic gain and loss recognition must be accomplished in part through accruals. This
gain and loss recognition role of accruals is a source of positive correlation between
accruals and current-period cash flow. They further propose that the positive correlation
between accruals and current cash flows is asymmetric due to the conservative nature of
accounting where economic losses are more likely to be recognized on a timely basis than
economic gains. Based on this discussion, the authors modify the Jones (1991) accrual
model by incorporating the loss recognition role of accruals.

To examine the change in conservatism around restatement announcements, I

employ the following model:

22

It

ACCjt=a0 +a1CFOj, +a2ASa/esJ-t +a3FASSETj, +a4DCFOJ-t +a5DCFOj, XCFOJ'I
+a6Restatej,(0r Tj)+ a7Restatej,(0r Tj)x CFOj, + agRestatej,(0r Tj)x ASaleSj,
+a9RestateJ-t (or Tj)>< FASSETJ-t +a10RestateJ-t (or Tj)>< DCFsz

+a1lRestatej,(0rTj)xDCF0jtxCFOjt+£jl (l)

where ACC is total accruals, which is defined as net income before extraordinary items
minus cash flow from operations; CF 0 is operating cash flow; ASa/es is change in sales;
FASSE T is gross property, plant and equipment; DCFO takes the value 1 if CF 0 <0 and
0 otherwise. All continuous variables are scaled by beginning total assets. To examine the
change in conservatism for restating firms, I adopt the design similar to Ball et al. (2008)
where the authors examine the change in conservatism for IPO firms. In particular, I first
compare restating firms to non-restating firms (more details in chapter 6). Restate takes
the value 1 for restating firms and 0 otherwise. I then examine the change in conservatism
within restating firms. T indicates the time period relative to the year of the restatement
announcement, FY t. For HI, I expect all to be significantly positive when comparing
restating firms to non-restating firms for the restatement announcement year and
significantly negative when comparing the announcement year (the benchmark year) to
other time periods, indicating a higher degree of conservatism in FY t. For H2, I expect
all to be insignificant when comparing restating firms to non-restating firms and
significantly negative when comparing restatement announcement year (the benchmark
year) to the subsequent period, indicating a reduced level of conservatism after the

announcements.

23

To examine the association between CEO compensation incentives and
accounting conservatism, I add the change in option and bonus awards to the above

regression and conduct the test within restating firms:

ACCJ', =a0 + a1CFOfl+ azASalesJ-t + a3FASSETfl + a4DCFOj, + asDCFOJ-t x CF01"
+ a6AOpti0n% _ Rank jt + a7 AOpti0n% _ Rank ft x CF 0 jt + agAOption% _ Rank jt x ASaIesJ-t
+ a9A0pti0n% _ Rank j! x FASSE T j! + a1 0A0pti0n% _ Rank jl x DCFOj,
+ a11A0p1i0n% _ Rank j! x DCFOJ', x CF 0 jt
+ a1 zABonus% _ Rank j! + a] 3ABonus% _ Rank jt x CF 0 jt + a14ABonus% _ Rank ft x ASa/esj-t
+ a15ABonus% _ Rank 1'! x FASSE T j! + a16ABonus% _ Rank j! x DCFOJ',
+ a17ABonus% _ Rank ji x DCFOJ-t x CF 0 jt + Control Variables + 8]" (2 )

where AOption%_Rank (ABonus%_Rank) represents the rank of the change of option
grants (bonus awards) as a percentage of CEO’s total compensation. A higher value in
this variable indicates a higher increase in option grants (bonus awards) fiom one period
to another. All the other variables are the same as above. The three-way interaction

terms, AOpti0n% _Rank x DCFO x CF 0 and ABonuS% _Rank x DCFO x CFO ,

represent the incremental impacts of option changes and bonus changes on the main

measure of accounting conservatism - DCFO x CF 0 . Under both H1 and H2, I expect

significantly positive all and 0:17. The difference is that for H1, the significance implies

that increases in CEO option grants and bonus awards from pre-announcement period to
the announcement year result in a higher degree of conservatism in the year of
restatement announcements. On the other hand, for H2, the significance implies that
decreases in CEO option grants and bonus awards from the announcement year to the
subsequent period result in a reduced degree of conservatism. In addition to these two

compensation measures, I also control for leverage, litigation risk, market-to-book ratio,

24

and firm size. All the control variables interact with the main variables in the accrual

model. The coefficients on their interactions with DCFO x CFO indicate their relations

to conservatism.

Leverage_rank is the rank of the firm’s long-term debt divided by average total
assets, standardized to the interval (0, 1). As discussed in chapter 2, Watts (2003) argues
that conservative accounting can reduce the likelihood of losses incurred by creditors.
Firms with a high level of leverage tend to have greater bond-holder and share-holder
conflicts. By requiring more conservative accounting, creditors can receive a more timely
signal of deteriorating financial performance through a tightening of covenants or a faster
triggering of covenant violations, thus improving the efficiency of debt contracting. For
example, Ahmed, Billings, Morton, and Harris (2002) hypothesize and find that when
firms face more severe bondholder-shareholder conflicts over dividend policy,
bondholders will .force the firm to adopt more conservative accounting. Zhang (2008)
documents that the likelihood of a covenant violation following a negative shock
increases in borrower conservatism and more conservative borrowers violate covenants
sooner. Therefore, creditors’ interests are protected ex ante by conservatism. Based on
these studies, I expect the leverage variable to be positively related to conservatism.

Lit is a dummy variable to control for firms with high litigation risk.
Overstatement of income and net assets is more likely to generate litigation costs. To
reduce expected litigation cost, firms adopt conservative accounting rules (Watts 2003).
Huijgen and Lubberink (2005) find that earnings of UK. firms cross-listed in US. are
significantly more conservative than earnings of UK. firms without a US. listing.

Because both types of firms report under UK. GAAP, they interpret the result as

25

indication of a stricter enforcement regime, implying a higher litigation risk faced by the
cross-listed firms. I control for litigation risk. by including a dummy variable that
classifies firms with four-digit industry codes between 2833-2836, 3570-3577, 3600-
3674, 5200-5961, and 7370-7374 as belonging to high litigation risk industries following
Francis, Philbrick, and Schipper (1994) and LaFond and Roychowdhury (2007). These
firms in general are in hi-tech and retail industries. I expect these firms to have a higher
level of conservatism.

As Roychowdhury and Watts (2007) demonstrate, one-year asymmetric-
timeliness measures, such as the Basu measure, are affected by the beginning
composition of equity value, including the effects of past asymmetric timeliness. To
explicitly control for the beginning level of conservatism, I include the beginning market-
to-book ratio following Roychowdhury and Watts (2007) and Lafond and Roychowdhury
(2008). MTB_rank is the rank of firms’ beginning market-to-book ratio and is
standardized to the interval (0, 1). I expect firms with high level of beginning MTB to
have lower conservatism.

Finally, I control for firm size. Early conservatism studies using size as a control
argue that larger firms could report losses in a more timely manner than small firms due
to differing agency costs or greater litigation risk. Large firms are also likely to face large
political costs that induce them to use more conservative accounting (Watts and
Zimmerman, 1978). Ball and Shivakumar (2005) report that conservatism increases with
firm size. But more recent developments in conservatism suggest the opposite. LaFond
and Watts (2007) argue that information asymmetry is often smaller for large firms

because they produce more public information, resulting in a reduced demand for

26

 

int’erent
annunt
hmn.

aggrega
l predit
average

4.2 The

“hero
Fear. .
momll

Othr

 

 

conservative accounting. Furthermore, Givoly et. al. (2007) demonstrate that the
aggregation of projects in large firms can lead to measurement error and incorrect
inferences regarding the level of conservatism. Both papers document that the
asymmetric timeliness of earnings is significantly smaller for large firms than for small
firms. Therefore, based on these recent studies, the information asymmetry and
aggregation effects might dominate concerns about agency costs and political costs. Thus
I predict size is negatively related to conservatism. Size_Rank is the rank of firms’
average total assets standardized to the interval (0, 1).
4.2 The Market-Based Measure of Conservatism

The market-based conservatism measure is from Basu (1997). Basu regresses
annual earnings on current annual returns. Due to accounting conservatism, earnings is
predicted to be more strongly associated with concurrent negative returns, a proxy for bad
news, than positive returns, which is a proxy for good news. Similar to my first measure,

the change in conservatism is tested using:

Ejt /Pjt—l =a0 “I'alet +6120le +a3DRjt ij,
+a4Restatej,(or Tj)+a5RestateJ-,(0r Tj)x Rj, + a6Restatej,(0r Tj)xDRj,
+a7Restatej,(0rTj)><DRJ-txRJ-t+£j, (3)

where E is the earnings per share, P is the price per share at the beginning of the fiscal
year, R is the buy-and-hold stock return on a firm for the 12-month period ended 3
months after a fiscal year end, and DR takes the value 1 if R <0 and 0 otherwise. All the
other variables are the same as previously defined. For HI, I expect a7 to be
significantly positive when comparing restating firms to non-restating firms and

significantly negative when comparing FY t to previous years, using FY t as the

27

 

henchr

IO IIOI

ClllllOUI

menu

 

benchmark year. For H2, I expect a7 to be insignificant when comparing restating firms

to non-restating firms and significantly negative when comparing restatement
announcement year to subsequent period, indicating a reduced level of conservatism after

the announcements.

28

CHAPTER 5: SAMPLE AND DATA
5.1 Sample of Restating Firms

My restatement announcement samples are selected from the 2002 and 2006
GAO reports. The 2002 GAO report contains 919 restatements announced by 845
companies from January 1997 through June 2002. The 2006 GAO report identifies 1,390
restatements announced by 1,121 public companies from July 2002 to September 2005.
These restatements include those due to an accounting irregularity, which the GAO
defines as “an instance in which a company restates its financial statements because they
were not fairly presented in accordance with generally accepted accounting principles
(GAAP). This would include material errors and fraud” (GAO 2002).

For the purpose of performing analyses subsequent to the announcements, I
restrict my restatement sample to the period from January 1997 to December 2004, which
contains 1,786 restatement announcements. Table 1 Panel A presents the restatement
sample reconciliation. 107 restatements could not be matched to the CRSP/Compustat
Merged Database using company names. 1,047 restating firms were not available on
ExecuComp. I further deleted 222 restatements due to multiple restatements by the same
firm. I also deleted 25 restatements by financial firms because their accrual behavior is
not well defined. By conducting a Lexis-Nexis search and reading the 8K5 and the
subsequent 10K or 10Q filings of each restatement, I further identified 31 announcements
for which no real restatements were issued. My final sample contains 354 restatements.

Panel B of Table 1 presents the distribution of restatements by year. There is a
trend of increasing restatements during the 8-year period. More than 50% of the

restatements are announced in the years 2002 to 2004. Panel C of Table 1 provides the

29

distribution of restatement characteristics by initiating entity, exchange listing of the
restating firm, and reason for the restatement. Company-initiated restatements account for
38.9% of the cases. More than half of my sample firms are listed on the NYSE. Revenue
recognition issues account for 43.0% of restatement cases.

As demonstrated by Hennes et al. (2008), the restatement data from the GAO
database might contain both irregularities (intentional misstatements) and errors
(unintentional misstatements). By distinguishing between errors and irregularities,
restatement studies can significantly enhance the power of tests when examining the
causes or consequences of restatements. Consistent with Efendi et al. (2007) and Hennes
et al. (2008), I also classify my restatement sample into fraud versus non-fiaud cases. By
reading the original restatement announcements from Lexis-Nexis, firms’ 8K3 and
subsequent 10K and/or 10Q filings, I classify those restatements as fraud when: 1)
variants of the words “irregularity” or “fraud” are used in describing the misstatement; 2)
SEC or DOJ investigations or charges are involved relating to the restatement; 3)
independent investigations are conducted (by audit committees, outside forensic firms,
etc.) relating to the restatement; 4) the restatements are prompted by auditors or SEC.
These procedures help me identify 120 restatement cases (Table 1 Panel C) as fraud. My
classification of each restatement is consistent with Hennes et al. (2008).

I also hand collect information about the number of years restated and the
magnitude of each restatement. Panel D of Table 1 shows that the median sample firm
restates two years of financial statements. For non-income increasing restatements, the
average effect of the restatement on net income is a reduction of about $95.24 million.

The median, however, is much lower, at $2.3 million. If a firm restated more than one

30

year, the effect on net income is the average annual effect on earnings over the entire
misreported period. The size of the restatement is a nontrivial percentage of net income.
For the average firm, the size of the restatement is about 42% of the absolute value of the
restated net income, which is calculated using the sum of the restated amount from each
restated year divided by the sum of the restated net income over the entire restated period.
The median value is lower at about 6.6% of restated net income.

Panel E of Table 1 indicates that my sample firms are from a broad spectrum of
industries. Panel F of Table 1 presents descriptive statistics for restating firms in the year
of the restatement announcement. The mean market value is $5.3 billion. The mean book
to market and leverage ratios were 0.492 and 0.232, respectively. These firms, on average,
were not profitable in the restatement announcement year. These descriptive statistics are
similar to those reported in Burns and Kedia (2006) and Chen and F arber (2008), except
that my firms are larger than those examined by Chen and Farber (2008), who hand-

collected compensation data for all of the restating firms in the GAO (2002) report.

31

CHAPTER 6: EMPIRICAL RESULTS

To test my hypotheses, I first examine changes in conservatism for restating firms
around restatement announcements. The results under the accrual—based measure are
presented first, followed by results using the market-based measure. I then document the
changes in CEO compensation. In the third part of this chapter, I investigate the
association between the observed changes of conservatism and the observed changes in
CEO compensation before and after restatement announcements.
6.1 Changes in Conservatism around Restatement Announcements
6.1.1 The Accrual-Based Measure

Table 2 presents results from tests of my first hypothesis using the accrual-based
measure. In Panel A, I estimate pooled cross-sectional regressions on restating firms and
all the non-restating firms on Compustat for event years tn-2, tn-l , t, t+1, and t+2, where t
is the fiscal year of the restatement announcement, and tn-l is the first fiscal year prior to
the earliest year restated. CF 0 is significantly negatively related to AC C in all the testing

years, consistent with the noise reduction role of accruals as proposed in Dechow (1994).
As expected, increases in sales are associated with higher levels of accruals ((12). Higher
values of PPE result in high depreciation expense and low earnings, which lead to low
accrual levels ((13). Turning to the conservatism indicator, as, I find that it is significantly
positive across all the periods, demonstrating the role of accruals in timely recognition of
bad news versus good news. These main effects are consistent with prior literature.
Because tn-l and tn-2 are years before the restated period, restating firms are

expected to exhibit normal financial reporting behavior during these years. Therefore I do

not expect them to have significantly different levels of conservatism when compared to

32

non-restating firms. My results, as evidenced by an insignificant coefficient on a“, are
consistent with my prediction. After restatement announcements, however, restating
firms become more conservative than non-restating firms, as indicated by a significant
positive coefficient on (111 in FY t and t+1. However, by the end of FY t+2, 0t” becomes
insignificant again.

In Panel B of Table 2, I compare the conservatism level of restating firms with a
group of non-restating firms operating in the same industries as the restaters and within
+/- 10% percent of the restating firm’s asset size in the same event year. Restating firms

who cannot be matched are excluded fi'om this analysis. The result is mostly in line with
my prediction and consistent with Panel A. (111 is insignificant in FY tn-2, indicating the

financial reporting behavior being in line with the matched non-restating firms in the
same period. But the restating firms behave more aggressively before the restated period,
’which is reflected in the significantly negative on in FY tn-l. In FY t, restating firms’
earnings become significantly more conservative when compared to this matched group,
consistent with Hla. Higher levels of conservatism continue for one more year (FY t+l),
and the two groups do not show a different conservatism level in FY t+2, consistent with
my prediction in H2a.

In Panel C of Table 2, I examine the changes in conservatism within restating
firms over time. FY t is my benchmark year. A positive (negative) coefficient on the

interaction term T xDCFO xCFO indicates that the event year is more (less)

conservative then FY t. I find that FY tn~2 has a similar conservatism level as FY t, but
FY tn-l is significantly less conservative than FY t, probably resulting from aggressive

accounting behavior in FY tn-l or changed financial reporting behavior in FY t (more

33

conservatism). Conservatism hits the highest level in FY t, and it declines over the two-

year period following the restatement announcement (significantly negative coefficient of
(111 for FY t+l and t+2).

Overall, the above results demonstrate changes in the financial reporting behavior
of restating firms after the restatement announcement, consistent with my predictions in
Hla and H2a. These firms become more conservative in FY t, and conservatism begins to
decline by the end of FY t+2, possibly due to the elimination of high incentives from
CEO compensation. There is also some evidence of aggressiveness in accounting even
before the restated period.

6.1.2 The Market-Based Measure
In Table 3, I conduct a sirrrilar analysis using the market-based Basu measure of

conservatism. In Panel A, I find that the coefficient on DR x R is significantly positive,

implying that bad news is recognized in a more timely fashion than good news. The Basu
measure shows that restating firms have less conservative financial reports even before
the restated period. However, I do not find that they report more conservatively than all
other non-restating firms after the restatement announcement.

Turning to Table 3 Panel B, my regression reveals that restating firms do not have

different conservatism levels when compared to industry, size and year matched control
group across all the five years, as demonstrated by the statistical insignificance of (17.
Although the insignificance for FY tn-l and tn-2 might be consistent with my expectation,
the result for FY t does not support my main hypothesis.

Table 3 Panel C presents the change in conservatism over time for restating firms

only. I find that both FY tn-2 and FY tn-l are less conservative than FY t (marginally

34

significant). If restating firms are expected to behave normally in these two years, the
result indicates changed financial reporting behavior in FY t. But contrary to my
expectation, it seems that the degrees of conservatism between FY t and the following
two-year period are not different from each other, as reflected by the insignificant

coefficient on T x DR x R in FY t+1 and FY t+2.

In summary, using the market-based Basu model, I do not find similar results as
those under the accrual-based model. In particular, the post-restatement financial
reporting behavior of restating firms does not differ from non-restating firms. In Givoly
et al. (2007), the authors argue that the Basu measure may contain a large measurement
error that prevents the detection of conservatism when it is likely to exist. I believe their
argument is particularly relevant to my tests. The authors argue that certain economic
events which will affect the current period’s returns will never, or only marginally, affect
current earnings. They demonstrate that firms with announcements of class action
lawsuits or SEC investigations have much lower conservatism using the Basu measure
when compared to matched firms during the announcement period. I believe the
restatement setting closely resembles these situations. A restatement announcement
represents an unfavorable economic event. The future effects of this current shock will
likely be immediately impounded in current stock prices. However, current earnings,
even with more conservative changes under the circumstance after the announcements,
will still have limited ability to completely reflect those effects. The fact that earnings do
not respond or only partially respond to this bad news will unduly lead to the inference of
aggressive, or at least non-conservative, accounting. Therefore, I believe that this

measurement error explains the weaker Basu measure results. To the extent that my study

35

aims at measuring the “discretionary” component of conservatism over which the
executives, who produce the financial reports, have control, the accrual-based measure,
which is not subject to the “biased” market impact, might perform a better job than the
Basu measure.

I also conduct further analysis to see if different types of restating firms exhibit
different “changing” behavior in conservatism. In particular, the descriptive statistics of
my sample in chapter 5 show that company-initiated restatements and restatements due to
revenue recognition issue account for a larger portion of my sample of restating firms.
Table 4 presents the results. It shows that the changes in conservatism after the
announcements, for Year t, t+l , t+2, are not different between company prompters and all
the other firms and between revenue-recognition restatements and other types of
restatements.

6.2 Changes in CEO Compensation
6.2.1 Changes in CEO Compensation

In Table 5, I present descriptive statistics on various compensation variables for
restating firms in the pre-restatement (FY tn-l, FY H and FY t-2) and post-restatement
periods (FY t through t+2), where FY t is the fiscal year in which the restatement is
announced. To calculate the changes from one year to another, I delete all the restating
firms which hire a new CEO in either year because the first-year compensation package
of new CEOs contains features (signing bonuses, one-time option awards, etc.) that are
different from CEO compensation during normal years. Panel A compares compensation

components across different time periods. Across all time periods, option grants represent

36

the largest component of CEO compensation. The values of option grants and restricted
stock are highly skewed.

Option grants increase from t-2 to t-l, although the increase is not significant.
There is a significant decrease in bonuses. From t-l to t, option grants experience an
insignificant drop while bonuses increase significantly. When I compare tn-l to t, there is
no significant change in bonuses as well as in option grants. In the two years after FY t,
the amount of option grants does not change significantly, but bonus awards increase
significantly’. Before the announcements, these changes have no impact on CEO total
compensation, which does not change significantly from tn-l to t. But there is a
significant increase in CEO total compensation after the announcements (from t to t+2).
Further analysis reveals that this increase is mainly driven by firms incurring CEO
turnover in FY t+1 (mean increase: $ 2.067 million for CEO-turnover firms versus 0.669
million for no-CEO—turnover firms; z-stat for mean difference test: 2.048).

Table 5 of Panel B presents the percentage change analysis for bonus and option
grants across the different time periods. The percentage number is defined as:

Annual bonus award or option grants (3) x 100

 

Bonus or option% = ,
Total compensation (S)

I then test the mean difference between two periods. The results indicate that
restating firms significantly cut CEO annual bonuses during the two years before FY t,
with significant increases after restatement announcements. On the other hand, option
grants exhibit the opposite trend. Restating firms started to cut the weight of option grants

in FY t-1, and this trend continues during the two years subsequent to the announcements.

 

9 Although this finding on the option grants is not consistent with my hypothesis and prior literature, this
analysis is on the change of the level of the option grants. When I analyze the change of option grants as a
percentage of CEOs’ total compensation, my finding is consistent with my hypothesis and prior literature.
In addition, 1 compare FY t+2 with FY I while prior literature compares FY t+2 with FY t-l.

37

The increase in bonuses and the decrease in option grants after the announcements might
indicate a possible substitution from options to bonus compensation.

Table 5 is largely consistent with findings from prior studies (Chen and Farber
2008; Collins, Reitenga, and Cuevas 2005). More importantly, these statistics show that
restating firms begin to re-contract with their CEOs after the announcements and
significantly reduce their option grants by the end of FY t+2. There is also some evidence
of a substitution effect between bonus and option grants.

6.2.2 The Determinants of the Changes in CEO Stock Option Grants

In chapter 3.2, I argue that having recognized the agency problem associated with
CEO high equity incentives, restating firms will reduce their use of option grants to
mitigate the problem, and their re-contracting process will likely take more than one year.
But the observation from chapter 6.2.1 shows that although the full effect of the
compensation re-contracting cannot be observed until FY t+2, some firms already started
to reduce CEO option grants in FY t. Thus the hypothesized effects are likely to be
stronger at firms who do not reduce their CEO option grants immediately after the
announcements.

In this chapter, I explore the firm characteristics that can potentially explain the
change in CEO option grants for the period from tn-l to t and for the period from t to t+2.
I adopt the approach from Cheng and Farber (2008) and the firm characteristics represent
a set of variables known to influence stock-based compensation from prior literature. The

following OLS regression is used:

38

AOption% = a0 + alAShare _ Own + azAExercisable _ Options + a3AUnarercisabIe _ Options
+ a4ASize + aSAB / M + a6AR & D + a7 ACash _ Constraint + agAEarn _ Constraint
+ agALeverage + aloAIdiosyncratic Risk + a] lACurrent Return + a1 2APast Return
+ a13ACash Compensation + a14ExIantCEO + a15Fraud + Year Dummies + 6‘

The dependent variable, AOption%, is the percentage change of CEO option
grants scaled by annual CEO total compensation from one year to another. All the
independent variables, except ExtantCEO and Fraud, are also measured as the change
from one year to another. Similar to Cheng and Farber (2008) and Core and Guay (1999),
I use the lagged changes of these variables. For example, when analyzing the change in
option grants from FY tn-l to t, the changes of these independent variables are measured
between FY tn-2 to t-l, except as indicated otherwise. Finally, I also include year
dummies to control for year-specific effects. To save space, I relegate the details for the
independent variables to Appendix A.

Table 6 presents the results for the above regression. For the period from tn-l to t,
restating firms that have reduced book-to-market ratio award their CEOs more option
grants, consistent with the prediction that managers of growth firms receive higher equity
incentive. Restating firms with increased leverage award few option grants to their CEOs,
consistent with debtholders’ demand for the reduced need of stock-based compensation. I
also find that during this period restating firms with increased CEO cash compensation
award fewer option grants, demonstrating the substitute effect between CEO cash
compensation and option grants. But contrary to my prediction, restating firms with

increased CEO stock ownership (AS/1ares_Own) award more option grants, while those

39

(4)

with increased past stock return award fewer CEO Option grants. More surprisingly, those
with increased cash constraint award even less CEO option grants during this period”).

For the period fiom t to t+2, the coefficients with a significant sign behave in the
same way as my prediction. Specifically, restating firms with increased CEO stock
ownership (A Unexercisable_0ptions) award less option grants while CEOs of firms who
increased their R&D expenditures (i.e., growth firms) and of firms with increased past
stock returns receive higher option grants.

In summary, from tn-l to t, the behavior of restating firms’ CEO option grants
seems to be quite “irrational”, which might be related to the “transitional” nature of this
period around the announcements. On the other hand, it seems that two years after the
restatement announcements, restating firms with high CEO stock ownerships, less growth
opportunities, and poor performance in past stock returns are more likely to reduce their
CEO option grants.

6.3 Changes in Conservatism and CEO Compensation

Having observed the changes in conservatism and the changes on CEO
compensation before and after restatement announcements separately, I now address the
question of whether the two changes are associated with each other. Due to the likely
measurement error issue in the Basu measure of conservatism (Givoly et al. 2007), I
focus my analysis on the accrual-based measure of conservatism.

6.3.1 Changes in CEO Compensation and the Level of Conservatism

I first examine the relationship between the change in CEO compensation and the

level of accounting conservatism at FY t and FY t+2, respectively. Based on the above

analysis of the change in conservatism and the analysis of the changes in bonus and

 

’0 Cheng and Farber (2008) find similar result on the cash constraint variable for the period tn-l to t-l.

40

option grants over time, FY t is the time when restating firms’ financial reporting
becomes the most conservative. Following the prediction in my hypotheses, this increase
is most likely a result of the high level of option grants utilized during the years leading
up to FY t. On the other hand, by the end of FY t+2, conservatism for restating firms
declines from its peak level in FY t. Given the decreased usage of option grants during
the same period, the reduction in conservatism is most likely a response to the reduced
option awards. In addition to these hypothesized relations, I also examine the impact of
changes in bonuses on the levels of conservatism.

Panel A of Table 7 provides the results for FY t. Column (i) shows that on

average, CFO is negatively related to ACC and the coefficient on DCFO xCFO is

significantly positive, indicating conservative accounting. In column (ii), I find that the
change in option grants from FY tn-l to t is positively associated with the level of
conservatism at FY t, as reflected by the significantly positive coefficient

of AOption%_Rank x DCFO x CFO. Thus an increase in CEO option grants from FY

tn-l to FY t is accompanied by a higher level of accounting conservatism. This is

consistent with Hla. The main conservatism interaction term, DCFO x CFO , is negative,

possibly indicating that for firms with the greatest reduction in option grants (lowest
value of A0ption%_rank), the demand for conservatism to control CEO’s compensation
incentive is low. Surprisingly, in column (iii), 1 find a significant negative association
between bonus increases and conservatism. In other words, firms that reduce bonus
awards the most from tn-l to t also have the highest level of conservatism (which is also

reflected by the significantly positive coefficient of DCFO x CFO). This is inconsistent

with H1.

41

In column (iv), I examine the changes in options and bonus compensation while
incorporating all of the control variables. Both the option and bonus coefficients remain
significant. The result also demonstrates that firms with high leverage ratios exhibit a
higher degree of conservatism, consistent with the demand of conservatism from
debtholders. For control variables, larger firms exhibit a lower degree of conservatism, on
average. This is consistent with the argument that the rich information environment of
large firms reduces the demand for conservatism. Overall, Panel A of Table 7
demonstrates that the high demand for conservatism at FY t is associated with the change
in CEO option grants, indicating that restating firms respond to the high CEO
compensation incentives awarded before the announcements.

Panel B of Table 7 presents the result for FY t+2. The main effect of CFO is

significant as presented in column (i) although the coefficient on DCFO >< CF 0 is not

significant. In column (ii), the change in option grants from t to t+2 has a significant
positive impact on conservatism at FY t+2. The more a restating firm decreased its option
grants, the less conservative its financial reports became at t+2. In column (iii), I do not
find a similar result for the change in bonus. Column (iv) presents the results when the
changes in both compensation variables are included in the same regression. Similar to
the results from previous columns, I find that change in option grants have a positive
(marginally significant) impact on the level of conservatism in FY t+2. The insignificant

coefficient on DCFO x CF 0 indicates that restating firms reduce CEO option grants the

most from t to t+2 exhibit no difference in the recognition of bad news versus good news.

Last, firms facing high litigation risk have a higher level of conservatism. However,

42

contrary to predictions, leverage seems to have a negative impact on conservatism at t+2.
Overall, the main results from Table 7 are consistent with my H2.
6.3.2 Changes in CEO Compensation and Changes of Conservatism

In addition to examining the impact of changing CEO bonus and option incentives
on the level of conservatism, I also investigate whether the change in CEO option grants
is related to the change in conservatism. To address this question, I split my sample firms
into two parts, based on the magnitude of the increase in option grants. The half of my
sample with the lowest increase in option grants (or high reduction in option grants from
tn-l to t) is named “the top 50%”; while the remaining half is named “the bottom 50%”.
Again, my analysis is conducted for two time periods and uses the accrual-based measure.

Panel A of Table 8 provides the results for period tn-l to t where I use FY t as my
benchmark year. Recall that in Chapter 6.1, I find that restating firms significantly
increase conservatism from tn-l to t, indicating changed financial reporting behavior. I
draw similar inferences fiom the same changes specification in the first two columns in
Panel A. The analysis on the two sub-samples is presented in the next columns.
Specifically, for the group with the largest reduction (or smallest increase) in CEO option
awards, I do not find a significant difference in conservatism between tn-l and t.
However, for the bottom 50% firms (the right two columns), conservatism was
significantly higher in FY t than in FY tn-l, as reflected by the significantly negative

coefficient on T x DCFO x CFO. The coefficients on CFO and DCFO x CFO are also

significant at 5% level. This result demonstrates that the increase in conservatism from

tn-l to t is exclusively driven by firms who increase their option grants the most in the

43

same period. When firms reduce high incentives of option grants from tn-l to t, higher
conservatism as a control mechanism is no longer necessary.

Panel B of Table 8 presents the results for the period from FY t to FY t+2.
Similar to the full sample results replicated in the first two columns, I find that for firms
with the largest option decreases, the degree in conservatism is significantly lower in FY
t+2 compared with FY t. But there were no differences in conservatism for the bottom
50% firms between the two years. Therefore, the overall decrease in conservatism from t
to t+2 is mostly driven by restating firms with the largest reductions in option grants. For
firms maintaining or increasing CEO option grants from t to t+2, a high level of
conservatism at FY t is maintained as a control mechanism. '

To summarize, using the accrual-based measure of conservatism, I find strong
evidence of a relationship between changes in CEO option grants and changes in
accounting conservatism. When CEO option grants are very high, resulting in high
incentives, restating firms increase conservatism. When firms begin to reduce option
grants, the accounting conservatism is also reduced. These results, together with those
presented in Table 7, demonstrate that investors demand a higher degree of conservatism
when facing a severe agency problem due to the high CEO option grants. As restating
firms start to alleviate the agency issue between shareholders and management through
re-contracting of CEO compensation, the demand for conservatism from shareholders
starts to decrease.

6.3.3 The Impact of Restatement Characteristics on the Change in Conservatism

Restatement studies have demonstrated that CEO option holdings are not only

related to the propensity of misreporting but also related to the magnitude of the

44

 

mag:

arm 0

eami

restat

COIISL'

 

estin

 

 

 

misreporting. Efendi et. al. (2007) show that CEO bonuses and in-the-money option
holdings are positively associated with income-decreasing restatements. Burns and Kedia
(2006) show that higher values of CEO option sensitivity lead to larger restatements. My
observation of the high degree of conservatism in FY t might also be a response to the
magnitude of restatements, independent of the change in CEO option grants before the
announcements. In other words, large corrections of earnings lead to more conservative
earnings in FY t.

Additionally, as presented in chapter 5, my restatement sample contains both
irregularities and errors. Although Table 6 shows that Fraud is not significantly related to
the changes in CEO option grants, to the extent that an accounting irregularity indicates a
greater agency problem between management and shareholders than other types of
restatements, restating firms with accounting irregularities may become more
conservative in FY t.

To address the impact of the restatement magnitude and the accounting
irregularities on the increase in conservatism in FY t, I incorporate two additional
variables into my analysis for the period from FY tn-l to t. Fraud takes the value of 1 if
the restatement is classified as an irregularity and 0 otherwise. Resize_rank is the rank of
the restatement magnitude, standardized to the interval (0, 1). Restatement magnitude is
estimated as the sum of the effect of the restatement on net income scaled by the sum of
the restated net income over the entire restated period. Similar to my analysis on option
grants, I first examine the association between the two variables and the level of
conservatism for all the restating firms at FY t. I then investigate the impact on the

change of conservatism from FY tn-l to FY t.

45

Table 9 presents the result of the analysis for the level of conservatism at FY t. Of
the 215 restating firms used in Table 7 Panel A, 28 firms are further deleted. These firms
either do not have enough information to determine restatement magnitudes or they
report income-increasing restatements. Therefore, only 187 restating firms are included in
the level analysis. After including the two additional controls, I rerun my regression on
the association between the change in option grants from tn-l to t and the level of
conservatism at FY t. Column (i) shows that the increase in option grants from FY tn-l to
t is associated with higher level of conservatism at FY t, as evidenced by the positive

coefficient of A Opti0n%_ Rank X DCFO X CFO. When I add the fraud indicator to

the regression, the effect of option grant change remains significant. But an accounting
irregularity does not induce more demand for conservatism in FY t (column (ii)). When I
incorporate the restatement size only into my regression (column (iii)), the impact of

options on conservatism goes away. But the coefficient of Resize_RankXDCF OX CFO

is not significant either. In column (iv), I run the full model with all the variables. Both

the coefficients on A Opti0n%_ Rank X DCFO X CFO and Resize _ Rank X DCFO
X CFO are insignificant. However, the coefficient on accounting irregularity is now

significant. Overall, Table 9 shows that the impact of change of CEO option grants on
conservatism in FY t disappears after controlling for the restatement magnitude and the
accounting irregularities. Moreover, it shows that accounting fraud has a dominant effect
in determining the level of conservatism for restating firms in FY t.

Due to the reduced sample size and the large number of independent variables
presented in the level regression, the insignificance of the coefficients of Resize _ Rank

>< DCFO X CFO and A Option%_ Rank X DCFO X CFO might be attributed to

46

multicollinearity. Therefore, a change analysis might provide us additional evidence. In
particular, if a large restatement indicates a higher degree of agency problems,
shareholders might demand more conservative accounting in FY t for high restatement-
size firms than for low restatement-size firms. Similarly, because fraud reveals a severe
agency problem, fraud firms might become more conservative in FY t than non-fraud
firms.

In Table 10, I examine the impact of the two variables on the change in
conservatism from FY tn-l to FY t. Panel A presents the results for fraud firms versus
non-fraud firms, while Panel B gives the results for the restatement magnitude. Similar to
my analysis on the change of option grants, column (i) of Panel A presents the change of
conservatism for the fraud sub-sample while column (ii) presents the results for the non-
fraud sub—sample. Both groups show an increase in the degree of conservatism from FY
tn-l to FY t, but the increase is marginally significant for the fraud firms. Fraud firms are
not more conservative than non-fraud firms. To the extent that option incentives drive
conservatism changes and there is no particularly significant association between change
of option grants and the type of restatements (Table 6), I expect those firms that increase
Option grants fi'om FY tn-l to FY t to be randomly distributed within each sub-group.
These firms thus drive the significant increase in conservatism for both groups. Although
this result contradicts the result in Table 9, Table 10 Panel A seems to be consistent with
my main results in Table 8.

For the impact of the restatement magnitude on the change in conservatism from
FY tn-l to FY t, I rank my sample firms based on the size of the restatements. The top

50% represents the half of my sample fimis with low ranks (i.e., small size of

47

a: II!

restatements) while the bottom 50% represents the half of the sample firms with high
ranks. Panel B of Table 10 shows that for firms with low magnitude of restatements (the
top 50%), their degree of conservatism increases significantly from FY tn-l to FY t
(column (i)). But for those firms with the large restatement sizes (the bottom 50%), there
is no difference in conservatism between FY tn-l and FY t.

This result seems to be inconsistent with the expectation that large magnitudes of
restatements lead to high conservatism after the announcements, however, it might be
consistent with my option incentive argument. In particular, higher option incentives lead
to higher magnitudes of restatements (Burns and Kedia (2006)) and firms with high CEO
option grants before the announcements are more likely to reduce grant size afier the
announcements (Cheng and Farber (2008)). Thus firms with large restatements in my
sample are more likely to be the firms that have high CEO option grants before the
announcements and that significantly reduce option grants at FY t. As already shown in
Table 8, restating firms with a large reduction in option grants do not show a different
degree of conservatism between FY tn-l and FY t. This potentially explains the
insignificance of conservatism between the two years for the group with large
restatements because these firms are more likely to reduce their CEO option grants at FY
t. In fact, further analysis reveals that after I split my sample into two subgroups based on
the size of the restatements (high-restatement group versus low-restatement group), the
high-restatement group exhibits a greater drop in CEO option grants (i.e., lower amounts
of change in CEO option grants) from tn-l to t than that of the low-restatement group (2-

stat: -1.55; one-sided p-value: 0.06).

48

In summary, the results from Table 9 and Table 10 indicate the important role of
the characteristics of restatements in explaining the increase in conservatism from FY tn-
1 to FY t. Although Table 10 seems to be in line with my hypothesis, by incorporating
the change of CEO option grants, the restatement magnitude, and fraud indicator in one
regression, Table 9 presents a more complete picture of the dominant effects of the
characteristics of restatements, especially accounting irregularities, in determining the

level of conservatism for restating firms in FY t.

49

CHAPTER 7: SENSITIVITY ANALYSES
7.1 Change in Conservatism and the “Big Bath” Theory

It has long been documented in accounting and finance research that new CEOs
are likely to boost future earnings at the expense of transition-year earnings by increasing
write-offs and income-reducing accounting accruals — “taking a big bath”. Moore (1973)
points out that through these discretionary accounting decisions, “the reported new
earnings may be blamed on the old management, and the historical bases for future
comparison will be reduced.” More importantly, “future income would be relieved of
these charges, so that improved earnings trends could be reported.” (Moore 1973: 101)
Moore (1973) compares a sample of management change companies with companies
having other personnel changes. He finds that incomes reducing discretionary accounting
decisions are more likely to be made in the year of the management change for his
sample of the management change companies.

Following Moore (1973), a stream of studies provides further support for this
hypothesis. Strong and Meyer (1987) compare a group of firms that record asset
writedowns with a group of non-writedown firms. They find that a change in senior
management is the most important factor observed for the writedown firms. Similarly, in
a sample of large discretionary write-offs examined by Elliott and Shaw (1988), 39
percent of the firms experience a change in the CEO, president, and/or CFO during the
year of the write-offs. Pourciau (1993) argues that the environment surrounding
nonroutine executive changes (i.e., no orderly and well-planned process of turnover)
provides strong incentives and opportunities for earnings management by the new

executives. Based on this argument, she finds that for a sample of firms having

50

experienced nonroutine executive changes, their earnings and accruals decrease
significantly in the year of the executive change and then reverse dramatically in the
following year. Murphy and Zimmerman (1993) provide further support. Through
analyzing a large sample of over 1,000 CEO departures from almost 600 firms, the
authors find that accounting accruals and growth in earnings are significantly lower in the
transition year than in previous years, and the incoming CEO’s first full year is associated
with a pronounced increase in earnings. They further document that when focusing on
cases of nonroutine, inferior departing CEOs, the accruals are significantly negative in
the transition year, although they do not find the dramatic reversal in accruals in the
following year as found in Pourciau (1993). Finally, Geiger and North (2006) document
that firms appointing new CFOs report significantly higher income decreasing
discretionary accruals in the first full reporting year under the new CFOs.

The above discussion demonstrates the “big bath” behavior in firms’ financial
reports immediately after the appointments of top executives. As presented in section
2.2.2, the restatement literature has documented that restating firms experience
significantly higher executive turnover after the announcements. Further analysis of my
sample finds that more than 39% of restating firms appoint new CEOs in the period from
FY t to FY t+2. If these new CEOs use write-offs and accruals to take an “earnings bath”
in the turnover years, then compared with non-restating firms, restating firms will appear
to have a more timely recognition of bad news, resulting in more conservative financial
reports. This potentially explains the observed increase in conservatism in FY t. Although
my analyses of the change in CEO compensation and the association between CEO

option incentives and conservatism already exclude the new CEO firm-year observations,

51

my analysis of the change in conservatism in section 6.1 might be influenced by the “big
ba ” behavior. To address this issue, I re-examine the change in conservatism around
restatement announcements using the accrual model and for each event year regression I
delete all the restating firms with CEO transitions in that year.

Table 10 presents the results. Panel A uses all the other firms in Compustat as the
control group. Compared with Table 2 Panel A, Table 10 Panel A loses about 40 to 60
restating firms across the five-year period. But the regression results are qualitatively
similar to those reported in Table 2 Panel A: restating firms become more conservative in
FY t and FY t+l and then they exhibit a similar conservatism level in FY t+2 when
compared with non-restating firms. Table 10 Panel B is also largely consistent with Table
2 Panel B: restating firms engage in aggressive financial reporting behavior in FY tn-l
while they report more conservatively in FY t and FY t+1. Interestingly and also different
from Table 2 Panel B, in FY t+2, restating firms report significantly less conservatively
than the matched group. This is consistent with “big bath” behavior. Specifically, if
restating firms that appoint new CEOs in FY t+2 are more conservative than other
restating firms due to a “big bath”, including these firms might potentially offset the
aggressive behavior shown in Table 10 Panel B, resulting in an insignificant coefficient

of Restatex DCFO x CFO as reported in Table 2 Panel B. Finally, Panel C of Table 10

is also consistent with Panel C of Table 2, which shows that FY t is the most conservative
year among all the five event years.

Overall, after deleting firms with CEO turnover in each event year, my
observation of the change in conservatism around restatement announcements

documented in section 6.1.1 is largely unchanged qualitatively. Furthermore, there is

52

some evidence supporting “big bath” behavior by firms experiencing CEO turnover,
especially in FY t+2.
7.2 The Relationship between Change in Conservatism and Restating Firms’
Performance

A significant portion of restating firms had very poor financial performance
before the restatement announcements based on my observation, and some of the
restating firms also significantly improved their performance by the end of the second
year after the announcements (e.g., Cheng and Farber 2008). An interesting issue is, thus,
whether the observed increase in conservatism in FY t is a reflection of the restating
firrns’ poor performance and the following decrease in conservatism by FY t+2 is a result
of their performance improvement. In other words, the conservatism measure itself might
be affected by firms’ performance, leading to the observed behavior of change in
conservatism for restating firms.

To address the concern, I first tried to match on performance (i.e. +/- 20% of
ROA) in addition to match on year, industry, and size when comparing the level of
conservatism between restating firms and the matched group of non-restating firms. But
adding performance matching significantly reduces my sample size. To the extent that the
conservatism measure itself might be affected by firms’ performance, I conduct separate
analysis to examine the effect of performance on the measure of conservatism in a
general setting. I investigate the accounting conservatism of all the non-financial firms in

Compustat for the period of 1997-2006. I employ the following model:

53

ACC/7:620 +alCFOj, +a2ASaIesJ-t +a3FASSETJ-t +a4DCFOj, +a5DCFOj, xCFOj,
+a6ROAJ-t +a7ROAj, x CFOj, +a8ROAJ-t xASa/esjt
+a9ROAj, xFASSETjt +a10ROAj, xDCFOj,
+al lROAj, xDCFOj, xCFOj, + ControlVariables +817 (5)

where performance is measured by return on assets (ROA), which is the standardized
decile rank of the firm’s net income before extraordinary items divided by its average
total assets. All of the other variables are the same as the variables defined in Equation (2)
from section 4.1, except that all of the continuous control variables are measured by using
standardized decile rank. All of the control variables interact with the main variables in
the accrual model. If a firm’s performance affects the conservatism measure, I expect

all to be significantly negative - increasing performance reduces conservatism.

My test result, presented in Table 12, does not indicate a significant impact of
ROA on the measure of conservatism. Although this test does not directly attack the
restating firms, it provides some evidence that the conservatism measured by the accrual
model is not a mechanistic reflection of firms’ performance.
7.3 CEO Compensation Incentives

As I noted earlier, prior research shows that the likelihood of a restatement is
increasing in the sensitivity of the value of the CEO stock option holdings. To the extent
that the agency problem between management and shareholders is exaggerated by high
CEO option incentives, the change in CEO option sensitivity might be a better measure to
capture the underlying change in the agency problem. Following Core and Guay (2002)
and Burns and Kedia (2006), I measure option sensitivity as the change in the value of
stock options held for a percentage change in firm value, which is obtained by

multiplying the option delta by 1% of the stock price and the number of options held.

54

 

sens
rhea
char
by r
rank

rehu

Tahl

l)(1
signi
resul
the s

14,:

lm‘es

 

I re-run equation (2) on all of the observations with available data on the option

sensitivity measure. I replace AOption%_Rank with AOptionsen_Rank , which

measures the change in option sensitivity from one year to another. I also include the
change in the sensitivity of CEO equity holdings - AEquitysen_Rank , which is obtained
by multiplying the number of shares of equity held by 1% of the stock price. Both are
ranked and standardized to the interval (0, 1). Both variables are expected to be positively
related to the level of conservatism in the event years tested. The results are presented in
Table 13.11

Column (i) of Table 13 presents the result for FY t. Although the main effect of
DCFOx CFO is significantly positive, none of the sensitivity variables have a
significant impact on conservatism. Column (ii) shows similar result for FY t+2. These
results are inconsistent with my hypotheses, but should be interpreted cautiously due to
the small sample sizes in each year.
7.4 Analysis of Discretionary Accruals around Restatement Announcements

In addition to restating firms’ conservatism changes, I also conduct analysis to
investigate restating firms’ discretionary accrual changes around the announcements. I
adopt the methodology similar to Ball and Shivakumar (2008) in which the authors
examine the changes of both accounting conservatism and discretionary accruals for UK
IPO firms. Normal accruals are estimated using Jones model modified by Ball and

Shivakumar (2006):

 

” I exclude bonuses in my analysis due to its weak impact on conservatism as observed in Table 7. I only
include the three-way interaction term of each control variables in my regression to increase the degree of
freedom for my regression. Including bonuses and all the complete interaction terms of the control
variables does not change my conclusion.

55

ACCit=aj0 + aleFOit + ajzASalesi, + aj3FASSETit + aj4DCFOl-t +
aj5DCF0i, x CFO“ + an (6)
Variables are as defined in Chapter 4. Model parameters are estimated separately
for each restating firm j from a cross-section of all non-restating firms 1' in its 2-digit SIC
with data for year I. Only industry-years with at least 10 observations are considered. 1%
on both extremes of each continuous variable are trimmed.

Abnormal accruals ABN _ ACC}, for restating firm j in year t are computed as the

difference between. the actual accruals and estimated normal accruals (A denotes
estimates):
ABN_ACCj-, = ACCj,—[a‘zj0 + dleFOj, + djzASaIesj, + dj3FASSETj,
+dj4DCF0j, +éj5DCFOJ-t xCFOj,] (7)

Table 14 presents summary statistics for restating firms for event years tn-2, tn-l,
t-2, H, t, t+1, and t+2. In FY tn-2 and tn-l , both mean and median abnormal accruals are
positive and statistically significant (1% to 2% of total assets), indicating that restating
firms already engage in earnings management activity even before the manipulated
period (but within GAAP). Abnormal accruals become statistically insignificant in FY t
and marginally significant in FY t+1 (the mean is negative in each year), demonstrating a
reduced level of earnings management immediately after the announcements. However,
by the end of the second year after the announcements, it seems that restating firms
resume their old earnings management behavior, with significantly positive median
abnormal accruals.

I also examine the abnormal accruals in FY t-2 and t-l. It seems that restating
firms engage in significant accrual management activity in t-2 but do not manage accrual

in t-1. These results, however, should be viewed cautiously because data originally

56

reported by restating firms in t-2 and H might be “contaminated” by restated data
updated through Compustat”.

My above finding for FY t and t+1 is consistent with Badertscher (2007) in which
the author shows that restating firms significantly reduce their accrual management
activity in the first year after the announcement. But more importantly, Table 14 also fits
well with my finding regarding the change in conservatism that restating firms are less
conservative in FY tn-l, become more conservative immediately after the announcements,

and then reduce conservatism in the two years afterwards.

 

’2 Based on discussions with Compustat, I learned that restated data is recorded in Restated data items, and
the original data is never changed. But my analysis finds that this policy is not consistently followed by
Compustat. For example, Net Income (data172) should record the original net income number as first
reported and subsequent restatements should be recorded into Net Income-Restated (Data 177). I randomly
picked 20 restating firms from my sample. I compared Compustat’s net income data during restated years
with the net income numbers in the lOKs and lOK/As filed through Edgar. I found that in 7 of 20 cases in
which a IOK/A is filed for a particular year, Compustat replaces data] 72 with the restated number for that
year. But in the 13 cases in which no 10K/As were filed (i.e., a firm’s 2001 restatement is reported through
its 2002 10K), Compustat changes Data] 77, not Data] 72 for 11 cases, and does not make any changes for 2
cases.

57

CHAPTER 8: CONCLUSION

In this paper, I investigate whether restating firms improve their financial
reporting quality after restatement announcements by examining one important
accounting quality dimension — conservatism — which assumes the role of a control
mechanism to curb managements’ opportunistic behavior. If restatements indicate
aggressive reporting behavior which implies a severe agency problem between
management and shareholders, I expect an increase in conservatism after a restatement
announcement. On the other hand, as restating firms take actions to address the agency
problem during the years after the announcements, I expect a subsequent decrease in
conservatism after the initial increase.

I further explore the change in conservatism around restatement announcements
by directly linking the change to CEOs’ compensation incentives. Empirical studies have
documented that high incentives from CEO option grants exaggerate the agency problem,
leading to aggressive accounting behavior, and restating firms significantly reduce their
CEO option grants two years after the announcements. Thus I expect cross-sectional
differences in the change in conservatism based on the changes in CEO option grants.
Changes in conservatism should be positively associated with the changes in CEO option
grants.

1 adopt two measures of conservatism: an accrual-based measure and a market-
based measure. Both measures build on the concept of earnings’ asymmetric timeliness in
bad news recognition. Using the accrual-based model, I find that restating firms become
more conservative in the announcement year, demonstrating a change in financial

reporting behavior. There is also evidence of decreased conservatism by the end of the

58

second year after a restatement announcement. When I incorporate changes in CEO
annual compensation into my analysis, I find that the increase in conservatism in the
restatement announcement year is significantly associated with the increased usage of
option grants during the prior period. After the announcement, the decrease in
conservatism is significantly related to the reduction in CEO option grants during the
following two years. In other words, restating firms, on average, started to reduce option
usage so as to reduce the undesirable incentive effects and address the agency problem.
Thus, as the severe agency problem is alleviated, investors’ demand for conservatism as a
control mechanism declines.

In addition to CEO option incentives, I also examine the impact of restatement
characteristics on the initial increase in conservatism. Specifically, I consider the
restatement magnitude and the presence of accounting irregularities. Large restatements
or an indication of accounting fraud may demonstrate a more severe agency problem to
shareholders, leading to an increased demand for conservative financial reporting
immediately after the announcements. Using the change specification, I do not find that
the increase in conservatism from FY tn-l to FY t is driven by firms with high
restatement magnitudes or by fraud firms. But when analyzing the level of conservatism
for FY t, the significant relationship between the change of CEO option grants and
conservatism disappears after controlling for the restatement magnitude and the
accounting irregularities. Interestingly, I find that fraud firms have a significantly higher
level of conservatism than non-fraud firms in FY t. Overall, these findings demonstrate
the important role of the characteristics of restatements in explaining restating firms’

financial reporting behavior immediately following the announcements.

59

My paper makes the following contributions. First, with respect to the
conservatism literature, the argument that there is a demand for conservatism arising
from compensation contracting is well established but empirical studies on this issue have
been lacking. By using the restatement setting, I can examine how the demand for
conservatism changes when there is a change in the agency problem between
management and shareholders. Second, most of the restatement studies that examine the
changes of restating firms’ financial reporting behavior fail to investigate the factors that
explain these changes. I not only examine a change in conservatism but also document an
association between the change and CEO compensation incentives, a major factor that
leads to restatements.

My results suggest several additional issues for further research. First, I find fraud
firms have a significantly higher level of conservatism than non-fraud firms in FY t. But
using the change specification, I find both fraud firms and non-fraud firms significantly
increase their accounting conservatism from FY tn-l to FY t. This finding needs to be
addressed in the future, and more analyses are necessary to fully evaluate the impact of
restatement characteristics on the change in accounting conservatism after the
announcements.

When I incorporate the restatement magnitude, the fraud indicator, and the change
in CEO option grants into the regression of the level of conservatism in FY t, the reduced
Sample size and the large number of independent variables and their interactions might
Create multicollinearity, potentially explaining the insignificant impact of the restatement
magnitude and the change in CEO option grants on conservatism. A larger sample size

Will increase the power of the regression analysis. On the other hand, a separate

60

 

ime
char
\th
COIli

fun}

flor
incl
con‘

(TEC

 

 

investigation can be conducted on the association between these restatement
characteristics and the change of CEO option grants. Such analysis will help to determine
whether the observed association between CEO option incentives and accounting
conservatism is influenced by the restatement characteristics. Moreover, this analysis will
fiirther our understanding of how CEO option grants change in the restatement setting.

Second, this dissertation does not consider the possible impact of corporate

governance on the change in accounting conservatism after restatement announcements.
Although prior literature does not find an association between corporate governance and
accounting conservatism for restating firms and there is still no strong empirical support
for the argument that poor corporate governance leads to earnings restatements, the
inclusion of corporate governance variables would provide a more complete and
convincing picture of the change in accounting conservatism and its association with
CEOs’ compensation contracts.

Finally, my results reveal that some restating firms reduce their CEO option
grants immediately afier the announcements, while other firms take more than a year to
re-contract with their CEOs. Further analysis could investigate whether the CEO option
incentives during the restated period predict firms’ financial reporting behavior and
Compensation contracting behavior after the announcements. In particular, restating firms
With relatively high CEO option incentives before the announcements might reduce their
CEO option grants immediately after the announcements, resulting in the unchanged
leVel of accounting conservatism in FY t; while restating firms with relatively low CEO

ODtion incentives might increase accounting conservatism first and re-contract with their

61

CEOs later. This analysis will also enrich the literature on the interaction between firms’

compensation contracts and their financial reporting choices.

62

 

2068:5055 EoEBSmom

Em Em Em Em Em Em Em
N: E E E 5m 3 E N-“ E 75 E as E

 

 

‘ _-----fi----fi-----m

/ x
<

wotom 3866M

 

 

 

.m So» Home
.6 cam 05 8 mmmcom cam m >m 53:8 wcugooow 958%? E wEmeco Ema mega“: 2t 8 Sta So» “we of 3 TE >m macaw ~
.35585058 “580832 2: co SPA Roma 2: m_ H >m 3088328,“ 3060338 a @588 who» ago 2: $8.32: Rama mutt

55952:: 30> =83—
“ oaswfi

 

63

mucus—852:8 8080882 23:2: 5;» WEE Sc 825882 “88— 2: 3:0 coax _ ._.

 

 

 

 

emm oEEmm Ems—0882 82E
NmE mm 3:: 868:5
_m 950.“ 825852 82 oZ
mmm ._. 3528882 29:2: :23 ESE
SA: mEouaoxm 80¢ 88 583388 802:3 EEK
no. 2388mm cowcoz mammsmEoUmeU 62m 88 868:: Emma 8923 ESE
amou—
ow: 33-32 coca 2: E O<O com $580882 co 238:2
“85 298m cocoEmom

 

:o_8_=o:ooom 298m n< 88m

2.25 2.25883— ..o 83.585 953.389 98 5:859:83— oEEam
— 038,—.

64

 

 

abode .XFMN $92 $53 $82 Xod fiend Rome .x._.m 800:0 X
emm vw mm. am we Nm mm 2 _ fl s
88H ~meow moon NOON _oom ooom 32 M32 baa—

 

uwotf «EOEOOCSOGC< afloEouwumoM—

 

0.808058% no 0055285 380% ”m 8.8m

€0.53 a 03.8

65

Table 1 (cont'd)

Panel C: Restatement Characteristics

 

 

Percentage
Frequency of total
Full sample 354 100%
Initiated by*
Company 138 38.9%
SEC 53 15.1%
Auditor 13 3.7%
FASB 4 1.0%
External/Media l 0.3%
Unknown 145 40.9%
Exchange Listing*
NYSE 183 51.7%
Nasdaq 156 44.0%
AMEX 15 4.4%
Reason "‘
Revenue recognitio 152 43.0%
Cost or expense 61 17.1%
Restructuring 56 15.8%
M&A 20 5.7%
In-process R&D 15 4.4%
Securities related 13 3.7%
Other 37 10.4%
Fraud or not
Fraud 120 33.9%
Not fraud 234 66.1%

 

* information is per GAO (2002 and 2006) report.

66

Table 1 (cont'd)

Panel D: Number of Years Restated and Size of Restatements

 

 

Size of restatement Size / Net income

Number of (non-income (non-income

years restated increasing) in 000's increasing) in %
Q1 1 14.67 0.15
Mean 2.15 95235.40 42.48
Median 2 2334.67 6.67
Q3 3 8738.00 38.19
N 354 296 296

 

*one extreme observation was deleted
for the above two columns.

67

Table 1 (cont'd)

Panel B: Industry Classification of Restatement Firms

 

 

 

2-Digit % of

Industry SIC code 11 % Execucomp*
Metal and Coal 10-12 6 1.7% 0.8%
Oil and Gas 13 12 3.4% 3.3%
Food Products 20 9 2.5% 2.7%
Paper and Paper Products 26-27 13 3.7% 3.8%
Chemical Products 28 24 6.8% 7.7%
Manufacturing 30-34 12 3.4% 5.8%
Computer Equipment and Services 35, 73 90 25.4% 18.3%
Electronic Equipment 36 23 6.5% 9.4%
Transportation 37, 39, 42 11 3.1% 4.8%
Scientific Instruments 38 21 5.9% 6.2%
Communications 48 12 3.4% 2.5%
Electric, Gas, Sanitary Service 49 29 8.2% 6.6%
Durable Goods 50 9 2.5% 2.4%
Nondurable Goods - wholesale 51 9 2.5% 1.1%
Eating and drinking 58 8 2.3% 2.2%
Retail 53-57, 59 29 8.2% 7.1%
Entertainment Services 78, 79 5 1.4% 1.0%
Health 80 6 l .7% 1.9%
Professional Services 87 6 1.7% 1.5%
All Others All others 20 5.6% 10.8%
Total 354 100.0% 100.0%

 

* Execucomp industry distribution is from fiscal year 2002.

68

Panel F: Restating Firrns' Financial Data in the Restatement Announcement Year

Table l (cont'd)

 

 

Variable Mean Std. Dev. Q1 Median Q3
Market Value ($ million) 5,335.1 13,328.3 323.7 867.2 3,664.0
Total Assets (S million) 5,533.1 11,876.2 423.3 1,257.7 4,437.4
Sales ($ million) 4,411.6 8,829.1 430.5 1,291.9 3,935.5
Book Value ($ million) 1,692.2 5,081.9 157.9 409.4 1,295.7
Net Income ($ million) -170.4 3,304.9 -33.4 20.7 94.8
Book-to-Market Ratio 0.492 1.019 0.252 0.460 0.729
Leverage 0.232 0.214 0.024 0.213 0.360

 

69

20:0: .8 _0>0_ $0: 05 8 008050me 8808:: 20m 0022.2 08 85:88-: 8320088200 8:2

Em 5:23 28 36:802—0208: 08280 :80 8 .XL .3 0082:: 08 :08 3883 0:88 809 @5880: an: @0800 08 838:3 8005800

=< 8:03:08 <03): 2805030 80300000 8:: 080% 080.: 8 :03 8 £28082 0:: 82.: 0002000 08 8:: 8:080: 0:0 =< 80> 80>0 0:: x:
02030 00:00 08: 028082 0:: mats: £00.: 8:882 0:: 8:. 20:00 080.: 8809000 0:: :8 88800 :0_:3 20:80 _2800 8 88 80> E0>0
:80 20.: 308802 008058 0: :0005 0: H .88.: 8882 20.: 008065 8 0: 033.0% 08 ”2:300:00 o 08 ov 0&0 : _ 008> 0:0 8:8 DRUG
A883 3008 0000.: m0 008.» :00: m: Hmmm/E x2083 0080 5 88:0 0: 88% A3883 0808088 30: :80 E0: 30: :80 @5280

0: 9:0 $808088 30: :80 82.: :20: A3883 8282000 80:, 30c :80 858 32883 80:80 8 000:0: 082000 :39 0: 00¢. 20:3

_. _. a a
0. 0+ H.OmUx 0. OmUQx 0. 0880M: :0+

_. _. a _.
0. OmUQ x 0. 08:03:20 + rummgsm x 0. 0883—00 +

a. 008m< x a. 0880de + H.aOnmmv x n. 088310 + an 0880M©0 +

a a a a a a q
a. 0.5.. a. 0.50? + .. 0505+ 02.88%: a. 888$. + a. 805+ can .. 00<

0:20 :2800 8 8:89:00 E 08:”: 20:00 05 :< was: 3. :08;

.0002 80.53.. aim: 80080000053. «00508801 0000.8 EFF: $550.3: 20.: 80.820800 :0 ”050,—.
N 030,—.

70

 

 

 

 

 

 

 

 

N050 300.. 00 _ 0.. w _ 000 0300 80.8?830 8.0. ..0 02

0.00m mmnmv 0mg? N0000 £00m 08035....— w=_.0.m0..u=0= ..0 .02

0M mm 30 mv 0.0 ~HOmUA— 50$ again—aw. $50.00.. ..0 .02

MR 8m 5 _ m 0% 3. m 280.5... 05.58. .0 dz

$00.0: $9.0: A0000.0: $3.: .000...— . 00.8000;—

0w.0 0N0 0N... 50.0 No.0” :0 w0.0- 0nd- 0v.0 0 : .0 OhU¥OmUG00§80-
0.0.0 No.0 090 No.0 wwd mod whd mod mm: 00.0 OhUn—«0.a.m0~—
00.0 .00 Ed 00.0 28 m0.0 00.0- _0.0- mm.~ 0.0.0 hmmw<000880m
00.. V0.0 3.0 5.0 wvd- No.0- Nwd- No.0- 8.0- 00.0 0289.03.00:
mwd no.0 mmd 00.0 00.0- mod- 0.0.: 00.0 m0. T 0: .0- 90.00.30.001
3.0- 5.0- 00.7 m0.0- 0v. _ - no.0- _m .0 00.0 m _ .0 00.0 0.0.8.:
5.00 00.0 3.... 00.0 00:0... 00.0 .m.0m 00.0 00.0w 00.0 OmUXE—UG
00.0- No.0- ~06- No.0- @50- Nod- N0.m- :00- 006- :00- Oh Un—
00.~—- mod- men- M00- 3.:- m0.0- m0.0—- 8.0- 0007 8.0- Pmmmafia
0nd. 00.0 0%: 00.0 5.0— 000 0043 .50 0m.0~ mod 8.00.4
3.00. 3.0- v0.0a. 30.0- 00.—.1 3.0- 0v.~m- _ m0- 00.0-m- Nw .0- 6.00
3.0- N0.0- meg.- .00- N00- :00- mm...- _0.0- 0m.m- 00.0 .0020...—
.0.m-. _ .....000 8.0-. fl .....000 8.0-. _ .....000 8.0-. _ .....000 8.0-. A .....000 0.00.05»

n+. >...— —+. >m . >m 7:. >-.— Nu... >0—
.80> .003.—

 

 

 

 

05=80 N as;

6.88. < _250

71

3:3 8 _o>o_ $3 2:
Ha oocmoficwi 38065 2.5 63.88 8m 85:33 3.59720320on 328 85 55:5 95 bfizmmwov—megoz 682% 2a _ 25% 02836
mkoou HEB 3283890 2: =< dour/tom «3.2 38$ch E0383: 8.: m8?» mEE mm :03 3 £258on 2: 8on 89:88 2m WEE
30:.ch 2: =< 53» 2:3 2: macaw «Em .33 98.5 ”5.38.. 2: .3 <1: .\+ :23: 98 mg: madame 2: mm 5.5255 253 05 E 3.5
35:8 :03? 9.on 35:00 a wig SPA :53 :03 H8 32888 368:3 E .255 2:. .< .053 N 2an E cosmos 8m why—amt? 2: =m 82—3

_. q a _.
H. w+ H.0mUx H.0mUDx H. gummy—H ~Ho+

a H H H
H. OHHUQ x H. 958%; + H. Emm< mx H. 898m? +

rs mflmmd x .m 0233de + H.aOnHU x Hm Bfimomhd + Hm Bfimodod +

a H. a a a a H
H. 95 x H. 80am: + H. Omuovu + H. Emmfima + H. mammqma + H. 05:. + can H. ou<

380 .8280 3:822 .35 28 .4339: .-Eo> was: ”m _ocmm

25:03 N 03.;

72

 

 

 

 

 

 

 

 

 

 

 

 

mmv _ 3 3m mmm o _ m 22.3.53: .33 3 62
5mm owm 5v Sm 5m 2835:.— 3:58.75: .3 62
mm Hm mm vm vm Tuck—D: 5; 9.33.5: 3538.. .8 dz
2 _ Tm. Hum _ _m H mm_ 2.3%—FE $558.. .3 dz
Xmflfi $02: Hme _ .wm $3; $3 .m nouns—5-:
cmd- mod- chm wmd wa.~ mm .N 2:- mod- 36 N_ .m OmU*O...—U:*85mom
and 86 _ _d 56 $.N mm .o :3 vod Sn.— 2 .o nah—09.238:
omd- mod- mmd- 8.0- nod 36 2 .o- cod 5.: wed hmmwflffiauom
E: mmd _m. _ - mod- on; wmd mmd- mod- ma; cmd 833%.88:
3.7 mmd- mod :6 de 56 mm; id mm._- omd- OmU«oHaH8~_
omd mod mud mod 33. mod- cod cod cod- 85. 358:
3% EH 2:. 93 mm; 36 _m._ 23 wad mod CRYOmUD
and- mod- mw.~.. o _ .o- :3. cod- 3.7 mod- 3.? E .o- Onon—
am; mod 36 mod mo. T mod- 5.7 86. S3. Ed- Pmmms—
w _ . T nod- 92 mod 96. mod- mm; mod mo. T n _ .o- 8.34
EA- 2 .o- 331 $6. HR. 36. 5.? and- Q3. Ed- OmD
ems“. mod- mw.m- mod. .3. T wed- 3. 7 mod- NV. _ - 86. 33.3.:—
233 flaws ~83 _ H38 .33 _ .208 :33 — 608 .33 _ ”:25 933:;

mi >...._ _+« >m a >...— 75 >m Nu... >m

23> .55—

GHHcoov m _ozmm

3.253 N 033—.

73

.553 8 _o>o_ $2 of 3 8:8me6 82865 San 58852 2.5 8:383 HmzpoH-b_o_Hmmvov_moHBoz 52205 05

_. 55% 858me5 £550 :53 22225530 05 =< 502285 <53; chomei £03855: 85 38% EE 3 :03 mm 5238?: 2: EB“ 525253
85 Etc .325ch 2: =< .855 0326.3 96: £023 mEE @5858 2.: :5 wEm: SPA H526 some 58 33838 539553 fl 359: 0:... .< 353
N 2an E 52605 25 335:? 650 05 =< .8?» :63 2: 85¢ fl 5539.530 :5 (t H .«o 2:? 05 megs 52:3 035:; 558:5 a E 5 25:3

H.H H.H . HH HH :
5+ 050x .8on h 5+
HH H H.H HH H H_. H H
0.505.. H. FoHHH+H .Emmfix .:oa+.8_mm<x H $51.05.. H. 955+ H 955+
H.H Ha m Ha v Ha m Ha N H.H H o H.H

OmUx .OmUD 5+ .On—UQ 5+ .Hmmmxxm 5+ .mflmmd 5+ OLD 5+ 5H UU<
mfibm wEHBmom 55:3 Emcmfiomcov E owcmau ”U _ocmm

3.253 N 935,—.

74

 

 

 

 

 

 

 

 

 

 

 

com omo mg 30 2.3%—FE 9558.. .33 .3 .cZ
Xvosm gomwm goodw RENE“ 3:53-:
:6- mmd- awn- omN- 3.0. 3.4m- mm. _ - ww. _- Omuamvnrr
3.8. 36. 5d- mmd- a. T o _ .o- 3.”- 3d. Oboe;
Ne. _- nod- owd- mod- 3.” mod- 36. mod- Hmmm<5P
$4 N _ .o w fl ._ 3d xv; mod c—.~ E .o 8.85;.
SW.— Zwo _m.N mvd afin wmd o _ .o- 8.0. 0”.“er
86 mod cmd- Ed. 5d mod nmd 86 .F
no.9 wwm 86 $6. 86 3.». no... de Oh—Ufivmun
bad mad afin mmd wm.~ mmd wmfi mmd Onon—
:._ mod 5: mod :4 med 5: mod hamms—
mod. 86. mod. 86. mod. .86. mod. mod- 8.24
321 :d- 3&1 K6- 3&1 :d- 3.7 :6- 0&0
mod. 23. mod- vod- mod. 36. wad- wed- 2.3.8::
:33 _ £25 .33 F H33 :3?“ _ H33 .33 _ H38 933...;

N+~ T3 75 Ni:

O .I. .38» aha—5.23m: 53> ~55—

??on U 3ch

8.253 N as;

75

.653 8 .05. $2 2: 8 852.:ch 3:865 3cm 68.5%: 2m 85:83 “mane“-cocflotoo 3:8 85 55:5 wan bficmmcoxmeouoz
2:838 some an XL .3 BEES 2m 83.2“? 952528 =< . 8:338 «322 EmoEcw_m E9525: 85 En?» acre 3 :03 mm .coamewm:
2: 80¢ cows—88 Em 2:5 328:: 2: =< use» Eo>o 05 .3 @2260 totem 2:: 3:00am“: 2: wctsv m8...“ 9583; 2: 52: $50
v.85 863800 2: =m 35:8 5:?» 29:3 75:00 a main So» Eo>o :28 SM 38883 888:3 E 688 2C. .mEE wccflme .5...
880:2: mm mm 888% was UflEofio o 98 ov k c: _ 2:? 05 mow—S «Q 65-39» Bow: 2: “Bum 3255 m 9 95-28» _momc 2: 283
3:58 0 80¢ 538 :85 22-3999 2: m_ % x3 _SmE SPA 38¢ 05 mo chEwon 2: 8 22m Ba 023 :86 2: fl TE AmmSmE
wEwcfiflso 3:23 5868 :38 3 @0236 $383 $5: meEombxo 883 2:85 3: ma venue—u 29% Ba 3553 2: 2 m 80:3

H.an..+ “mm x :de “a Bflmomnd+
:. MD x H.H uufimomcd + ram x a. BfiwoMmd+ rm ouSmodv5+
a .2 n a I _. _.

9.80 3550 3 663800 E mafia $50 2: :< wEmD u< ESE

.252 :25 uni: 3:0E3:=c==< «588.38% 2:5: «Eh...— wfiafimuy— .5. Emu-3.32:0 .3 wfiamoh
m 03.;

76

 

 

 

 

 

 

 

 

 

 

 

 

3.283 m 23h

mchm mnoov flomv o _ 3m 3me 2.23.233 .88 .3 .32

3me $53 82v 933 2: mm 2:335 “53.3.73: .3 .32

m3 2: mm _ w: 3. # Fun .53 2.333;... 9553.. .3 52

8m mom 2% gm 2 m 233:... “5.38. 3 52

.396 go? $3.0 $3 $3.» 8333.

n _ .o mod mmd- mod- _ _.o- Ed- and.- nmd- mafi- mmd- 1*mn—333z

36. Ed- 3.6 56 Ed mod ood- cod cod :3 wan—$3.3:

amd- :5. mad :3 mud 33 3.9 vod 86 mod 1.3.53:

E..— 86 mmd 5d 2: 53 am; 3.0 3; .86 3581

3.3 8.6 ma.~n mvd 3.0m $6 3.2” $6 cmdm Nvd mL—n

mm; Ed vfim mod am; So mc.~ _od cad _od . m:

5.2.. cod- 2.2- vod- 3.5- vod- 3.2- 36- 23.. mod- m

vm... 56 an... 55 3mm 36 m2. Nod 8.x mod 33.33—

33... _ H33 .83 F H33 .33 — H33 33-. _ H33 33-. _ H33 333...;
«+3 >m :. >..._ a >h TE >m «-5 >m

33> 33m
8.303 < 33m

77

.339 3 33— $3 2: 3 3:35ch 33035
2.5 3839: 03 3333-. 338-333.:3 313 8.5 5:35 33 923333—3332 .3333 03 _ 0253 3:333 M300 53,
933233 2: =< 32353 <32 33an3 30333: 35 m3?» WEE 3 :03 3 .3333“: 05 Got 33388 03 mafia 33:55 2:
=< v.3» 2:3 05 waist on? 323 $8....“ 3:333 2: .3 3.: 4+ 553» 23 BE: 9:338 05 m3 333:: 2:3 2: E £36 3333
323 3on 3.33 3 mag 39A :33 33 3% b03233 33:53 2 33:. 05. .< 353 m 2me E 35.33 83 mpg—33> 05 :3 03:3

3w+ 3M x 3&on H.HBvaMhd+
3 MD x 3.3 283M05+ 3d x 3.3 853mm“. + t. oafimoMVd+
f? Adam: “$599+ f}... can 7:333

@380 .8.on 3332 -ofim E3 .4033:— r30> chD “m. 33.—

APHEBV n 033,—.

78

 

 

 

 

 

 

 

 

 

 

 

 

2v mmm w _ m m _ m mdm 3323.82: _83 a: 62

an as E m: mom raga... usage-5:. .s .2

M? S 8 we 6 Ta: .5: $35.: “.558. s 52

2 d E o: o: E 2.25.5... “5.58. 3 .2

$8.— 33 .m 331 $3 d A.\..cdd nausea-m

dmd mdd 8d wmd wwd :4 mod- mod- 3d- Ndd- m-mn-SSRE

_dd ddd _md vod mmd o _ d mm; odd 5d- wdd- yum-.3881

w d d- Sd- dmd- 5.? dm. T de- a: mod mm; mod mEuSmom

Ed odd 5.? ddd mmd mdd 3d- 8d- mmd- odd 888m

vcfi wmd mm; wdd No.7 2.7 ddé Ed mwd 2d ~79:—

wmd mod Ed- mdd- wad- nod- be. _- mdd- mod add a:

vwd- mdd- Ed 5d mad 5 _ d 25- Vdd- mud- 5.? ~—

Nod- mdd- dwd- mdd- Ed 5d 33 mod 3.». vod 2.3.8.:—

Eum-a fibocu «Sm-u _ H33 .33 _ H33 .33 _ £98 :33 _ £98 033:;
Ni >n— —+u >m a E 75 >h N-E >m

23> 33m
9253 m 653

€283 n 2...:-

79

.5on ho _o>o_ $2 2: 8 oocwomeG 3:865 Eom detoue 8a 859%

A $38- bfitwmdoxmobaoz flea—ow 2w ~ 03% 855% mkoou 5?» 3033.830 05 =< .moEZHom «3%: EmoEcwE
5958:: 85 an?» 8.5 mm :03 mm eBay-Ewe 2: 50¢ down—88 Ea mama _Eocmcc 2: =< Sad ozngm «>2

:28: 9:5 958%: 2: =m wEm: SPA :55 some “8 32838 38:53 E 558 2:- .< _ocmm m 2an E 3:ch 2m
833:3 $50 2: =< .30» E26 2: Sod fl comatose cm ,2 _ do 02? on. moxfi 2023 mini? 53:56 m fl 5 80:3

H.nm..+ Hmax “TmQx “mu-55+

H.H a

re Hmd+ H939+

Bed + ram x
:

f5 x

as \Q

.MX .MQMG-T MDNU-T “fix—5+OGH ~|u_.n_\u_.m—

ESE wcssmom 5:23 Emsmtumcou E amigo ”U _ocmm

€233 n 2.3.

80

 

 

 

 

 

 

 

 

 

 

 

an wwm Sn 9% eaghflésme .33 1... 52
.xnng Xmm; fiwmdd RAVE 69:53;—
vod Ed mud mmd 3.7 Ed- n.2- Rd- wry—9L.
_wd omd dwd omd dad- vod- NWT vod- mag-F
mud- ~add. mmd- _dd- mNd ddd 2d ddd mi.
a: mdd mod- ddd mmd 3d wwd :3 h
2 d vod w _ d 3d mbd ovd wed wvd yr»:—
E-d- vmd- wnd- mmd- an; odd mm; odd m:
3d- 3d. mwd- Ndd- mmd- _dd- vod odd ~—
54 _dd mm; mdd hm.~ mdd Sta Nod 33.3.:—
ufim-u _ €93 tam-«J H33 .83 _ 508 :3... _ £03 «33:5

N+~ ~+u 75 «:5

O n 53> 3.3.5.235 39> 2.23m

6.283 U _uawm

5.55 m 2...?

81

Table 4

Test of the Change in Conservatism within Restating Firms

Panel A: Revenue Recognition Restatement vs. All the Other Restatements

ACCJ-t =a0 + aICFO 't + azASales

J

x CFO 't + a8RevRestate

J
J

jt
+ a6RevRestate
t x ASales

x F ASSET 't + a1 ORevRestate 't x DCFO-

+ (l3FASSETjt + a4DCFO '

J

jt + a7RevRestate

+ a9RevRestate t

_]t

Jt

)1

+ a1 lRevRestate jt x DCFOJ»t x CFOJ-t + gjt

where RevRestate is an indicator for restating firms engaging in revenue recognition
restatement. All the other variables are defined in Table 2. The model is estimated separately
for each event year using a control sample which contains all the restating firms other than
the revenue-recognition restating firms during the respective time period covered by the event
year. All the financial firms are excluded from the regression, as well as firms years that
underwent significant M&A activities. Observations with Cook's distance above 1 are deleted
as outliners. Heteroskedasticity and within firm serial correlation-robust t-statistics are
reported. Bold indicates significance at the 10% level or better.

 

 

 

 

 

 

 

 

Event Year
FY t FY t+1 FY t+2

Variable coeff. [ t—stat coeff. I t-stat coeff. I t-stat
Intercept -0.03 -2.24 -0.04 -2.03 -0.04 -2.50
CFO -0.57 -7.55 -0.46 -4.79 -0.34 -4.68
Asales 0.10 3.11 0.08 2.03 0.11 2.14
FASSET 0.01 0.61 0.02 1.25 -0.01 -0.66
DCFO 0.00 0.00 0.01 0.22 -0.07 -1.45
DCFO*CFO 1.21 1.85 0.86 2.63 0.26 1.66
RevRestate -0.01 -0.62 0.02 0.58 0.03 1.17
RevRestate*CFO 0.34 1.96 0.1 l 0.91 -0.06 -0.48
RevRestate*Asales -0.04 -0.78 0.01 0. l 9 0.00 0.01
RevRestate*FASSET -0.03' -0.82 -0.02 -1.06 -0.02 -0.78
RevRestate*DCFO 0.04 0.51 -0.04 -0.86 -0.01 -0.16
RevRestate*DCFO*CFO 0.66 0.89 -0.41 -1.02 -0.47 -0.61
R-squared 28.10% 17.62% 13.36%

No. of RevRestate firm/years 128 123 l 10

No. of other restating firm/years 184 177 161

No. of total observations 312 300 271

 

82

 

Table 4 (cont'd)

Panel B: Company-Initiated Restatements vs. All the Other Restatements

ACCJ-t =00 + (IICFOjt + azASales jt + 013FASSETJ-t + org-DCFOJ-t
+ GSDCFOjt x CFOjt + a6ComRestate jt + a7ComRestate jt
x CFOjt + a8CornRestate jt x ASales jt

x FASSETjt + alOComRestate jt x DCFOJ-t

+ a1 1ComRestate jt x DCFOJ-t x CFOjt + Sjt

+ a9ComRestate jt

where ComRestate is an indicator for company-initiated restatement. All the other variables
are defined in Table 2. The model is estimated separately for each event year using a control
sample which contains all the restating firms other than company prompters during the
respective time period covered by the event year. All the financial firms are excluded from
the regression, as well as firms years that underwent significant M&A activities. Observations
with Cook's distance above 1 are deleted as outliners. Heteroskedasticity and within firm
serial correlation-robust t-statistics are reported. Bold indicates significance at the 10% level
or better.

 

 

 

 

 

Event Year
FY t FY t+1 FY t+2

Variable coeff. I t-stat coeff. I t-stat coeff. rt-stat
Intercept -0.04 -2.22 -0.06 -2.23 -0.03 -2.15
CFO -0.43 -3.77 -0.31 -3.20 -0.35 -4.04
Asales 0.09 2.69 0.07 2.14 0.04 1.14
FASSET 0.01 0.58 0.01 0.54 -0.01 -0.34
DCFO 0.00 0.08 0.07 1.71 0.03 0.46
DCFO*CFO 1.49 5.34 0.96 4.92 0.57 1.99
ComRestate -0.01 -0.33 0.04 1.61 0.01 0.29
ComRestate*CFO -0.09 -0.61 -0.22 -l.72 -0.07 -0.55
ComRestate*Asales 0.00 -0.03 0.01 0.32 0. 15 2.55
ComRestate*FASSET -0.01 -0.21 -0.01 -0.32 -0.01 -0.23
ComRestate*DCFO 0.00 0.05 -0.1 l -2.30 -0. 18 -2.00
ComRestate*DCFO*CFO 0.15 0.32 -0. 12 -0.38 -0.90 -1.51
R-squared - 30.42% 20.51% 16.21%

No. of ComRestate firm/years 160 153 135

No. of other restating firm/years 153 145 136

No. of total observations 313 298 271

 

 

 

 

83

 

Table 5

Distributional Statistics of Annual CEO Compensation and Its Components
for Restating Firms

Panel A

This table reports descriptive statistics of CEO compensation and its components in the pre- and
post-restatement periods for restating firms. All variables are measured in $000. The Wilcoxon
rank tests report the two-sided p-values for the testing of the difference in the medians. To
calculate the changes from one year to another, I delete all the restating firms which hire a new
CEO in either year. The numbers of restating firms deleted are: from t-2 to t-l: 77; from t-l to t:
70; from tn-l to t: 49; from t to t+2: 43.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Option Restricted Other Total
Event Years Salary Bonus Grants Stocks Comp comp

t-2 Mean 612.04 587.79 2916.89 406.25 216.99 4739.96

Q3 743.75 706.62 3073.45 0.00 113.69 5480.58

Median 550.00 297.50 874.62 0.00 38.80 2313.18

Q1 359.49 75.50 79.16 0.00 5.92 1024.86

t-l Mean 629.92 476.49 3072.53 521.90 236.21 4937.05

Q3 800.89 595.75 2817.31 0.00 138.05 5086.46

# of restaters: Median 584.77 227.34 837.75 0.00 37.91 2229.69

240 Q1 392.78 0.00 0.00 0.00 8.31 1104.87

Mean difference 17.87 -111.29 155.64 115.65 19.22 197.10
Wilcoxon rank

test p—value <.0001 0.049 0.678 0.268 0.001 0.694

t-l Mean 619.01 507.58 2618.85 229.35 141.61 4116.40

Q3 775.03 600.00 2566.46 0.00 128.94 4574.42

Median 584.77 253.69 852.94 0.00 34.57 2133.40

Q1 397.09 0.00 0.00 0.00 6.05 1153.69

t Mean 645.10 544.31 2123.73 283.93 229.35 3826.42

Q3 807.87 783.86 2254.20 0.00 156.52 4581.80

# of restaters: Median 600.00 288.01 867.40 0.00 42.50 2391.71

226 Q1 422.50 0.00 0.00 0.00 7.16

Mean difference 26.09 36.73 —495.12 54.57 87.74 -289.99
Wilcoxon rank

test p-value <.0001 0.031 0.545 0.316 0.085 0.594

 

 

 

 

 

 

 

 

84

 

Table 5 (cont'd)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Panel A (cont'd)
tn-l Mean 591.06 591.55 3494.16 271.45 136.39 5084.62
Q3 768.68 750.30 2757.31 0.00 115.64 4252.56
Median 550.00 315.43 . 708.20 0.00 33.39 1859.05
Q1 350.00 50.00 48.35 0.00 5.98 1007.64
t Mean 660.70 603.73 2255.24 269.37 255.56 4044.59
Q3 854.06 848.15 2511.93 0.00 162.75 4818.14
# of restaters: Median 630.00 340.99 928.51 0.00 50.76 2422.73
222 Q1 445.00 0.00 0.00 0.00 8.00 1135.88
Mean difference 69.64 12.18 -1238.92 -2.09 119.16 -1040.03
Wilcoxon rank
test p—value <.0001 0.399 0.573 0.064 0.005 0.189
t Mean 625.40 509.31 2010.45 198.70 275.54 3619.41
Q3 812.50 755.20 2458.45 0.00 168.67 4530.00
Median 593.58 240.51 756.62 0.00 32.76 2387.58
Q1 402.12 0.00 7.17 0.00 6.77 966.74
t+2 Mean 688.11 804.60 2260.89 462.84 395.59 4612.02
Q3 908.31 937.39 2790.09 353.50 245.99 6066.34
# of restaters: Median 687.27 476.32 619.22 0.00 57.26 2937.94
164 Q1 444.06 35.95 0.00 0.00 10.31 1135.80
Mean difference 62.70 295.29 250.43 264.13 120.05 992.61
Wilcoxon rank
test p-value <.0001 <.0001 0.787 0.000 0.025 0.003

 

 

 

 

 

 

 

 

85

 

 

Table 5 (cont'd)
Panel B
This table reports the mean difference in CEO bonus and option grants between two

years. Option (bonus) is measured as the value of CEO annual stock option grants
(bonus awards) scaled by the value of CEO total annual compensation.

 

Mean Difference between

 

t-2 & t-l t-l & t t & t+2 tn-l and t

bonus% ~2.60% 0.27% 3.76% -2.28%
[0.0551] [0.9451] [0.0076] [0.143]

option% -0.41% -4.58% -5.31% -2.52°/o
[0.6274] [0.07] [0.0067] [0.1306]

 

86

Table 6
Regression Analyses of Changes in CEO Option-Based Compensation

This table reports regression results of the following equations for the period tn-l to
t and for the period t to t+2, where t is the restatement announcement year:

AOption% = a0 + alAShare_Own + aZAExercisab le_Options
+ a3AUnexercis able_Optio ns + a4ASize + aSAB/M
+ a6AR & D + a7ACash_Cons traint + agAEam_Cons traint
+ a9ALeverage + a] OAIdiosyncr atic Risk + a1 lACurrent Return
+ alePast Retmn + a1 3ACash Compensation + a14ExtantCEO

+ (115me + Year Dummies + e

where the independent variable, A$Option%, is the difference in the dollar
value of annual CEO option grants scaled by annual CEO total compensation
(in percent) between years tn-l to t and t to t+2, where year t is the
restatement year. All continuous dependent variables are measured as the
differences (A) between years tn-2 to t-l when examining tn-l to t and
between years H to t+l when examining t to t+2, except for current return,
which is measured as the difference between the years examined. Shares_Own
(%) is CEO ownership in shares scaled by outstanding shares;
Exercisable__0ptions (%) is the CEO’s exercisable options in shares scaled by
outstanding shares; Unexercisable_Options (%) is the CEO’s unexercisable
options in shares scaled by outstanding shares; Size is the natural log of sales
(in Smillion, Compustat datal2); B/M is the book-to-market ratio, measured
as book value (data 60) divided by market value (data25*data199); R&D is
research and development expense (data46) scaled by total assets (data6);
Cash_Constraint is measured as common and preferred dividends (data127)
plus net cash flow used in investment activities (data311) minus net cash flow
from operations (data308), divided by total assets (data6); Earn_Constraint
equals one if there is an operating loss (i.e., if data178 is negative) and zero
otherwise; Leverage is measured as long-term debt (data9) divided by total
assets (data6); Idiosyncratic Risk is the standard deviation of the residual from
the market model using weekly returns over 12 months; CurrentReturn is the
accumulated monthly stock return for the current year; PastReturn is the
accumulated monthly stock return for the last year; CashCompensation is the
sum of salary and bonus scaled by sales. ExtantCEO is a dummy variable
taking the value of 1 if the restating firm did not experience CEO turnover
during the period examined and 0 otherwise; Fraud is a dummy variable
taking the value of 1 if the restatement is classified as a fraud case and 0
otherwise. Bold indicates significance at the 10% level or better. The results
for year dummies are omitted for brevity.

87

Table 6 (cont'd)

 

 

 

 

 

 

 

 

 

tn-l to t t to t+2
Predicted

Variables sign Coeff. t-stat. Coeff. t-stat.
Intercept ? 0.04 0.60 0.04 0.58
AShares_Own - 1.33 1.70 0.07 0.17
AExercisable_Options - 1.19 0.53 -6.68 -1.46
AUnexercisable_Options - 1 .74 0.29 -12.96 -2.30
ASize + -0.12 -1.58 0.14 1.29
AB/M - -0.06 -2.46 0.00 0.06
AR&D + -0.24 -0.25 1.62 2.20
ACash_Constraint + -0.39 -1.81 —0.02 -0.09
AEarn_Constraint ? 0.01 0.10 0.03 0.34
ALeverage - -0.83 -3.31 0.03 0. 1 5
AIdiosyncratic Risk + -0.05 -O.56 —0.63 -0.97
ACurrent Return + 0.00 0.03 0.00 0.08
APast Return + -0.12 -3.56 0.04 1.71
ACash Compensation ? -0.02 -1.85 0.03 1.41
ExtantCEO - -0.05 -0.65 -0.12 -1.50
Fraud - 0.1 1 1.49 -0.04 -0.56
n 151 121

Adj. R-squared 8.78% 11.84%

 

88

 

00:03 00 _0>0_ $8 05

:0 00:02.:ch 300205 Eom 00:09: 0.5 8508?“ “manorbfitmfluflmobfim .N 030% E 005.800
0.8 3300:; 0050 05 =< 080—00 000 300500 63.34 Emomfiwmm 00030000: :20 300.»-805 .AmEE
m8 0000—00 0.3 Ni >m 00 0 Cr.» 5 Duo 30: 0 0:: 0:3 £00m wccfime .m 05$ 00m .955 08
0000.00 0.3 u >m 00 75 >u~ 5 CPU 30: 0 0:: 0:? mafia $50500 .< BEE 00m ._0008 33.800 0.0 E
3300:; £08 05 505 “0000:: 333.5; _00000 05 =< .2 .8 33005 05 2 00300005.“; .8908
£030 039 0w000>0 $8.5 .00 x03 05 2 VESI0N5 w: .8 .0302: 0.0 00 0030000005 .30808 .3300
.00 020> x003 .3 00030 fimfimfoafimE 3:60 mo 02? 000—008 mm 000.000 008 0.009 9 “00.00:.
00 x000 05 fl 00:2 mtz $255nt 000 .330on .vboméog .Emménmm .0mwm-mmwm $0000 2m
.3 00.00002 3305 9023:— m 5 mm? 8.5 0 8 000 00000 2 05 m: .8 0.30:: 05 00 003030003
.3808 $080 _800 .3 00230 $330+ 08008 500 .800 m0 005.000 008 0m80>0_ .2020 .30

x00.— 05 fl xsmul0w80>2 ”00203 332:3 _05000 .0wcwso 000 $0006 «00¢ E .0350? 03:05 03,—.
MC .8 .0300: 05 00 0030.80.85 6003000800 13:00 .800 OmD E 03.50 3009 OmU .«0 098:0
093000000 00 M08 05 fl xcfilfiomscom m2 .8 03.35 05 0“ 0030000003 605800.300 30:00
:39 OmU E mucmnw 00:00 0.00% OmU .30 0wcmno 083000009 .00 x00.— 05 mm x000I$00300 00:3

0 + 830. 528.80 + .3000 x @0000 x f: «01.83801... +

0&8 X «av—d §|0\0WSGOMH<©~5 + ufihtmmm<n~ X “five élfimfifiom<m~d +
a 0... £59832... +

.3000 x .3 0.: 8.1.8.383... r. +

.3
..

H.H 3.me x H.H 0.: §I§mscom<30 +

:05 x @0000 x .3 0...... 01:05:54. :0 +
.5000 x ..q 0:... 01.85.0042... + .fimmmfi x .3 0...... «$3308... +
a. 834 x .4 0.5. 01.85308... + .3000 x .4 :5 01.8.8001... + .q x5... 0:36.020... +

a

.5304... + .fimmmfimu + .3 8329. + 00.00:. + o...“ .. 00...

.3000 x 00.50? +

”mm 003503 9:03 cosmfiwe 2E.

mow—5:0 500:0 0:0 3:8— .mOmU 0:0 8030:0200 C0 .30.— :00300: 130.000.:3— 2:.
N. 030,—.

89

 

wvd
de-
mvd
om.—
mm. _ -
de
and-
d _ .m-
mud-
_Nd
mud-
00.—
aha
end
3 . T
Sud-
dm d-
vod
and
3.8
N. _ .—
om.—
n 0 d-
mm.~-
05.0-.

 

MdN
add-
mod
_Nd
Ed-
Nod
mad-
bmd-
odd-
mdd
mmd-
d_ d
umd
00d
8 d-
add-
2d-
2 d
3...
Odd
:d
dmd
N_ d-
8d-

H.000

EV

mmdu
05.”
8 d-
odd
mm d-
mm d

Sdn
N—.N
nmd-
7:..—
05.7
C . _ -
«Sun.

 

mwdT
Nod-
5d-
3d
Cd-
mod

00¢.
NNd
Nod-
8d
0md-
mod-
E000

and
add
5.8
wd. T
Nod
m _ ._
00.7
dN.N-
Ndd
w—fi
3d-
wd. _ -
~30-“

 

wdd
End
wdd-
NNd-
mm d
nod
EN-
wmd-
ddd
Rd
mod-
vod-
£000

0

00.0
and
3 d
V”;
2:1
mo. _ -
230-.

 

8mm

nmd

_dd

2 d

mmd-

mod-
H.000

8

 

000.0000. isnomeu$0
OmUQ .. 0.000930051—
hflmmflm .. 0.003030031—
000200 0. x000lowa00>04

OmU 0.0.000I0w00051—
0.000I0wa0054

000.0000. 0.233.222
0000 . 0.5300222
0.0.030 .. 0530.222
00.004 .. 0:00Ic\cm=:00<

000 .. 0:..-{322
0.000.600.0034
000.0000. 0:2-.0...¢00o<
OKUG .. 0.000I300300<
0003.0 .. gazu$=o=03
00.34 0. 0.000Io\a:0000<
0.00 .. :0a0le\..:0_000<
0.:a0lo\e:0000<
9.000900:

OED:

hmmmtfl

00.03

0.00

000008:—

2.3.5

 

 

05:80 0 2...;

0 2.50. :0 8000300000 2. _0000

90

 

 

 

 

 

 

 

 

2m 2m 2m 2m 23323.... .33: .az
cameo $2“. a $3; $8.3 3.2.5.:

:3- «S. 9.90%? 3.33%
84 mg 95: .. 35.185
2:- mod- Emmfi .. x535
2:”- :3- 8.52 .. fielfim
NE. N2 95 .. €2-35
3o 86 €2.95
3.. _- 32. 96.559, €33.92
8.? Ed. 96: .. :52:th
m3 :3 .559; « €212:
ad. 85. 8.53 V. iguana:
mg 85 95 .. €219;
Ed 8d €215:
NS Nam 055,59, .5
:2 m2 95a .. .5
85. 8d Emmi ., .5
N3 25 8:8 15
85 So 96 u. .5
Ed. 8d- .5
5:53 < $93

€283 h «Bah

91

 

 

 

 

 

 

 

 

N3 _Nd o5 .. .5
3d- 2d- .5
m2- 83. 399,59 éeuoueozd
m: and 95: .. 33.09.23
3d- wdd- Saws. .. fizdu§>3
om. T 3d- 8E3 * x=ahlouauo>od
2d. 3d- 9.5 3.55.323
a. .o Ed x=aulow§o>od
d2- 3%. ad 2 .N 9.9.959, fienezdddfi
8..- d3- 2.. 5d 95m .. €E|§==B<
NE. 8d- Ed- mdd- Emmi ., €2J§§=3<
3d- 3d- mud- ddd- 8.34 V. gueuecsdzq
8d- 2 d- wvd- 8d. 95 .. gueuxuédfi
NE 8d m5 2d €21§==3<
:.._ dvdm .3 «3 899,69, z=a._u..\..._o=%<
a: x: 8d 3d 95: .. xdeucddeaé
Rd- ddd- ddd odd 5%.; .fi 45.1.3832
dmd- 8d- 8d- odd- 8:3 ., gangesaé
3d- 3d. odd- Rd- 9.6 ., isacdgié
a: d. d Sd 3d g=a.|.x.=d_.dd<
2d- a... 2 d- mmd- dd. _- d3- VS Ndd 9.99.59
:3- ddd- «d..- 2 d- Ed- m _ d- 2 d- _dd- 9.8:
:2 ddd 9d 8d 8d- 3.9 Sd- 3d. 533.
8d wvd 8d ddd NS .3 3: odd 8:8
3d Ed 6d. 2 d- Ed- odd- 2d- dd. 95
23. 2d- 3d. d_ d- :.._- 8d- 3d- 3d- 585.:

«at; _ .393 33-. _ E03 :33 _ E03 «3?. _ £08 933...;

Ed :5 Ed 6

 

 

23:33 h 03.;

mi 08:13 Emzmtowcou Hm $ch

92

 

 

 

 

 

 

 

 

5:83 N. 22F

.5. d2 d2 5 2.2.2.3.... .53... 5:

$5.5 $5.2 :3 .2 53.5 3.23-5

8... 8.5. 95,959, €255
E..- N... _- 95: ., €2.25
2d. 8d. 5555— . €2.25
d5. _- :d- 858 h. €2|25
5:. 5d- 9.5 ., €2.25
55 3d €2IS5
dd; 2.8 95559, 5.2:th
add 3: 9.5: . U52:52
5d- 8d- 5555. .. €255
mad dmd 858 .. €255
5d- 2 d- 9.5 ., 5.2..th
5d 8d €2|52
.2 5.3 05.95:. .5
52 3d 0.5: 5 .5
d2 wdd 555$. .. .5
5."- 5d- 8:2 * .5
6.83 m .25

93

The regression being estimated is:

ACCjt =00 + GICFOjt + a2ASales
+ a6T- + a7Tx CFO

J
J

+ (111zj DCFOjt X CFO

jt +08
+ a9T- x FASSETjt + alOT- x DCFOJ-t

J
jt

jt

Table 8
The Relationship between Change in Conservatism
and Change in CEOs' Option Grants

+ a3FASSET

J

t + 0.41)c:1=ojt + asDCFO .t x CFOJ-t

J

where all the variables are defined in Table 2. The benchmark year is Year t. Finn-years
that underwent significant M&A activities are excluded. For Panel A, restating firms who
hire a new CEO in FY tn-l or FY t are deleted (49 firms). For Panel B, restating firms who

hire a new CEO in FY t or FY t+2 are deleted (43 firms). Heteroskedasticity-robust t-

statistics are reported. Bold indicates significance at the 10% level or better.

Panel A: Change in Conservatism from tn-l tot

 

 

 

 

 

Top 50% firms Bottom 50% firms

whole sample with low Aoption% with high Aoption%
Variable coeff. I t-stat coeff. I t-stat coeff. J t-stat
Intercept -0.03 -1.01 -0.03 -1.55 -0.03 -0.41
CFO -0.52 -3.99 -0.56 -5.21 -0.56 -2.22
Asales 0.18 1.91 0.14 2.66 0.24 1.36
FASSET 0.00 0.03 -0.01 -0.42 0.01 0.10
DCFO 0.30 2.67 -0.10 -1.24 0.47 2.71
DCFO*CFO 3.49 4.71 0.28 0.85 3.81 5.79
T 0.03 0.82 0.03 1.07 0.01 0.13
T*CFO 0.22 1.43 0.17 1.04 0.34 1.31
T*Asales —0. 1 3 -1.29 -0.17 -1.96 -0. 14 -0.79
T*FASSET -0.06 -0.92 -0.03 -0.91 -0.06 -0.46
T*DCFO -0.27 -2.16 0.09 0.93 -0.44 -2.18
T*DCFO*CFO -3.20 -3.89 -0.61 -0.84 -3.47 -4.70
R-squared 63.55% 24.70% 71.98%
No. of total observations 428 214 214

 

 

 

94

 

Table 8 (cont'd)

Panel B: Change in Conservatism from t to t+2

 

 

 

Top 50% firms Bottom 50% firms

whole samm with low Aoption% with high Aoption%
Variable coeff. I t-stat coeff. [ t-stat coeff. I t-stat
Intercept -0.04 -2.28 -0.02 -0.62 -0.05 -2.06
CFO -0.43 -3.69 -0.75 -5.73 -0.27 -1.75
Asales 0.11 2.72 0.10 2.10 0.13 2.17
FASSET 0.00 0.18 0.02 0.77 -0.01 -0.40
DCFO 0.03 0.52 -0.07 -0.70 0.09 1.96
DCFO*CFO 2.1 1 6.37 1.91 4.08 2.32 10.49
T 0.00 -0.04 -0.04 —1.07 0.03 0.93
T*CFO 0.14 0.99 0.62 3.18 -0.11 -0.56
T*Asales -0.05 -O.78 -0.02 -0.20 -0.06 -O.72
T*FASSET -0.02 -l.04 -0.04 -l.66 -0.02 -0.56
T*DCFO 0.00 0.05 -0.02 -0. 15 0.03 0.32
T*DCFO*CFO -1.68 -2.00 ~2.96 -3.33 0.36 0.87
R-squared 23.23% 25.66% 36.43%
No. of total observations 318 160 158

 

 

 

 

 

95

 

6050002 03 505235

-2 53055625008502.0203 20:3 20 _0>2 $2 05 8 oocwoficwmm 530202 20m 02200 03 522220.... «5.2
250$:me 20032000: :2: 580%-:Em 20008 33000 05 E 50322? 22: 05 :23 68022 5032:; 35:00 05

=< 023.850 o 0:0 022m 50 02.25220 2 00080352 05 t: _ .20 022, 05 5028 033.2 60500 00852 02:20 05 20>0
2:005 80 00852 05 .«0 :25 05 .3 00220 .80» 08052 2000 80¢ 25080 080502 05 .«0 :25 05 $22. 0220200
2 00:2:me #:082053— .2 .8 12225 05 02 00222285 .2032:me 20082052 05 .20 0.52 05 2 2528253“
”00550232 3000 05 5 5032.23 32:00 0208 025 00202 _ K 030,—. E 02500 50322? 05 :2“ 9 002200 5

2a

. w + 5052 820950 + a 2.5

020 x 0. 0.200 x 2 £35280? :0 +
20200 x 0.. 2. «212.2830 :0 + 2.25555 x 0.. € 5.12.2835 :0 +
H.H 52mm< x H.H x: malfomzcom<30 + “Hon—U x 2. x: mal§5scom<20 + A x: «alfamscomdm _0 +
H.4020 x H.2950 x 0.. 25 212.8302 _5 +

0.200 x 2.. 2.. 21282225 + :Emmé x 2.. 28 212.88%? +

0.. 23% x 2.. €a 21.852255 + 0.2020 x a as. 2.2.320th + 0.. 25 22.32%? +

H€20 x H20.602. + 20.20050 + 0.2.5555? + 0.2

2..

H..
2:599. + H520:0 + can HJo...

2 0038250 wasp 5550232 05.
a 05; an Emzmtemcoo *0 .054 05 :0

320.32... m:==:000< 05 0:0 mouagcmas 22:23.00”. 05 no 30008. 0:...
a 030k

96

 

 

w I ~0.~ 80 0: ~00 £0 30 000 00000000.. 0.53ng
_2- $0- 21- 30. _~._- 60. :0- 000. 0000 .. 050100232
8.0 :0 2: S 0 m: 20 000 20 00220 .. 05.30232
0~0 000 0.00 80 $0 00 0 000 ~_ 0 8.30 .. 0.521002%.
00.~- mm. 0- 00.0- 03- 00.0- 2:- 0:- 02- 000 .. 0:200:22
S0 000 000 80 R0 000 ~00 80 0:21.00232
02- $07 2:- $0- 2 0 20 000 2 0 00000000... 0:213:33
~0.~- $0- 0m.~- :0- :.~- 30. 0...~- 000- 0000 .. 0.02.3353
00. 0- 20- ~00- 20- ~0. _- 20- ~00- E 0- 00030 .. 0:21.0855
£0 .00 :0 80 0~0 80 0~0 80 833 0 05308550
000- $0. .000 ~00 ~00- -0- t 0 00.0 000 .. fieufimécfi
vm.~ 2 0 .0.~ ~00 0E 2 0 m0.~ S 0 0.02:..0382
2.0 2 .~ 2: 3.0 03 ~00 m~.~ w~0 00000000... 0:2..82002
$0- 0. 0- ~00 ~00 00.0 -0 8.0 :0 0000 .. 0.02:2...8000
0S- :0- -._- 2 0- 05. E 0- 8.0- 20. 0002.0 .. 0:00:905002
0~._ m~0 000 :0 ~00 20 £0 E 0 8.30 .. 0.50:0.5008
~00- 30- ~00- 20- $0- $0- $0- 80. 000 .. 0.5.5.5002
S.~ 0 _ 0 0~.~ 20 G.~ 20 ~m.~ :0 0.5.1.0822
00.0 00.2 000 $0 $0- 00.~- 30. 00.0- 00000000
2: 000 000 ~00 :0 0~0 £0 ~0 0 0000
$0 80 0~0 000 0~._ _ _0 02 0 _0 0.0030
0~0 80 000 20 02 2 0 2: 20 8.34
000 000 w~0 v~0 000 $0 -0 20 000
2..~- -0- 05- £0- 00.~- E0- ~0.~- 0~0- .0855

030.0 _ £000 .30.. A H.000 .33 _ H.000 030.0 _ H.000 0.0305»

 

 

20

 

 

00

 

00

 

 

 

8.280 0 29¢

97

 

 

Sho-
mud
3.0-
omd
3.0
00.0
_m. _-
MFA-
two
m0. _-
00.0-
mm.—
wwd
~0.N
00.0
ovd
end-
0m._-

 

3.0-
hmd
no.0-
3.0
3nd
cod
0&3-
00.0-
mod
@00-
0nd-
no.0
Nod
N50
006
v0.0
00.0-
no.0-

w _ .o-
E:
ood-
who-
00.0
mmd-
0 _ . T
00.7
00.0
Na;-
o0.o-
mod
31
«3.0
3.0
10.0
v0.0
ood-

 

0 —._-
00.0
00.0-
mod-
Nvd
5.0-
04 _-
5.0-
v0.0
000-
Rd-
000
Nvd
m0.o
Nod
v0.0
3.0
mod-

95
and
00.7
mmd-
mud
mmd
wad-
_ :-
3.0
m0;-
$0-
00.—
m—.N
Nb.—
2:
3 .o
wmd-
0N2-

 

Maw.—
00.0
_ :0-
mod-
end
No.0
ons-
00.0-
mod
3nd-
omd-
mod
\- 0 .0
m0.o
mod
3.0
:d-
vod-

00.0
w0.~
on;-
3.0-
m0.o
omd
No.0-
00. _-
mwd
00.7
010-
00.0
m0.~
m0;
Sud
vvd
mod-
No.0-

 

_~.N
:20
0 0.0-
m“ .o-
vvd
No.0
_ 0.0-
Exo-
00.0
0nd-
2.0-
mod
56
wmd
No.0
#06
No.0-
8.0-

 

000-0000- 0.52-Em
0000 .. 0.52-Em
0.000-«0 .. 0.02-000
0.203 0, 0.00.. 005

000 .. 0.02-cam
1:5.— 0N5
000-0000... 03.2-00.2,
0000 .. 0.52-0.02
0.0.0000 - 0.52-0.02
mo_aw< .0 02:; 9%:
000 - 0.52-002
0.52-002
00000000.. .5
0000 15

000200 .13

8.03 15

Oahu 0 =1—

.5

 

 

00.2.80 0 20¢

98

 

 

2 . 0 -
cod
No.0
:20
_ To-
00.0-
N—.N
00.0
wad
«0&-
m 0 .o-
and

 

hf
£0003

0N0-
cod
nod
m _ .0
mod-
vod-
3 .0
_ _.o
00.0
~Nd-
00.0-
No.0

 

mwd-
nmd
mm;
wfio
£6
:2-

0w.
£08.00

05v-
ad
2 .0
No.0
mod
no.0-

N0.—
00.0
Swo-
a0.-
de-

 

0wd

D:
33.3

nod
5.0
No.0-
wfio-
ood-
mod

 

hw—
€05.00

 

0005:0800 .80. ._0 .0 Z
00000—004—

000-0000... 0.50-3.80

0000 .. 0.02-0080 .

0.000-«0 ... 0.02-0080
@0136. 0» 3:3.— @500“
000 0 0.53080
2:97.503—
00000000- .520
0000 0 0520
0000.00 .. 0:20
00.02 0, 000.5

000 ... .520

0:00.0—

 

 

05080 0 2000

99

Table 10
The Impacts of the Restatement Magnitudes and the Accounting Irregularity
on the Change in Conservatism

The regression being estimated is:

ACC-

Jt=a0 +(2t1CFO-t +a2ASales +0L3FASSET-t +a4DCFO-

J J't J Jt

+ (ISDCFOjt x CFOjt + a6Tj + (17zj CFOjt

+ agTj x ASales jt + a9Tj x FASSETjt

+(110Tj XDCFOjt +(111zjDCFOjthFOjt +8jt

where all the variables are defined in Table 2. The benchmark year is Year t. Firm-
years that underwent significant M&A activities are excluded. Heteroskedasticity-
robust t-statistics are reported. Bold indicates significance at the 10% level or better.

Panel A: Change in Conservatism from tn-l to t: Fraud Firms vs. Non-Fraud Firms

 

 

 

 

 

 

 

Fraud Non-fraud

Variable coeff. ] t-stat coeff. I t-stat
Intercept -0.03 -l .5 1 —0.03 -0.72
CFO -0.63 -4.91 -0.54 -3.14
Asales 0.06 1.60 0.26 1.78
FASSET 0.01 0.47 0.00 -0.01
DCFO 0.08 2.53 0.37 2.55
DCFO*CFO 3.51 1.52 3.59 4.84
T 0.06 1.91 0.00 0.10
T*CFO 0.19 1.20 0.33 1.76
T*Asales -0.05 -1.02 -0.23 -l .41
T*FASSET -0.08 -2.01 -0.03 -O.41
T*DCFO -0.02 -0.31 -0.35 -2.26
T*DCFO*CFO -3.90 -1.67 -3.28 -4.09
R-squared 60.68% 65.05%

No. of total observations 122 306

 

 

100

 

Panel B: Change in Conservatism from tn-l to t:

Table 10 (cont'd)

High Restatement Size Firms vs. Low Restatement Size Firms

 

 

 

whole sample Top 50% Bottom 50%
low restatement high restatement

Variable coeff. L t-stat coeffj t-stat coeff.? t-stat
Intercept -0.07 -2.13 -0.07 -1. l 1 -0.07 -3.51
CFO -0.45 -3.05 -O.61 -3.16 -0.39 -1.84
Asales 0.08 1.88 0.08 1.08 0.1 1 2.54
FASSET 0.06 1.03 0.1 l 0.92 0.04 1.23
DCFO 0.24 2.83 0.45 2.41 -0.05 -0.87
DCFO*CFO 3.58 6.06 4.06 9.09 0.06 0.20
T 0.07 1.83 0.08 1.08 0.05 1.56
T*CFO 0.16 0.96 0.31 1.37 0.17 0.63
T*Asales -0.04 -0.65 -0.02 -0.26 -0. l 2 -1.21
T*FASSET -0. 12 -1.89 -O.17 -l .31 -0.08 -2.02
T*DCFO -0.18 -1.63 -O.43 -1.84 0.08 0.74
T*DCFO*CFO ~3.20 -4.65 ~3.44 -6.97 -0.59 -0.93
R-squared 78.12% 89.31% 18.18%

No. of total observations 372 186 186

 

 

 

 

101

 

00309 no _0>0_ £2 05 8 00:00E:w_m 88065 20m 602800 03 85:93 53835520050

300m SE 2223 can 365800—0820: 06058 £000 6 .x: 3 00585 08 0030:? 3035280 =< .3355 303-600 85 no“ 02200 08 30>
E0>0 £000 5 OmU 30: a 0:: £033 085 was—30.8 05 =< 0023000 {6002 EmoEcw_m 8030025 85 0.80% 085 m0 :03 mm 606000.000. 05 50¢
000280 03 2:5 32:35 05 =< .50» E0>0 05 E @0550 00:3 08: 03002300 05 9:50 0.85 $59000 05 :05 0050 8.5 ~905an0

05 :0 35:00 £033 03:80. 33:00 a main .00» Eo>0 £000 08 38838 0000850 m_ 602: 05. .N 030,—. 5 008.00 000 830:? 05 :0 00:3

a... A H.. 0.. :
1.. ODUX On—UQX 9350M 5+

4 a a a
.. 0.50 x a. 258% _u + a. Hmmm/E x a. 298er +

H.H mofim< x H.H 00800de + H.mOmU x #0080315 + a. 003.00%06 +

a “a _. _. _. n we
0. OmUx .OmUDmd+ u. OmUQVU+ H.Hmmw<mmd+ «. mflmm<md+ H.Onmnvfc + o5” .UU<

.380 .8200 mm “835800 E 25E 0050 05 =< wEmD H< Baum

mua0>-E.._m OmUn30 Z “500.09 00:0
3602 35.53.. aim: 3:0E00=:0=:< Eon—0.5001 “:50...“ 0:5,; $5300.: .8.— Em_§>..0m=00 a: $50.0...

: 03.;

102

 

 

 

 

 

 

 

 

 

 

 

 

00 _00 808 808 880 0200 8003.800 .32 00 .02

0 _ 00m 880 85. 8000 00000 8885... 000380-00: 00 .02

mm 0m 0m 0m 8 FORE 5E 8835.... 0.5580 00 .02

$8 80. 000. 80 00m 888...... 0:058. 00 .02

$8.2 $8.2 $82 $00.: $8.: 022.03.

00. _- 00.0- 8.0 00.0 00.0 00.0 8.0. 000- 02 00.0 0005009388:

8._- 8.0- 00 .0 00.0 8.0 8.0 8.0 8.0 80 8.0 805858.:

8.0 8.0 02 8.0 03 8.0 8.0- _00- 8% 8.0 50828038:

00.0 8.0 00.0- 8.0- 8.0- 8.0- 80- 8.0- 00 . _- 8.0- 8.83388:

00.0 00.0 t .0 _00 8.0- 2.0- 8.0 :0 8.0- :0- 958.380.

00.0- _00- 8.0- 8.0- 8.0- 00.0- $0 _00 3.0 _00 888:

00.8 00.0 00.00 8.0 00.00 00.0 01:. 00.0 00.0.. 08.0 955.5:

8.0. 8.0- 8.0- 80- 00.0- 80- 2.0- 8.0- 8.0- 8.0- 9.0:

000—- 80- 00.2- 8.0- 00.:- 80- 00.00- 8.0- 00.2- 8.0- Emmi

8.: 8.0 80— 80 00.0. 8.0 00.8 8.0 8.3 8.0 8.08

8.00. 00.0- 8.8- 00.0- 030- 00.0. 3-00- 00.0- 00.8. 8.0- 9.0

.00.? 8.0- t00.0. 8.0- 00.0- 8.0- 8.0- 8.0- 8.- 00.0 0088::

_ 23m... ”:25 :3qu F390 «wanna Human 23min ”Coco «aumia ”Coco Snataxr
_ 3 >0 .8 >0 0 >0 0.... >0 «-5 >-._

_ 30> E0>H
€383 < $53

80080 : 03.2.

103

 

000000 00 _0>0_ £03 0:0 00 0000000000000 00000600 28— 000000000 0000 00000000000 003.000-22003080

_00000 800.0 03:3 00:0 >0_0_000_000_0o0000: 00002000 000 _ 0500 0000006 0.0.ooU 503 0020020090 05 =< 00002000 000 00000—000 00000-00005 £000

E OmU 30: 0 000: .0023 0:00: wE000000 0:0 =< 000030000 <Q2 0:02-0:0w00 000300000: 00:0 00090 0.00-00.0 00 :03 00 .020000w00 0:0 880 0000020000
000 0500-0 0000:0000 0:0 =< .0090 0:000 0000 wctzc 050 00000 0.80.0 $500000 05 0o 0\0o_ -\+ E00003 000 0800.0 wE000000 0:0 00 500305 0:000 05

E 0:00: 0:000:00 £053 955 300:8 0 @000: 80> E03 :000 00.0 300000000 000080000 00 .0002: 00:- .N 0300- 5 0005-0000 000 00300000, 0:0 :0 000003

0 0 0 0
0. 0+ 0. OmUx 0. OmUQx 0. 0000031 00+

0 0 0 0
0. DRUG x 0. 00000004000000 + 0. Hmmmxxm x 0. 00000000000 +

0.0 0£0m< x 0.0 000000de + 000-00 x 0.0 000000-00 0.0 + 0_. 000003005 +

0 0 0 0 0 0 0
0. 000 x 0. 000000 + 0. 000000 + 0. 0000.000: + 0. 8000000 + 0. 00000 + 00“ 0. 000.

9500 300000 00050002 -0N0m 0000 5000205 r000> mam: ”m .0000

8.0080 00 0.00-0

104

0—0._:¢Uv —_ 22.:-

 

 

 

 

 

 

 

 

 

 

 

 

 

30.00800 : 030,—.

0... 00.. 00.. 8.. 0... 80.09880 .33 .0 .07.
000 0. 0 000 000 .00 880E... 050380-000 .0 .0z
0. 0. 0. 0m 0. .u000: 5.: 8800...... 00.00.80 .0 .02
.0. 8. 00. 00. 0... 88005... 00.038. .0 .0z
$.20 $0.... $00.... $00.0. $00.0. 02.2.03.
00..- 00.0- 00... .00 00.0 00.. 00.0. 00..- 0.0 80 0000000900880
00.0- 8.0- m. .0- .00- 00.0 0.0 0.0 .00 0.... 0.0 00092080
00.0- 8.0- 000- .00- 00.0 .00 8.0- 8.0- mm. .- 8.0- 0.000.008.0080
00.0 00.0 00... 8.0- 00.0 2.0 00.0- 8.0- 000 00.0 8.038.580
00..- 00.0. 000- 8.0- 8.0- 8.0- 0.... 00.0 00.0- . .0- 000.8380
00.. 00.0 :0 8.0 :0 00.0 00.0 .00 00.. 8.0 000.80
00.0 00.. 8... $0 00.0 00.0 2.0 8.0 00.0 00.0 000.0000
8..- 8.0- 8.0- 8.0- 00..- 00.0- 00..- 8.0- 00... 8.0- 00.0:
8.. 8.0 0... 8.0 8..- 8.0- 3...- 8.0- 8.0 .00 0.00000
0 . .0 .00 00.. 0.0 00.0- .00- 00.0 00.0 0.0- .00- 8.84
00.0- 0.3- _0....- 0.0.0- 2.0- 8.0- 0.0.0. ....0- 00.0- 00.0- 000
:0...- 80- 8.0- 00.0- 00.0. ..0.0- 0....- 80- 8.0- 8.0- 0088....
0.30-.0 ”Coco 050.6 ”Coco «flaw-.0 ”Coca «Gum-.0 ”Coco usmnu .Cocu 033.30»

0+. >0 .0 >0 0 >0 0.... >0 0.... >0

33> «:93.—

Gucoov m 3:000

105

.q’ .. I'.\ ‘.fi

 

.0022. 00 3.6. 000: on. .0 80.00.0890 00.86:. 20m 000.00%. 20 02.0.8.0

-. .038- b.o..mmwov_0o0.2o: .0222. 000 _ 0250 80.0.06 0.0.80 5.? 32.02030 2.. =< .0222. 20 03> 0008:0000. on.

.o .00» .520 0;. 0056 E OmU >5: 0 BE 5055 080... 950.00.. 05 0.03008 000%-.005 some 0000 02.30.00 .0332 2.00....ch
.5202: .00.. 0.0.90 E0... 00 :95 00 0.200230 0... :5... 020288 20 00.0.... 30:00.... 00.. :< 0.0.. 030.6% 02.: :023 00.00...
950.08 2.. :0 mafia .090 .026 some 00.. 380000.00 008080.00 fl 6on on... .N 030,—. E cacao—0 000 003000? on. :0 82.3

0 0 0 0
..0+ ..O0Ux ..O0UQx ..0..0+
0 0 ..0 .0 ..0 0 ..0 0 0
.. 0000.X . 00.0+. 0000.00 x 0000+. 00.0.3VA . 000+. .O0Ux . 000+ . 000+
..0 ..0 m ..0 0 .0 , 0 .0 0 .0 . 0 .0

OmUx OmUD 5+ OLUD 5+ .mem<n~ 5+ .mo—wm4 6+ .0000 6+ d” .UU<

0:00.00 9.0.0.030 55.3 Swamtomcou E omcmno ”U .2000.

20.0.53 : 030,—.

106

 

 

 

 

 

 

 

 

 

 

 

 

0.0.080 .. 0.000

mom :w .mn wmm 00090050: W533...— _53 ..0 .07.
£02.06 .xbvdc 0x.— _.mo $3.: “09:350-:
NNé. 30m- Ema- 3N- 3.0.. 0.0.0.. mo. 7 av... 0&0000—0909
Sam- wmd- 2:- mmd- 3. T a _ .o- Nwéu NNd- 9.00900.
no. T mod- 36. mod. mud- mod- mod- cod hamwflmfr
.3.— o _ .o mm. mod 0 _ ._ .36 mm. mod 00.0300.
:3 20.0. :d 92. 0:6 and 26 .86 00.000.
cod- mod. N _ . _ - vod- ood- cod nod cod a.
mwé no...“ mwé $.m vwé 0.0.0. and. SM OhU0OmUQ
2: mmd 2: mmd acfi mmd aQN mmd OmUG
mud mod 36 mod mud mod mud mod Hmmm<m
26. Nod- 36. Nod- 36. Nod- 36. mod. 0200<
50%.. 3.0. chm- _wd- Ffi- 50.. Ram. _wd- 0&0
86. Ed. cud- 3.0.. end. _od- ON? Ed. 3308:—
0300 .0003 050-0 .0008 050-0 ”:03 050-0 .0008 033.80»

Ni :0 75 «-5

c u 03> 0.03—£2.05 03> 0:95—

5:53 U 6:00—

107

Table 12
The Relationship between the Accrual Measure of Conservatism
and Firm's Performance

The regression being estimated is:

ACCjt :00 + (XICFOjt + (12ASalesjt + G3FASSETjt + (l4DCFOjt

+0.5DCFOjt XCFOjt +a6ROAjt +0.7ROAjt XCFOjt I“
+ aSROA jt >< ASales jt + 09110.0jt x FASSETJ-t if;
+0.10ROAjt XDCFOjt +a11ROAjt XDCFOjt xCFOjt i'
+ ControIVar iables + e jt

where ACC is total accruals; CFO is operating cash flow; Asales is change

 

in sales; F ASSET is book value of fixed assets; DCFO takes the value 1 if V
CFO <0 and 0 otherwise. They are defined in Table 2. ROA is the
standardized decile rank of net income before extraordinary items (data18)
divided by average total assets (data6). Other control variables include:
Leverage is the standardized decile rank of leverage ratio defined in Table 7
Lit is coded one if the firm was in a litigious industry defined in Table 7;
MTB is the standardized decile rank of market to book ratio defined in
Table 7; Size is the standardized decile rank of average total assets defined
in Table7. These control variables also interact with the main variables in
the accrual model. Standardized decile rank is determined by first ranking
observations each year into 10 groups, and then scaling the ranking by 9.
All the financial firms are excluded from the regression, as well as firm-
years that underwent significant M&A activities. The sample period is
1997-2006. Pooled regression is used for the estimation. Heteroskedasticity
and within firm serial correlation-robust t-statistics are reported. Bold
indicates significance at the 10% level or better.

108

Table 12 (cont'd)

 

 

 

Variable coeff. I t-stat
Intercept 0.00 —1.03
CFO -0.57 -l3.27
Asales 0.05 11.08
FASSET -0.02 -6.47
DCFO 0.05 11.17
DCFO*CFO 0.52 10.52
ROA 0.05 1.39
ROA * CFO 0.35 1.32
ROA * Asales 0.12 8.07
ROA * FASSET 0.01 0.59
ROA * DCFO 0.57 14.37
ROA *DCFO*CFO 0.11 0.42
Leverage 0.00 -0.33
Leverage" CFO -0.09 -l.44
Leverage * Asales -0.02 -1.53
Leverage * FASSET -0.02 -1.79
Leverage * DCFO -0.01 -1.28
Leverage *DCFO*CFO 0.11 1.76
Lit -0.01 -2.57
Lit * CFO 0.04 1.69
Lit * Asales 0.01 1.47
Lit * FASSET -0.01 -1.47
Lit * DCFO 0.02 4.38
Lit *DCFO*CFO -0.02 -O.65
MTB 0.00 0.01
MTB * CFO 0.00 0.19
MTB * Asales 0.00 -0.51
MTB * FASSET 0.00 -0.01
MTB * DCFO 0.00 0.94
MTB *DCFO*CFO 0.00 0.23
Size 0.00 -2.93
Size * CFO 0.00 2.24
Size * Asales 0.00 -4.79
Size * FASSET 0.00 3.51
Size * DCFO 0.00 -5.53
Size *DCFO*CFO 0.00 -5.49
R-squared 49.36%

No. of total observations 40442

 

 

109

 

Table 13
The Relationship between Level of Conservatism and
Changes of CEO Option and Equity Sensitivities

The regression being estimated is:

ACC-

jt

+ 0.4DCFO '

=00 + GICFO 't + (leSfliCS 't + GBFASSET Jt

J J it
+ (ISDCFOJ-t >< CFOjt + a6AOptionsen_Rank jt

+ a7AOptionsen _Rank - x CFO 't + a8AOptionsen_Rank -

.1t .1 .1t
x ASales jt + a9AOptionsen _Rank jt x FASSET-)-t
+ alOAOptionsen _Rank jt x DCFOjt + a1 IAOptionsen _Rank jt
X DCFOjt X CFOjt + aleEquitysen _Rank jt

+ a13AEquitysen _Rank jt x CFOj.t + a14AEquitysen _Rank jt
x ASales jt + alsAEquitysen _Rank jt x FASSETjt
+ a16AEquitysen _Rank jt x DCFOJ-t + a17AEquitysen _Rank -

3t
x DCFOj,t x CFOjt + ControIVar iables + Sjt

where Aoptionsen_rank is the rank of the change of CEO option sensitivity,
standardized to the interval (0, 1); AEquitysen__rank is the rank of the change
of CEO stock equity sensitivity, standardized to the interval (0, l). CEO
Option sensitivity is measured by multiplying the option delta by 1% of the
stock price and the number of option held. CEO stock equity sensitivity is
calculated by multiplying the number of shares of equity held by 1% of the
stock price. All the other variables are as defined in Table 6, and only the three-
way interaction terms of the control variables are included. F inn-years that
underwent significant M&A activities are deleted. Bold indicates significance
at the 10% level or better. Heteroskedasticity-robust t-statistics are reported.

110

 

Table 13 (cont'd)

 

 

 

 

 

 

(i): FY t (ii): FY t+2

Variable coeff. I t—stat coeff. I t-stat
Intercept 0.00 -0.09 -0.20 -3.86
CFO -0.90 -1.55 0.50 1.55
Asales 0.57 3.68 0.11 1.10
FASSET -0.07 -1.01 0.04 1.11
DCFO 0.04 0.35 0.02 0.24
DCFO*CFO 9.08 4.83 -1.56 -2.52
Aoptionsen_rank -0.05 -1.04 0.06 1.15
Aoptionsen_rank * CFO 0.86 1.51 -0.28 -0.78
Aoptionsen_rank * Asales 0.24 1.22 -0.11 -0.52
Aoptionsen_rank * FASSET 0.04 0.61 0.01 0.19
Aoptionsen_rank * DCFO 0.18 0.52 0.29 2.18
Aoptionsen_rank *DCFO*CFO -8.89 -1.22 2.15 1.61
Aequitysen_rank 0.07 1 . 19 0.23 2.91
Aequitysen_rank * CFO -0.02 -0.03 -0.95 -2.57
Aequitysen_rank * Asales -0.34 -1.62 -0.01 -O.10
Aequitysen_rank * FASSET -0.09 -l .19 -0. 13 -1.87
Aequitysen_rank * DCFO -O.12 -O.66 -O.75 -4.37
Aequitysen_rank *DCFO*CFO -7.41 -1.38 -1.53 -1.24
leverage_rank *DCFO*CFO 18.09 3.42 —17.74 -14.23
Lit *DCFO*CFO -4.20 -0.92 4.19 4.35
MTB_rank *DCFO*CFO -11.95 -2.78 2.59 10.26
Size_rank *DCFO*CFO -O.59 -0.08 6.22 4.38
R-squared 72.42% 64.56%

No. of total observations 138 130

 

 

 

 

111

 

 

 

mmoo- god- mmm.m- Nmm. .- www.mr Evo- momé- 000003005

omoo wood wovo vooo ooo.o .oo.o .ooo.v €30.00: .000 0.0—0.0-3:”? 00880:?»

009$ Xmom $0.0m 005% 0\0N.No 0\0N.oo 0\0m.mo ..0—0.0000 .02—0.50:.0 030.02. 0\0

wood wood o.o.o vooo m.o.o o.o.o o.o.o ..0—0.0000 ..0—00.00:...w 0.0.3:.

mooo mooo- ©.o.o- wood. wood oooo m.o.o ..0—0.0000 ..0—00.0250 .0005.

mom mom m . m Nmm .mm vmm N. m 00.0 ..0 67.
Ni 0+. 0 .-0 Nu. 75 Nu...

 

 

 

 

..0000 x .. 00000000+ 0.0000000...+ .

..

0000.00.00 + 002000000 +

..0

000.00 + 0.021.000... u .0ou<uzm<

”0.020000 .0800: E00 .0300 00003.00. 0000000.... 0... 00 00.0.9008 000 . 000.. .0. . E0... 9030.000 00.. ..0UU<IZm<
0.00.0000 .0885? 030.00., 0.00.00.00.00 0.000 .0 0.0.0 00....0 .00 £0. 00000.00 0... 0.000.800 . .N 0.0.0... .0. 0000.00 00 000 00.0.0.0?

00.. ..< 000020000 000 0:0..030030 o. .000. .0 0.0.3 000000-930... 3:0 .0 000.. 0.0000000006000000 0... 00.. 0.0.. 0.0.3 U.m

..w..0-m 0.. 0.. 0.0.0... we..0.000-.0o0. 00.. ..0 ..o 00.600.000.00 0 0.000.. .0 .000... @5000. 0.000 00.. 30.000000 0000.00.00 000 0.0.0 o. .30

00+ 0000.. 00000000 0+ 00000.00 0+ 000000000 0+ 002000.003 0000.0 0+ 0.0 u ..ou...

”Goomo 00.003.0>.0.m .000 ..0m. >0. .00....on 6.00.00 00.00.. n0.0.0.. 000005.00 000 0.00.0000 .0800 Z .N+. .000 ..+. .0 ..0 .N

-. . .-0. .N.:. 000.. .:0>0 00.. 00.. 00:0... w0.00.000 00. 0.050000 Ab0co..0000..0o .050250 00. 00..0..0.0 b08050 0.000000. 0.0.0. 0......

00:0500:=0:=< 0.00—00000.03— 6.0.5.00 0:00.0— $50.03. .00.. 0.0.0.0030 000.00.00.80;—
v. 0.0.0...

112

APPENDIX A
DEFINITION OF VARIABLES

Following Cheng and Farber (2008), I adopt their model to analyze the change in
CEO option grants. Prior research finds that CEOs’ stock or option ownership is
negatively related to annual CEO option or stock grants: when a CEO’s stock or option
ownership is low, the firm tends to award its CEO more option or stock grants and vice-
versa (Core and Guay 1999; Bryan et al. 2000). To measure CEO ownership, I use the
actual number of shares owned (Shares_Own) and the number of both exercisable
(Exercisable_0pti0ns) and unexercisable options (Unexercisable_Options), all scaled by
shares outstanding. I predict negative signs on the coefficients of these variables.

Core and Guay (1999) argue that the optimal level of equity incentives increases
with firm size. The larger the firm, the more complex it becomes, giving rise to agency
conflicts. Also, CEOs of large firms tend to be wealthier and need more stock-based
compensation to be motivated to work in the interests of shareholders. I therefore predict
a positive sign on the coefficient of Size, which I measure as the natural logarithm of
sales.

Prior literature proposes that because it is difficult for shareholders to determine
the appropriate corporate/operational strategy for a growth firm, it is likely that growth
firms provide their managers with higher equity incentives to align their interests with
those of shareholders (Smith and Watts 1992; Gaver and Gaver 1993; Core and Guay
1999; Hanlon et al. 2003). Consistent with these studies, I use the book-to-market ratio
(B/M) and research and development intensity (R&D) to proxy for growth opportunities. I

predict a negative coefficient on _B/M and a positive coefficient on R&D.

113

 

‘1'—

Compared to cash-based compensation, such as salary and bonus, stock-based
compensation exerts relatively little pressure on a company’s current cash flow. Thus,
firms with cash constraints are more likely to use stock-based compensation (Y ermack
1995; Dechow et al. 1996). I measure cash constraint (Cash_Constraint) as common and
preferred dividends plus cash flows used in investment activities minus cash flows from
operations, divided by total assets. I predict a positive sign on the coefficient of this
variable.

Prior literature argue that firms with earnings constraints (Earn_Constraint) are
also more likely to use option-based compensation, especially prior to FAS 123R (fiscal
2006) when option-based compensation was not required to be expensed. I use a dummy
variable to indicate firms with an operating loss. However, the empirical evidence on the
relation between Earn Constraint and stock-based compensation is mixed (Y ermack 1995;
Core and Guay 1999; Bryan et al. 2000). Thus similar to Cheng and Farber (2008), I
make no directional prediction for the coefficient on this variable.

If a CEO’s stock-based compensation induces risk-taking, then shareholders
receive a benefit over debtholders. It follows that shareholders will bear this debt agency
cost in the form of higher interest. Therefore, stock-based compensation will be
negatively related to debt. In addition, debt financing also serves as a monitoring
mechanism that can reduce the need for stock-based compensation. Consistent with these
arguments, Bryan et al. (2000) identify a negative relation between incentive-intensity
and leverage. I measure Leverage as long-term debt divided by total assets and predict a

negative sign on the coefficient of this variable.

114

 

Prior research finds a positive relation between equity incentives and a firm’s
idiosyncratic risk (Idiosyncratic Risk) (e.g., Core and Guay 1999; Hanlon et al. 2003).
When the uncertainty surrounding a firm’s performance is higher, it is more difficult for
shareholders to monitor managers, thus making it more likely that the firm will use
option-based compensation to motivate managers. I measure idiosyncratic risk as the
standard deviation of the residual from the market model over the prior 12 months and
predict a positive sign on its coefficient.

Prior research finds a positive relation between a firm’s current returns (Current
Return) and CEO compensation (Hanlon et al. 2003), consistent with CEOs being
compensated for good firm performance. Hanlon et al. (2003) also find that firms with
greater lagged stock returns (Past Return) grant more stock options to their CEOs. I use
these variables and predict positive signs on their coefficients.

With greater cash compensation - a proxy for outside wealth -managers can
reduce their risk-aversion through better diversification, thus reducing the need for using
options grants to encourage managers to invest in risky projects (Guay 1999). In support
of this prediction, Hanlon et al. (2003) find that firms with greater CEO cash
compensation (Cash Compensation) grant fewer stock options. However, as proposed in
Cheng and Farber (2008), this relation can be positive as firms might award managers
additional cash compensation to offset the additional risk they bear through increased
option grants. I therefore do not predict a sign on the coefficient on this variable.

Prior studies show that a restatement is frequently associated with CEO turnover
(e.g., Desai et al. 2006). As proposed in Cheng and Farber (2008), new CEOs might

prefer more cash-based compensation in lieu of stock-based compensation because of the

115

risk of further decline in the company’s stock price subsequent to the restatement. On the
other hand, because new CEOs likely have a lower level of ownership in the company
than former CEOs, restating firms may actually award new CEOs more Option grants
than those given to former CEOs. Finally, it is also possible that restating firms are more
likely to reduce option grants for extant CEOs than for new CEOs as a punishment for the
wrongdoings associated with the restatements. But as documented in Cheng and Farber
(2008), both restating firms with CEO turnover and those without CEO turnover are
equally likely to reduce their CEO option grants after the announcements. To test these
effects, I include a dummy variable, ExtantCEO, which takes the value of 1 if the
restating firm retains its CEO during the period examined and 0 otherwise. I predict a
negative sign on the coefficient of this variable as an indication of the punishment effect
associated with the restatements.

Finally, I examine whether the nature of restatements is associated with the
changes in CEO option grants. Specifically, if restatements classified as fraud are
associated with severe agency problem between management and shareholders, I expect
fraud firms are more likely to reduce their CEO option grants after the announcements
than restatements classified as technical restatements. Fraud is a dummy variable taking
the value of 1 if a restatement is classified as fraud and 0 otherwise. I predict a negative

sign on the coefficient of this variable.

116

BIBLIOGRAPHY

Ahmed, A. S., B. K. Billings, R. M. Morton, and M. Stanford-Harris. 2002. The role of
accounting conservatism in mitigating bondholder-shareholder conflicts over dividend
policy and in reducing debt costs. The Accounting Review 77 (4): 867-890.

Ahmed, A. S. and S. Duellman. 2007. Accounting conservatism and board of director
characteristics: an empirical analysis. Journal of Accounting and Economics 43 (2-3):
411-437.

Armstrong, C. S., A. D. Jagolinzer, and D. F. Larcker. 2009. Chief executive officer
equity incentives and accounting irregularities. The Wharton School Working Paper.

Badertscher, B. 2007. Evidence on the pecking order of earnings management. Doctoral
Dissertation. University of Iowa.

Ball, R. 2001. Infrastructure requirements for an economically efficient system of public
financial reporting and disclosure. Brookings-Wharton Papers on Financial Services:
127-182.

Ball, R., S. P. Kothari, and A. Robin. 2000. The effect of international institutional
factors on properties of accounting earnings. Journal of Accounting and Economics 29
(1)21-51.

Ball, R. and L. Shivakumar. 2005. Earnings quality in UK. private firms: comparative
loss recognition timeliness. Journal of Accounting and Economics 39 (1): 83-128.

Ball, R. and L. Shivakumar. 2006. The role of accruals in asymmetrically timely gain and
loss recognition. Journal of Accounting Research 44 (2): 207-255.

Ball, R and L. Shivakumar. 2008. Earnings quality at initial public offerings. Journal of
Accounting and Economics 45 (2-3): 324-349.

Basu, S. 1997. The conservatism principle and the asymmetric timeliness of earnings.
Journal of Accounting and Economics 24. (1): 3-37.

Bergstresser, D. and T. Philippon. 2006. CEO incentives and earnings management.
Journal of Financial Economics 80 (3): 511-529.

Bryan, S., LeeSeok, H., Lilien S., 2000. CEO stock-based compensation: An empirical

analysis of incentive-intensity, relative mix, and economic determinants. Journal of
Business 73(4): 661-693.

117

Bushman, R. and J. Piotroski. 2006. Financial reporting incentives for conservative
accounting: the influence of legal of political institutions. Journal of Accounting and
Economics 42 (1-2): 107-148.

Burns, N. and S. Kedia. 2006. The impact of performance-based compensation on
misreporting. Journal of Financial Economics 79 (1): 35-67.

Chandra, U., C. E. Wasley, and G. B. Waymire. 2004. Income conservatism in the US
technology sector. Simon School Working Paper No. FR 04-01.

Cheng, Q. and D. B. Farber. 2008. Earnings restatements, chanage in CEO compensation,
and frirn performance. The Accounting Review 83 (5): 1217-1250.

Chen, Q., T. Hemmer, and Y. Zhang, 2007. On the relation between conservatism in
accounting standards and incentives for earnings management. Journal of Accounting
Research 45 (3): 1—25.

Cheng, Q. and T. D. Warfield. 2005. Equity Incentives and earnings management. The
Accounting Review 80 (2): 441-476.

Collins, D., A. L. Reitenga, and J. M. Cuevas. 2005. The managerial consequences of
earnings restatements. Working paper. The University of Memphis.

Core, J. and W. Guay. 1999. The use of equity grants to manage optimal equity incentive
levels. Journal of A ccounting and Economics 28 (2): 151-184.

Core, J., W. Guay, and D. Larcker. 2003. Executive equity compensation and incentives:
A survey. FRBNY Economic Policy Review (Apr. 2003): 27-50.

Core, J. and D. Larcker. 2002. Performance consequences of mandatory increases in
executive stock ownership. Journal of Financial Economics 64 (3): 317-340.

Dechow, P., Shoan, and R., Sweeney, A., 1996. Causes and consequences of earnings
manipulation: An analysis of firms subject to enforcement actions by the SEC.
Contemporary Accounting Research 13 (Spring): 1-36.

Desai, H., C. E. Hogan, and M. S. Wilkins. 2006. The reputational penalty for aggressive

accounting: earnings restatements and management turnover. The Accounting Review 81
(1): 83-112.

Efendi, J ., A. Srivastava, and E. P. Swanson. 2007. Why do corporate managers misstate
financial statements? The role of in-the-money options and other incentives. The Journal

of Financial Economics 85(3): 667-708.

Elliott, J. A. and W. H. Shaw. 1988. White-offs as accounting procedures to manage
perceptions. Journal of A ccounting Research 26(Supplement): 91-1 19.

118

Farber, D. B. 2005. Restoring trust after fraud: Does corporate governance matter?. The
Accounting Review 80 (2): 539-561.

Francis, J., D. Philbrick, and K. Schipper, 1994. Shareholder litigation and corporate
disclosures. Journal of Accounting Research 32 (2): 137-164.

Gaver, J. J. and K. M. Gaver, 1993. Addition evidence on the association between the
investment opportunity set and corporate financing, dividend, and compensation policies.
Journal of Accounting and Economics 16: 125-160.

Gaver, J. J ., K. M. Gaver, and J. R. Austin, 1995. Additional evidence on bonus plan and
income management. Journal of Accounting and Economics 19 (1): 3-28.

Giannetti, M. and A. Simonov. 2006. Which investors fear expropriation? Evidence from
investors’ portfolio choices. Journal of Finance 61 (3): 1507-1547.

Givoly, D. and C. Hayn. 2000. The changing time series properties of earnings, cash flow
and accruals. Journal of Accounting and Economics 29 (3): 287-320.

Givoly, D., C. Hayn, and A. Natarajan. 2007. Measuring reporting conservatism. The
Accounting Review 82 (1): 65-106.

Guay, W., 1999. An empirical analysis of the convexity of between stock price and
CEO’s wealth. Journal of Accounting and Economics 53: 43-71.

Guidry, F., A. J. Leone, and S. Rock. 1999. Earnings-based bonus plans and earnings
management by business-unit managers. Journal of Accounting and Economics 26 (1-3):
113-142.

Hanlon, M., Rajgopal, S., and Shevlin, T., 2003. Are executive stock options associated
with fiJture earnings? Journal of Accounting and Economics 36: 3-43.

Healy, P. M. 1985. The effect of bonus schemes on accounting decisions. Journal of
Accounting and Economics 7 (1-3): 85-107.

Hermes, K. M., A. J. Leone, and B. P. Miller. 2008. Management intent and CEO and
CFO turnover around earnings announcements. The Accounting Review 83 (6): 1487-
1519.

Holthausen, R. W., D. F. Larcker, and R. Sloan. 1995. Annual bonus schemes and the
manipulation of earnings. Journal of Accounting and Economics 19 (1): 29-74.

Hribar P. and N. T. Jenkins. 2004. The effect of accounting restatements on earnings

revisions and the estimated cost of capital. Review of Accounting Studies 19 (2-3): 337-
356.

119

Huijgen, C. and M. J. P. Lubberink. 2005. Earnings conservatism, litigation and
contracting: the case of cross-listed firms. Journal of Business Finance & Accounting 32
(7-8): 1275-1309.

Huang, Y. and W. Zhang, 2009. Earnings restatements and subsequent accounting
conservatism. Working paper. University of Kansas.

Jensen, M., 2005. Agency costs of overvalued equity. Financial Management 34, 5-19.

Johnson, S. A., H. E. Ryan, JR., and Y. S. Tian. 2009. Managerial incentives and
corporate fraud: the sources of incentives matter. Review of Finance (2009) 13: 115-145.

Klein, A. and C. A. Marquardt. 2006. Fundamentals of Accounting Losses. The
Accounting Review 81 (1): 179-206.

Kwon, Y. K., D. P. Newman, and Y. S. Sub. 2001. The demand for accounting
conservatism for management control. Review of Accounting Studies 6 (1): 29-51.

Larcker, D. F., S. A. Richardson, and I. Tuna. 2007. Corporate governance, accounting
outcomes, and organizational performance. The Accounting Review 82 (4): 963-1108.

LaFond, R. and S. Roychowdhury. 2008. Managerial ownership and accounting
conservatism. Journal of A ccounting Research 46 (1): 101-135.

LaFond, R. and R. Watts. 2008. The information role of conservatism. The Accounting
Review 83 (2): 417.

LaGore, W. 2008. Conditional and unconditional conservatism following a fincial
reporting failure: an empirical study. Doctoral dissertation. Florida State University.

Moore, E. A. and R. J. Pfeiffer. 2004. The effects of financial statement restatements on
finns’ financial reporting strategies. Working paper. University of Massachusetts.

Moore, M. L., 1973. Management changes and discretionary accounting decisions.
Journal of Accounting Research 11 (1): 100-107.

Murphy, K. J. and J. L. Zimmerman. 1993. Financial performance surrounding CEO
turnover. Journal of Accounting and Economics 16 (1993): 273-315.

Richardson, S. A., A. 1. Tuna, and M. Wu. 2003. Predicting earnings management: the
case of earnings restatements. Working paper. The Wharton School.

Palmrose, Z., V. J. Richardson, and S. Scholz. 2004. Determinants of market reactions to
restatement announcements. Journal of Accounting and Economics 37 (1): 59-89.

120

Palmrose, Z. and S. Scholz, 2004. The circumstances and legal consequences of Non-
GAAP reporting: evidence from restatements. Contemporary Accounting Research 21 (1):
142-180.

Plumlee, Marlene and Teri Yohn, 2009. An analysis of the underlying causes of
restatements. Working paper, University of Utah.

Pourciau, S., 1993. Earnings management and nonroutine executive changes. Journal of
Accounting and Economics 16 (1993): 317-336.

Roychowdhury, S. and R. Watts. 2007. Asymmetric timeliness of earnings, market-to-
book and conservatism in financial reporting. Journal of Accounting and Economics 44
(1): 2-31.

Sloan, R. 1993. Accounting earnings and top executive compensation. Journal of
Accounting and Economics 16 (1-3): 55-100.

Smith, C. and Watts, R., 1992, The investment opportunity set and corporate financing,
dividends, and compensation policies. Journal of Financial Economics 32: 263-292.

Srinivasan, S. 2005. Consequences of financial reporting failure for outside directors:
evidence from accounting restatements and audit committee members. Journal of
Accounting Research 43 (2): 291-334.

Strong, J. S. and J. R. Meyer, 1987, Asset writedowns: managerial incentives and security
returns. Journal of Finance 42(3): 643-661.

Financial Accounting Standards Board, 1978. Statement of Financial Accounting
Concepts No. 1.

US. General Accounting Office (GAO), 2002. Financial statement restatements: trends,
market impacts, regulatory responses, and remaining challenges. GAO-03-138.

US. General Accounting Office (GAO), 2006. Financial statement restatements: update
of public company trends, market impacts, and enforcement activities. GAO-0-678.

Watts, R. 2003. Conservatism in accounting Part I: explanations and implications.
Accounting Horizon 17 (3): 207-221.

Watts, R. and J. L. Zimmerman. 1978. Toward a positive theory of the determination of
accounting standards. The Accounting Review 53 (1): 112-134.

Wilson, W. 2008. The information content of earnings following restatements. The
Accounting Review 83 (2): 519-548.

121

Zhang, J. 2008. The Contracting Benefits of Accounting Conservatism to Lenders and
Borrower. Journal of Accounting and Economics 45 (1): 27-54.

Yerrnack, D., 1995. Do corporation award CEO stock options effectively? Journal of
Financial Economics 39 (2-3): 237-269.

122

        

    

    
  

        

   

ERS TY L BR

63 0200

ARIES

 

mlllllllllllllllllll mill
3 1 2 9 3 0 3 0