Amsterdam Business School
Real earnings management and executive compensation
and the impact of the financial crisis at U.S. stock listed companies (2005-2012)
Name: Gino van Heusden Student number: 10291601 Date: August 2014
Supervisor: ir. drs. A.C.M. de Bakker Second supervisor: dr. ir. S.P. van Triest
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Abstract
Existing research have documented there seems to be a positive relationship between equity-based executive compensation and Accrual equity-based earnings management. This research will empirically re-establish the relationships between equity-based executive compensation components and earnings management. However this research will contribute to previous studies by specifically focus on the relationship between executive compensation and Real earnings management. No recent research focused on the relation financial crisis /economic recession with earnings management, therefore this research will contribute by specifically examine the effect of the financial crisis/ economic recession on the previously determined relationship between executive compensation and earnings management.
This research provides a sample of 9.208 CEO year observations, taken from companies listed on the U.S. stock exchange over the years 2005-2012. To measure the use of real earnings management at U.S stock companies, this research will use an existing research model to perform the regression analysis. This model has been designed and developed by Dechow, Kothari and Watts (1998) and was first implemented by Roychowdhury (2006).
The results do not show a positive significant relationship between the equity-based executive compensation in a fiscal year and the use of real earnings management. However the results do show a positive significant relationship between the total executive compensation of CEO’s in a fiscal year and the use of real earnings management as measured by abnormal cash flow from operations and discretionary expenses. Furthermore the results show that the financial crisis period (2008-2009) is positively associated with use of real earnings management as measured by abnormal cash flows from operations.
Keywords: Accrual based earnings management, Real earnings management, Executive compensation and regression analysis.
Preface
All master studies are completed with a final scientific research. This research is performed to complete my master Accounting and Control at the University of Amsterdam.
For me there are three important conditions on which I choose my thesis topic. The first condition was to choose a thesis topic where my interest lies, the second conditions was to choose a topic with actuality and the last condition was to choose a topic which is frequently used during my master courses.
In this thesis I have examined the relationship between Real earnings management and executive compensation and the effect of the financial crisis/economic recession on this relationship for US stock exchange companies over the year 2005-2012. In my point of view this topic satisfies all described conditions and is therefore a very interesting topic to me.
I have learned a lot about the possibilities of executives to engage in earnings management and the different forms of CEO compensation after completion of writing this thesis. This thesis will not only help me to better understand how executives and investors look to and use the company’s annual report, but it will also help me becoming a better accountant as I now have a better understanding how earnings can be “wrong” in practice.
This research is conducted under the supervision of ir. drs. A.C.M. de Bakker – from the University of Amsterdam. I would like to take the opportunity to thank ir. drs. A.C.M. de Bakker for his supervision during this period.
Finally I would like to thank my employer Deloitte Accountants B.V. for the opportunity to attend this Master Accounting and Control at the University of Amsterdam and the opportunity to write my thesis.
Gino van Heusden Amsterdam, August 2014 The Netherlands
Table of contents Abstract ... 1 Preface ... 2 1 Introduction ... 5 1.1 Background ... 5 1.2 Research question ... 6
1.3 Contribution of the research ... 6
1.4 Structure of the research ... 7
2 Literature review and hypotheses development ... 8
2.1 Agency Theory ... 8
2.2 Stewardship theory ... 9
2.3 The Positive accounting theory ... 9
2.4 Earnings management ... 10
2.4.1 Definition and different forms of Earnings management ... 10
2.4.2 Why would managers get involved in earnings management? ... 11
2.4.3 Prior literature Real Earnings Management ... 12
2.5 Executives compensation structure ... 14
2.6 Evidence from prior studies: The relation between Executive compensation and earnings management ... 15
2.7 Financial crisis ... 17
3 Hypotheses development ... 18
4 Research Methodology ... 20
4.1 The research model ... 20
4.2 Determination of the abnormal values ... 20
4.2.1 Cash flow from operations (CFO) ... 21
4.2.2 Discretionary expenses ... 22
4.2.3 Overproduction ... 22
4.2.4 Definition of dependent variables ... 23
4.2.5 Composition of independent variables for abnormal value regressions ... 23
4.3 Regression model for hypotheses testing ... 24
4.3.1 Regression model ... 24
4.3.2 Dependent Variables ... 25
4.3.4 Control variables (Independent variables) ... 26
4.4 Summary regressions ... 27
5 Data and sample selection ... 28
5.1 Database sources ... 28
5.2 Data collection process ... 28
6 Research findings ... 30 6.1 Regression results ... 30 6.2 Descriptive statistics ... 31 6.3 Correlation matrix ... 33 6.4 Multicollinearity ... 34 6.5 Hypotheses testing ... 35 6.5.1 Testing hypothesis 1 ... 35 6.5.2 Testing hypothesis 2 ... 36 6.5.3 Testing hypothesis 3 ... 38 6.5.4 Testing hypothesis 4 ... 39
6.6 Summary of research results ... 40
7 Conclusions, limitations and suggestions for future research ... 42
7.1 Conclusions ... 42
7.2 Limitations and suggestions for future research ... 43
References ... 44
Appendix A - Assessment of natural logarithm executive compensation ... 46
1 Introduction
1.1 Background
Over the last few decades, much research has been performed on earnings management. In the beginning, this research was mainly focused on accruals-based earnings management. Recently due to the stringent regulations, such as the introduction of the Sarbanes-Oxley Act (SOX)1 it
has become more difficult for managers to manage their earnings by means of accruals. Partly because of these stricter regulations, there has been a shift from accrual-based earnings management to real earnings management (Cohen et. al, 2008). Real based earnings management is generally harder to detect and prevent compared to accruals-based earnings management, plus does not reverse automatically over time. Real earnings management generally involves changes in the timing or structuring of operations, investments, and/or financing transactions with cash flow consequences (Xi Li, 2010).
Besides the shifting from accrual-based earnings management to real-based earnings management, previous studies shown that executive compensations has become a hot topic in recent years. Mainly driven by the ongoing civil discussion about the height of executive compensation and related new laws and strict regulation. The focus on executive compensation is mainly driven by the current debate whether stock options and large bonuses are associated with potential of engaging in earnings management.
This research will focus on the effect of executive compensation structure on the use of real earnings management for US Stock listed companies over the years 2005 to 2012 and will specifically investigate whether the relation between equity-based executive compensation and real earnings management is affected by the financial crisis/ economic recession period. In this research the financial crisis periods are divided in: the pre-crisis years (2006-2007), crisis years (2008-2009) and post-crisis years (2010-2012).
The financial crisis started in the US in December 2007 and lasted till June 2009 (National Bureau of Economic Research, 2013) and was able to spread out rapidly worldwide. The financial crisis had significant effect on company’s profits and revenues worldwide and is therefore interesting as a research subject. The purpose of including the financial crisis in this research is to analyze whether the financial crisis affect the use of earnings management. Prior research by Johl et al. (2003) on earnings management during the financial crisis in Asia in 1997,
concludes that the financial crisis had a significant effect on the use of earnings management at Asian companies.
Previous research has found there is a significant relationship between the compensation structure of executives and earnings management (Kuang, 2008). This study shows that managers are more likely to manage earnings if a greater proportion of total compensation consists of performance vested stock options. Other prior research has found there is a positive relationship between equity-based executive compensation and earnings management (Gao & Shrieves, 2002; Bergstresser & Philippon, 2006; Cheng & Warfield, 2005).
1.2 Research question
Based on the background as described in paragraph 1.1, the research question can be formulated as follows:
“How does the crisis affect the relationship between equity-based executive compensation and real based earnings management?”
To answer the research question, the following sub-questions will be answered:
1. What is real earnings management, what methods of earnings management are there and how can real earnings management be measured?
2. Why would executives engage in earnings management? 3. Of which variables consists executive compensation?
4. What is the influence of executive compensation structure on real earnings management? 5. What research methodology is best suitable in order to measure the use of real earnings
management in times of a financial crisis / economic recession?
1.3 Contribution of the research
Prior research has found there is a positive relationship between equity-based executive compensation and earnings management (Gao & Shrieves, 2002; Bergstresser & Philippon, 2006; Cheng & Warfield, 2005). This paper will contribute to existing research as it will examine specifically the effect of executive compensation components on real based earnings management for US Stock listed companies in the period 2005-2012.
Previous research mainly focused on the relation between executive compensation and accrual based earnings management and are mostly performed before the financial crisis. This research will specifically contribute by focusing on the pre-crisis years (2006-2007), crisis years
(2008-2009) and post-crisis years (2010-2012) and will examine the effect of executive compensation on the use of real earnings management at US Stock listed companies.
1.4 Structure of the research
Chapter two of this research will deal with the literature review and will discuss prior literature. Chapter three consists of the hypotheses development. Chapter four will deal with the research model, the different variables and the research methodology. Chapter five includes the data and sample selection and chapter six will discuss the research results. Finally Chapter seven will deal with the conclusion and limitations of the research.
2 Literature review and hypotheses development
This chapter will firstly describe the Agency theory and the stewardship theory. Secondly earnings management will be explained in more detail, this includes the definition and different methods of earning management applied in practice. Next the various forms of executive compensation will be explained. Furthermore this chapter will describe what prior research has shown on the relationship between executive compensation and real earnings management.
2.1 Agency Theory
There is in a conflict of interest between the agent and the principal according to the Agency Theory. Jensen and Meckling (1976) are among the first researchers who examined the Agency Theory. Within the agency theory, the shareholders who are the owners or principals of the company, hire agents to perform work for them. Basically the agency theory is based on three assumptions: (1) conflicting goals between the agent (company managers/executives) and the principle (such as shareholders). The principle is unable to verify what the agents are doing. (2) The Agent and the principle have different attitudes towards risk, the agent is generally more risk averse compared to the principle, this may result in taking different actions in similar situations, this is normally not good for creating company value. (3) Managers always try putting their own interest first while performing their work, this conflict of interest can result in choices made by management which are not the best in the long term for the company.
As described above, according to the agency theory, the agent and the principle differ in goals. One of the main goals of the Agent (Manager) is to gain the highest possible bonus or equity incentive, while the principal has as main goals to gain and increase long term company value and long term profits which result to higher stock prices and thus on the long term higher dividend distributions to the shareholders. This conflict of interest between the agent and the principle may possibly result in earnings management (Jensen and Meckling 1976). Managers for example are willing to increase or decrease current investments that will result in an increase or decrease in current year earning if this would result in higher equity compensations packages or bonus payment (focus on self-interest. This is a typical example of Real earnings management in practice. Agents are generally less concerned about future period actions as they will don’t expect to be with the company anymore by then (Lambert, 2001).
In practice it is impossible for the principle to permanently check and/or review the work, actions and decisions made by their agents. Therefore agents and principles generally sign up for
a contract which includes the main goals for the agent. This process will lead to goals which are desired by both parties.
2.2 Stewardship theory
The stewardship theory is an alternative view of the agency theory. The stewardship theory is defined by Davis, Schoorman & Donaldson (1997) as follows “a steward protects and maximizes shareholders wealth through firm performance, because by so doing, the steward’s utility functions are maximized”. Within the stewardship theory, stewards are managers and executives working for the shareholders, Stewards protects shareholders value and make profits for the shareholders. The agency theory assumed managers to act in their own self-interests at the expense of shareholder, where the stewardship theory assumes that managers will indeed act as responsible stewards of the company they control on behalf of the shareholders.
According to the stewardship theory, stewards will be satisfied and motivated when organizational success is attained. Organizations are more likely to achieve success when managers are financially rewarded or when shares are offered when certain goals are achieved. In order to achieve these goals set by the stewards, managers might possibly engage in earnings management.
2.3 The Positive accounting theory
The positive accounting theory is a descriptive theory which describes particular aspects of financial accounting practice and is initially developed by Watts en Zimmerman (1986). The positive accounting theory is a theory that seeks to explain, observe and predict a particular phenomenon (Deegan and Unerman, 2006). For example the positive accounting theory seeks to explain why managers and/or accountants choose to adopt particular accounting methods in preference over others. The positive accounting theory of Watts and Zimmerman’s (1986) will be used in this research in paragraph 2.4.2 to describe why managers are willing to engage in earnings management in practice.
The normative accounting theory is the opposite equivalent of the positive accounting theory. The normative accounting theory is more descriptive theory and tells what individuals or organizations should do in certain situations. In contrast, the positive accounting theory is a theory that explains and observes particular behaviour, and predict what an individual or organization is going to do in a given situation.
2.4 Earnings management
In this paragraph the definition and different forms of Earnings management will be described. Thereafter will be explained why managers engage / get involved in earnings management. Subsequently prior literature on Real Earnings Management will be outlined.
2.4.1 Definition and different forms of Earnings management
In prior studies different definitions has been used to define earnings management. K. Schipper (1989) defines earnings management as: “A purposeful intervention in the external financial reporting process, with the intent of obtaining private gain (as opposed to, say, merely facilitating the neutral operation)”, which can be interpreted as intentional influencing the financial reporting of a company.
A more detailed definition is given in the research of Healy and Wahlen (1999) “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 practices”. Above definition of earnings management by Healy and Wahlen (1999) indicates there are main two motives for managers to manage their earnings: (1) misleading stakeholders about the actual economic performance and (2) influencing contractual outcomes which rely on the reported figures.
Basically there are four patterns of earnings management according to Scott (2009): (1) taking a bath, (2) Income minimization, (3) Income smoothing and (4) income maximization. The earnings management pattern chosen is related to the goals included in the bonus plan of executives. In practice managers will use mostly income maximizing.
Managers engage in earning management in two possible ways: accrual based earnings management and real earnings management. Accrual based earnings management typically involves accounting choices and reverses itself over time. Accounting numbers contains cash components and accrual components. However in practice managers are able to influence accrual components more compared to cash components (Dechow & Dichev, 2002). Real earnings manipulation is defined by Roychowdhury (2006) as “departures from normal operational practices, motivated by managers’ desire to mislead at least some stakeholders into believing certain financial reporting goals have been met in the normal course of operations.” Gunny (2005) defines real earnings management as “Real earnings management occurs when
managers undertake real business actions that deviate from the first best practice to increase reported earnings”.
Because of the stricter regulations such as SOX, there has been a shift from accrual-based earnings management to real earnings management (Cohen et. al, 2008). Previous research found that managers adjust the level of accrual-based earnings management according to the level of real earnings management applied (Zang, 2012). The research of Bruns and Merchant (1990) and Graham et al (2005) indicates that executives have a greater willingness to manipulate their earnings through real activities rather than through accruals.
Practical examples of accrual based earnings management are: bad debt expenses and delaying asset write-offs. Examples of real earnings management are changes in the timing and/or structuring of operations, investments, and/or financing transactions and have direct cash flow consequences (Li, 2010).
Earning management is generally associated with negative behavior, such us “misleading stakeholders”. However according to the theory in Scott (2009) earnings management can sometimes be used in a positive way, this is called “Signaling”. Signaling is an example of earnings smoothing where the company gives a positive signal to the market regarding future profitability or stock market expectations.
2.4.2 Why would managers get involved in earnings management?
The positive accounting theory of Watts and Zimmerman’s (1986) provides a framework which explains why managers are willing to make use of and engage in earnings management. Watts and Zimmerman (1986) conclude that individuals are mainly focused on self-interest behavior. As previously mentioned: the positive accounting theory is a theory that explains and observe particular behavior, and predict what an individual or organization is going to do in a given situation.
The Positive accounting theory investigated three hypotheses which attempt to explain why managers choose particular accounting options/methods over others. The following hypotheses have been investigated: (1) Bonus plan hypotheses, (2) debt covenant hypotheses and (3) political costs hypotheses.
The hypotheses used in above described research that is interesting for this research is the bonus plan hypotheses. Watts and Zimmerman (1986) examine how the choice for accounting methods could lead to opportunistic behavior among managers within companies. The bonus plan hypotheses is interesting to this research as it focus as the bonus plan
hypotheses assumes that opportunistic behavior takes place when individuals attempt to maximize their own compensation. According to the research performed by Watts and Zimmerman (1986), managers have incentives to apply for the accounting method which will lead to maximum self-interest (i.e. highest compensation for the individual).
2.4.3 Prior literature Real Earnings Management
Besides using accrual based earnings management, managers are likely to engage in manipulation of real operational activities in order to achieve certain earnings levels (Healy and Wahlen, 1999; Fudenberg and Tirole, 1995; Dechow and Skinner, 2000).
Prior literature in which real earnings management has been examined, mainly focused on three forms of real earning management. Dechow, Kothari and Watts (1998) developed an empirical model, which was first implemented by Roychowdhury (2006) and later by Cohen et.al (2008), to detect real earnings manipulations. According to Roychowdhury (2006) the variables as described below capture the effect of real operations better than accruals would do. The studies of Cohen et.al (2008) and Roychowdhury (2006) focused on three main manipulation methods:
1. Cash flow from operations: Acceleration of the timing of sales through increased price discounts or more lenient credit terms. (Managers use sales manipulations in order to achieve certain sales targets which will lead to higher earnings levels and thus managers are able to achieve their bonus plans)
2. Overproduction: Reporting of lower cost of goods sold through increased production. (Managers use production manipulations by increasing the production levels in the current period, which result in lower cost of goods sold, which leads to higher margins and thus increased short-term earnings.)
3. Decreasing discretionary expenses: Decreases in discretionary expenses which include advertising expense, research and development and SG&E expenses (Selling, General and Administrative Expenses). (Managers use discretionary expenses manipulations by decreasing: current period advertising expense, research and development and SG&E expenses, which leads to increased short term earnings in the current period).
Roychowdhury (2006) used in his research the abnormal level of (1) Cash flow from operations, (2) Overproduction and (3) Decreasing discretionary expenses to measure the degree of real earnings management. The above three measures for real earnings management and the research model as applied by Roychowdhury (2006) will be used in this research in chapter four.
Managers attach high value to meeting earning targets such as such as: zero earnings, previous period’s earnings, and analyst forecasts, according to the research of Roychowdhury (2006). In addition Roychowdhury (2006) finds that real earnings management can reduce firm value, as actions taken in the current period to manipulate earnings has a negative effect on the future period earnings and cash flows. Examples are: (1) price discounts can lead to customers expecting certain discounts in future period as well, (2) overproduction leads to higher inventory levels which could increase inventory storage cost on the long term. Roychowdhury (2006) concludes that managers prefer real earnings management over accrual based earnings management. This evidence can be confirmed by the outcome of the research by Graham et al (2005), who concludes that executives have a greater willingness to manipulate their earnings through real activities rather than accruals.
Cohen et.al (2008) investigated the level of accrual based and real earnings management Pre- and Post-Sarbanes Oxley Periods over the years 1987-2005. They found that the level of real earnings management declined prior to the SOX passage in 2002, however the use of real earnings management increased significantly after the introduction in 2002. Controversy the level of accrual based earnings management increased prior to the SOX passage and decreased after this introduction in 2002. Since the introduction of stringent regulations such as SOX it became harder to engage in earnings management for managers by using accruals. Therefore managers start to use real earnings management more often. Real based earnings management is generally harder to detect and prevent compared to accruals-based earnings management, plus does not reverse automatically over time.
In addition Cohen et.al (2008) found that less accruals and more real earnings management was used at firms who just achieved important earnings benchmarks after the introduction of SOX compared to the same firms before the introduction of SOX. Furthermore Cohen et.al found that the increase of accrual based earnings management before the introduction of SOX were in line with the increase of equity-based compensation of executives.
Other previous studies on real earnings management were mainly focusing on R&D expenses. Dechow and Sloan (1991), Baber et al. (1991) and Bushee (1998) find evidence that executives reduce R&D expenses towards the end of their tenure to increase current period earnings to meet certain earnings benchmarks. The study of Burgstahler and Dichev (1997) provide evidence that executives manage real earnings to meet zero earnings threshold. Finally the survey by Graham et al (2005) find evidence that managers prefer reducing discretionary expenses and capital expenses over other manipulation methods. Graham et al. (2005) reported
the following in their survey: “[W]e find strong evidence that managers take real economic actions to maintain accounting appearances. In particular, 80% of survey participants report that they would decrease discretionary spending on R&D, advertising, and maintenance to meet an earnings target. More than half (55.3%) state that they would delay starting a new project to meet an earnings target, even if such a delay entailed a small sacrifice in value….”
2.5 Executives compensation structure
Over the last couple years executive compensation structure has undergo major changes. Figure 1 below describes the change of equity pay and non-equity pay components from 1970 to 2010 based on average CEO pay (Murphy, 2012) (all CEO’s included in the S&P 500, data used from Forbes and ExecuComp). Figure 1 shows that CEO equity pay didn’t even exist until the early 1980’s. Since then equity-based compensation at CEO’s started to become the more and more important. Nowadays Equity-based compensation is seen as the most important compensation element for executives.
Figure 1 Figure 2
According to figure 2 (Murphy, 2012) finds that executive compensation packages consists of the following five main elements:
1. Salary: fixed compensation for executives, not directly tied to financial and non-financial performance measures;
2. Bonuses: A sum of money executives receive based on their performance within one year, usually in appreciation for work done, length of service and based on financial and non-financial performance measures;
3. Stock options: executives receive “options” which give them the right to buy company stocks during a certain period of time at a predetermined price;
4. Restricted stock options: Executives receive stock of a company that is not completely transferable until certain conditions have been met, such as: continued employment or certain earnings target; and
5. Other compensations: includes all other types of compensation CEO’s receive, such as perquisites, signing bonuses, termination payments, long term retirement plans and above-market interest paid on deferred compensation.
However, according to the research of Murphy (1999), executive compensation components can be divided into three main categories: (1) Fixed salary, (2) Bonus and (3) long term compensation such as stocks, stock options and other long term compensations.
Therefore this research will focus on Salary, bonuses and the total of equity-based compensation. Equity compensation components in this study is the total of stock option compensation and stock compensation packages.
According to the study of Murphy (1999) executive bonus plans are based on three basic components: Performance measures, performance standards and pay performance structure. Performance measures are used by stakeholder to reward executives for behavior which is good for the company (based on the agency theory). Performance measures can be financial (i.e. Earnings before tax) or non-financial performance measures (customer satisfaction rates). Performance standards are based on the main objectives of the bonus plan. Bonus objectives shouldn’t be too easy or too hard to achieve as executives need achievable bonus targets or they might get demotivated (Fisher, 2003). The pay performance structure refers to the height of the bonus paid to executives when certain target have been achieved (Murphy, 1999).
Earlier studies found that real earnings management destroys long-term company value and is bad for organization (Roychowdhury, 2006). Therefore we can assume that executives reduce the amount of real earnings management when he/she has more interest in the long term perspective of compensation (equity-based compensations).
2.6 Evidence from prior studies: The relation between Executive compensation and earnings management
Previous studies have found there is a significant relationship between executive compensation and earnings management (Kuang, 2008). Kuang (2008) investigated for the 240 largest UK firms the effects of performance-vested stock options on the tendency of managers to engage in earnings management. This study concludes that managers are more likely to engage in earnings management when they hold large portions of performance vested stock options.
Other prior research has found there is a positive relationship between equity-based executive compensation and earnings management (Gao & Shrieves, 2002; Bergstresser & Philippon, 2006; Cheng & Warfield, 2005). Above studies indicate that equity-based
compensation packages for CEO’s can have a negative impact on the behavior of these company executives. For example when executives hold large portions of stock options, they are able to influence the stock market by reducing or increasing the current period earnings. Executives can use their information advantage on other investors and ability to engage in earnings management to increase their personal benefit.
Cheng & Warfield (2005) find that managers with relatively high equity incentives are less likely to report large positive earnings surprises, are they are sensitive for future stock performance. These actions can lead to increased reserving of current earnings to avoid future earnings disappointments (Cheng & Warfield, 2005).
Gao & Shrieves (2002) investigated whether “discretion over accounting accruals gives managers a potentially valuable timing option that will lead to strategies for maximizing their compensation”. Gao & Shrieves (2002) found that accrual based earnings management is positively related to manager’s compensation design. Additionally they found that bonuses and options are positively related to earnings management, whether Normal salaries are negatively related to earnings management (as measured by abnormal discretionary current accruals).
Bergstresser & Philippon (2006) found in their research that executives exercise large amount of options in years with relatively high accruals. In addition they find that during years of high accruals CEOs sell large amounts of shares (which indicates they increase the share price with these discretionary accruals in order to boost the share price).
Baker et al.’s (2003) investigated whether the structure of executive compensation is associated with opportune use of discretionary accruals in company’s annual report. They found that high portion executive option compensation is associated with accruals which result in lower reported earnings.
Chi et al. (2011) examined whether managers resort to real earnings management when the ability to engage in accrual based earnings management is constrained by high quality auditors. They found that firms who meet or just beat earnings benchmarks and have city level auditors are associated with higher level real earnings management. Furthermore they found that longer audit tenures are associated with higher levels of real earnings management.
Furthermore prior studies from Ball and Brown (1968) found there is a positive relation between earnings and stock price. Which indicate that managers are able to influence the stock price by engaging in earnings management.
2.7 Financial crisis
The financial crisis/ economic recession started in the US in December 2007 and lasted till approximately June 2009 (National Bureau of Economic Research, 2013). The financial crisis/ economic recession was mainly caused by increasing interest on mortgages at the U.S. property market, which resulted to homeowners who were no longer able to afford the these higher mortgage costs. US Banks had took huge risks supplying risky mortgages to people in the U.S. as the people they sold the mortgages to were allowed to handover the mortgage to the banks when they were no longer able to pay for the mortgage. The US banks had lots of bad/risky loans re-sold to European and other banks outside the US, the problem quickly became a worldwide problem.
The financial crisis had significant effects on company’s profits and revenues worldwide and is therefore interesting as an addition to this research. The purpose of including the financial crisis in this research is to analyze whether the financial crisis affect the use of earnings management. Prior research by Johl et al. (2003) related to earnings management during the financial crisis/ economic recession in Asia in 1997, concludes that the financial crisis had a significant effect on the use of earnings management in this period of crisis.
In this research the financial crisis periods are divided in: the pre-crisis years (2006-2007), crisis years (2008-2009) and post-crisis years (2010-2012). Above determined periods are based on the US Gross Domestic Product (GDP) and gross national product (GNP) figures from the National Bureau of Economic Research (NBER).
The National Bureau of Economic Research found evidence that the financial crisis/ economic recession started in the last quarter of 2007 and ended in the second quarter of 2009 in the US. Therefore the years 2008 and 2009 are determined as the crisis period, 2005-2007 as pre-crisis years and 2010-2012 as post-crisis years in this research.
3 Hypotheses development
Chapter 3 will outline the hypotheses development. The hypotheses are based on the previous chapters and paragraphs.
Investors use companies annual reporting’s to predict future cash flows and to make expectations about future firm performance in order to make investment decisions. Executives are primary responsible for the figures in the annual report of the company. To derive to the figures in the annual report, executives have some level of discretion in making different accounting choices within the boundaries of accounting regulation. Executives are often awarded with (equity) incentives when they make certain accounting choices that result in an increased share price or increased earnings level. These incentives may cause managers to manage their earnings in order to achieve the targets set. As previous described in paragraph 2.1 the main goal of the Agent (Manager) is to gain the highest possible bonus or equity incentives while the principal has as main goal to gain or increase company value.
As previously described in paragraph 2.6 prior research has indicated there is a positive relationship between executive compensation structure and earnings management (Gao & Shrieves, 2002; Bergstresser & Philippon, 2006; Cheng & Warfield, 2005). And that real earnings management destroys long-term company value and is bad for organization (Roychowdhury, 2006).
Based on the evidence from prior research and taking into account the three measurements for real earnings management as defined by Cohen et. al, (2008) and Roychowdhury, (2006), the first three hypotheses of this research are formulated as follows: H1: Equity-based CEO compensation packages are positively associated with real earnings management as measured by abnormal cash flows from operations.
H2: Equity-based CEO compensation packages are positively associated with real earnings management as measured by abnormal discretionary expenses.
H3: Equity-based CEO compensation packages are positively associated with real earnings management as measured by abnormal production costs.
In order to measure the effect the financial crisis/economic recession had on the relation between executives’ compensation structure and real earnings management. A separate hypothesis is included in this research to measure the effect of financial crisis/economic recession on this relation. This hypothesis will be carried out using a dummy variables for the “period” of the CEO years. The period will be divided in: the pre-crisis years (2006-2007), crisis years (2008-2009) and post-crisis years (2010-2012).
H4: The crisis has a positive moderating effect on the relationship between executive compensation and real earnings management.
The above period hypothesis will be measured by including Crisis and post-Crisis dummy variables in all three real earnings management regressions as used for H1, H2 and H3. The variables will be tested for multicollinearity.
4 Research Methodology
In this chapter, the research model of this research will be described. The purpose of the research model is to investigate whether there is a relationship between executive compensation and real earnings management. First will be described how the abnormal values for the dependent variables are determined and which variables are used as dependent and independent variables. Second, the control variables will be described. The assumptions and the chosen variables ultimately result in a research model, the regression model is described in detail in paragraph three of this chapter.
4.1 The research model
This research model in this paper contains several independent variables and dependent variables, furthermore the research model contains a moderator (Financial crisis) which affect the relation between Real earnings management and Executive compensation. An abbreviated schematic representation of the research model is given figure 3 below. Paragraph 4.3 will outline the complete regression model; including schematic overview with all the dependent and
independent variables.
Figure 3
4.2 Determination of the abnormal values
This paper will make use of real earnings management as a proxy for earnings management. This research will rely on prior studies where Dechow, Kothari and Watts (1998) developed a model to determine the extent of real earnings management, this model was first implemented by Roychowdhury (2006).
This paragraph will outline how the abnormal values of real earnings management can be measured, this is needed in order to determine the dependent variables required for the final research model in paragraph 4.
The model of Roychowdhury (2006) measures the degree of real earnings management using the abnormal levels of Cash flow from operations, discretionary expenses and production costs. These manipulation methods that will be used are outlined below:
1. Cash flow from operations (Acceleration of the timing of sales through increased price discounts or more lenient credit terms)
2. Decreasing discretionary expenditures (Decreases in discretionary expenses which include advertising expense, research and development, and SG&A expenses).
3. Overproduction (Reporting of lower cost of goods sold through increased production). The normal levels of Cash flows from operations, production costs and discretionary expenses are calculated using the model as described above (Roychowdhury, 2006).
In order to calculate the abnormal levels of the three variables for real earnings management, the following formula will be used: Abnormal value= Actual value – normal value.
The normal levels are unknown and will be estimated using a multiple regression model. These regression models will be explained below per category.
4.2.1 Cash flow from operations (CFO)
Normal cash flows from operations (CFO) will be expressed as a linear function of sales and mutation in sales. Abnormal CFO is actual CFO minus the normal level of CFO calculated using the estimated coefficient (Roychowdhury, 2006) and (Cohen et al. 2008). The abnormal CFO will be calculated for every CEO year.
Where:
= Operational cash flows of organization ‘i’ in year ‘t’ = Total assets of organization ‘i’ in year ‘t’.
= Total assets of organization ‘i’ in prior year.
= Total sales of organization ‘i’ in year ‘t’.
= Mutation in sales (S t - St-1) of organization ‘i’ in period ‘t’.
4.2.2 Discretionary expenses
According to Roychowdhury (2006) discretionary expenses function as a model of “lagged sales” and “lagged assets” and estimate the following model to derive ‘normal’ levels of discretionary expenses. Where:
= Are the discretionary expenses of organization ‘i’ in period ‘t’. = Total assets of organization ‘i’ in prior year.
= Total sales of organization ‘i’ in year ‘t’.
= error
Discretionary expenses are defined by Roychowdhury (2006) as the sum of advertising expenses, R&D expenses and Selling, General and Administrative expenses (SG&A). Abnormal
Discretionary expenses are the actual Discretionary expenses minus the normal level of Discretionary expenses, calculated using the estimated coefficient (Roychowdhury, 2006) and (Cohen et al. 2008).
4.2.3 Overproduction
Production costs are defined as the sum of cost of goods sold (COGS) and change in inventory during the year. The normal level of production costs are calculated using the regression model as described below. Where:
= Production costs of organization ‘i’ in period ‘t’. = Total assets of organization ‘i’ in prior year. = Total sales of organization ‘i’ in year ‘t’.
= Mutation in sales (S t - St-1) of organization ‘i’ in period ‘t’.
= Mutation in sales prior year (S t-1 - St-2) of organization ‘i’ in period ‘t’.
The abnormal production costs are calculated by: Actual production costs minus normal production costs.
4.2.4 Definition of dependent variables
As described earlier in this chapter, three linear regressions are required to measure the abnormal values of real earnings management. The three dependent variables needed to measure real earnings management are included in table 1 below.
Table 1 – Total overview of dependent variables for abnormal value regressions
# Dependent variables Definition Database
1. Cash flow from
operations
CFO / Total assets prior year. Compustat
2. Discretionary expenses Sum of advertising expenses, R&D expenses and Selling, General and Administrative expenses / Total assets prior year.
Compustat
3. Production costs. Sum of cost of COGS and change in inventory in the year / Total assets prior year.
Compustat
4.2.5 Composition of independent variables for abnormal value regressions
Included in table 2 below are all the independent variables needed to measure the abnormal values of real earnings management.
Table 2 – Total overview of independent variables of abnormal value regressions
# Independent variables Definition Database
1.
1/ Total assets prior year. Compustat
2.
Sales / Total assets prior year. Compustat
3.
Current year change in sales / Total assets prior year. Compustat
4.
4.3 Regression model for hypotheses testing
In this paragraph the regression model for hypotheses testing, the dependent variables, the independent variables and the control variables of the regression model are outlined.
4.3.1 Regression model
This research will make use of a regression model to test the research question and hypotheses. The following empirical model will be used to test for hypothesis 1, 2, 3 and 4 (as described in chapter 3):
Regression model
Above regression model includes several independent variables. Four independent variables are included for executive compensation of which one is total compensation. It is possible that one or more variables will be eliminated in the final model applied due to inter correlation of the variables. This will be tested for in the multicollinearity paragraph of chapter 6.
Figure 4
Real earnings management 3 independent variables:
H1: Abnormal Cash flow from operations
H2: Abnormal discretionary expenses H3: Abnormal production costs.
(1) Total compensation (2) Salary-based compensation (3) Bonus-based compensation (4) Equity-based compensation
(5) SIZE
(6) Market to book ratio (7) Return on assets
Below I will outline the dependent variables as used in this research per hypothesis
H1: Equity-based CEO compensation packages are positively associated with real earnings management as measured by abnormal cash flows from operations.
H2: Equity-based CEO compensation packages are positively associated with real earnings management as measured by abnormal discretionary expenses.
H3: Equity-based CEO compensation packages are positively associated with real earnings management as measured by abnormal production costs.
4.3.2 Dependent Variables
The dependent variables used in this regression model are based on the abnormal level of Cash flow from operations, discretionary expenses and production costs. The abnormal levels are calculated using a regression (as described in paragraph 2 of this chapter). Included in table 3 below are all the dependent variables used in the regression model.
Table 3 – Total overview of dependent variables
# Dependent variables Definition Database
1. Abnormal Cash flow from
operations
Abnormal CFO = Actual CFO -/- normal CFO. Compustat
2. Abnormal discretionary
expenses
Abnormal DiscExp = Actual DiscExp -/- normal DiscExp.
Compustat
3. Abnormal production costs Abnormal PROD = Actual PROD -/- normal PROD.
4.3.3 Independent variables
This paper will make use of equity-based compensation and bonus compensation as proxies for executive compensation. Executive compensation components that can be categorized as equity are stock option compensation and restricted stock compensation. Table 4 includes all the independent variables used in the regression model.
Table 4 – Total overview of independent variables
# Independent variables Definition ExecuComp Database
1. Total Executive compensation TDC1: Total Compensation ExecuComp
2. Fixed annual executive salary Salary ExecuComp
3. Bonus compensation Bonus ExecuComp
4. Equity-based compensation: the total
of stock and options compensation
Grant Date Fair Value of Options Granted + Grant Date Fair Value of Stock Awarded Under Plan-Based Awards
ExecuComp
4.3.4 Control variables (Independent variables)
Several control variables will be included in the empirical regression model. This paragraph will describe all the control variables; what they mean and how they are calculated.
1. SIZE
First total assets (SIZE) will be included as a proxy for the logarithm size of a company using the natural log of the total assets of the firm. When firms become larger, the potential to engage in real earnings is presumed to increase. Prior research has shown a negative association with the levels of earnings management when firms become larger (Watts & Zimmerman, 1990) as larger firms are under more supervision of the outside world.
2. Market-to-book ratio (MBR)
Next, the market-to-book ratio will be included as a control variable to the empirical regression model to control for company growth (MBR). The market–to-book ratio will be measured by the market value of equity divided by the book value of equity (total ordinary shares x book value per share).
3. Return on Assets (ROA)
Third, the yearly return on assets as measured by earnings divided by lagged total assets (ROA) will be included to control for differences in performance. The return on assets will be measured by “Income before extraordinary Items” divided by the total assets of the firm.
4. Period
Finally this research includes period as a dummy control variable for the financial crisis. The years 2005-2007 are defined as pre-crisis years, the period 2008-2009 is defined as crisis years and the period 2010-2012 is defined as post-crisis years.
Table 5 includes all the independent control variables used in the regression model. Table 5 – Total overview of control variables
# Control variables
Definition Database
1. SIZE Total assets size, calculated using natural log of total assets: LN (total Assets) Compustat
2. MBR Market to book ratio: market value of equity divided by the book value of equity.
Compustat
3. Return on
assets
Earnings divided by lagged total assets Compustat
4. Period Crisis_Dummy Post_crisis_Dummy
2008-2009 2010-2012
The default period for the “period” variable is the pre-crisis period from 2005 to 2007.
4.4 Summary regressions
For this research a multiple regression will be used. Since the research model consist of 3 different dependent variables (described in more detail in paragraph 4.2). First three separate regressions will determine the normal level of real earnings management for all three dependent variables. Second the abnormal values of real earnings management is calculated using the outcome from the normal value level regression. The abnormal levels of real earnings management are the dependent variables for the hypotheses testing of regressions.
5 Data and sample selection
This chapter will outline which database resources have been used for this research. Furthermore this chapter will deal with the data collection process.
5.1 Database sources
For this research the CEO compensation variables have been collected from ExecuComp databases, which are available at the Wharton Research Data Services (WRDS). The real earnings management data and other variables are available at the Compustat databases which are also available at WRDS.
5.2 Data collection process
The sample used in this research will be based on firms listed on the U.S Stock exchange market for the years 2005 till 2012. From now on a data point of this sample will be referred to as “CEO year”. The coverage of the ExecuComp database roughly corresponds with the S&P 1500. Executive compensation data is included in the ExecuComp database for the five most earning/most important employees of the firms, however this research will specifically focus on only the CEO compensation data. To obtain only the CEO compensation data, “Annual CEO Flag” has been selected as a variable of the data set, in addition the following data has been selected: Company name, Executive name, Fiscal year, Salary, Bonus, TDC1 (Total compensation), Grant Date Fair Value of Options Granted, Grant Date Fair Value of Stock Awarded Under Plan-Based Awards and the “GVKEY” (Global Compustat Company Key). The GVKEY will later be as a unique key used to merge the ExecuComp database with the Compustat database. Now the total dataset consists of 15.357 CEO years.
Next four extra compensation variables will be added to the dataset for the executive compensation t+1. This means for example that the 2013 executive compensations are mentioned at the 2012 CEO years. Executive compensations of the current year generally relate to prior year achieved results. Now the executive compensation is linked to the year the compensation has been “earned” (CEO year 2013 was initially taken into the sample as this was needed to obtain the 2012 compensations). This step results in a total of 3.525 CEO years which are eliminated for missing t+1 CEO compensation data. The executive compensation variables are normalized by applying the natural logarithm of these variables in order to obtain more normally distributed data, Appendix A support evidence why this has been done. First all the compensation variable with a 0 value balance are transformed into a value 1, as it is not
negative compensation variables. Next all the finance industry firms will be eliminated from the sample (SIC codes between 6000 and 6999). Finance industry firms are in general strong regulated industries, and executives could potentially have different motivations at finance industry firms to manage their earnings (Cheng & Warfield, 2005). A total of 2.451 Finance industry CEO years is eliminated. Next all CEO years (60 CEO years with 0 total compensation in that year are eliminated from the sample (TDC1 = 0). Next the database has been analyzed for outliers. To check for outlier, the database has been sorted on total compensation. Based on this selection five CEO years has been eliminated as their total compensation for that year was above USD one billion. Furthermore 41 CEO years has been eliminated as their total compensation was below USD 100.000. The final ExecuComp dataset contains 11.106 CEO years. Next the Compustat database has been used to collect the real earnings management and control data. The following data has been selected: Company name, Fiscal year, GVKEY and several company reporting figures (such as total assets, sales, R&D expense). When all the data was collected, the databases has been merged (ExecuComp and Compustat) to derive to a final dataset. 66 CEO years have been eliminated from the sample as these years has missing Compustat values. The final sample used in this research is 9.208 CEO years. Finally two dummy variables have been included for “period”. As earlier described this research will specifically focus on the periods pre-crisis (2005-2007), crisis (2008-2009) and post-crisis (2010-2012). The crisis dummy variable is 1 when the period is 2008 – 2009 and 0 otherwise. The post-crisis dummy variable is 1 when the period is 2010 – 2012 and 0 otherwise.
Below in table 6 is briefly summarized what steps has led to the final dataset. Table 6 – Data used in this research
Data CEO Years
Total observations dataset ExecuComp 15.357
out: Missing CEO compensation variables for (T+1) -3.525
out: finance industry firms - SIC 6000-6999 -2.451
out: Missing CEO compensation (TDC1 = 0) -60
out: Identified outliers -46
out: CEO-years with incomplete or missing real earnings management information Compustat -66
Total CEO years observations useful for regression analysis 9.208
6 Research findings
In this chapter the collected dataset from ExecuComp and Compustat for fiscal years 2005 to 2012 will be analyzed. First the three dependent variable predictions will be examined. Next the descriptive statistics of the research will be outlined. Subsequently the correlation analysis will be performed and will be tested for multicollinearity of the variables. Finally the regression analysis will be performed and the hypotheses will be tested, the results are summarized in the last paragraph of this chapter.
6.1 Regression results
The dependent variables of real earnings management are measured based on the abnormal values of cash flow from operations, discretionary expenses and production costs, as previously mentioned. The abnormal values are predicted based on the previously described formulas in SPSS. Table 7 below shows the coefficients and result that have been obtained from SPSS using the 9208 CEO years and multiple regression.
Table 7 – Regression parameters
A B C Stats Coeff. (p-value) Coeff. (p-value) Coeff. (p-value) Constant 0,090** (,000) 0,141** (,000) -0,209** (,000) -0,995** (,000) 3,442** (,000) -0,245 (,000) 0,022** (,000) 0,130** (,000) 0,841** (,000) 0,001** (,000) n/a -0,172** (,000) n/a n/a -0,001 (0,235) R2 0,042 0,157 0,873 Adjusted R2 F-Test (sign) 0,042 0,000 0,157 0,000 0,873 0,000
*. Significant at the 10% level **. Significant at the 5% level
Table 7 shows the estimated parameters as obtained from the following regression models: A:
The results show for both the abnormal cash flow from operations and the abnormal discretionary expenses a relatively low adjusted R square. This indicates that the dependent variable is only for a small fraction explained by the independent variables. However this is not completely in line with expectations as the adjusted R squares from Roychowdhury (2006) were higher for CFO and discretionary expenses. Below in table 8 a comparison is included between the adjusted R square from this research and the adjusted R square from the result of Roychowdhury (2006).
Table 8 –Adjusted R square Comparison Roychowdhury (2006)
Adjusted R2 this research Adjusted R2 Roychowdhury 0,042 0,45 0,157 0,38 0,873 0,859 Difference 0,408 0,223 -0,014 6.2 Descriptive statistics
Table 9 reports the Descriptive statistics for all variables used in all three regression models. The total dataset consists of 9.208 CEO years over the years 2005-2012. Table 4 reports the following statistics: N (total CEO years), minimum values, maximum values, variable mean values and standard deviations. For CEO compensation included are: the absolute values and the logarithmic values, the absolute values are shown in thousands (x1000).
Table 9 – Descriptive statistics for variables used in the analyses (N = 9208)
Description Min Max Mean Std. Dev
Abn_CFO -1,85 1,02 0,00 0,10650
Abn_DiscExp -3,58 4,66 0,00 0,26640
Abn_Prod -3,02 3,03 0,00 0,25537
TDC1 (USD x1000) 111,0 96.161,0 5.570,62 6286,62
LN_TDC1 4,71 11,47 8,17 0,98527
Total Salary (USD x1000) 1,00 8111,0 822,76 426,88
LN_SalaryComp 0,00 9,00 6,57 0,66928
Total Bonus (USD x1000) 1,00 76.951,0 246,54 1646,68
LN_BonusComp 0,00 11,25 1,35 2,58987
Total equity comp (USD x1000) 2,00 91.694,0 3031,57 4779,31
LN_EquityComp 0,69 11,43 6,30 2,94585
Total assets (USD 1.000.000) 0,00 797.769,0 8.317,73 30.106,74
SIZE 0,00 13,59 7,56 1,66154
MBR (688,33) 759,62 2,62 17,15702
ROA (-16,00) 1,28 0,03 0,22105
Crisis_Dummy 0,00 1,00 0,34 0,47244
Post_Crisis_Dummy 0,00 1,00 0,37 0,48306
Table 9 shows an average CEO total compensation of USD 5.570 k, minimum CEO total compensation of USD 111 k and a maximum total compensation of USD 96.161 k. Furthermore table 4 shows that some CEO’s don’t have equity or bonus compensation (distinguished by the 0 value). The abnormal levels of cash flow from operations and the abnormal discretionary expenses are relatively normal distributed, for which we refer to appendix A. The SIZE variable (log assets) has a mean value of 7.56, which corresponds with an average company size (mean total assets) of about USD 8.317 million. The market-to-book ratio (MBR) has a mean value of 2,62 and shows large positive maximums values and negative minimums values. This is related to negative equity balances in the dataset. The firm performance is measured by the return on assets (ROA) which has a mean of 0,03. Finally the Crisis variables are shown in table 9, as expected they show a minimum of 0 and a maximum of 1 with both a mean of approximately 0,3.
6.3 Correlation matrix
In this paragraph the research variables are tested for correlation coefficients. The Pearson correlation coefficients are expressed in a value between -1 and 1, the higher the value the stronger the correlation between both variables (-1 is a strong negative correlation and +1 is a strong positive correlation). Table 10 reports Pearson correlation coefficients. The results show several significant correlations. The strongest significant negative correlation (-,823) is shown between abnormal production and abnormal discretionary expenses, this is probably related to the fact that the formulas for abnormal production and abnormal discretionary expenses contain matching components.
Table 10 – Pearson correlation coefficients
1 2 3 4 5 6 7 8 9 10 11 12 1 Abn_CFO 1 2 Abn_DiscExp ,063** 1 3 Abn_Prod -,404** -,823** 1 4 LN_TDC1 ,125** -,116** ,049** 1 5 LN_SalaryComp ,008 -,119** ,074** ,541** 1 6 LN_BonusComp ,010 -,003 ,004 ,024* ,025* 1 7 LN_EquityComp ,013 -,014 ,004 -,021* ,005 -,108** 1 8 SIZE ,047** -,321** ,236** ,702** ,481** ,036** ,030** 1 9 MBR ,061** ,031** ,053** ,028** ,008 -,011 -,005 ,011 1 10 ROA ,337** -,019 -,180** ,135** ,089** ,016 ,007 ,134* ,036** 1 11 Crisis_Dummy ,014 ,017 ,000 -,061** -,032** -,009 -,010 -,056** -,028** -,079** 1 12 Post_Crisis_Dummy -,008 -,014 -,012 ,093** ,040** ,003 -,001 ,051** ,001 ,064** -,546** 1 * = significant at 0.05 level ** = significant at 0.01 level
The abnormal CFO, discretionary expenses and productions costs has a significant correlation with the total CEO compensation. Additionally can be noted that “crisis dummy” variable has negative significant correlation with total compensation (-,061) and salary compensation (-,032). Subsequently the post crisis dummy variable has significant positive correlation with total compensation (,093) and salary compensation (-,040). Finally SIZE, market-to-book ratio and ROA also show significant correlations with total compensation and salary compensation. The values of the correlation matrix do not indicate multicollinearity.
6.4 Multicollinearity
The independent variables of the research model are deeper tested for multicollinearity. This test will indicate whether the independent variables explain the same variance of the dependent variables. It might happen that an independent variable do contribute to the explanation of the research phenomenon, but that this variable is incorrectly not seen as significant. This can happen when independent variables show a linear interrelation (this called to as multicollinearity). Table 11 shows the tolerance and VIF values. When the VIF value is greater than 5, there evidence of multicollinearity in the research model (O’Brien, 2007).
Table 11 – Collinearity statistics
Variable Tolerance VIF
LN_TDC1 0,449 2,226 LN_SalaryComp 0,686 1,458 LN_BonusComp 0,987 1,014 LN_EquityComp 0,987 1,013 SIZE 0,490 2,039 MBR 0,997 1,003 ROA 0,972 1,029 Crisis_Dummy 0,698 1,433 Post_Crisis_Dummy 0,696 1,436
Table 11 shows there is no indication for multicollinearity in the research model. All variables can be used in the final regression model.
6.5 Hypotheses testing
In this chapter the formulated hypotheses from chapter three will be tested using three regression analyses. The results will be examined for each individual hypothesis using the regression model as per below. The abnormal levels of CFO, discretionary expenses and production costs are relatively normal distributed, included in appendix B of this research are the histograms that support this statement.
6.5.1 Testing hypothesis 1
H1: Equity-based CEO compensation packages are positively associated with real earnings management as measured by abnormal cash flows from operations.
Table 12 includes the regression results for the dependent variable abnormal cash flow from operations. The regression model shows an adjusted R square of 0,134 which indicates that the model explains 13,4% of the total variance of abnormal cash flow from operations. In other words, there are other factors that affect the use of real earnings management as measured by abnormal cash flow from operations which have not been included in this research model. Table 12 – Regression results for abnormal CFO
Coeff. Error Std. t-value p-value
Intercept -,051 0,12 -4,361 ,000 LN_TDC1 ,020 ,002 12,957 ,000** LN_SalaryComp -,013 ,002 -6,775 ,000** LN_BonusComp ,000 ,000 ,775 ,438 LN_EquityComp ,000 ,000 1,364 ,173 SIZE -,006 ,001 -6,381 ,000** MBR ,000 ,000 4,805 ,000** ROA ,160 ,005 33,810 ,000** Crisis_Dummy ,008 ,003 2,916 ,004** Post_Crisis_Dummy -,005 ,003 -1,796 ,073 Result regression Adjusted R Square 0,134 F-value 158,715 p-value ,000** * = significant at 0.05 level ** = significant at 0.01 level
![Bull, E. 1991. Aided immigration from Britain to South Africa 1857 to 1867. [Boek resensie]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACH5BAEAAAAALAAAAAABAAEAAAICRAEAOw==)