Conditional and Unconditional Conservatism Following a Financial Reporting Failure: An Empirical Study

Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2008

Conditional and Unconditional

Conservatism Following a Financial

Reporting Failure: An Empirical Study

William Delton LaGore

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected]

FLORIDA STATE UNIVERSITY

COLLEGE OF BUSINESS

CONDITIONAL AND UNCONDITIONAL CONSERVATISM FOLLOWING A

FINANCIAL REPORTING FAILURE: AN EMPIRICAL STUDY

WILLIAM DELTON LAGORE

A Dissertation submitted to the

Department of Accounting

in partial fulfillment of the

requirements for the degree of

Doctor of Philosophy

Degree Awarded:

Spring Semester, 2008

The members of the Committee approve the dissertation of William Delton LaGore

defended on January 11, 2008.

__________________________

Richard Morton

Professor Directing Dissertation

__________________________

Gary Benesh

Outside Committee Member

__________________________

Bok Baik

Committee Member

__________________________

Bruce Billings

Committee Member

Approved:

_________________________________________

Caryn L. Beck-Dudley, Dean, College of Business

The Office of Graduate Studies has verified and approved the above named committee

members.

ACKNOWLEDGMENTS

I thank and acknowledge God from whom all blessings flow.

I thank my mother and grandparents who have always believed in me, encouraged me

and supported me my entire life. I thank my father for his support and help during my

time at Florida State.

I thank the members of my dissertation committee, Rick Morton (Chair), Bruce

Billings, Bok Baik, and Gary Benesh. You have given me invaluable guidance and

assistance throughout this project and during my time at Florida State.

I especially want to acknowledge my chairman, Dr. Rick Morton, for his invaluable

guidance and assistance. Thank you very much for always being there to help and guide

me through this process.

I thank Rose McGorey for her research assistance in collecting data. I want to thank

Linwood Kearney and my fellow doctoral students for their friendship and

encouragement throughout this lengthy process.

Finally, I would like to thank the faculty members, Nancy LaPorte and Donna

Arnold, the accounting department, the College of Business, and the many people at

Florida State University for giving me the opportunity and assistance to earn a Ph.D.

from Florida State University.

iii

TABLE OF CONTENTS

LIST OF TABLES......................................................................................... vi

ABSTRACT................................................................................................... viii

1. INTRODUCTION AND MOTIVATION ............................................... 1

2. BACKGROUND AND HYPOTHESES DEVELOPMENT................... 7

2.1 Conservatism and Contracting..................................................... 7

2.2 Financial Reporting Failures and Contracting ............................. 9

2.3 Hypotheses Development ............................................................ 12

3. RESEARCH DESIGN............................................................................. 21

3.1 Conditional Conservatism............................................................ 21

3.2 Test of H1 .................................................................................... 22

3.3 Unconditional Conservatism........................................................ 24

3.4 Tests of H2 and H3 ...................................................................... 24

3.5 Tests of H4 through H9................................................................ 29

4. EMPIRICAL RESULTS AND ANALYSIS ........................................... 33

4.1 Sample Selection ......................................................................... 33

4.2 Descriptive Statistics.................................................................... 34

4.3 Tests of Hypothesis 1................................................................... 36

4.4 Unconditional Conservatism........................................................ 37

4.5 Tests of Hypotheses 2a and 2b..................................................... 38

4.6 Tests of Hypotheses 3a and 3b..................................................... 39

4.7 Tests of Hypotheses 4a and 4b..................................................... 41

4.8 Tests of Hypotheses 5a and 5b..................................................... 41

4.9 Tests of Hypotheses 6a and 6b..................................................... 42

4.10 Tests of Hypotheses 7a and 7b................................................... 42

4.11 Tests of Hypotheses 8a and 8b................................................... 43

4.12 Tests of Hypotheses 9a and 9b................................................... 43

4.13 Sensitivity Analysis ................................................................... 44

5. CONCLUSION........................................................................................ 93

5.1 Results ......................................................................................... 93

5.2 Contributions ............................................................................... 94

5.3 Limitations and Suggestions for Future Research ....................... 95

TABLE OF CONTENTS

REFERENCES .............................................................................................. 98

BIOGRAPHICAL SKETCH ......................................................................... 102

LIST OF TABLES

Table 4.1: Sample Attrition....................................................................................48

Table 4.2: Descriptive Statistics ............................................................................49

Table 4.3: Pearson (above the diagonal) and Spearman (below the diagonal)

Correlation Matrix for the Primary Variables......................................52

Table 4.4: Tests of Conditional Conservatism, Pre and Post Restatement for

the Period 1994–2004 ..........................................................................54

Table 4.5: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on the Demand for External Financing for the

Period 1994-2004.................................................................................56

Table 4.6: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on Bonus Compensation for the Period 1994-2004.........59

Table 4.7: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Outside Directors for the Period

1994-2004 ............................................................................................62

Table 4.8: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Audit Committee Meetings for the

Period 1994-2004.................................................................................65

Table 4.9: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Independent Directors on Audit

Committee for the Period 1994-2004 ..................................................68

Table 4.10: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Financial Experts on Audit

Committee for the Period 1994-2004 .................................................71

Table 4.11: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Top Management for the

Period 1994-2004................................................................................74

Table 4.12: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Auditor for the Period 1994-2004..........77

LIST OF TABLES

Table 4.13: Tests of Conditional Conservatism, Pre and Post Restatement,

Separated into Pre- and Post-SOX Periods for the Period

1994–2004............................................................................................80

Table 4.14: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on Outside Director Percentage for the Period

1994-2004 ...........................................................................................83

Table 4.15: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on Audit Committee Meeting Variables for Year

t+3

for the Period 1994-2004 ....................................................................86

Table 4.16: Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on the Financial Expert Percentage on Audit

Committee for the Period 1994-2004 .................................................89

Table 4.17: Tests of Conditional Conservatism, Pre and Post Restatement,

Industry-Adjusted for the Period 1994-2004 for the Period

1994 – 2004.........................................................................................92

vii

viii

ABSTRACT

The quality of financial reporting has come under increased scrutiny in recent years

because of high-profile financial reporting failures, such as Enron and WorldCom, and

the significant increase in the number of financial restatements. While most of the

attention following these financial reporting failures has focused on the capital market

consequences and the negative impact on investor confidence, very little analysis has

adopted a contracting perspective. As accounting information is often used in contracts,

the reliability and quality of the accounting information plays a critical role. A financial

reporting failure casts doubt on the quality of the financial information, and consequently,

leads to greater contracting uncertainty. I posit one means of restoring credibility of the

accounting numbers following a financial reporting failure is to take a more conservative

approach in reporting firm performance. Therefore, this study investigates whether firms

report more conservatively following a financial restatement in order to restore the trust

and confidence of the contracting parties and to protect the interests of the contracting

parties.

I specifically examine whether firms report more conservatively in the three years

following a restatement announcement, and whether an increase in conservatism is

related to the extent of debt and compensation contracts. I then examine various

corporate governance variables as possible explanations for an increase in conservatism

following a restatement. I find restating firms significantly increase conditional

conservatism in the three year period following the restatement announcement year. In

addition, I find that both the level and change in conditional conservatism following a

restatement is greater for firms where debt and compensation contracting are more

important. These results are consistent with the contracting explanation of conservatism

provided in Watts (2003). Lastly, I find little support for corporate governance variables

as explanations for the increase in conservatism.

CHAPTER 1

INTRODUCTION AND MOTIVATION

The quality of financial reporting has come under increased scrutiny in recent years

because of high-profile financial reporting failures, such as Enron and WorldCom, and

the significant increase in the number of financial restatements. An October 2002

General Accounting Office (GAO) report documents that the number of financial

restatements has increased 145 percent from 1997 to 2001 and that publicly traded

companies lost billions in market capitalization in the days and months following a

restatement announcement. The GAO report further concludes that the increase in

restatements has negatively impacted investor confidence. For example, the GAO’s

October 4, 2002 letter to Senator Paul Sarbanes states the following:

“The growing number of restatements and mounting questions about certain corporate

accounting practices appear to have shaken investors’ confidence in our financial

reporting system... empirical research studies and academic experts generally suggest

accounting issues have negatively affected overall investor confidence and raised

questions about the integrity of U.S. markets.”

While most of the attention has focused on the increased restatement activity from a

capital market perspective, very little analysis has adopted a contracting perspective.

Financial reporting serves an important role in contracting as accounting information is

often used in the contract terms and in monitoring of the contract terms (Watts and

Zimmerman 1986). For contracts to be enforceable, the accounting information used in

the contracts must be reliable (Watts 2003a). A financial reporting failure

undermines

the reliability of the accounting numbers leading to greater contracting uncertainty. Like

investors, contracting parties likely lose confidence in the firm’s financial reporting, and

may view future financial reports as less credible. This becomes a significant problem

I define a financial reporting failure as an accounting fraud or restatement. A fraud is defined as

“intentional non-GAAP financial reporting” (Palmrose, Richardson, and Scholz 2004). A restatement is an

admission that previously issued financial statements were not in accordance with GAAP (Palmrose and

Scholz 2004), and can occur for a variety of reasons, ranging from fraud to an honest mistake.

for the firm, as contracting parties need to have confidence and trust in a firm’s financial

reporting process.

I expect that firms will attempt to mitigate the increased contracting uncertainty

following a financial reporting failure. One means of restoring credibility of the

accounting numbers is to take a more conservative approach in reporting firm

performance. Therefore, this study investigates whether firms report more conservatively

following a financial reporting failure in order to restore the trust and confidence of

contracting parties and to protect their interests. Specifically, I examine accounting

conservatism for a sample of restatement firms using both “levels” and “changes”

analyses. I identify restatement firms from the GAO’s 2002 database of financial

statement restatements for the period 1997 - 2001.

Watts (2003a) suggests conservatism is beneficial to contracting parties because it

decreases the probability of distributions that violate contracts and reduce firm value.

Accordingly, managers of firms with reporting failures can voluntarily report

conservatively as a “signal” to contracting parties that their interests are protected and the

financial reports are credible. Alternatively, contracting parties may demand more

conservative reporting following a reporting failure in order to ensure their interests are

protected. Both alternatives lead to the same prediction – firms report more

conservatively following a financial reporting failure.

Recent studies describe two types of conservatism, conditional and unconditional

(e.g. Ball and Shivakumar 2005; Beaver and Ryan 2005). Beaver and Ryan (2005)

define unconditional conservatism as when net assets are understated due to

“predetermined aspects of the accounting process.” They define conditional

conservatism as when net assets are “written down under sufficiently adverse

circumstances, but not up under favorable circumstances.” Conditional conservatism

corresponds to the Basu (1997) notion of conservatism, that is, earnings reflect ‘bad

news’ more quickly than ‘good news’. Basu (2005) describes the key distinction between

the two forms of conservatism as follows: unconditional conservatism utilizes

information known at the inception of assets and liabilities; whereas, conditional

conservatism utilizes new information as it is received in future periods.

Ball and Shivakumar (2005) argue that unconditional conservatism does not enhance

contracting efficiency, as they state it is “inefficient or at best neutral in contracting.”

More specifically, unconditional conservatism is just a downward accounting bias that

investors and other contractual parties can adjust for ex ante; whereas, conditional

conservatism protects contracting parties as book values are written down when

economic losses become known. Therefore, conditional conservatism has the potential to

improve contract efficiency for both debt and compensation contracts. For debt

contracts, recognizing bad news in earnings sooner than good news protects lenders as it

more quickly triggers debt covenant violations. For compensation contracts, conditional

conservatism makes it less likely that managers receive excess compensation due to

upwardly biased financial information. Therefore, managers would likely be motivated

to increase conditional conservatism following a financial reporting failure in order to

reduce contracting uncertainty and restore credibility of the accounting numbers. In

addition, these incentives would likely depend on the extent of contracting and the

corporate governance environment.

Whereas conditional conservatism may have originated to facilitate contracting, Basu

(2005) suggests that unconditional conservatism arose mainly for regulatory or tax

reasons, such as accelerated depreciation for tax purposes or immediate expensing of

R&D expenses as required by FASB. The regulatory or GAAP-based rules that lead to

unconditional conservatism are not affected by a restatement; thus, it is unlikely that

unconditional conservatism would change following a financial reporting failure.

In contrast to the above arguments, there are still strong economic incentives for firms

to report aggressively, such as to meet or beat analyst forecasts or to increase

compensation. Thus, it is an empirical question whether or not firms report more

conservatively following a financial reporting failure. The first research objective of this

paper is to determine if there is a change in conditional and unconditional conservatism

following one type of financial reporting failure, a restatement. Given the need to restore

credibility to the reporting process, I predict that conditional conservatism increases

following a financial restatement. To test this prediction, I examine conditional

conservatism in restating firms for the three years before and after the announcement year

of a financial reporting failure. Conditional conservatism is calculated based on the Basu

(1997) asymmetric timeliness of earnings perspective of conservatism. Unconditional

conservatism is measured as the Basu (1997) regression intercepts, as suggested in Basu

(2005). I make no prediction regarding a change in unconditional conservatism as I

expect the restatement to have no effect on this type of conservatism.

I find restating firms significantly increase conditional conservatism in the three year

period following the restatement announcement. These findings suggest firms

proactively take steps to restore trust and confidence in their financial reporting following

a restatement. I find no change in unconditional conservatism following the restatement

announcement. This is consistent with unconditional conservatism being a downward

accounting bias not affected by a restatement.

As conditional conservatism has the potential to improve contract efficiency, the

second research objective of this paper is to examine whether the increase in conditional

conservatism following a restatement is related to a firm’s contracting. I find the increase

in conditional conservatism following a restatement is greater for firms where debt and

compensation contracting are more important. Based on sensitivity tests, the results

relating to compensation contracting are relatively weaker than the results for debt

compensation. Overall, these findings suggest firms report more conservatively in order

to reduce contracting uncertainty and restore the trust and confidence of the contracting

parties.

An increase in conditional conservatism following a financial reporting failure may

be linked to various intermediate changes in a firm’s financial reporting environment,

such as changes in management, board of directors, audit committee, or the auditor. For

example, Farber (2005) provides evidence that firms take actions to improve corporate

governance following a fraud; such as increasing the number of audit committee

meetings and the number and percentage of outside board members. However, very little

research examines whether these actions lead to an increase in conservatism. Therefore,

the third research objective of this paper is to examine the determinants of a change in

accounting conservatism following a restatement by including variables relating to

changes in management, board of directors, audit committee, and audit firm. This

analysis provides evidence on whether changes in corporate governance mechanisms

following a restatement are associated with an increase in conservatism.

Consistent with Farber (2005), I find firms take actions to improve corporate

governance following the restatement. However, I find these changes in corporate

governance mechanisms are not related to the increase in conservatism following a

restatement. Surprisingly, improving corporate governance mechanisms did not lead to a

change in how firms report financial performance following a restatement. Thus, my

findings suggest that greater levels of contracting, and not necessarily changes in

corporate governance, motivate an increase in conservatism following a restatement.

This study contributes to the literature on restatements and accounting conservatism

in several ways. First, the study provides a potential explanation for the findings in

Moore and Pfeiffer (2004). They find no change in conservatism following financial

statement restatements and conclude that restatement firms do not reform their financial

reporting. However, they use total accruals as their primary measure of conservatism and

make no attempt to isolate conditional conservatism. This study may help explain why

they find no change in conservatism following restatements, as I find it is only

conditional conservatism that changes following a restatement.

This study also contributes to the financial reporting failure literature as Farber (2005)

finds that fraud firms take measures to improve corporate governance following a fraud;

however, he does not address the issue of conservatism. I find no evidence that

improvements in corporate governance following a restatement affect a firm’s financial

reporting choices.

Finally, this study furthers our understanding of accounting conservatism in at least

three ways. First, this study provides evidence regarding the contracting explanation for

conservatism, as I examine the impact of debt and compensation contracting on firms’

financial reporting choices. My findings are consistent with the contracting explanation

of conservatism. Second, I empirically examine changes in the two types of

conservatism, conditional and unconditional, following a financial reporting failure. As

expected, I find only conditional conservatism changes following a restatement. Third,

this study provides insight into what may be the driving force(s) behind a change in

conservatism by examining changes in the financial reporting environment following a

financial reporting failure. Knowledge of which monitoring mechanisms (i.e. board of

directors, audit committee, and auditor) lead to increased conservatism is important to

regulators and contracting parties. Surprisingly, I find no relation between changes in

corporate governance and a change in conservatism following a restatement.

The rest of the paper is organized as follows. Chapter 2 reviews the literature and

develops hypotheses. Chapter 3 describes the research design, including the data sources,

methodology, and hypotheses testing. Chapter 4 presents the empirical results and

analysis. Chapter 5 offers concluding remarks, along with limitations and suggestions for

future research.

CHAPTER 2

BACKGROUND AND HYPOTHESES DEVELOPMENT

2.1 Conservatism and Contracting

Although there is no agreed-upon formal definition of conservatism, there have been

several definitions of conservatism brought forth in the literature. Watts (2003a) notes

early accounting textbooks, such as Bliss (1924), define conservatism as “anticipate no

profit, but anticipate all losses.” Givoly and Hayn (2000) define conservatism as

selecting accounting principles that minimize “cumulative reported earnings by slower

revenue recognition, faster expense recognition, lower asset valuation, and higher

liability valuation.” Basu (1997) views conservatism as when earnings reflect ‘bad news’

more quickly than ‘good news’. Watts (2003a) ultimately defines conservatism as the

“differential verifiability required for recognition of profits versus losses,” which is

consistent with the Basu (1997) definition. This definition means there is a higher degree

of verification required for recognizing good news as gains than for recognizing bad

news as losses (Basu 1997). Even though there is no single agreed-upon definition of

conservatism, one common result arises from conservative reporting - conservatism leads

to a persistent understatement of net assets.

As for the reasons for the existence of accounting conservatism, Watts (2003a)

suggests four potential explanations: contracting, shareholder litigation, taxation, and

accounting regulation. He argues that contracting is the most likely explanation for the

origins of accounting and conservatism centuries ago because only contracting existed

centuries ago, whereas litigation, taxation, and accounting regulation have only occurred

more recently.

Penman (2003) describes the important role accounting serves in contracting as

follows:

“…accounting is concerned with the division of property rights between

shareholders and bondholders, between shareholders and management, and

between shareholders and auditors – and with the efficient contracting over those

property rights that is so essential for the functioning of an economy.”

Holthausen and Watts (2001) find empirical evidence consistent with the contracting

explanation of conservatism. They find conservatism existed before formal standard

setting and before litigation became a major factor in the late 1960’s.

Watts (2003a) explains why conservatism arose for contracting purposes as follows.

In debt contracts, conservatism makes it less likely that management upwardly biases

earnings and “make what is effectively a liquidating dividend payment to shareholders at

the expense of debt holders”. Ahmed, Billings, Morton, and Stanford-Harris (2002)

provide empirical evidence of the role conservatism plays in debt contracting. They

argue that conservatism mitigates the conflicts of interest over dividend policy between

shareholders and bondholders. They find that firms with more severe conflicts of interest

over dividend policy use more conservative accounting, and that conservatism is

associated with a lower cost of debt. For compensation contracts, conservatism reduces

the likelihood that management overstates financial measures in order to take excess

compensation. Conservative earnings tend to defer a manager’s earnings-based

compensation until the full effects of the manager’s actions are recognized in earnings

(Holthausen and Watts 2001). Therefore, this study focuses on two contracting reasons

for conservatism: debt and management compensation contracts.

Conditional and unconditional conservatism are two distinct types of conservatism

recognized in recent literature (Ball and Shivakumar 2005; Beaver and Ryan 2005).

Beaver and Ryan (2005) define unconditional conservatism as when net assets are

understated due to “predetermined aspects of the accounting process.” This type of

conservatism is also known as “ex ante” or “news-independent” conservatism. Beaver

and Ryan (2005) describe accelerated depreciation and immediate expensing of R&D as

examples of unconditional conservatism. The subsequent write-off of assets under this

type of conservatism will be independent of future events, thus the term “news-

independent.” They define conditional conservatism as when net assets are “written

down under sufficiently adverse circumstances, but not up under favorable

circumstances.” This type of conservatism is also known as “ex post” or “news-

dependent” conservatism. Examples of conditional conservatism given by Beaver and

Ryan (2005) are the lower of cost or market rule for inventory and impairment

accounting of long-term tangible and intangible assets, such as impairment accounting for

goodwill. The write-off of assets under conditional conservatism is triggered by future

events, thus the term “news-dependent.”

As for the two types of conservatism, Ball and Shivakumar (2005) argue that only the

conditional form of conservatism should affect contract efficiency because it incorporates

new information that could produce contracting responses. The unconditional form of

conservatism does not employ new information and is merely a downward accounting

bias that contracting parties can adjust for ex ante. Contract terms can be written to undo

the effect of accounting methods. For example, Ball and Shivakumar (2005) point out

that if book values of assets are systematically understated by a known amount, then debt

covenants can require an increase to the book values by this amount when determining

leverage and the amount a firm could borrow. Basu (2005) points out that historical

evidence suggests unconditional conservatism arose as a response to regulatory or tax

incentives. For example, Basu (2005) suggests the early development of conservative

depreciation methods was influenced by corporate income taxes as firms attempted to

maximize the depreciation deduction. In summary, unconditional conservatism serves a

different function than conditional conservatism from a contracting perspective, as only

conditional conservatism has the potential to improve contract efficiency. Conditional

conservatism allows for contracting responses to new information, such as bad news in

earnings triggering a debt covenant violation. Conversely, unconditional conservatism

does not allow for contracting responses to new information because it is a downward

accounting bias independent of new information.

2.2 Financial Reporting Failures and Contracting

Financial reporting failures include both frauds and restatements. For the purpose of

this study, I examine only accounting restatements. Restatements are an admission that

previously issued financial statements were not in accordance with GAAP (Palmrose and

Scholz 2004). Early research focused on characteristics of restating firms. For example,

Kinney and McDaniel (1989) find restatement firms are smaller, less profitable, have

higher debt, and are slower growing. DeFond and Jiambalvo (1991) find earnings

overstatements are more likely for firms with diffuse ownership and lower growth in

earnings, and less likely for firms with audit committees.

Recent studies document significant negative economic consequences related to

financial reporting failures. Palmrose et al. (2004) find a mean abnormal return of -9.2

percent in the two-day window (day 0, 1) around a restatement announcement, with more

negative returns for restatements involving fraud (-20 percent). Palmrose and Scholz

(2004) find the negative market reaction is greater for restatements of core earnings (i.e.

pre-tax earnings from primary operations) than for non-core earnings (i.e. all other

earnings). Anderson and Yohn (2002) document the long-term economic consequences

of restatements by finding average cumulative abnormal returns of -7.97 percent for the

period from three days before the restatement announcement through three days after the

restatement filing with the SEC.

There are also legal consequences to a financial reporting failure, as Palmrose and

Scholz (2004) find that 38 percent of the companies in their restatement sample

subsequently faced civil litigation. They find companies with restatements of core

earnings (primarily revenue restatements) and pervasive restatements (i.e. more than one

accounting item restated) are more likely to be subject to litigation.

A financial reporting failure also damages the reputation of the firm, auditors,

management, and the board of directors. For example, Srinivasan (2005) finds outside

board members experience significant reputational costs following accounting

restatements. Srinivasan finds significant turnover of board members in the three years

following the restatement, including director turnover for 48 percent of firms that restate

earnings downward. The likelihood of director turnover increases if the board member is

also on the audit committee. The study also finds outside directors lose positions on

other boards following a restatement.

Another study, Desai, Hogan, and Wilkins (2006) examines the reputational penalties

to managers of restating firms, and finds that 60 percent of restating firms experience

management turnover in the two years following a restatement as compared with 35

percent for a control sample. An audit firm’s reputation can be damaged by a financial

reporting failure, as evidenced by the demise of Arthur Andersen. Barton (2005)

examines the demand for auditor reputation by examining the client defections from

Arthur Andersen. Barton (2005) finds firms that are more visible in the capital markets

switched sooner to another Big 5 auditor, as they were concerned about their auditor’s

reputation and the credibility of their financial reporting.

As discussed above, accounting numbers used in contracts must be verifiable for the

contract to be enforceable in court (Watts 2003a). Based on prior literature, it is

reasonable to assume that a financial reporting failure leads to greater uncertainty about

the reliability and verifiability of the accounting numbers used in contracts. As a

restatement casts doubt on the quality of the financial reports and increases the risk to the

contracting parties, shareholders and lenders will demand an increased risk premium

following a reporting failure. For example, empirical studies find that frauds and

accounting restatements lead to an increased cost of capital (e.g. Dechow, Sloan and

Sweeney 1996; Hribar and Jenkins 2004). Investors demand a higher rate of return to

compensate for the perceived riskiness of the firm due to less reliable accounting

numbers following a reporting failure.

As for debt contracts, Sengupta (1998) suggests one factor likely used by lenders in

calculating default risk is the quality of financial reporting. Sengupta (1998) finds firms

with high disclosure quality ratings from financial analysts are charged a lower cost of

debt, and the importance of disclosure is greater when there is greater market uncertainty

as measured by the variance of stock returns. Thus, lenders likely demand a higher risk

premium following a reporting failure in part due to the perceived decrease in quality of

the accounting reports.

The risk premium demanded by shareholders and debt holders also increases

following a reporting failure because of the increased uncertainty about the future

profitability and economic prospects of restatement firms. Palmrose et al. (2004) find a

significant downward revision in earnings forecasts following restatements and find a

significant increase in analyst forecast dispersion (a proxy for uncertainty). Hribar and

Jenkins (2004) find accounting restatements lead to decreases in expected future

earnings.

In summary, restatements can have numerous negative effects. These include

economic losses to investors; damage to the reputations of the firm, auditors,

management, and the board of directors; an increase in the cost of capital; and a negative

impact on future earnings power.

2.3 Hypotheses Development

As discussed in the previous section, there are significant negative economic, legal,

reputational, and contractual consequences to a financial reporting failure. A financial

reporting failure is evidence that previously issued accounting reports were incorrect,

thus creating uncertainty about the credibility and verifiability of financial reports after

the reporting failure. This uncertainty likely increases the perceived information

asymmetry between contracting parties (i.e. management, debt holders, and

shareholders). Watts (2003a) suggests conservative accounting is one way to address the

moral hazard problem created by the information asymmetry between contracting parties.

This is because conservatism constrains management and protects the interests of the

contracting parties. In debt contracts, Watts (2003a) suggests conservatism decreases the

probability that management makes a “liquidating dividend payment to shareholders at

the expense of debt holders.” For compensation contracts, conservatism reduces the

likelihood that management overstates financial measures in order to take excess

compensation.

I take the Ball and Shivakumar (2005) view that only the conditional form of

conservatism affects contracting efficiency. They view unconditional conservatism as a

downward accounting bias that contracting parties can adjust for ex ante. When new

information comes along, this form of conservatism does not react to the news. Thus,

there is no advantage from an efficient contracting perspective for firms to become more

unconditionally conservative following a reporting failure. Therefore, I expect no change

in unconditional conservatism following a financial reporting failure.

However, under the conditional form of conservatism, firms exercise judgment in

how they react to and incorporate new information into their financial reporting, which

offers ex post protection to contracting parties. The primary example of conditional

conservatism is impairment accounting. Riedl (2004) models asset impairments as a

“function of economic factors and reporting incentives.” While the economic factors that

lead to asset impairments may be beyond the firms’ control, firms do exercise significant

judgment and discretion when reporting asset impairments. Managers can exercise

discretion over asset impairments because SFAS No. 144 (which superceded SFAS No.

121) requires firms to forecast future performance in order to estimate undiscounted

future cash flows. This estimation of future cash flows requires significant judgment,

estimates, and assumptions, which allows firms to behave more or less conservatively.

Therefore, firms have the ability to become more conservative and improve contract

efficiency following a reporting failure. For example, in debt contracting, conservatism

benefits borrowers as they can borrow at lower interest rates (Zhang 2004; Ahmed et al.

2002), and benefits lenders as conservatism more quickly triggers debt covenant

violations (Zhang 2004). Conservatism can lead to a lower cost of capital as

shareholders’ interests are more protected from managers taking excess compensation.

As discussed above, lenders and shareholders demand a greater risk premium

following a reporting failure because of the uncertainty surrounding the credibility of the

financial reports and the uncertainty about the future economic prospects of the firm. I

predict firms increase conditional conservatism following a financial reporting failure in

order to reduce contracting uncertainty, offset or minimize increases in the cost of debt

and capital, and to restore trust and confidence in their financial reports. This results in

the following hypothesis, stated in the alternate form:

H1: Restatement firms increase conditional conservatism following a financial

reporting failure.

Watts (2003b) summarizes the empirical literature on conservatism and finds results

consistent with all four explanations (i.e. contracting, litigation, taxation, and regulation),

but suggests contracting is the most important explanation for accounting conservatism.

This is because only contracting can explain the existence of conservatism centuries ago,

as litigation, taxation, and accounting regulation have only occurred more recently.

As for debt contracting, agency theory predicts a moral hazard problem created by the

information asymmetry between shareholders and debt holders (Jensen and Meckling

1976). Watts (2003a) suggests conservative accounting is one way to address this moral

hazard problem. Watts suggests that in debt contracts, conservatism reduces the

probability that management will “overstate earnings and assets, and make what is

effectively a liquidating dividend payment to shareholders at the expense of debt

holders.” Accounting conservatism reduces the earnings and net asset measures used in

debt contracts, and thus more quickly triggers debt covenant constraints. Therefore, debt

holders demand conservative accounting in order to protect their interests and ensure

payment of the loan. As a result, I expect that the more important debt contracting is to a

firm, the more conservative it will be following a reporting failure and the greater the

increase in conservatism relative to before the failure. This results in the following

hypotheses concerning both the level and change in conservatism:

H2a: Post restatement, conditional conservatism is greater for firms who have

more of a demand for debt contracting.

H2b: An increase in conditional conservatism following a restatement is greater for

firms who have more of a demand for debt contracting.

As for compensation contracts, agency theory predicts moral hazard problems related

to the separation of ownership and control (Jensen and Meckling 1976). This moral

hazard problem arises from the information asymmetry between the principal

(shareholders) and the agent (managers). The bonus hypothesis predicts managers of

firms with bonus plans are more likely to bias accounting numbers upward to increase

their compensation (Watts and Zimmerman 1986). Watts (2003a) suggests this upward

bias leads to excessive management compensation, which reduces shareholder value.

Watts (2003a) further suggests that in compensation contracts, conservatism reduces the

likelihood that management overstates earnings and net assets in order to take excess

compensation. Conservative earnings defer a manager’s earnings-based compensation

until the full effects of the manager’s actions are recognized in earnings (Holthausen and

Watts 2001). Consequently, I expect shareholders of firms with earnings-based

managerial compensation plans to demand more conservative accounting to offset this

upward bias. As a result, I predict the more important performance-related compensation

contracting is to a firm, the more conservative it will be following a reporting failure and

the greater the increase in conservatism relative to before the failure. This leads to the

following hypotheses concerning both the level and change in conservatism:

H3a: Post restatement, conditional conservatism is greater for firms with more

bonus compensation.

H3b: An increase in conditional conservatism following a restatement is greater for

firms with more bonus compensation.

To my knowledge there is only one study, Moore and Pfeiffer (2004), which

examines whether a firm reports more conservatively following a restatement. They

examine several measures to proxy for a firm’s financial reporting aggressiveness,

including total accruals, working capital accruals, patterns in earnings, and patterns in

analysts’ forecast errors. They measure these proxies before and after the restatement

and find no change in accounting aggressiveness. They conclude that firms do not

reform their financial reporting by reporting more conservatively following a restatement.

However, the Moore and Pfeiffer study does not mention the two types of conservatism

and does not use the Basu (1997) measure of conditional conservatism. Givoly, Hayn,

and Natarajan (2007) find the Basu measure of conservatism is negatively correlated with

other measures of conservatism, such as the market-to-book ratio, conservatism due to

investments in positive net present value projects and accounting rules (Easton and Pae

2004), and conservatism as measured as the sum of three reserve components: inventory,

R&D and advertising (Penman and Zhang 2002). Givoly et al. (2007) suggest the Basu

measure of conditional conservatism is capturing a unique aspect of conservatism not

captured by the other measures. This study may help explain Moore and Pfeiffer’s lack

of results as H1 predicts only conditional conservatism changes after a restatement. In

addition, this study views contracting as the primary explanation for conservatism, and

thus H2 and H3 provide more powerful tests by examining whether a change in

conservatism following a reporting failure is greater for firms where contracting is more

important.

Conservative reporting following a financial reporting failure may be the result of a

chain of events. It would be insightful to examine the antecedents to conservatism

following a restatement. Restatement firms typically receive increased scrutiny from

regulators and experience significant operational, structural, and corporate governance

changes, such as turnover in management, board members, audit committee, and auditors.

These changes may be mechanisms that constrain management’s opportunistic behavior

and lead to more conservative reporting. The importance of enhancing corporate

governance to improve financial reporting practices can be explicitly seen in the recent

passage of the Sarbanes-Oxley Act of 2002 (SOX). SOX contains many provisions to

advance corporate governance with the explicit purpose to “protect investors by

improving the accuracy and reliability of corporate disclosures made pursuant to the

securities laws, and for other purposes” (SOX 2002). Therefore, I next examine the

determinants of an increase in conditional conservatism following a financial reporting

failure.

Beasley (1996) finds that firms with more outside board members significantly

reduce the likelihood of fraud. So a likely response a firm can take following a financial

reporting failure is to increase the number of outside members on the board of directors.

This response can be seen in Farber (2005), who finds that fraud firms increase the

number and percentage of outside directors in the three years following the fraud.

Farber’s study does not examine accounting conservatism; therefore, it is still an open

empirical question as to whether an improvement in corporate governance mechanisms

following a reporting failure leads to an increase in conservatism.

The literature on board composition and financial reporting has primarily focused on

earnings management, and not conservatism. For example, Klein (2002) finds a negative

relation between board independence and abnormal accruals and concludes a more

independent board is more effective in monitoring financial reporting. One study,

Beekes, Pope, and Young (2004) directly examines the relation between board

composition and conditional conservatism in the U.K. They use the Basu (1997) method

to measure conditional conservatism and find that U.K. firms with a higher proportion of

outside board members report more conservatively. Based on the findings of these

studies, I expect firms that increase the number of outside directors following a reporting

failure report more conservatively and have a greater increase in conservatism relative to

before the failure. This leads to the following hypotheses:

H4a: Post restatement, conditional conservatism is greater for firms that increase

the number of outside directors.

H4b: An increase in conditional conservatism following a restatement is greater for

firms that increase the number of outside directors.

Abbott, Parker, and Peters (2004) examine how certain audit committee

characteristics are associated with both fraud and restatement companies. They find a

fraud/restatement is less likely when the audit committee is more independent, meets at

least four times a year, and includes at least one member with financial expertise. So a

likely response to a fraud or restatement is to improve the monitoring function of the

audit committee by having more independent members, meetings, and financial experts

on the committee.

Farber (2005) provides evidence of this as he finds fraud firms significantly increase

the number of audit committee meetings in the three years after the fraud. However,

Farber (2005) finds no significant change in the number of independent members or

financial experts on the audit committee following a fraud. This is surprising based on

the results of Abbott et al. (2004), who find fraud is less likely for firms with audit

committees that are more independent and have at least one financial expert.

Furthermore, DeFond, Hann, and Hu (2005) find a positive market reaction to the

appointment of accounting financial experts to the audit committee. Since the market

places value on audit committee members who are financial experts, it would be

reasonable to expect a firm to increase the number of financial experts following a

reporting failure in order to restore investor confidence. Therefore, I predict an audit

committee that is more diligent, more independent, and has more financial expertise

following a reporting failure demands more conservative financial reporting. These firms

will exhibit a greater level and greater change in conservatism following a restatement.

This leads to the following hypotheses:

H5a: Post restatement, conditional conservatism is greater for firms that increase

the number of audit committee meetings.

H5b: An increase in conditional conservatism following a restatement is greater for

firms that increase the number of audit committee meetings.

H6a: Post restatement, conditional conservatism is greater for firms that increase

the number of independent directors on the audit committee.

H6b: An increase in conditional conservatism following a restatement is greater for

firms that increase the number of independent directors on the audit

committee.

H7a: Post restatement, conditional conservatism is greater for firms that increase

the number of financial experts on the audit committee.

H7b: An increase in conditional conservatism following a restatement is greater for

firms that increase the number of financial experts on the audit committee.

Desai et al. (2006) find significant reputational penalties to managers of restatement

firms. They find that 60 percent of restatement firms experience turnover of at least one

top manager in the two years following the restatement, and they find 85 percent of

displaced managers are unable to find comparable employment. In addition, research

finds a change in top management is associated with write-offs or “big bath” charges

(Pourciau 1993; Francis, Hanna, and Vincent 1996). One interpretation of these findings

is that new management is taking big bath charges and attributing them to prior

management. Alternatively, Riedl (2004) suggests these write-offs “may reflect true

economic changes,” such as a change in business strategy or increased attention to

existing assets. Both alternatives suggest that a change in top management leads to more

conservative reporting. Based on these results, I predict new managers of restatement

firms report more conservatively in order to protect their reputations and “signal” to

contracting parties that their interests are protected and the financial reports are credible.

This prediction, with respect to both the level and change in conditional conservatism, is

hypothesized as follows:

H8a: Post restatement, conditional conservatism is greater for firms with a change

in top management.

H8b: An increase in conditional conservatism following a restatement is greater for

firms with a change in top management.

Farber (2005) finds that fraud firms are more likely than a control sample to switch

auditors following a fraud. It seems reasonable to believe that the new auditor of a

restatement firm brings increased skepticism over the reliability and credibility of the

firm’s financial reports. The new auditor may require more conservative reporting in

order to protect their reputation and to reduce the likelihood of litigation and the

reoccurrence of a fraud or restatement. For example, empirical studies find that auditors

can mitigate their litigation risk by requiring their clients to report more conservatively

(e.g. Cahan & Zhang 2006; Krishnan 2005). Specifically, Cahan and Zhang (2006) find

the former Arthur Andersen clients had lower levels of abnormal accruals and greater

decreases in abnormal accruals in the first year of being audited by the successor audit

firm. Francis and Krishnan (1999) find that auditors respond to the increased risk of

high-accrual firms by reporting more conservatively through the issuance of more

modified audit reports. DeFond and Subramanyam (1998) find that negative

discretionary accruals are greater for clients that pose the greatest litigation risk.

Therefore, I expect the new auditor of a restating firm to demand more conservative

reporting in order to protect their reputation and reduce the likelihood of litigation. This

leads to the following hypotheses concerning both the level and change in conservatism:

H9a: Post restatement, conditional conservatism is greater for firms with a change

in auditor.

H9b: An increase in conditional conservatism following a restatement is greater for

firms with a change in auditor.

In summary, Hypotheses 4 through 9 predict changes in several corporate governance

variables are related to both the level and change in conditional conservatism following a

restatement. The corporate governance variables included in this study are the number of

outside directors, the number of audit committee meetings, the number of independent

directors and financial experts on the audit committee, and the change in top management

and audit firm.

CHAPTER 3

RESEARCH DESIGN

This chapter presents the research design to test the hypotheses presented in the

previous chapter. Section 3.1 first describes the Basu (1997) methodology of capturing

conditional conservatism, which is a basis of my analysis. Section 3.2 describes an

alternative model to Basu’s model developed to test the first hypothesis. Section 3.3

describes the test of unconditional conservatism. Section 3.4 describes the tests of the

contracting hypotheses, H2 and H3. Finally, Section 3.5 describes the methodology used

to test hypotheses 4 through 9.

3.1 Conditional Conservatism

This study relies on a measure of conditional conservatism to investigate financial

reporting changes following a restatement. I adopt the Basu (1997) approach for

identifying conditional conservatism, under the assumption that earnings reflect ‘bad

news’ more quickly than ‘good news’. This approach is consistent with the Watts

(2003a) view of conservatism, which he defines as the “asymmetrical verification

requirements for gains and losses.” Therefore, conditional conservatism in this study is

based on the Basu (1997) model as follows:

it-1

= α

+ α

D + β

+ β

D*R

, (Basu 1997) (1)

where X

it-1

is current period earnings scaled by beginning of period price, R

is the 12-

month return beginning nine months prior to fiscal year end to three months after fiscal

year end, and D is an indicator variable set equal to one when R

is negative, and zero

otherwise. Earnings are measured as earnings per share (Compustat data item 58). The

coefficients capture the mapping of news (returns) into current earnings. β

reflects the

mapping of good news into current earnings. β

captures the incremental effect of

earnings reflecting bad news relative to good news, and under conservative accounting,

is expected to be greater than zero. The ratio (β

+β

)/β

captures the sensitivity of

earnings to bad news relative to good news and is expected to be greater than one under

conservative accounting.

Givoly et al. (2007) examine the power and reliability of the Basu measure of

reporting conservatism and conclude its use “in different research settings should be more

selective and qualified.” They find the use of aggregated measures of earnings and

returns has an excessive influence on the magnitude of the Basu measure of

conservatism. They identify characteristics of the firm’s information environment that

are unrelated to conservatism that can affect the Basu measure; however, they freely

admit that controlling for these characteristics “are not easy to implement.” They also

find the Basu measure is negatively correlated with other measures of conservatism and

suggest it is only capturing one source or dimension of conservatism. However, an

express purpose of this study is to distinguish differential timeliness from other measures

of unconditional conservatism. Thus, while the criticisms of the Basu approach

expressed by Givoly et al. (2007) highlight limitations of the measure, they do not

invalidate this basic method of capturing conditional conservatism.

3.2 Test of H1

I predict in this study that firms increase conditional conservatism following a

restatement to help restore the credibility of their financial reports. With respect to

equation (1), this implies that the coefficient β

increases following a restatement. I

define the pre-restatement period to be the three years prior to the year the reporting

failure is announced. The post-restatement period includes the three years following the

year of the restatement announcement. I exclude the year of the restatement

announcement from my regression analyses because of the confounding effects of the

restatement announcement on the earnings/returns relationship.

I first consider the Basu (1997) approach and include incremental dummy variables

(e.g. bad news dummy) interacted with returns; however, preliminary data analysis

indicates a multicollinearity problem. The interaction of incremental dummy variables

with returns leads to much commonality among the independent variables. To mitigate

this multicollinearity problem, I instead define mutually exclusive dummy variables for

the conditional analysis throughout this study. I include separate PRE and POST dummy

variables to represent the pre/post-restatement periods. The PRE and POST variables are

interacted with returns and with dummy variables representing good and bad news

periods (POS and NEG). This results in mutually exclusive partitions of the sample and

the avoidance of repeated values among the explanatory variables. Preliminary data

analysis using this approach indicates lower standard errors and no multicollinearity

problem. Thus, I test H1 using the following pooled cross-sectional regression:

it-1

= β

PRE*POS + β

PRE*NEG + β

POST*POS + β

POST*NEG

+ β

PRE*POS*R

+ β

PRE*NEG*R

+ β

POST*POS*R

+ β

POST*NEG*R

+ ε

, (2)

where

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

= the 12-month return beginning nine months prior to fiscal year end to three

months after fiscal year end;

= error term;

it = subscript for firm i in year t;

= good news in pre-restatement period earnings;

= bad news in pre-restatement period earnings;

= good news in post-restatement period earnings;

= bad news in post-restatement period earnings;

Conditional Conservatism – difference between bad news and good news

– β

: Conditional conservatism in the pre-restatement period.

– β

: Conditional conservatism in the post-restatement period.

H1: (β

– β

) > (β

– β

)

H1 predicts an increase in conditional conservatism in the period following a financial

reporting failure, which is the change in conditional conservatism from the pre- to post-

restatement periods. Therefore, H1 predicts (β

– β

) > (β

– β

3.3 Unconditional Conservatism

Basu (2005) suggests using the intercepts from the Basu (1997) asymmetric

regression as a measure of unconditional conservatism. Basu (2005) suggests that the

intercept can be used to measure unconditional conservatism since it reflects average

earnings. Since unconditional conservatism leads to a downward bias in earnings, then a

lower intercept provides evidence of unconditional conservatism. Therefore, the

intercepts from equation (2) are used as my measure of unconditional conservatism.

Even though the model from equation (2) allows for separate intercepts for good and bad

news, since unconditional conservatism is news independent, I am able to aggregate the

good and bad news intercepts into the pre- and post restatement periods as follows: (β

) and (β

+ β

). However, this is not a formal hypothesis test as I expect the null

condition to be true.

As discussed before, I expect to find no change in unconditional conservatism before

and after the restatement, because unconditional conservatism is just a downward

accounting bias that should not be affected by a reporting failure. Therefore, I expect an

insignificant shift in the post-fraud/restatement period intercepts

from equation (2), which

is tested as (β

+ β

) = (β

+ β

3.4 Tests of H2 and H3

Hypotheses 2 and 3 predict that the level and the increase in conditional conservatism

following a reporting failure are greater for firms where debt and compensation

contracting are more important. To measure the importance of debt contracting to a firm,

I use the ex ante demand for external financing measure from Dechow et al. (1996). An

alternative measure considered for the importance of debt contracting is a firm’s level of

debt or the debt-to-asset ratio. However, these measures do not capture a firm’s need for

external financing over the three-year period following a restatement announcement. It is

very important for potential investors/creditors of restating firms to have trust and

confidence in the financial reporting of these firms. Thus, I expect firms that require

external financing soon after a restatement are more motivated to restore the credibility of

their financial reports by reporting more conservatively. The use of the debt level or the

debt-to-asset ratio may not effectively capture this motivation to report conservatively

brought on by the desire to attract external financing at the lowest cost possible. Thus, I

consider the Dechow et al. (1996) measure to be a more powerful measure of the

importance of debt contracting following a restatement.

The Dechow et al. (1996) measure is expressed as follows:

FreeC

t-1

2-t

2- t to4-t1-t

AssetsCurrent

esexpenditur capital Average - operations fromCash

The intuition behind this measure is as follows. If cash from operations is insufficient

to cover a firm’s desired investment level (i.e. measured as average capital expenditures),

then additional financing is needed. Thus, a negative numerator represents the amount of

financing needed. The denominator represents current assets available to fund the

desired investment level. Once the internal funds are consumed, then external financing

is needed.

The absolute value of the ratio

1-t

FreeC

when FreeC

t-1

is negative indicates the

remaining years the firm can go without needing external financing. For example, if

FreeC

t-1

equals -0.5, then the absolute value of the ratio

1-t

FreeC

equals 2. This means a

firm will consume its current assets in two years and thus require external financing.

Dechow et al. (1996) set FreeC

t-1

equal to -0.5 as the cutoff point between firms that have

a greater (lesser) demand for external financing. They also point out that since firms can

not exist without current assets and their model assumes all current assets are converted

to cash, their measure likely understates the need for external financing. Therefore,

assuming this relation carries over to the post-restatement period, and following Dechow

et al. (1996), I code the variable EXTFINHI as 1 if FreeC

t-1

is less than -0.5, and 0

otherwise. I code EXTFINLO as 1 if is greater than or equal to -0.5, and 0 otherwise. I

use HI/LO variables for EXTFIN in order to partition the sample into mutually exclusive

groups, where the separate relations will be captured by the regression coefficients.

Finally, I measure FreeC

t-1

in the year prior to the restatement announcement to ensure

the measure is not affected by the restatement announcement.

One contracting theory of earnings manipulation brought forth in the literature is the

bonus hypothesis (Watts and Zimmerman 1986), which predicts managers of firms with

bonus plans are more likely to manipulate earnings upward to increase their

compensation. Therefore, I measure the importance of compensation contracting to a

firm by using the variable Bonus from the Compustat Executive Compensation database.

Compustat defines Bonus as “the dollar value of a bonus (cash and non-cash) earned by

the named executive officer during the fiscal year.” I aggregate the bonuses of all

executives of the firm and compute the average for each year in the sample period. I then

scale the firm’s average bonus by end-of-year total assets to control for firm size effects

on compensation. I use HI/LO variables for BONUS in order to partition the sample into

mutually exclusive groups, which helps avoid a multicollinearity problem. The variable

BONUSLO (BONUSHI) is set equal to 1 if a firm’s average bonus is less than (greater

than or equal to) the median bonus of all sample firms, otherwise BONUSLO

(BONUSHI) is set equal to 0. Finally, the use of the variable Bonus does have potential

limitations as some firms may use stock-based incentive plans rather than earnings-based

incentive plans. However, the failure to include other incentive plans works against me

finding significant results in support of H3, as my tests may not capture all of the

compensation contracting effects on conservatism following a restatement.

I extend equation (2) to include the EXTFIN and BONUS HI/LO variables and

estimate the following pooled cross-sectional regressions:

it-1

= β

PRE*EXTFINHI*POS + β

PRE*EXTFINLO*POS

+ β

POST*EXTFINHI*POS + β

POST*EXTFINLO*POS

+ β

PRE*EXTFINHI*R*POS + β

PRE*EXTFINLO*R*POS

+ β

POST*EXTFINHI*R*POS + β

POST*EXTFINLO*R*POS

+ β

PRE*EXTFINHI*NEG + β

PRE*EXTFINLO*NEG

+ β

POST*EXTFINHI*NEG + β

POST*EXTFINLO*NEG

+ β

PRE*EXTFINHI*R*NEG + β

PRE*EXTFINLO*R*NEG

+ β

POST*EXTFINHI*R*NEG + β

POST*EXTFINLO*R*NEG

+ ε

, (3)

it-1

= β

PRE*BONUSHI*POS + β

PRE*BONUSLO*POS

+ β

POST*BONUSHI*POS + β

POST*BONUSLO*POS

+ β

PRE*BONUSHI*R*POS + β

PRE*BONUSLO*R*POS

+ β

POST*BONUSHI*R*POS + β

POST*BONUSLO*R*POS

+ β

PRE*BONUSHI*NEG + β

PRE*BONUSLO*NEG

+ β

POST*BONUSHI*NEG + β

POST*BONUSLO*NEG

+ β

PRE*BONUSHI*R*NEG + β

PRE*BONUSLO*R*NEG

+ β

POST*BONUSHI*R*NEG + β

POST*BONUSLO*R*NEG

+ ε

, (4)

where

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

>=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

EXTFINHI = 1 if firm’s FreeC

t-1

is less than -0.5, and 0 otherwise. Where FreeC

t-1

defined as:

FreeC

t-1

2-t

2- t to4-t1-t

AssetsCurrent

esexpenditur capital Average - operations fromCash

EXTFINLO = 1 if firm’s FreeC

t-1

is greater than or equal to -0.5, and 0 otherwise;

BONUSHI = 1 if a firm’s dollar value of a bonus is greater than or equal to the median

bonus of all sample firms, otherwise BONUSHI = 0. A firm’s bonus is

measured as the average bonus of the firm’s executives scaled by end-of-

year total assets (Compustat Data 6). Bonus data is obtained from the

Compustat Executive Compensation database;

BONUSLO = 1 if a firm’s dollar value of a bonus is less than the median bonus of all

sample firms, otherwise BONUSLO = 0;

= the 12-month return beginning nine months prior to fiscal year end to three

months after fiscal year end;

= error term;

it = subscript for firm i in year t.

I next describe what the coefficients in equations 3 and 4 are capturing; including the

specific coefficients I use to test H2a-b and H3a-b. The regression equations follow a

parallel structure such that the effect of high and low external financing needs in equation

3 correspond to the high and low bonus conditions in equation 4. Thus, for simplicity, I

describe the coefficients only once and note the alternative interpretations for equation

3(4) in the description.

Good news reflected in earnings

= good news in pre-restatement period earnings for high financing (bonus) firms;

= good news in pre-restatement period earnings for low financing (bonus) firms;

= good news in post-restatement period earnings for high financing (bonus) firms;

= good news in post-restatement period earnings for low financing (bonus) firms;

Bad news reflected in earnings

= bad news in pre-restatement period earnings for high financing (bonus) firms;

= bad news in pre-restatement period earnings for low financing (bonus) firms;

= bad news in post-restatement period earnings for high financing (bonus) firms;

= bad news in post-restatement period earnings for low financing (bonus) firms;

Conditional Conservatism – difference between bad news and good news

– β

Conservatism for high financing (bonus) firms in pre-restatement period;

– β

: Conservatism for low financing (bonus) firms in pre-restatement period;

– β

Conservatism for high financing (bonus) firms in post-restatement period;

– β

: Conservatism for low financing (bonus) firms in post-restatement period;

Change in Conditional Conservatism

(β

- β

) – (β

- β

): Change in conditional conservatism from the pre- to post

restatement period for high financing (bonus) firms;

(β

- β

) – (β

- β

): Change in conditional conservatism from the pre- to post-

restatement period for low financing (bonus) firms;

H2a (H3a): (β

- β

) > (β

- β

)

H2b (H3b): [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)]

H2a (H3a) predicts that in the post-restatement period conditional conservatism for

high external financing (high bonus) firms is greater than conditional conservatism for

low external financing (low bonus) firms. Thus, (β

- β

) > (β

- β

). H2b (H3b)

predicts that the increase in conditional conservatism from the pre- to post-restatement

periods is greater for high external financing (high bonus) firms as compared with low

external financing (low bonus) firms. Therefore, [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)].

3.5 Tests of H4 through H9

I next examine changes in the financial reporting environment as potential

explanations for an increase in conditional conservatism following a financial reporting

failure. I test hypotheses 4 – 9 by extending equation (2) to include variables relating to

changes in the board of directors, audit committee, management, and auditor. I test each

potential determinant separately by estimating a pooled cross-sectional regression that

takes the following general form:

it-1

= β

PRE*detyes*POS + β

PRE*detno*POS

+ β

POST*detyes*POS + β

POST*detno*POS

+ β

PRE*detyes*R*POS + β

PRE*detno*R*POS

+ β

POST*detyes*R*POS + β

POST*detno*R*POS

+ β

PRE*detyes*NEG + β

PRE*detno*NEG

+ β

POST*detyes*NEG + β

POST*detno*NEG

+ β

PRE*detyes*R*NEG + β

PRE*detno*R*NEG

+ β

POST*detyes*R*NEG + β

POST*detno*R*NEG + ε

, (5)

where the variables detyes and detno represent the six potential determinants of a change

in conservatism as discussed for hypotheses 4 – 9. The other variables are defined as

before. detyes is coded as 1 when the determinant variable has increased or changed;

otherwise detyes is coded as 0. detno is coded as 1 when the specific determinant

variable has not increased or changed; otherwise detno is coded as 0. The pooled cross-

sectional regression in equation (5) is estimated six times for each of the six determinant

variables. The six determinant (det) variables partitioned into yes/no conditions are as

follows:

OutDirYes = 1 if there is an increase in the number of outside directors from the year

prior to the third year following the restatement, 0 otherwise;

OutDirNo = 1 if there is no increase in the number of outside directors from the year

prior to the third year following the restatement, 0 otherwise;

ACMeetYes = 1 if there is an increase in the number of audit committee meetings from

the year prior to the third year following the restatement, 0 otherwise;

ACMeetNo = 1 if there is no increase in the number of audit committee meetings from

the year prior to the third year following the restatement, 0 otherwise;

ACIndDirYes = 1 if there is an increase in the number of independent members on the

audit committee from the year prior to the third year following the

restatement, 0 otherwise;

ACIndDirNo = 1 if there is no increase in the number of independent members on the

audit committee from the year prior to the third year following the

restatement, 0 otherwise;

ACFinExpYes =1 if there is an increase in the number of financial experts on the audit

committee from the year prior to the third year following the

restatement; 0 otherwise;

ACFinExpNo =1 if there is no increase in the number of financial experts on the audit

committee from the year prior to the third year following the

restatement; 0 otherwise;

MgmtChgYes = 1 if there is a change in top management following the restatement, 0

otherwise;

MgmtChgNo = 1 if there is no change in top management following the restatement, 0

otherwise;

AudChgYes = 1 if there is a change in the auditor following the restatement, 0

otherwise;

AudChgNo = 1 if there is no change in the auditor following the restatement, 0

otherwise;

Following the same methodology discussed previously to test H2 and H3, H4-H9 are

tested as follows:

H4a, H5a, H6a, H7a, H8a, H9a: (β

- β

) > (β

- β

);

H4b, H5b, H6b, H7b, H8b, H9b: [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)].

The interpretation of these test specifications is similar to Hypotheses 2 and 3. For

example, H4a specifies that in the post restatement period, conditional conservatism is

greater for firms that increase the number of outside directors. Thus, H4a predicts (β

) > (β

- β

). H4b specifies that an increase in conditional conservatism following a

restatement is greater for firms that increase the number of outside directors. Thus, H4b

predicts [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)]. The remaining hypotheses, H5-

H9, are tested using the same methodology just described for H4.

In addition to the hypothesis tests described above, I perform some sensitivity

analysis on the data. First, I estimate the primary regression separated into the pre- and

post-SOX periods, as the increased regulatory attention and rules brought forth by SOX

may affect a firm’s decision to report more conservatively following a reporting failure. I

classify firms in the pre-SOX period if the year of the restatement announcement is 1999

or earlier. Thus for a restatement in 1999, the three-year period following the restatement

announcement ends in 2002. (Congress passed SOX in July of 2002.)

Second, in an attempt to increase the sample size, I set the pre and post-restatement

periods to be two years before and after the restatement announcement year. This

sensitivity test is a tradeoff because I sacrifice time-series data in order to increase the

number of sample firms. Sacrificing time-series data makes it more difficult to detect

changes in reporting behavior.

Third, I replace the dichotomous measures for an increase in the number of outside

directors, audit committee members, and financial experts on the audit committee with

dichotomous measures for the level of these variables compared to the sample median in

the third year following the restatement announcement. For example, the variable

OutDir%

is set equal to 1 if a firm’s outside director percentage in the third year

following the restatement announcement is greater than the median outside director

percentage for all samples firms over the same time period. This alternative approach

reflects the level of corporate governance following a restatement, rather than the change.

It is unclear whether a change in corporate governance or the level of corporate

governance following a restatement prompts a change in a firm’s financial reporting.

As a fourth sensitivity analysis, I test all hypotheses using market-adjusted returns in

addition to the raw returns indicated in the research design. Market-adjusted returns

controls for the overall market performance and better isolates the firm-specific news and

performance in returns which proxies for good and bad news periods in the model.

Finally, to control for economic events I use industry-adjusted measures with industry

defined as sample firms with the same two-digit Standard Industrial Classification (SIC)

code. I measure earnings and returns variables relative to their industry mean. The

industry-adjusted measures control for economic events that may vary across industries

and controls for conservatism that may arise from industry-specific GAAP or regulations.

CHAPTER 4

EMPIRICAL RESULTS AND ANALYSIS

This chapter provides a description of the empirical results. First, I describe the

sample selection process, sample attrition, and descriptive statistics of the primary

variables of interest. Next, I present the results of the hypotheses tests described in the

previous chapters. Last, I present the results of sensitivity tests.

4.1 Sample Selection

The initial restatement sample is obtained from the GAO-03-395R Financial

Statement Restatement Database. The GAO database identifies 919 restatements from

the period January 1, 1997 to June 30, 2002. The GAO database provides a

comprehensive list of restatements resulting from accounting irregularities, including

fraud, aggressive accounting, intentional and unintentional misuse of facts, and

misinterpretation or oversight of accounting rules. The GAO database excludes

restatements due to normal business activities, such as mergers and acquisitions, stock

splits, stock dividends, discontinued operations, currency-related issues, change in

business segment definitions, changes due to transfers of management, litigation

settlements, general accounting changes under GAAP, and general bookkeeping errors.

Thus, the GAO database of restatements due to accounting irregularities provides an

appropriate sample to study the impact of restatements on firms’ financial reporting

choices.

Stock return data is obtained from the Center for Research in Security Prices (CRSP)

data files. Financial data such as earnings, long-term debt, and total assets are obtained

from the Compustat 2004 annual database, which includes the PST, Full Coverage, and

Research files. I use proxy statements or SEC Form 10-K Annual Reports to obtain data

for variables relating to the board of directors, audit committee, management, and

auditor. Finally, a firm’s bonus data is obtained from the Compustat Executive

Compensation database.

I summarize sample attrition in Table 4.1. Of the initial sample of 919 restating

firms, 40 firms were eliminated because no ticker symbol or CNUM could be found.

Ninety-three firms were deleted because of multiple restatements. Approximately 72

percent (659 firms) of the initial restatement sample of 919 firms did not have sufficient

returns or financial data from CRSP and Compustat to be included in the final sample.

The missing data does not appear to be clustered in either the pre- or post-restatement

period. The number of firms with missing Compustat data in the pre-restatement period

is comparable to the post-restatement period, with 200 and 174 firms, respectively. In

addition, 48 firms were missing Compustat data in both the pre- and post-restatement

period. This study requires stock and financial data for the three years prior to and the

three years following the restatement announcement year. A likely explanation for the

loss of these firms is due to the fact that many restating firms declare bankruptcy or are

delisted following the restatements. This could potentially lead to a survivorship bias,

which may prevent the results from generalizing to the overall set of publicly traded

firms. Finally, 15 firms were eliminated because their returns and earnings data are

considered outliers with studentized residuals greater than the absolute value of three.

Outliers are observations that are extreme or appear inconsistent with the remaining data.

This leaves a final sample of 112 firms.

4.2 Descriptive Statistics

Table 4.2 presents descriptive data for the sample firms averaged over the six-year

sample period, three years pre and post the restatement announcement year. Panel A of

Table 4.2 provides descriptive statistics regarding the main explanatory variables, and

Panel B provides descriptive statistics for the corporate governance variables used to test

hypotheses 4 through 9.

As presented in Panel A of Table 4.2, the sample firms represent a wide cross-section

of performance and size, but the distribution appears to be skewed towards larger firms.

As for size, total assets range from $6.04 million to $208,504 million with a mean

(median) of $5,005 million ($713.93 million). This skewness towards large firms is

likely a result of the research design of this study where financial data is needed for the

three years prior to and following a restatement announcement, which requires seven

consecutive years of financial data. It stands to reason that larger firms are better able to

withstand the negative effects of a restatement and continue in business without being

delisted or declaring bankruptcy. However, this leads to a possible survivorship bias,

which may reduce the external validity of the results.

As presented in Table 4.2, Panel B, I find the mean number of outside directors

the board in the year prior to the restatement announcement is 6.28, and 6.54 in the third

year following the restatement announcement, a statistically significant difference (t =

1.65, p=0.05). This is consistent with Farber (2005), who finds fraud firms increase the

number of outside directors on the board.

Table 4.2, Panel B also provides descriptive statistics for audit committee variables

and auditor changes

. Results indicate that sample firms increased significantly the mean

number of audit committee meetings from 3.40 in the year prior to restatement

announcement to 7.55 in the third year following the restatement announcement (t =

27.23). Results in Table 4.2, Panel B also indicate the sample firms increased the

average number of audit committee members and the average number of outside directors

and financial experts

on the audit committee from the year prior to the third year

following the restatement announcement. The mean number of audit committee

members increased from 3.51 to 3.61 (t = 1.66), the mean number of outside directors on

the audit committee increased from 3.29 to 3.56 (t = 4.35), and the mean number of

financial experts on the audit committee increased from 0.74 to 1.38 (t = 11.52).

However, the median values for these variables did not change.

In summary, the changes in the financial reporting environment discussed above are

consistent with previous corporate governance research. However, it is still an empirical

question whether or not these corporate governance changes affect conservatism.

A potential problem with regression analysis is multicollinearity. Multicollinearity

occurs when some of the independent variables are highly correlated. Multicollinearity

produces coefficient estimates that are unstable and have large standard errors. I test for

Consistent with Farber (2005), an outside director is defined as a director who is not a present or former

firm employee and whose only formal connection to the firm is service as a director on the board.

The number of observations for the variable AudChg is different due to the inclusion of five additional

sample firms with available auditor change data.

Following Farber (2005), I define a financial expert as someone who has accounting or financial

management expertise (i.e., CPA, CFO, PFO, CAO, PAO, controller, auditor, accounting or finance

professor).

the presence of multicollinearity by examining correlations between the independent

variables and by using the variance inflation factor (VIF). The variance inflation factors

(untabulated) show no significant correlations between the independent variables. I

present the Pearson and Spearman correlation coefficients in table 4.3 for the primary

variables of interest

. The correlations between most of the independent variables are

insignificant. For the correlations that are statistically significant, the correlations are

very low. Based on the results of the correlation analysis and the variance inflation

factors, multicollinearity among the independent variables does not appear to be a

problem.

4.3 Tests of Hypothesis 1

Hypothesis 1 predicts that restatement firms increase conditional conservatism

following a financial reporting failure. This hypothesis is tested using a pooled cross-

sectional regression analysis of beginning-of-fiscal year price deflated earnings on

positive and negative returns with pre/post-restatement year interactions.

The regression results for equation (2), which tests hypothesis 1 for the restatement

sample, are presented in Table 4.4. Hypothesis 1 is tested using three alternative returns

measures. In addition to annual raw returns, I also examine equally weighted market-

adjusted and value weighted market-adjusted annual returns. I perform an F-test which

imposes a linear restriction on the coefficients and tests whether the sum of squared

residuals is significantly greater than the sum of squared residuals from the unrestricted

model. The F-test is a test of equality (e.g. (β

– β

) = (β

– β

), even though all of my

hypotheses are directional. Therefore, I present in the tables the differences in

coefficients for both sides of the equation to verify the difference is in the predicted

direction. In panel A, using annual raw returns, the results support hypothesis 1 as

conditional conservatism in the post-restatement period (coefficient = 0.194) is

significantly greater than in the pre-restatement period (coefficient = 0.048), (F-statistic =

8.84, p=0.003).

The variable BONUS is omitted from the table because the number of observations is much lower at 360.

I ran a separate correlation analysis including BONUS and find the correlations between BONUS and the

other variables in Table 4.3 are insignificant or very low.

Results presented in Panel B of Table 4.4 using equally weighted market-adjusted

annual returns also support hypothesis 1. The coefficients for conditional conservatism

in the pre- and post-restatement periods are 0.026 and 0.160, respectively. The difference

is statistically significant (F-statistic = 12.70, p=0.0004). In panel C of Table 4.4, using

value weighted market-adjusted returns, conditional conservatism in the post-restatement

period (coefficient = 0.203) is significantly greater than in the pre-restatement period

(coefficient = 0.040), (F-statistic = 13.88, p=0.0002). The use of market-adjusted returns

in the regressions provides higher F-statistics and stronger support for H1 than does the

use of annual raw returns. Basu (1997) finds a higher adjusted R

using market-adjusted

returns as compared to annual raw returns. I find a higher adjusted R

using value-

weighted market-adjusted returns, but not equal-weighted market-adjusted returns. The

use of value-weighted market-adjusted returns likely provides more explanatory power

because it controls for news events that affect overall market performance. As returns are

used to proxy for good and bad news periods, controlling for the overall market

performance will better isolate firm-specific news and performance and thereby reduce

measurement error for the good and bad news dummy variables. In summary, the results

presented in Table 4.4 provide strong support for the first hypothesis. Restating firms

significantly increase conditional conservatism in the three-year period following the

restatement announcement year. These results provide the basis for further investigation

into the underlying reasons or determinants of the increase in conservatism.

4.4 Unconditional Conservatism

As discussed in Chapter 3, I expect to find no change in unconditional conservatism

before and after the restatement, because unconditional conservatism is just a downward

accounting bias. Therefore, I expect managers to be less motivated to change

unconditional conservatism following a restatement. Following Basu (2005), the

intercepts from equation (2) are used as a measure of unconditional conservatism. Basu

(2005) suggests the intercept can be used to measure unconditional conservatism since it

reflects average earnings. Therefore, I expect an insignificant shift in the post-

fraud/restatement period intercepts

from equation (2), which is tested as (β

+ β

) = (β

As expected, the untabulated results show no significant intercept shift from the pre-

to post-restatement period using all three returns measures. For annual raw returns, (β

) = 0.066 and (β

+ β

) = 0.095, for which the difference is statistically insignificant (F-

statistic =1.84, p=0.18). For equally weighted market-adjust returns, (β

+ β

) = 0.064

and (β

+ β

) = 0.091, a statistically insignificant difference (F-statistic = 1.75, p=0.19).

For value weighted market-adjust returns, (β

+ β

) = 0.073 and (β

+ β

) = 0.084, a

statistically insignificant difference (F-statistic = 0.31, p=0.58). To the extent the

intercepts from equation (2) capture unconditional conservatism, the results appear

consistent with unconditional conservatism being unaffected by a restatement. This

supports the Ball and Shivakumar (2005) view that unconditional conservatism is just a

downward accounting bias that contracting parties can adjust for ex ante. Thus,

managers have less incentive to increase unconditional conservatism because it does not

enhance contracting efficiency. In contrast, conditional conservatism does have the

potential to improve contracting efficiency for both debt and compensation contracting,

which I test next.

4.5 Tests of Hypotheses 2a and 2b

The second hypothesis examines debt contracting as an explanation for the increase in

conditional conservatism following a restatement. This hypothesis predicts the level and

change in conditional conservatism following a restatement is greater for firms where

debt contracting is more important. To measure the importance of debt contracting to a

firm, I use the ex ante demand for external financing measure from Dechow et al. (1996).

Table 4.5 presents the regression results for equation (3) pertaining to hypothesis 2.

In Panel A, conditional conservatism for high financing firms in the post-restatement

period (coefficient = 0.456) is significantly greater than for low financing firms

(coefficient = 0.192) (F-statistic = 5.45, p=0.02). However, the results in Panel A do not

support H2b, as the change in conditional conservatism from the pre- to post restatement

period for high financing firms is not significantly different than the change for low

financing firms (F-statistic = 2.52, p=0.113). The results using annual raw returns

suggest the level, but not necessarily the change, in conditional conservatism is greater

for firms that have a stronger need for debt financing.

Panel B of Table 4.5 provides the results of testing hypothesis 2 using market-

adjusted returns. The results using market-adjusted returns provide strong support for

hypothesis 2. In Panel B using equal weighted market-adjusted returns, H2a is strongly

supported as conditional conservatism in the post-restatement period is significantly

greater for high financing firms (coefficient =0.484) than for low financing firms

(coefficient = 0.149). This difference is highly significant (F-statistic = 8.33, p=0.004).

H2b is also supported using equal weighted market-adjusted returns at the 10% level (F-

statistic = 2.89, p=0.09). The use of value weighted market-adjusted returns in the model

provides the strongest support for hypothesis 2. In Panel B, conditional conservatism

following a restatement is significantly greater for high financing firms (coefficient =

0.702) as compared with low financing firms (coefficient = 0.184). The difference is

highly significant (F-statistic = 13.76, p=0.0002). Furthermore, the change in conditional

conservatism from the pre- to post-restatement period for high financing firms is

significantly greater than for low financing firms (F-statistic = 13.33, p=0.0003).

The results in Table 4.5 provide strong support for the second hypothesis. The results

suggest that both the level and change in conditional conservatism following a

restatement is greater for firms where debt contracting is more important. These results

are consistent with the contracting explanation of conservatism brought forth in Watts

(2003). The results suggest restating firms attempt to restore the credibility of their

financial reports and the trust of the contracting parties by reporting more conservatively.

Conservative reporting constrains management’s opportunistic behavior and protects the

interests of the contracting parties.

4.6 Tests of Hypotheses 3a and 3b

Hypothesis 3 examines compensation contracting as an explanation for an increase in

conditional conservatism following a restatement. The third hypothesis predicts that the

more important compensation contracting is to a firm, the more conservative it will be

following a reporting failure. I measure the importance of compensation contracting to a

firm by using the variable Bonus from the Compustat Executive Compensation database.

I measure the variable BONUS used in equation (4) as the average bonus of the firm’s

executives scaled by end-of-year total assets (Compustat Data 6). The bonus hypothesis

predicts managers of firms with bonus plans are more likely to bias accounting numbers

upward to increase their compensation (Watts and Zimmerman 1986). Watts (2003a)

suggests that conservatism reduces the likelihood that management overstates earnings

and net assets in order to take excess compensation. Therefore, I predict restating firms

with earnings-based managerial compensation plans are more likely to report

conservatively in order to signal that the financial reports are reliable and trustworthy.

The regression results for equation (4) pertaining to hypothesis 3 are presented in

Table 4.6. Panel A of Table 4.6 presents the results using annual raw returns.

Surprisingly, neither H3a nor H3b are supported (F-statistic = 0.48 and 0.14). However,

in Panel B of Table 4.6, results based on market-adjusted returns provide strong support

for hypothesis 3. Using equal weighted market-adjusted returns, H3a is significant (F-

statistic = 7.41, p=0.007), and H3b is significant (F-statistic = 8.24, p=0.004). Using

value weighted market-adjusted returns, H3a and H3b are significant at the 5% and 1%

levels with F-statistic = 4.45 and 7.32, respectively.

The use of market-adjusted returns in the model testing H3 appears to provide a better

fit and higher explanatory power. The adjusted R

s in Panel B of Table 4.6 using the two

measures of market-adjusted returns are 25.71% and 25.74%, while the adjusted R

using

annual raw returns is 22.6%. As discussed previously, the use of market-adjusted returns

in testing H1 and H2 provides higher F-statistics and stronger support than does the use

of annual raw returns. Overall, it appears that market-adjusted returns are better at

capturing the level and changes in conditional conservatism for the sample firms. This is

likely because removing the effect of news that affects overall market performance better

isolates firm-specific news in returns which reduces the measurement error in the good

and bad news dummy variables used in the model.

Thus, to the extent that using market-adjusted returns captures the Basu (1997)

measure of conditional conservatism, hypothesis 3 is supported. The results suggest that

both the level and change in conditional conservatism following a restatement are greater

for firms where compensation contracting is more important. The results for both

hypotheses 2 and 3 appear to be consistent with the contracting explanation of

conservatism put forth by Watts (2003).

4.7 Tests of Hypotheses 4a and 4b

Hypotheses 4 through 9 examine several corporate governance variables as possible

explanations for an increase in conditional conservatism following a restatement.

Hypothesis 4 posits firms with more outside directors following a reporting failure report

more conservatively.

The results of equation (5) pertaining to hypothesis 4 are presented in Table 4.7. In

Panel A using annual returns, the F-test is significant at the 10% level (F-statistic = 3.11,

p=0.078), but is not in the predicted direction. Therefore, H4a is not supported. Panel A

also shows no support for H4b (F-statistic = 1.04, p=0.308). In Panel B using market-

adjusted returns, neither H4a nor H4b are supported.

Therefore, the results from Table 4.7 do not support hypothesis 4. I am unable to

document any consistent relationship between an increase in the number of outside

directors on the board and the level or change in conditional conservatism following a

restatement. This is not too surprising since the increase in the average number of

outside directors following a restatement was very small. As reported in Table 4.2, the

average number of outside directors increased from 6.28 to 6.54 following the

restatement. In addition, the median number of outside directors remained at 6 following

the restatement.

4.8 Tests of Hypotheses 5a and 5b

Hypotheses 5 through 7 examine whether certain audit committee characteristics

explain a change in conditional conservatism following a restatement. Specifically, I

predict an audit committee that meets more frequently, is more independent, and has

more financial expertise following a reporting failure demands more conservative

financial reporting.

Hypothesis 5 predicts that the level and change in conditional conservatism following

a restatement is greater for firms that increase the number of audit committee meetings

following a restatement. The results for equation (5) pertaining to hypothesis 5 are

presented in Table 4.8. In Panel A, the results do not support hypotheses 5a or 5b as the

F-statistics are 0.33 and 0.29, respectively. In Panel B, the results using equal weighted

market-adjusted returns also do not support hypotheses 5a or 5b as the F-statistics are

0.32 and 1.54, respectively. However, the results in Panel B of Table 4.8 using value

weighted market-adjusted returns do strongly support hypothesis 5. H5a and H5b are

both significant at the 5% level, with F-statistic =5.19 for H5a and F-statistic =4.25 for

H5b.

In summary, the results are mixed for hypothesis 5 concerning the relation between

conservatism and the frequency of audit committee meetings. This hypothesis is strongly

supported using value weighted market-adjusted returns; however, the results using equal

weighted market-adjusted and annual returns do not support hypothesis 5.

4.9 Tests of Hypotheses 6a and 6b

The sixth hypothesis predicts the level and change in conditional conservatism

following a restatement is greater for firms that increase the number of independent

directors on the audit committee. Table 4.9 presents the regression results pertaining to

the tests of hypothesis 6. In Panel A using annual returns, the results are insignificant and

do not support hypotheses 6a or 6b. Panel B using market-adjusted returns also yields

insignificant results.

The results from Table 4.9 do not support hypothesis 6. There is no significant

relationship between an increase in the number of outside directors on the audit

committee and the level or change in conditional conservatism following a restatement.

The lack of results may be explained by the small increase in the mean number of outside

directors on the audit committee following a restatement. The mean number of outside

directors on the audit committee as presented in Panel B of Table 4.2 increased from 3.29

to 3.56, while the median remained the same at 3.

4.10 Tests of Hypotheses 7a and 7b

The seventh hypothesis predicts the level and change in conditional conservatism

following a restatement is greater for firms that increase the number of financial experts

on the audit committee.

The results of equation (5) pertaining to hypothesis 7 are presented in Table 4.10. In

Panel A using annual returns, both H7a and H7b are not supported. Likewise, in Panel B

using equal weighted market-adjusted returns, both H7a and H7b are not supported with

F-statistics of 1.48 and 2.17, respectively. However, Panel B using value weighted

market-adjusted returns does provide support for hypothesis 7. Both H7a and H7b are

supported at the 5% level with F-statistics of 5.18 and 4.55, respectively.

These results are similar to the results for hypothesis 3, in that annual returns provide

insignificant results, but market-adjusted returns provide strong support for the

hypothesis. As discussed previously, it appears using market-adjusted returns

(particularly with the value weighted index) in the model improves the fit and increases

the explanatory power of the model. This can be seen in Table 4.10 where the value

weighted model has the highest adjusted R

(17.10%).

4.11 Tests of Hypotheses 8a and 8b

The eighth hypothesis examines a change in top management

as a potential

explanation for a change in conditional conservatism following a restatement.

Approximately 63% (64 out of 102 firms) of the sample firms experienced a change in

top management. This change in top management is consistent with Desai et al. (2006),

who find that 60 percent of restatement firms experience turnover of at least one top

manager in the two years following the restatement.

Table 4.11 presents the regression results that test hypothesis 8. In Panel A using

annual returns, neither H8a (F-statistic = 2.00, p=0.158) nor H8b are supported (F-

statistic = 1.18, p=0.278). In Panel B of Table 4.11 using market-adjusted returns, the

results are all insignificant. Therefore, the results from Table 4.11 do not support

hypothesis 8. There appears to be no significant association between a change in top

management and a change in conditional conservatism following a restatement.

4.12 Tests of Hypotheses 9a and 9b

The ninth hypothesis examines a change in auditor as a potential explanation for the

level and change in conditional conservatism following a restatement. Empirical studies

find that auditors can mitigate their litigation risk by requiring their clients to report more

conservatively (e.g. Cahan & Zhang 2006; Krishnan 2005). Therefore, I predict the new

Following Desai et al. (2006), top management is defined as Chairman, CEO, and/or President.

auditor of a restating firm to demand more conservative reporting in order to protect their

reputation and reduce the likelihood of litigation.

I find approximately 32% (34 out of 107 firms) of the sample firms experienced a

change in auditor. This is comparable to Farber (2005), who finds 25 percent of fraud

firms switch audit firms. The results from Table 4.12 are all statistically insignificant and

do not support hypothesis 9. Thus, I find no evidence of an association between a change

in auditor and the level or change in conditional conservatism following a restatement.

4.13 Sensitivity Analysis

As a sensitivity analysis, I estimate equation (2) for the pre- and post-SOX periods, as

the increased regulatory attention and rules brought forth by SOX may affect a firm’s

decision to report more conservatively following a reporting failure. I classify firms in

the pre-SOX period if the year of the restatement announcement is year 1999 or earlier.

Thus for a restatement in 1999, the third year following the restatement announcement is

2002. Since Congress passed SOX in July of 2002, the majority of the three year period

following a 1999 restatement is pre-SOX.

The results of separating the equation (2) regression into pre- and post-SOX periods

are presented in Table 4.13. Of the 112 sample firms, 45 firms (270 observations) are in

the pre-SOX period and 67 firms (402 observations) are in the post-SOX period. In Panel

A using annual returns, hypothesis one is supported in the pre-SOX period with F-

statistic = 3.59 (p=0.059). In Panels B and C using equally and value weighted market-

adjusted returns, hypothesis one is strongly supported in the pre-SOX period with F-

statistics of 6.79 and 5.38, respectively. These results suggest that firms significantly

increase conditional conservatism in the three year period following the restatement

announcement prior to the passage of SOX. As for the post-SOX period, I find mixed

results as hypothesis one is only supported with the use of value weighted market-

adjusted returns (F-statistic = 3.41, p = 0.066). Importantly, the primary results of the

study do not appear to be driven by the passage of SOX.

In an attempt to increase the sample size, I set the pre and post-restatement periods to

be two years before and after the restatement announcement year. This results in an

additional 84 sample firms, bringing the total sample firms to 196. This robustness test is

a tradeoff because I sacrifice time-series data in order to increase the number of sample

firms. Sacrificing time-series data makes it more difficult to detect changes in reporting

behavior. I retest the first two hypotheses using the larger sample size of 196 firms. In

untabulated results, both hypotheses are still supported using market-adjusted returns.

For the test of H1, the F-statistics using both industry-adjusted and market-adjusted

returns are statistically significant (Equal weighted: F-statistic = 5.97, p=0.015; Value

weighted: F-statistic = 7.02, p=0.008). H2a is supported at the 10% significance level

using equal-weighted market-adjusted returns (F-statistic = 3.09, p=0.079) and at the 5%

level using value-weighted market-adjusted returns (F-statistic = 4.62, p=0.032). H2b is

supported at the 10% significance level using equal-weighted market-adjusted returns (F-

statistic = 3.62, p=0.058) and at the 5% level using value-weighted market-adjusted

returns (F-statistic = 6.32, p=0.012).

I set the pre and post-restatement periods to be two years before and after the

restatement announcement year to test hypothesis 3. This results in 29 additional firms

for a total of 89 firms. Surprisingly, the untabulated results are statistically insignificant

for the tests of H3a and H3b. Therefore, the results for H3 appear to be less robust.

As another sensitivity test, I replace the variable OUTDIR which is set equal to 1 if

there is an increase in the number of outside directors from the year prior to the third year

following the restatement with a measure of the firm’s outside director percentage. The

variable OUTDIR% is set equal to 1 if a firm’s outside director percentage in the third

year following the restatement announcement is greater than the median outside director

percentage for all sample firms over the same time period. This alternative measure tests

the relative level of this variable, rather than the absolute change. The level may be a

better measure as a firm with a large number of outside directors on the board may be

forced to report more conservatively following a restatement, even if the number of

outside directors did not increase.

The results of using the outside director percentage variable in equation (5) are

presented in Table 4.14. The results in Table 4.14 are all insignificant and do not support

hypothesis 4. This is consistent with the results from Table 4.7 using the variable

OUTDIR. Thus, results still fail to support a relationship between outside directors on

the board and the level or change in conditional conservatism following a restatement.

I also test hypothesis 5 using an alternative measure for audit committee meetings.

Instead of using a dichotomous variable for an increase in the number of meetings, I

measure the variable ACMEET

t+3

, which is set equal to 1 if the firm’s number of audit

committee meetings in the third year following the restatement announcement is greater

than the median number of audit committee meetings for all sample firms over the same

time period. The results using this alternative measure of audit committee meetings are

presented in Table 4.15. The results are all insignificant except for the test of hypothesis

5a using equal weighted market-adjusted return (F-statistic = 9.54, p<0.001). However,

this significant result is not in the predicted direction. Thus, the results using this

alternative measure for audit committee meetings do not support hypothesis 5.

I also test the seventh hypothesis by using an alternative measure for the financial

expertise of the audit committee. I set the variable ACFINEXP% equal to 1 if the firm’s

percentage of audit committee members who are financial experts during the third year

following the restatement is greater than the median percentage for all sample firms for

the same time period. The results using this alternative measure are presented in Table

4.16. The results using all three returns measures are insignificant. Thus, the results

using this alternative measure for financial expertise on the audit committee provide no

support for hypothesis 7.

Finally, I use industry-adjusted earnings and returns variables to control for economic

events that may vary across industries. This approach also controls for conservatism that

may arise from industry-specific GAAP or regulations. I measure earnings and returns

variables relative to industry mean for all sample firms in the firm’s industry. I define

industry using the two-digit Standard Industrial Classification (SIC) code.

The results using the industry-adjusted measures are reported in Table 4.17. The first

hypothesis is still supported using industry-adjusted EPS and annual returns (F-statistic =

5.97, p<0.007). However, the explanatory power of the model decreases significantly

with an Adjusted R

of 3.93%. The same model without adjusting for industry effects

has an Adjusted R

of 13.28% (Table 4.4, Panel A). In untabulated results using equal

and value weighted market-adjusted returns, hypothesis one is not supported after

adjusting for industry. Adjusting for both the market and industry produces a model with

insignificant explanatory power. The F-statistics for the models using both industry-

adjusted and market-adjusted returns are insignificant (Equal weighted: F-statistic = 0.54,

p=0.8274; Value weighted: F-statistic = 0.58, p=0.7964).

Table 4.1

Sample Attrition

No. of Firms

Initial Restatement Sample

919

Less:

Firms with no Ticker Symbol or CNUM 40

Firms with more than one restatement 93

Firms with missing Compustat data 138

Firms with missing Compustat data in post-restatement period 174

Firms with missing Compustat data in pre-restatement period 200

Firms with missing Compustat data in both pre- and post-

restatement periods 48

Firms with missing CRSP data 99

Outliers – firms with studentized residuals greater than +/-3 15

(807)

Final Restatement Sample

112

The initial restatement sample is obtained from the GAO-03-395R Financial Statement

Restatement Database. The GAO database identifies 919 restatements from the period

January 1, 1997 to June 30, 2002.

A significant portion of the missing data items results from the need to obtain stock and

financial data for the three years prior to and the three years following the restatement

announcement year. In addition, capital expenditures from the fourth year prior to the

restatement announcement year are needed to compute the ex ante demand for external

financing measure from Dechow et al. (1996).

Table 4.2

Descriptive Statistics

Panel A: Main Explanatory Variables

Standard

Variable

n Mean Deviation Median Min. Max.

Earnings ($million) 672 151.043 501.751 18.019 -4,895.0 4,220.0

Earnings per share 672 0.899 2.816 0.740 -9.0 35.83

EPS/Price 672 0.022 0.092 0.035 -0.421 0.352

Return 672 0.220 0.899 0.099 -0.860 14.192

EWMAR 672 0.065 0.860 -0.061 -1.293 13.646

VWMAR 672 0.117 0.874 -0.015 -1.039 13.930

Total Assets ($million) 672 5,005 12,773 713.93 6.04 208,504

Debt ($million) 672 1,224 2,614 79.23 0 22,792

Bonus ($thousand) 360 333.441 403.387 211.225 0 4,584.00

Bonus/Total Assets 360 0.242 0.776 0.072 0 9.295

FreeC

t-1

112 -0.030 0.848 0.055 -7.904 1.850

Panel B: Corporate Governance Variables

Standard

Variable

n Mean Deviation Median Min. Max.

Board of Directors

Dir

t-1

102 8.98 3.07 9.0 4.0 17.0

Dir

t+3

102 9.04 2.66 8.5 4.0 16.0

OutDir

t-1

102 6.28 2.88 6.0 1.0 15.0

OutDir

t+3

102 6.54 2.53 6.0 2.0 13.0

Audit Committee

ACMeet

t-1

102 3.40 1.68 3.0 1.0 9.0

ACMeet

t+3

102 7.55 3.37 7.0 1.0 18.0

ACMem

t-1

102 3.51 1.22 3.0 2.0 8.0

ACMem

t+3

102 3.61 0.81 3.0 3.0 6.0

ACIndDir

t-1

102 3.29 1.26 3.0 1.0 8.0

ACIndDir

t+3

102 3.56 0.82 3.0 2.0 6.0

ACFinExp

t-1

102 0.74 0.85 1.0 0 3.0

ACFinExp

t+3

102 1.38 1.05 1.0 0 5.0

MgmtChg 102 0.627 0.484 1.0 0 1.0

AudChg 107 0.318 0.466 0 0 1.0

Table 4.2 (Continued)

Variables are averaged over the six-year sample period (three years before and after

restatement announcement year) and are defined as follows:

Earnings are net income before extraordinary items (Compustat Data 18);

Earnings per share (EPS) is Compustat Data 58;

Price is beginning of fiscal year stock price;

Returns are the compounded monthly returns beginning nine months prior to fiscal year

end to three months after fiscal year end;

EWMAR are computed as the difference between the annual return and the mean return of

all firms from the CRSP equally-weighted market index.

VWMAR are computed as the difference between the annual return and the mean return of

all firms from the CRSP value-weighted market index.

Total Assets is Compustat Data 6;

Debt is total long-term debt (Compustat Data 9);

Bonus is measured as the average bonus of the firm’s executives in each sample year

scaled by end-of-year total assets (Compustat Data 6). Bonus data is obtained from the

Compustat Executive Compensation database and is defined by Compustat as “the dollar

value of a bonus (cash and non-cash) earned by the named executive officer during the

fiscal year”;

Bonus/Total Assets is variable Bonus divided by end-of-year total assets (Compustat Data

6);

FreeC

t-1

2-t

2- t to4-t1-t

AssetsCurrent

esexpenditur capital Average - operations fromCash

Dir

t-1

is the number of directors on the board in the year prior to the restatement year;

Dir

t+3

is the number of directors on the board in the third year following the restatement

year;

OutDir

t-1

is the number of outside directors on the board in the year prior to the

restatement year. An outside director is defined as a director who is not a present or

former firm employee and whose only formal connection to the firm is service as a

director on the board;

OutDir

t+3

is the number of outside directors on the board in the third year following the

restatement year;

ACMeet

t-1

is the number of audit committee meetings in the year prior to the restatement

year;

ACMeet

t+3

is the number of audit committee meetings in the third year following the

restatement year;

ACMem

t-1

is the number of audit committee members in the year prior to the restatement

year;

ACMem

t+3

is the number of audit committee members in the third year following the

restatement year;

Table 4.2 (Continued)

Variable definitions are as follows (continued):

ACIndDir

t-1

is the number of outside directors on the audit committee in the year prior to

the restatement year.

ACIndDir

t+3

is the number of outside directors on the audit committee in the third year

following the restatement year;

ACFinExp

t-1

is the number of financial experts on the audit committee in the year prior to

the restatement year. A financial expert is defined as someone who has accounting or

financial management expertise (i.e. CPA, CFO, PFO, CAO, PAO, controller, auditor,

accounting or finance professor);

ACFinExp

t+3

is the number of financial experts on the audit committee in the third year

following the restatement year;

MgmtChg is coded 1 if there is a change in top management following the restatement, 0

otherwise. Top management is defined as Chairman, CEO, and/or President;

AudChg is coded 1 if there is a change in auditor following the restatement, 0 otherwise.

Table 4.3

Pearson (above the diagonal) and Spearman (below the diagonal) Correlation

Matrix for the Primary Variables

(n=612)

EPS/

Price

Return EXT

FIN

OutDir AC

Meet

IndDir

FinExp

Mgmt

Chg

Aud

Chg

EPS/Price

0.141

0.001

-0.177

<0.0001

0.051

0.206

0.063

0.121

-0.010

0.797

0.045

0.270

-0.006

0.884

0.008

0.852

Return

0.307

<0.0001

0.013

0.744

0.027

0.508

0.001

0.985

0.034

0.407

0.020

0.617

-0.034

0.395

-0.034

0.409

ExtFin

-0.124

0.002

0.028

0.487

-0.134

0.001

-0.029

0.468

-0.030

0.457

0.064

0.114

-0.152

0.000

0.004

0.931

OutDir

0.026

0.520

-0.043

0.290

-0.134

0.001

0.036

0.373

0.355

<0.0001

0.209

<0.0001

0.016

0.692

-0.144

0.000

ACMeet

0.077

0.056

0.015

0.715

-0.029

0.468

0.036

0.373

-0.115

0.005

0.178

<0.0001

-0.112

0.006

0.060

0.140

ACIndDir

-0.001

0.978

-0.015

0.718

-0.030

0.457

0.355

<0.0001

-0.115

0.005

0.058

0.154

-0.004

0.920

-0.003

0.941

ACFinExp

0.025

0.534

0.011

0.793

0.064

0.114

0.209

<0.0001

0.178

<0.0001

0.058

0.154

0.101

0.012

-0.041

0.312

MgmtChg

-0.006

0.875

0.013

0.747

-0.152

0.000

0.016

0.692

-0.112

0.006

-0.004

0.920

0.101

0.012

-0.025

0.532

AudChg

0.013

0.752

-0.022

0.596

0.004

0.931

-0.144

<0.0001

0.060

0.140

-0.003

0.941

-0.041

0.312

-0.025

0.532

The p-values are reported below each correlation coefficient.

Variable definitions are as follows:

EPS = Earnings per share (Compustat Data 58);

Price = beginning of fiscal year stock price;

Returns = the compounded monthly returns beginning nine months prior to fiscal

year end to three months after fiscal year end;

EXTFIN = 1 if firm’s FreeC

t-1

is less than -0.5, and 0 otherwise. Where FreeC

t-1

defined as:

FreeC

t-1

2-t

2- t to4-t1-t

AssetsCurrent

esexpenditur capital Average - operations fromCash

OutDir = 1 if there is an increase in the number of outside directors from the year

prior to the third year following the restatement, 0 otherwise. An outside

director is defined as a director who is not a present or former firm

employee and whose only formal connection to the firm is service as a

director on the board.

Table 4.3 (Continued)

Variable definitions are as follows (continued):

ACMeet = 1 if there is an increase in the number of audit committee meetings from

the year prior to the third year following the restatement, 0 otherwise;

ACIndDir = 1 if there is an increase in the number of independent members on the

audit committee from the year prior to the third year following the

restatement, 0 otherwise;

ACFinExp= 1 if there is an increase in the number of financial experts on the audit

committee from the year prior to the third year following the restatement;

0 otherwise. A financial expert is defined as someone who has accounting

or financial management expertise (i.e. CPA, CFO, PFO, CAO, PAO,

controller, auditor, accounting or finance professor);

MgmtChg = 1 if there is a change in top management following the restatement, 0

otherwise. Top management is defined as Chairman, CEO, and/or

President;

AudChg = 1 if there is a change in auditor following the restatement, 0 otherwise.

Table 4.4

Tests of Conditional Conservatism, Pre and Post Restatement for

the Period 1994–2004

it-1

= β

PRE*POS + β

PRE*NEG + β

POST*POS + β

POST*NEG

+ β

PRE*POS*R

+ β

PRE*NEG*R

+ β

POST*POS*R

+ β

POST*NEG*R

+ ε

, (2)

Panel A: Annual Returns (n = 672)

Adj. R

Coefficients 0.027 0.039 0.038 0.057 0.015 0.063 -0.004 0.190

13.28%

(t-value)

3.29* 3.00* 5.15* 4.19*

1.70*** 1.79***

-0.78 5.79*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conditional Conservatism Post-restatement, β

– β

= 0.194

Conditional Conservatism Pre-restatement, β

– β

= 0.048

Test H1: (β

– β

) = (β

– β

), F-statistic = 8.84 p=0.003

Panel B: Equally Weighted Market-Adjusted Annual Returns (n = 672)

Adj. R

Coefficients 0.022 0.042 0.037 0.054 0.022 0.048 -0.003 0.157 12.43%

(t-value)

2.53** 3.72* 4.25* 5.12* 2.19** 1.99** -0.45 6.04*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conditional Conservatism Post-restatement, β

– β

= 0.160

Conditional Conservatism Pre-restatement, β

– β

= 0.026

Test H1: (β

– β

) = (β

– β

), F-statistic = 12.70 p=0.0004

Table 4.4 (Continued)

Panel C: Value Weighted Market-Adjusted Annual Returns (n = 672)

Adj. R

Coefficients 0.031 0.042 0.038 0.046 0.015 0.055 -0.003 0.200 13.72%

(t-value) 3.23* 3.70* 5.17* 3.95* 1.48 2.10** -0.63 6.00*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conditional Conservatism Post-restatement, β

– β

= 0.203

Conditional Conservatism Pre-restatement, β

– β

= 0.040

Test H1: (β

– β

) = (β

– β

), F-statistic = 13.88 p=0.0002

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.5

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on the Demand for External Financing for the

Period 1994-2004

it-1

= β

PRE*EXTFINHI*POS + β

PRE*EXTFINLO*POS

+ β

POST*EXTFINHI*POS + β

POST*EXTFINLO*POS

+ β

PRE*EXTFINHI*R*POS + β

PRE*EXTFINLO*R*POS

+ β

POST*EXTFINHI*R*POS + β

POST*EXTFINLO*R*POS

+ β

PRE*EXTFINHI*NEG + β

PRE*EXTFINLO*NEG

+ β

POST*EXTFINHI*NEG + β

POST*EXTFINLO*NEG

+ β

PRE*EXTFINHI*R*NEG + β

PRE*EXTFINLO*R*NEG

+ β

POST*EXTFINHI*R*NEG + β

POST*EXTFINLO*R*NEG

+ ε

, (3)

Panel A: Annual Returns (n = 672)

Adj. R

Coefficient t-value

-0.065 -1.77*** 20.62%

0.032 4.02*

0.157 3.29*

0.042 5.72*

0.006 0.16

0.018 2.05**

-0.275 -5.49*

-0.001 -0.20

-0.041 -0.83

0.044 3.42*

0.025 0.59

0.059 4.35*

0.032 0.25

0.065 1.87***

0.181 1.89***

0.191 5.75*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for high financing firms in post-restatement period, β

– β

= 0.456

Conservatism for high financing firms in pre-restatement period, β

– β

= 0.026

Conservatism for low financing firms in post-restatement period, β

– β

= 0.192

Conservatism for low financing firms in pre-restatement period, β

– β

= 0.047

Test H2a: (β

- β

) = (β

- β

), F-statistic 5.45 p=0.02

H2b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 2.52 p=0.113

Table 4.5 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 672)

Equal Weighte

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

-0.144 -4.21* 20.66% -0.126 -2.35** 21.84%

0.034 3.86* 0.037 4.05*

-0.025 -0.54 0.072 2.04**

0.045 5.42* 0.043 6.00*

0.080 2.11** 0.061 1.18

0.021 2.05** 0.018 1.82***

-0.191 -3.17* -0.245 -5.24*

-0.002 -0.28 -0.001 -0.17

0.051 0.91 -0.055 -1.31

0.042 3.83* 0.048 4.32*

0.092 2.35** 0.135 2.49**

0.052 4.95* 0.042 3.65*

0.145 1.22 -0.025 -0.27

0.044 1.87*** 0.061 2.35**

0.293 3.06* 0.457 3.58*

0.147 5.75* 0.183 5.54*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for high financing firms in post-restatement period, β

– β

= 0.484

Conservatism for high financing firms in pre-restatement period, β

– β

= 0.065

Conservatism for low financing firms in post-restatement period, β

– β

= 0.149

Conservatism for low financing firms in pre-restatement period, β

– β

= 0.023

Test H2a: (β

- β

) = (β

- β

), F-statistic 8.33 p=0.004

H2b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 2.89 p=0.090

Value Weighted Market-Adjusted Returns

Conservatism for high financing firms in post-restatement period, β

– β

= 0.702

Conservatism for high financing firms in pre-restatement period, β

– β

= -0.086

Conservatism for low financing firms in post-restatement period, β

– β

= 0.184

Conservatism for low financing firms in pre-restatement period, β

– β

= 0.043

Test H2a: (β

- β

) = (β

- β

), F-statistic 13.76 p=0.0002

H2b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 13.33 p=0.0003

Table 4.5 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

EXTFINHI = 1 if firm’s FreeC

t-1

is less than -0.5, and 0 otherwise. Where FreeC

t-1

defined as:

FreeC

t-1

2-t

2- t to4-t1-t

AssetsCurrent

esexpenditur capital Average - operations fromCash

EXTFINLO = 1 if firm’s FreeC

t-1

is greater than or equal to -0.5, and 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.6

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on Bonus Compensation for the Period 1994-2004

it-1

= β

PRE*BONUSHI*POS + β

PRE*BONUSLO*POS

+ β

POST*BONUSHI*POS + β

POST*BONUSLO*POS

+ β

PRE*BONUSHI*R*POS + β

PRE*BONUSLO*R*POS

+ β

POST*BONUSHI*R*POS + β

POST*BONUSLO*R*POS

+ β

PRE*BONUSHI*NEG + β

PRE*BONUSLO*NEG

+ β

POST*BONUSHI*NEG + β

POST*BONUSLO*NEG

+ β

PRE*BONUSHI*R*NEG + β

PRE*BONUSLO*R*NEG

+ β

POST*BONUSHI*R*NEG + β

POST*BONUSLO*R*NEG

+ ε

, (4)

Panel A: Annual Returns (n = 360)

Adj. R

Coefficient t-value

0.032 3.01* 22.6%

0.037 3.53*

0.070 5.77*

0.024 1.82***

0.020 1.67***

-0.005 -0.26

-0.023 -1.26

-0.026 -1.31

0.022 0.97

0.013 0.68

0.028 1.43

0.039 2.34**

-0.050 -0.67

0.022 0.44

0.047 0.91

0.097 2.08**

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for high bonus firms in post-restatement period, β

– β

= 0.070

Conservatism for high bonus firms in pre-restatement period, β

– β

= -0.070

Conservatism for low bonus firms in post-restatement period, β

– β

= 0.123

Conservatism for low bonus firms in pre-restatement period, β

– β

= 0.027

Test H3a: (β

- β

) = (β

- β

), F-statistic 0.48 p=0.489

H3b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.14 p=0.710

Table 4.6 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 360)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.026 2.30** 25.71% 0.035 2.97* 25.74%

0.044 3.73* 0.044 3.55*

0.060 4.62* 0.071 5.89*

0.024 1.41 0.017 1.40

0.028 2.01** 0.022 1.59

-0.022 -1.11 -0.022 -1.08

-0.020 -0.90 -0.028 -1.39

-0.029 -1.17 -0.018 -0.86

0.045 2.34** 0.033 1.98**

0.029 2.05** 0.036 2.42**

0.095 6.52* 0.076 4.64*

0.016 1.37 0.030 1.92***

0.007 0.15 -0.009 -0.19

0.039 1.37 0.055 1.54

0.187 5.38* 0.240 4.48*

0.019 0.57 0.078 1.43

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for high bonus firms in post-restatement period, β

– β

= 0.207

Conservatism for high bonus firms in pre-restatement period, β

– β

= -0.021

Conservatism for low bonus firms in post-restatement period, β

– β

= 0.048

Conservatism for low bonus firms in pre-restatement period, β

– β

= 0.061

Test H3a: (β

- β

) = (β

- β

), F-statistic 7.41 p=0.007

H3b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 8.24 p=0.004

Value Weighted Market-Adjusted Returns

Conservatism for high bonus firms in post-restatement period, β

– β

= 0.268

Conservatism for high bonus firms in pre-restatement period, β

– β

= -0.031

Conservatism for low bonus firms in post-restatement period, β

– β

= 0.096

Conservatism for low bonus firms in pre-restatement period, β

– β

= 0.077

Test H3a: (β

- β

) = (β

- β

), F-statistic 4.45 p=0.036

H3b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 7.32 p=0.007

Table 4.6 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

BONUSHI = 1 if a firm’s dollar value of a bonus is greater than the median bonus of all

sample firms, otherwise BONUSHI = 0. BONUS is measured as the

average bonus of the firm’s executives scaled by end-of-year total assets

(Compustat Data 6). Bonus data is obtained from the Compustat

Executive Compensation database and is defined by Compustat as “the

dollar value of a bonus (cash and non-cash) earned by the named

executive officer during the fiscal year”;

BONUSLO = 1 if a firm’s dollar value of a bonus is less than the median bonus of all

sample firms, otherwise BONUSLO = 0;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.7

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Outside Directors for the Period

1994-2004

it-1

= β

PRE*OUTDIRYES*POS + β

PRE*OUTDIRNO*POS

+ β

POST*OUTDIRYES*POS + β

POST*OUTDIRNO*POS

+ β

PRE*OUTDIRYES*R*POS + β

PRE*OUTDIRNO*R*POS

+ β

POST*OUTDIRYES*R*POS + β

POST*OUTDIRNO*R*POS

+ β

PRE*OUTDIRYES*NEG + β

PRE*OUTDIRNO*NEG

+ β

POST*OUTDIRYES*NEG + β

POST*OUTDIRNO*NEG

+ β

PRE*OUTDIRYES*R*NEG + β

PRE*OUTDIRNO*R*NEG

+ β

POST*OUTDIRYES*R*NEG + β

POST*OUTDIRNO*R*NEG

+ ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.034 2.82* 15.48%

0.025 2.23**

0.052 4.73*

0.024 2.01**

0.017 1.48

0.010 0.71

-0.008 -1.35

0.017 1.23

0.046 2.29**

0.040 2.09**

0.025 1.18

0.077 4.26*

0.068 1.25

0.076 1.46

0.102 2.02**

0.245 5.78*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for OutDirYes firms in post-restatement period, β

– β

= 0.110

Conservatism for OutDirYes firms in pre-restatement period, β

– β

= 0.051

Conservatism for OutDirNo firms in post-restatement period, β

– β

= 0.228

Conservatism for OutDirNo firms in pre-restatement period, β

– β

= 0.066

Test H4a: (β

- β

) = (β

- β

), F-statistic 3.11 p=0.078

H4b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 1.04 p=0.308

Table 4.7 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.031 2.22** 15.54% 0.027 1.75*** 16.50%

0.018 1.51 0.037 2.95*

0.044 3.42* 0.040 3.70*

0.016 1.20 0.029 2.52**

0.021 1.68*** 0.023 1.81***

0.022 1.33 0.001 0.08

-0.007 -1.21 -0.006 -1.08

0.040 2.60* 0.022 1.60

0.044 2.59* 0.052 3.22*

0.050 3.22* 0.039 2.41**

0.050 2.91* 0.046 2.58*

0.062 4.53* 0.050 3.26*

0.035 1.03 0.055 1.51

0.082 2.31** 0.066 1.75***

0.123 2.85* 0.154 3.07*

0.193 5.94* 0.246 5.70*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for OutDirYes firms in post-restatement period, β

– β

= 0.130

Conservatism for OutDirYes firms in pre-restatement period, β

– β

= 0.014

Conservatism for OutDirNo firms in post-restatement period, β

– β

= 0.153

Conservatism for OutDirNo firms in pre-restatement period, β

– β

= 0.060

Test H4a: (β

- β

) = (β

- β

), F-statistic 0.17 p=0.682

H4b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.09 p=0.759

Value Weighted Market-Adjusted Returns

Conservatism for OutDirYes firms in post-restatement period, β

– β

= 0.160

Conservatism for OutDirYes firms in pre-restatement period, β

– β

= 0.032

Conservatism for OutDirNo firms in post-restatement period, β

– β

= 0.224

Conservatism for OutDirNo firms in pre-restatement period, β

– β

= 0.065

Test H4a: (β

- β

) = (β

- β

), F-statistic 0.88 p=0.348

H4b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.12 p=0.724

Table 4.7 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

OutDirYes = 1 if there is an increase in the number of outside directors from the year

prior to the third year following the restatement, 0 otherwise. An

outside director is defined as a director who is not a present or former

firm employee and whose only formal connection to the firm is service

as a director on the board;

OutDirNo = 1 if there is no increase in the number of outside directors from the year

prior to the third year following the restatement, 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.8

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Audit Committee Meetings for the

Period 1994-2004

it-1

= β

PRE*ACMEETYES*POS + β

PRE*ACMEETNO*POS

+ β

POST*ACMEETYES*POS + β

POST*ACMEETNO*POS

+ β

PRE*ACMEETYES*R*POS + β

PRE*ACMEETNO*R*POS

+ β

POST*ACMEETYES*R*POS + β

POST*ACMEETNO*R*POS

+ β

PRE*ACMEETYES*NEG + β

PRE*ACMEETNO*NEG

+ β

POST*ACMEETYES*NEG + β

POST*ACMEETNO*NEG

+ β

PRE*ACMEETYES*R*NEG + β

PRE*ACMEETNO*R*NEG

+ β

POST*ACMEETYES*R*NEG + β

POST*ACMEETNO*R*NEG

+ ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.036 1.85*** 14.79%

0.048 1.68***

0.055 2.86*

0.038 1.13

0.023 1.20

0.066 0.89

-0.013 -0.92

0.213 2.61*

-0.002 -0.29

-0.002 -0.81

-0.004 -0.60

0.002 0.70

-0.002 -0.24

0.002 0.51

0.002 0.09

-0.004 -0.40

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for ACMeetYes firms in post-restatement period, β

– β

= 0.015

Conservatism for ACMeetYes firms in pre-restatement period, β

– β

= -0.025

Conservatism for ACMeetNo firms in post-restatement period, β

– β

= -0.217

Conservatism for ACMeetNo firms in pre-restatement period, β

– β

= -0.064

Test H5a: (β

- β

) = (β

- β

), F-statistic 0.33 p=0.568

H5b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.29 p=0.588

Table 4.8 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.023 2.47** 14.77% 0.034 3.42* 16.36%

0.018 0.69 -0.001 -0.02

0.041 4.38* 0.041 5.25*

0.022 0.85 0.030 1.26

0.022 2.16** 0.014 1.35

0.019 0.50 0.042 0.94

-0.003 -0.44 -0.003 -0.51

0.015 0.80 0.006 0.33

0.046 3.83* 0.045 3.73*

0.049 1.27 0.045 1.35

0.061 5.41* 0.056 4.50*

0.036 1.03 -0.007 -0.21

0.041 1.57 0.050 1.77***

0.149 1.93** 0.132 1.72***

0.171 6.33* 0.235 6.70*

0.131 1.36 0.016 0.18

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for ACMeetYes firms in post-restatement period, β

– β

= 0.174

Conservatism for ACMeetYes firms in pre-restatement period, β

– β

= 0.019

Conservatism for ACMeetNo firms in post-restatement period, β

– β

= 0.116

Conservatism for ACMeetNo firms in pre-restatement period, β

– β

= 0.130

Test H5a: (β

- β

) = (β

- β

), F-statistic 0.32 p=0.572

H5b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 1.54 p=0.215

Value Weighted Market-Adjusted Returns

Conservatism for ACMeetYes firms in post-restatement period, β

– β

= 0.238

Conservatism for ACMeetYes firms in pre-restatement period, β

– β

= 0.036

Conservatism for ACMeetNo firms in post-restatement period, β

– β

= 0.010

Conservatism for ACMeetNo firms in pre-restatement period, β

– β

= 0.090

Test H5a: (β

- β

) = (β

- β

), F-statistic 5.19 p=0.023

H5b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 4.25 p=0.04

Table 4.8 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

ACMeetYes = 1 if there is an increase in the number of audit committee meetings from

the year prior to the third year following the restatement, 0 otherwise;

ACMeetNo = 1 if there is no increase in the number of audit committee meetings from

the year prior to the third year following the restatement, 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.9

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Independent Directors on Audit

Committee for the Period 1994-2004

it-1

= β

PRE*ACINDDIRYES*POS + β

PRE*ACINDDIRNO*POS

+ β

POST*ACINDDIRYES*POS + β

POST*ACINDDIRNO*POS

+ β

PRE*ACINDDIRYES*R*POS + β

PRE*ACINDDIRNO*R*POS

+ β

POST*ACINDDIRYES*R*POS + β

POST*ACINDDIRNO*R*POS

+ β

PRE*ACINDDIRYES*NEG + β

PRE*ACINDDIRNO*NEG

+ β

POST*ACINDDIRYES*NEG + β

POST*ACINDDIRNO*NEG

+ β

PRE*ACINDDIRYES*R*NEG + β

PRE*ACINDDIRNO*R*NEG

+ β

POST*ACINDDIRYES*R*NEG

+ β

POST*ACINDDIRNO*R*NEG + ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.035 2.49** 14.61%

0.024 2.41**

0.045 3.98*

0.040 3.51*

0.009 0.64

0.017 1.55

-0.004 -0.61

-0.006 -0.46

0.019 0.88

0.059 3.25*

0.041 1.81***

0.065 3.70*

0.027 0.50

0.103 1.99**

0.150 3.04*

0.214 4.86*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for ACIndDirYes firms in post-restatement period, β

– β

= 0.154

Conservatism for ACIndDirYes firms in pre-restatement period, β

– β

= 0.018

Conservatism for ACIndDirNo firms in post-restatement period, β

– β

= 0.220

Conservatism for ACIndDirNo firms in pre-restatement period, β

– β

= 0.086

Test H6a: (β

- β

) = (β

- β

), F-statistic 0.97 p=0.325

H6b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.00 p=0.994

Table 4.9 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.027 1.78*** 14.09% 0.027 1.60 15.46%

0.020 1.84*** 0.033 2.84*

0.046 3.57* 0.050 4.20*

0.029 2.17** 0.033 3.12*

0.017 1.05 0.017 1.01

0.025 1.99** 0.014 1.18

-0.004 -0.60 -0.004 -0.65

0.011 0.79 0.002 0.17

0.049 2.59* 0.049 2.78*

0.044 2.99* 0.040 2.65*

0.066 3.26* 0.029 1.52

0.056 4.38* 0.061 4.10*

0.074 1.97** 0.073 1.95***

0.037 1.12 0.045 1.19

0.193 4.04* 0.149 2.97*

0.156 4.97* 0.248 5.68*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for ACIndDirYes firms in post-restatement period, β

– β

= 0.197

Conservatism for ACIndDirYes firms in pre-restatement period, β

– β

= 0.057

Conservatism for ACIndDirNo firms in post-restatement period, β

– β

= 0.145

Conservatism for ACIndDirNo firms in pre-restatement period, β

– β

= 0.012

Test H6a: (β

- β

) = (β

- β

), F-statistic 0.75 p=0.387

H6b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.01 p=0.933

Value Weighted Market-Adjusted Returns

Conservatism for ACIndDirYes firms in post-restatement period, β

– β

= 0.153

Conservatism for ACIndDirYes firms in pre-restatement period, β

– β

= 0.056

Conservatism for ACIndDirNo firms in post-restatement period, β

– β

= 0.246

Conservatism for ACIndDirNo firms in pre-restatement period, β

– β

= 0.031

Test H6a: (β

- β

) = (β

- β

), F-statistic 1.87 p=0.172

H6b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 1.78 p=0.182

Table 4.9 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

ACIndDirYes = 1 if there is an increase in the number of independent members on the

audit committee from the year prior to the third year following the

restatement, 0 otherwise. An independent director is defined as a

director who is not a present or former firm employee and whose only

formal connection to the firm is service as a director on the board;

ACIndDirNo = 1 if there is no increase in the number of independent members on the

audit committee from the year prior to the third year following the

restatement, 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.10

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Financial Experts on Audit

Committee for the Period 1994-2004

it-1

= β

PRE*ACFINEXPYES*POS + β

PRE*ACFINEXPNO*POS

+ β

POST*ACFINEXPYES*POS + β

POST*ACFINEXPNO*POS

+ β

PRE*ACFINEXPYES*R*POS + β

PRE*ACFINEXPNO*R*POS

+ β

POST*ACFINEXPYES*R*POS + β

POST*ACFINEXPNO*R*POS

+ β

PRE*ACFINEXPYES*NEG + β

PRE*ACFINEXPNO*NEG

+ β

POST*ACFINEXPYES*NEG + β

POST*ACFINEXPNO*NEG

+ β

PRE*ACFINEXPYES*R*NEG + β

PRE*ACFINEXPNO*R*NEG

+ β

POST*ACFINEXPYES*R*NEG

+ β

POST*ACFINEXPNO*R*NEG + ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.025 2.23** 15.05%

0.031 2.59*

0.046 4.73*

0.023 1.70***

0.017 1.26

0.013 1.08

-0.007 -1.30

0.021 1.44

0.052 2.89*

0.029 1.34

0.057 3.05*

0.053 2.61*

0.088 1.76***

0.045 0.78

0.163 3.68*

0.210 4.40*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for ACFinExpYes firms in post-restatement period, β

– β

= 0.170

Conservatism for ACFinExpYes firms in pre-restatement period, β

– β

= 0.071

Conservatism for ACFinExpNo firms in post-restatement period, β

– β

= 0.189

Conservatism for ACFinExpNo firms in pre-restatement period, β

– β

= 0.032

Test H7a: (β

- β

) = (β

- β

), F-statistic 0.07 p=0.785

H7b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.30 p=0.581

Table 4.10 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.019 1.57 15.15% 0.020 1.56 17.10%

0.026 1.99** 0.042 3.05*

0.041 3.44* 0.039 3.92*

0.022 1.49 0.031 2.57**

0.027 1.83*** 0.025 1.72***

0.018 1.36 0.007 0.53

-0.006 -0.95 -0.006 -0.98

0.033 2.07** 0.020 1.45

0.044 2.86* 0.046 3.10*

0.051 2.94* 0.043 2.42**

0.073 5.19* 0.077 5.21*

0.039 2.36** 0.004 0.20

0.041 1.29 0.051 1.46

0.079 2.02** 0.066 1.68***

0.179 5.22* 0.254 6.00*

0.150 3.74* 0.125 2.45**

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for ACFinExpYes firms in post-restatement period, β

– β

= 0.185

Conservatism for ACFinExpYes firms in pre-restatement period, β

– β

= 0.014

Conservatism for ACFinExpNo firms in post-restatement period, β

– β

= 0.117

Conservatism for ACFinExpNo firms in pre-restatement period, β

– β

= 0.061

Test H7a: (β

- β

) = (β

- β

), F-statistic 1.48 p=0.225

H7b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 2.17 p=0.142

Value Weighted Market-Adjusted Returns

Conservatism for ACFinExpYes firms in post-restatement period, β

– β

= 0.260

Conservatism for ACFinExpYes firms in pre-restatement period, β

– β

= 0.026

Conservatism for ACFinExpNo firms in post-restatement period, β

– β

= 0.105

Conservatism for ACFinExpNo firms in pre-restatement period, β

– β

= 0.059

Test H7a: (β

- β

) = (β

- β

), F-statistic 5.18 p=0.023

H7b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 4.55 p=0.033

Table 4.10 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

ACFinExpYes =1 if there is an increase in the number of financial experts on the audit

committee from the year prior to the third year following the

restatement; 0 otherwise. A financial expert is defined as someone who

has accounting or financial management expertise (i.e. CPA, CFO, PFO,

CAO, PAO, controller, auditor, accounting or finance professor);

ACFinExpNo =1 if there is no increase in the number of financial experts on the audit

committee from the year prior to the third year following the

restatement; 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.11

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Top Management for the

Period 1994-2004

it-1

= β

PRE*MGMTCHGYES*POS + β

PRE*MGMTCHGNO*POS

+ β

POST*MGMTCHGYES*POS + β

POST*MGMTCHGNO*POS

+ β

PRE*MGMTCHGYES*R*POS + β

PRE*MGMTCHGNO*R*POS

+ β

POST*MGMTCHGYES*R*POS

+ β

POST*MGMTCHGNO*R*POS + β

PRE*MGMTCHGYES*NEG

+ β

PRE*MGMTCHGNO*NEG + β

POST*MGMTCHGYES*NEG

+ β

POST*MGMTCHGNO*NEG + β

PRE*MGMTCHGYES*R*NEG

+ β

PRE*MGMTCHGNO*R*NEG

+ β

POST*MGMTCHGYES*R*NEG

+ β

POST*MGMTCHGNO*R*NEG + ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.031 3.10* 14.64%

0.022 1.51

0.038 3.59*

0.043 3.57*

0.009 0.78

0.025 1.70***

0.001 0.10

-0.005 -0.83

0.039 2.24**

0.049 2.13**

0.067 3.75*

0.040 1.85***

0.058 1.20

0.092 1.53

0.229 5.32*

0.128 2.57**

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for MgmtChgYes firms in post-restatement period, β

– β

= 0.228

Conservatism for MgmtChgYes firms in pre-restatement period, β

– β

= 0.049

Conservatism for MgmtChgNo firms in post-restatement period, β

– β

= 0.133

Conservatism for MgmtChgNo firms in pre-restatement period, β

– β

= 0.067

Test H8a: (β

- β

) = (β

- β

), F-statistic 2.00 p=0.158

H8b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 1.18 p=0.278

Table 4.11 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.031 2.86* 14.82% 0.040 3.50* 15.76%

0.005 0.32 0.010 0.60

0.027 2.12** 0.036 3.50*

0.044 3.20* 0.039 3.26*

0.010 0.78 0.003 0.25

0.044 2.70* 0.037 2.28**

0.020 1.46 0.008 0.67

-0.005 -0.89 -0.004 -0.73

0.039 2.74* 0.033 2.31**

0.062 3.09* 0.063 3.33*

0.065 4.98* 0.050 3.35*

0.042 2.14** 0.044 2.31**

0.038 1.24 0.033 0.94

0.087 2.13** 0.095 2.39**

0.184 6.12* 0.216 5.48*

0.120 2.27** 0.181 3.00*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for MgmtChgYes firms in post-restatement period, β

– β

= 0.164

Conservatism for MgmtChgYes firms in pre-restatement period, β

– β

= 0.028

Conservatism for MgmtChgNo firms in post-restatement period, β

– β

= 0.125

Conservatism for MgmtChgNo firms in pre-restatement period, β

– β

= 0.043

Test H8a: (β

- β

) = (β

- β

), F-statistic 0.38 p=0.537

H8b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.41 p=0.523

Value Weighted Market-Adjusted Returns

Conservatism for MgmtChgYes firms in post-restatement period, β

– β

= 0.208

Conservatism for MgmtChgYes firms in pre-restatement period, β

– β

= 0.030

Conservatism for MgmtChgNo firms in post-restatement period, β

– β

= 0.185

Conservatism for MgmtChgNo firms in pre-restatement period, β

– β

= 0.058

Test H8a: (β

- β

) = (β

- β

), F-statistic 0.10 p=0.755

H8b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.30 p=0.585

Table 4.11 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

MgmtChgYes = 1 if there is a change in top management following the restatement, 0

otherwise. Top management is defined as Chairman, CEO, and/or

President;

MgmtChgNo = 1 if there is no change in top management following the restatement, 0

otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.12

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on a Change in Auditor for the Period 1994-2004

it-1

= β

PRE*AUDCHGYES*POS + β

PRE*AUDCHGNO*POS

+ β

POST*AUDCHGYES*POS + β

POST*AUDCHGNO*POS

+ β

PRE*AUDCHGYES*R*POS + β

PRE*AUDCHGNO*R*POS

+ β

POST*AUDCHGYES*R*POS + β

POST*AUDCHGNO*R*POS

+ β

PRE*AUDCHGYES*NEG + β

PRE*AUDCHGNO*NEG

+ β

POST*AUDCHGYES*NEG + β

POST*AUDCHGNO*NEG

+ β

PRE*AUDCHGYES*R*NEG + β

PRE*AUDCHGNO*R*NEG

+ β

POST*AUDCHGYES*R*NEG + β

POST*AUDCHGNO*R*NEG

+ ε

, (5)

Panel A: Annual Returns (n = 642)

Coefficient Adj. R

t-value

0.028 1.61 14.34%

0.026 2.75*

0.008 0.59

0.046 5.00*

0.015 0.57

0.015 1.58

0.038 3.10*

-0.013 -2.13**

0.050 2.35**

0.031 1.74***

0.075 3.14*

0.052 3.00*

0.091 1.67***

0.038 0.75

0.226 3.53*

0.178 4.46*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for AudChgYes firms in post-restatement period, β

– β

= 0.188

Conservatism for AudChgYes firms in pre-restatement period, β

– β

= 0.076

Conservatism for AudChgNo firms in post-restatement period, β

– β

= 0.191

Conservatism for AudChgNo firms in pre-restatement period, β

– β

= 0.023

Test H9a: (β

- β

) = (β

- β

), F-statistic 0.00 p=0.965

H9b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.25 p=0.614

Table 4.12 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 642)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.022 1.14 13.55% 0.021 0.96 15.06%

0.021 2.00** 0.033 2.95*

0.022 1.43 0.028 2.15**

0.040 3.59* 0.040 4.32*

0.037 1.25 0.034 1.06

0.021 1.91*** 0.013 1.18

0.036 2.70* 0.030 2.49**

-0.010 -1.65*** -0.011 -1.83***

0.042 2.14** 0.060 3.03*

0.043 3.05* 0.033 2.29**

0.057 2.84* 0.026 1.09

0.061 4.65* 0.056 4.06*

0.051 1.23 0.088 2.11**

0.048 1.53 0.033 0.96

0.187 3.17* 0.179 2.26**

0.165 5.43* 0.216 5.77*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for AudChgYes firms in post-restatement period, β

– β

= 0.151

Conservatism for AudChgYes firms in pre-restatement period, β

– β

= 0.014

Conservatism for AudChgNo firms in post-restatement period, β

– β

= 0.175

Conservatism for AudChgNo firms in pre-restatement period, β

– β

= 0.027

Test H9a: (β

- β

) = (β

- β

), F-statistic 0.13 p=0.722

H9b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.01 p=0.908

Value Weighted Market-Adjusted Returns

Conservatism for AudChgYes firms in post-restatement period, β

– β

= 0.149

Conservatism for AudChgYes firms in pre-restatement period, β

– β

= 0.054

Conservatism for AudChgNo firms in post-restatement period, β

– β

= 0.227

Conservatism for AudChgNo firms in pre-restatement period, β

– β

= 0.020

Test H9a: (β

- β

) = (β

- β

), F-statistic 0.80 p=0.371

H9b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 1.07 p=0.302

Table 4.12 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

AudChgYes = 1 if there is a change in the auditor following the restatement, 0

otherwise;

AudChgNo = 1 if there is no change in the auditor following the restatement, 0

otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.13

Tests of Conditional Conservatism, Pre and Post Restatement,

Separated into Pre- and Post-SOX Periods for the Period

1994–2004

it-1

= β

PRE*POS + β

PRE*NEG + β

POST*POS + β

POST*NEG

+ β

PRE*POS*R

+ β

PRE*NEG*R

+ β

POST*POS*R

+ β

POST*NEG*R

+ ε

, (2)

Panel A: Annual Returns

PRE-SOX (n=270) POST-SOX (n=402)

Adj. R

=14.5% Adj. R

=13.6%

Coefficient t-value Coefficient t-value

0.028 2.32** 0.026 2.38**

0.035 1.51 0.041 2.63*

0.021 1.64 0.041 4.10*

0.041 2.17** 0.060 3.07*

0.010 0.71 0.019 1.62

0.061 0.91 0.065 1.58

-0.006 -0.97 0.005 0.49

0.199 4.70* 0.139 2.60*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

PRE-SOX:

Conditional Conservatism Post-restatement, β

– β

= 0.205

Conditional Conservatism Pre-restatement, β

– β

= 0.051

Test H1: (β

– β

) = (β

– β

), F-statistic = 3.59 p=0.059

POST-SOX:

Conditional Conservatism Post-restatement, β

– β

= 0.134

Conditional Conservatism Pre-restatement, β

– β

= 0.046

Test H1: (β

– β

) = (β

– β

), F-statistic = 1.61 p=0.205

Table 4.13 (Continued)

Panel B: Equally Weighted Market-Adjusted Annual Returns

PRE-SOX (n=270) POST-SOX (n=402)

Adj. R

=18.3% Adj. R

=13.0%

Coefficient t-value Coefficient t-value

0.029 2.28** 0.017 1.46

0.040 2.01** 0.044 3.20*

0.030 2.33** 0.033 2.83*

0.045 2.51** 0.048 3.70*

0.009 0.59 0.032 2.41**

0.053 1.03 0.049 1.78***

-0.007 -1.12 0.016 1.37

0.209 5.49* 0.067 1.84***

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

PRE-SOX:

Conditional Conservatism Post-restatement, β

– β

= 0.216

Conditional Conservatism Pre-restatement, β

– β

= 0.044

Test H1: (β

– β

) = (β

– β

), F-statistic = 6.79 p=0.010

POST-SOX:

Conditional Conservatism Post-restatement, β

– β

= 0.051

Conditional Conservatism Pre-restatement, β

– β

= 0.017

Test H1: (β

– β

) = (β

– β

), F-statistic = 0.46 p=0.496

Table 4.13 (Continued)

Panel C: Value Weighted Market-Adjusted Annual Returns

PRE-SOX (n=270) POST-SOX (n=402)

Adj. R

=13.7% Adj. R

=14.5%

Coefficient t-value Coefficient t-value

0.034 2.47** 0.028 2.15**

0.034 1.86*** 0.047 3.28*

0.025 1.97*** 0.039 4.07*

0.019 1.04 0.060 3.70*

0.007 0.41 0.020 1.57

0.036 0.83 0.067 2.04**

-0.006 -0.95 0.009 0.83

0.172 3.93* 0.181 3.16*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

PRE-SOX:

Conditional Conservatism Post-restatement, β

– β

= 0.178

Conditional Conservatism Pre-restatement, β

– β

= 0.029

Test H1: (β

– β

) = (β

– β

), F-statistic = 5.38 p=0.021

POST-SOX:

Conditional Conservatism Post-restatement, β

– β

= 0.172

Conditional Conservatism Pre-restatement, β

– β

= 0.047

Test H1: (β

– β

) = (β

– β

), F-statistic = 3.41 p=0.066

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.14

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on Outside Director Percentage for the Period

1994-2004

it-1

= β

PRE*OUTDIR%YES*POS + β

PRE*OUTDIR%NO*POS

+ β

POST*OUTDIR%YES*POS + β

POST*OUTDIR%NO*POS

+ β

PRE*OUTDIR%YES*R*POS + β

PRE*OUTDIR%NO*R*POS

+ β

POST*OUTDIR%YES*R*POS + β

POST*OUTDIR%NO*R*POS

+ β

PRE*OUTDIR%YES*NEG + β

PRE*OUTDIR%NO*NEG

+ β

POST*OUTDIR%YES*NEG + β

POST*OUTDIR%NO*NEG

+ β

PRE*OUTDIR%YES*R*NEG + β

PRE*OUTDIR%NO*R*NEG

+ β

POST*OUTDIR%YES*R*NEG

+ β

POST*OUTDIR%NO*R*NEG + ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.032 2.44** 15.38%

0.031 2.85*

0.030 2.25**

0.049 5.03*

-0.011 -0.56

0.019 1.95***

0.000 0.01

-0.005 -0.90

0.011 0.52

0.067 3.65*

0.065 3.14

0.048 2.58**

-0.001 -0.02

0.128 2.60*

0.208 4.07*

0.169 4.01*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for OutDir%Yes firms in post-restatement period, β

– β

= 0.208

Conservatism for OutDir%Yes firms in pre-restatement period, β

– β

= 0.010

Conservatism for OutDir%No firms in post-restatement period, β

– β

= 0.174

Conservatism for OutDir%No firms in pre-restatement period, β

– β

= 0.109

Test H4a: (β

- β

) = (β

- β

), F-statistic 0.24 p=0.622

H4b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 1.59 p=0.209

Table 4.14 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.018 1.34 14.63% 0.042 2.79* 16.42%

0.034 2.72* 0.032 2.46**

0.029 1.68*** 0.036 2.64*

0.042 3.88* 0.042 4.36*

0.002 0.10 -0.026 -1.17

0.023 2.04** 0.023 2.10**

0.013 0.52 0.005 0.23

-0.002 -0.35 -0.003 -0.50

0.038 2.08** 0.018 1.07

0.054 3.54* 0.065 4.30*

0.047 3.27* 0.036 2.15**

0.073 4.51* 0.060 3.74*

0.034 0.98 0.012 0.30

0.077 2.20** 0.096 2.81*

0.145 3.81* 0.190 3.58*

0.194 5.30* 0.224 5.32*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for OutDir%Yes firms in post-restatement period, β

– β

= 0.132

Conservatism for OutDir%Yes firms in pre-restatement period, β

– β

= 0.032

Conservatism for OutDir%No firms in post-restatement period, β

– β

= 0.196

Conservatism for OutDir%No firms in pre-restatement period, β

– β

= 0.054

Test H4a: (β

- β

) = (β

- β

), F-statistic 1.17 p=0.279

H4b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.27 p=0.603

Value Weighted Market-Adjusted Returns

Conservatism for OutDir%Yes firms in post-restatement period, β

– β

= 0.185

Conservatism for OutDir%Yes firms in pre-restatement period, β

– β

= 0.038

Conservatism for OutDir%No firms in post-restatement period, β

– β

= 0.227

Conservatism for OutDir%No firms in pre-restatement period, β

– β

= 0.073

Test H4a: (β

- β

) = (β

- β

), F-statistic 0.35 p=0.554

H4b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.01 p=0.932

Table 4.14 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

OUTDIR%YES = 1 if a firm’s outside director percentage in the third year following

the restatement announcement is greater than the median outside

director percentage for all sample firms over the same time period,

and 0 otherwise. An outside director is defined as a director who is

not a present or former firm employee and whose only formal

connection to the firm is service as a director on the board;

OUTDIR%NO = 1 if a firm’s outside director percentage in the third year following

the restatement announcement is less than or equal to the median

outside director percentage for all sample firms over the same time

period, and 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.15

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on Audit Committee Meeting Variables for Year

t+3

for the Period 1994-2004

it-1

= β

PRE*ACMEETYES

t+3

*POS + β

PRE*ACMEETNO

t+3

*POS

+ β

POST*ACMEETYES

t+3

*POS + β

POST*ACMEETNO

t+3

*POS

+ β

PRE*ACMEETYES

t+3

*R*POS + β

PRE*ACMEETNO

t+3

*R*POS

+ β

POST*ACMEETYES

t+3

*R*POS

+ β

POST*ACMEETNO

t+3

*R*POS + β

PRE*ACMEETYES

t+3

*NEG

+ β

PRE*ACMEETNO

t+3

*NEG + β

POST*ACMEETYES

t+3

*NEG

+ β

POST*ACMEETNO

t+3

*NEG + β

PRE*ACMEETYES

t+3

*R*NEG

+ β

PRE*ACMEETNO

t+3

*R*NEG

+ β

POST*ACMEETYES

t+3

*R*NEG

+ β

POST*ACMEETNO

t+3

*R*NEG + ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.022 1.90*** 15.31%

0.031 2.69*

0.045 3.42*

0.044 4.21*

0.031 2.17**

0.005 0.47

-0.016 -0.90

-0.003 -0.50

0.060 2.92*

0.029 1.53

0.054 2.83*

0.055 2.76*

0.099 1.84***

0.047 0.90

0.141 3.01*

0.215 4.77*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for ACMeetYes firms in post-restatement period, β

– β

= 0.157

Conservatism for ACMeetYes firms in pre-restatement period, β

– β

= 0.068

Conservatism for ACMeetNo firms in post-restatement period, β

– β

= 0.218

Conservatism for ACMeetNo firms in pre-restatement period, β

– β

= 0.042

Test H5a: (β

- β

) = (β

- β

), F-statistic 0.81 p=0.368

H5b: [(β

- β

) – (β

- β

)] = [(β

- β

) – (β

- β

)], F-statistic 0.73 p=0.395

Table 4.15 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.012 0.91 16.17% 0.029 2.26** 16.16%

0.029 2.38** 0.030 2.17**

0.033 2.09** 0.032 2.60*

0.040 3.46* 0.048 4.51*

0.048 2.98* 0.030 1.93***

0.007 0.59 0.007 0.53

0.010 0.48 0.003 0.16

-0.002 -0.33 -0.003 -0.62

0.046 2.99* 0.040 2.47**

0.048 2.83* 0.048 3.04*

0.039 2.67* 0.073 3.78*

0.076 4.92* 0.030 2.05**

0.037 1.11 0.040 1.05

0.074 2.06** 0.074 2.08**

0.078 2.06** 0.233 3.94*

0.241 6.74* 0.190 4.82*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for ACMeetYes firms in post-restatement period, β

– β

= 0.068

Conservatism for ACMeetYes firms in pre-restatement period, β

– β

= -0.011

Conservatism for ACMeetNo firms in post-restatement period, β

– β

= 0.243

Conservatism for ACMeetNo firms in pre-restatement period, β

– β

= 0.067

Test H5a: (β

- β

) > (β

- β

), F-statistic 9.54 p=0.002

H5b: [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)], F-statistic 1.54 p=0.215

Value Weighted Market-Adjusted Returns

Conservatism for ACMeetYes firms in post-restatement period, β

– β

= 0.230

Conservatism for ACMeetYes firms in pre-restatement period, β

– β

= 0.010

Conservatism for ACMeetNo firms in post-restatement period, β

– β

= 0.193

Conservatism for ACMeetNo firms in pre-restatement period, β

– β

= 0.067

Test H5a: (β

- β

) > (β

- β

), F-statistic 0.25 p=0.615

H5b: [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)], F-statistic 1.04 p=0.309

Table 4.15 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

ACMEETYES

t+3

= 1 if the firm’s number of audit committee meetings in the third

year following the restatement announcement is greater than the

median number of audit committee meetings for all sample firms

over the same time period, 0 otherwise;

ACMEETNO

t+3

= 1 if the firm’s number of audit committee meetings in the third

year following the restatement announcement is less than or equal

to the median number of audit committee meetings for all sample

firms over the same time period, 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.16

Tests of Conditional Conservatism, Pre and Post Restatement,

Conditional on the Financial Expert Percentage on Audit

Committee for the Period 1994-2004

it-1

= β

PRE*ACFINEXP%YES*POS + β

PRE*ACFINEXP%NO*POS

+ β

POST*ACFINEXP%YES*POS + β

POST*ACFINEXP%NO*POS

+ β

PRE*ACFINEXP%YES*R*POS + β

PRE*ACFINEXP%NO*R*POS

+ β

POST*ACFINEXP%YES*R*POS + β

POST*ACFINEXP%NO*R*POS

+ β

PRE*ACFINEXP%YES*NEG + β

PRE*ACFINEXP%NO*NEG

+ β

POST*ACFINEXP%YES*NEG + β

POST*ACFINEXP%NO*NEG

+ β

PRE*ACFINEXP%YES*R*NEG + β

PRE*ACFINEXP%NO*R*NEG

+ β

POST*ACFINEXP%YES*R*NEG

+ β

POST*ACFINEXP%NO*R*NEG + ε

, (5)

Panel A: Annual Returns (n = 612)

Coefficient Adj. R

t-value

0.023 2.09** 16.68%

0.030 2.52**

0.046 4.99*

0.013 0.98

0.030 2.27**

0.002 0.15

-0.010 -1.86***

0.037 2.77*

0.061 3.60*

0.004 0.15

0.045 2.65*

0.070 3.14*

0.095 2.01**

0.010 0.17

0.155 3.74*

0.230 4.50*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conservatism for ACFinExp%Yes firms in post-restatement period, β

– β

= 0.165

Conservatism for ACFinExp%Yes firms in pre-restatement period, β

– β

= 0.065

Conservatism for ACFinExp%No firms in post-restatement period, β

– β

= 0.193

Conservatism for ACFinExp%No firms in pre-restatement period, β

– β

= 0.008

Test H7a: (β

- β

) > (β

- β

), F-statistic 0.17 p=0.681

H7b: [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)], F-statistic 0.65 p=0.419

Table 4.16 (Continued)

Panel B: Market-Adjusted Annual Returns (n = 612)

Equal Weighted

Value Weighted

Coefficient Adj. R

Adj. R

t-value Coefficient t-value

0.013 1.05 16.25% 0.022 1.70*** 17.49%

0.028 2.15** 0.035 2.54**

0.040 3.84* 0.037 4.09*

0.012 0.71 0.027 2.01**

0.045 3.10* 0.034 2.32**

0.004 0.28 0.001 0.05

-0.007 -1.23 -0.007 -1.31

0.043 2.83* 0.034 2.47**

0.057 3.99* 0.054 3.84*

0.031 1.60 0.030 1.56

0.054 3.74* 0.062 4.10*

0.060 3.75* 0.030 1.70***

0.053 1.77*** 0.056 1.75***

0.058 1.38 0.060 1.35

0.143 4.04* 0.227 5.17*

0.189 5.01* 0.185 3.80*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Equal Weighted Market-Adjusted Returns

Conservatism for ACFinExp%Yes firms in post-restatement period, β

– β

= 0.150

Conservatism for ACFinExp%Yes firms in pre-restatement period, β

– β

= 0.008

Conservatism for ACFinExp%No firms in post-restatement period, β

– β

= 0.146

Conservatism for ACFinExp%No firms in pre-restatement period, β

– β

= 0.054

Test H7a: (β

- β

) > (β

- β

), F-statistic 0.01 p=0.936

H7b: [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)], F-statistic 0.42 p=0.515

Value Weighted Market-Adjusted Returns

Conservatism for ACFinExp%Yes firms in post-restatement period, β

– β

= 0.234

Conservatism for ACFinExp%Yes firms in pre-restatement period, β

– β

= 0.022

Conservatism for ACFinExp%No firms in post-restatement period, β

– β

= 0.151

Conservatism for ACFinExp%No firms in pre-restatement period, β

– β

= 0.059

Test H7a: (β

- β

) > (β

- β

), F-statistic 1.53 p=0.217

H7b: [(β

- β

) – (β

- β

)] > [(β

- β

) – (β

- β

)], F-statistic 1.81 p=0.179

Table 4.16 (Continued)

where

it-1

= Earnings per share (Compustat Data 58) divided by beginning of fiscal year

stock price;

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

ACFINEXP%YES = 1 if the firm’s percentage of audit committee members who are

financial experts during the third year following the restatement is

greater than the median percentage for all sample firms for the

same time period; 0 otherwise. A financial expert is defined as

someone who has accounting or financial management expertise

(i.e. CPA, CFO, PFO, CAO, PAO, controller, auditor, accounting

or finance professor);

ACFINEXP%NO = 1 if the firm’s percentage of audit committee members who are

financial experts during the third year following the restatement is

less than or equal to the median percentage for all sample firms for

the same time period; 0 otherwise;

= the 12-month compounded return beginning nine months prior to fiscal year end

to three months after fiscal year end. Equally weighted market-adjusted returns

are calculated as the difference between the firm’s annual return R

and the

mean return on an equally-weighted market portfolio. Value weighted market-

adjusted returns are calculated as the difference between the firm’s annual

return R

and the mean return on a value-weighted market portfolio;

= error term;

it = subscript for firm i in year t.

Table 4.17

Tests of Conditional Conservatism, Pre and Post Restatement,

Industry-Adjusted for the Period 1994-2004 for the Period

1994 – 2004

IndX

it-1

= β

PRE*POS + β

PRE*NEG + β

POST*POS + β

POST*NEG

+ β

PRE*POS*IndR

+ β

PRE*NEG*IndR

+ β

POST*POS*IndR

+ β

POST*NEG*IndR

+ ε

, (2)

Panel A: Industry-Adjusted Annual Returns (n = 460)

Adj. R

Coefficients 0.006 -0.012 -0.002 0.007 0.020 0.010 -0.005 0.046

3.93%

(t-value)

0.46 -1.19 -0.14 0.56*

1.63 0.75

-0.81 2.71*

* Significant at 0.01. ** Significant at 0.05. *** Significant at 0.10.

Conditional Conservatism Post-restatement, β

– β

= 0.051

Conditional Conservatism Pre-restatement, β

– β

= -0.010

Test H1: (β

– β

) > (β

– β

), F-statistic = 5.97 p=0.015

where

IndX

it-1

= firm i’s earnings per share (Compustat Data 58) divided by beginning of

fiscal year stock price minus the mean X

it-1

for all samples firms in the same

industry. Industry is defined using the 2-digit SIC code.

PRE = 1 if the year is before the restatement announcement year, and 0 otherwise;

POST = 1 if the year is after the restatement announcement year, and 0 otherwise;

POS = 1 if R

is >=0, and 0 otherwise;

NEG = 1 if R

is <0, and 0 otherwise;

IndR

= firm i’s 12-month return beginning nine months prior to fiscal year end to three

months after fiscal year end minus the mean return for all sample firms in the

same industry;

= error term;

it = subscript for firm i in year t.

CHAPTER 5

CONCLUSION

This chapter concludes the dissertation with a discussion of the results, contributions,

limitations and possible extensions.

5.1 Results

This study examines whether firms report more conservatively in the three years

following a restatement announcement, and whether an increase in conservatism is

related to the extent of debt and compensation contracts. I find restating firms

significantly increase conditional conservatism in the three year period following the

restatement announcement year. The results are unchanged using three alternative

returns measures, controlling for the passage of the Sarbanes-Oxley Act of 2002, and

controlling for industry effects. The results suggest that firms attempt to restore the

credibility of their financial reporting by taking a more conservative approach in

reporting firm performance following a restatement.

As expected, I find no change in unconditional conservatism from the pre- to post-

restatement period. The results appear consistent with the Ball and Shivakumar (2005)

view that unconditional conservatism is simply a downward accounting bias that

contracting parties can adjust for ex ante.

The results also suggest that the increase in conditional conservatism is related to the

importance of debt and compensation contracting to a firm. I find that both the level and

change in conditional conservatism following a restatement is greater for firms that have

a greater demand for external financing. Thus, debt contracting plays a role in

determining a level of conservatism. I find weaker evidence that both the level and

change in conditional conservatism following a restatement is greater for firms where

compensation contracting is more important. Specifically, I find very strong results using

market-adjusted returns, but insignificant results using raw returns. In addition,

sensitivity analysis shows the compensation contracting results to be less robust as

compared to the debt contracting results. Overall, these results appear to be consistent

with the contracting explanation of conservatism provided in Watts (2003).

I examine several corporate governance variables as possible explanations for an

increase in conditional conservatism following a restatement. Specifically, I examine the

number of outside directors and audit committee meetings, the number of independent

directors and financial experts on the audit committee, and a change in top management

and/or audit firm. I find that firms do take steps to improve corporate governance

following a restatement; however, I fail to find consistent support that the improved

corporate governance is associated with a change in conservatism.

Overall, the results suggest that firms do change their financial reporting following a

restatement by reporting more conservatively and this change is more likely for firms

where debt and compensation contracting are more important. I find no consistent

support for an improvement in corporate governance being associated with the increase in

conservatism. Thus, it appears the contracting explanation is the most likely reason for

the increase in conservatism. This suggests that firms report more conservatively

following a financial reporting failure in order to restore the trust and confidence of the

contracting parties and to protect the interests of the contracting parties.

5.2 Contributions

This study furthers our understanding of how firms respond to a financial reporting

failure and contributes to the restatement, corporate governance, and accounting

conservatism literature. This study contributes to the conservatism literature in three

ways. First, I extend the literature that distinguishes between the two types of

conservatism, conditional and unconditional, and make distinct predictions about

expected changes following a specific event, a reporting restatement. As expected, I find

only conditional conservatism changes following a restatement as firms increase

conditional conservatism. The results may provide an explanation to the findings in

Moore and Pfeiffer (2004). They find no change in conservatism following restatements;

however, their study did not use the Basu (1997) measure of conditional conservatism.

This study may help explain their lack of results as I find it is only conditional

conservatism that changes following a restatement. Second, this study provides evidence

regarding the contracting explanation for conservatism, as I examine the impact of debt

and compensation contracting on firms’ financial reporting choices. I find the increase in

conditional conservatism following a restatement is significantly associated with the

extent of debt and compensation contracting. This is consistent with the contracting

explanation of conservatism provided in Watts (2003) and provides insight by explaining

variation within firms after a restatement.

This study’s third contribution to the conservatism literature is to examine various

corporate governance variables as determinants of an increase in conditional

conservatism following a reporting failure. It is important for regulators, financial

analysts, investors and contracting parties to know what mechanisms in the financial

reporting environment lead to more conservative reporting. Does an improvement in

corporate governance following a reporting failure lead to more conservative reporting?

Does a new auditor require more conservative reporting following a restatement? Does

new management report more conservatively to “signal” to financial statement users and

contracting parties that the financial statements are reliable and trustworthy? This study

finds that firms do improve their corporate governance variables in the three years

following a restatement announcement. This is consistent with Farber (2005), who uses a

fraud sample and finds firms improve their corporate governance mechanisms in the three

years following fraud detection. However, this study finds no consistent relationship

between the changes in corporate governance following a restatement and the increase in

conservatism.

5.3 Limitations and Suggestions for Future Research

This study does have potential limitations. The primary focus of this study is to

examine changes in conditional and unconditional conservatism following a financial

reporting failure. Therefore, the results are based on how well my measures capture the

two types of conservatism. Givoly et al. (2007) find the Basu measure may not be

appropriate in all research setting because of some limitations of the measure. For

example, they find characteristics of the firm’s information environment that are

unrelated to conservatism may unduly affect the Basu measure. In addition, they find the

Basu measure is negatively correlated to other measures of conservatism, so they suggest

using multiple measures of conservatism in studies. I recognize in this study that the

Basu measure is capturing a unique dimension of conservatism apart from other

measures. As one of the purposes of this study is to examine two different dimensions of

conservatism (conditional and unconditional) following restatements, it is necessary to

use the Basu measure to capture the conditional dimension of conservatism. I leave the

use of alternative measures of conservatism to future research. In addition, the Basu

(1997) approach to measure conservatism has been widely used and accepted in the

literature; therefore, I believe this is an appropriate measure of conservatism for this

study. However, the limitations of the measure as pointed out in Givoly et al. (2007)

need to be considered when interpreting the results of this study.

A second limitation is that I do not test the interactions between the two types of

conservatism nor the specific predictions as modeled in Beaver and Ryan (2005). For

example, they point out that conditional conservatism resets the cost bases of net assets

when assets are written down, which then affects subsequent unconditional conservatism.

As this is an initial study on the change in conservatism following a reporting failure, I

leave these other issues to future research.

Another limitation is the external validity of the study as the results may not

generalize to the overall population of publicly traded companies. There are some

possible selection biases in my final sample of restatement firms. This is because of the

research design which requires each sample firm to have data for a consecutive 7-year

period, the three years prior and after the restatement announcement. Thus, the final

sample tends to include surviving and larger firms that maybe perceived as more reliable.

On the other hand, larger firms receive more media coverage and regulatory attention

than smaller firms and therefore may be under more pressure to change financial

reporting following a restatement in order to restore the public’s trust in their financial

reporting.

Givoly et al. (2007) find the use of aggregated data over longer time periods may

unduly affect the magnitude of the Basu measure. As this study uses annual data, future

research could use quarterly data to control for this “aggregation effect.” In addition,

future research can examine specific acts of conditional conservatism following a

restatement, such as asset impairment accounting and the lower of cost or market rule for

inventory. It would be interesting to determine if the way conditional conservatism is

manifested has an affect on a firm’s financial reporting choices following a financial

reporting failure.

Finally, future research can explore different aspects of compensation and debt

contracting to further refine our understanding of the contracting effects on financial

reporting choices following a restatement. For example, Beatty, Ramesh, and Weber

(2002) find a wide variation in the use of GAAP in debt covenants, with many requiring

the use of GAAP in existence at the contract date, and others allowing accounting

flexibility with the use of mandatory and/or voluntary accounting changes after the

contract date. It would be interesting to know if the inclusion or exclusion of mandatory

or voluntary accounting changes in debt covenants affects a firm’s conservatism

following a restatement.

REFERENCES

Abbott, L., S. Parker and G. Peters. 2004. Audit committee characteristics and

restatements.

Auditing: A Journal of Practice & Theory, v. 23(1): 69-87.

Ahmed, A., B. Billings, R. 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, v. 77(4): 867-890.

Anderson, K. and T. Yohn. 2002. The effect of 10-K restatements on firm value,

information asymmetries, and investors’ reliance on earnings. Working Paper,

Georgetown University.

Ball, R. and L. Shivakumar. 2005. Earnings quality in UK private firms: comparative

loss recognition timeliness.

Journal of Accounting and Economics, v. 39(1): 83-128.

Barton, J. 2005. Who cares about auditor reputation?

Contemporary Accounting

Research, v. 22(3): 549-586.

Basu, S. 1997. The conservatism principle and the asymmetric timeliness of earnings.

Journal of Accounting and Economics, v. 24(1): 3-37.

Basu, S. 2005. Discussion of “Conditional and Unconditional Conservatism: Concepts

and Modeling.”

Review of Accounting Studies, v. 10: 311-321.

Beasley, Mark. 1996. An empirical analysis of the relation between the board of director

composition and financial statement fraud.

The Accounting Review, v. 71(4): 443-465.

Beatty, Anne, K. Ramesh, and Joseph Weber. 2002. The importance of accounting

changes in debt contracts: the cost of flexibility in covenant calculations.

Journal of

Accounting and Economics, v. 33: 205-227.

Beaver, W. and S. Ryan. 2005. Conditional and unconditional conservatism: Concepts

and modeling.

Review of Accounting Studies, v. 10: 269-309.

Beekes, W., P. Pope and S. Young. 2004. The link between earnings timeliness,

earnings conservatism and board composition: evidence from the U.K.

Corporate

Governance, v. 12 (1): 47-59.

Bliss, J.H. 1924. Management through accounts. The Ronald Press Company, New

York, NY.

Bonner, S., Z. Palmrose and S. Young. 1998. Fraud type and auditor litigation: An

analysis of SEC Accounting and Auditing Enforcement Releases.

The Accounting

Review, v. 73(4): 503-532.

Cahan, S. and W. Zhang. 2006. After Enron: Auditor conservatism and ex-Andersen

clients.

The Accounting Review, v. 81(1): 49-82.

Dechow, P., R. Sloan and A. Sweeney. 1996. Causes and consequences of earnings

manipulations: An analysis of firms subject to enforcement actions by the SEC.

Contemporary Accounting Research, v. 13(1): 1-36.

DeFond, M. and J. Jiambalvo. 1991. Incidence and circumstances of accounting errors.

The Accounting Review, v. 66(3): 643-655.

DeFond, M. and Subramanyam. 1998. Auditor changes and discretionary accruals.

Journal of Accounting and Economics, v. 25: 35-67.

DeFond, M., R. Hann and X. Hu. 2005. Does the market value financial expertise on

audit committees of boards of directors?

Journal of Accounting Research, v. 43(2): 153-

193.

Desai, H., C. Hogan, and M. Wilkins. 2006. The Reputational Penalty for Aggressive

Accounting: Earnings Restatements and Management Turnover.

The Accounting Review,

v. 81(1): 83-112.

Easton, P. and J. Pae. 2004. Accounting conservatism and the relation between returns

and accounting data.

Review of Accounting Studies, v. 9: 495-521.

Farber, D. 2005. Restoring trust after fraud: does corporate governance matter?

The

Accounting Review, v. 80 (2): 539-561.

Francis, J., J. Hanna and L. Vincent. 1996. Causes and effects of discretionary asset

write-offs.

Journal of Accounting Research, v. 34: 117-134.

Francis, J. and J. Krishnan. 1999. Accounting accruals and auditor reporting

conservatism.

Contemporary Accounting Research, v. 16(1): 135-165.

General Accounting Office. 2002. Financial Statement Restatements: Trends, Market

Impacts, Regulatory Responses, and Remaining Challenges. GAO-03-138.

Givoly, D. and C. Hayn. 2000. The changing time-series properties of earnings, cash

flows and accruals: Has financial reporting become more conservative?

Journal of

Accounting and Economics, v. 29(3): 287-320.

Givoly, D., C. Hayn and A. Natarajan. 2007. Measuring Reporting Conservatism.

The

Accounting Review, v. 82 (1): 65-106.

Holthausen, R. and R. Watts. 2001. The relevance of the value-relevance literature for

financial accounting standard setting.

Journal of Accounting and Economics, v. 31: 3-75.

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

revisions and the estimated cost of capital.

Review of Accounting Studies, v. 9: 337-356.

Jensen, M. and W. Meckling. 1976. Theory of the firm: managerial behavior, agency

costs and ownership structure.

Journal of Financial Economics, v. 3: 305-360.

Kinney, W. and L. McDaniel. 1989. Characteristics of firms correcting previously

reported quarterly earnings.

Journal of Accounting and Economics, v. 11(1): 71-93.

Klein, A. 2002. Audit committee, board of director characteristics, and earnings

management.

Journal of Accounting and Economics, v. 33(3): 375-400.

Krishnan, G. 2005. Did earnings conservatism increase for former Andersen clients?

Working Paper, George Mason University.

Lee, T., R. Ingram, and T. Howard. 1999. The difference between earnings and

operating cash flow as an indicator of financial reporting fraud.

Contemporary

Accounting Research, v. 16 (4): 749-786.

Moore, E. and R. Pfeiffer Jr. 2004. The effect of financial statement restatements on

firms’ financial reporting strategies. Working Paper, Lehigh University.

Palmrose, Z., V. Richardson, and S. Scholz. 2004. Determinants of market reactions to

restatement announcements.

Journal of Accounting and Economics, v. 37(1): 59-89.

Palmrose, Z. and S. Scholz. 2004. The circumstances and legal consequences of non-

GAAP reporting: Evidence from restatements.

Contemporary Accounting Research, v.

21(1): 139-180.

Penman, S. and X. Zhang. 2002. Accounting conservatism, the quality of earnings, and

stock returns.

The Accounting Review, v. 77(2): 237-264.

Penman, S. 2003. The quality of financial statements: Perspectives from the recent stock

market bubble.

Accounting Horizons, v. 17: 77-96.

Pourciau, S. 1993. Earnings management and nonroutine executive changes.

Journal of

Accounting and Economics, v. 16: 317-336.

Riedl, E. 2004. An examination of long-lived asset impairments.

The Accounting

Review, v. 79(3): 823-852.

Sarbanes-Oxley Act. 2002. Public Law 107-204, 107

Congress.

http://www.sec.gov/about/laws/soa2002.pdf

100

Sengupta, P. 1998. Corporate disclosure quality and the cost of debt.

The Accounting

Review, v. 73(4): 459-474.

Srinivasan, S. 2005. Consequences of financial reporting failure for outside directors:

Evidence from accounting restatements and audit committee members.

Journal of

Accounting Research, v. 43(2): 291-334.

Watts, R. and J. Zimmerman. 1986. Positive Accounting Theory. Prentice Hall.

Watts, R. 2003a. Conservatism in accounting Part I: Explanations and implications.

Accounting Horizons, v. 17 (3): 207-221.

Watts, R. 2003b. Conservatism in accounting Part II: Evidence and research

opportunities.

Accounting Horizons, v. 17 (4): 287-301.

Zhang, J. 2004. Efficiency gains from accounting conservatism: benefits to lenders and

borrowers. Working Paper, Massachusetts Institute of Technology.

101

102

BIOGRAPHICAL SKETCH

William D. LaGore is an Assistant Professor of Accounting at Eastern Michigan

University. He is currently teaching the introductory financial accounting course at the

undergraduate level and the graduate-level advanced financial accounting course. In

addition, he has taught courses in intermediate accounting, managerial accounting,

governmental and nonprofit accounting, and financial statement analysis. His research

interests are in the area of financial accounting and reporting, including accounting

conservatism and the value-relevance of accounting information. He also has several

years experience in public accounting, including audits of governmental and nonprofit

entities.

William D. LaGore is a Registered CPA in the State of Michigan. He has a Master’s

Degree in Business Administration from the University of Wyoming and a Bachelor’s

Degree in Business Administration from Adrian College in Adrian, Michigan.