The effects of budget control tightness and the use of own-level financial performance measures on real earnings management

(1)

Amsterdam Business School

The effects of budget control tightness and

the use of own-level financial performance measures

on real earnings management

Master Thesis

Name: Martijn van Deun Student number: 10695257 Date: June 12, 2015

MSc Accountancy & Control, specialization Control

Faculty of Economics and Business, University of Amsterdam 1st_{Supervisor: Dr. ir. S.P. van Triest}

2nd_{Supervisor: Dr. ir. B.A.C. Groen}

(2)

Abstract

In this master thesis I examine the relationship between budget control tightness and real earnings management. Furthermore I examine the effect of the use of own-level performance measures on this relationship. With a survey among 98 business unit managers I find evidence that budget control tightness is positively related to real earnings management. I don’t find evidence that the use of own-level performance measures increases this relationship. In the past, many researchers have been reporting on the adverse effects of tight management control- and tight budget control systems. Nevertheless, to my best knowledge, this is the first study, which explicitly examines the relationship between budget control tightness and real earnings management. The results of this study are interesting and call for more research on this relationship.

Keywords Earnings Management, Real Earnings Management, Tightness, Budgets, Management control systems, performance measures

Statement of Originality

This document is written by student Martijn van Deun 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)

Acknowledgements

I am very grateful to my supervisor dr. ir. Sander van Triest for his encouragement, suggestions, inspiring ideas and the pleasant meetings we had. I also thank dr. ir. Bianca Groen for introducing me into different paradigms of management control research during the MCR lectures, which led to the first ideas for this thesis. I thank Wim Peeters for giving me the opportunity to complete this study. His views on management and his encouragement to think outside “the box” inspired me when writing this thesis.

Of course, completing this research was only possible because of the willingness and openness of the business unit managers who completed the survey.

Although I enjoyed travelling to Amsterdam on Fridays, the last few years have been quite a challenge. Work- and personal life were often out of balance. Completing this study would not have been possible without the positivism, caring and loving support of my wife Leonie.

Martijn van Deun Amsterdam, June 2015

(4)

1 Introduction

In this master thesis I investigate the impact of budget control tightness on real earnings management. In addition, I examine the impact of the use of own-level performance measures on this relationship. Most prior research on earnings management has been focused on firm level.1_{When studying earnings management on firm level, “important information is lost}

through aggregation” (Guidry et al 1997, p. 140). I will conduct survey research on a more operational level: a survey among “local” or business unit managers. Studying business unit managers enables me to zoom in on the effects of management control systems (MCS), such as budget systems.

Agency theory teaches us “that agents (employees) will not always act in the best interest of the principal (owners of a firm)” (Jensen and Meckling 1976, p. 5). Management control systems are crucial to align interest of employees and owners of a firm because “failure can lead to large financial losses, reputation damage, and possibly even to organizational failure” (Merchant and Van der Stede 2012, p. 3). Commonly used MCS are budget systems. These systems are often combined with incentive systems such as bonuses. Although MCS are crucial to answer agency problems, prior research has shown that these systems can have adverse effects (Healy 1985, Guidry 1997, Lau 1999, Van der Stede 2000, Graham et al 2005, Cohen et al 2008). One of these effects is manipulating the earnings figure also known as earnings management. This can be done by manipulating accruals - which is called accrual based earnings management - or by real activities manipulation - which is called real earnings management - such as accelerating a sale or deferring expenditure (Roychowdhury 2006). Recent studies find evidence that, because of “the stigma attached to accounting fraud in the post-Enron environment” (Graham et al 2005, p. 5) and the impact of the Sarbanes & Oxley act (Cohen et al 2008), managers will easier engage in real earnings management instead of accrual based earnings management.

Researchers (e.g. Healy 1985, Guidry et al 1997, Graham et al. 2005) report that ‘meeting the benchmark” (e.g. analysts’ consensus earnings estimate or prior year earnings) is one of the motives why executives engage in earnings management. The desire of meeting the benchmark is driven by reputation and career concerns on the one hand and on the other hand by incentive contracts such as bonuses. For business unit managers, I expect that ‘meeting the benchmark’ will in many cases mean ‘deliver the budget’. When management does not easily accept budget deviations and puts to much pressure on meeting the budget, hence when a tight budget control

(7)

system is applied, I expect a higher risk of “gaming the numbers”. Furthermore, I expect that the size and nature of this relationship will be influenced by the composition of the measures on which managers’ performance is evaluated. The most common performance measures used are financial performance measures. Abernethy et al. (2012) show that aggregated or above-level performance measures are positively related to a more ethical working climate where there is less manipulation. My prediction is that when a firm uses a tight budget control system together with performance evaluation based on own-level thus “local” performance measures, real earnings management will be larger than when a firm uses a tight budget control system together with above-level performance measures.

The purpose of this paper is to contribute to existing knowledge regarding the negative side effects of tight MCS design, in particular tight budget control systems. According to Merchant and Van der Stede (2012) control tightness is a subject that has received relatively little attention in the literature. Nevertheless, this study builds on prior research conducted by Van der Stede (2000), in which evidence of a positive relationship between budget control tightness and managerial termism is found. In this research I move the focus from managerial short-termism to real earnings management. Although, researchers have intensively been reporting on the side effects of overly tight control systems, to my best knowledge, this is the first study that explicitly examines the relationship between budget control tightness and real earnings management. Today, this is relevant and important because of the shift from accrual based- to real earnings management due the tightened financial reporting legislation and renewed focus on corporate governance. Furthermore, the additional focus on own-level financial performance measures on business unit level will give this research an extra dimension.

In the remainder of this paper I first discuss existing theory regarding MCS, the views on MCS and the side effects of MCS choices. Next I will zoom in on the concept of budget control tightness and differences between above- and own-level performance measures. Based on this theoretical framework and review of existing research regarding these subjects I will present my hypothesis and conceptual model. In the third chapter of this paper I discuss the survey characteristics and how the hypothesis will be operationalized. Furthermore, in this chapter I assess the reliability and internal validity of this operationalization. In the fourth chapter the results of the analysis are presented. Finally, in the last chapter the main findings are discussed in detail. This chapter will conclude with an overview of the limitations of this research and a discussion of the possibilities for future research regarding the subject of this paper.

(8)

2 Theory

In the first part of the theory section I will give an overview of the current state of knowledge regarding earnings management, MCS design and performance measures composition implications. I start with an explanation of basic agency theory. This is important to understand the concepts of MCS design and its side effects. Next, I will discuss earnings management, the difference between accrual based- and real earnings management and earnings management motives. After that, I will describe the concepts of MCS, MCS design choices and its side effects. Subsequently I zoom in on the ideas behind tightness, more specifically budget control tightness. The first section of this literature review ends with an overview of the composition of performance measures and an explanation of how this composition can increase the risk of earnings management. Based on the concepts discussed in the literature review the hypothesis and conceptual model are presented in the second part of this chapter.

2.1 Literature review 2.1.1 Agency theory

Basic agency theory addresses the problem that arises when firm-owners (principals) delegate authorities to managers (agents). The theory predicts goal conflicts between principal and agent, because people will always act to increase their own wealth: self-interest. Key mechanism in agency theory is information asymmetry where the agent has better information than the principal. This information asymmetry combined with the assumption that people act to increase their own wealth, will lead to high cost for the principal. According to Jensen and Meckling (1976) the principal has to establish incentives for the agent. Furthermore, he has to invest in systems to monitor actions and activities of the agent. According to Eisenhardt, “the focus of the principal-agent literature is on determining the optimal contract, behavior versus outcome between the principal and the agent” (1989, p. 60). Behavior-based systems are for example diagnostic control systems such as budget systems. These systems help closing the information gap and give guidance on how organizational objectives have to be achieved. Outcome based systems are for example bonus or incentive contracts. Based on the economic theory of self-interest, these systems are often used “when principals cannot rely on the prevailing social norm to make agents to do the right thing” (Abernethy et al 2012, p. 11).

Prior research has shown that both behavior based- as well as outcome based contracts can potentially lead to undesirable actions by managers, such as earnings management.

(9)

2.1.2 Earnings Management

Graham et al (2005) report that the earnings figure is the most important performance measure for executives. Due to the importance of this figure, it is to be expected that the likelihood of manipulation is substantial. Over the years there has been much research in the field of earnings management. Several definitions of earnings management emerged from these studies; the definition by Healy and Wahlen (1999, p. 368) is often cited:

“Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company, or to influence contractual outcomes that depend on reported accounting numbers.”

In contrast to Healy and Wahlen, Schipper (1989) makes an explicit distinction between disclosure management or accrual based earnings management and real earnings management.

2.1.2.1 Accrual based earnings management

The definition of accrual based earnings management according to Schipper (1989, p. 92): “Within the opportunities offered by the accounting system, managers could manage earnings by selecting methods within the Generally Accepted Accounting Principles (GAAP)”.

In other words, when managers use accruals or make decisions regarding accounting policies in order to influence the earnings figure, we speak about accrual based earnings management.

Manipulation of accruals has no direct cash flow consequences as they only influence the timing of booking revenues or costs. “Examples include under-provisioning for bad debt expenses and delaying write-offs” (Roychowdhury 2006, p. 336). Within accrual based earnings management there is a thin line between what is legitimate and what is fraud. Making accounting decisions outside GAAP can be seen as fraudulent actions. The past and more recent scandals have shown that these actions can have detrimental effects on credibility, firm value and even destroy large organizations (e.g. Enron, WorldCom).

2.1.2.2 Real earnings management

Roychowdhury (2006) defines real earnings management as “departures from normal operational practices motivated by managers’ desire to mislead at least some stakeholders into believing certain reporting goals have been met in the normal course of operations” (2006, p. 337). Graham et al (2005) define real earnings management as “managers’ willingness to give up real economic value to manage financial reporting outcomes” (2005, p. 6). Examples of real earnings

(10)

activities are “temporary increase sales, overproduction to report lower cost of goods sold” (Roychowdhury 2006, p. 335), “postpone or eliminate hiring, R&D, advertising, cutting the travel budget, delaying or cancelling software spending, or deferring maintenance spending” (Graham et al 2005, p. 40).

Real earnings management has nothing to do with fraud. Real earnings management decisions can be perfectly legitimate, but as opposed to accrual based earnings management, they do have cash flow consequences and impacts long-term value. For instance, accelerating a sale with the use of discounts will indeed boost current year earnings figure, but in the long run the given discount will end up as a cost that wasn’t completely necessary. Furthermore, it is expected that deferring maintenance expenses will have a positive impact on the current year earnings figure, but in the long run it can reduces productivity and even impact technical life span of machinery. In this sense it can be argued that real earnings management is more undesirable than accrual based earnings management within GAAP. Hence, 78% of the surveyed executives in the study of Graham admit that they would sacrifice economic value in order to smooth earnings. Because of “the stigma attached to accounting fraud” (Graham et al 2005, p. 5) and the impact of the Sarbanes & Oxley act (Cohen et al 2008), we could expect that managers will easier engage in real earnings management instead of accrual based earnings management.

2.1.2.3 Earnings management motives

Researchers have identified several motives why managers engage in earnings management. Meeting or beating earnings benchmarks is one of the most important motives. “Failure to hit the benchmark creates uncertainty about a firm’s prospects, and raises the possibility of hidden, deeper problems at the firm” (Graham et al 2005, p 66). This benchmark can be the analysts’ consensus forecast, prior year earnings or certain financing covenants. Meeting the benchmark is often achieved with earnings management actions, but also setting next years target is a motive for earnings management. Leone and Rock (2001) find for instance evidence of business units, who are managing earnings by making income decreasing discretionary accruals in order to reduce next year’s target. Besides meeting or beating the benchmark, Cohen et al (2008) find evidence that avoiding reporting losses is also an important motive for earnings management.

Not only firms’ reputation is at stake when targets are not achieved, personal credibility and career concerns are important motives for earnings management. Van der Stede (2000) reports that the (in) ability of meeting the budget, or meeting the benchmark, directly affects managers’ salary and career prospects. Graham et al (2005) report that more than three-fourth of their survey respondents agree that “external reputation helps explain their desire to hit the earnings

(11)

benchmark” (2005, p. 28). Also from a personal reputation point of view, Godfrey et al (2003) find evidence of earnings management in the case of CEO changes: downward earnings management in the first year and upward earnings management in the next year after the change. Besides firm or personal reputation, there are also pecuniary motives. Strong evidence is found on the positive relationship between bonus compensation and earnings management (Healy 1985, Guidry 1997, Abernethy et al 2012). Healy (1985) reports “bonus schemes do create incentives for managers to select accounting procedures and accruals to maximize the value of their bonus onwards” (1985, p. 106). He finds evidence of executives, who are manipulating earnings figures in order to meet bonus targets. Guidry (1997) finds similar evidence on business unit managers’ level. Abernethy et al (2012) show that the level of manipulation depends on performance measures choices used to evaluate and determine managers’ bonus.

Considering the above, there is a wide range of earnings management motives. They can however be roughly classified as “externally driven motives” and “internally driven motives”. Meeting analysts’ consensus forecast, meeting debt covenants, “making the new CEO look good” are typically externally driven motives. Internally driven motives are meeting the budgeted target, meeting the bonus target and reporting good results from an internal career perspective. This paper focuses on the determinants of internally driven motives such as management control systems and it’s particular elements. The next paragraph contains an in-depth discussion of these systems.

2.1.3 Management Control Systems

As pointed out in the introduction, management control systems (MCS) should be an answer to agency problems. In this section I first discuss MCS in a broad sense and present reasons why MCS can be an answer to the agency problem. Next I will zoom in on more specific MCS such as (tight) budget control systems and types of performance measures.

In the literature it is hard to find a single definition of MCS. There are even different terms used: some authors speak about Management Accounting Systems (MAS), Management Accounting and Control Systems (MACS) or Management Control Systems (MCS). Despite the different terms and definitions used, in general they all have the same meaning. MCS are used to align the interests of employees (or managers) with organizational interests by influencing behavior. Simons’ definition of MCS relates explicitly to the information asymmetry problem as discussed in the agency theory paragraph (1995, p. 3):

(12)

“The formal, information-based routines and procedures managers use to maintain or alter patterns in organizational activities”.

Forms of formal routines and procedures are for instance budget systems. With information-based routines the organizations’ objectives and goals are communicated and the current state (e.g. variances) is monitored and evaluated. According to Simons (Simons 1995, p. 5) management control is operationalized by four different systems; also know as “the levers of control”:

1. Beliefs systems: in order to inspire and direct; 2. Boundary systems: in order to set limits;

3. Diagnostic control systems: in order to motivate, monitor and reward achievement of goals;

4. Interactive control systems: in order to stimulate organizational learning.

Merchant and Van der Stede (2012) distinguish three types of MCS: Results controls, Action controls and Personal/Cultural controls. Results controls are often associated with bonuses or pay for performance. There are however other incentives for delivering desired results such as promotion or demotion in the case the results are not as expected. Also the risk of dismissal can be an incentive to deliver desired results. In contrast to results controls, with action controls there is a direct focus on actual behavior of employees. This can be achieved with behavioral constraints like physical closure of certain areas, system authentication or segregation of duties. Pre action reviews, like approving the budget before spending, are also a form of action controls. Another form of action controls is employee accountability. By evaluating actual behavior against desired actions (instructions) people can be held accountable for their behavior. Besides results- and actions controls Merchant and Van der Stede speak about personal and cultural controls like selection and placement, training, corporate culture, code of conduct and tone at the top. In the spirit of COSO, these types of controls relate to the internal environment of a firm.

In their model to evaluate MCS characteristics within service Firms, Auzair and Langfield-Smith (2005) use the following dimensions: action versus result controls, formal versus informal controls, tight versus loose controls, restricted versus flexible controls and impersonal versus interpersonal controls. Combining these dimensions, they distinguish two types of MCS (2005, p. 401/402): more bureaucratic MCS that are based on actions-, formal-, tight-, restricted- and impersonal controls. Vice versa less bureaucratic MCS are based on results-, informal-, loose-, flexible- and interpersonal controls. These less bureaucratic MCS are based on interpersonal relationship, autonomy and trust.

(13)

Considering the above, when designing a MCS, management has to decide on the level of bureaucratic- and more tight controls. Prior research has shown that the choice for a more- or less bureaucratic MCS depends on many factors such as organizational size, level of decentralization, level of interdependencies, industry (from a contingency theory perspective, Chenhall 2003), strategy (Simons 1987), stage in organizational life cycle (Auzair and Langfield-Smith 2005), past performance, legislation, financing structure, etcetera. MCS can although be very costly. Direct costs related to MCS are for instance cost for information systems, cost of control and auditing personnel and the cost of incentive contracts. Next to these explicit costs, there are also indirect mostly “hidden” costs caused by the negative side effects of MCS. These costs are often the result of bureaucratic and tight MCS. The terms bureaucratic, formal and tight are often used interchangeably. In the next section I will refer to these elements as “tightness”.

2.1.3.1 Views on MCS, tightness and side effects

According to Merchant and Van der Stede “tighter control should provide more assurance that employees will act in the organization’s best interest” (2012, p. 123). With for instance frequent budget reviews, an organization is able to adapt more quickly to new situations. Furthermore, it is expected that tight budgets combined with a low degree of tolerance to deviate from the budgeted figure will lead a higher level of cost control and consequently better financial results. Merchant and Van der Stede believe that tighter MCS are an answer to agency problems.

Besides MCS from the agency perspective, Macintoch and Quatrrone (2011) present some other views on MCS and tightness. The first one is based on research carried out by Mintzberg (1975), where managers were asked what they really did. In contrast to basic managerial ideas – that organizational objectives are achieved via plans and formal actions - Mintzberg discovered the opposite. Managers’ work seems to be highly fragmented. Information gathering and processing is carried out through cognitive processes rather than using formal information systems. This point of view is interesting considering the basic ideas of rationality, formal rules and routines as the building blocks of MCS and can have consequences for MCS’ success.

Another view on MCS is that information gathered by these systems can be used to influence and gain power. Hopper and Armstrong (1991) present a more radical view, based on labor process theory which suggests that MCS are used to dominate, deskill and control the workforce in order to create more output.

Although many researchers, such as for instance Merchant and Van der Stede, have a positive stance regarding formal MCS, they do present negative side effects of management control

(14)

systems and (overly) tightness. Merchant and Van der Stede (2012) distinguish the following (negative) side effects:

• Behavioral displacement • Gamesmanship

• Operating delays • Negative attitudes

Behavioral displacement occurs when certain elements of MCS like rules, procedures and incentives

are not in line (congruent) with the organizations’ objectives. In these situations, though with best intentions, managers can act in a harmful way. For instance, when a purchasing manager has a target on purchasing against the lowest unit cost, the quality of these units can affect the production process and finished goods quality in a negative way. This can even result in customer claims and higher cost than the initial saving just on the purchase price.

When managers consciously engage in actions that could be harmful for the company, we speak about gamesmanship. Examples of gamesmanship are earnings management, but also undesirable effects during budget-target-setting process like “under-promise and over-deliver” tactics, target ratcheting (Leone and Rock 2001, Bouwens and Kroos 2011) and effort reduction (Bouwens and Kroos 2011) when the periodic target has already been met.

Operating delays are the consequence of formal and bureaucratic systems. Examples are

beforehand approval systems and tight system based segregation of duties, which can delay the sales order process or even reduce sales because of the inability to response quickly to customer demands.

The last side effect of MCS Merchant and Van der Stede mention, concerns negative attitudes like tension, conflicts and frustration that arises as a result of certain MCS design choices. Although these side effects are hard to measure, they can be very disruptive.

One side effect of tight MCS, which is not mentioned explicitly by Merchant and Van der Stede, is that MCS can harm entrepreneurship and constrain innovation. One of the goals of MCS is to reduce risk-taking behavior, but doing business is taking risks. In this sense, MCS that are overly tight can reduce innovation (Auzair and Langfield-Smith 2005), reduce flexibility, delay sales processes and consequently can be perceived as systems that “kill entrepreneurship”. A less bureaucratic MCS can affect innovation and entrepreneurship in a positive way.

Of the side effects presented above, in this paper I will mainly focus on the side effect “gamesmanship”. More specifically the relation of budget control tightness and the composition

(15)

of performance measures on real earnings management. In line with the behavioral- and outcome contracts as discussed in the section about agency theory, in the next section I first will discuss the characteristics of budget control tightness. Second I will present some theory regarding own-level performance measures.

2.1.3.2 Budget control tightness

Budgeting is commonly used for both planning as well as performance evaluation; hence it consists roughly of two elements, namely planning and control. During the planning phase the budget and targets are developed. During the control phase monitoring activities are being executed to assess actual- versus budgeted performance. But when do we speak about budget control tightness? “In a tight control philosophy, the budget target is considered to be a first commitment against which the manager is evaluated” (Van der Stede 2001, p. 122). More specifically, Van der Stede defines the following characteristics of tight budgeting (2001, p. 122):

• Amount of emphasis on attaining budget targets

• Degree of budget commitment, whether budget revisions during the year are allowed • Amount of detail of interim budget reviews

• Degree of tolerance for interim budget deviations

• Degree of involvement of top management in the subordinates’ business

Despite of the benefits expected in a tighter control environment, prior research has shown that too much emphasis on meeting the budget, hence budget control tightness, can have adverse effects. These effects occur both during the budgeting and target setting process, as well as during the “achievement” phase. Merchant (1985) reported that tight budgeting has an impact on inefficient choices managers make, also known as “slack”. Budgetary slack is the “intentional biasing of performance targets below the expected levels” (Webb 2002, p. 361). Lau (1999) also finds evidence that too much emphasis on meeting budget targets will lead to slack creation. Van der Stede (2000) reports budgetary slack and managerial short-termism as a result of tight budgeting.

Budgeting in general has received much criticism during the last decades. Jensen (2001) writes that budgeting “consumes a huge amount of executives’ time, forcing them into endless rounds of dull meetings and tense negotiations. It encourages managers to lie and cheat, lowballing targets and inflating results, and it penalizes them for telling the truth. (2001 p. 96)”. Proponents of “beyond budgeting”, Hope and Fraser (2003), write that budgeting is too expensive, it doesn’t meet the needs of managers and it results in “gaming the numbers”.

(16)

Marginson and Ogden (2005) wonder about the lack of publications on the positive side of budgeting, because budgets can be useful to present managers a clear and structured path to achieve companies’ objectives. The problem with budgeting, however, arises because it is used for both planning as well as for performance evaluation. Performance evaluation can be done with different forms of indicators. In this paper I will focus on the use of “local” or own-level budgeted targets. In the next section I will discuss this concept in more detail.

2.1.3.3 Own-level performance measures

Performance evaluation can be done in various ways. For instance, based on objective measures, but also based on (subjective) perceptions of the superior. Performance evaluation is often combined with a bonus. “Pay for performance is a prominent example of a type of control that can be called result controls” (Merchant and Van der Stede, 2012 p. 123). The performance measures used occur in different forms, i.e. financial versus non-financial, long term versus short term, aggregated versus non-aggregated. Specifically, regarding to aggregation, Bouwens et al (2011) distinguish “above-level measures” and “own-level measures”. Above-level measures are aggregated performance measures, which summarize total firm performance. Examples include share price, Return On Capital Employed (ROCE) or Earnings Before Interest and Taxes (EBIT) on firm level. Own-level measures or non-aggregated measures that provide an image of local-, for instance business unit or department performance. Next to profit or return measures, also measures such as budgeted versus actual costs and product cots could be used. In the remaining part of this research I will use the terms above- and own-level performance measures.

For business unit managers, above-level performance measures are more “noisy” than own level performance measures because they are more difficult to influence directly. Hence, from an agency theory- and controllability perspective it is not very likely that a higher weight on these kinds of measures will solve agency problems. On the other hand, manipulation of these measures will also be less likely.

According to Abernethy et al “aggregate2_{accounting performance measures are expected to}

be used to influence managers to ‘think of others’” (2012, p. 12). A higher weight put on above-level performance measures, thus performance measures at firm above-level instead of performance measure solely at business unit level, is positively associated with a more ethical working climate where we can expect less manipulation. Vice versa, this would imply that a higher weight put on

(17)

own-level performance measures, thus performance measures on business unit level would create a less ethical working climate with a higher chance of manipulation.

2.2 Hypothesis development

Based on the theory and prior research set out in the previous paragraph, below I will summarize the most important elements in order to build up the hypothesis.

Management control systems should increase the likelihood that organizational objectives will be achieved (Merchant & Van der Stede, 2012). Although they are implemented to control managers and to align the goal of the principal and the agent, many researchers have shown that these systems can have adverse effects such as the manipulation of the earnings figure, or earnings management. Specifically related to budget systems Jensen (2002) writes, “these systems are based on the premise that managers should be rewarded for achieving their targets and punished for missing them” (2012, p. 382). As mentioned earlier the problem with budget systems arises because they are used for both planning as well as performance evaluation. According to Graham et al. (2005), the importance of meeting the benchmark is a reason why executives engage in (real) earnings management. For business unit managers, meeting the benchmark will in most cases mean: “deliver the budgeted earnings figure”. The managers’ perceived pressure to perform increases when the superior does not easily tolerate budget deviations or accept budget revisions, hence budget control tightness, will potentially lead to dysfunctional behavior. Van der Stede (2000) reports dysfunctional behavior and adverse effects, such as managerial short-term orientation, when superiors put too much emphasis on budgeting, ergo tight budgeting. Jensen (2002) reports gaming practices during the planning phase, but also earnings management activities such as accelerating a sale in order to meet the budget target.

Considering the above, it is expected that managers will engage in undesirable “gaming” behavior, such as earnings management, in order to meet their target. Given the shift from accrual based- to real earning management in this post-Enron and Sarbanes and Oxley era (Graham et al 2005, Roychowdhury 2006, Cohen et al 2008), my first hypothesis is:

(18)

Budget Control Tightness Real Earnings Management Own-level Performance Measures

As outlined in the theory paragraph, important motives for meeting the benchmark are firm- and personal reputation on the one hand, and the managers’ bonus compensation on the other hand. In line with the findings presented by Abernethy et al (2012), my assumption is that the chance of earnings manipulation is lower when a firm uses more above-level measures to evaluate managers’ performance. Above-level performance measures are noisier and more difficult to influence directly. Putting high emphasis on these kinds of measures when assessing managers’ performance will reduce necessity for the manager to influence this measure with for instance earnings management. Vice versa, putting high emphasis on own-level performance measures, which the manager can easily manipulate will increase the risk of earnings management. In this sense, I think that the combination of the pressure to perform which leads to persistence to deliver the budgeted earnings figure, from a personal reputation drive, together with the use of own-level performance measures to assess managers’ performance, will lead to higher real earnings management.

Ergo, my second expectation is that the size and nature of the relationship as presented under H1 changes as a function of the use of own-level performance measures:

H2: Budget control tightness together with the use of own-level performance measures will lead to more real earnings management.

The model of the hypothesis above is structured as follows:

+ / +

(19)

3 Sample selection, survey design and variable measurement

To test the above hypothesis I have joined the thesis survey project of the Amsterdam University. First, because of the explicit focus on business unit managers. Second, with the construct sources3_{used in the questionnaire I am able to operationalize the hypothesis. Third, I}

think it is quite ambitious, if not impossible, to obtain sufficient responses on my own. And last but not least, I think survey research is well suitable to examine effects of MCS on business unit level. Even though most earnings management research is carried using large databases. To detect accrual based earnings management, for instance the Jones model (1991) is frequently used. For detection of real earnings management the model of Dechow et al (1998) can be used. However, with the use of these models it is still difficult to find convincing evidence (Healy and Wahlen 1999, Grahem et al 2005).

In the next section I first discuss the sample selection. Thereafter I present the survey design. In the last part the operationalization of the hypothesis is presented together with an in-depth discussion of the variables and reliability- and validity tests.

3.1 Sample selection

Together with fellow students of the Amsterdam University I obtained data from a survey amongst business unit managers. The data was collected between February and May 2015. By submitting at least 10 completed surveys, access was given to the entire dataset.

We approached managers who have substantial responsibilities, take decisions relatively independently and are held accountable for the results of their organizational unit. This unit should be operational rather than supportive and the sector could be anything: profit as well as non-profit. The unit could be an entire business unit, a division or a department. We aim for business unit managers who are responsible for at least 20 FTE. Size, measured by FTE is meaningful because MCS become more important as the number “agency problems” increase analogously with the number of employees. Finally, because the respondent needs to be subject to the management control system, the respondent could not be the owner or CEO of the company.

The initial dataset consist of 112 responses. In total I excluded 14 responses. 10 respondents did not answer the real earnings management questions. One respondent did not answer the budget control tightness questions. Two respondents did not meet the business unit manager

(20)

function criterion (1 CEO and 1 Financial Controller). One respondent did not answer the questions related to size; neither FTE nor sales value. For this respondent I cannot determine to what extend a MCS influences his decisions. 19 respondents did not meet the FTE criterion. After robustness checks4_{- analysis with and without these respondents that showed no}

significant differences between the groups - I decided to leave these responses in the dataset. Consequently the final dataset consist of 98 responses.

Table 1 presents the characteristics of the business units and their managers. The average age of the managers is 48 years. They have been working for 11 years in their current business unit or department and 5 years in their current position. The business unit has 70 FTE on average (median 32 FTE) and the average annual sales value or budgeted costs are 35 million euro (median 6.6 million euro).

When examining industry classification, 41% of the respondents work in a non-profit environment such as government, education or health care. 23% of the business units work in the professional services field such as advisory, engineering, legal and auditing. 19% work in the commercial services field such as wholesale, retail, hotels and restaurants. 12% are manufacturing companies and the remaining 5% are financial intermediation companies such as banking- and insurance companies and pension funds.

Table 1 Respondents descriptive statistics

Variable No of

observations Min Max Average Median St. dev

Panel A: Business Unit

Age of business unit 93 0.5 440 26 12 62.4

Annual Sales / Budget (M euro) 88 0.4 1200 35.6 7.0 132416163

Panel B: Business Unit Manager

Age of manager 98 25 65 48 47 9

Years working in business unit 98 0.5 42 11 8 8.9

Years working in position 98 0.1 27 5 4 5.5

FTE reporting to manager 98 4 450 70 32 82.5

Highest level of education (< Bachelor = 1, Bachelor = 2, Master or higher = 3)

97 1 3 2.5 2.5 0.6

(21)

3.2 Survey design

The survey is designed to gather information about the way organizations design their MCS, how managers use these MCS and what side effects of those MCS are to be expected. The survey covers the following areas:

- Business characteristics, growth and uncertainty; - Interdependencies and managers’ level of autonomy; - Leadership and culture;

- Managers’ characteristics and risk taking behavior; - Design and effects of the MCS.

In this study the primary focus is on the last area: the design and effects of MCS.

3.3 Variable measurement

As mentioned in the previous section, the survey contains different areas. The primary focus of this study is the design of MCS, in particular “budget control tightness”, “composition of performance measures” and the effects of MCS, specifically “real earnings management”. The operationalization of these variables is discussed below. The descriptive statistics of the specific questions and constructed variables can be found in paragraph 4.1.

3.3.1 Dependent variable: real earnings management

The dependent variable in this research is real earnings management. The constructed variable is called REM. The constructs to measure REM are based on prior research by Graham et al (2005). For the dependent variable the manager has to indicate how frequently the following actions are taken in order to comply with financial controls:

• Deferring or accelerating a needed expenditure • Deferring or accelerating a sale

• Book revenues now rather than next period (if justified in either period) • Book costs next period rather than now (if justified in either period)

• Provide incentives (e.g. price discounts) to customers to buy more this period • Decrease discretionary spending (R&D, advertising, maintenance)

• Delay or accelerate starting a new project (IT, support) • Delay or accelerate hiring an employee

(22)

Note that the third and fourth action can be seen as accrual based earnings management (within the GAAP) instead of real earnings management. Therefore, these questions will be excluded from the analysis. As a result, the measurement of real earnings management is based on 6 items.

The scale used to indicate how frequent managers use certain actions to manage earnings is a five-point Likert-scale. A score of 1 indicates that a certain action is never used. Vice versa, a score of 5 indicates that a certain action is used very frequently. When a manager is not able to use a certain action, for instance within a non-profit organization it is not very likely that earnings are managed by deferring or accelerating a sale, a score of 0 can be used. Obviously, these 0-scores don’t have to be taken into account. Therefore I will construct the REM variable based on a conditional average where only the answers of 1 and higher are taken into account. For example, if a respondent has 1 or higher on all 6 items, I take the average of all 6 items, but if 2 items are assessed as non-applicable with an answer of 0, I take the average of the remaining 4 items. Because of this calculation of real earnings management, it is not possible to evaluate the quality of the scale using Cronbach’s alpha, since this requires all items to have valid answers for all respondents. I interpret this measure as a formative rather than a reflective measure (See e.g. Bisbe et al 2007, p. 799).

3.3.2 Independent variable: budget control tightness

The independent variable in this research is budget control tightness. To measure this, the construct sources developed by Van der Stede (2001) are used. For the independent variable the manager has to answer five questions related to the behavior of his superior when it comes to budgeting:

• Puts much emphasis on meeting the budget

• Does not easily accept budget revisions during the year • Has a detailed interest in specific budget line items

• Does not lightly tolerate deviations from interim budgets targets • Is intensively engaged in budget-related communication

The first, second and fourth question are more related to the concept of “pressure to perform” where the budget acts like a diagnostic control system. The third and fifth question is more related to an interactive control system.

Before constructing this new variable by averaging the answers to the budget control tightness questions, reliability has to be determined. Reliability is tested with cronbach’s alpha. Cronbach’s alpha has a value between 0 and 1. Although there is no general rule as to which

(23)

level classifies as acceptable, the rule of thumb is that the cronbach’s alpha should be above 0.55_.

Cronbach’s alpha of the questions related to budget control tightness is on a good level of 0.74. This figure cannot be improved by deleting one or more questions indicating a high consistency between the questions. By averaging the answers to the tightness questions, the variable TIGHT has been constructed.

3.3.3 Validity analysis

To test if the variables for real earnings management and budget control tightness each have one underlying factor that explains the items scores, I first test convergent validity per variable with a “varimax rotated” factor analysis. Second I test whether there are indeed two different factors relating to real earnings management on the one hand and budget control tightness on the other hand. This test performed is by adding all the questions related to real earnings management and budget control tightness to the factor analysis. The test is called discriminant validity.

3.3.3.1 Convergent validity

For real earnings management I find two underlying factors with an eigenvalue higher than one. The rotated component matrix, Table 2, panel A, shows the loadings for each question. A possible explanation for the loadings of questions 97 and 98 might be that these activities both relate to “revenue based” earnings management. The fact that question 101 also loads on this factor seems to be peculiar because this question relates to “cost based” earnings management and is of the same kind as questions 102 and 103. Given the relatively high cronbach’s alpha, I will not exclude questions based on this factor analysis. However, to assess validity I will perform robustness checks, which are discussed in paragraph 4.4.

For budget control tightness I find one underlying factors with an eigenvalue higher than one. (Table 2, panel B) This single factor explains 50% of the variance. The rotated component matrix shows that all questions load (> 0.5) on a single factor. Ergo, convergent validity of the TIGHT variable is guaranteed.

3.3.3.2 Discriminant validity

When running the factor analysis with the questions of both real earnings management and budget control tightness I find three underlying factors with an eigenvalue higher than one. The first factor (eigenvalue of 3.1) clearly relates to budget control tightness. Just like the results of the convergent validity test for real earnings management, the second (eigenvalue of 1.6) and

(24)

third (eigenvalue of 1.2) factor relates to the real earnings management questions. Considering the fact that all questions related to budget control tightness load on 1 single factor and that the questions related to real earnings management - analogously to the convergent validity test - load on 2 factors, I conclude that discriminant validity is safeguarded6_.

Table 2 Rotated Component Matrix factor loadings PANEL A: Real Earnings Management

Question

Component 1 Component 1

Please indicate how frequently you, or someone in your business unit, took the following actions in order to comply with financial controls ( 0= option is not available to you, you cannot perform such an action)

96. Deferring or accelerating a needed expenditure 0.808 0.177

97. Deferring or accelerating a sale 0.299 0.681

100. Provide incentives (e.g. price discounts) to customers to buy more this

period 0.195 0.864

101. Decrease discretionary spending (R&D, advertising, maintenance) 0.272 0.563 102. Delay or accelerate starting a new project (IT, support) 0.805 0.162 103. Delay or accelerate hiring an employee 0.579 0.019

PANEL B: Budget Control Tightness Question

Component 1

To what extend do you agree with the following statements on the behavior of your superior?

79. Puts much emphasis on meeting the budget 0.570 80. Does not easily accept budget revisions during the year 0.801 81. Has a detailed interest in specific budget line items 0.657 82. Does not lightly tolerate deviations from interim budget targets 0.833 83. Is intensively engaged in budget related communications 0.631

3.3.4 Moderating variable: own-level performance measures

To measure the weight put on own-level financial performance measures (or local), the manager has to indicate the weight his superior puts on each measure when evaluating his performance. These measures are based on prior research by Bouwens and Van Lent (2007) and make a distinction between summary measures like share price or EBIT at a higher organizational level and

(25)

level, or local, measures at business unit level. Furthermore, at business unit level a distinction between financial (EBIT, ROI) and non-financial (customer, process) is made.

I construct the moderating variable (OWNLEVELPERF) by summing the local financial performance measures.

3.3.5 Robustness checks and control variables

A survey in general, but specifically the comprehensive design of this survey, offers many opportunities to control for other factors and thus improve internal validity. Obviously size, measured by FTE will be taken into account. As discussed earlier in this section, size is a meaningful variable because MCS become more important as the number of employees that “have to be managed” increases. Because the distribution of the individual item scores is skewed, the variable is constructed by a log transformation of the FTE variable.

Many researchers have shown that meeting the bonus target is an important motive to engage in earnings management. That’s why a dummy variable will be taken into account: BONUSDUMMY returns 1 when respondents have a financial bonus and consequently 0 when they have not. Another motive for earnings management, is reporting “good” results. Therefore, I expect that when the organization performance well, there will be less need to manage earnings. In the questionnaire, there are three questions about organizational performance7_.

Hence, a variable ORGPERF will be constructed out of three questions. Factor analysis and cronbach’s alpha of 0.88 confirm the validity and reliability of this variable.

Regarding to the business unit manager characteristics, a TENURE variable will be taken into account. Managers who have many years of experience in their current position have more knowledge regarding to when and how to manage earnings. On the other hand, from a career concerns perspective it can be argued that a manager who is new in his job directly wants to report good results. When the results are disappointing, the risk of gaming the numbers could be higher.

Finally, because non-profit organizations don’t have a natural “profit-test” like commercial organizations, I expect that the effects of budget control tightness on real earnings management will be lower for non-profit organizations. That’s why a dummy variable is created and controlled for. The variable is called PROFITDUMMY and returns 1 when the organization is a profit firm and 0 for a non-profit organization

(26)

3.3.6 Model

I test the hypothesis with a multiple regression model based on ordinary least squares (OLS). Based on the variables as mentioned in the previous paragraphs, the model for H1 is:

H1 REM = β 0 + β 1 TIGHT + β 2 LnFTE + β 3 ORGPERF + β 4 TENURE + β 5

BONUSDUMMY + β 6 PROFITDUMMY + ε

H2 tests the impact of own-level performance measures on the relationship found under H1. For this test I use a moderated OLS regression analysis. To do so I have to construct a variable for the moderation. Doing this by simply multiplying TIGHT with OWNLEVELPERF, will result in multicollinearity problems. Therefore new standardized variables, or z-scores, of TIGHT and OWNLEVELPERF have been constructed.8_{The variable OWNLEVELPERF and}

the moderating term based on the new variables zTIGHT and zOWNLEVELPERF are added to the regression model:

H2 REM = β 0 + β 1 zTIGHT + β 2 LnFTE + β 3 ORGPERF + β 4 TENURE +

β 5 BONUSDUMMY + β 6 PROFITDUMMY + β 7 zOWNLEVELPERF +

β 8 (zTIGHT x zOWNLEVELPERF) + ε

The results of the regressions analysis are discussed in the next chapter.

(27)

4 Results

This section starts with an overview of the descriptive statistics. To understand the key concepts of real earnings management and budget control tightness, I think it is valuable to present not only the statistics of the constructed summary variables, but also the statistics of the answers to the individual questions. Second a correlation table is presented which gives the first insights in possible relations between the variables. Third, the results of the multiple regression analysis are discussed. Finally additional analyzes with some robustness tests are presented.

4.1 Descriptive statistics

Tabel 3 shows the statistics for the dependent, independent and moderating variable.

Out of the 98 responses, only 1 business unit manager stated that (s)he never undertakes real earnings management actions. Table 3, panel A shows that the average score on earnings management is 2.8. Considering the 5-point scale, this indicates that real earnings management occurs on a regular basis, but not very frequently. There is no significant difference between nonprofit- and profit organizations.

With an average score of 3.1, the highest score is on question 103: “delay or accelerate hiring an employee”. The lowest score is on question 100: “provide incentives to customers to buy more this period”. A possible explanation for these high and low scores is that the effects of question 100 are directly resulting in a lower cash inflow, which can be perceived as directly sacrificing value. Vice versa, the effects of question 103 are resulting in less cash outflow, which can be seen as more legitimate.

On average budget control tightness is on a moderate level (3.1). There is no significant difference between nonprofit- (average of 3) and profit organizations (average of 3.1). Tables 3, panel B shows the statistics per question. The highest score (3.4) is on question 79 which relates to the pressure to achieve budget goals. Although this indicates a relatively high level of pressure to perform, the other questions related to pressure (80 and 82) have a score below average.

Table 3, panel C shows the composition of the performance measures. On average, 27.7% of the measures used are local financial performance measures. The scores on summary performance measures are low: 11.7%. Mainly due to the low score on the share price measure9_.

(28)

Table 3 Statistics dependent, independent and moderating variable (N=98) PANEL A: Real Earnings Management

Question N Answers 0 (N/A) N Answers 1 or higher

Min Max Average Median Std.

Please indicate how frequently you, or someone in your business unit, took the following actions in order to comply with financial controls ( 0= option is not available to you, you cannot perform such an action)

96. Deferring or accelerating a needed

expenditure 7 91 1.0 5.0 2.9 3.0 1.12

97. Deferring or accelerating a sale 22 76 1.0 5.0 2.6 3.0 0.99 100. Provide incentives to customers to

buy more this period 29 69 1.0 5.0 2.4 2.0 1.12

101. Decrease discretionary spending

(R&D, advertising, maintenance) 15 83 1.0 5.0 2.6 3.0 0.90 102. Delay or accelerate starting a new

project (IT, support) 7 91 1.0 4.0 3.0 3.0 0.98

103. Delay or accelerate hiring an

employee 3 95 1.0 5.0 3.1 3.0 0.90

Variable REM 1.0 4.3 2.8 2.5 0.61

PANEL B: Budget Control Tightness

Question Min Max Average Median Std.

To what extend do you agree with the following statements on the behavior of your superior?

79. Puts much emphasis on meeting the budget 1.0 5.0 3.4 4.0 0.97 80. Does not easily accept budget revisions during the year 1.0 5.0 2.9 3.0 1.11 81. Has a detailed interest in specific budget line items 1.0 5.0 2.8 3.0 0.99 82. Does not lightly tolerate deviations from interim budget targets 1.0 5.0 2.9 3.0 0.95 83. Is intensively engaged in budget related communications 1.0 5.0 3.2 3.0 1.01

Variable TIGHT 1.5 4.8 3.1 3.0 0.70

PANEL C: Composition of performance measures

Item Average

70. Summary level: Share Price 0.3 71. Summary Financial Performance Measures 11.4 72. Local Financial Profit Measures 17.0 73. Local Financial Return Measures 10.7 74. Local Non-financial customer measures 21.0 75. Local Non-financial process measures 12.5 76. Local Non-financial learning measures 15.3

77. Other 6.7

(29)

Tabel 4 shows the statistics of the control variables. Organizational performance is on a moderate level (3.1). On average the managers have been working 5 years in their current position. The business unit has 70 FTE on average (median 32 FTE). 60% of the organizations are profit firms. Surprisingly only 32% of the managers indicate that they have a financial bonus agreement. Perhaps, there was some reluctance when answering the question related to bonus pay.

Table 4 Statistics control variables (N=98)

Variable Min Max Average Median Std.

ORGPERF: Organizations Performance 1.0 5.0 3.1 3.0 0.83 TENURE: Years working in current position 0.1 27.0 5.0 4.0 5.52 FTE: Number of employees in business unit 4.0 450.0 70.5 32.0 82.5

N %

PROFITDUMMY: Profit organizations 59 60.2 BONUSDUMMY: Managers with financial bonus 32 32.3

4.2 Correlation analysis

Table 5 shows the Pearson correlations between dependent-, independent- and control variables. As expected, there is a significant positive relationship between REM and TIGHT (r = 0.224, p = 0.027). Also as explained in the control variables paragraph, there is a significant negative relationship between ORGPERF and REM (r = -0.338, p = 0.001). Furthermore there is a positive relationship between the use of own-level financial performance measures (OWNLEVELPERF) and profit organizations (r = 0.384, p = 0.000). But also between OWNLEVELPERF and the use of financial bonuses (r = 0.330, p = 0.001). This makes sense because it is expected that the use of local financial performance measures (such as EBITDA) and the use of financial bonuses are higher in profit organizations. These correlations underpin the need of the variable PROFITDUMMY in the multiple regression analysis.

The correlation table shows no high correlations between independent (TIGHT) and control variables. This is a first indication that there are no multicollinearity problems. Multicollinearity will be further tested during the regression analysis.

(30)

Table 5 Pearson Correlations between all variables (N=98)

REM TIGHT ORG

PERF FTE Ln LEVEL OWN PERF

TENURE PROFIT

DUMMY DUMMY BONUS

REM 1 TIGHT 0.224 1 0.027* ORGPERF -0.338 0.140 1 0.001* 0.169 LnFTE -0.066 0.049 0.013 1 0.516 0.635 0.895 OWNLEVELPERF -0.155 0.021 0.230 0.268** 1 0.128 0.835 0.023* 0.008 TENURE -0.079 0.041 -0.057 -0.197 -0.119 1 0.440 0.691 0.577 0.520 0.242 PROFITDUMMY -0.130 0.098 0.170 0.019 0.384** -0.100 1 0.201 0.339 0.095 0.850 0.000 0.329 BONUSDUMMY -0.186 0.112 0.151 0.057 0.330** 0.046 0.433 1 0.067 0.273 0.136 0.579 0.001 0.654 0.000** P value in italic.

**. Correlation is significant at 0.01 level (2-tailed) *. Correlation is significant at 0.05 level (2-tailed)

4.3 Regression analysis

Table 6 shows the results of the OLS multiple regressions. Multicollinearity is assessed by examining the variance inflation factors (VIF). The VIF values for H1 are at an acceptable level between 1.0 and 1.3, so there are no multicollinearity problems. When testing H2 with the interaction term based on z-scores, I also find acceptable VIF values. In particular the VIF value of the moderating term is 1.1.

H1 predicted that budget control tightness (TIGHT) is positively related to real earnings management (REM). The outcomes of the regression analysis for H1 are presented in table 6, panel A. The analysis of variance (ANOVA) shows a highly significant model fit (p = 0.00), which explains 18.6% (R2_{) of the variance in REM. As predicted, the OLS regression provides}

strong evidence that TIGHT is positively related to REM (β 1 = 0.26, p = 0.00). As expected the regression also provides strong evidence that organizational performance (ORGPERF) is

(31)

negatively related to REM (β 3 = -0.27, p = 0.00). The other control variables don’t have a significant impact on REM.

H2 predicted that the size and nature of the relationship found under H1 changes as a function of the use of own-level performance measures. Therefore moderation term zTIGHT x zOWNLEVELPERF together with the standardized variables zTIGHT and zOWNLEVELPERF are added to the regression model. As shown in table 6, panel B, the results of this regression analysis suggests that the use of own-level performance measures does not impact the relationship between TIGHT en REM. Ergo, I find no support for H2.

4.4 Additional analysis

The results as presented in table 6 clearly show that the control variables, other than organizational performance and profit organizations, don’t have a significant impact on the relationship between TIGHT and REM. Controlling for these other variables adds credibility to the statement that there is a highly significant positive relationship between REM and TIGHT. However, the adjusted R2_{(19%) of the total model leaves 81% of the variance in REM}

unexplained. The variance in REM has to be explained by other factors, that we missed in our survey and analysis. Therefore, I performed additional analysis10_{on both H1 and H2. First to see}

whether the results as presented above are still valid when modifying the regression model by changing, splitting-up, adding or deleting variables. Second to attempt to improve R2_{of the total}

model.

The first additional analysis is driven by the results of the factor analysis on REM. In this analysis I found two underlying factors for REM. For these factors new variables were constructed: REMCOST based on questions 96, 102, 103 which are more related to “cost based real earnings management” and REMREVENUE based on questions 97 and 100 which are more related to “revenue based real earnings management”11_{. The model with REMCOST as}

dependent variable resulted in an adjusted R2_{of 17% again with a highly significant relationship}

between REM and TIGHT and REM and ORGPERF. The model with REMREVENUE as dependent variable resulted in an adjusted R2_{of 21% also with a significant relationship between}

REM and TIGHT. No significant relationship with ORGPERF was found. The results of this first additional analysis show that splitting up the dependent variable does lead to a higher R2_for

10_{The details of the additional analysis can be found in appendix B}

(32)

“cost related” REM. Hence, in both cases the relationship between REM and TIGHT remains significant.

The second additional analysis is related to size measure. The survey aimed for business unit managers with at least 20 FTE reporting to them. 19 respondents did not meet this criterion. To test whether these “small sized” units influenced the results, I ran a regression without these respondents. The model for H1 remained significant with an adjusted R2_{of 15% and the impact}

of TIGHT and ORGPERF on REM remained significant. The model for H2 again showed no significant impact of own-level performance measures.

The third additional analysis is related to the nature of the organization. Individual analyzes for both profit- as non-profit organizations are performed by splitting up the dataset based on the PROFITDUMY. This resulted in a regression on profit organizations (N=59) and a regression on non-profit organizations (N=39) with interesting results. The model for non-profit organizations is not significant, but the model for profit organizations is significant (p = 0.00) with an adjusted R2_{of 32%. Taking only profit organizations into account improves the model}

significantly: an increase in R2_{of 13% compared to the initial analysis. Furthermore, the positive}

relationship between REM and TIGHT remained significant (p = 0.00) with an improved beta coefficient: 0.46 compared to 0.26 as obtained in the initial analysis. The negative relationship between REM and ORGPERF also remained significant (p = 0.00) with a slightly improved beta coefficient: -0.29 compared to -0.27. The results of this additional analysis are striking. First, because the model for profit organizations explains approximately one third of the variance in REM. Second, the improved beta coefficient on TIGHT shows that the effects of TIGHT on REM are greater within profit organizations. Third, although the results of this survey show that REM occurs in both profit- as well as non-profit organizations12_{, within non-profit organizations}

REM cannot be explained by TIGHT. It seems that the pressure to manage figures in non-profit organizations is driven by other factors, which I did not capture in this study. Apparently, the pressure felt by business unit managers within the profit sector, as a result of tightness, makes them more susceptible to manage earnings.

The fourth additional analysis relates to H2 and examines whether changing the nature of the construct for own-level performance measures (OWNLEVELPERF), results in other findings for H2. In the initial analysis OWNLEVELPERF was constructed as a scale variable based on the accumulation of percentages given to the specific local performance measures items. For the additional analysis I constructed two dummy variables. The first is OWNLEVELPERFDUMMY

The effects of budget control tightness and the use of own-level financial performance measures on real earnings management