Masters Thesis of Accounting

Master’s thesis accounting, auditing and control
Equity incentives and earnings management
In partial fulfillment of the requirements for the degree of

Master of Science in Economics and Business
Erasmus University Rotterdam

Department: Business Economics
Section: Accounting, Auditing & Control
Course code: FEM 11032‐11
Supervisor: Dr. C.D. Knoops
Written by: Winfred Damler
Student nr: 295272
Date: July 2012
Abstract
This master’s thesis examines the relation between equity incentives and earnings
management. It extends prior research by providing a more detailed insight on the
relation between discretionary accruals and equity incentives. The study finds evidence
for a significant relation between discretionary accruals calculated by a linear Kothari
accrual model and equity incentives, in a pre‐Sarbanes Oxley sample. It shows that this
relation is stronger for CFO equity incentives than for CEO equity incentives. The study
finds a significant positive relation between earnings management and total equity
incentives; it also shows such a positive relation for option‐based equity incentives. For
stock‐based equity incentives no such positive relation is found. The third finding is that
the relation between earnings management and equity incentives changes before and
after the major accounting scandals and introduction of the Sarbanes Oxley act.
Abbreviations
CEO Chief executive officer
CFO Chief financial officer
GAAP Generally accepted accounting principles
IRS Internal revenue service
M&A Mergers and acquisitions
ROA Return on assets
R&D Research and Development
SEC Securities and Exchange Commission
SIC Standard industry classification
SOX Sarbanes Oxley act
US United States of America
2
Table of Contents
ABSTRACT...........................................................................................................................................................2
ABBREVIATIONS...............................................................................................................................................2
CHAPTER 1 INTRODUCTION ........................................................................................................................5
1.1 INTRODUCTION ........................................................................................................................................................... 5
1.2 PURPOSE OF THE THESIS AND RESEARCH QUESTION............................................................................................ 8
1.3 RELEVANCE AND CONTRIBUTION ............................................................................................................................ 9
1.4 STRUCTURE OF THE THESIS ....................................................................................................................................10
CHAPTER 2 EARNINGS MANAGEMENT, THE THEORY...................................................................... 11
2.1 INTRODUCTION AND THE REASON FOR EARNINGS MANAGEMENT ...................................................................11
2.2 WHAT DO WE CONSIDER EARNINGS MANAGEMENT? .........................................................................................13
2.3 MEASURING EARNINGS MANAGEMENT WITH ACCRUALS ...................................................................................17
2.4 WHO COMMITS EARNINGS MANAGEMENT? .........................................................................................................18
2.5 SUMMARY DEFINITION EARNINGS MANAGEMENT ..............................................................................................18
CHAPTER 3 ACCRUAL MODELS................................................................................................................. 20
3.1 ACCRUALS .................................................................................................................................................................. 20
3.2 THE HEALY MODEL 1985.......................................................................................................................................22
3.3 THE DE ANGELO MODEL 1986..............................................................................................................................23
3.4 JONES MODEL 1991.................................................................................................................................................23
3.5 MODIFIED JONES MODEL 1995 .............................................................................................................................26
3.6 TIME‐SERIES VERSUS CROSS SECTIONAL JONES MODELS ...................................................................................27
3.6.1 Time‐Series designs with the Jones model.................................................................................................28
3.6.2 Cross‐sectional designs with the Jones model..........................................................................................29
3.7 DIFFERENCE BETWEEN BALANCE SHEET ACCRUALS AND CASH FLOW ACCRUALS.........................................30
3.8 IMPROVED VERSIONS OF THE JONES MODEL ........................................................................................................32
3.9 THE FORWARD‐LOOKING MODEL 2003 ...............................................................................................................32
3.10 CASH FLOW JONES MODEL 2002 ........................................................................................................................34
3.11 LARCKER AND RICHARDSON 2004 ....................................................................................................................37
3.12 PERFORMANCE MATCHING MODEL 2005 .........................................................................................................38
3.13 THE BUSINESS MODEL 2007 ..............................................................................................................................41
3.14 RECENT LITERATURE ON ACCRUAL MODELS .....................................................................................................44
3.15 CHAPTER 3 SUMMARY...........................................................................................................................................45
CHAPTER 4 ESTIMATING THE EQUITY INCENTIVES......................................................................... 47
4.1 BOUNDARIES OF BONUS SCHEMES..........................................................................................................................47
4.2 MAXIMIZING EARNINGS IN JAPAN ..........................................................................................................................47
4.3 PROXY FOR EQUITY INCENTIVES ............................................................................................................................48
4.4 SUMMARY .................................................................................................................................................................. 49
CHAPTER 5 EMPIRICAL RESEARCH ON EARNINGS MANAGEMENT DUE TO EQUITY
INCENTIVES ..................................................................................................................................................... 50
5.1 INTRODUCTION ........................................................................................................................................................50
5.2 REMUNERATION .......................................................................................................................................................50
5.3 EQUITY INCENTIVES .................................................................................................................................................54
5.4 CEO AND CFO EQUITY INCENTIVES ......................................................................................................................59
5.5 SUMMARY .................................................................................................................................................................. 64
CHAPTER 6 HYPOTHESIS............................................................................................................................ 66
6.1 HYPOTHESIS 1 ..........................................................................................................................................................66
6.2 HYPOTHESIS 2 ..........................................................................................................................................................66
3
6.3 HYPOTHESIS 3 ..........................................................................................................................................................66
6.4 HYPOTHESIS 4 ..........................................................................................................................................................67
CHAPTER 7 RESEARCH DESIGN AND METHODOLOGY ..................................................................... 69
7.1 INTRODUCTION .........................................................................................................................................................69
7.2 ACCRUAL MODEL ......................................................................................................................................................69
7.3 MEASURE FOR EQUITY INCENTIVES.......................................................................................................................71
7.4 ESTIMATING THE RELATION BETWEEN EARNINGS MANAGEMENT AND EQUITY INCENTIVES .....................73
7.5 SAMPLE ......................................................................................................................................................................73
7.6 DESCRIPTIVE STATISTICS ........................................................................................................................................74
CHAPTER 8 FINDINGS .................................................................................................................................. 79
8.1 INTRODUCTION .........................................................................................................................................................79
8.2 HYPOTHESIS 1 ..........................................................................................................................................................82
8.3 HYPOTHESIS 2 ..........................................................................................................................................................83
8.4 HYPOTHESIS 3 ..........................................................................................................................................................85
8.5 HYPOTHESIS 4 ..........................................................................................................................................................86
8.6 SUMMARY .................................................................................................................................................................. 88
CHAPTER 9 LIMITATIONS .......................................................................................................................... 90
CHAPTER 10 CONCLUSION ......................................................................................................................... 93
SUMMARY AND MAIN CONCLUSIONS .............................................................................................................................93
RECOMMENDATIONS .......................................................................................................................................................93
BIBLIOGRAPHY .............................................................................................................................................. 95
APPENDIX 1 ..................................................................................................................................................... 97
APPENDIX 2 ..................................................................................................................................................... 99
APPENDIX 3 ...................................................................................................................................................100
APPENDIX 4 ...................................................................................................................................................101
4
Chapter 1 introduction
1.1 Introduction
Management compensation has been a much‐discussed item over the last decade.
Different accounting scandals, like Enron, Ahold and Parmalat have damaged trust in
executive managers and financial reports. Due to these scandals, there has been a lot of
discussion about remuneration of executives. Stock and option‐based compensation has
increased strongly during the 1980’s and the 1990’s (Bergstresser & Philippon, 2006).
Before that time managers had little or no incentive to maximize the firms performance.
Since that time the use of equity incentives has increased for a number of reasons. The
most obvious reason is to align the interests of the owners and the managers of
companies. Because interests of managers deviated from the interests of the owners of
firms, firms were not effectively managed from an owner’s point of view. An example of
this management behavior that is not line with owner’s interests is the fruitless “empire‐
building” as described in the study by Jensen (1991); too many mergers and takeovers
led to large firms, instead of enhancing performance this led to declining corporate
efficiency and destroying value.
Until the 1980’s not much performance enhancing incentives were provided to
management, this led to behavior from managers that was not in line with the interests
of stockholders. To provide management with an incentive to increase firm performance
companies started using more equity‐based incentives. Mehran’s (1995) study
demonstrates that providing performance enhancing incentives can work; his study
shows that firm performance is enhanced by providing management with stock or
option‐based compensation. Not only equity‐based incentives were introduced,
performance related bonuses where introduced as well. While the purpose of stock and
option‐based compensation plans was to align the interest of management with the
interests of the owners of the company, this also opened the door to opportunistic
behavior from management, as they could influence their remuneration by maximizing
the performance of the company. Healy (1985) is one of the first to provide proof that
managers use earnings management techniques to maximize their income.
5
Due to accounting scandals rewarding executives with equity incentives has become a
much‐discussed topic. For this discussion it is important to know what the effects of
equity incentives are, and how the relation between earnings management and equity
incentives works. This master’s thesis examines this relation for a sample of large firms
that are listed in the United States and are part of the S&P 500.
A first aspect this master’s thesis focuses on is the difference between CEO and CFO
equity incentives. Much of the prior research on this subject has focused on the relation
between the total equity incentives rewarded to the CEO and earnings management. But
there is more in it than just that. It is very well possible that the CFO has more influence
on accounting and accrual decisions than the CEO. As the CFO is the one responsible for
the financial administration of the firm and he is the one in charge of composing the
financial statements. Therefore it is useful to examine the relation between equity
incentives and earnings management for both the CEO and the CFO as it might be
possible that awarding equity incentives to the CFO, who is responsible for the financial
statements leads to more earnings management than equity incentives awarded to the
CEO, as the CEO cannot influence the financial statements as directly as the CFO can.
A second aspect this study examines is the effects of the different kinds of equity
incentives. Executives can be rewarded with different equity incentives, it is likely that
these different incentives have different effects on the behavior of the executives
because the characteristics of the equity incentives differ. There are more remuneration
incentives that can have an influence on management behavior like bonuses; this
master’s thesis will be limited to equity incentives. Equity incentives can be based on
stocks or derivates from stock, like options. This master’s thesis focuses on share‐ and
option‐based incentives. An important characteristic of options is that most options
have an expiration date, after this date the option has no value anymore. As a result
options are relatively short time incentives. Options motivate managers to increase
earnings until the expiration date of the options. Due to this, option incentives are by
definition incentives to increase short‐term firm performance. Stock‐based incentives
have no expiration date; a manager can benefit from both short‐ and long‐term firm
performance. As stocks do not have an expiration data they are a more permanent
incentive than options, the incentive only ends if the shares are sold. Another
characteristic of options is that executives can benefit from an increase in the stock price
6
due to earnings management, but that his wealth does not suffer much if the stock price
declines (Burns & Kedia, 2006). Therefore option‐based incentives can lead to managers
taking more risk and to use earnings management, as their wealth is not affected so
much if things go wrong. This is different for share‐based incentives.
A third aspect this study looks into is the change over time of the relation between
equity incentives and earnings management. Scandals like Enron and Ahold at the start
of the last decade have led to heaps of public attention on management compensation;
this might have led to companies changing their remuneration policies in order to keep
their reputation intact. Another reaction is that the scandals have led to legislation on
reporting details of management compensation. An example of such legislation is the
Sarbanes Oxley act in 2002. Some of the managers involved in accounting scandals have
been convicted, this in combination with new legislation and more public attention on
the subject may have led to a situation where managers are more careful to use earnings
management. They are more in the spotlight these days and are possibly more aware of
the consequences of earnings management. I examine the relation between earnings
management and equity incentives over a 10 years period. Starting in 1999, two years
before the major accounting scandals, until 2009. It is useful to examine if the relation
between earnings management and equity incentives changes over time, as it indicates
the effect changes following the accounting scandals have had. It is possible that
companies use different forms of remuneration nowadays, for instance more long‐term
incentives. This change in equity incentives is probably due to the accounting scandals
of the early 2000’s. In the 1980’s and 1990’s option‐based equity incentives were the
most important equity incentives, I expect however that the use of options as equity
incentives has declined and that share‐based equity incentives are more important
nowadays. This expectation is supported by the Global Equity incentives survey by PWC
(2011). This survey shows that performance‐based shares and share units are now more
used than stock options. It could also be the fact that managers do not want to use
earnings management too much anymore as they are afraid for the consequences. It is
useful to see if and how managers and firms reacted to the changed situation or that
there is not much difference between 1999 and 2009 despite all the changes in the
environment.
7
1.2 Purpose of the thesis and research question
This master’s thesis examines the relation between earnings management and equity
incentives. I intend to more precisely examine if this relation is different for incentives
awarded to the CEO and the CFO and if there are different effects for option‐ and stock‐
based equity incentives. I examine this relation over a ten‐year period (1999‐2009)
covering major accounting scandals, the years preceding these scandals and the
aftermath of those scandals. This leads to a more detailed insight on the effect of equity
incentives and provides information on the effect of measures taken in response to
accounting scandals on the relation between earnings management and equity
incentives.
My main research question is:
What is the relation between earnings management and equity incentives awarded to
CEO’s and CFO’s?
To analyze this relation further I examine the following sub questions:
‐ Is this relation different for incentives awarded to a CFO than for incentives
awarded to a CEO?
‐ Does this relation defer for stock or option‐based incentives?
‐ Do these relations change in the 10 years period from 1999 to 2009?
The goal of this master’s thesis is to provide more detailed insight in the relation
between earnings management and equity incentives. By answering these research
questions I provide insight in the difference in the relation between earnings
management and equity incentives for the CEO and the CFO. Much of the previous
research has focused on this relation for the CEO only, while this relation for the CFO
might even be stronger. One can imagine that the CFO has a big influence on accounting
decisions. As the CFO is responsible for the financial statements it might not be a good
idea that his personal wealth depends on the earnings of the company. Because the
financial administration is the responsibility of the CFO it could be that the CFO is the
manager who takes most of the accounting decisions. Therefore it could be the fact that
equity incentives for CFO’s have more influence on earnings management than equity
incentives for CEO’s. One could argue that it would be wise to have a CFO whose
8
personal wealth does not depend on firm performance. Especially in a situation where
the CEO’s remuneration does depend on the performance of the company this could be
important. The financially independent CFO can in such a situation prevent the CEO
from opportunistic behavior. This master’s thesis examines if there is a positive relation
between earnings management and equity incentives for the CFO and if this relation is
stronger or weaker than the relation of the CEO.
As mentioned in the previous section stock and option‐based equity incentives may have
a different effect than share‐based incentives. Because option‐based incentives are
expected to provide a short‐term incentive due to the expiration date of the options
while the incentive for stock‐based remuneration has a more long‐term effect as stocks
do not have such an expiration date.
The third sub question focuses on the change of this relation over time; it provides
information if the relations described above have changed over the years and if the
measures taken in the aftermath of accounting scandals had an effect on these relations.
For this I examine a sample of companies that are part of the S&P 500 as the needed
data is available for these companies in the “compustat” database. I use an accrual model
to measure earnings management and compare this accrual model with the dependence
of a manager’s income on the stock price. This master’s thesis contains a literature study
that covers prior research on measuring earnings management and equity incentives
and it contains an empirical research to answer the research question.
1.3 Relevance and contribution

This master’s thesis contributes to the field of research because it provides a more
specified insight in the relation between equity incentives and earnings management.
Where much of the prior research focused on the role of the CEO and at equity
incentives as a whole, this master’s thesis examines the role of the CEO and the CFO and
examines whether short‐term option‐based incentives have different effects on the
behavior of management than stock‐based incentives.
The second point why this master’s thesis is relevant is that it helps understanding the
relations between equity incentives and earnings management in more detail. This
makes it possible to provide managers in the future with adequate remuneration plans
that will maximize their productivity but do not create an incentive for opportunistic
9
behavior. It provides knowledge needed, not only to create better future remuneration
plans but also to provide information that is useful in the discussion around
management remuneration and creating legislation on management remuneration.
This master’s thesis also provides insight in the question if the relation between equity
incentives and earnings management has changed due to accounting scandals and the
measures taken in the aftermath of these scandals. It shows whether the scandals and
the measures taken after these scandals have changed the effect of equity incentives and
it will show if this is different for short‐term option‐based incentives and for the more
long‐term stock‐based incentives. It helps to analyze the effect of legislation and other
measures taken considering management remuneration.
1.4 Structure of the thesis

To examine the subject and to find an answer to the research question this master’s
thesis proceeds as follows: Chapter two discusses what earnings management entails
and why it can be triggered by equity incentives. Chapter three and four describe the
literature on measuring earnings management with accruals accounting and measuring
equity incentives respectively. Chapter five presents the hypotheses for the empirical
part of the master’s thesis, chapter six discusses the methodology and the research
design and the sample used. Chapter seven presents the results of the empirical
research and chapter eight discusses the limitations of the research. The last chapter,
chapter nine, presents the conclusions, a summary and recommendations for further
research.
10
Chapter 2 earnings management, the theory
2.1 introduction and the reason for earnings management

In this chapter I discuss: what earnings management is, why it exists, how it can be
measured and who uses earnings management. In this master’s thesis I look at earnings
management by board members. To understand the idea of earnings management it is
important to know why people take the effort to manage these earnings.
A well‐known theory on decision‐making is the utility maximizing theory. This theory

was designed in the 18th and 19th century by Jeremey Benthem (1789) and John Steward
Mill (1863). It says that society has as ultimate goal to maximize the utility of all
individual members of society. Individual members of society will maximize their own
utility; therefore a manager also looks to maximize his own utility. How the utility of a
manager is maximized will differ from person to person.
For a manager of a company who is trying to maximize his utility different factors might
be important, for instance: his social status, the fact that he wants to keep his job, his
remuneration and the amount of effort he has to put in his job. For these factors other
underlying factors might be important: For his social status it might be important the
company does well or that the press writes positive articles about the company. For his
remuneration it might be important the company is profitable or that the stock price
rises.
Another theory that comes into play is the ‘Agency Theory’ originally introduced by
Adam Smith (Smith, 1776). This thesis considers publicly held companies, in those
companies there is a possible difference in interest between the owners of the company
and the managers. In a publicly held company the owner, or owners, are the
stockholders. I assume that in a publicly held company it are the managers who take
most of the decisions. Because the managers and the owners are often different people
there can be a difference of interest between he manager (the agent) and the owner (the
principal). Both want to maximize their personal utility, but as their interests are not
always in line this can lead to difficulties. Because the utility maximizing manager does
not what the owners of the company, who hire the manager, want him to do. This is
called the principal agent dilemma (Jensen & Meckling, 1976). One of the solutions used
11
to mitigate this problem is to try to align the interest of the principal and the agent. A
common way to do this is to provide management with stock and option‐based
remuneration.1 Thereby a part of their remuneration depends on the performance of the
company on the stock markets. This brings the interests of the management more in line
with the interest of the owners of the company, who are also dependent on the
performance on the stock marked. The idea is that a manger whose interests are in line
with the interests of the owner of the company makes decisions that are beneficial from
the owners’ point of view. In this situation maximizing stock value or paying dividends
is now favorable for both managers and owners.
In this master’s thesis I look at managers who, as I assume, want to maximize their
remuneration. In line with Healy’s earnings maximizing hypothesis (Healy, The effect of
bonus schemes on accounting decisions, 1985). A manager will try to maximize his own
wealth despite possible negative effects for the company. We look at the case where the
remuneration depends for a certain amount on the stock price of the company. In this
master’s thesis I focus on equity‐based remuneration in the form of stocks and options.
There are however other ways to bring managers interest more in line with that of the
owner of the company, for instance bonuses that depend on the performance of the
company or on the relative performance of the company in a peer group.
A manager who wants to maximize his remuneration will, if the height of his
remuneration correlates strongly with the stock price, try to maximize the stock price.
As I assume the stock price depends on the performance of the company, as earnings are
an important indicator for the company’s performance the manager will try to maximize
the earnings, because this is in line with his interest2. He maximizes his utility by
maximizing the company’s stock price. Mehran (1995) finds that this actually works. He
finds that: “firm performance is positively related to the share of equity held by
managers, and the share of management compensation that is equity‐based”.
On the other hand if his remuneration does not depend so strongly on the company’s
stock price the manager might be driven by other incentives. He might maximize his
utility in another way and not spend so much effort on maximizing the stock price. He
1 See Hall and Liebman (1998), who find that the effect of the value of a firm on the wealth of the CEO
has tripled between 1980 and 1994.
2 See Ronen and Yaari (2008), chapter 1, for the question why earnings are important.
12
then might choose to spend more time relaxing, spending time with his family or reach
other targets that for instance increase his bonus or status. This does not mean that in
those cases he will not use earnings management. Remuneration is not the only
incentive that could lead to earnings management. Other well‐known examples are:
earnings management to keep within the limits of contracts, for example debt contracts.
A company might want to reach a certain level of performance to prevent it has to pay a
higher interest rate (Stolowy & Breton, 2004). Another reason for earnings management
can be that a company wants to maintain a stable dividend policy or just present a stable
performance over time, therefore they might use income smoothing (I explain income
smoothing later in this chapter). An example of this is provided in a study of Kasanen,
Kinunnen and Niskanen (1996); they provide evidence of earnings management in
Finland to keep dividend payment up with the expectations of their large institutional
shareholders. Stolowy and Breton (2004) also state that some managers manage the
earnings down to pay less tax or to obey certain regulations.
In this thesis I assume that a manager whose remuneration depends on the company’s
stock price wants to present earnings the best way possible. He might be able to do this
by working very hard to try to use the firms’ potential to a maximum, and therefore be
able to present a proper profit. However he can also (next to this) try to manage the
earnings so he can present them in the best (to his interests) possible way. This is called
earnings management. In section 2.2 I discuss the definition of earnings management.
2.2 What do we consider earnings management?

There is a vast amount of literature about what is considered earnings management. In
this section I discuss this literature and come to a definition of earnings management
that I use in this paper.
Earnings management has different names, some stand for special kinds of earnings
management; others contain all sorts of earnings management. Stolowy & Breton (2004)
present a framework to understand accounts manipulation. They use accounts
manipulation as the general term. Illegal accounts manipulation is called fraud, accounts
manipulation within boundaries of the law is divided into earnings management (in a
broad sense) and creative accounting. Earnings management in the broad sense exists of
income smoothing, big bath accounting and earnings management (narrow sense). Their
definition of accounts manipulation is:
13
“ The use of management’s discretion to make accounting choices or to design transactions
so as to affect the possibilities of wealth transfer between the company and society
(political costs), funds providers (cost of capital) or managers (compensation plans).”
To my opinion it is often difficult to determine when accounts manipulation is legal or

not. It is even more difficult to determine whether the managers’ intentions are
opportunistic or not. This is due to the discretion managers have and the flexibility in
accounting regulations. As accounting is no exact science there is no absolute truth,
management has a certain degree of freedom use accrual accounting and to design the
transactions they make. Stolowy and Breton (2004) describe that: “When accounts
manipulation is used, the financial position and the results of operations do not fall into
the fair presentation category of the figure below”. That does not directly mean that the
actions are illegal. According to Stolowy and Breton (2004): “To be legal, interpretations
may be in keeping with the spirit of the standard, or at the other extreme, clearly stretch
that spirit while remaining within the letter of the law. They may be erroneous, but
never fraudulent”.
Figure 1
Stolowy and Breton 2004
Thereby it is to my opinion important to know why someone took a certain decision

before you can say if something is done legally or illegally. There are many different
definitions of earnings management Ronen and Yaari (2008) divide a couple of these
definitions in three groups: white, gray and black. In the white group: earnings
management is taking advantage of the flexibility in choice of accounting treatment to
14
signal the manager’s private information on future cash flows. In the gray group:
earnings management is choosing an accounting treatment that is either opportunistic
(maximizing the utility of management only) or economically efficient, maximizing the
utility of the firm. In the black group: Earnings management is the practice of using
tricks to misrepresent or reduce transparency of the financial reports (Ronen & Yaari,
2008). This indicates there are many different views on earnings management. As I use
accrual accounting in this thesis it is good to look at a definition that uses accrual
accounting.
Dechow & Skinner (2000) explain earnings management from the perspective of accrual
accounting. Accrual accounting tries to relate expenses, income, revenues, gains and
losses to a certain period. This is done to provide better or more complete information
about a company’s performance. In order to do this, choices have to be made to allocate
certain cash flows to certain periods. Revenues and costs have to be matched and
choices about depreciation of investments have to be made. The good thing about
accrual accounting is that it provides better information about the company’s
performance. The reported earnings using accruals accounting will be smoother and; if
done well, will provide a more realistic view of a company’s performance than the
underlying cash flows (Dechow & Skinner, 2000). On the hind side the choices made with
accrual accounting influence the view, this makes the financial statements subjective.
People who make the financial statements have an influence on the outcome; it is often
difficult to say whether they are trying to provide a realistic view or that they have other
plans with the financial statements. This can be a dangerous side of accrual accounting.
This is the grey area I mentioned before in this section. It is almost impossible to see
whether managers who use accruals accounting make choices that help investors get a
realistic view of the performance of the company or that they make choices that are in
their own interest. Because there are a lot of accrual decisions to be made it is difficult to
monitor whether this is correctly done. As the choices are subjective there is no absolute
truth. Therefore there is a very vague and thin line. It depends on your definition of
earnings management from what point you call this earnings management.
Healy & Wahlen (1999) give a definition on earnings management in line with this. They
do not mention the fact whether earnings management is legal or not. They set the line
15
at the point where the accounting decisions are no longer made to give a realistic view
of the company’s performance:
“Earnings management occurs when managers use judgment in financial reporting and in
structuring transactions to alter financial reports to either mislead some stakeholders
about the underlying economic performance of the company, or to influence contractual
outcomes that depend on reported accounting numbers”
This is therefore in my opinion a good definition of earnings management. However

Ronen and Yaari (2008) who call this definition of earnings management the best
definition in the literature point out two weak points in this definition. The first one is
that this definition does not set a clear boundary between earnings management and
normal activities that have an influence on earnings. The second point is that earnings
management does not have to be misleading, certainly not all the earnings management.
An example of this is that investors would like to see persistent earnings separated form
one‐time shocks. Therefore firms manage earnings in order to allow investors to
distinguish between the two sorts of earnings (Ronen & Yaari, 2008).
Ronen and Yaari (2008) present a definition of earnings management that takes these
weaknesses into account. Their definition is:
“Earnings management is a collection of managerial decisions that result in not reporting

the true short‐term, value‐maximizing earnings as known to management.
Earnings management can be:
Beneficial: it signals long‐term value;
Pernicious: it conceals short‐ or long‐term value;
Neutral: it reveals the short‐term true performance.
The managed earnings result from taking production/investment actions before earnings
are realized, or making accounting choices that affect the earnings numbers and their
interpretation after the true earnings are realized.”
Although maybe more complete I consider the definition of Healy and Wahlen (1999)
more clear because it is more concise and therefore better to understand.
16
There are different forms of earnings management, sometimes with different names that
fall under the broader definition of earnings management. These are for example:
income smoothing, big bath accounting, creative accounting and earnings management
due to accrual accounting. For more information about these different kinds of earnings
management see amongst others: Stolowy and Breton (2004), Ronen and Yaari (2008)
and Healy (1985).
2.3 Measuring earnings management with accruals

There are different ways to indicate earnings management. In this thesis I focus on
earnings management indicated by accruals. Accruals are defined as the difference
between the reported net income and the cash flow of a company. Each company has
accruals; that is perfectly normal. How much accruals a company normally has depends
amongst other things on the size of the company. Examples of accruals that each
company has are accruals due to depreciations or normal income smoothing (following
accounting rules). A part of the accruals are subjective, like the valuation of assets for
example or they can be influenced by management. These accruals are called the
discretionary accruals. The discretionary accruals are the accruals that indicate earnings
management.
Accruals accounting is something that is normally used in everyday practice. Accruals

are therefore not always wrong or suspected. A manager uses accruals to transfer the
company’s cash flows into an annual profit or loss. Without accruals this would not be
possible as I explained before. Accruals can also be used for the more dark sides of
earnings management, for instance to make a company’s performance look better than it
is, this is what happened at Enron. A danger of accrual accounting is that it is vulnerable
for opportunistic behavior.
Healy (1985) started a discussion on measuring earnings management with accruals

and the effect of management incentives on earnings management.. After Healy’s article
much has been written about the subject. People have designed different models to
indicate earnings management with accruals and to calculate accruals the best way
possible. In the next chapter I take a closer look at some of these models. I discuss the
early Healy (1985) and d’Angelo (1986) models, The Jones (1991) and modified Jones
model (1995) and a number of models that refine and improve the Jones and modified
17
Jones models. As the forward‐looking model by Dechow et al. (2003), the Kothari et al.
(2005) performance model and the syntheses model by Ye (2007).
2.4 Who commits earnings management?

When looking at the relation between equity incentives and earnings management it is
important to realize who are the people that take the accrual decisions. Bergstresser and
Philippon (2006) find proof for a positive relation between CEO equity incentives and
earnings management.
Jiang, Petroni and Wang (2010) find that the equity incentives for de CFO are more
important than equity incentives given to a CEO. Because the CFO is the one responsible
for presenting the annual numbers in a reliable way you could argue that it would not be
a good idea that his personal wealth depends on the way he presents the accounting
report of the company. As Katz (2006) describes IRS commissioner Mark Everson
suggested in front of the Senate committee that CFO’s should be rewarded with a fixed
payment.
It is important when using equity incentives to know how decisions are made within a
company. Because with this knowledge incentives can be used in a more effective way,
whether these are equity‐based or not. It probably differs from company to company
how decisions are made. In companies with a very strong CEO the rest of the
management might not have so much influence. But as one might imagine there are
other companies that work more on basis of mutual consensus or where for instance;
the rest of the board does not bother about the financial part and leaves that to the CFO.
Taken this into account it is important not to focus solely on the CEO when looking at
earnings management. Because it is possible other members of the board can be
triggered by equity incentives as well.
2.5 Summary Definition earnings management

In the first chapter of this master’s thesis I discuss what earnings management is, why
managers use earnings management, and which people use earnings management. I also
discuss earnings management that is due to accrual accounting, as it is that form of
earnings management I use in my master’s thesis. It is important to understand that
accrual accounting is not per definition something that is bad. It is used in everyday
practice; to translate the cash flows into an annual profit or loss. A problem can be that
18
accrual accounting is sensitive for opportunistic behavior. As discussed managers strive
to maximize their own utility. By granting them equity incentives their utility becomes
dependent on the stock price. It then depends of the manager, how far he will go to
maximize his utility, if he is opportunistic he can use earnings management to generate
more income for himself. As equity incentives are rewarded to more people than the
CEO alone it is important to think about which people have influence on the accounting
numbers, to know how incentives can be rewarded in a more effective way. While at the
same time lowering the risk of opportunistic behavior.
19
Chapter 3 accrual models
This chapter discusses literature on how accruals are used to measure earnings
management. Measuring accruals has developed over time; in this chapter I discuss how
the methods to measure earnings management have developed from simple models
measuring total accruals to more complex models separating accruals in discretionary
and non‐discretionary accruals while taking into account characteristics of the firm and
its environment.
When using earnings management managers try to influence the accounting numbers of
a firm. They can do this by using real transaction‐based earnings management.
Examples of real transaction‐based earnings management are: “providing price
discounts or cutting discretionary expenses” (Bartov & Cohen, 2008). While doing that,
the profit will increase but it does not say much about the real performance of the
company. These methods are easy to detect for analysts and stakeholders. Another
method to influence the accounting numbers is using accrual accounting, this method is
more difficult to detect. Measuring earnings management by using accrual accounting is
discussed in this chapter.
3.1 Accruals
The earnings of a company contain cash flows and accruals.
Earnings = cash flow + accruals
Management can influence accrual accounting. Management has a certain degree of

discretion when making accruals decisions. This discretion can be used
opportunistically. Accrual accounting has to be used according to accounting regulations
as IFRS. Accrual accounting in itself is therefore not mischievous but it can be used in an
opportunistic way. The alternative for accrual accounting is cash flow accounting. Cash
flow accounting is not in line with the accounting rules. Managers can influence accrual
decisions to their own interest. An example of this is maximizing their bonus as
described in the thesis by Watts and Zimmerman (1986)
Examples of influencing the accounting report using accrual manipulation are for
instance:
20
Trade receivables: the account “trade receivables” is subjective, because management
has to estimate the amount of the receivables that will actually be paid and the amount
that is qualified as bad debt. Management can therefore manipulate the valuation of this
item, for instance by changing the bad debt policy.
Stock: Another highly subjective item on the balance sheet is stock. The valuation of the
trade stock can be influenced, managers can decide whether it is necessary to depreciate
the stock or not.
Current assets: Current assets can be used to move cost to a subsequent period, by
capitalizing a certain amount instead of taking the costs at once.
Fixed assets: fixed assets as real estate, machines and other equipment have to be
measured. This can be subjective. Besides that, certain costs related to the fixed assets
can be capitalized and depreciated at the discretion of management.
For example: A manager wants to manipulate the company’s profit in a certain year
because he wants to maximize the value of his equity incentives; the manager can decide
to change the bad debt policy. By changing the bad debt policy a manager can classify a
smaller or larger amount of the debt as bad debt. Thereby he is able to manage the
earnings of the firm upwards or downwards.
Mohanram (2003) defines accruals as the revenues and costs that make up the difference
between the reported profit as the cash flow of the company. Accounting profit can be
divided into three parts: the operational cash flow, the non‐discretionary accruals and
the discretionary accruals. Therefore:
Earnings = cash flow + normal accruals + discretionary accruals

Discretionary accruals = earnings – cash flow – normal accruals
The non‐discretionary accruals are accounting changes that are imposed by accounting
regulations. For instance booking expenses at the moment they are realized according to
accounting regulations but before the cash flow takes place. The discretionary accruals
are the accounting decisions the manager can influence. He can for example decide if he
wants to capitalize cost related to the fixed assets and decide how he depreciates these
capitalized costs. These accruals are therefore called discretionary; the discretionary
accruals are used to measure earnings management. As discretionary accruals are used
21
as measure for earnings management one has to separate these accruals from the total
earnings. There are different models designed that try to separate accruals or
discretionary accruals from total accounting profit. Some of the early models only
separate earnings in total accruals and the operating cash flow. Later models also
separate discretionary accruals from the non‐discretionary accruals. In the following
sections of this master’s thesis I discuss the different models used to separate accruals
from the total profit.
3.2 The Healy model 1985

Healy’s (1985) model is one of the first accrual models. Healy measures earnings
management while using accruals. He tries to find evidence for earnings management
around the top and bottom level of bonus schemes. He expects that managers, with
bonus schemes that depend on the company’s profit, influence the profit in a way that
maximizes the manager’s bonus.
Healy (1985) defines accruals as the difference between reported earnings and the
operational cash flow. He uses total accruals as indicator for discretionary accruals, as
he does not separate the total accruals in discretionary and non‐discretionary accruals.
He states it is not possible to identify the non‐discretionary accruals. He does separate
the total accruals into “normal” accruals and “abnormal” accruals. He uses the abnormal
accruals as proxy for discretionary accruals
Total accruals are estimated by the difference between reported accounting earnings
and cash flow from operations (Healy, The effect of bonus schemes on accounting
decisions, 1985):
TAi,t = ( CAi,t – CLi,t – Cashi,t + STDi,t – Depi,t) / Ai,t – 1
TAi,t Total accruals of firm i at time t

CAi,t The change in the current assets of firm i at time t
CLi, The change in current liabilities of firm i at time t
Cashi,t The change in cash holdings of firm i at time t
STDi,t The change in long term debt in current liabilities of firm i at time t
Depi,t Depreciation and amortization expense of the firm of firm i at time t
Ai,t – 1 Lagged size (in assets) of firm i at time t‐1
Healy (1985) estimates the “abnormal” accruals as the difference of the total accruals of
the current year and the “normal” accruals of that year. The “normal” accruals are the
average total accruals of the years prior to the current year scaled by total assets. You
22
could say that average accruals of the previous years are used as proxy for non‐
discretionary accruals (Dechow, Sloan, & Sweeney, 1995)
DAt = TAt – TAa
DAt Discretionary accruals of year t scaled by lagged total assets

TAt Total accruals of year t scaled by lagged total assets
TAa Average total accruals of the 10 years prior to year t scaled by lagged total assets
This means that if non‐discretionary accruals are constant over time and the
discretionary accruals have a mean of zero in the estimation period, then the model
measures nondiscretionary accruals without error. But if non‐discretionary accruals
change from year to year then the non‐discretionary accruals will not be measured
without error. The assumption that non‐discretionary accruals are constant is most
possibly not realistic, because non‐discretionary accruals change in response to changes
the economic circumstances and with firm characteristics (Dechow, Sloan, & Sweeney,
1995).
3.3 The De Angelo model 1986

The model by De Angelo (1986) can be considered as a special version of the Healy
(1985) model. De Angelo (1986) describes, like Healy, the “abnormal” accruals as the total
accruals minus the normal accruals. She uses the accruals of the preceding year as the
“normal” or “expected” accruals. These normal accruals could be seen as proxy for non‐
discretionary accruals and the abnormal accruals as proxy for discretionary accruals.
His formula for discretionary accruals is:
DAt = TAt – TAt‐1

TAt‐1 Total accruals of the year prior to year t scaled by lagged total assets
When using this model one assumes that accruals are constant over time and have a
mean of zero in the estimation period. Because of these assumptions the model does not
take into account changes in the performance and economic circumstances of the firm
(1995).
3.4 Jones model 1991

The Jones model (1991) is an important improvement on the previous models. The
improvement Jones (1991) makes it that she takes into account the effect of the
23
contemporaneous sales revenue and the fixed assets on the non‐discretionary accruals.
The Healy (1985) and De Angelo (1986) models ignore the influence of changes in sales
and the fixed assets on working capital accounts and thereby on accruals. If non‐
discretionary accruals depend for example on the revenues, than a change in accruals
can be caused by changes in non‐discretionary rather than discretionary accruals (1991).
Therefore the model to measure non‐discretionary accruals must correct for the
influence revenues have on the non‐discretionary accruals.
Using the De Angelo (1986) model one assumes that the difference between current and
prior‐year accruals is due to changes in discretionary accruals only. One assumes
thereby that non‐discretionary accruals are constant from period to period. Jones
controls for changes in revenue in her model, with this she eases the assumption that
non‐discretionary accruals are constant.
The Jones (Jones, 1991) model can be divided into three stages. She first calculates the
total accruals. With the total accruals she estimates the coefficients in the formula for
non‐discretionary accruals. With these coefficients the non‐discretionary accruals in the
event year can be calculated and with the non‐discretionary accruals we can find the
discretionary accruals. The discretionary accruals are used as proxy for earnings
management.
The first stage is to calculate the total accruals. As definition for total accruals Jones
(1991) uses the changes in the non‐cash working capital before income taxes payable
less total depreciation expense.
TAi,t = ( CAi,t – CLi,t – Cashi,t + DD1i,t – Depi,t) / Ai,t – 1
TAi,t Total accruals of firm i at time t scaled by lagged total assets

CLi,t The change in current liabilities of firm i at time t
DD1i, t The change in long term debt due in one year of firm i at time t
It is also possible to calculate total accruals using cash flow data as indicated by Hribar
and Collins (2002). I discuss this later in this chapter. When using this method only the
24
first step of the Jones model changes. The calculated total accruals are used in the next
two steps of the Jones (1991) model to find discretionary accruals.
The second stage is to estimate the coefficients in the equation for non‐discretionary
accruals using the total accruals calculated in stage one. Jones (1991) uses a regression
model to estimate the coefficients in the formula for non‐discretionary accruals. The
Jones (1991) model is an event model; it assumes that firms do not manage earnings in
the years before the event. The time‐series of the firms earnings can be separated in an
estimation period where discretionary accruals are zero and the event period (Ronen &
Yaari, 2008).
To estimate these coefficients total accruals are used as dependent variable in the
regression analysis. The coefficients of the formula can be estimated using a time series
model or a cross‐sectional model. I will further explain the difference between these two
and the advantages and disadvantages of both later in this chapter. In both versions the
coefficients are estimated on an estimation sample, this can be the years prior to the
event period (time‐series) or other companies in the industry (cross‐section).
The first part of the second stage is to estimate the coefficients in the formula using total
accruals as dependent variable.
TAi,t = 1× ( 1/ Ai,t‐1) + 2 × (∆REVi,t) + 3 × (PPEi,t) +ei,t
TAi,t Total accruals scaled by lagged total assets of company i in year t

Ai,t‐1 Lagged total assets of company i
∆REVi,t The change in revenue scaled by lagged total assets of company i in year t
PPEi,t The gross value of property, plant, and equipment in year t for firm i
ei,t Residual of the model
In this equation the change in revenues and gross property plant and equipment are
included in the model. Jones (1991) ads these variables to control for changes in non‐
discretionary accruals caused by changing conditions in the environment of the
company. The equation is estimated with an OLS‐regression. When using a time‐series
approach coefficients are estimated on basis of a time‐series prior the year in which one
wants to measure earnings management. For this estimation data is needed of the years
preceding the event year. One needs approximately 10 years prior to the event year.
Though normally the equation is estimated on the longest time series of observations
25
available prior to year t‐1 for each firm. Jones states that using the longest time series of
observations improves estimation efficiency but the downside of a long estimation
period is that the likelihood of structural changes in the period increases. Structural
changes in the company can contaminate the model because the accruals before and
around the structural change are no good measure for the accruals in the event year.
Jones (1991) uses the change in revenue as a control variable because total accruals
contain changes in working capital accounts. For example the change in accounts
receivable, the change in inventory and the change in accounts payable. These items
depend to a certain degree on the changes in revenue. By adding the change in revenue
as an independent variable Jones (1991) controls for this effect, assuming that the
change in revenue is an objective measure of the firms’ operations before earnings
management takes place. This assumption might not be completely justified; therefore
this is later changed in the modified version of the Jones model, discussed in the next
part of this chapter.
Gross property, plant and equipment is the other independent variable Jones (1991) uses
in her regression model. Gross property plant and equipment is included as independent
variable to control for the part of total accruals that is due to regular (non‐discretionary)
depreciation expenses. Jones (1991) uses Gross property plant and equipment and not
the change in gross property plant and equipment because the total annual depreciation
expense is part the total accruals model. All variables in the formula are scaled by lagged
assets, this is done to mitigate the effect of heteroscedasticity.
The third stage is to derive the discretionary accruals. As total accruals contain
discretionary accruals and non‐discretionary accruals one can easily derive the
discretionary accruals, as non‐discretionary accruals are known.
DAi,t= TAi,t – NDAi,t
DAi,t Discretionary accruals scaled by lagged total assets

TAi,t Total accruals scaled by lagged total assets
NDAi,t Non‐discretionary accruals scaled by lagged total assets
3.5 Modified Jones model 1995

A problem with the Jones (1991) model is that the sales of the company are used to
control for the firm’s economic circumstances. Credit sales however can be subject to
earnings management themselves, because credit sales can be managed by moving them
26
from period to period or by changing the bad debt policy. Dechow et al. (1995) present a
model that solves this problem. They use cash sales instead of total sales; they obtain
cash sales by deducting the change in accounts receivable from the change in revenue.
When using time‐series version of the modified Jones model the total accruals (stage
one of the Jones model) are calculated with the normal Jones model and the coefficients
are estimated with the normal Jones model as well (stage two of the normal Jones
model). In the event period the non‐discretionary accruals are then calculated with the
modified version of the Jones model. When using the modified Jones model in a cross
sectional research design the modified Jones model is used for the estimation of normal
accruals and for calculating non‐discretionary accruals. This is the formula of the
modified Jones model:
NDAi,t = 0+ 1× ( 1/ Ai,t‐1) + 2 × (∆REVi,t ‐ ∆AR i,t)+ 3 × PPEi,t

NDAi,t Non‐discretionary accruals scaled by lagged total assets of company i in year t
Ai,t‐1 Lagged total assets of company i
∆REVi,t The change in revenue scaled by lagged total assets of company i in year t
∆ARi,t The change in accounts receivable scaled by lagged total assets of company i in year t.
PPEi,t The gross value of property, plant, and equipment scaled by lagged assets in year t for firm i.
When using the modified version of the Jones model one assumes that changes in
revenue les changes in accounts receivable are free from earnings management. One
also assumes that changes in accounts receivable are abnormal, because changes in
credit‐sales are seen as discretionary.
Although the Jones and the modified Jones models have their shortcomings they have
been important in this field of research. They have been used frequently and different
people have improved and extended the models. The next two sections discuss using the
models in different ways: time series versus cross sectional and the use of balance sheet
accruals versus cash flow accruals, after these sections I discuss improved accrual
models most of them based on the Jones model.
3.6 Time‐series versus cross sectional Jones models

There are two possible research designs in which the different versions of the Jones
models can be used. The first one uses the Jones model in a time‐series design. The other
option is to use the Jones model in a cross sectional research design. When using a time‐
series research design one uses the data of a company from the years prior to the year
27
where one wants to measure earnings management to estimate the coefficients in the
formula for non‐discretionary accruals. The period in which one wants to measure
earnings management is the event period. The years before that are called the
estimation period. In a cross‐sectional research design one estimates the normal
accruals on a sample of companies from the same industry. In the cross‐sectional design
data from the same year but from other companies in the industry is used to estimate
the normal accruals.
3.6.1 Time‐Series designs with the Jones model

When using the Jones model in a time‐series version we distinguish two periods. The
event year, that is the year in which one wants to measure earnings management and
the estimation period, these are the years, preferably 10 or more, prior to the event year.
The data from these years is used to estimate the coefficients of the formula for non‐
discretionary accruals. If one uses the model in this way one makes the implicit
assumption that there is no earnings management in the estimation period. “Earnings
management in the estimation period contaminates the test (Ronen & Yaari, 2008)”,
because that level of earnings management will be seen as normal and will therefore be
no part of the discretionary accruals. Another implication of the time‐series model is
that is requires a long series of observations. Therefore it is likely that a firm adapts its
business‐and accruals policies in that time. Changing business and accrual policies will
have an effect on the accruals; therefore the measurement will be influenced.
Using a time series model may also create a selection bias, because firms have to survive
at least 10 years to be able to carry out a time series research. The bias arises because
such firms are more likely to be bigger more mature firms. These firms have other
priorities than smaller younger firms. For example: Established firms may have a
carefully build reputation which they don’t want to risk by using earnings management
(Ronen & Yaari, 2008).
The most important argument in favor of using a time‐series model is that it uses data of
the company where we want to test earnings management to estimate the coefficients in
the formula for non‐discretionary accruals. The advantage of using data from the same
company is that it is much more likely that the data is better comparable with the data of
the event year were we want to measure earnings management because this
information is firm‐specific. Because a company stays, unless major changes, the same
28
company from year to year we can predict the normal or expected accruals more
accurately using this method. While using information from other companies will always
have the problem that not two companies are the same.
3.6.2 Cross‐sectional designs with the Jones model

When used in a cross‐sectional design the non‐discretionary accruals are estimated on a
sample of firms in the same industry, instead of data of the same firm from prior years.
This implies that the coefficients obtained are now industry‐specific instead of firm
specific.
With the coefficients estimated one is than able to obtain the non‐discretionary accruals
of the company of which we want to measure earnings management. The sic codes are
often used for grouping the firms per industry; to obtain a larger sample the two digit sic
code is often used.
A cross‐sectional research design has disadvantages as well. It is questionable whether

the benchmark used: the other companies in the industry, is an appropriate benchmark.
The cross‐sectional design assumes that there is homogeneity within the industry; for
instance that companies have the same operating technology which gives the same level
of normal accruals for a certain level of performance. It could be the fact that the other
companies deviate too much from the company tested. Therefore those companies have
different normal accruals. This contaminates the estimation of the coefficients in the
accrual formula. The grouping of companies per industry is important; hence how
industries are defined, because aggregating companies that have little in common will
negatively affect the result as in that case the bench mark where the coefficients are
estimated on is not representative (Ronen & Yaari, 2008).
Another possible problem with a cross‐sectional design is that the observations used for
estimating the coefficients may include managed earnings. If other companies in the
industry manage earnings as well this will be the fact, normal accruals then contain a
part of the discretionary accruals. The Jones model will in such a case only recognize
earnings management if the earnings management is relatively high compared to the
earnings management used in the industry. In times of economic prosperity companies
may decide to smooth the earnings, therefore the “normal” accrual in this industry will
29
be negative. Only if the accruals of the tested company are more negative than those in
the industry earnings management is recognized.
A weak point in the sample selection is that normally only industries with more than 8
or 10 observations are used; therefore some industries are excluded from the test.
Whether to use a cross‐sectional or a time‐series design depends on the assumptions

one wants to make. It depends on which assumptions are the most realistic for the
particular research design. If a cross sectional sample contains a large number of firms
that are similar to the firm where we want to measure earnings management in the
sense that they are for example all mature firms and have a comparable operating cycle,
the cross sectional approach will suit. If we want to measure earnings management in a
relatively new company there will be no data for the time series approach, than the
cross sectional method is the solution. Cross sectional designs sometimes considered as
the better designs as these models have greater power due to lager samples (Ronen &
Yaari, 2008).
3.7 Difference between balance sheet accruals and cash flow accruals
Earnings consist of a cash component and an accrual component. The first step of the
Jones (1991) model is to calculate the total accruals, hence the accrual component of
earnings. There are two main approaches to determine the accrual component of
earnings: the balance sheet approach and the cash flow approach. The balance sheet
approach calculates accruals using information from the balance sheet. These accruals
are derived from the change in non‐cash working capital. The other approach is to
calculate total accruals using cash flow information.
Hribar and Collins (2002) show that when using time series models in combination with
balance sheet information problems can occur around non‐articulation dates. The
balance sheet approach is an indirect approach to calculate accruals, when using this
approach one assumes that there is articulation between the accrual component of
revenues and expenses in the income statement and changes in balance sheet working
capital items. This assumption however does not hold for non‐operating events. These
are for instance: mergers, acquisitions, reclassifications, accounting changes,
divestitures and foreign currency translations (Hribar & Collins, 2002).
The following formula is used to calculate balance sheet accruals:
30
TAi,t = ( CAi,t – CLi,t – Cashi,t + STDi,t – Depi,t) / Ai,t – 1
TAi,t Total accruals of firm i at time t lagged by total assets

CLi,t The change in current liabilities of firm I at time t
STDi,t The change in long term debt in current liabilities of firm i at time t
Changes in working capital due to a non‐operating event are visible in the balance sheet
but do not flow through the income statement. Therefore, some of the changes in
balance sheet working capital accounts relate to the non‐operating events. These could
falsely be shown as accruals when using balance sheet approach (Hribar & Collins, 2002).
In the case that mergers and acquisitions (from here on M&A) increase working‐capital
accruals, the normal accruals are estimated higher than they should. Which leads to a
positive bias in the normal accruals and thereby to a negative bias in discretionary
accruals. Divestitures have the opposite effect on the accruals (Ronen & Yaari, 2008).
Foreign currency translations have no effect on the earnings reported in the income
statement as those are only recognized in the comprehensive income on the balance
sheet. The bias in balance sheet accruals caused by foreign currency translations
depends on whether the main currency of the company strengthens or weakens.
To mitigate the bias around the non‐articulation events as M&A and divestitures one can
use a measure for total accruals that is based on cash flow data instead of balance sheet
information. Using cash‐flow accruals solves the problem around non‐articulation
events because the changes in investment activities around non‐articulation events do
not flow through the cash‐flow statement.
The total accruals are calculated as the difference between earnings before
extraordinary items and discontinued operations – operating cash flows from
continuing operations scaled by total assets.
TAi,t = (EBXIi,t – CFOi,t)/ Ai,t‐1
TAi,t Total accruals of firm i at time t scaled by lagged total assets.

EBXIi,t Earnings before extraordinary items and discontinued operations
31
CFOi,t Operating cash flows from continuing operations
Ai,t ‐1 Total assets
Hribrar and Collins (2002) conclude that it is prudent for researchers to rely on accrual
measures taken directly from the cash flow statements, because these do not
contaminate the test around non‐articulation events. A possible problem with cash flow
accruals in certain research designs can be that cash flow data is only available from
1987 onwards.
3.8 Improved versions of the Jones model

The Jones (1991) and the modified Jones model attempt to separate total accruals into
non‐discretionary (normal) and discretionary (abnormal) accruals. The Jones (1991)
model is criticized for not correctly separating the accruals in non‐discretionary and
discretionary accruals. This is due to the fact that the model for non‐discretionary
accruals in incomplete (Bernard & Skinner, 1996). It is very difficult or maybe impossible
to exactly determine the normal accruals, but since the Jones (1991) model more
advanced models have been designed which improve the estimation of the non‐
discretionary accruals. In this part I discuss different models that are improved versions
of the Jones (1991) model. They mitigate some of the weak points of the Jones (1991)
model and therefore generate a more reliable measure of discretionary accruals than the
Jones and modified Jones (1991) models.
3.9 The forward‐looking model 2003

Dechow, Richardson and Tuna (2003) improve the cross‐sectional version of the
modified Jones model. They present three improvements on the Jones (1991) model that
enhance the reliability of the model. De model designed by Dechow et al. (2003) is called
the forward‐looking model.
The first improvement they introduce is that they make an adjustment for the expected
increase in credit sales. The modified Jones (1991) model treats all credit sales as
discretionary; this might not be entirely just. It causes a correlation between
discretionary accruals and sales growth. Dechow et al. (2003) do a regression to
estimate a sales coefficient that the expected change in accounts receivable for a certain
change in sales. The model treats this expected change in accounts receivable as non‐
32
discretionary. Where the modified Jones (1991) model would treat al credit sales as
discretionary.
Dechow et al. (2003) use the following regression to estimate the sales coefficient:
∆AR = + K × ∆sales + e
∆ AR The change in accounts receivable
K Coefficient for the expected change in accounts receivables for a given change in sales
∆ sales The change in sales
K measures the sensitivity of the change in non‐discretionary accounts receivable to

sales. If K is 1, 100% of the change in accounts receivable is non‐discretionary if K is 0
100% of the change in accounts receivable is discretionary. Adding K to the Jones (1991)
model adds non‐discretionary accounts receivable in sales to the cash sales (both
discretionary). The slope coefficient is added to the cross‐sectional modified Jones
model (Dechow, Sloan, & Sweeney, 1995).
The second improvement Dechow et al. (2003) apply in their model is that they control
for so called ‘reversals’. Reversals are the consequences of accounting decisions made in
the previous period. Accruals by definition reverse trough time, they are less persistent
than cash flows. To control for this effect the lagged value of total accruals is added to
the model.
Accruals are used to turn the continuous cash flow of the company in an annual income
statement. Therefore accruals are per definition designed to smooth the reporting of the
financial performance of the company. For example a firm that grows and therefore
anticipates a future sales grow will increase its inventory. In this case the increase in
inventory is not due to earnings management. But the Jones model would classify this as
discretionary accruals. To control for this mistake Dechow et al. (2003) include a
measure of future sales growth.
The change in sales from the current year to the subsequent year scaled by current sales
is used as measure for the future sales growth. Therefore this item is only available if the
financial statement of the year following the event year is available. If this is not the case
the lagged value of this measure can be used, as Dechow et al. (2003) do.
GR_Salest = (salest+1 – salest) / salest
GR_Sales Measure for the future sales growth
sales i,t+1 Sales of the year after the event year of company t
sales i,t Sales in the event year of company t
33
With these modifications Dechow et al. (2003) compose the following model:
TAi,t= 1 + 2((1+K) ∆salesi,t ‐ ∆ARi,t) + 3PPEt + 4 (TAi,t,‐1/ Ai,t‐2) + 5 GR_salesi,t
TA Total accruals in the current year of firm i, scaled by year t‐1 total assets.
∆sales Change in sales, scaled by year t‐1 property plant and equipment
∆AR The change in accounts receivable, scaled by year t‐1 total assets.
K The slope coefficient from a regression of ∆AR on ∆sales
PPE Gross property plant and equipment
TAi,t,‐1/ Ai,t‐2 Total accruals from the year before the current scaled by total assets of two years ago for
firm i.
GR_sales The change in sales for firm i, from year t to t+1, scaled by year t sales.
3.10 Cash flow Jones model 2002

Dechow and Dichev (2002) focus on the quality of earnings and accruals. They state that
the quality of measured accruals depends on the mistake in predicting cash flows. This is
due to the fact that accruals separate the timing of cash flows and their recognition in
the financial statements. If an expense is booked before the cash flow takes place an
accrual is created. This accrual is the predicted future cash flow.
Until the Dechow and Dichev (2002) thesis most models did not control for cash flow in
their accruals model. This had to do with the possible simultaneity problem. As cash
flows are by definition the difference between earnings before extraordinary items and
accruals. This causes the simultaneity problem especially if the cash flow method is used
to calculate the total accruals (Ronen & Yaari, 2008).
They define earnings as cash flows and accruals:
Earnings= cash flow + accruals

From an accounting point of view there are two important events for each cash flow.
The first is the receipt or payment of the amount. The second is the recognition of this
amount as profit or loss.
Cash flows for a certain period “t” can be divided into three groups:
CFt‐1t= cash receipts or payments of amounts accrued at t‐1
CF t,t = current cash flows
CF t+1,t= cash flows postponed to the next period
34
As you can see there is a sub script and a super script behind the variable, the sub script
is the period in which the amounted is booked as earning or expense. The super script
refers to the period in which the amount is received or paid.
The total cash flow for a period “t” consists of:
CFt = CFt‐1t + CFtt + CF t+1t
Dechow and Dichev (2002) assume that working capital accruals are solved within the
next period. For example if in period “t” a sale on credit is done and the earnings are
realized, than in the next period the payment is done and the accrual is solved.
Therefore the change in working capital in a certain year is influenced by the cash flow
of the prior year, the current cash flow and the cash flow of the next year.
In case the cash flow follows after the revenue or expense is recognized managers have
to estimate what cash flow will follow in the future. The real cash flow might defer from
the cash flow that was anticipated. If the realized cash flows defer from the accruals
estimations than the opening accruals (the accruals made when the future cash flow was
estimated) contain an error that is corrected by the closing accrual (the accrual made
when the final cash flow takes place).
Dechow and Dichev (2002) conclude about accruals:
‐ That accruals are temporary adjustments that delay or anticipate the recognition
of realized cash flows plus an estimation error term.
‐ Accruals are negatively related to current cash flows and positively related to
past and future cash flows.
‐ The error term measures to what extent the cash flow realizations matches the
accruals; it can be used of a measure of accrual and earnings quality.
Dechow and Dichev (2002) express accruals with the following equation:
TAt= CFt‐1t – (CFtt+1 + CFtt‐1) + CFt+1t + εt+1t – εtt+1

TAt Current total accruals recognized in period t
CFt s Cash from operations realized in period t and recognized in period s
Εt s Error term, for accruals recognized in period s and cash flows realized in period t
35
The error term is this regression represents accruals that are not related to the
realizations of predicted cash flows. This error term is used as measure for the quality of
earnings and accruals.
The following formula is made by, Dechow and Dichev (2002) to find a practical measure
for working capital accruals. They use the equation in a time series test.
∆WCt= 0 + 1 CFOt‐1 + 2 CFOt + 3 CFOt+1+ t
∆ WC The change in working capital3

CFO Cash flow from operations
t An error term, used to measure the quality of earnings
The first of these formulas (the theoretical) uses components of the past, present and
future cash flows. Because it is not possible to identify these components of cash flows,
the second (more practical) formula is made that uses cash flow from operations. Since
cash flow from operations includes cash flows realized in a certain period that are
recognized in different periods this leads to an estimation error4. The standard deviation
of the residual is used as measure for earnings quality.
This error term does not only contain earnings management but can also reflect firms
that report honestly but face uncertain economic environments and are therefore not
able to properly predict future cash flows or firms whose managers are not good at
estimating future cash flows (McNichols, 2002).
McNichols (2002) discusses the Dechow and Dichev (2002) thesis in her work. She
recognizes a couple drawbacks of the model. Dechow and Dichev (2002) assume that
accruals are solved within the next period; therefore the model only works for short‐
term accruals. That means that their measure is not suited to use for companies that
work with a production cycle that takes more time than one accounting period.
A second drawback of the model is that the two estimation errors are assumed to be
independent of each other and of the cash flow realizations. However estimation errors
in the case of discretionary accruals are likely to depend on each other and on the cash
3 The change in working capital, measured as the sum of the change in accounts receivable + the change
in inventory minus the change in accounts payable minus the change in tax payable plus the change in
other net assets net of liabilities. All these variables are scaled by average total assets (Ronen and Yaari
2008).
4 See Dechow and Dichev (2002), the coefficients are likely to be biased towards 0 and the R2 is
lower.
36
flow realizations. Therefore the model might not apply in the context where
management uses earnings management (McNichols, 2002).
McNichols (2002) links the work of Dechow and Dichev (2002) with the literature on
discretionary accruals and especially the Jones model. Where the Jones model focuses on
separating total accruals into discretionary‐ and non‐discretionary accruals the Dechow
and Dichev (2002) thesis focuses on total accruals. A weakness of the Jones model is that
discretionary accruals estimated with the Jones model are not pure discretionary
accruals, hence they also contain in part accruals that are due to other things than
earnings management. McNichols (2002) points out two examples of factors that
contaminate the discretionary accruals in the Jones model. The first is that the Jones
model assumes that accruals react to the current change in sales, but does not take into
account future and lagged changes in sales. The second factor that McNichols (2002)
points out in her study is that the Jones model ignores the expected future sales growth.
McNichols (2002) finds, as Dechow and Dichev (2002) do that: prior, current, and
subsequent years cash flow from operations is correlated with the Jones model.
Ye (2007) points at another weakness in the model by Dechow and Dichev (2002). The
model shows the relation between accruals and cash flows. A problem with the measure
proposed is that it excludes earnings management committed by shifting income from
period to period, as that shifting satisfies the measure proposed in the thesis. However
this can be an important earnings management tool.
3.11 Larcker and Richardson 2004

Larcker and Richardson (2004) improve the measurement on non‐discretionary accruals
by adding the book‐to‐market ratio and current operating cash flows to the model.
The book‐to‐market ratio is included to control for the expected growth in a company’s
operations. This variable is added because growing firms show high levels of accruals.
Growing companies are expected to invest more in inventory and other assets.
Therefore an increase in inventory for a growing company is probably not due to
opportunistic behavior. To control for wrongly qualifying these accruals as discretionary
the book‐to‐market ratio is added to the model. The book‐to‐market ratio is calculated
as the book value of common equity divided by the market value of common equity.
37
Dechow et al. (1995) conclude that firms with extreme performance are more likely to
have miss‐specified discretionary accruals. To control for this miss specification the
current operating cash flow is used as extra independent variable in the model. The
model Larcker and Richardson (2004) use is the following:
TAi,t = 0 + 1× ( ∆REVi,t− ∆ARi,t ) + 2 × PPEi,t + 3 × BMi,t + 4× CFOi,t + εi,t
TAi,t Total accruals of company i at time t scaled by average total assets

∆REVi, The change in revenue of company i at time t scaled by average total assets
∆ARi,t The change in accounts receivable of company i at time t scaled by average total assets
PPEi,t The gross value of property, plant and equipment of company i at time t scaled by
average total assets
BMi,t The book to market ratio company i at time t scaled by average total assets
CFOi,t The current operating cash flow company i at time t scaled by average total assets
ε Residual
Larcker and Richardson (2004) use the cash flow approach to calculate the total accruals.
They state that it is better to use the extra control variable in the accrual formula than in
the final regression analysis were accruals are linked with another measure, depending
on the research. This is better because adding he control variables in the cross‐sectional
accrual model makes it possible to identify industry year specific coefficients for these
variables.
3.12 Performance matching model 2005

Kothari, Leone and Wasley (2005) present another much used improvement on the Jones
model. Their model controls for performance related accruals. The accruals this model
controls for are performance related accruals that are classified as discretionary under
the modified Jones Model. These accruals are however not due to earnings management
but are a result of the performance of the company. As in this master’s thesis I want to
measure earnings management due to equity incentives it is extra important to control
for firm performance, because of the correlation between firm performance and
compensation (Ronen & Yaari, 2008).
Accruals are correlated by performance, in other words there is a relation between the
firms current and past performance and the accruals. Kothari et al. (2005) describe this
relation in their thesis; if forecasted sales changes are not zero or when other
parameters affecting accruals change than the forecasted accruals are non‐zero. When
sales changes are predictable, earnings changes will also be predictable and expected
38
accruals will therefore be non‐zero. The predictability in future performance leads to
predictable future accruals. The accruals model needs to filter these performance‐
related component out of the discretionary accruals. Otherwise this leads to false
discretionary accruals.
Kothari et al. (2005) present two ways to do this. The first one is adding return on assets
to the modified version of the Jones model to control for the performance of the
company. Thereby improving the linear accrual model. The second method is to
calculate discretionary accruals by comparing the discretionary accruals two
comparable companies, where one of the two is expected to manage earnings and the
other not. In this method Kothari et al. (2005) abandon the path of the linear regression
model to calculate discretionary accruals.
The essence of the second method is that there are two groups of firms, two samples:
The measurement sample, this is the sample were one wants to measure earnings
management and the control sample in which no earnings management takes place.
Firms from the measurement sample have to be matched with firms from the control
sample that have a similar return on assets. This is the control for performance, the
firms are similar and have similar performance therefore similar accruals are expected.
The Jones model is used to calculate discretionary accruals for both sets of firms, the
accruals in the control sample al seen as the normal accruals. The abnormal accruals are
than calculated as the difference between the discretionary accruals of the firm in the
measurement sample and the discretionary accruals of the firm in the control sample by
deducting the discretionary accruals of the firm in the control sample from the
discretionary accruals of the matched firm in the measurement sample. To successfully
do this it is important that firms are matched with a comparable firm. Kothari et al.
(2005) therefore match firms on industry by using the two digit sic code and on return
on assets. The goal is that the matched firms are almost identical except the fact that one
is in the measurement group and the other in the control sample. An essential
assumption in this model is that the firms in the control sample are not involved in
A second assumption that Kothari et al. (2005) make is that firms (that are similar) with
the same performance have the same discretionary accruals. The firms in the
39
measurement sample are in that sample because earnings management is expected for
these firms for example because these firms have gone through a certain event where
earnings management is expected, or they have certain characteristics that the firms in
the control sample do not have. The event or the characteristics are the reason for
expected earnings management. For instance the measurement firms use equity
incentives and the control firms do not.
For this method it is important to be able to match the firm with a control firm that is
comparable. One assumes that there is no earnings management in the control firm. For
this way of measuring accruals firms one needs a reliable control sample of which one
expects no earnings management. Therefore one needs a sample of firms that are similar
but have not gone through that certain event or do not have those characteristics.
In their linear approach Kothari et al. (2005) improve the modified Jones model in two
ways; they add an intercept to the model and control for the company’s performance by
adding return on assets to the model. The first term in the modified Jones model is the
inverse of lagged assets, the inverse of lagged assets is used in the model to mitigate
heteroscedasticity. Despite this, heteroscedasticity was still in a problem in the Jones
(1991) model, Kothari et al. (2005) ad an extra intercept to further reduce
heteroscedasticity.
In the Linear model return on assets is added as independent variable to control for the
effect of prior performance on the discretionary accruals. Kothari et al. (2005) choose to
use return on assets as control variable for prior firm performance because: “by
definition earnings deflated by assets equals return on assets, which in turn measures
performance.” The other argument to use return on assets Kothari et al. (2005) bring
forward is that prior research proves that return on assets is better specified and more
powerful than other control variables for firm performance.
An important question is whether controlling for performance does not lead to over
controlling. Controlling comes with the risk that we remove in part, discretionary
accruals that result from earnings management. This could be the case because the firms
in the industry where the cross‐sectional model is estimated on might have the same
incentives to manage earnings when compared to the treatment firms. This is a possible
weakness in the model.
40
The Kothari et al. (2005) linear model:
NDAi,t= 0 + 1 (1/Ai,t ‐1) + 2( REVi,t ‐ ARi,t) + 3 PPEi,t + 4ROAi,t‐1
NDA Non‐discretionary accruals scaled by lagged total assets

0 A constant, the intercept of the regression
1/Ai,t ‐1 1 divided by lagged total assets
REV The change in revenue scaled by lagged assets
AR The change in accounts receivable scaled by lagged assets
PPE Gross property plant and equipment scaled by lagged assets
ROAi,t‐1 Lagged rate of return on assets
3.13 The Business Model 2007

Ye (2007) introduces a model that combines many of the improvements made in the
models discussed in this chapter. Ye (2007) takes together the improvements of the
different improved variants of the Jones model to make a new model. Besides that he
provides important insight as he shows that even limited tweaking of the Jones model
enhances the power of the model. When improving the Jones (1991) model numerous
possibilities arise, Ye (2007) states that it is important to stop adding extra variables at a
certain point.
Ye (2007) also finds a solution that mitigates a drawback of cross‐sectional models, in

that cross‐sectional models do not take into account firm‐specific parameters. Ye (2007)
ads three additional firm‐specific variables to mitigate this problem.
In his thesis Ye (2007) shows how basic firm specific characteristics influence accruals.
The measures he discusses are: “the level of beginning abnormal working capital,
working capital intensity and the useful life of a firm’s long term assets”. Ye (2007) states
that these factors are not the same within a industry and also change over time for a
company.
The first measure Ye (2007) ads tries to mitigate the weakness caused by reversing
accruals. Ye (2007) uses the reversion of abnormal noncash working capital as his first
measure. He states that in other models people have added lagged accruals to the Jones
model to control for reversing accruals, an example of such a model is the forward
looking model discussed in this chapter. This however does not improve the Jones
model according to Ye (2007), because when high lagged accruals occur, they either
compensate for an abnormally low working capital in the previous year, or they raise
41
the working capital to an abnormal level. The accruals are only in the second case likely
to reverse. Therefore the accrual reversal should be based on the abnormal balance of
working capital instead of lagged accruals. Ye (2007) uses the lagged abnormal balance
of working capital. The abnormal balance of working capital is the amount of which
contemporaneous working capital defers form normal working capital. Normal working
capital is estimated with the historical values of non‐cash working capital. The
abnormal balance can be caused by a fluctuation in business activities or by earnings
management during the previous period. An abnormal balance is expected to vanish in
one operating cycle.
The measure for the lagged balance of the abnormal non‐cash working capital is:
ANCWCi,t‐1 = NCWCi,t‐1 – NNCWCit
ANCWC Abnormal non‐cash working capital

NCWC Non‐cash working capital
NNCWC Normal non‐cash working capital
The second measure Ye (2007) introduces is non‐cash working capital intensity. This is
measured as the ratio of noncash working capital to sales or assets. Ye (2007) states that
different firms have a different working capital intensity. The working capital intensity
depends on the nature of the business and the policy of the firm. These aspects can
change over time and can be different for firms in the same industry. Ye (2007) gives as
example that: “a one dollar increase in revenue leads to a higher increase in receivables
for a firm that mostly makes credit sales than for a firm that makes cash sales”.
Therefore the effect of a change in revenue on accruals depends not just on the amount
of revenue change, but also on its working capital intensity. The second measure Ye
(2007) introduces mitigates the error due to differences in working capital intensity
among firms and for the change over time within one company. The effect of an increase
in the firm sales on current accruals is proportional to: lagged non‐cash working capital
multiplied by the change in revenue, therefore this is the second measure proposed by
Ye (2007). This measure uses an up‐to‐date indicator for non‐cash working capital of a
company. It therefore covers both to time series as the cross‐sectional difference in non‐
cash working capital between companies.
Second measure:
42
NCWCi,t ‐1 × ∆ REVi,t
The third measure Ye (2007) uses are non‐current accruals. As the depreciation expense
of a company depends on more than only the amount of assets. Ye (2007) states that the
depreciation expense also depends on the average depreciation rate. A depreciation rate
can be different between different companies in the industry but can also change within
a firm. The depreciation rate changes depending on the stage of the company in the life
cycle, changes in the life of assets, deprecation methods and the way the asset portfolio
is made up. The Jones (1991) model assumes the deprecation rate is constant, this is
however not realistic. To reckon this, the lagged depreciation rate of a firm is added to
estimate the current depreciation rate. Ye (2007) defines the depreciation rate as
follows:
DEPRATi,t = DEPi,t / PPEi,t
DEPRAT Depreciation rate

DEP Depreciation expense
Ye (2007) discusses the problem of unusual observations. Unusual observations may

cause problems when using accrual models, even when using large datasets. The
problem becomes more severe when using firm or industry specific estimations because
the sample size reduces in those cases.
To mitigate this problem Ye (2007) proposes to winsorize the dependent variable on the
first and the 99th percentile. The independent variables are truncated at the first and
99th percentage.
Kothari et al. (2005) point at the possibility of a nonlinear relation between accruals and
performance variables as ROA. As solution to this problem they propose performance
matching, as discussed in this chapter. Ye (2007) states that this method is only effective
if there is only nonlinearity in one variable. As there might be non‐linearity in more
variables the matching method is not possible. Ye (2007) states that an effective method
to deal with the nonlinearity is to winsorize the independent variable further. “For
example winsorizing return on assets to a range of a normal return on assets ‐0.1 to 0.3
reduces the nonlinearity in return on assets” (Ye, 2007).
43
Ye (2007) combines his new variables with the linear version of the Kothari et al. model,
which itself is a further developed Jones model. This is the model Ye (2007) designed:
TAi,t =βo + β1 /Ai,t −1 + β2 ∆REVi,t + β3 PPEi,t + β4 ROAi,t−1 + β5 NCWCi,t −1 + β6 NNCWC,t + β7

NCWCi,t−1 × ∆REVi,t + β8 DEPRAT,t −1 + β9 DEPRAT,t−1 × PPE i,t .
TA Total accruals
∆REV The change in revenue
ROA Return on assets
NCWC Non‐cash working capital
NNCWC Normal non‐cash working capital
DEPRAT Depreciation rate (the depreciation divided by gross property plant and equipment)
A Total assets
i Subscript indicates the company
t Subscript indicates the year
3.14 Recent literature on accrual models

In addition to the accrual models discussed in this chapter I would like to point out two
recent articles that discuss accrual models. The first is a model by Dechow, Hutton, Kim
and Sloan (2012), their model takes into account that accruals managed in one period
will reverse in a later period. Including this characteristic of accruals “improves the
power and the specification of the tests for earnings management” (Dechow et al., 2012).
In order to include the reversal of accruals in the tests for earnings management
researchers have to be able to identify a period where accruals are expected to be
managed and the periods in which these accruals are predicted to reverse.
The second is the article by Gerakos (2012). This article discusses the new accrual model
introduced by Dechow et al. (2012). Gerakos (2012) analyses the model introduced by
Dechow et al. (2012). Gerakos (2012) states that Dechow et al. (2012) make an important
improvement to the way discretionary accruals are measured, their most important
improvement is that they introduce the dynamics of accruals to the discretionary
accrual measures. These dynamics are the reversing of the accruals that Dechow et al.
(2012) use in their model. Besides introducing the reversal of accruals to the model they
also identify a number of fundamental questions that are of interest for future research.
44
3.15 Chapter 3 summary
In this chapter I discussed a number of accrual models that can be used to measure
earnings management. Most of the more advanced models are based on the Jones model
introduced in 1991. The Jones (1991) model adds to the research as it finds a way to
separate the total accruals in discretionary‐ and non‐discretionary accruals. In the
newer models, based in the Jones (1991) model, the way non‐discretionary accruals are
estimated is more refined. This leads to a better measure of discretionary and non‐
discretionary accruals.
The improvement Jones (1991) makes, is that she takes into account the effect of the
contemporaneous sales revenue and the fixed assets on the non‐discretionary accruals.
In 1995 the modified Jones model was introduced by Dechow et al. (1995), in the
modified Jones model cash sales are used as control variable instead of total sales.
Because credit sales can be managed these are excluded from the model. The hindsight
of this adoption is that the modified Jones model assumes that all credit sales are
discretionary.
The Jones model and the models based on the Jones models can be used in a cross‐
sectional and in a times series test. Both have their strong and weak sides, the cross‐
sectional model is however often described as the better of the two as it creates greater
power due to larger possible sample sizes. In a time series model one assumes that
there is no earnings management in the estimation period, however this assumption is
not always realistic. Another possible problem in a time series model is that there can be
structural changes in the estimation period, these contaminate the test. Time series
models require a long series of observations, the model can only be used for companies
that have such a time series available and this may lead to a selection bias. The main
advantage of a time series analysis is that company specific data is used. A cross‐
sectional model assumes that there is homogeneity within the industry; because the
accruals are estimated on a sample of firms in the same industry similarity is very
important. A possible problem in the cross‐section can be that companies in the group
where the coefficients are estimated on also commit earnings management.
There are two ways to calculate accruals: one can use balance sheet information or cash
flow information. Hribrar and Collis (2002) find that using cash flow information leads to
better results around non‐articulation events. A reason not to use cash flow information
45
can be that this data is only available from 1987 onwards, which can be a problem in
certain research designs.
In this chapter I discussed the forward‐looking model (Dechow, Richardson, & Tuna,
2003), the cash flow model (Dechow & Dichev, 2002), the Larcker and Richardson model
(2004), the Kothari et al. (2005) model and Ye’s syntheses model (2007).
The forward looking model (2003) is an improvement on the Jones (1991) model, it
controls for the future sales growth and it refines the adoption made in the modified
Jones model, by not excluding all credit sales but only the unexpected credit sales from
the model.
The cash flow model is not based on the Jones (1991) model, it uses total accruals, but it
does show that there is a relation between accruals and the preceding, current and
subsequent cash flows.
Larcker and Richardson (2004) improve the modified Jones model in that they ad the
book to market ratio to control the effects of firm growth. They also ad cash flow from
operations to control for firms with extreme performance.
Kothari et al. (2005) provide two ideas to improve the Jones model; one is an
improvement on the linear model where the return on assets is used to control for the
performance of a firm and an extra intercept is added to mitigate heterostcedasticity.
The other method abandons the linear path and measures discretionary accruals as the
difference between the accruals of two similar firms. One of those is expected to use
earnings management and the other is not.
Ye (2007) introduces a model that combines much of the improvements on the Jones
model. He uses the linear model made by Kothari et al. (2005) as basis. With that he
introduces an alternative way to cope with the non‐linearity problem and he ads three
firm specific variables. These variables reduce the weakness, caused the fact that cross‐
sectional models do not use firm specific information to estimate accruals.
46
Chapter 4 estimating the equity incentives
Different measures have been used to examine the relation between earnings
management and incentives caused by remuneration. The proxy’s for equity incentives
defer per research design and the availability of information.
4.1 boundaries of bonus schemes

As proxy for equity incentives the boundaries of bonus schemes have been used by for
example Healy (1985) and Holtausen et al. (1995). They expect that managers manage
earnings around the upper and lower bounds of bonus schemes to maximize their bonus
over the years. For example if earnings exceed the upper bound of the bonus scheme
managers are expected to manage earnings down in the current period to be able to use
these earnings in the next period to reach the bonus target. Healy (1985) compares
accruals with estimated boundaries of bonus plans. He estimates the upper and lower
boundary of bonus plans and examines the accruals around these boundaries. Healy
(1985) assigns observations to different categories; earnings can be under the bonus
range, in the bonus range or above the bonus range. He uses the estimated boundaries of
bonus schemes to assign observations to the different categories. Because the estimated
boundaries of bonus schemes are used, mistakes are made, this makes the model les
reliable. Holthausen et al. (1995) use a similar design, the difference is that they use the
real upper and lower bounds of bonus plans to examine accruals. Holthausen et al.
(1995) are able to do this as he uses inside information from human resource‐consulting
firms. Healy’s method is weak because the boundaries of the bonus plans are estimated,
as no data is available on the real boundaries of the bonus plans. This method does not
lead to reliable results unlike the real boundaries of bonus plans can be used, therefore
this data needs to be available.
4.2 Maximizing earnings in Japan

Shuto (2007) examines the relation between earnings management and maximizing
management earnings in Japan. Management remuneration in Japan is different from
that in the United States. In Japan it is not common to use earnings based compensation
for management, according to Japanese accounting rules companies are not required to
disclose information on executive compensation. This makes it more challenging to
measure executive compensation. Otomasa (2000) explains that in Japan variable
remuneration is different from the bonuses and stock remuneration we know in Europe
47
in the United States. In Japan no bonuses are awarded but executives are penalized
when the firm performance worsens. The only information disclosed considering
remuneration in Japan is the total cash compensation for the complete board. This is
used as proxy for executive compensation; this total compensation is compared to the
discretionary accruals calculated with the Jones model. In other words Shuto (2007) uses
the total remuneration to management as proxy for incentives to manage earnings.
Shuto (2007) finds that when discretionary accruals increase the executive
compensation increases as well and that in years with low payment earnings decreasing
accruals and extraordinarily items can be found.
4.3 Proxy for equity incentives

The measures discussed in the last two sections focused on bonus plan based or total
earnings incentives. This master’s thesis focuses on stock‐based incentives. Bergstresser
and Philippon (2006) introduce a measure for the power of equity incentives. They
measure this as the: “dollar change in the value of a CEO’s stock and options holdings
that would come from a one percentage point increase in the company stock price”.
They call this measure ONEPCT, and compute it with data from the “compustat executive
compensation” database.
ONEPCT i,t = 0.01 × PRICE i,t × (SHARES i,t + OPTIONS i,t)
ONEPCT The dollar change in the value of a CEO’s stock and options holdings that would
come from a one‐percentage point increase in the company stock price.
PRICE The company share price
SHARES The number of shares held by the CEO
OPTIONS The number of options held by the CEO
Bergstresser and Philippon (2006) use this measure to calculate their measure for equity
incentives. They state that: “this is a measure of incentives normalized in a way that
captures the share of a hypothetical CEO’s value in the equity of his or her company”.
INCENTIVE RATIO i,t = ONEPCT i,t / ( ONEPCT i,t + SALARY i,t + BONUS
i,t)
INCENTIVE RATIO The share of a CEO’s total compensation that would come from a one‐
percentage point increase in the value of the equity of his or her company.
ONEPCT The dollar change in the value of a CEO’s stock and options holdings that
would come from a one‐percentage point increase in the company stock price.
SALARY The cash salary awarded to the CEO
BONUS The cash bonus awarded to the CEO
48
This measure implicitly assumes that if the share price changes the change of the value
of the options is the same as the change in value of the shares. However the change in
the price of options depends on several factors and is not the same as the change in the
price of shares.
The price of an option depends amongst other things, on the exercise price of that
option, the price of the underlying stock, the expiration date, the volatility of underlying
stock, the interest rate and expected dividend policy. Therefore the change of the price
of an option is a bit more complex than assumed in the model by Bergstresser and
Philippon (2006), this makes their proxy quite rough. Bergstresser and Philippon (2006)
use a technique introduced by Core and Guay (2002) to mitigate this problem, Core and
Guay (2002) use an option delta that compensates for the different behavior of the value
of the options taken into account the option portfolio owned by the manager. This
option delta represents the sensitivity of the price of the options to the price of the
stock; it is calculated using the Black‐Scholes‐Merton formula. Without using the option
delta one would implicitly assume that all options are deep in the money, because one
would assume that the option delta is one, which is true for deep in the money options.
ONEPCToptions = 0,01 × share price × option delta × number of options
ONEPCTshares = 0,01 × share price × number of shares
4.4 Summary
Different proxies have been used to estimate the incentive caused by remuneration of
the management. Healy and Holthausen (1995) investigate earnings management
around the boundaries of bonus plans, where Healy estimates these boundaries
Holthousen (1995) uses the real boundaries of bonus plans, he is able to do this because
he has obtained this data for his sample from human resource consulting firms. Shuto
(2007) uses the total cash remuneration awarded to the entire management, he
examines if there is more earnings management in years that the total cash
remuneration is higher. Bergstresser and Philippon (2006) introduce a model that
measures the incentive created by stock and option‐remuneration. They use the share of
CEO’s total compensation that that would come from a one percentage point increase in
the value of the equity of his or her company.
49
Chapter 5 empirical research on earnings management due to
equity incentives
5.1 Introduction
Prior research provides evidence for a positive relation between earnings management
and equity incentives. This chapter is divided into five sections. After this introduction
the second part (5.2) contains so called remuneration studies. These studies examine
earnings management caused by bonus schemes. These schemes award a bonus when a
certain performance‐related target is met. The third section (5.3) contains equity
incentive studies, these studies examine the relation between incentives that are due to
equity holdings and earnings management. The fourth section (5.4) contains studies that
measure if the relation between equity incentives and earnings management deviates
for the CEO and the CFO. This chapter ends with a summary in section 5.55.
5.2 Remuneration
This part contains studies that measure the relation between remuneration‐based
incentives and earnings management.
5.2.1 Bonus‐based equity incentives and the Healy accrual model

Healy (1985) finds evidence for earnings management around the upper boundaries of
bonus plans using his early accrual model discussed in section 3.2 of this master’s thesis.
Healy (1985) measures if earnings management is caused by bonus contracts.
Incentive model
Healy (1985) assigns the observations of earnings to a part of the bonus contract; this
can be under the lower bound of the bonus range, in the bonus range or above the bonus
range. Healy estimates the lower and upper bounds of the bonus contracts. Hereby
Healy (1985) classifies firms with low earnings as firms that are under the lower bound
of bonus contracts and firm with high earnings as over the upper bounds of the bonus
plan.
5
A summary of this chapter is included in appendix 1
50
Bonus Plan
Bonus
Above the upper boundary of bonus plan
Within the boundaries of the bonus plan
Under the lower boundary of the bonus plan

Earnings
Expectations
Healy (1985) expects that managers maximize their bonus; this is called the bonus‐
maximizing hypothesis. He expects that managers manage earnings downwards when
under the lower boundary of bonus plans; this is called taking a bath. Managers are
expected to manage earnings down in the current period because if they realize they will
not make the lower bound of the period, they will manage earnings down in the current
period so they can manage the earnings upwards in the next period. Above the upper
boundary of the bonus plans managers are expected to manage earnings downwards,
because more earnings do not increase their bonus in that situation. By managing the
earnings down in this period he will be able to manage the earnings up in the next
period, and he will thereby be able to maximize his bonus. Healy (1985) therefore
expects negative accruals under the lower bound of the bonus contract, negative
accruals above the upper boundaries of the bonus contracts and positive accruals in
between the boundaries of the bonus scheme.
Accrual model
Healy (1985) measures total accruals as the estimated difference between reported
accounting earnings and cash flow from operations.
TA t = – DEPt – XIt ×D¹ + ARt + ∆ INVt – ∆ APt – (∆ TPt + DEFt) ×D²
51
DEPt Depreciation
XIt Extraordinary items in year t
∆ ARt Accounts receivable in year t less accounts receivable in year t‐ 1
∆ INVt Inventory in year t less inventory in year t ‐ 1
∆ APt Accounts payable in year t less accounts payable in year t‐1
∆ TPt Income taxes payable in year t less income taxes payable in year t‐1
DEFt Deferred income tax (expense) credit for year t
D1 1 if the bonus plan earnings are defined after extraordinary items
0 if the bonus plan earnings are defined before extraordinary items
D2 1 if bonus plan earnings are defined after income taxes
0 if bonus plan earnings are defined before income taxes
Healy’s (1985) accrual model is described in section 3.2 of this master’s thesis; he
calculates abnormal accruals as the total accruals minus the normal accruals. As proxy
for normal accruals Healy (1985) uses the average total accruals of the last 10 years.
Discretionary accruals:
DAt = TAt – TAta

TAta Average total accruals of 10 years prior to year t
Tests
Healy compares the accruals with the place the firms’ take in the bonus contracts to test
his hypotheses. He uses a sample of the 250 largest U.S. listed industrial firms running
from 1930 until 1980. Healy finds prove for his bonus maximizing hypothesis, he finds
that managers manage earnings downward under the lower bound of the bonus plan
and downward above the lower bounds of the bonus plans.
Dechow et al. (1995) show that the way Healy assigns the earnings to the parts of the
bonus contract biases the model. They show that if one takes a sample of 1000 firm
years randomly selected from firms with low earnings one finds negative accruals. The
other way around if one selects a sample from companies with high operating cash
flows. Healy assigns companies with low earnings or high operating cash flows to a
certain group; the mistake in the model is that these low earnings or high operating cash
flows do not indicate if that performance is a relatively good or bad performance for that
firm. Healy’s results might therefore be biased, because Healy (1985) estimates the
lower and upper bound of the bonus contract and assigns observations on basis of high
52
or low earnings or operating cash flow without knowing whether this is a relatively
good or bad performance for the firm.
5.2.2 Bonus‐ based equity incentives and the Jones accrual model
Holthausen et al. (1995) conduct a study that is similar to Healy (1985), the difference is
that Holthausen et al. (1995) use the Jones and modified Jones model to measure
accruals and that they use an improved way to measure the equity incentives. Their
study examines if the same results are found with a better accrual model and a better
model to estimate the equity incentive.
Model
Holthausen et al. (1995) use the Jones and modified Jones model to measure accruals and
find somewhat different results than Healy (1985). The Jones and modified Jones model
are described in section 3.4 and 3.5 of this master’s thesis.
To estimate the equity incentive Holthausen et al. (1995) use the same measure Healy
used (1985). They also assign observations to parts of the bonus schemes; like in Healy’s
study observations can be under the lower boundary of the bonus scheme, in between
the boundaries of the bonus scheme or above the boundaries of the bonus scheme. The
important difference is that Holthausen et al. (1995) have insider information about the
bonus schemes to their disposal. They obtain this confidential information from two
human resource management‐consulting firms. Therefore Holthausen et al. (1995) are
able to use the real boundaries of the bonus plans, where Healy (1985) had to estimate
this. Therefore Holthausen et al. (1995) have a more precise measure to assign
observations to the different categories. Holthausen et al. (1995) obtain their sample
from two different human resources management consulting firms, they test a sample
running from 1982 to 1984 and 1987 to 1991.
Findings
Holthausen et al. (1995) find evidence, like Healy (1985) for management managing
earnings downward above the upper limits of the bonus schemes, but they do not find
evidence for earnings being managed downward under the lower bound of the bonus
scheme. Thereby the evidence Holthausen et al. (1995) find is not as strong as the
evidence Healy finds. Holthausen et al. (1995) state that the reason for this weaker
relation is due to the different models used. Healy estimates the upper and lower
bounds of bonus plans where Holthausen et al. (1995) use confidential insider
53
information on bonus plans to assign the observations. Holthausen et al. (1995) state
that the method used by Healy (1985) to assign observations to the three different
categories has influenced his results.
5.3 Equity incentives

Where the early research on management incentives due to remuneration incentives
focused on bonus plan‐based incentives more recent studies have examined equity‐
based management incentives like I use in this master’s thesis. In the next section I
discuss a number of models that study the relation between equity incentives and
5.3.1 Equity incentives and the Jones accrual model

Bergstresser and Philippon (2006) measure the relation between equity incentives and
earnings management. They use an accrual model to measure earnings management and
an incentive ratio as proxy for the total equity incentive.
Expectations
Bergstresser and Philippon (2006) expect that CEO’s manage earnings upwards in order
to maximize the gains from equity incentives; they expect that CEO’s whose
remuneration is more sensitive to the company’s stock price are more likely to use
Accrual model
Bergstresser and Philippon (2006) use the Jones and Modified Jones model to estimate
the discretionary accruals. The Jones and modified Jones model are discussed in section
3.2 and 3.3 of this master’s thesis. Bergstresser and Philippon (2006) use both the cash
flow method as the balance sheet method to calculate the total accruals, how these
methods work is explained in section 3.7. Bergstresser and Philippon (2006) use a time‐
series model to estimate the non‐discretionary accruals. They use data back to 1979 to
estimate the coefficients in the Jones model.
Equity incentives
Bergstresser and Philippon (2006) examine the relation between equity incentives and
earnings management by estimating the relation between discretionary accruals and an
incentive ratio. This ratio represents the sensitivity of the total CEO remuneration to a
change in stock price. I discussed this incentive ratio in section 4.3.
54
They also use this incentive ratio adding the calculated option delta to the formula, as
suggested by Core and Guay (2002), this option delta is discussed in section 4.3.This
delta is calculated using the Black‐Scholes‐Merton formula, and indicates the relation
between the share price and the value of the options. In the model without the option
delta the share price multiplied with the number of options is used as proxy for the
value of options. When using this proxy one assumes that the option delta is one.
Bergstresser and Philippon (2006) find similar results with both these measures.
Sample
Bergstresser and Philippon (2006) test the relation between their incentive ratio and
discretionary accruals over a period from 1994 to 2000. The sample they use consists of
publicly held firms, they split the sample in 4199 firm year observations with firms that
have 1996 lagged assets above one million dollars and 4671 firm year observations with
firms that have 1996 lagged assets below one million dollars. They obtain their data on
CEO remuneration from the “execucomp” database and obtain information used to
calculate accruals from the “compustat” database.
Model
Bergstresser and Philippon (2006) examine the relation between equity incentives and
earnings management with the following equation.
Absolute value of discretionary accruals = + × Incentive_Ratioi,t + i,t × Controls + i,t
Bergstresser and Philippon (2006) use different regressions for firms with total assets
under one million dollars and firms with total assets over one million dollars.
Bergstresser and Philippon (2006) winsorise all variables on the 1st and 99th percentiles.
The Controls in the model represents a set of control variables that control for firm
characteristics as size, leverage, firm age, and volatility of sales growth. Bergstresser and
Philippon (2006) use different regressions to test their model as they use both the
absolute value of discretionary accruals and the absolute value of total accruals, and use
accruals calculated with cash flow information and accruals calculated with balance
sheet information, they test the equity incentive in two versions: one with the option
delta proposed by Core and Guay (2002) and one without this option delta. All methods
lead to similar results.
55
Findings
Bergstresser and Philippon (2006) find that in firms where the CEO’s total remuneration
is more sensitive to the stock price, discretionary accruals are more actively used to
manage earnings.
5.3.2 Equity incentives and analyst forecasts

Cheng and Warfield (2005) examine the relation between equity incentives and earnings
management. The equity incentives they examine are equity incentives that come from
stock ownership and stock and options awarded.
Cheng and Warfield (2005) measure earnings management as the likelihood of managers
meeting or beating the analyst forecast; they compare this measure with their measure
of equity incentives. They also examine if managers who manage earnings sell more
shares after meeting or just beating the analyst forecast. Besides the meeting or beating
of the analyst forecast they also use abnormal accruals as measure for earnings
management. The abnormal accrual part of their study is the part that is relevant for this
thesis as accrual models are used in this master’s thesis as well.
Expectations
The first and main hypothesis of their study is:
“The incidence of earnings management is positively related with equity incentives”.
Besides the main hypotheses the second relation Cheng and Warfield (2005) examine is
that: “firms with high equity incentives are more likely to sell shares in the future and
that motivates these managers to engage in earnings management to increase the value
of the shares to be sold”. They expect that there is a difference in CEO’s who receive
persistent high equity incentives and CEO’s whose equity incentives are less persistent.
They expect that CEO’s whose equity incentives are less persistent are more likely to
report surprisingly high or low earnings and that CEO’s whose incentives are persistent
do not report shock earnings but more constant earnings. Managers who have high
equity incentives over a long period of time are expected to smooth earnings, and
therefore not report shock earnings but a more constant amount of earnings. The third
relation Cheng and Warfield predict is that managers sell more shares in the period after
the announcement of earnings.
56
Equity incentives
As measure for equity incentives Cheng and Warfield (2005) use a ratio of five equity
elements divided by the total outstanding shares of the firm. The fife elements of equity
incentives they use are: “option grants in the current period, un‐exercisable options
(excluding option grants in the current period), exercisable options, restricted stock
grants, and stock ownership.” They deflate these indicators by total outstanding shares
of the firm. Cheng and Warfield (2005) measure the five different elements separately as
well; however in most analyses the ratios of restricted stock grants are combined to one
measure with share ownership and the option grants are combined to one measure with
un‐exercisable options. Cheng and Warfield (2005) use these ratios because the benefit
of possible gains due to increased short‐term stock prices is shared by all shareholders.
Therefore the gain acquired by the CEO is proportional to the ratio of the equity
incentives in the total shares outstanding.
Equity incentives and earnings management measured with abnormal accruals

Cheng and Warfield (2005) state that equity incentives can be an incentive to use income
increasing accruals but that income smoothing incentives may lead to the use of income
decreasing accruals. That means that the relation between discretionary accruals and
equity incentives is positive or negative depending on the fact which of the two
incentives is stronger (Cheng & Warfield, 2005).
Cheng and Warfield (2005) define abnormal accruals as the difference between total
accruals and normal accruals. They estimate the normal (non‐discretionary) accruals
with the cross‐sectional version of the Jones (1991) model as described in section 3.4
and 3.6.2 of this master’s thesis. Cheng and Warfield (2005) use the two digit SIC code to
group the firms into industry categories and delete all industry groups with less than
seven observations.
They use the following regression to estimate the relation between equity incentives
and earnings management:
Abnormal accrualsi,t = 0 + ß Eq_incent,t + 1 sizei,t + 2 Leveragei,t + 3 Bonusi,t + 4 Riski,t + 5
Growthi,t + 6 NOAi,t‐1 + 7 Litigationi,t+ 8 Implicit claimi,t + I,r
Eq_incenti,t Equity incentives measured in shares in the year t scaled by the number of
outstanding shares. The measure contains five components explained
above.
57
Sizei,t Natural logarithm of the market value of the firm at the end of the year
Leveragei,t Long term debt divided total assets at the end of the year
Bonusi,t Dummy variable is 1 for years that CEO’s are awarded a bonus and 0 if they are
not awarded a bonus.
Riski,t Market beta in fiscal year t estimated from weekly returns
Growthi,t The book to market ratio at the end of the fiscal year
NOAi,t‐1 Net operating assets at the end of fiscal year t‐1 scaled by sales for fiscal year t‐1.
Litigationi,t Dummy variable is one if the firm part of one of the following industries:
pharmaceutical/biotechnology, computer, electronics, or retail and zero if the
firm is in another industry.
Implicit claimi,t The implicit claim is proxied by the intensity of labour. This is calculated as one
minus the ratio of gross property plant and equipment and total assets.
Cheng and Warfield (2005) add the above‐mentioned control variables as these might
have an influence on the abnormal accruals. They add firm size to control for managers
that tend to manage earnings to decrease political cost. Leverage is added to the model
to control for the effects of debt covenants. To control for earnings management due to
bonus‐related incentives Cheng and Warfield (2005) add a dummy variable that is 1 if a
positive bonus is awarded and zero otherwise. Growth is added to the model because
higher accruals are expected fast growing firms.
CEO Trading
Cheng and Warfield (2005) also examine the share sales after the earnings
announcement; they examine share sales in the first six months and second six months
period after earnings announcements. This data is obtained from the SEC ownerships
reporting file. Cheng and Warfield measure CEO trading as:
Net CEO sales = Open market sales ‐ (Open market purchases + Options exercised).
The net CEO sales are scaled by the firm’s market value at the end of the fiscal year.
Earnings surprises
A third measure Cheng and Warfield (2005) use are earnings surprises. Earnings
surprises are measured as: “the difference between the actual earnings and the earnings
analysts predicted”, these earnings are measured per share. As measure for the analyst
forecast Cheng and Warfield (2005) use the most recent analyst forecast within a three‐
months period before the earnings announcement.
Sample
Cheng and Warfield (2005) use a sample that contains companies with information on
stock‐ and option‐based remuneration available over the 1993‐2000 period. They
58
obtain information on executive compensation from the standard & poor’s “execucomp”
database. Cheng and Warfield use a sample of 9472 firm years. Financial institutions and
firms active in the utilities sectors are not included in the sample because these
industries are highly regulated and they expect that these managers might have
different incentives to manage earnings.
Findings
Cheng and Warfield (2005) find that: “CEO’s that have high equity incentives use more
income‐increasing abnormal accruals than CEO’s with low equity incentives”. This
means that the incentive to increase earnings is stronger than the incentive to reserve
earnings for the future. Reserving earnings for the future would be in line with the
income‐smoothing hypothesis. Cheng and Warfield (2005) find that these results are
mainly due to managers that have less persistent equity incentives, these managers have
less incentives to reserve earnings for the future and are affected less by accrual
reversals. They state that if the sample is split based on the average ranking of equity
incentives over the sample period they find that the relation between equity incentives
and discretionary accruals is not significant for firms with persistent equity incentives
and that the relation is significant for companies whose equity incentives are less
persistent. They also find that share sales by CEO’s increase after the earnings are
presented in periods with high income increasing abnormal accruals.
Using their other indicator for earnings management Cheng and Warfield (2005) find
that managers who have persistent high equity incentives report earnings that are in
line or slightly better than the analyst forecasts. Managers with consistently high equity
incentives are less likely to report surprisingly high earnings. This is consistent with
what they expect, managers with constant high equity incentives are expected to
manage earning in a way that results in smooth earnings, as this provides them with the
most profit from their current and future equity incentives.
5.4 CEO and CFO equity incentives

5.4.1 CEO and CFO equity incentives and accrual management
Jiang, Petroni and Wang (2010) examine the relation between CFO equity incentives and
earnings management, they compare their results with the relation between CEO equity
incentives and earnings management to find which relation is stronger. By doing this
they continue on the studies of Bergstresser and Philippon (2006) and Cheng and
59
Warfield (2005). They use two measures for earnings management: accruals and the
likelihood of beating analyst forecast. The findings found with the accrual model are
most relevant for this thesis as an accrual model is used in this master’s thesis as well.
Expectations
Jiang et al. (2010) expect that CFO equity incentives have more influence on earnings
management than CEO equity incentives. To examine this they measure the relation
between equity incentives and earnings management for the CEO and the CFO
separately and they also measure the joint relation. Jiang et al. (2010) also expect that
the Sarbanes Oxley act has led to a weakening in the relation between equity incentives
and earnings management, therefore they measure these relations in the period prior to
the Sarbanes Oxley legislation and the period after the Sarbanes Oxley legislation.
Equity incentives
As measure for equity incentives Jiang et al. (2010) use the equity incentive ratio by
Bergstresser and Philippon (2006). This ratio is described in section 4.3 of this thesis.
Kim et al. (2011) use this measure as well. Kim et al. (2011) separate the equity incentive
ratio in an option incentive ratio and a stock incentive ratio, Jiang et al. (2010) do not do
this, they use the total equity incentive ratio. They add, like Kim et al. (2011) and
Bergstresser and Philippon (2006) do, the option delta to the model as proposed by Core
and Guay (2002). The ratio contains the value of: the shares a manager owns, the newly
granted options, unexercised‐exercisable options, and un‐exercisable options.
Accrual model
Jiang et al. (2010) estimate the relation between equity incentives and earnings
management using accrual models.
They calculate the total accruals as the “difference between earnings before
extraordinary items, scaled by lagged total assets”. In this way they calculate the total
accruals using cash flow data as described in section 3.7 of this master’s thesis.
With the total accruals they calculate the discretionary accruals. They use the Forward
looking Jones model by Dechow et al. (2003) to calculate discretionary accruals, this
model is discussed in section 3.9 of this master’s thesis.
TAi,t= 1 + 2((1+K) ∆salesi,t ‐ ∆ARi,t) + 3PPEt + 4 (TAi,t,‐1/ Ai,t‐2) + 5 GR_salesi,t
60
TA Total accruals in the current year of firm i, scaled by year t‐1 total assets.
∆sales Change in sales, scaled by year t‐1 property plant and equipment
∆AR The change in accounts receivable, scaled by year t‐1 total assets.
K The slope coefficient from a regression of ∆AR on ∆sales
TAi,t,‐1/ Ai,t‐2 Total accruals from the year before the current scaled by total assets of two years ago for
firm i.
GR_sales The change in sales for firm i, from year t to t+1, scaled by year t sales.
Jiang et al. (2010) calculate the discretionary accruals by deducting the non‐
discretionarily accruals from the total accruals. They use the absolute value of
discretionary accruals in their final analysis.
Final model
With the absolute value of discretionary accruals and the equity incentives Jiang et al.
(2010) estimate the following model:
Absolute value discretionary accruals = 0+ 1 incentive ratio + controls +
Jiang et al. (2010) use the absolute value of total and discretionary accruals instead of
the normal value as they expect that equity incentives increase extreme accruals. They
control for variables as: firm size, a dummy for firm age, the volatility of sales growth
and the leverage.
Sample
The sample Jiang et al. (2010) use consists of S&P 1500 firms that are part of the
“execucomp” database and have data for CEO and CFO compensation available. They find
a total of 17542 firm years with the data available on compensation of the CEO and the
CFO. They test their sample over the period from 1993 to 2006.
Findings
Jiang et al. (2010) find that there is a positive relation between both CEO and CFO equity
incentives in the pre‐Sarbanes Oxley 1994‐2000 period. The relation for CFO equity
incentives they find is stronger than the relation for CEO equity incentives. In the post‐
Sarbanes Oxley period they do not find a positive relation between CEO or CFO equity
incentives and earnings management. In the period after the implementation of the
Sarbanes Oxley legislation they even find a negative relation between CFO equity
incentives and earnings management. Jiang et al. (2010) contribute to the research of
Cheng and Warfield (2005) and Bergstresser and Philippon (2006) by showing that the
61
relation between CFO equity incentives and earnings management is stronger than the
relation between CEO equity incentives and earnings management. They also show that
this relation changes for a pre and post‐Sarbanes Oxley sample. In the post‐Sarbanes
Oxley sample they even find a negative relation between CFO equity incentives and
earnings management. This is possibly due to the fact that in the period after the
introduction of the Sarbanes Oxley act the markets reacted negatively to positive
earnings surprises if the firms’ management has high equity incentives.
5.4.2 CEO and CFO equity incentives and crash risk

Kim, Li and Zang (2011) examine the relation between CEO and CFO equity
compensation and firm crash risk. They find proof that the sensitivity of the CFO’s
compensation to the stock price is positively related to firm’s crash risk. They however
do not find positive evidence of the impact of the sensitivity of the CEO’s compensation
and crash risk.
Expectations
Kim et al. (2011) predict to find a positive relation between equity incentives and crash
risk. They test whether this relation is stronger for the CFO or for the CEO. They also
expect that options have a more powerful impact on manager’s behavior than stock
based incentives.
Crash risk
Kim et al. (2011) examine the relation between equity incentives and crash risk.
Kim et al. (2011) estimate the firm‐specific weekly returns for each firm and year; they
need this measure to be able to measure the firm specific crash risk. They define crash
weeks in a given fiscal year for a given firm, as those weeks in a year where the company
experiences weekly returns 3.2 standard deviations below the mean.
Their first measure of crash probability is a dummy that is 1 in years where firms
experience one or more crash weeks and zero in years where the firms do not
experience crash risk.
The second measure of crash risk they use is the negative conditional return skewness:
“This is estimated by taking the negative of the skewness of firm‐specific weekly returns
for each sample year and dividing that by the standard deviation of firm‐specific weekly
returns raised to the third power”.
62
The third measure of crash risk Kim et al. (2011) use is calculated by “separating all the
weeks with firm‐specific weekly returns below the annual mean from weeks with firm‐
specific returns that are higher than the yearly mean. They then calculate the standard
deviation for each of these subsamples separately. The final measure is the logarithm of
the ratio of the standard deviation in the weeks with a mean below average to the
standard deviation of the weeks above average.”
Equity incentives
Kim et al. use the measure proposed by Bergstresser and Philipon (2006) to estimate
equity incentives. This measure is discussed in section 4.3 of this master’s thesis. They
first calculate ONEPCT this is the: “dollar change in the value of an manager’s option
holdings that would come from a one percentage point increase in the company stock
price”. They also calculate this measure for the stock holding
The difference with Bergstresser and Philippon (2006) is that they calculate this measure
for both option and stock holdings together. Kim et al. (2011) calculate this measure
separately for the option and stock holdings to be able to test if the influence of option
or stocks is stronger. Kim et al. (2011) estimate ONEPCT with the option delta as
proposed by Core and Guay. The option delta is explained in section 4.3.
Sample
The sample Kim et al. (2011) use consists of US‐based firms obtained from the
“Compustat Executive Compensation” database. Their sample runs from 1993 to 2009.
They delete observations with missing data in the “compustat” database and in the
Research in Security Prices databases. They also exclude firms whit a year‐end share
price that is lower than 1 dollar. The sample consists of 29638 firm year observations.
Main regression
Kim et al. estimate the relation between crash risk and equity incentives with the
following regression:
CrashRisk = o + 1 × Incentive + × Control Variables +
Kim et al. (2011) use the measure of the dependent variable in year t and the measure of
the independent variable in year t‐1. They add a number of control variables that are
seen as possible predictors of the crash risk. They also control for abnormal accruals
calculated with the modified Jones model; this is done to ensure that the relation
63
between equity incentives and crash risk is not due to managers using accrual
accounting, because they expect that accruals accounting increases crash risk.
Findings
Kim et al. (2011) find that CEO and CFO stock incentives are not related to the future
crash risk. But that CEO and CFO option incentives are positively related to crash risk.
The relation for the CEO is however less significant than the relation for the CFO. They
find that: “CFO option incentives dominate CEO option incentives in predicting the
future crash risk”. Kim et al. (2011) use like Jiang et al. (2010) both the CEO and CFO
equity incentives. They find that there is no relation between stock incentives and
earnings management but a positive relation between option incentives and earnings
management. They find this relation over the period 1993 to 2009. These findings in
combination with the Jiang et al. (2010) paper leads to a number of things to further
examine. I discuss this in the next chapter. A possible problem with the study by Kim et
al. (2011) is that it covers the 1993‐2009 period, this period contains a period before the
implementation of the Sarbanes Oxley act and a period after the Sarbanes Oxley act. This
might contaminate their research as Jiang et al. (2010) show that the relation between
equity incentives and earnings management is different before and after the
implementation of Sarbanes Oxley. This might also hold for the relation between equity
incentives and crash risk.
5.5 Summary
This chapter discusses empirical research on equity incentives and earnings
management. The first part discusses research on remuneration incentives. The first of
the remuneration papers is the paper by Healy (1985), he finds evidence for his bonus
maximizing hypothesis. Healy (1985) finds that CEO’s manage earnings downward
under the lower boundary of bonus plans and above the upper boundaries of bonus
plans. The article by Holthausen et al. (2005) is the second remuneration article
discussed. Holthausen et al. (1995) find a little different results, they find like Healy
(1985) evidence for managers managing earnings downward from above the upper
boundary of bonus plans but they do not find evidence for managers managing earnings
downwards below the lower bound of the bonus plan.
The second part of this chapter discusses three papers that examine the relation
between equity incentives and discretionary accruals. Bergstresser and Philippon
64
(2006) find evidence for a positive relation between discretionary accruals and equity
incentives. As measure for equity incentives they use a ratio that measures the
sensitiveness of CEO income to the stock price of the company. Cheng and Warfield
(2005) measure the relation between equity incentives and CEO’s meeting or just
beating the “analyst forecasts” and the relation between equity incentives and
discretionary accruals. Cheng and Warfield (2005) find that CEO’s who have high equity
incentives use more income increasing abnormal accruals. They also find that managers
with persistent high earnings are less likely to present surprisingly high earnings than
managers with less persistent equity incentives. And they find that CEO’s with persistent
high earnings are likely to meat or just beat the “analyst forecasts”.
The third part of this chapter discusses studies that measure the relation between
equity incentives and earnings management for both the CEO and the CFO. Jiang et al.
(2010) find that there is a positive relation between both CEO and CEO equity incentives
and earnings management. They find that the relation between CFO equity incentives
and earnings management is stronger than the relation between CEO equity incentives
and earnings management. They find these positive relations in a sample running before
the introduction of the Sarbanes Oxley act; in a sample running in the period after the
introduction of the Sarbanes Oxley act, they find no positive relation between equity
incentives and earnings management. Kim et al. (2011) measure the relation between
equity incentives and crash risk. Kim et al. (2011) examine if this relation is different for
CEO equity incentives or CFO equity incentives. They find that the relation between
equity incentives and crash risk is stronger for the CEO than for the CFO. They also
examine if there is a difference between stock and option based equity incentives. They
find no positive relation between stock‐based equity incentives and crash risk, they do
find a positive relation between option based equity incentives and crash risk.
In the next chapter I present my expectations and the hypotheses of my master’s thesis.
These expectations are based on the prior research I discussed in this chapter.
65
Chapter 6 Hypothesis
6.1 Hypothesis 1
I expect to find a positive relation between earnings management and equity incentives.
I expect that when the compensation of an executive is more sensitive to the stock price
he uses earnings management to increase the performance of the company. Prior
research shows evidence for this relation as well. Examples of these prior studies are
Cheng and Warfield (2005) and Bergstresser and Philippon (2006). They find a positive
relation between accrual‐based earnings management and measures for equity
incentives.
H1: There is a positive relation between earnings management and equity incentives.
6.2 Hypothesis 2
Because the main job of a CFO is to take care of the financial administration I expect that
the CFO has more influence on earnings management than the CEO has. Providing the
CFO with a compensation that is highly sensitive to the stock price of the firm is
therefore, in my opinion, asking for trouble. Because the CFO is the manager that is
responsible for composing the financial statements and for making accrual decisions it is
very easy for him to manage earnings. As the CFO can manage earning with less effort
than the CEO the CFO could be tempted to use earnings management quicker than a
CEO. This expectation is supported by the findings of Jiang et al. (2010) , discussed in the
previous chapter. They find a positive relation between discretionary accruals and
earnings management; they find a stronger relation for CFO equity incentives than for
CEO equity incentives.
H2: The positive relation between earning management and equity incentives is stronger
for CFO’s than for CEO’s
6.3 Hypothesis 3
I also expect to find a stronger relation for option‐based incentives than for share‐based
incentives. This is due to the character of options as short‐time incentive. It is easy for
management to make quick gains with options in years where many options can be
exercised. Due to the reversing character of accruals they are well suited to use in
combination with short‐term incentives. This is because when using accruals earnings
can only be managed upwards for a limited amount of time, after this accruals will
66
reverse. This suits the character of options, as earnings only have to be managed
upwards for a short period of time to be able to exercise the options at a good price. Kim
et al (2011) find that there is no relation between stock‐based equity incentives and
crash risk over the 2003‐2009 period but they do find that there is a relation between
option‐based equity incentives and crash risk. This supports my expectation to find a
stronger relation between option‐based equity incentives and earnings management
than between stock‐based equity incentives and earnings management. Cheng and
Warfield (2005) find that managers with less persistent equity incentives are more likely
to present surprisingly high earnings than managers with more persistent equity
incentives. As option incentives are less persistent than stock‐based equity incentives
this leads to the expectation that option‐based equity incentives are short time
incentives and lead to management managing earnings up in periods when there are
many option‐based incentives available.
H3: The relation between equity incentives and earnings management is stronger for
option‐based incentives than for share‐based incentives.
6.4 Hypothesis 4
I expect to find a positive relation between earnings management and equity incentives.
This is in line with the findings of Jiang et al. (2010) in their study; they find that the
relation between equity incentives and discretionary accruals weakens in the period
after the implementation of the Sarbanes Oxley act. I expect to find similar results. I also
examine if this relation is different for option‐based and share‐based incentives. I expect
to find that the decline of this relation is stronger for the short‐term option‐based
incentives than for the share‐based incentives. As to my opinion these short‐term
option‐based incentives are more sensitive for opportunistic use. Managers can more
easily manage earnings upwards for a short period of time than they can for a longer
period. Another reason that the positive relation between earnings management and
option‐based incentives might have declined is that the market reacts negatively to
positive earnings surprises in the period after the Sarbanes Oxley act has come into
effect, this is shown by Kothari et al. (2006). Earnings growth to maximize the rewards
from option incentives is therefore not effective anymore if the earnings growth comes
by surprise. Steady long‐term growth however may still lead to a positive relation from
the markets. Another possible explanation why the relation for option‐based incentives
67
should decline faster than the relation for stock‐based incentives is that firms may have
become more careful awarding managers which options‐based incentives, because the
short term incentives may have caused harmful earnings management, as well as that
option‐based incentives do not succeed to align the interests of management with the
owners of the firm as well long term incentives do. As stockholders are often better off
by long‐term growth than short‐term earnings shocks, this applies in particular for large
institutional shareholders like pension funds. Another reason is that option‐based
incentives might have become less popular due to the effect that option‐based incentives
lead to risk‐seeking behavior as I explained in the introduction to this master’s thesis.
The risk seeking management might not be in line with the interest of stockholders.
H4: The positive relation between equity incentives and earnings management declines
after the major accounting scandals of the early 2000s, this decline is stronger for option‐
based incentives.
68
Chapter 7 Research design and methodology
7.1 introduction
This chapter describes the empirical research of this master’s thesis; it explains what
methods are used to obtain the results presented in the next chapter. As described in the
introduction of this master’s thesis I want to measure the relation between equity
incentives and earnings management and thereby focus a possible different relation for
CEO and the CFO, and different effects of option‐based and stock‐based incentives. I test
this relation over a ten‐year period running from 1999 to 2009. I choose this period
because it contains accounting scandals that have had a major impact on the accounting
world. I use an accrual model to measure earnings management and an incentive ratio
to measure the equity incentive. This chapter will first describe the accrual model and
the incentive ratio I use and then describe the sample used.
7.2 Accrual model

To measure earnings management I use the linear cross‐sectional performance model
by Kothari et al. (2005) I described this model in section 3.12. I use this model because if
one measures earnings management caused by equity incentives, it is especially
important to control for firm performance and that is what the Kothari et al. (2005)
model does. The goal of the earnings management is to enhance firm performance, there
is a correlation between firm performance and management compensation (Ronen &
Yaari, 2008), therefore controlling for firm performance is important when using
earnings management in combination with equity incentives.
I use the linear model introduced by Kothari et al. (2005) rather than the non‐linear
Kothari et al. (2005) model, as there is no reliable way of matching at my disposal for this
sample. Such a method would for instance be available if one wants to measure earnings
management around a certain event, where one group experiences such an event and
another group of companies does not, it is than possible to match companies from the
first group with companies from the second group. This is however not possible in my
research design, therefore I will use a linear approach. The advantage of the non‐linear
model is that it mitigates the non‐linearity problem of the return on assets. Ye (2007)
however introduces another way to mitigate the non‐linearity problem in the return on
assets, he winsorizes the return on assets to normal values between ‐0.1 and 0.3. I use
69
the method Ye (2007) proposes, to mitigate the possible non‐linearity problem in return
on assets. I also, winsorize the accruals and incentive rations on first and ninety‐ninth
percentile to exclude effects of extreme performance. The model is estimated in a cross‐
sectional way because a time‐series model assumes an earnings management free
estimation period, which is not a realistic assumption for this research design. The two
digit sic codes are used to divide the companies into their industry groups in order to be
able to perform the cross‐sectional test. I use the two digit sic code instead of the full sic
code to obtain larger groups, because larger groups lead to better results. The hindsight
of having larger groups is that separating the firms into groups is done less precise: the
groups are larger therefore firms are put into groups with firms that can be quite
different.
I calculate the total accruals using cash flow information. Hribar and Collins (2002) show
in their study that this leads to better results around non‐articulation events. Total
accruals are calculated as net income minus cash flow from operations.
TAi,t = (NIi,t – CFOI,t)/ A i,t ‐1
TAi,t Total accruals calculated with cash flow

information of firm I at time T scaled by lagged
total assets
NI i,t Bottom line net income
CFO i,t Operating cash flows
A i,t ‐1 Total Assets
I use the total accruals to estimate the coefficients in the formula for nondiscretionary
accruals. I estimate the coefficients using the cross‐sectional linear version of the
Kothari et al. (2005) model:
TAi,t = 0 + 1 (1/Ai,t ‐1)+ 2( REVi,t ‐ ARi,t )/Ai,t‐1) + 3 (PPEi,t/ Ai,t‐1) + 4 ROAi,t‐1

REVi,t The change in revenue
ARi,t The change in accounts receivable
PPEi,t Gross property plant and equipment
Ai,t‐1 Total assets
With the estimated coefficients the non‐discretionary accruals can then be calculated:
NDAi,t= 0 + 1 (1/Ai,t ‐1)+ 2( REVi,t ‐ ARi,t )/Ai,t‐1) + 3 (PPEi,t/ Ai,t‐1) + 4 ROAi,t‐1
70
NDAi,t Non‐discretionary accruals scaled by lagged total assets
REVi,t The change in revenue
ARi,t The change in accounts receivable
PPEi,t Gross property plant and equipment
Ai,t‐1 Total lagged assets
Once the non‐discretionary accruals are calculated the discretionary accruals can be
derived:
DAi,t = TAi,t – NDAi,t
DAi,t Discretionary accruals scaled by lagged total assets

NDAi,t Non‐discretionary accruals scaled by lagged total
assets
The discretionary accruals are used as measure for earnings management and will be
compared with equity incentives to examine the relation between equity incentives and
earnings management and thereby answering the research questions.
7.3 Measure for equity incentives

In this master’s thesis I want to measure the relation between earnings management
and equity incentives. As measure for the incentive created by stock and option‐based
remuneration I use effect of a change in the stock price on total remuneration. This
measure is based on the measure used by Bergstresser and Philippon (2006), it is
however different than the measure they use.
The modification is that the option part of the formula, uses the estimated value of
options, were Bergstresser and Philippon (2006) use the number of options multiplied
by the share price. With this Bergstresser and Philippon (2006) implicitly assume that
the when the stock price changes the value of options change in the same way as the
value of stocks. This is however a quite rough assumption as the valuation of options is
more complex.
As explained in the first chapter of this master’s thesis, options create a short‐term
incentive for managers. This is due to the fact that options have an expiration date, after
this date they are valueless. Managers are triggered to maximize the value of options
71
before the expiration date. In this way they are able to sell or exercise the options for the
best price possible.
As measure for the effect of the change in the value of a CEO’s stock or options portfolio
due to a one percent change in the stock price, I use the following measure.
EQVALi,t = 0.01 × PRICEi,t × SHARESi,t + 0.01 × OPTIONVALi,t
EQVAL The equity value ratio measured as the dollar change in the value of a CEO’s stock and
options holdings that would come from a one‐percentage point increase in the
company stock price.
PRICE Company share price
SHARES Number of shares held by the executive
OPTIONVAL Value of options held by the executive + value realized on options exercise during the
year.
This equity value ratio differs from the ratio used by Bergstresser and Philippon (2006)
on one important point as this ratio uses the estimated value of options + the value
realized on option exercise during the year. Where Bergstresser and Philippon (2006)
use the number of options multiplied by the share price. It still measures the change in
value of the stock and option portfolio held by the executive due to a one percent change
in the firms’ stock price as the option value also contains the stock price of the firm. The
value realized on option exercise is added to the model, as this value is part of the
incentive for an executive to manage earnings during a year. The intrinsic value of the
options is derived from the “compustat execucomp” database and is: the estimated
intrinsic value of the unexercised exercisable options as reported by the company6. The
non‐exercisable options are not included in the model as they are irrelevant in the short
term. The options will automatically be included in the model from the year they
become exercisable, because from that moment they directly influence the executives
income.
This measure improves the Bergstresser and Philippon (2006) measure, as the change in
the value of the options is measured more precise. Because the method Bergstresser and
Philippon (2006) use is a rough estimate for the value of options.
6From 2006 onwards this measure is changed in the intrinsic value of exercisable options, calculated
based on the difference between the exercise price of the options and the close price of the
company’s primary issue of stock.
72
With the calculated equity value ratio the incentive ratio is calculated. This incentive
ratio represents the: “hypothetical CEO’s total compensation that would come from a
percentage point increase in the value of the equity of the firm” (Bergstresser &
Philippon, 2006).
INCENTIVE RATIO i,t = EQVAL i,t/(EQVAL i,t + SALARY i,t + BONUS i,t)
INCENTIVE RATIO The hypothetical CEO’s total compensation that would come from a one
percentage point increase in the value of the equity of the firm
EQVAL The equity value ratio measured as the dollar change in the value of a CEO’s
stock and options holdings that would come from a one‐percentage point
increase in the company stock price.
SALARY The fixed salary awarded to the executive
BONUS The cash bonus awarded to the executive
This incentive ratio is calculated in three different ways, the first one as described above
taking into account total equity incentives. The other two measures focus on the
influence of option or stock‐based equity incentives. In those models the EQVAL ratio in
the nominator is calculated using only the option part of the EQVAL formula or only the
stock part of the EQVAL formula. In that way the influence from the options‐ and stock‐
based incentives can be measured separately.
7.4 Estimating the relation between earnings management and equity incentives
To answer the research questions I examine the relation between earnings management
and equity incentives. Therefore the incentive ratios and the discretionary accruals are
compared in a regression. In order to provide a more detailed view of this relation I use
different regressions using data for the CEO, the CFO and both of them together. Each of
the three tests are carried out three times each; once with the total equity incentives
ratio, once the option incentive ratio and once with the stock incentive ratio.
DAi,t = α 0 + α1 × INCENTIVE RATIO + αx + Controls
DAi,t Absolute value of discretionary accruals scaled by lagged total assets

INCENTIVE RATIO The hypothetical CEO’s total compensation that would come from a one
percentage point increase in the value of the equity of the firm
Controls Control variables, discussed in descriptive statistics
7.5 Sample
In this section I discuss the sample selection. The sample consists of firms listed in the
United States that are part of the S&P 500 index. The sample period is 1999 until 2009, I
73
use this period because it covers important accounting scandals, years before these
scandals and the years after these scandals, I choose to take a longer period after the
scandals as I expect that it will take some time for the dust to settle after these events.
The companies need to meet the following requirements to be included in the sample:
‐ The firm is part of the S&P 500
‐ The firm does not have any missing values for the variables required
‐ Firm total assets are more than one million dollars.
‐ Firms are in industry groups of ten firms or more
The requirement that firm total assets have to be more than one million dollars is
because firm total assets are used as scaling variable. Therefore a small value may
generate unusual observations that are of little significance (Ye, 2007). The data to
estimate the accruals is derived from the comp stat database. Data about executive
remuneration is derived from the “execucomp” database.
7.6 Descriptive statistics

The sample used in this study contains information derived from the “compustat North
America” database and the “execucomp” database. The variables used for this study are
described in appendix 2.
For tests over the total research period running from 1993 until 2009 a total of 12587
firm years are included in the sample. For the period prior to the introduction of the
Sarbanes Oxley act 3168 firm years are included in the sample. In the period from 2003
until 2009 8346 firms years are included in the sample.
To identify the CEO’s in the database the variable “annual CEO flag” was used, this points
out which of the executives in the database was CEO of the firm in the specific firm year.
Such a variable is available as well for the function of CFO This variable however is not
suited for my research design as this variable is not introduced until 2006. Therefore the
CFO’s have to be identified in different way. To identify the CFO the variable “Title_Ann”
is used. This variable indicates the name of the function of the executive per year. To
identify the CFO’s all executives are classified as CFO if their title contains one of the
74
following words: CFO, chief financial officer, treasurer, controller, finance, vp‐finance
and administrative. This method is identical as the method used by Kim et al. (2011).
7.6.1 Calculation of discretionary accruals

Appendix 2a shows the descriptive statistics for the calculation of the discretionary
accruals. The discretionary accruals are calculated as described in section 7.2. First the
total accruals and independent variables are used in a regression to estimate the
coefficients. Then the coefficients are used in the formula to calculate non‐discretionary
accruals. With the non‐discretionary accruals the total accruals are calculated by
deducting the non‐discretionary accruals from the total accruals. The definitions of the
variables used to calculate discretionary accruals are included in appendix 2b.The
average of the total accruals is similar to the average of the total accruals found by
Bergstresser and Philippon (2006) in their study. They find a mean of 0,062 for the
absolute value of total accruals for a sample of companies with lagged 1996 assets above
1 billion dollars. Their standard deviation is lower than the standard deviation in my
sample: 0,054 compared to 0,075. This difference is probably due to the fact that the
Bergstresser and Philippon (2006) sample only contains firms with lagged 1996 assets
above 1 billion dollar; my sample contains firms with total assets above and below one
billion dollars. The study by Bergstresser and Philippon (2006) also includes a sample of
firms with total assets below 1 billion dollar, for that sample the average of the total
accruals are higher: 0,089 and the standard deviation is higher as well: 0,271. This could
explain a higher standard deviation for my sample because the sample I use contains
both firms with assets above 1 billion dollar and firms with assets below one billion
dollar. This is supported by Kothari et al. (2005), they state that accruals of small firms
are more volatile than accruals of large firms.
Compared to Jiang et al. (2010) the average of the total accruals in this master’s thesis
are a bit lower. Jiang et al. (2010) use a sample of S&P 1500 firms and they find an
average of the total accruals of 0,082 and a standard deviation of 0,078. Their higher
average of the total accruals could be explained by the fact that their S&P 1500 sample
contains more small firms than the S&P 500 sample I use; of the two samples
Bergstresser and Philippon (2006) use, the sample of smaller firms has higher
discretionary accruals as well.
75
The discretionary accruals I find are a bit lower than the average discretionary accruals
Bergstresser and Philippon (2006) find; 0,031 in my study compared to 0,041 in the
study by Bergstresser and Philippon (2006). A possible explanation for this difference is
that the discretionary accruals in this master’s thesis are estimated using a linear
Kothari model while Bergstresser and Philippon (2006) use the Jones and modified
Jones models. Another possible explanation is that Bergstresser and Philippon have a
different sample period; their sample runs from 1993 to 2001 and my sample runs from
1999 to 2009. Chart 1 shows that the average discretionary accruals in this master’s
thesis never reach the average value Bergstresser and Philippon (2006) find. The
average discretionary accruals Jiang et al. (2010) find are higher than the discretionary
accruals in this master’s thesis, Jiang et al. (2010) find average discretionary accruals of
0,072 compared to the average of 0,031 in my sample. This difference is probably due to
the fact their sample contains S&P 1500 firms, that means that their average firm will be
smaller than the average firm in my sample.
Chart 1
average absolute discretionary accruals
Chart 1 shows the development of the average value of absolute discretionary accruals
over the sample period used in this master’s thesis. It shows that the average absolute
value of discretionary accruals declines after 2002, the year in which the Sarbanes Oxley
act was introduced. It shows a peak again in 2008, in 2009 it returns to a value around
the normal value for the period 2003‐2007. The peak in 2008 could be explained by the
76
financial crises that started with the bankruptcy of Lehman Brother in September 2008.
Due to this crisis many firms had to revalue their assets and liabilities, this could explain
the relatively high discretionary accruals in 2008.
7.6.2 Descriptive statistics of the equity incentives

Appendix 3a and 3b give an overview of the yearly development of executive
remuneration in the sample. It contains the average and median value of the options and
shares held by the CEO and the CFO. It also shows the average cash remuneration. The
cash remuneration contains both salary and a bonus. The last three columns show the
average of the option, the stock and the total incentive ratio of each year.
The information in the tables shows that the incentives ratios are considerably higher
for CEO’s than for CFO’s. This also holds for the cash pay and the value of the stock and
options held. Equity incentives awarded to CEO’s are considerably higher than equity
incentives awarded to CFO’s, the cash payment of CEO’s is higher as well. The average
cash pay awarded to the CEO is about three times as high as the average cash pay
awarded to CFO’s, the difference in average and median options and stock held is even
larger. CEO cash pay steadily increases until 2005, in 2006 the average cash CEO
remuneration is somewhat lower and after 2006 the average CEO pay further declines.
The CFO cash pay shows a similar development until 2006 but the average CFO pay does
not decline after 2006. The levels of options held increase from 1999 to 2000, then they
decline a bit in 2001 and show a real dip in 2002. After 2002 the value of the options
held increases again until 2007, 2007 shows a decline in options followed by very strong
decline in 2008. In 2009 the value of options held recovers again. The value of the stock
held by both the CEO’s and the CFO’s follows a similar path as the value of options
though its development is more gradually.
7.6.3 Descriptive statistics of the discretionary accruals and equity incentives

Appendix 4 shows the descriptive statistics for the data used in the regressions between
discretionary accruals and equity incentives. There are six tables included in the
appendix. These tables show the descriptive statistics of the regressions between the
absolute discretionary accruals and the different equity incentive ratios. For both the
CEO and the CFO three tables are included in the appendix, these tables show the
descriptives of the statistics of three different samples: the total sample running from
1999 to 2009, the pre‐Sarbanes Oxley sample running from 1999 until 2001 and the
77
post‐Sarbanes Oxley sample running from 2003 until 2009. The tables show the
descriptive statistics of the different regressions between absolute discretionary
accruals and the equity incentive ratio as described in section 7.4. Following Jiang et al.
(2010) control variables for firm size, firm age, the leverage, the volatility of sales and a
firm year dummy are included in the regressions. To control for the size of firms the
natural logarithm of lagged total assets is used. Following the example of Jiang et al.
(2010) I add control variables for firm age using a dummy variable that is one if the
company is included in the “compustat” database for 20 years or more and is zero
otherwise. Firms are included in the “compustat” database from the moment they are
listed. This proxy for firm age is used because the real firm age is not available in the
“compustat” database. To control for the leverage of firms I use the total liabilities of a
firm divided by the total assets. As control variable for the volatility of sales I include the
standard deviation of the total revenue of the current firm year and the four previous
years. To control for the firm year effects 11 firm year dummy variables are included
that are 1 in a single sample year and zero otherwise. Controlling for the size of firms,
their age and the sales volatility indicates that the results are not due to volatile
environments in which firms that use a lot of equity based compensation might be in
(Bergstresser & Philippon, 2006). For instance starting firms are often more volatile and
have a good reason to use equity based compensation as they do not normally possess
large amounts of cash. Following Jiang et al. (2010) and Bergstresser and Philippon
(2006) the results for the firm age dummy variable are not included in the descriptive
statistics for the reason of brevity. The definitions of all variables are explained in
appendix 4g.
78
Chapter 8 Findings
8.1 Introduction
This chapter discusses the findings of the empirical study described in chapter 7 and
tries to answer the hypotheses discussed in chapter 6. I discuss the findings of the study
using the hypotheses presented in chapter 6 and I discuss my findings in relation to the
prior research as discussed in chapter 4.
The results of the study are presented in table 1 and 3. Appendix 4 contains the
descriptive statistics of the different regressions used to examine the relation between
discretionary accruals and earnings management. Appendix 4g contains the definitions
of the variables used in table 4 and 5. Table 1 presents the results of the regression
between equity incentives and earnings management for the CEO and table 2 presents
these results for the CFO. Both tables present the relation between equity incentives and
earnings management over the period 1999‐2009, the period 1999‐2001 (pre‐Sarbanes
Oxley act) and the period 2003‐2009 (post‐Sarbanes Oxley act). The year 2002 is left out
of both periods, as this is the year the Sarbanes Oxley act was accepted. 2002 is
therefore expected to be a year with many changes in accounting policies that could
contaminate the tests. The results for the different periods are presented for the three
different incentive ratios. The first is the incentive ratio represents the total
(stock+option) equity incentive ratio, the second represents the option‐based equity
incentive ratio and the third the stock‐based equity incentive ratio.
79
Table 1
Results regression CEO equity incentives and earnings management
The relation between CEO equity incentives and earning management. Measured over three different periods; a sample from 1999 to 2009 a pre‐SOX
sample running from 1999 until 2001 and a post‐SOX sample running from 2003 until 2009. For all periods three regressions are performed, absolute
discretionary accruals are used as dependent variable against a total incentive ratio an option incentive ratio and a stock incentive ratio. The t values are
reported in parenthesis. The 10%, 5% and 1% significance levels are indicated by respectively *,**,***. The variables are defined in appendix 4g.
Model: Discretionary accruals = α0 + α1 × incentive ratio × Controls
Total period 1999‐2009 Pre‐SOX 1999‐2001 Post‐SOX 2003‐2009
Total Option Stock Total Option Stock Total Option Stock

incentive incentive incentive incentive incentive incentive incentive incentive incentive
variables ratio ratio ratio ratio ratio ratio ratio ratio ratio
Inc. ratio 0,003* 0,014*** 0,000 0,011*** 0,032*** 0,002 0,000 0,005 ‐0,001
(1,89) (5,27) (‐0,22) (3,35) (6,61) (0,47) 0,01 (1,46) (‐530)
Oldfirm ‐0,010*** ‐0,010*** ‐0.011*** ‐0.016*** ‐0,015*** ‐0,018*** ‐0,007*** ‐0,006*** ‐0,007***
(‐8,50) (‐8,08) (‐8,87) (‐5,96) (‐5,58) (‐6,63) (‐4,95) (‐4,77) (‐5,07)
Size ‐0,002*** ‐0,002*** ‐0,001*** ‐0,003*** ‐0,003*** ‐0,003*** ‐0,001*** ‐0,001*** ‐.001***
(‐4,25) (‐4,50) (‐4,09) (‐4,28) (‐4,61) (‐4,15) (‐2,90) (‐3,03) (‐2,83)
Leverage ‐0,002*** ‐0,004** ‐0,004*** ‐0,001 ‐0,001 ‐0,001 ‐0,005*** ‐0,005*** ‐0.006***
(‐2,46) (‐2,43) (‐2,46) (‐0,26) (‐0,52) (‐0,45) (‐2.78) (‐2,68) (‐2,83)
Sales vol. 0,008*** 0,008*** 0,008*** 0,002 0,001 0,003 0,009*** 0,009*** 0,009***
(4,09) (3,95) (4,16) (0,52) (0,30) (0,71) (3,37) (3,34) (3,37)
Year dum. Yes Yes Yes Yes Yes Yes Yes Yes Yes
Constant 0,064*** 0.065*** 0,064*** 0,101*** 0,104*** 0,102*** 0,066*** 0,066*** 0,065***
(9,66) (9,80) (9,67) (7,87) (8,18) 8,06 (8,46) (8,50) (8,43
Adj. R2 0,038 0,042 0,038 0,061 0,070 0,046 0,024 0,024 0,024
N 6042 6042 6042 1675 1675 1675 3815 3815 3815

Table 2
Results regression CFO equity incentives and earnings management
The relation between CFO equity incentives and earning management. Measured over three different periods; a sample from 1999 to 2009 a pre‐SOX
sample running from 1999 until 2001 and a post‐SOX sample running from 2003 until 2009. For all periods three regressions are performed, absolute
discretionary accruals are used as dependent variable against a total incentive ratio an option incentive ratio and a stock incentive ratio. The t values are
reported in parenthesis. The 10%, 5% and 1% significance levels are indicated by respectively *,**,***. The variables are defined in appendix 4g.
Model: Discretionary accruals = α0 + α1 × incentive ratio × Controls
Total period 1999‐2009 Pre‐SOX 1999‐2001 Post‐SOX 2003‐2009
Total Option Stock Total Option Stock Total Option Stock

incentive incentive incentive incentive incentive incentive incentive incentive incentive
variables ratio ratio ratio ratio ratio ratio ratio ratio ratio
Inc. ratio 0,014*** 0,031*** 0,002 0,038*** 0,069*** 0,010 0,003 0,013** ‐0.002
(3,81) (5,84) (0,46) (5,11) (6,69) (0,94) (0,75) (2,03) (‐0,36)
Oldfirm ‐0.009*** ‐0,009*** ‐0,010*** ‐0,017*** ‐0,016*** ‐0,020*** ‐0,005*** ‐0,005*** ‐0,006***
(‐8,31) (‐7,99) (‐8,69) (‐6,04) (‐5,72) (‐7,06) (‐4,42) (‐4,29) (‐4,48)
Size ‐0,002*** ‐0,002*** ‐0,002*** ‐0,004*** ‐0,004*** ‐0,004*** ‐0,002*** ‐0,002*** ‐0,002***
(‐5,72) (‐5,59) (‐5,49) (‐4,88) (4,89) (‐4,74) (‐4,27) (‐4,28) (‐4,13)
Leverage ‐0,003** ‐0,003** ‐0,004*** 0,002 0,002 0,001 ‐0,005*** ‐0,005*** ‐0,005***
(‐2,29) (‐2,13) ‐2,47 (0,53) (0,73) 0,35 (‐3,00) (‐2,90) (‐3,06)
Sales vol. 0,009*** 0,008*** 0,009*** 0,003 0,002 0,003 0,010*** 0,010*** 0,010***
(4,46) (4,38) (4,41) (0,90) (0,62) (0,84) (3,82) (3,81) (3,81)
Year dum. Yes Yes Yes Yes Yes Yes Yes Yes Yes
Constant 0,073*** 0,071*** 0,073*** 0,114*** 0,113*** 0,119*** 0,063*** 0,062*** 0,063***
(11,82) (11,51) (11,93) (8,06) (8,05) (8,30) (8,85) (8,76) (8,78)
Adj. R2 0,041 0,044 0,039 0,089 0,101 0,074 0,026 0,027 0,026
N 6545 6545 6545 1493 1493 1493 4531 4531 4531
81
8.2 Hypothesis 1
The main hypothesis of this thesis is:
H1: There is a positive relation between earnings management and equity incentives.
This hypothesis is tested using the following equation:
DAi,t = α 0 + α1 ×Total INCENTIVE RATIO + Controls
This relation is tested for both the CEO and the CFO over a sample period running from
1999 until 2009. The descriptive statistics of the regressions are presented in appendix
4a for the CEO and 4d for the CFO.
The results of the empirical study presented in table 1 and 2 show a significant positive
relation between discretionary accruals calculated with the linear Kothari model and the
total equity incentive ratio. This relation is positive for both the CEO and the CFO,
however the coefficient and the explanatory value of the CFO equity incentives are
higher. The relation between absolute discretionary accruals total CEO equity incentives
is rather weak and only significant at the 10% level. Because I use a 5% significance
level I do not classify this as a significant positive relation. These findings can be
explained by prior research discussed in chapter 5. Bergstresser and Philippon (2006)
show a significant positive relation between CEO equity incentives and discretionary
accruals calculated with the Jones model and modified Jones model over the period
1994‐2000. They find higher coefficients and the explanatory value for CEO equity
incentives and equity incentives they find is much higher than the explanatory value
found in this master’s thesis. This is possibly due to the different period used, as they
measure the relation between their incentive ratio and discretionary accruals in the
period prior to the major accounting scandals of the early 2000’s. The descriptive
statistics discussed in chapter 7 show that the discretionary accruals over the sample
period used by Bergstresser and Philippon (2006) are higher than the discretionary
accruals in the sample period I used. That indicates that the use of discretionary accruals
has declined. This is supported by the findings of Jiang et al. (2010). They find no
positive relation between discretionary accruals and earnings management for a post‐
Sarbanes Oxley act sample, I find similar results to Jiang et al. (2010) for a post‐Sarbanes
Oxley act sample, these results are discussed in section 8.5 The fact that my total sample
period contains this post‐Sarbanes Oxley act period could explain the fact that I do not
find a 5% significant relation between absolute discretionary accruals and CEO equity
incentives but that Bergstresser and Philippon (2006) do find a positive relation.
Descriptive statistics in appendix 2 and 4 show that discretionary accrual levels are
lower than the discretionary accruals Bergstresser and Philippon find and that
discretionary accruals are lower in the post‐Sarbanes Oxley act sample than in the pre‐
Sarbanes Oxley act sample. This indicates that earnings management declines over the
sample period I use. Hypothesis 1 is only partially accepted, the results show a
significant positive relation between absolute discretionary accruals and the total CFO
equity incentive ratio over the total sample period. I find no significant positive relation
between absolute discretionary accruals and the total CEO equity incentives over the
period 1999‐2009.
8.3 Hypothesis 2
The second hypothesis examined in this thesis is:
H2: The positive relation between earnings management and equity incentives is stronger
for CFO’s than for CEO’s.
The second hypothesis is tested with the following equation:
DAi,t = α 0 + α1 × INCENTIVE RATIO + Controls
Different regressions are done with this formula, using both CEO and CFO data to be able
to compare the differences between CEO and CFO’s. Regressions are performed for three
sample periods, the pre‐Sarbanes Oxley act period from 1999 until 2001, the post‐
Sarbanes Oxley period from 2003 until 2009 and the total period from 1999 until 2009.
As the hypothesis states, the relation between CFO equity incentives and earnings
management is expected to be stronger than the relation between CEO equity incentives
and earnings management. This relation is tested over three different periods, in two of
those periods a significant relation is found between CFO equity incentives and earnings
management. These are the total research period running from 1999 until 2009, and the
period prior to the introduction of the Sarbanes Oxley act running from 1999 until 2001.
For CEO total equity incentives I only find a significant positive relation in the pre‐
Sarbanes Oxley act period. In the pre‐Sarbanes Oxley act period I find a higher
coefficient, higher t‐values and a higher adjusted R2 for the CFO than for the CEO equity
83
incentives; this is shown in table 1 and 2. In the period before the introduction of the
Sarbanes Oxley act, the coefficient for the CFO is almost 3,5 times higher than the
coefficient for the CFO.
Table 1 and 2 show a significant positive relation between absolute discretionary

accruals and the total CFO equity incentives over the total research period, however this
relation for the CEO is not significant at the 5% level. The second main columns of table
1 and 2 present the same relations for the period before the introduction of the
Sarbanes Oxley act. The explanatory value in these relations is higher than explanatory
value over the total period. In this period the coefficient and the explanatory value for
the CFO are again higher than those for the CEO. This implies that for the CFO, the model
explains more of the earnings management than for the CEO.
Table 1 and 2 also show the relation between the discretionary accruals and option‐
based equity incentives. I find a significant positive relation between discretionary
accruals and the option‐based equity incentive. These relations have again higher
coefficients and higher explanatory value for the CFO than for the CEO. This is despite
the fact that the different incentive ratios of the CFO are much lower than the average
incentive ratios of the CEO’s, as can be seen in the tables included in appendix 2. The
results of this study indicate that these lower CFO incentives have a bigger influence on
accrual accounting than the higher CEO incentives. This could be due to the fact that a
CFO has a more direct influence on the accounting decisions than the CEO.
These findings confirm the hypothesis that the positive relation between equity
incentives and earnings management in this model is stronger for CFO equity incentives
than for CEO equity incentives. These findings are in line with the findings of Jiang et al.
(2010) who examine the relation between discretionary accruals calculated with the
forward looking discretionary accrual model and equity incentives calculated with the
incentive ratio as discussed in section 4.3. They also find that the CFO coefficient is
about three times larger than the coefficient for the CEO in the period before the
introduction of the Sarbanes Oxley act. The results found by Kim et al. (2011) point in the
same direction as my results, they measure the relation between equity incentives and
firm crash risk and find evidence for a positive relation between CFO equity incentives
and crash risk. They find only a weak relation between CEO equity incentives and firm
crash risk, similar to my results as I only find a positive relation between discretionary
84
accruals and CEO equity incentives in the pre‐Sarbanes Oxley act period, but no 5%
significant relation in the total research period.
8.4 Hypothesis 3
The third research question tested in this thesis is the following:
H3: The relation between equity incentives and earnings management is stronger for
option‐based incentives than for share‐based incentives.
This hypothesis is tested with the following two formulas:
DAi,t = α 0 + α1 × Option INCENTIVE RATIO + Controls
And
DAi,t = α 0 + α1 × Stock INCENTIVE RATIO + Controls
The first equation measures the relation between discretionary accruals and option‐
based equity incentives. The second equation measures the relation between
discretionary accruals and stock‐based equity incentives. These relations are estimated
over the three sample periods for both the CEO and the CFO.
This hypothesis comes from the expectation that short‐term option‐based equity
incentives have more influence on accruals than longer term share‐based equity
incentives. The results of my study show a significant relation between earnings
management and option‐based equity incentives. This relation is significant for both
CEO and CFO equity incentives. These findings apply for the total research period and
the period before the introduction of the Sarbanes Oxley act. For stock‐based equity
incentives no significant results are found with this study, as can be seen in table 1 and
2.
Table 1 and 2 show that the relation between discretionary accruals and stock‐based
equity incentives is never significant, however the relation for option‐based equity
incentives is highly significant in all four tests. This means that this study shows a
relation between discretionary accruals calculated with the linear Kothari model and the
option‐based equity incentive ratio. The study shows no such relation for the stock‐
based equity incentive ratio. It therefore supports the hypothesis that the relation
between option‐based equity incentives and earnings management is stronger than the
85
relation between share‐based equity incentives and earnings management. These
results are similar to the results found by Kim et al. (2011). Kim et al. (2011) test the
relation between CEO and CFO equity incentives and crash risk. They use the
measurements by Bergstresser and Philippon (2006) to measure the equity incentives;
this ratio is discussed in section 4.3. Kim et al. (2011) find significant positive relations
between option‐based equity incentives and crash risk for both the CEO and the CFO but
they do not find a significant relation between stock‐based equity incentives and crash
risk. The results of this master’s thesis suggest that the findings of Kim et al. (2011) for
crash risk also hold for earnings management measured with discretionary accruals.
8.5 Hypothesis 4
The last research question examined in this master’s thesis is the following:
H4: The positive relation between equity incentives and earnings management declines
after the major accounting scandals of the early 2000’s, this decline is stronger for option‐
based incentives.
To test the fourth hypothesis the following regression is used:
DAi,t = α 0 + α1 × INCENTIVE RATIO + Controls
Regressions are made with all three incentive ratios (the total incentive ratio, the
option‐based incentive ratio and the stock‐based incentive ratio) for both the CEO and
the CFO.
This fourth hypothesis is based on the influence of major accounting scandals at the
start of the early 2000’s and the reaction on these scandals; these circumstances have
possibly changed the environment where companies operate. As discussed in the
previous couple of sections the results of this study show a significant positive relation
between discretionary accruals and the equity incentive ratio for the period before the
introduction of Sarbanes Oxley and for CFO equity incentives over total research period
from 1999 to 2009 as well. For the period 2003‐2009 however, this study only finds a
5% significant relation between discretionary accruals and CFO option‐based equity
incentives. The third main columns in table 1 and 2 show the results of the test for the
period 2003‐2009.
86
These results suggest that the relation between discretionary accruals and equity
incentives changes between the periods 1999‐2001 and 2003‐2009. The model finds a
highly significant positive relation between discretionary accruals and all types of equity
incentives for both CEO and CFO in the first period but only finds a significant relation
between discretionary accruals and CFO option‐based equity incentives in the second
period.
Jiang et al. (2010) do not find a positive relation between discretionary accruals and
equity incentives for the post‐Sarbanes Oxley period either. They however do find a
negative relation between discretionary accruals and CFO equity incentives in this
period. Jiang et al. (2010) state that a possible explanation for this negative relation is
that after the Sarbanes Oxley act: “CFO’s believe investors penalize earnings
management by executives with high equity incentives”. This negative relation could be
explained by the model they use as well, because the incentive ratio used by Jiang et al.
(2010) contains both exercisable as un‐exercisable options. This could lead to a
situation where the equity incentive ratio is high but the executive is not able to cash
this value as the options have not‐yet vested. Therefore they could be motivated to
manage earnings downwards over that period, to be able to manage earnings upwards
again once these options vest.
Besides the effects of the accounting scandals of the early 2000’s, another possible
reason for the change in the relation between equity incentives and earnings
management is the financial crises that started in 2008. The tables in appendix 2 show
that the value of option and stock held by executives plummets in 2008, therefore the
equity incentive ratios are rather low in 2008 and 2009. However it is imaginable that
managers manage earnings upwards in these years to smooth their earnings or just to
limit the damage of the financial crises. Chart 1 shows a peak in discretionary accruals in
the year 2008, this in combination with low equity incentives could explain the findings
by Jiang et al. (2010). A reason that Jiang et al. (2010) find a negative relation for CEO
equity incentives and I don’t find a relation can be that Jiang et al. (2010) include the
year 2002 in their post‐Sarbanes Oxley act sample. This year is like 2008 a year with
relatively high discretionary accruals and relatively low equity incentives. My sample
contains only one such year (2008), the sample by Jiang et al. (2010) contains two of
87
those years (2002 and 2008), that could be the reason Jiang et al. (2010) find a negative
relation over their sample period and I do not find a relation of my sample period.
One would possibly expect that the use of short‐term option‐based equity incentives
would decline after the accounting scandals of the start of the 21th century, this could
also be a possible explanation for the fact that the relation between discretionary
accruals and equity incentives changes after 2002. However the tables in appendix 4
show that the value of the options held by the managers does not substantially decline in
this sample until 2008 when the financial crisis starts. This is therefore not a likely
explanation for the change in this relation in my model.
The second part of the hypothesis states that: it is expected that the relation between
equity incentives and earnings management declines more for the CFO than for the CEO.
The study however does not present strong evidence for this expectation. It only finds
one significant relation for the post‐Sarbanes Oxley act period 2003‐2009. This is the
relation between CFO option‐based equity incentives and earnings management. This
could indicate that the relation between CFO option‐based equity incentives and
earnings management is stronger in this period than the relation between CEO option‐
based equity incentives and earnings management. The tests do not provide information
about the relation between total and stock‐based equity incentive ratios and earnings
management as I do not find significant results for these tests.
8.6 Summary
The goal of this master’s thesis is to provide more detailed information about the
relation between equity incentives and earnings management. This chapter shows the
findings of my study. It shows that; for my study with this sample and this model there is
a significant positive relation between CFO equity incentives and earnings management
over the total research period. I do not find a 5% significant relation between total CEO
equity incentives and earnings management over the total research period. It shows that
there is a positive relation for both the total equity incentive ratio as for the option‐
based incentive ratio, the relation for the total equity incentive ratio over the total
research period is however only significant for CFO equity incentives. This study does
not find a positive relation for the stock‐based incentive ratio. It also shows a higher
coefficient in the relation between discretionary accruals and equity incentives for CFO
equity incentives than for CEO equity incentives. This study finds a positive relation
88
between discretionary accruals and equity incentives in the period before the
introduction of the Sarbanes Oxley act. For the total research period from 1999 to 2009
it only shows a positive relation for CFO equity incentives. For the period after the
introduction of the Sarbanes Oxley act this study only finds a positive relation between
CFO option‐based equity incentives and earnings management.
89
Chapter 9 limitations
This chapter discusses several limitations of the study; these are due to data availability
and choices made in the research design. It is important to take these limitations into
account when interpreting the results of this study.
One of the limitations in this research is caused by the sample used. The sample consists
of listed US companies that are part of the S&P 500. The S&P 500 contains the largest
listed companies in the United States. This means that the results found with this sample
do not necessarily hold for companies with different characteristics than these S&P 500
firms. This can be because the firms in this sample are the largest companies in the
United States. It is imaginable that these companies are more in the spotlight than other
companies. That could be a reason for the management to adopt behavior in order to
maintain the reputation of a company. The situation could however be quite different
for smaller firms which are less in the spotlight and therefore it might be possible that
other results are found when testing a different kind of sample.
Another possible limitation one has to consider is the sample period. The period from
1999 to 2009 has been quite eventful from an accounting point of view. First there were
the accounting scandals of the early 2000’s followed by a financial crisis several years
later. This all led to quite some changes in financial reporting and possibly in the
behavior of executives as well, as financial reporting was suddenly under public
scrutiny. It could be that when these tests are performed for a different sample of firm
years one may find different results because each period has its own characteristics that
could influence the behavior of CEO’s and CFO’s.
A limitation of the model is that the incentive ratio used in this study splits executive
remuneration in options, shares, salary and bonus. It does take into account the
difference between vested and not vested options but it does not take the details of all
the different remuneration contracts into account. This is a limitation to the model as
the behavior of executives is influenced by their specific remuneration; these
remuneration plans could contain certain aspects that are not taken into account by this
model. Examples of that are possible restrictions on shares awarded to management, or
specific targets that have to be met to receive a bonus.
90
Another possible limitation that has to be considered is the use of the “execucomp”
database. Until 2005 the “execucomp” database only reports the five highest paid
managers, these five do not necessarily include the CFO, this could lead to a possible
sample selection, as companies where the CFO is not among the five highest paid
executives might have a CFO who is less influential than companies were the CFO is
among the five highest paid executives. These possibly less influential CFO’s are not
included in my sample.
Appendix 2 shows the average value of shares and the average value of options held by
executives. These values are rather high, certainly when compared to the mean values of
these observations. These high average values are possible distorted because of a
number of executives that have extremely high amounts of shares and options, some
having multiple billion dollars’ worth of shares and options of their companies. These
observations lead to the enormous difference between the average and the median of
these values. These observations are however excluded from my actual tests, as the
incentive ratios are winsorized on first and 99th percentile. These high amounts of
shares and options are normally caused by executives that were the founders of those
companies and therefore have large amounts of shares in their firms.
Another limitation of the study is a possible bias created by omitted variables. I use a
number of control variables for firm age, leverage, sales volatility, firm size and for firm
year effects. It is possible that other variables might have an influence as well. Jiang et al.
(2010) for example use more control variables, like firm corporate governance, industry
dummies and the volatility of cash flows.
Finally when interpreting the results of this master’s thesis one should consider that
these results were found with this particular research design on this particular sample.
Results could differ when one uses another research model or another sample.
Therefore one should be careful generalizing this model. When the research design was
made choices had to be made about what models would be used. These choices could
influence the outcome of the study. For instance a linear version of the Kothari model
has been used to estimate discretionary accruals. This model estimates discretionary
accruals and is widely used, that however does not imply that this estimation of
discretionary accruals is perfect. It could be that results are different in another model.
It is therefore important to be careful when generalizing the results of this study. Taking
91
into account these limitations the results of this study do provide evidence for a positive
relation between discretionary accruals and equity incentives for a sample of large,
listed US firms. These results are supported by findings in prior literature.
92
Chapter 10 conclusion
Summary and main conclusions
This master’s thesis provides more detailed information on the relation between
earnings management and equity incentives, for a sample of large US listed firms. This
insight can help to understand the reaction of managers to their remuneration
incentives and is useful when future remuneration plans are designed.
This master’s thesis shows a relation between discretionary accruals and equity
incentives. It gives this relation for total equity incentives and for option‐based equity
incentives, the relation for option‐based equity incentives is stronger than for total
equity incentives. The master’s thesis show no significant relation for total equity
incentives awarded to the CEO over the total research period, it does show this relation
for total equity incentives awarded to the CFO. It also finds that the relation between
discretionary accruals and earnings management holds for both the CEO and the CFO in
the pre‐Sarbanes Oxley act period and that this relation is stronger for CFO equity
incentives. This study shows that the positive relation found over the total research
period and over the period from 1999 to 2001 changes for the period 2003 to 2009 as
for this period the tests find only a significant relation between discretionary accruals
and CFO option‐based equity incentives. These findings confirm prior research that
there is a positive relation between earnings management and equity incentives before
the introduction of the Sarbanes Oxley act, and that this relation changes after the
introduction of the Sarbanes Oxley act. Lastly it results in the finding that there is a
significant relation between option‐based equity incentives and earnings management
while the study finds no such relation between stock‐based incentives and earnings
management.
Recommendations
The results of this study lead to some questions that could be examined further in the
future.
This study focuses on option and stock‐based equity incentives in general, it shows that
there is a difference between stock‐ and option‐based equity incentives but it does not
pay attention to the specific details of the different forms of remuneration, there are
however many different sorts of option‐ and stock‐based incentives and there are other
93
forms of remuneration as well. It is useful to take a closer look at the effects of the
different characteristics of specific forms of remuneration and to take more details into
account as these might influence the decisions managers make.
Another interesting aspect for further research could be the connection in the relation
between CEO equity incentives and CFO equity incentives. To fully understand the effect
of equity incentives it is important to understand how the decision making process in
the board works. It could be that different effects are found in cases where for example
CEO does receive equity incentives but the CFO doesn’t, or where the one receives
different kind of equity incentives than the other executives. Examining the coherence in
the relations between the CEO and the CFO can provide further information on the
effects of equity incentives.
94
Bibliography
Bartov, E., & Cohen, D. A. (2008). The "Numbers Game" in the Pre‐and Post‐Sarbanes‐Oxley Eras. Journal of
Accounting, Auditing & Finance, 24(4), 505‐534.
Bentham. (1789). An Introduction to the principles of morals and legislation.
Bergstresser, D., & Philippon, T. (2006). CEO incentives and earnings management. journal of financial
economics, 80, 511‐529.
Bernard, V. L., & Skinner, D. J. (1996). What movivates managers' choice of discretionary accruals? Journal
of Accounting and Economics, 22, 313‐325.
Burns, N., & Kedia, S. (2006). The impact of performance‐based compensation on misreporting. Journal of
Financial Economics, 79, 35‐67.
Cheng, Q., & Warfield, T. D. (2005). Equity incentives and Earnings Management. The accounting review,
80(2), 441‐476.
Core, J., & Guay, W. (2002). Estimating the value of employee stock option protfolios and their sensitivities
to price and volatility. Journal of Accounting Research, 40, 613‐630.
DeAngelo, L. E. (1986). Accounting numbers as Market Valuation Substitutes: a study of management
buyouts of public stockholders. The Accounting Review, 61, 400‐420.
Dechow, P. M., & Skinner, D. J. (2000). Earnings management reconciling the views of accounting
academics, practitioners, and regulators. Accounting Horizons, 14(2), 235‐250.
Dechow, P. M., Hutton, A. P., Kim, J. H., & Sloan, R. G. (2012). Detecting Earnings Management: A new
approach. Journal of Accounting Research, 50(2).
Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995). Detecting earnings managment. The Accounting
Review, 70, 193‐225.
Dechow, p., & Dichev, I. (2002). The quality of accruals and earnings: The role of accrual estimation errors.
The Accounting Review, 77(supplement), 35‐59.
Dechow, p., Richardson, S., & Tuna, I. (2003). Why are earnings kinky? An examination of the earnings
management explanation. Review of accounting studies, 8, 355‐384.
Gerakos, J. (2012). Discussion of detecting earnings management: a new approach. Journal of accounting
Research, 50(2).
Healy, P. M. (1985). The effect of bonus schemes on accounting decisions. Journal of Accounting and
Economics, 7, 85‐107.
Healy, P. M., & Wahlen, J. (1999, December). A review of of the earnings management literature and
its implications for standard setting. Accounting Horizons, 365‐383.
Holthausen, R. W., Larcker, D. F., & Sloan, R. G. (1995). Annual bonus schemes and the manipulation of
earnings. Journal of Accounting and Economics, 19, 29‐74.
Hribar, P., & Collins, D. (2002). Errors in estimating accruals: implications for emperical research. Journal
of Accounting Research. 40, 105‐134.
Jensen, M. C. (1991). Corporate control and the politcs of finance. Journal of applied corporate finance, 4(2),
13‐33.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: managerial behavior, agency costs, and
ownership structure. Journal of financial economics, 3, 305‐360.
Jiang, J. X., Petroni, K. R., & Wang, I. Y. (2010). CFOs and CEOs: Who have the most influence on earnings
management? Journal of Financial Economics, 96, 513‐526.
Jones, J. (1991). Earnings management during import relief investigations. Journal of Accounting Research,
29, 193‐228.
Kasanen, E., Kinunnen, J., & Niskanen, J. (1996). Dividend‐based earnings management:
Empirical evidence from Finland. Journal of Accounting and Economics, 22, 283‐312.
Katz, D. M. (2006, september 6). IRS Chief: CFO pay should be fixed. Retrieved may 2012, from
www.CFO.com: http://www.cfo.com/article.cfm/7880243/c_7932402
Kim, J. B., Li, Y., & Zhang, L. (2011). CFO's versus CEO's: equity incentives and crashes. journal of financial
economics, 101, 713‐730.
Koh, K., Matsumoto, D., & Rajgopal, S. (2006). Meeting or beating analyst expectations in the post‐
scandals world: changes in stock market rewards and managerial actions. Working papar
university of
Washington.
95
Kothari, S., Leone, A., & Wasley, C. (2005). Performance Matched Discretionary Accrual Measures.
Journal of Accounting & Economics, 39(1), 163‐197.
Larcker, D., & Richardson, S. (2004). Fees paid to audit firms, accrual choices and corporate governance.
Journal of Accoutning Research, 42(3), 625‐657.
McNichols, M. F. (2002). The quality of accruals and earnings: The role of accrual estimation errors. The
Accounting Review, 77 suplement, 61‐69.
Mehran, H. (1995). Executive compensation sturcture, ownership and firm performance. Journal of
Financial Economics, 38, 163‐184.
Mill, J. S. (1863). Utilitarianism.
Mohanram, P. S. (2003). How to manage Earnings Management? Accounting World,october 2003 1‐11.
Otomasa, S. (2000). Cut in bonus and managerial accounting discretion. Accounting(158), 43‐54.
PWC. (2011). Global Equity Incentives Survey. Retrieved 07 19, 2012, from www.pwc.com:
http://www.pwc.com/en_US/us/hr‐
management/assets/2011_Global_Equity_Incentives_Survey_fv.pdf.
Ronen, J., & Yaari, V. (2008). Earnings Management, Emerging insights inThory, Practice and Research.
Springer.
Shuto, A. (2007). Executive compensation and earnigns management: Emperical evidence from Japan.
Journal of International Accounting, Auditing and Taxation, 16, 1‐26.
Smith, A. (1776). Wealth of nations.
Stolowy, H., & Breton, G. (2004). Accounts manipulation: a literature review and proposed
conceptual framework. Revieuw of accounting & finance, 3(1), 5‐66.
Watts, R., & Zimmerman, J. (1986). Positive Accounting theory. Englewood‐Cliffs, USA: prentice‐hall.
Ye, J. (2007). Accounting accruals and tests of earnings management. working paper Baruch college.
96
Appendix 1
Literature review summary

Year Author Object of study Sample Methodology Outcome
1985 Healy The effect of bonus Sample: listed firms Using his own Healy model, Healy finds evidence that managers
schemes on earnings from fortune directory comparing total accruals as proxy manipulate earnings downwards when
management of 250 biggest US for earnings management with the the results are under the lower bound
industrial firms place of the firm’s performance in or above the upper bound of the bonus
running from; 1930‐ the bonus scheme: under, in the scheme. This proves his bonus‐
1980 margin or above the estimated maximizing hypothesis.
boundaries of their bonus schemes.
1995 Holthausen, Annual bonus Sample of firms from Discretionairy accruals are Holthausen finds evidence for earnings
Larcker and schemes and the confidential estimated with the Jones model. He maximizing at the upper bound but
Sloan manipulation of compensation compares these with observations unlike Healy he finds no evidence for
earnings databases provided by assigned to categories (below lower managing earnings downwards under
two different human bound, above upper bound or in‐ the lower bound of the bonus. This
resource consulting between) of the bonus scheme. supports the income‐smoothing
firms from 1982‐1984 Because of his confidential data he hypothesis instead of the bonus
and 1987‐1981 has detailed information about the maximizing theory. He shows that
real boundaries of the bonus Healy’s results here are likely to be
schemes. caused by his methodology.
2006 Bergstresser and CEO equity Compustat dataset of Measuring earnings management Find significant evidence that earnings
Philippon incentives and publicly held using Accruals. A version of the management is positively related to
earnings corporations over a Jones model and the modified Jones CEO stock ownership. And in periods
management period from 1994‐2000 model are used to calculate the of high accruals there is more
Measure of CEO Accruals. And management execution of options and selling of
incentives and incentives with a formula to shares by top management.
measures of CEO measuring the dollar change in the
option exercises are value of a CEO’s stock and options
from the Executive portfolio that would result from a
Compensation one percentage point increase in the
database. stock price.
2005 Cheng and CEO equity Execucomp dataset Earnings management as the CEO’s with high equity incentives use
Warfield incentives and US‐based publicly held likelihood of meeting and beating more earnings management. CEO’s
earnings firms existing of 9472 analyst forecast and they use with less persistent equity incentives
management firm years over the abnormal accruals calculated with are more likely to report surprising
period 1993‐2000. cross‐sectional Jones model earnings. CEO share sales increase in
Only firms with data the period after the earnings are
on stock and option‐ presented in periods with high income
based remuneration. increasing accruals.
2010 Jiang, Petroni CFO and CEO equity US‐based S&P 1500 Two measures for earnings A positve relation between both CFO
and Wang incentives and firms that are part of management, beating the analyst and CEO equity incentives pre‐
earnings the execucomp forecasts and the forward‐looking Sarbanes Oxley. Stronger relation for
management. database and have Jones accrual model. As measure for the CFO. Post‐Sarbanes Oxley there is
Comparing the data available on CEO equity incentives they use the no positive relation between equity
relation between and CFO incentive ratio by Bergstresser and incentives and earning management,
CFO and CEO and compensation. Sample Philippon. there is a negative relation between
examining pre and runs 1993‐2006 CFO equity incentives and earnings
post‐Sarbanes Oxley management.
2011 Kim, Li and CEO and CFO equity Publicly held US‐based As measure for equity incentives the CEO and CFO stock‐based incentives
Zhang incentives and crash firms from the incentive ratio by Bergstresser and are not linked with future crash risk in
risk. Examine the executive Philippon is used. This measure is the 1993‐2003 period. Option‐based
difference between compensation split to measure stock and option‐ incentives are positively related with
option and stock‐ database. The sample based incentives separately. No future crash risk over this period. The
based incentives. runs from 1993‐2009. accrual model used the equity relation for the CFO is much stronger
Only firms with data incentives are compared with crash than the relation for the CEO.
on CEO and CFO risk.
compensation
available. Total of
29638 firm year
observations.
98
Appendix 2
Appendix 2a
Descriptive statistics calculation discretionary accruals
Accrual model: TAi,t= 0 + 1 (1/Ai,t ‐1)+ 2(( REVi,t ‐ ARi,t )/Ai,t‐1) + 3 (PPEi,t/ Ai,t‐1) + 4 ROAi,t‐1
Variables N Mean Median Std. Dev. Min Max
Total accruals/Ai,t ‐1 6514 ‐0,063 ‐0,052 0,075 ‐0,359 0,173
Independent variables
1/ Ai,t ‐1 6514 4,33E‐10 1,72E‐10 9,90E‐10 5,18E‐12 1,39E‐08
(∆ revenue ‐ ∆ receivables)/ Ai,t ‐1 6514 0,072 0,044 0,207 ‐0,467 1,144
PPE/ Ai,t‐1 6514 0,574 0,465 0,430 0,011 2,009
ROAi,t‐1 6514 0,053 0,049 0,070 ‐0,100 0,266
coefficients
α0 6514 ‐0,040 ‐0,033 0,189 ‐4,275 2,416
α1 6514 ‐2,58E7 ‐3,24E6 5,70E8 ‐6,99E9 1,42E10
α2 6514 ‐0,015 ‐0,003 0,448 ‐6,121 7,632
α3 6514 ‐0,032 ‐0,022 0,261 ‐2,701 3,511
α4 6514 0,144 0,068 3,351 ‐59,076 154,323
Estimated accruals
Non‐Discretionary accruals/Ai,t ‐1 6514 ‐0,063 ‐0,055 0,050 ‐0,232 0,065
Discretionary accruals/Ai,t ‐1 6514 8,01E‐4 0,048 ‐0,191 0,157
3.75E‐16
Abs. Discretionary accruals/Ai,t ‐1 6514 0,031 0,018 0,038 0,000 0,191
Appendix 2b
Description of variables used to calculate discretionary accruals and incentive ratios
Indicator in thesis Description database Description database
NI NI Bottom line net income
CFO ONACF Operating activities net cash flow
A AT Assets Total
REV REVT Revenue Total
AR RECT Receivables Total
PPE PPEGT Property Plant and Equipment Total (Gross)
ROA NI/AT Net income/assets total
Price PRCC_F Price Close Annual Fiscal
Shares SHROWN_EXCL_OPTS Shares owned options excluded
Optionval ‐ OPT_UNEX_EXER_EST_VAL + OPT_EXER_VAL
Value options held OPT_UNEX_EXER_EST_VAL Estimated intrinsic value of in the money

unexercised exercisable options
Value options exercised OPT_EXER_VAL Value realized on option exercise during the
year
Appendix 3
Appendix 3a
Development by Year equity incentives and cash pay CEO
Average Median Average Median Average Average Average Average
N Options Options Stock held Stock Held cash pay Incratio Incratio Incratio
Year held held options stock total
1999 569 33.469.911 327.000 378.863.789 987.741 1.285.479 0,1357 0,2126 0,2815
2000 559 32.761.469 464.719 222.870.902 849.206 1.301.469 0,1482 0,2043 0,2843
2001 547 19.882.525 409.264 172.846.020 1.078.522 1.345.452 0,1141 0,1906 0,2547
2002 552 12.043.306 137.000 149.686.931 694.882 1.435.542 0,0770 0,1634 0,2017
2003 555 18.854.333 716.504 166.922.442 1.168.580 1.597.005 0,1113 0,1772 0,2367
2004 559 22.688.179 891.073 144.763.226 1.482.018 1.727.544 0,1165 0,1645 0,2323
2005 549 24.188.955 916.551 94.055.185 1.562.937 1.715.038 0,1275 0,1706 0,2425
2006 557 24.113.646 773.008 159.536.016 1.727.214 1.216.754 0,1571 0,2048 0,2948
2007 557 21.063.207 375.000 193.818.196 1.693.487 1.209.321 0,1513 0,2042 0,2868
2008 541 6.682.137 0 81.633.173 933.047 1.185.864 0,0712 0,1491 0,1899
2009 497 7.992.999 46.894 155.774.217 1.338.473 1.169.578 0,0690 0,1587 0,2004
Appendix 3b
Development by Year equity incentives and cash pay CFO
Average Median Average Median Average Average Average Average
Options Options Cash Incratio Incratio Incratio
N Stock Held Stock held
Year held held Pay options stock total
1999 479 5.308.818 327.000 6.088.963 987.741 553963 0,0689 0,0617 0,1190
2000 511 3.702.600 464.719 5.731.416 849.206 522069 0,0687 0,0537 0,1143
2001 503 2.377.201 409.264 5.288.444 1.078.522 563800 0,0479 0,0505 0,0931
2002 530 1.780.305 137.000 3.795.769 694.882 615905 0,0297 0,0385 0,0639
2003 551 2.951.554 716.504 5.938.947 1.168.580 643388 0,0488 0,0480 0,0909
2004 569 2.822.296 891.073 4.600.927 1.482.018 703879 0,0526 0,0500 0,0945
2005 570 3.354.244 916.551 4.709.435 1.562.937 745941 0,0539 0,0533 0,0983
2006 707 3.800.942 773.008 5.588.860 1.727.214 562744 0,0700 0,0637 0,1219
2007 737 3.100.615 375.000 6.289.965 1.693.487 569871 0,0664 0,0681 0,1238
2008 732 739.678 0 2.913.469 933.047 558931 0,0259 0,0398 0,0624
2009 665 1.749.098 46.894 9.848.747 1.338.473 621557 0,0255 0,0480 0,0698
100
Appendix 4
Appendix 4a
Descriptive Statistics of the regression with discretionary accruals and CEO equity incentives for
the total sample period 1999‐2009
DAi,t = α 0 + α1 ∙ INCENTIVE RATIO + α2 ∙ oldfirm + α3 ∙size + α4 ∙leverage + α5 ∙sales volatility + αx ∙

year dummies
Variables are explained in appendix 4g

Accruals
Abs. Discretionary accruals 6042 0,030 0,018 0,038 0,000 0,191
Equity incentives
Incentive Ratio Total 6042 0,246 0,139 0,266 0,000 0,994
Incentive Ratio Option 6042 0,117 0,039 0,179 0,000 0,880
Incentive Ratio Stock 6042 0,181 0,073 0,253 0,000 0,993
Controls
Oldfirm 6042 0,775 1,000 0,417 0,000 1,000
Size 6042 17,976 17,887 1,416 12,211 23,140
Leverage 6042 0,668 0,641 0,317 0,000 4,955
Sales volatility 6042 0,190 0,120 0,251 0,000 4,857
Appendix 4b
Summary Statistics of the regression with discretionary accruals and CEO equity incentives for the
Pre‐SOX sample period 1999‐2001

year dummies

Accruals
Equity incentives
Controls
Oldfirm 1675 0,810 1,000 0,392 0,000 1,000
Size 1675 17,641 17,551 1,435 12,211 22,198
Leverage 1675 0,703 0,657 0,378 0,000 4,955
Appendix 4c
Summary Statistics of the regression with discretionary accruals and CEO equity incentives for the
post‐SOX sample period 2003‐2009

year dummies

Accruals
Equity incentives
Controls
Oldfirm 3815 0,758 1,000 0,428 0,000 1,000
Size 3815 18,141 18,038 1,379 12,573 23,140
Leverage 3815 0,655 0,634 0,290 0,000 4,630
Appendix 4d
Summary Statistics of the regression with discretionary accruals and CFO equity incentives for the
total sample period 1999‐2009

year dummies

Accruals
Equity incentives
Controls
Oldfirm 6554 0,769 1,000 0,421 0,000 1,000
Size 6554 18,006 17,921 1,410 12,211 23,140
Leverage 6554 0,673 0,647 0,310 0,000 4,630
102
Appendix 4e
Summary Statistics of the regression with discretionary accruals and CFO equity incentives over
the Pre‐SOX sample period 1999‐2001

year dummies
Variables are explained in appendix4g

Accruals
Equity incentives
Controls
Oldfirm 1493 0,806 1,000 0,395 0,000 1,000
Size 1493 17,583 17,502 1,398 12,211 21,844
Leverage 1493 0,703 0,657 0,362 0,000 3,550
Appendix 4f
Summary Statistics of the regression with discretionary accruals and CFO equity incentives over
the post‐SOX sample period 2003‐2009

year dummies
Variables are explained in appendix4g

Accruals
Equity incentives
Controls
Oldfirm 4531 0,755 1,000 0,430 0,000 1,000
Size 4531 18,165 18,061 1,387 12,573 23,140
Leverage 4531 0,663 0,642 0,292 0,000 4,630
103
Appendix 4g
Description of variables in regression between discretionary accruals and equity incentives
Indicator in thesis Description
Abs. Discretionary accruals The absolute value of discretionary accruals calculated with a linear
Kothari model as describes in chapter 7 scaled by lagged total assets
Incentive Ratio Total Equity incentive ratio containing both stock and equity incentives. This
ratio is explained in chapter 7
Incentive Ratio Option Equity incentive ratio containing only option incentives, this is further
explained in chapter 7
Incentive Ratio Stock Equity incentive ratio containing only stock incentives, this is further
explained in chapter 7
Oldfirm Dummy variable for firm age, this variable is one if the company is part of
the “compustat” database for 20 years or more.
Size The natural logarithm of lagged total assets.
Leverage Total firm liabilities deflated by total assets.
Sales volatility This is the standard deviation of revenue deflated by total assets over the
current year and previous four years
Year dummy A dummy variable for the firm year.
104

Masters Thesis of Accounting

Uploaded by

Copyright:

Available Formats

Masters Thesis of Accounting

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Masters Thesis of Accounting

Uploaded by

Copyright:

Available Formats

Master’s thesis accounting, auditing and control

Equity incentives and earnings management

In partial fulfillment of the requirements for the degree of

Erasmus University Rotterdam

My main research question is:

To analyze this relation further I examine the following sub questions:

‐ Does this relation defer for stock or option‐based incentives?

‐ Do these relations change in the 10 years period from 1999 to 2009?

1.3 Relevance and contribution

1.4 Structure of the thesis

2.1 introduction and the reason for earnings management

A well‐known theory on decision‐making is the utility maximizing theory. This theory

2.2 What do we consider earnings management?

To my opinion it is often difficult to determine when accounts manipulation is legal or

Stolowy and Breton 2004

Thereby it is to my opinion important to know why someone took a certain decision

This is therefore in my opinion a good definition of earnings management. However

“Earnings management is a collection of managerial decisions that result in not reporting

Earnings management can be:

Beneficial: it signals long‐term value;

Pernicious: it conceals short‐ or long‐term value;

Neutral: it reveals the short‐term true performance.

2.3 Measuring earnings management with accruals

Accruals accounting is something that is normally used in everyday practice. Accruals

Healy (1985) started a discussion on measuring earnings management with accruals

2.4 Who commits earnings management?

2.5 Summary Definition earnings management

Earnings = cash flow + accruals

Management can influence accrual accounting. Management has a certain degree of

Earnings = cash flow + normal accruals + discretionary accruals

3.2 The Healy model 1985

TAi,t = ( CAi,t – CLi,t – Cashi,t + STDi,t – Depi,t) / Ai,t – 1

TAi,t Total accruals of firm i at time t

DAt Discretionary accruals of year t scaled by lagged total assets

3.3 The De Angelo model 1986

DAt = TAt – TAt‐1

DAt Discretionary accruals of year t scaled by lagged total assets

3.4 Jones model 1991

TAi,t = ( CAi,t – CLi,t – Cashi,t + DD1i,t – Depi,t) / Ai,t – 1

TAi,t Total accruals of firm i at time t scaled by lagged total assets

TAi,t = 1× ( 1/ Ai,t‐1) + 2 × (∆REVi,t) + 3 × (PPEi,t) +ei,t

TAi,t Total accruals scaled by lagged total assets of company i in year t

DAi,t= TAi,t – NDAi,t

DAi,t Discretionary accruals scaled by lagged total assets

3.5 Modified Jones model 1995

NDAi,t = 0+ 1× ( 1/ Ai,t‐1) + 2 × (∆REVi,t ‐ ∆AR i,t)+ 3 × PPEi,t

3.6 Time‐series versus cross sectional Jones models

3.6.1 Time‐Series designs with the Jones model

3.6.2 Cross‐sectional designs with the Jones model

A cross‐sectional research design has disadvantages as well. It is questionable whether

Whether to use a cross‐sectional or a time‐series design depends on the assumptions

The following formula is used to calculate balance sheet accruals:

TAi,t Total accruals of firm i at time t lagged by total assets

TAi,t = (EBXIi,t – CFOi,t)/ Ai,t‐1

TAi,t Total accruals of firm i at time t scaled by lagged total assets.

3.8 Improved versions of the Jones model

3.9 The forward‐looking model 2003

K measures the sensitivity of the change in non‐discretionary accounts receivable to

TAi,t= 1 + 2((1+K) ∆salesi,t ‐ ∆ARi,t) + 3PPEt + 4 (TAi,t,‐1/ Ai,t‐2) + 5 GR_salesi,t

3.10 Cash flow Jones model 2002