The relation between CEO stock option grants and real earnings management

(1)

0

The Relation Between CEO Stock Option Grants and Real

Earnings Management

Master Thesis Accountancy & Control

Bas Bovenlander

10253270

June 25th 2017, Final Version

M.Schabus

Universiteit van Amsterdam, Amsterdam Business School

(2)

1

Statement of Originality

This document is written by student Bas Bovenlander who declares to take full

responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that

no sources other than those mentioned in the text and its references have been

used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision

of completion of the work, not for the contents.

(3)

2

Statement of Originality

This document is written by student Bas Bovenlander who declares to take full

responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that

no sources other than those mentioned in the text and its references have been

used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision

of completion of the work, not for the contents.

Abstract

In this research I find support for a negative relation between CEO stock option grants and

real earnings management. We model the effect of CEO option grants, according to the

Black-Scholes model, on the deviations of the predicted discretionary expenses. These

abnormal discretionary expenses serve as my measure for real earnings management. Our

prediction, taken out of former research, is that abnormal discretionary expenses and CEO

stock option grants are negatively related. Therefore, a positive relation between CEO stock

option grants and real earnings management is expected. The results of my model disclose the

opposite. These results support a positive relation between CEO option grants and abnormal

discretionary expenses. This study therefore, supports a negative relation between CEO stock

option grants and real earnings management.

(4)

3

1.Introduction

“Grants of stock options represent the largest component of CEO compensation in large publicly traded corporations in the United States, and therefore, it’s of big importance to efficiently allocate pay to performance”. A phrase that Hall & Liebman conclude out of their

paper, published in 1997. However, it was only after the recent financial crisis, that CEO and executive pay became fairly criticized within the economic world and within the academic world. Especially stock options granted to CEO‟s became subject to discussion (Laux, 2012). There are many insights and conclusions on whether granting stock options to CEO‟s are beneficial for a firm or just beneficial for the CEO itself. Jensen and Murphy state in their paper published in 1990 that the optimal situation within a company applies when the agency problem is decreased, towards the point that the incentives of shareholders and managers are aligned. The basic idea of the agency theory sees managers as more risk averse than

shareholders because managers are more depending on a single company (Hall & Liebman 1997). To cope with this risk aversion and align the risk appetite of managers and

shareholder, a company has the possibility to grant equity incentives to management. This can be either through stocks or stock options. In the 90‟s, many companies followed this advice and granted these equity incentives to managers (Hall & Liebman 1997).

Later, there came a big discussion about these equity incentives. Critics said that the idea around granting equity incentives is not bad but it encourages a fixation on stock prices and incentivizes managers to act selfish (Hall & Murphy. 2003, Scott. 2011). Recent

accounting scandals like Enron, WorldCom, Global Crossing and other companies link stock options to investment decisions that contained excessive fixation on stock prices. All

allegedly caused by the escalation in options (Hall & Murphy, 2003).

According to Zingales (2000), success for a firm nowadays, is increasingly depending on intangible assets such as research & development. However, managers fail to invest in discretionary accounts because they are too fixated on maximizing firm's short term stock price. Former research found substantial evidence that this fixation on short term stock prices tempts managers to manage earnings (Roychowdhury, 2006). One means of earnings

management is by manipulation of accruals with no direct cash flow consequences. In this study, the focus is on the earnings management that does directly influence cash flow namely

(6)

5

real earnings management, hereafter also referred to as REM. We aim solely on REM because Cohen, Dey & Lys (2008) and Cheng & Warfield (2005) do not find a significant association between accrual-based earnings management and exercisable vested stock options.

This study builds on the foundation of Core & Guay (1999) and Roychowdhury (2006). Core & Guay (1999) try to determine if stock options influence the incentives of CEO‟s when making investment decisions. In this study I will also discuss the effect of stock options on managerial investment decisions. There has been a lot of research done on the effect of equity incentives on earnings management after 2006. However these studies discuss earnings management of equity incentives as a whole, only for CFO‟s or do not link stock options to REM. My paper contributes to the literature on earnings management

because I will narrow my research to the effect of solely stock options granted to CEO‟s, and whether the incentive to maximize short term stock prices, influences discretionary

investments like R&D, advertising and SG&A. The main focus will be on total discretionary expenses, but I will make a distinction between R&D, advertising, selling, general and administrative as well. Advertising and SGA expenses are seen less important than R&D but managing these expenses is still an effective way to manage short term earnings

(Roychowdhury. 2006). I will not discuss the equity incentives of stocks granted to CEO‟s itself. The reason is that, options do not bear the downside of a drop in stock price, and I want to examine if this incentive is enough to engage in earnings management.

In the next paragraph I will further elaborate on the agency theory, the theory behind the incentives of earnings management and earnings management itself. I will discuss what is already examined and build on that. After that, I will draft hypotheses and develop a

regression that will be tested according to a quantized test in paragraph 4. In the last

paragraph I will draw a conclusion on the results of my test. The research question that I will conclude is: Does granting stock options to CEO’s enhance Real Earnings Management, by

(7)

6

2. Theory

2.1 Features of stock options

Hall & Murphy (2003) describes a stock option as a contract that gives an employee, or in this case a CEO, the right to buy a share in the company at a pre-specified exercise amount. By giving managers stock options, the company is trying to attract and motivate managers without spending any direct cash. These options are structured to keep managers and

employees in the firm by providing them with retention incentives. The manager, or CEO, is incentivized to stay at the firm because the right can only be exercised after a period as agreed upon on the date the option was granted. This period is called a vesting period. A normal stock option for employees has an expiring date of about ten years. The option cannot be exercised immediately but only over time. This means that for example 10% could be exercised every year for the upcoming 10 years. When the stock option is exercised then the option is said to be vested.

When a CEO or an employee exercises the stock option, the company issues new shares. These shares increase the total number of shares outstanding. Some companies on the other hand decide to offer the CEO or employee a cashless exercise program. In this case the CEO or employee does not pay for the stock options granted but simply receives the

difference in market price of the shares and the exercise price in cash. Because the stock option cannot be exercised all at once, the CEO or the employee could make a profit every year by exercising a percentage of the stock option in cash. Which percentage can be exercised has to be determined on forehand on the grant date (Hall & Murphy 2003).

Stocks granted to CEO‟s are seen to have similar incentives as granting stock options to CEO‟s. Wu & Tu (2006) discuss this in their research. They state that stock ownership has a linear relationship with stock price, as does stock options. The difference is that the

downside risk of a drop in share prices may discourage executives from risk-taking decisions or earnings management when granted stock. Stock options, on the other hand, do not result in real and immediate wealth reduction given that the stock price is dropping. Thus, the bar to accomplish earnings management is way lower when granted stock options opposite to stocks itself.

(8)

7

2.2 Agency Theory.

As mentioned in my introduction, Jensen and Murphy (1990) state in their paper that stock options are mainly granted to decrease the agency problem within a firm. There are two important players within firms. Management acts as an agent and shareholders act as the principal. The agent acts on behalf of the principal and is supposed to make decisions that increase the wealth of the principal, next to maximizing his own wealth. This means that Shareholders are trying to maximize their profit, by making an effort to raise share prices as much as possible. Managers on the other hand, try to support and improve day to day operations so that the firm maximizes profit (Tzioumis, 2008).

The agency problem comes about when information asymmetry exists. Information asymmetry means that management has a lot more information about the day to day

operations than shareholders. Shareholders are not able to monitor if the right decisions are made by management to maximize their wealth. There is a chance that because of this difference in information and lack of monitoring, management only tries to enhance his own wealth instead of making decisions that are in the best interest of the shareholders. Jensen & Meckling (1976) call this the problem of adverse selection. In this situation, the primary objective of managers is to mislead shareholders about the underlying economic performance of the firm. A conflict arises when management has private information and earnings

management is unlikely to be transparent to outsiders. This manipulation is called moral hazard and could be deadly for the firms‟ performance when management uses information asymmetry to pursue negative NPV projects that benefit themselves but harm the firm (Jensen & Meckling. 1976).

Agency theory is based on the fact that managers are seen to be more risk averse than shareholders. CEO‟s depend on a single firms‟ welfare. Not only their economic welfare is tied to the firm but also their reputation is closely tied to the wellbeing of the managed firm. The reputation of management will take a bigger hit than shareholders‟ reputation when their company stumbles or fails. Shareholders spread their risk by engaging in a broader and widely diversified investment portfolio. Therefore, shareholders are less affected by the downfall of one single company. Overall shareholders tend to be risk neutral and understand that risk is required to be taken for a higher return. Therefore they try to influence

management to pursue less risk averse projects (Sanders & Hambrick. 2007).

(9)

8

aversion of management so that the risk appetite and incentives of shareholders are aligned with those of management. Hall & Liebman state that granting stock options to CEO‟s is therefore the most direct solution to the agency problem. Theorists have shown that stock options shape managerial perceptions and behavior within the company. Devers, Wiseman, and Holmes (2007) examined how option holders subjectively revise their valuations of their options by looking at the interaction of companies' recent stock price trends and volatility. Out of this finding the important insight can be concluded that managers continuously update their assessments of the value of their options.

2.3 Earnings Management

2.3.1 Types of Earnings Management

As stated in the introduction, Hall & Murphy (2003) name granting stock options to CEO‟s as an instrument to cope with the agency problem but it also encourages earnings management. By providing stock options, the firms create a direct link between realized compensation and company stock price performance. Problem hereby is that CEO‟s start to develop an

excessive fixation on stock prices.

Scott (2011) states in his paper, that because of the new fixation on stock prices, managers are given incentives to be selfish. Selfish incentives can be seen as a higher reputation, a higher bonus, showing good performance or in this case, receiving a higher compensation out of stock options. Managers are determined to reach benchmarks because they are rewarded for reaching these targets. To reach these benchmarks they are tempted to manipulate the numbers of the period to make the financial reporting look better to exactly reach the target. Healy (1985) even says that in some situations earnings are deferred when target is already reached to prevent impossible target in upcoming years.

There are two types of earnings management namely accruals manipulation and REM. One means of earnings management is by manipulation of accruals. Accruals are accounts on a balance sheet that represent liabilities and non-cash-based assets used in accrual based accounting. These accounts include, among many others, accounts payable, accounts receivable and future interest expense. Accruals are characterized by the fact that it has no direct cash flow consequences. It is used by accountants to disclose revenue that has been

(10)

9

earned but is not yet recorded in the accounts and expenses that have been incurred but are not yet recorded in the accounts. The opportunity to manage earnings with accruals comes about when the reported income includes both cash flows and changes in firm value that are not yet reflected in current cash flows (Bergstresser & Philippon, 2006, Gunny, 2005). Real earnings management on the other hand, is significantly different than accrual-based, as real based earnings management has direct cash flows effects. It is accomplished when managers undertake actions that deviate from the first best practice to increase reported earnings or when managers time investment decisions to alter and subset earnings (Gunny, 2005). Cohen & Zarowin (2010) analyzed the trend of earnings management and names that the aftermath of accounting scandals like Enron and Worldcom imposed by Sarbanes-Oxley Act (SOX) may have changed managers‟ preferences on earnings management. After SOX they find that option-based incentives actually decreased after SOX but managers preferred real actions managed earnings above accounting manipulation. He survey of Graham et al. (2005) adds, that managers indeed prefer real activities manipulation, by such means as reducing discretionary expenditures or capital investments, over accruals manipulation as a way to manage earnings. One reason is that even though real earnings management is more costly it is also harder to detect. In the next paragraph I will discuss the former research on real earnings management in relation to discretionary expenditures.

2.3.2 Former research on REM in relation with discretionary expenditures

In the last century, firms are predominantly capital-intensive but nowadays the higher competitiveness for firms, has increased the dependency on other factors and forms of investments as well. According to Zingales (2000), success for a firm nowadays, is increasingly depending on intangible assets such as research & development. However, managers fail to invest in discretionary accounts because they are too fixated on maximizing firm's short term stock price. Zingales (2000) even calls this the first order problem of firms nowadays. Investing in long term intangible assets brings short term costs, which depress short term earnings, so that managers can‟t reach short term benchmarks (Edmans, 2009). Roychowdhury (2006) agrees to this. He finds support for the fact that firms use real earnings manipulation so that financial reporting benchmarks are reached. Discretionary expenditures such as R&D, advertising and SG&A are generally expensed in the same period that they occur and do not immediately generate income. Therefore a firm can cut

(11)

10

discretionary expenditures to temporarily boost income and improve reported margins. Roychowdhury (2006) names Graham et al (2005) as supporter of this theory. In the survey of Graham et al (2005) a large number of respondents admit that they prefer reducing discretionary expenditures than other capital investments.

Furthermore, Harvey, and Rajgopal's (2005) state in their paper that an increase in CEO incentives, goes hand in hand with a decrease in long term investments. Their survey finds that 78% of executives would sacrifice long-term value to meet short earnings targets. CEO behavior towards research and development (R&D) spending may provide insight into CEO pay deviations and CEO behavioral issues, given that R&D spending is directly

influenced by the CEO and is related to both firm performance and CEO pay (Cheng, 2004). Fong (2010) hereby names that R&D expenses are negatively associated with CEO pay. The CEO has the obligation towards shareholders to reach short benchmarks because shareholders want to raise stock prices to maximize their profit. Fong (2010) states that the CEO is

pressured into making short term decisions to maximize these short term earnings. If the incentive of stock options to the CEO is being added, the CEO also has the intrinsic pressure to boost his compensation in the short term. Therefore CEO‟s manipulate discretionary expenses expenses to maximize their compensation.

Last, Edmans, Fang & Lewellen examined managerial investment behavior in their paper from 2013, linking stock options granted to CEO‟s, with the amount of short term investment opposite to long term investments. They came to the conclusion that especially the year before the company allows the CEO to exercise the option, long term discretionary expenses like R&D and Advertising declined significantly. Next to the fact that in these years‟ executives were more likely to meet or just slightly beat the earnings forecasts of analysts. According to Edmans, Fang & Lewellen this suggests that CEO‟s tried to manage real earnings in the year before exercising to maximize profit out of the option.

2.4 Hypotheses

In former paragraphs I explained that sometimes a conflict of interest exists between managers and shareholders within a firm. This occurs because managers act like agents for the shareholders but overall managers are more risk averse than shareholders. To make managers less risk averse, stock options are granted to align the risk appetite of managers and shareholders. Scott (2011) adds that stock options also give an incentive to act selfish and

(12)

11

even manage earnings in their favor. According to Hall & Murphy (2003) this excessive fixation on stock prices even leaded to recent accounting scandals like Enron, WorldCom, Global Crossing and other companies. Within these scandals, stock options are linked to investment decisions that contained excessive fixation on stock prices. All allegedly caused by the escalation in option grants. This fixation on short term stock prices tempts CEO‟s to act into REM by decreasing discretionary expenses temporarily to boost short term earnings and therefore will boost short term stock prices. Higher stock prices mean a higher

compensation, after every period that a part of his option is vested.

In this paper I use data from public traded US firms from the timeframe 2003 till 2006. I use this timeframe because 2003 is the first year that SOX became applicable to public traded companies in the US. From this point, US public firms had to disclose a lot more information in their financial report (Cohen & Zarowin, 2010). Next to that, the years 2003 till 2006 have data available about the Black-Scholes value of options granted to CEO‟s as used by Core & Guay (1999). The years after 2006 show zero observations on the Black-Scholes value of options grants to CEO‟s that‟s why will not use these years. I use US firms because big scandals like WorldCom, Enron and Global Crossing happened in the US and I want to determine if CEO‟s investment behavior is indeed affected by stock options in this timeframe and country (Hall & Murphy, 2003).

To come to a conclusion on whether investment behavior of CEO‟s is influenced by stock options granted, I compose one hypothesis and three sub hypotheses. Out of the hypotheses, I determine the presence of REM. I can support that REM has a significant relation with CEO option grants, when the results support the hypotheses.

Roychowdhury (2006) states, that discretionary expenditures, are seen as the sum of R&D, advertising and SG&A. These are all expenditures that are generally expensed in the same period and do not generate income in the same period or short term. Therefore firms can engage into real earnings management by reducing these expenditures in order to decrease short term expenses and boost short term income. The first hypothesis that I test is therefore:

H1: CEO stock options grants of a public traded US firms, have a positive relation to REM by decreasing the total abnormal discretionary expenses

For my sub hypotheses I want to support the findings of H1 by splitting the discretionary expenses into three expenses namely R&D expenses, advertising expenses and SG&A

(13)

12

expenses. By looking at the discretionary expenses separately, and as a total, my findings will not only represent the effect of CEO stock grants on the total amount of discretionary

expenses, but also the expenses on its own. This makes my results stronger.

According to Zingales (2000), success for a firm nowadays is increasingly depending on investment in R&D. Thus it is important that these investments are not manipulated by CEO‟s because it could harm the firm. The first sub hypothesis is therefore tested on research & development. Here research & development is tested on whether it decreases significantly when CEO‟s are given short term incentives in the forms of stock options. My first

hypothesis is therefore:

H1A: CEO stock options grants of a public traded US firms, have a positive relation to REM by decreasing the abnormal R&D expenses.

Second, success for a firm nowadays is increasingly depending on investment in R&D. On the other hand, I can only conclude that REM fully exists when H1 is accompanied by a significant decrease in advertising and selling, general & administration expenses, when stock options are granted to CEO‟s. My second and third sub hypotheses are therefore:

H1B: CEO stock options grants of a public traded US firms, have a positive relation to REM by decreasing the abnormal advertising expenses.

H1C: CEO stock options grants of a public traded US firms, have a positive relation to REM by decreasing the abnormal SG&A expenses.

(14)

13

3. Data and methodology

The sample used in my test consist of all US public traded firm from after the introduction of SOX. This is the timeframe of the years 2003 till 2006 and is gathered from COMPUSTAT integrated with the EXUCOMP database. I keep all observations with data on CEO options grants. The companies where the financial data did not match with the information provided about the CEO‟s, were eliminated. Next to that, I Winsorize all the data used in my model to eliminate outliers. My sample consists of 1310 US public traded firms and 2854 observations over 4 fiscal years.

3.1 Independent variable

In a former research, Core & Guay (1999) defines equity incentives as the change in the dollar value of the CEO‟s stock options for a change of 1% change in the stock price. It is not too difficult to compute this measure of incentives for stockholdings, because stock value increases by 1% for each 1% increase in the stock price. Disclosing the incentives provided by stock options is the harder part. The value change percentage of an option is less than the value change percentage of the stock price. This lower value increase exists because

managers are seen more risk averse than shareholders and employee stock options are non-transferable. Therefore managers are more likely to exercise their options sooner than is assumed by the Black-Scholes model. Next to that, the value of a stock option also depends on the parameters embedded in the option contract which gives it more restrictions (Core & Guay. 1999).

Core & Guay (1999) name that there are two ways to measure the stock options incentive, given to managers to be less risk averse and to maximize short term share price. First measure that Core & Guay use in their paper is the sensitivity of an options‟ value, as the representative of the incentive. Core & Guay (1999) call this the Option Delta. The second measure is the options grants value according to the Black-Scholes model. Core & Guay (1999) name this the appropriate risk neutral valuation for an executive stock option to account for dividend payouts. This model calculates the percentage value stock price opposite to stock the price by using six inputs: stock price, exercise price, time-to-maturity, expected stock-return volatility, expected dividend yield and the risk-free rate. To get the dollar value

(15)

14

of the stock option, the outcome of the Black-Scholes model is multiplied the firms‟ stock price. When calculating the new incentives given to CEO‟s, it is required to take measures of new equity incentives granted during the year. New equity incentives granted during the year are equal to the sum of options granted to CEO‟s, in the year according to the Black-Scholes model. Core & Guay (1999) find, that the average increase in value for a typical newly granted long term executive stock options is $0.75, when a stock price increases with $ 1.00. In this paper I will not calculate the Scholes value myself but I will use the Black-Scholes value provided by Execucomp. The Black-Black-Scholes value shall be scaled by lagged assets to fit the ratio presented by the dependent variable. This will be my independent

variable for testing the incentive given to CEO‟s to act into REM. In my model I will call this variable CEO_OptGrants.

3.2 Dependent variables

According to Roychowdhury (2006), discretionary expenditures are seen as the sum of R&D, advertising and SG&A. These are all expenditures that are generally expensed in the same period and do not generate income in the same period or short term. Therefore firms can engage into real earnings management by reducing these expenditures to decrease short term expenses and boost short term income.

In the paper of Roychowdhury (2006) abnormal discretionary expenses are used as the dependent variable to determine the effect on REM. Abnormal discretionary expenses are the actual discretionary expenses minus, the „„normal‟‟ discretionary expenses calculated using estimated coefficients from the corresponding industry year model and the firm sales t-1

and assets t-1. This means that abnormal discretionary expenses can be measured as the

deviation from the predicted values or normal values from the current fiscal year, also expressed as the residual of the „„normal‟‟ discretionary expenses. The normal discretionary expenses could be expressed as a linear function firms‟ sales t-1 and assets t-1. The function of

the „„normal‟‟ total discretionary expenses would then be: ( ) ( )

(16)

15

period, the function could show unusual low residuals. To avoid this problem, total

discretionary expenses are expressed as a function of lagged sales. The regression that will be run is: ( ) ( )

where At-1 are the total assets at the end of last fiscal year also known as the lagged assets. St-1

are the sales at the end of last fiscal year also known as the lagged sales. DiscExpTott are the

total discretionary expenses in period t (Roychowdhury. 2006, Dechow et all. 1998).

After that I split the discretionary expenses into research & development expenses (XRD), advertising expenses (XAD) and selling, general & administration expenses (XSGA). With these linear functions I can eventually test H1A, H1B and H1C on whether CEO option grants affect expenses separately. The function for the separate “normal” R&D expenses to test H1A is therefore:

( ) ( )

The function for the “normal” advertising expenses to test H1B is therefore: ( ) ( )

“normal” selling, general and administration expenses to test H1C is therefore: ( ) ( )

I will take these residuals ( ) that represent the abnormal discretionary expenses ratio opposite to lagged assets. I will use this abnormal expense in my regression as my dependent

(17)

16

variables. With these abnormal discretionary expenses I can test if stock options granted incentivizes CEO‟s to engage in REM by significantly decreasing discretionary expenses within a fiscal year. I use this abnormal total discretionary expenses ratio to test H1 and name this dependent variable AbnDiscExpTot. I use the abnormal R&D expenses ratio to test H1A and name this dependent variable AbnXRD. I use the abnormal advertising expenses ratio to test H1B and name this dependent variable AbnXAD. I use the abnormal SG&A expenses ratio to test H1C and name this dependent variable AbnXSGA.

The expected effect of an increase in abnormal discretionary expenses is that it decreases net income because higher expenses mean a lower net income. A decrease in net income would mean that REM decreases as well. In my paper, I am testing whether there is an increase of REM, when CEO‟s are granted with stock options. To make my regression more fitting to my hypotheses I need to take in consideration that an increase in

AbnDiscExpTot, AbnXRD, AbnXAD and AbnXSGA means that earnings decrease and

therefore has a negative effect on REM. This means that my hypotheses are only supported when the relation between CEO options grants and abnormal discretionary expenses is negative. This creates a situation where the relation towards an increase in REM can be tested.

I expect that my research will be more precise than former research because I am testing the effect on discretionary expenses as a total and separately. I can support that there is a significant relation to all of the discretionary expenses and CEO stock option grants when all hypotheses are supported.

3.3 Control variables

In this paragraph I discuss the control variables needed to control for size and growth so that my test is not affected by these factors and therefore is more precise. To fit my control variables to my dependent variables I scale most of my control variables by lagged assets to present ratios that can be compared with the ratio of abnormal discretionary expenses opposite to lagged assets.

Assetst-1: Are the total assets of a firm within fiscal year t-1. This control variable is a good indicator to control for size because a firm with higher assets is expected to have higher discretionary expenses overall (Roychowdhury, 2006). Here the variable Assetst-1 is logged because I expect that there is a big difference between the small observations and the big

(18)

17

observations.

Emp: is the size of a company based on the amount of employees that a firm has

multiplied by 1000. I think that the amount of employees within a firm is a good indicator to show the size of the firm. A firm that has more employees needs to make higher investment to cope with employee costs. Therefore, Emp is expected to have a positive relation with discretionary expenses.

NetIncome: Is the net income of a firm within the fiscal year t. According to

Roychowdhury (2006) the NetIncome variable is a good indicator to control for growth and size. Next to that it is expected, that there is a positive relation between the net income generated within last fiscal year, and discretionary expenses overall. This means that a firm with higher net operating income in t-1 means a firm will invest more in R&D, Advertising and XSGA (Roychowdhury, 2006). This can only be the case when scaled by lagged assets so it is similar to the return on assets (ROA). Therefore net income is also scaled by lagged assets in this paper (Edmans, 2009).

DivTott: Are the total dividends payed out to stock holders within the fiscal year t. Core & Guay (1999) name the Black-Scholes model the appropriate risk neutral valuation for an executive stock option to account for dividend payouts. In my model I expect that stock influences abnormal discretionary expenses. Therefore it can be expected that dividends on have an influence on discretionary expenses as well. I scale the DivTott by lagged assets fit the dependent variable and scale for size.

CEObonus: the control variable CEObonus is used to control for the fact that it can be

expected that the incentive to engage in earnings management is higher when a CEO earns a lower bonus last fiscal year and therefore has the incentive to reach a higher bonus

(Edmans, 2009). To take last year into consideration, the variable is scaled by lagged assets . EPS: stand for the earnings per share ratio. It is expected that this has a negative

relation with abnormal discretionary expenses because expenses decrease income

(Roychowdhury, 2006). Edmans (2009) expects that stock option grants increases the CEO‟s stock price concerns. The CEO will try particularly hard to avoid announcing earnings per share below analyst expectations, since doing so typically leads to a large price drop. Thus, the CEO should be particularly likely to beat the consensus forecast, either by cutting investment or by engaging in other short term actions, such as managing discretionary expenses (Edmans, 2009).Edmans (2009) uses a dummy in his paper to measure the effect of stock options when EPS is above or below mean analyst consensus forecast but I will use the overall EPS because my paper does not take analyst forecasts into consideration.

(19)

18

CeoAge: Is the age of the CEO. Former studies show that an older CEO‟s has higher moral standards and overall has more experience in the field to deal with incentives.

Therefore an older CEO is expected to be less sensitive to given incentives and therefore REM (Bergstresser & Philippon, 2006).

DummyGender: Reflects whether the CEO is male or female. Peni & Vähämaa (2010)

find that that female CEO‟s are following a more conservative earnings management

strategy. This research finds that men are more vulnerable to engage in earnings management and manipulate investments to boost their own compensation.

3.4 Econometric model

The regression that will be tested in this paper to test H1, H1A, H1B and H1C is:

Where the dependent variable:

-Y1a = the total abnormal discretionary expenses lagged by assets (AbnDiscExpTot) in fiscal

year t when H1 is tested

-Y1a = the abnormal R&D expenses lagged by assets (AbnXRD) in fiscal year t when H1A is

tested

-Y1b= the abnormal advertising expenses lagged by assets (AbnXAD) in fiscal year t when

H1B is tested.

-Y1c= the abnormal selling, general and administration expenses lagged by assets

(AbnXSGA) in fiscal year t when H1C is tested.

If the p-values of my regression show that p<0.1, I can reject the H0 of my model. From here out I conclude whether I support the existence of a relation between CEO options grants and REM by decreasing discretionary expenses.

(20)

19

3.5 Descriptive statistics

3.5.1 Descriptive statistics total discretionary expenses

Table 1a disclose the descriptive statistics for abnormal total discretionary expenses and their hypothesized determinants. In the descriptive statistics the number of observations, the mean, standard deviation, skewness, kurtosis, and the value of the percentile 1, 5, 50, 95, 99 of every variable is shown.

First it is good to notice that half the variables stated in the regression and in table 1a, are ratios opposite to the assets of the fiscal year t-1. We use the ratios because this is

supported by the paper of Roychowbury (2006). The value mean of the ratio observations, are expected to be between 0 and 1. This is also what is represented in table 1a when looking at the average value (Mean) and the middle value (Median) of every variable. The mean of the ratio variables AbnDiscExpTot, NetIncome, DivTot and CEObonus, are respectively 0.606, 0.108, 0.023 and 0.851. CEO_OptGrants is also a ratio but the mean of this variable is 2.138. This means that on average, the Black-Scholes value of stock options granted to CEO‟s is on average more than two times as big as the Assetst-1. This represents that the

Black-Scholes of options is a sufficient amount. Before Winsorizing the variable, the mean was, 3.019 which was even higher. Next to that, the mean value of the Black-Scholes options is 2683.152 and the mean of Assetst-1 is 13829.2. This would have been a ratio of 0.194 when

means were taken separately. On average the Black-Scholes value of CEO options grants opposite to Assetst-1 would then have been 19.4%. The high mean of the CEO_OptGrants

variable can be explained by firms with extremely high options values opposite to Assetst-1.

Even after Winsorizing thisis still present.

The mean of the other ratios seem logical. The average value of the total dividends is 2.3% of lagged assets. This seems pretty low but can also be explained by the fact that a lot of firms did not pay out dividends at all. These firms were given the value 0, which has a decreasing effect on the DivTot ratio and therefore the mean. This is also why the p01 and the p5 are 0 for this variable. For CEObonus, the mean is 0.851. This is not an oddly high

number because there a US public firms that give out high bonuses overall especially when you take in consideration that the Black-Scholes value of CEO option grants are pretty high as well.

The mean and median of the independent variable AbnDiscExpTot are also positive, namely 0.606 and0.091. As stated in paragraph 3.2. Important is, that the variable represents

(21)

20

a positive value, so that an increase in abnormal discretionary expenses should decrease income and therefore represents an decrease of REM. Important to know is that only when

AbnDiscExpTot decreases, an increase in REM could be present.

The value of the mean itself does not present an odd value. When assets increase by 1 unit, discretionary expenses increase by 0.606 units. This is logical because it is likely that a bigger company has more overall discretionary expenses.

For the variable Assetst-1 the Mean and the Median are way higher than the rest of the variables. This can easily explained by the fact that this is the total value of lagged assets without manipulating the numbers or scaling it. The Mean and Median are not considered unreasonably high. In the paper of Roychowbury (2006) the total assets Mean and Median are also around that size. Here Assetst-1 is the variable that especially controls the other variables on size. We log the lagged assets because the Skewness and the Kurtosis are way too high. According to Trochim & Donnelly, 2006, the acceptable limit of skewness contains a value between 2 or -2. If the skewness is higher than 2 or lower than -2, the data needs be adjusted or explained. The Skewness and Kurtosis of lagged assets is 0.278, respectively 2.689 after the observations are logged.

Furthermore, DummyGender is a dummy variable, thus it can only contain a value of 0 or a 1. The mean of the DummyGender variable, shows that the vast majority of CEO‟s are male. On average that 98.1% of the CEO‟s within a public traded US firms is male. This can be considered a high percentage. This is also the reason why the variable observations are negatively skewed because the values are leaning towards 1. Martin, Nishikawa & Williams (2009) see this in their paper as well. They specify that only eight Fortune 500 firms in 2005 had women CEOs. This represents a slight increase in female CEOs opposite to the year 1997. During this year, only one of the fortune 500 firms had a female as their CEO. An average of 98.1% is therefore not a strange outcome.

3.5.2 Descriptive statistics separate discretionary expenses

In the former paragraph, I discussed the descriptive statistics of the total discretionary expenses. In this paragraph I will focus on the separate abnormal discretionary expenses, by splitting the total abnormal expenses in R&D, advertising and XSGA. The separate expenses are scaled by lagged assets so they also present a ratio opposite to lagged assets.

(22)

21

expenses. The p.1 of the R&D and advertising expenses are both 0 because in some cases one separate expense was 0 but another expense was above 0. A 0 value still gives information when calculating the total expenses; therefore I did not drop de 0 value.

Table 1a shows that the mean of the Abnormal R&D ratio is much lower than the mean of the abnormal XSGA ratio and abnormal advertising ratio. This represents that abnormal R&D cost item overall have a lower value opposite to assets than XSGA and advertising. This does not support the theory of Zingales (2000), described in paragraph 2.4. In his paper, he states that success for a firm nowadays is increasingly depending on

investment in R&D. Our numbers do not explicitly support that because on average the Mean of R&D expenses, opposite to lagged assets are lower than the other discretionary expenses. This is also the reason why I do not only test R&D expenses but all discretionary expenses. The descriptive statistics confirm the importance of all these investments not being

manipulated by CEO‟s.

(23)

22

Table 1a, descriptive statistics of the variables, total discretionary expenses and control variables

Variable Obs Mean Median Standard

deviation Skewness Kurtosis P1 P5 P95 P99 AbnDiscExpTot/At-1 2,854 0.606 0.091 1.439 3.987 21.805 -0.139 -0.138 3.155 9.598 CEO_OptGrants/At-1 2,854 2.138 0.489 4.983 4.419 24.883 0 0 9.683 33.967 Assetst-1 2,854 13829.2 1900.449 77049.94 18.44972 421.9648 46.129 150.89 44260.3 189286 EMP 2,854 16.794 5.016 33.050 3.836 19.890 0.140 0.336 75.036 214.500 NetIncome/At-1 2,854 0.108 0.0535 0.301 5.354 36.510 -0.373 -0.066 0.352 2.314 DivTot/At-1 2,854 0.023 0.002 0.084 6.778 51.633 0 0 0.066 0.704 CEObonus/At-1 2,854 0.851 0.327 1.672 4.448 25.765 0.003 0.014 3.178 11.607 EPS 2,854 1.695 1.450 2.111 1.081 7.071 -4.430 -1.090 5.500 10.380 CEOage 2,854 55.131 55 6.705 -0.010 2.523 40 44 66 71 DummyGender 2,854 0.981 1 0.138 -6.994 49.910 0 1 1 1

(24)

23

Variable Obs Mean Median Standard

deviation

Skewness Kurtosis P1 P5 P95 P99

AbnXRD/At-1 2,854 0.062 0.0605 0.033 2.265 13.782 0 0 0.119 0.230

AbnXAD/At-1 2,854 0.291 0.026 0.019 1.270 5.595 0 0.005 0.066 0.891

AbnXSGA/At-1 2,854 0.321 0.136 0.564 3.868 20.948 -0.031 -0.011 1.326 3.724

(25)

24

4. Results

4.1 Bivariate analysis

4.1.1 Correlation total abnormal discretionary expenses model

Table 2a describes the correlation between the independent variables of the regression. Multicollinearity is present when two or more independent variables are highly correlated with each other. This happens when the variables have a strong linear association between 2 them, so that they strongly influence another. Huber & Stephens assume in their paper published in 1993, that the acceptable limit of correlation is between -0.5 and 0.5. When the correlation is above 0.75 or lower than -0.75, the correlation is not tolerated. They make an exception for “tolerance”. This is accepted when 1 - multicollinearity is higher than 0.25 or lower than -0.25.

In table 2a the correlation between the independent variables is disclosed. Almost all correlations are within the tolerated interval. The correlation between AbnDiscExpTot and

Assetst-1 is above the limit of 0.5 namely 0.665. This correlation is pretty high and the existence of multicollinearity can be considered here. A good explanation for this high correlation is that AbnDiscExpTot present a ratio, scaled by lagged assets. The strong

correlation between the variables is therefore not strange. Next to that, a high correlation does not have to mean that multicollinearity exists. An additional check, next to correlation, can be done via the variance inflation factors (VIF‟s). VIF‟s are often used to detect

multicollinearity. When a VIF of an independent variable is above 5 or above 10 it is highly possible that multicollinearity exists. The VIF of Assetst-1, calculated in my model, is 1.79. This VIF is much lower than the critical value of 5 or 10 so it can be supported that no multicollinearity is present here.

Overall, all variables scaled by lagged assets, seem to have a higher correlation. This is present between CEObonus and AbnDiscExpTot with a correlation of -0.6392, between

CEO_OptGrants and CEObonus with a correlation of 0.631, between DivTot and NetIncome

with a correlation of 0.706 and CEObonus and between NetIncome with a correlation of 0.514. All correlations are higher than 0.5 but this can be explained by the fact that these variables are scaled by lagged assets.

(26)

25

A more logical explanation for the high correlation between CEO_OptGrants and

CEObonus is that CEO_OptGrants is a part of the total CEO bonuses. CEO bonuses are

namely the sum of cash bonuses and stock bonuses given to CEO‟s. Hall & Liebman (2003) also state that stock options contain a big part of total CEO compensation. Therefore it is not odd that CEO_OptGrants and CEObonus strongly correlate. The VIF is 2.11, so smaller than the limit of 5. This supports the statement that no multicollinearity exists between these variables

The high correlation between DivTot and NetIncome can also be explained because dividends are strongly depending on net income. Without a form of income, a firm is not able to give out dividends. If a firm has a higher net income there is more liquidity to payout dividends. I my model it is not a bad thing that CEO_OptGrants and CEO_OptGrants are highly correlated. This means that in my model these two variables vary together. In my model the sign of the correlation is negative which means that high scores on the first thing are associated with low scores on the second. When the CEO_OptGrants has high scores

AbnDiscExpTot has high scores. This is in line with my H1 because this effect would mean

higher earnings and therefore would support an increase in REM.

Overall, I can be support that even though correlation between some variables is higher than normal, multicollinearity is not present between these variables. This is

confirmed by the variance inflation factors (VIF). The highest VIF calculated in my model is the VIF of NetIncome. This VIF is calculated at a value of 2.70. This VIF is much lower than the critical value of 5 or 10 so it can be supported that no multicollinearity is present in my model at all.

4.1.2 Correlation separate abnormal discretionary expenses model

Table 2b describes the correlation between the control variables of the my regression and the separate abnormal discretionary expenses variables. In paragraph 4.1.1, mainly the ratios presented high correlations. This is supported by table 2b. In table 2b especially

CEO_OptGrants, CEObonus and Assetst-1 show a high correlation. Explanation is similar to the one from paragraph 4.1.1. These high correlations are present, because the variables are all scaled by lagged assets.

The positive correlation of AbnXAD and EMP also stands out because AbnbXRD and AbnXSGA have a negative correlation opposite to the amount of employees. Advertising

(27)

26

increases when there are more employees, while R&D and SGA decrease when there are more employees in the firm.

Further, nothing really stands out. Only that the correlation of DummyGender is really low opposite to all the separate abnormal discretionary expenses. This is similar as to the total abnormal expenses. This represents that the separate abnormal discretionary expenses are not hugely influencing each other.

(28)

27

Table 2a, correlation between the variables, total abnormal discretionary expenses model

Variable AbnDiscExpTot CEO_OptGrants Assetst-1 EMP NetIncome DivTot CEObonus EPS CEOage DummyGender

AbnDiscExpTot 1.000 CEO_OptGrants 0.573*** 1.000 Assetst-1 -0.665*** -0.444*** 1.000 EMP -0.151*** -0.036* 0.369*** 1.000 NetIncome 0.287*** 0.457*** -0.216*** 0.177*** 1.000 DivTot 0.216*** 0.301*** -0.132*** 0.210*** 0.706*** 1.000 CEObonus 0.639*** 0.631*** -0.488*** -0.020 0.514*** 0.387*** 1.000 EPS -0.184*** -0.100*** 0.267*** 0.127*** 0.231*** 0.075*** -0.003 1.000 CEOage -0.051*** -0.068*** 0.100*** 0.076*** 0.029 0.045** 0.029 0.103*** 1.000 DummyGender -0.019 -0.024 0.017 0.005 0.001 0.004 -0.024 0.044** 0.078*** 1.000

(29)

28

Table 2b, correlation between the variables, separate abnormal discretionary expenses model

Variable AbnXRD AbnXAD AbnXSGA

CEO_OptGrants 0.484*** 0.313*** 0.573*** Assetst-1 -0.545*** -0.513*** -0.683*** EMP -0.177*** 0.031*** -0.143*** NetIncome 0.213*** 0.220*** 0.293*** DivTot 0.165*** 0.153*** 0.219*** CEObonus 0.509*** 0.399*** 0.643*** EPS -0.189*** -0.076*** -0.183*** CEOage -0.027 -0.049*** -0.053*** DummyGender -0.016 -0.002 -0.018

(30)

29

4.2 Multivariate analysis

This section will discuss the results of the multivariate regression. In paragraph 4.2.1, I will discuss the results of the regression, with the total abnormal discretionary expenses as my dependent variable. In paragraph 4.2.2, I will discuss the results of the regression with the separate abnormal discretionary expenses as my dependent variable. From these results I will draw my conclusion on whether CEO stock options grants relates to REM by decreasing the discretionary expenses.

4.2.1 Total abnormal discretionary expenses regression results

In this paragraph, the results of the relation between the dependent variables and the

independent variables, including the control variables, are presented. In other words, we are testing the relation between total abnormal discretionary expenses (AbnDiscExpTot) with the CEO stock options grants (CEO_OptGrants) and the control variables. With these results we can conclude whether H1 is supported or not.

First, Table 3a shows that six out of the 10 variables are significantly related worth

AbnTotDiscExp. This means that between the independent variables CEO_OptGrants, Assetst-1, EMP, NetIncome, CEObonus, CEObonus, CEOage and the dependent variable

AbnTotDiscExp a correlation exists.

Table 3a shows that that AbnDiscExpTot and CEO_OptGrants are influencing each other significantly as well. The coefficient of 0.057 indicates that abnormal total discretionary increase by 0.057 when CEO option grants increase with 1 unit. Next to that, the coefficient of 0.057 is positive. This means that there is a positive relation between CEO option grants and total discretionary expenses. As stated in former paragraphs, a negative relation was needed to support my H1. Now an increase in CEO stock options grants lead to a decrease in REM because a higher abnormal total discretionary expenses have a decreasing effect on earnings. Cohen & Zarowin (2010) find a different result. They find that stock options granted to executives are positively related to REM management by an abnormal decrease in discretionary expenses. The difference is that Cohen & Zarowin test the REM hypotheses solely around seasoned equity offerings, which makes their test slightly different than mine. I can conclude out of my results, that there is enough evidence, not to support Cohen &

(31)

30

CEO_OptGrants. I also have enough evidence not to support H1. In my test I cannot reject

H0 because there is a positive relation and a negative significant relation was needed, between AbnDiscExpTot and CEO_OptGrants, to support a decrease in REM. I support the, with a significance level of p<0.01, that REM decreases instead of increases, when CEO stock option grants increases.

The coefficient of NetIncome is negative and is significant for p<0.01. The coefficient of -0.418 declares that in our sample, abnormal total discretionary expenses scaled by lagged assets goes down with 0.418 when NetIncome scaled by lagged assets increases with 1. This means that net income and discretionary expenses are well influencing each other.

Roychowbury (2006) confirms this relation and names the negative relation logical because overall when expenses are lower, income is higher. In his paper the coefficient is -0.097 and is significant with p<0.05. If you compare the results of Roychowbury (2006) to mine, the outcome is pretty similar but the coefficient in my test is much higher. This means that in my test, the reaction on abnormal discretionary total expenses is stronger.

I notice that the t-statistics of the variables CEObonus and Assetst-1 stand out because they are much bigger than the t-statistics of the other variables, repectively 19.75 and -27.08. For Assetst-1 this is expected because CEO_OptGrants is scaled by lagged assets the ratio therefore is partly reflected by assets. This is also reflected in the p-value. Both p-values are significant for a p<0.01. This means that both variables are strongly related to abnormal total discretionary expenses. The difference is that Assetst-1 has a negative relation and CEObonus has a positive relation. For Assetst-1 this is quite logical because when lagged assets increase, the ratio of abnormal total discretionary expenses divided by lagged goes down. In table 3a is disclosed that when lagged assets goes up by 1 unit, AbnDiscExpTot goes down with 0.358 unit. I don‟t draw conclusions towards REM out of the Assetst-1 variable because this variable is mainly in my model to control for size.

The outcome for CEObonus, on the other hand, are is applicable towards REM. There is a positive relation between CEObonus and AbnDiscExpTot with a coefficient of 0.294 and a p-value<0.01. Edmans (2009) find a negative significant relation between CEO bonuses and discretionary expenses with a coefficient of -0.006 and a significance of p<0.01. The output of my model gives me enough evidence to say that I do not support this statement of Edmans (2009). Out of my model comes forward that when CEO bonuses increases, abnormal discretionary expenses increase as well. I can conclude that this increase in expenses brings along a decrease in income. Therefore CEO bonuses actually have a decreasing effect on REM.

(32)

31

Next to Assetst-1, EMP is also a variable to control for size. EMP is the total number of employees within a public traded US firm multiplied by 1000. There is a positive

significant relation but the coefficient is low, namely 0.002. Our expectation that a firm with more employees needs to make higher investment to cope with employee costs is confirmed. The link between REM and the amount of employees is not as clear as CEObonus but it does influence abnormal total discretionary expenses and therefore REM. EMP is used as a good indicator to control for firm size and the positive coefficient. This variable reflects the expected relation between size and amount of abnormal expenses. Similar to Roychowbury (2006), when size goes up expenses go up as well.

The DummyGender variable is one of the variables that is not significant with a p-value of 0.786. This means that I don‟t find a relation between gender and REM in my model. Therefore I cannot support the statement of Peni & Vähämaa (2010) that female CEO‟s are following a more conservative earnings management strategy and are less vulnerable to engage in earnings management. REM is not related to gender according to my model. Peni & Vähämaa (2010) state in their paper that they use a timeframe from 2003-2007 and that only 3% of the tested US public firms have female CEO‟s. This timeframe and percentage is almost the same as in my test. This makes my findings even more interesting when I compare the outcomes and conclude that I do not support the statement of Peni & Vähämaa (2010). The variable DivTot has a positive coefficient of 0.415 and is not significant with a p-value of 0.158>0.1. Core & Guay (1999) name the Black-Scholes model the appropriate risk neutral valuation for an executive stock option to account for dividend payouts. Expected was that dividends on its own would influence the discretionary expenses, when looking at

incentives to improve compensations. Also because it is a part of the valuation of the Black-Scholes value and the CEO_OptGrants is significant as well. But the total dividends do not only consist of options grants to CEO‟s, but also includes dividends on normal shares and option grants to other employees. Out of my test is concluded that there is no relation between the total dividends and abnormal total discretionary expenses, and therefore REM. This means that dividend payouts on its own do not influence CEO‟s to engage in REM. The EPS variable has a coefficient of -0.024 and is significant with a p-value of 0.009<0.01. A negative coefficient of -0.024 discloses a negative relation between the earnings per share ratio and the abnormal total discretionary expenses. Out of my test comes forward that we cannot support the statement of Edmans (2009), because there is enough evidence with a p<0.01 that an increase of the earnings per share ratio of 1unit, decreases the abnormal total discretionary expenses by 0.024 unit. This is not a huge influence but enough

(33)

32

to support the statement of Roychowdhury (2006). He states that abnormal discretionary expenses decrease EPS. My conclusion is that an increase in EPS actually increases REM engaged by CEO‟s.

The R-squared in my model is 0.598. So when controlling the regression, abnormal discretionary expenses is for 59.8% explained by the independent variables and control variables of my regression. The other 40.2 % is not explained by my model, and is influenced by something other than my model. This R-squared is pretty high compared to other papers where the R-squared is around 40% to 50% (Core & Guay .1999, Roychowdhury. 2006). When I take CEO_OptGrants out of my regression, and run solely it on my control variables, my R-squared is 0.577. The regression ran on my control variables only, explains 57.7% of my dependent variable. This is a difference of 2.1%, which is solely explained by

(34)

33

Variable Coefficient T-value P-value 95% confidence interval

lower bound

95% confidence interval upper bound CEO_OptGrants 0.057 12.1 0.000 0.048 0.066 Assetst-1 -0.358 -27.08 0.000 -0.385 -0.333 EMP 0.002 2.76 0.006 0.001 0.003 NetIncome -0.418 -4.48 0.000 -0.601 -0.235 DivTot 0.415 1.41 0.158 -0.161 0.992 CEObonus 0.294 19.75 0.000 0.264 0.323 EPS -0.024 -2.62 0.009 -0.415 -0.006 CEOage -0.001 -0.2 0.840 -0.006 0.005 DummyGender 0.034 0.27 0.786 -0.211 0.279 Observations 2854 R-squared 0.598

(35)

34

4.2.2 Separate abnormal discretionary expenses regression results

In this paragraph, the results of the relation between the dependent variables and the

independent variables, including the control variables, from the sub regression are presented. In other words, we are testing the relation between separate abnormal discretionary expenses (AbnXRD, AbnXAD, AbnXSGA) with the CEO stock options grants (CEO_OptGrants). With these results I can eventually determine if H1A, H1B and H1C can be supported. Table 3b and 3d show that that AbnXRD and AbnXSGA have a relation opposite to CEO_OptGrants with a p<0.01. This means that the CEO stock option grants are well

influencing these abnormal discretionary costs. As determined in in paragraph 4.2.1 as well, a positive relation is present. This is reflected in the coefficient of 0.001 in table 3b,

respectively 0.021 in table 3d. In my model, when CEO option grants increase with 1, abnormal R&D expenses go up with 0.01 and XSGA go up with 0.021. In paragraph 3.2 I determined that there was a negative relation needed to increase REM. The conclusion that can be drawn out of results is that enough evidence exists, not to support my H1A and H1C. In both cases an increase in CEO stock options, actually have a decreasing effect in net income and therefore REM.

Next to that, table 3b shows the information needed to determine the support for H1B. As former hypotheses, a negative relation is needed between the independent and the

dependent variable. From the output of Table 3b can be concluded that the needed significant relation between CEO_OptGrants and abnormal advertising expenses, does not exists in my model. Even with a p<0.1, CEO_OptGrants is not significant with a p-value of 0.950. This makes me wonder if Cohen & Zarowin (2010) & Roychowbury (2006) need to include advertising expenses as a measure for REM incentives. This is difficult to find out because these papers only discuss discretionary expenses as a whole and not separate. Out of my results can be concluded that there is not enough evidence to support H1B with a significance level p<0.1. H0 is not rejected.

Further, it stands out that the R-squared, in comparison with AbnXSGA, is much higher than the R-squared of the AbnXRD and AbnXAD output. The R-squared from the sub regression including AbnXSGA, is 0.617. The R-squared of the sub regression including AbnXAD and AbnXRD are 0.334 respectively 0.406. I conclude that the biggest part of the AbnDiscExpTot regression of 0.598 is therefore explained by the abnormal selling, general and administrative expenses.

(36)

35

There is a difference for NetIncome between the AbnXAD and the other separate expenses but also compared to the abnormal discretionary expenses as a total. For AbnXAD, NetIncome is not significant with a p-value of 0.621 while for the other dependent variables NetIncome is significant with a p<0.01. This implies that abnormal advertising expenses are not related to net income. When only looking at this variable, you could have already

concluded that advertising and REM have no relation with each other because for this a relation with net income is necessary and that lacks here.

It stands out that in Table 3a, 3b and 3c, the age of a CEO has no significant relation with any of the separate abnormal discretionary expenses. Nowhere are the variables

significant to each other, the coefficient of the variable CEOage is 0.000 opposite towards each separate abnormal discretionary expense. Out of my separate output, I cannot support the statement of Bergstresser & Philippon (2006), that a CEO is less sensitive to REM when he or she is older.

Last, the results for my hypotheses H1A, H1B and H1C should be taken with care. Measuring REM with separate abnormal discretionary expenses has not been established in prior literature and hence these measures may not be entirely reliable. Former research mainly uses the abnormal total discretionary expenses as a measure but I wanted to test if separate abnormal expenses would have the same outcome.

(37)

36

Variable Coefficient t-statistics P-value 95% confidence interval

lower bound

95% confidence interval upper bound CEO_OptGrants 0.001 10.43 0.000 0.001 0.002 Assetst-1 -0.006 -16.28 0.000 -0.007 -0.006 EMP 0.0001 -2.09 0.037 0.000 0.000 NetIncome -0.009 -3.43 0.001 -0.148 -0.004 DivTot 0.013 1.56 0.119 -0.003 0.030 CEObonus 0.005 12.01 0.000 0.004 0.006 EPS -0.001 -4.04 0.000 -0.002 -0.001 CEOage -0.000 1.43 0.154 -0.000 0.0003 DummyGender 0.001 0.15 0.879 -0.007 0.008 Observations 2854 R-squared 0.406

(38)

37

lower bound

95% confidence interval upper bound CEO_OptGrants 0.0005 -0.06 0.950 -0.0001 0.0001 Assetst-1 -0.006 -26.37 0.000 -0.006 -0.005 EMP 0.0001 13.62 0.000 0.0001 0.0002 NetIncome -0.0008 -0.49 0.621 -0.004 0.002 DivTot 0.005 -0.99 0.320 -0.14 0.005 CEObonus 0.002 7.07 0.000 0.001 0.002 EPS -0.0004 2.6 0.009 -0.0001 0.000 CEOage -0.000 -1.39 0.166 -0.0001 0.005 DummyGender 0.001 0.65 0.519 -0.003 0.005 Observations 2854 R-squared 0.334

(39)

38

lower bound

95% confidence interval upper bound CEO_OptGrants 0.021 11.73 0.000 0.018 0.025 Assetst-1 -0.152 -29.90 0.000 -0.161 -0.142 EMP 0.001 4.27 0.000 0.001 0.001 NetIncome -0.156 -4.37 0.000 -0.226 -0.086 DivTot 0.146 1.29 0.196 -0.075 0.366 CEObonus 0.113 19.83 0.000 0.102 0.124 EPS -0.008 -2.17 0.030 -0.014 0.001 CEOage -0.000 -0.37 0.711 -0.002 0.002 DummyGender 0.016 0.34 0.734 -0.077 0.110 Observations 2854 R-squared 0.617

(40)

39

Conclusion

In the late 90‟s, many firms started granting equity incentives to executives. These incentives where granted to incentivize executives to bind an executive to a firms for a longer time but it was mainly used to align the incentives of shareholders and managers (Hall & Liebman, 1997). According to Hall & Liebman, the optimal situation for a firm is when the agency problem is decreased towards the point that the incentives and risk appetite of shareholders and managers are aligned. On the other hand, equity incentives are often linked to an excessive fixation on stock prices (Hall & Murphy, 2003) and eventually real earnings management to boost short term stock prices (Roychowdhury, 2006).

In my research I use CEO stock option grants and tried to determine the relation between this equity incentive and real earnings management done by CEO‟s. Hereby, I have tested whether there was an abnormal decrease in discretionary expenses when CEO options grants increased. I made the assumption that a decrease in abnormal discretionary expenses measures an increasing incentive to act into REM. This assumption was also used by Roychowdhury (2006) as a measure for REM. In other words, a negative relation between CEO option grants and abnormal discretionary expenses would have meant an increase in REM. Eventually I ran a regression, where tested the total abnormal discretionary expenses opposite to CEO option grants. After that I ran three sub regressions, where I tested the separate abnormal discretionary expenses opposite to CEO option grants. These separate expenses where: R&D, advertising and SGA.

Out of my results comes forward that, CEO stock options grants has a positive relation with total abnormal discretionary expenses, abnormal R&D expenses and abnormal SGA expenses. I found no significant relation towards abnormal advertising expenses. In my model, CEO stock option grants do not influence abnormal advertising expenses. REM can therefore not be influenced by CEO stock option grants through a decrease in advertising expenses. The conclusion that can be drawn about the other dependent variables is that my output does not support my hypotheses and sub hypotheses. Out of my results I find that CEO option grants have a decreasing effect on REM instead of an increasing effect on REM. I can therefore not reject my hypothesis H0 and support that stock options is seen to be a good way to decrease agency problem without the risk of an increase in REM.

The relation between CEO stock option grants and real earnings management