How budget authority and the mode of reporting are able to alter decision frames

(1)

Amsterdam Business School

How budget authority and the mode of reporting are able to alter decision

frames

Thesis

‘’History shows that where ethics and economics come in conflict, victory is always with economics. Vested interests have never been known to have willingly divested themselves unless

there was sufficient force to compel them.’’

― B.R. Ambedkar

Name: Patrick Doornekamp Student number: 10867856 Date: 16 June 2016

Word count: 12610

MSc Accountancy & Control, specialization Control

Faculty of Economics and Business, University of Amsterdam

I would like to thank Peter Kroos for buying me one beer.

Thank you, Elise Briedé for putting up with me for around 5 years.

Also I would like to thank Robbert the Hunter, for paying me to thank you.

Last but not least I thank Arjen van Soest for believing in me, inspiring me, mentally supporting me and calling me to go to the gym.

(2)

Statement of Originality

This document is written by student Patrick Doornekamp 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)

Abstract

Participative budgeting is often used by superiors to gain private information of subordinate managers. But when budgets are used for performance evaluation, it could entail an important side-effect of budgetary slack. When managers are able to make realistic and attainable targets for themselves they could be incentivized to misreport their budget and consume excess resources. Previous literature suggest that the intrinsic motivation to report honestly is crowded out by the extrinsic motivation of superior budget authority. In this budget-context superiors are able to accept or reject budget proposals of the subordinate. This paper examines if managers tendency to misreport depends on the allocation of final budget authority. In my 2x2 experiment managers either decided themselves whether they want a project to start or this authority is allocated to the superior. Also, I investigate whether the mode of reporting has an effect on the opportunistic behavior of the subordinate. The budget reporting either requires a factual assertion or none. Although the results of this study are in line with previous literature, it shows no main effects of final budget authority and the mode of reporting. Less slack is created when superiors have final authority due to fairness concerns of the

subordinate than when subordinates hold this authority. Besides this, subordinates who have to report their budgets based on factual assertions build less slack than when no factual assertions are required. This study contributes to a better understanding of the underlying motivations behind the deleterious side-effect of slack building behavior.

(4)

1. Introduction

Budgeting is one of the most important tools in management control frameworks (Liessem et al., 2015). It is used for both coordination and motivation and is an important control

mechanism in evaluating performance to align individual and corporate interests. When budgets are used for incentivizing, budgetary slack becomes an important issue. Budgetary slack occurs when a subordinate understates their capabilities in their budget report (Hobson et al., 2011). The subordinate has private information about factors that are influencing his or her performance (Waller, 1988). This private information needs to be communicated to the superior using budgets. When this information is used to set a norm for performance and evaluation purposes, a subordinate can be incentivized to bias this communication through distorting their capabilities. This problem is mostly recognized in participative budget-setting, where subordinates are given the opportunity to set relatively easy targets. This creation of budgetary slack is a type of opportunistic behavior and reduces the value of the budgeting process. It can impose an ethical dilemma on the individual when it is created, as the

subordinate receives excess resources in such a way that violates social norms (Hobson et al., 2011). Evidence suggest that subordinates strategically manipulate this budget-setting process which is detrimental for a firm (Douthit & Stevens, 2015). Previous research shows that this manipulation can be affected by the use of formal budget authority and the mode of reporting that influences the framing of the situation as an ethical dilemma (Douthit & Stevens, 2015). When ethical considerations come into play, honest behavior may become more prevalent, decreasing the likelihood of strategic behavior that increases budgetary slack. Though, there hasn’t been much research on this moral framing of individuals in relation to formal control systems.

1.1 Background on formal budget authority and the mode of reporting

How individual’s react in budget settings is strongly related to the way they perceive the ethicality of the situation. However, prior literature is quite inconclusive about the role of ethics in a budgetary context. According to Jensen (2003) subordinates do not believe

themselves that they are lying or behaving unethically when building slack into a budget. The budget system creates expectations of managerial behavior which supports concealing

information. A result of this is that subordinates view the budget-setting process as a game rather than an ethical dilemma. How individuals judge a certain setting is called ‘Framing’ and this judgment can be divided into an ethical and a business decision frame (Tenbrunsel & Messick, 1999). Besides Jensen, also Schatzberg & Stevens (2008) find no evidence that

(7)

ethical concerns are raised when creating budgetary slack. They explain that high budgetary slack may decrease ethical concerns. When more and more slack is built into budgets, social norms are created for such behavior confining any ethical considerations making them to see the situation as a business decision. However, a number of studies contradict these findings by suggesting that budgetary slack is sometimes viewed as a moral dilemma where ethical

considerations are raised and an ethical decision frame is adopted. This causes individuals to judge budgetary slack as unethical, reducing slack building as a consequence of their moral judgment (Hobson et al., 2011). The study of Douglas and Wier (2000) indicate that there is a significant correlation between the ethical position of an individual and budgetary slack in a way that slack building behavior decreases (increases) when the respondent is more (less) ethical.

Final budget authority seems to have an effect on this relationship between budgetary slack and an individual’s (ethical) judgment. Final budgetauthority is the authority of the superior or subordinate to approve or reject a budget proposal. Rankin et al. (2008) concluded that when superiors have final budgetauthority in the budget-setting process, slack was significantly less apparent than when subordinates had this budget authority. In this experiment two different budget-setting procedures were set up: one where the superior had final authority over the budget and another where the subordinate has final authority. They explain these results as a consequence of the ‘’crowding out’’ theory (Rankin et al., 2008). This theory suggest that extrinsic control such as superior authority, crowds out the intrinsic motivation to behave honestly in a budget-setting process and activates strategic thinking. But this doesn’t explain the higher level of slack when subordinate authority is applicable

compared to superior authority. The study of Hobson et al. (2011) explains this: when superiors are not allowed to reject budget proposals this may increase the moral judgment of the subordinates regarding slack and generate a moral frame where every individual reacts differently to various moral frames. However, this is contradicting with findings of Stevens (2002) where individuals who judged slack as unethical build less slack into their budgets.

Not only the allocation of budget authority but also the mode of reporting budgets seems to have a an effect on the amount of budgetary slack that is built. When explicit budget communication is required by making factual assertions, subordinates are more refrained to show dishonest behavior resulting in less budgetary slack (Douthit & Stevens, 2015; Rankin et al., 2008). This can be explained because explicit reporting provides the subordinate with the opportunity to make untrue statements which raises awareness of the importance of honesty making individuals adapting an ethical decision frame. But with implicit reporting

(8)

without factual assertions this opportunity is not recognized by the subordinate and a business decision frame is taking limiting the importance of honesty and making slack building more prevalent.

1.2 Research question

Given the conflicting findings of prior studies I will conduct an experiment where I investigate the following research question:

Does final budget authority and the mode of reporting have an effect on the ethical decision-making process of individuals in a budget-context?

The theory that I test with using an experiment is the framing theory and the crowding-out theory. The framing theory is based on the ethical decision-making model of James Rest (1986) that comprises of a four-step process: moral awareness, moral judgment, moral intent and moral behavior (O’Fallon & Butterfield, 2005). The theory suggests that an individual has to frame the situation as having a moral issue (moral awareness) before deciding whether a decision is morally right or wrong (moral judgment) (Hobson et al., 2011). In this way different frames can be taken, each leading to a different decision. It explains how intrinsic motivations can be driven out by extrinsic ones depending on the frame that is taken (Rankin et al., 2008). Thus, based on these theories, I expect that budget authority and the mode of reporting will have an effect on the ethical behavior of individuals influencing the amount of budgetary slack.

To test this research question I will conduct a budget reporting experiment under 60 both students and non-students. In this experiment the respondents are assigned a role as subordinate manager who is given the task to implement a project on behalf of the superior. The subordinate manager has private information about the estimated cost, and therefore the estimated profit, required for implementation of the project which needs to be reported to the superior manager. The superior is the residual claimant and the subordinate receives a fixed salary and any potential surplus extracted due to strategic reporting. The amount of the reported budget is measured on a continuous scale. The budget authority (superior versus subordinate authority) and the mode of reporting (explicit or implicit reporting) will be manipulated. In the superior authority case the superior has the final authority to reject the project based on the reported budget of the subordinate. In the other case, the subordinate holds this position. Explicit reporting means that the respondent has to

(9)

report the actual cost of the project that is given in the case. With implicit reporting not the actual costs are reported, but indirectly by calculating and reporting the estimated profit of the project.

1.3 Relevance

As previous studies are not conclusive about the relationship between formal authority and ethical behavior and not much research has been done on formal authority in relation to the decision-making process I would like to contribute to this new field of research with the use of the aforementioned research question in a numeral ways. First, given the conflicting findings of prior studies on the relationship between formal budget authority and ethical decision making, I aim to contribute to this stream of literature by re-examining this research question. Second, I also extend prior research by also examining how the mode of reporting (explicit versus implicit) affects the ethical dimension of reporting choices. Third, I contribute to the literature on budgetary slack. That is, given the private information that often resides with lower-level employees it is often argued that firms can benefit when they make use of their knowledge, for example through participative budgeting. It is, however, also argued that this may be costly as lower-level employees may want to increase budgetary slack. This study contributes by examining how allocation of formal budget authority and the mode of reporting affects budgetary slack. Fourth and final, this study also features a societal contribution given the prevalence of budget systems in organizations. That is, this study may contribute by deepening our understanding in important dysfunctional side-effects of budgets like budgetary slack.

1.4 Structure

The remainder of this paper will be structured as follows. Chapter two will consist of a literature review that will contribute to the development of my hypothesis. Chapter three will consist of an outline of the research methodology that is used to investigate the research question and test the hypotheses. In the fourth chapter the findings of the research will be explained. The final chapter provides a discussion of the main conclusions, research limitations and opportunities for future research.

(10)

2. Literature review and hypothesis

2.1 Budget-setting

Budgeting may be characterized as one of the most important tools of an organization and can be described as the cornerstone of management control systems (Hansen et al., 2003). The use of budgets by organizations can be divided into four main categories: operational planning, performance evaluation, communication of goals and strategy formulation (Hansen & Stede, 2004). Lower-level employees (now: subordinates) often have important information about their subunit’s circumstances that upper management does not have (Church et al., 2012). This information asymmetry can be harmful for organizations as this private information is vital in order to improve the efficiency of the aforementioned categories.

Information asymmetry in general can be described as the discrepancy in the possession of information by the superior and the subordinate (Lau & Eggleton, 2003). In a budget-setting process it often arises when the subordinate has private information that is relevant for the decision process related to budgeting (Dunk, 1993). To constrain this

information asymmetry, subordinates are often asked to communicate the private information by incorporating it into budgets using participation. Participation is the amount of influence and involvement in determining budgetary targets a subordinate is granted by its superior (Kramer & Hartmann, 2014). Besides limiting information asymmetry, it also enables subordinates to make sure that budgets are realistic and attainable (Lau & Eggleton, 2003). But participation also entails gamesmanship by creating slack in such a way that revenue, expenditures and investments are skillfully timed to easily achieve budgets (Hansen et al., 2003).

Different definitions explain budgetary slack: agency theory researchers describe slack as ‘’the tendency to shirk’’ or “excess consumption of perquisites” and behavioral researchers as ‘’managerial biasing’’ (Merchant, 1985). As for rational economic subordinates it may be in their best interest to create slack, it is not in the best interest of the overall

organization as strategically manipulating budget processes can even be detrimental for a firm (Church et al., 2012; Merchant, 1985; Douthit & Stevens, 2015). Budgetary slack is often recognized when employees are involved in budget-setting through participation (Lukka, 1988; Lau & Eggleton, 2003; Dunk, 1993). Besides participation, this slack building behavior will mostly occur in those situations where budgets are used for performance evaluation because it gives subordinates incentives to act strategically in budget reporting (Christensen, 1982; Dunk, 1993; Baiman & Lewis, 1989).

(11)

The amount of which slack is successfully created depends on information asymmetry. When information asymmetry is high, the likelihood to create slack is also high, but this relationship may be influenced by several contextual factors (Fisher et al., 2002). One example of a

contextual factor is an environment where a coercive control system is applicable with a tight budget emphasis (Onsi, 1973). In this way, participation doesn’t allow for enabling means to provide attainable targets but rather provides a way to circumvent very strict targets.

2.2 Ethical decision making

The study of ethics can be broadly divided into two branches: descriptive and normative ethics (Tenbrunsel & Smith‐Crowe, 2008). Descriptive ethics describes and studies what people do and tries to explain moral behavior without taking a moral position (Tenbrunsel & Smith‐Crowe, 2008; De Graaf, 2006). Normative ethics is prescriptive in nature and describes what people should do by establishing norms or rules to weigh against behavior to determine what is right and wrong (Tenbrunsel & Messick, 1999). The definition of ‘Ethical’ and ‘unethical’ will be a combination of these two ethical branches. I will use the definitions of ethical and unethical decisions as explained by Jones (1991). According to him, decisions are ethical when it is both legal and morally acceptable in the broader community and in this way an unethical decision is illegal or morally unacceptable in the broader community.

But before a decision is made, which may be both ethical or unethical, a reasoning process is undertaken by the individual. This process is defined by James Rest (1986) who divides an individual’s reasoning process in four stages that an individual passes when he or she is confronted with an ethical issue: moral awareness, moral judgment, moral intention and moral behavior (Rottig et al., 2011). This means that an individual first has to recognize the ethical issue before the moral decision process commences, which is triggered by moral awareness.

Moral awareness is described as the ability of an individual to recognize a situation that contains a moral issue. With that moral issue, a potential harm or benefit to other people of this individual’s action is associated. Moral judgment is the process in which an individual formulates and evaluates possible solutions that are most morally justified. Coming towards this possible solution, a process of moral reasoning is undertaken. Moral intention refers to the intention of an individual to choose the most ethical solution over another solution that represent a different value (for example power) to that individual. This individual’s intention is ethical when it chooses the most moral one. Moral action consists of

(12)

the individual’s behavior and refers to the action he or she takes which forms the result of the moral reasoning process.

2.3 Framing theory

In order to achieve moral awareness and thus initiate the moral decision-making process, it is essential to understand how decision makers code or categorize their decision (Tenbrunsel & Smith‐Crowe, 2008). This coding of their decision or ‘decision frame’ is contingent on the situational context and determines the reasoning process within that frame. Decision makers first determine the type of the decision and the frame that matches it. Decision frames are often divided into business decision frames and ethical decision frames (Kouchaki et al., 2013; Tenbrunsel & Messick, 1999; Tenbrunsel & Messick, 2004; Tenbrunsel & Smith‐ Crowe, 2008). Within ethical decision frames, ethical considerations are raised and thus the ethical decision-making process is initiated. Business decision frames on the other hand, activate a calculative cost-benefit analysis where self-interest is more important than the interest of others regardless of the ethical consequences (Kouchaki et al., 2013). In this case, no ethical decision-making process is activated and decision-makers are not morally aware. The subsequent cognitive process which results in certain behavior depends on the frame that is adopted. Within an ethical frame, the processing is more automatic, directing to do the right thing. The processing in a business decision frame is more complex and depends on the chance to be punished or rewarded (Kouchaki et al., 2013; Tenbrunsel & Messick, 2004).

Previous studies investigated the effect of the decision frame on the behavior of individuals. Kouchaki et al. (2013) investigated this using four studies. In these studies, it became clear that business decision frames led to selfish and unethical behavior. One of the studies presented a hiring situation where the respondents were responsible for hiring a new assistant manager. It indicated that the respondents were more likely to hire candidates who held confidential information valuable for the firm when a business decision frame1 was taken than when such a frame was not taken, implicating selfish behavior. Another study contained a visual perception task in which respondents were presented two squares with 20 dots scattered inside of it. It required respondents to identify in which of the two presented triangles more dots appeared, and based on this report they would earn money. More money was granted when respondents indicated that the right triangle had more dots, thus more money was earned with lying. Within the business decision frame, respondents were behaving

1

No distinction was made in this study between business decision frames and ethical decision frames. Situations are compared between business decision frames and no frame taken at all which can be compared to an ethical frame, as the ethical position of the individual mainly determines the reasoning process within this frame.

(13)

more unethical as they were more likely to indicate that the right side had more dots when clearly the left side had.

2.4 Motivational crowding theory

The economic law that raising incentives leads to an increase in effort is invalidated by the theory of motivational crowding (Frey & Jegen, 2001)2. Because when this crowding-effect occurs, effort will decrease rather than increase when incentives are raised due to the fact that intrinsic motivation is crowded out by extrinsic motivation. Extrinsic motivation is defined as an incentive coming from outside the individual that is motivated (Frey & Jegen, 2001). Intrinsic motivation comes from the activity itself that is performed and no reward is received other than from this activity. The idea that intrinsic motivation can be crowded out by

extrinsic motivation, such as extrinsic control systems or monetary incentives, originates from a theory of Titmuss (Frey & Jegen, 2001; Georgellis et al., 2011). He argued that paying for blood donations would reduce the supply of donated blood because the monetary reward to donate blood would undermine the underlying social values reducing the willingness to donate it. While he lacked any empirical evidence, a large amount of later studies were able to provide this evidence.

The effect is caused by two psychological processes: impaired self-determination and impaired esteem (Frey & Jegen, 2001). When a person feels impaired in their self-determination because of external interference they are forced to act in a specific way, their internal motivation is substituted by this external control. But also when a person gets the notion that their intrinsic motivation in the form of involvement or interest is not

acknowledged he or she will give less value to this intrinsic motivation affecting their self-esteem and focus. The emphasis here is placed upon the word ‘impaired’, because only when this external intervention is seen as controlling, crowding-out will appear.

The process in which intrinsic motivation is crowded out by extrinsic control is demonstrated by a neurological study of Ma et al. (2014). The participants had to stop a watch from counting at a 5 seconds time point during three stages. During stage one and three they were incentivized with a fixed payment, but only during stage two a small monetary reward was introduced for every successful hit that was removed in stage three. To investigate the crowding out effect an EEG was carried out whereby feedback-related negativity (FRN) is used to measure the affection of the participant to a certain activity where a high FRN

2

This economic law that higher incentives leads to an increasing effort can also be defined as the opposite of crowding-out: crowding-in.

(14)

indicates high motivation. Evidence for this crowding effect comes from differences in this FRN after introduction of the small extrinsic reward in stage two and removing it in stage three. The participants were less affective to the outcome and thus gave less importance to their task after the reward was introduced. Their original intrinsic motivation, playing for fun, diverted to the extrinsic motivation of completing the task to gain the monetary reward. Removing the incentive again didn’t have any effect on the extrinsic focus.

A variety of examples were delivered by numeral studies. Carpenter and Myers (2010) found that volunteer firefighters that were rewarded with a small pay, were less likely to respond to emergency calls. According to Frey and Goette (1999) volunteers were showing less effort when they received a financial reward. The number of late-coming parents who collect their children at day-care centers increased when a fine for coming late was

introduced, and was still that high when the fee was removed (Gneezy & Rustichini, 2000). And once children are being paid for mowing the lawn, will never do this, or any other housework, for free again (Frey, 1997).

2.5 Hypotheses development 2.5.1 Final budget authority

Firms differ in the degree in which they delegate budget authority to lower-level employees and allow them to participate in budget-setting. An example of a participative budget setting comes from an agency model of Antle and Eppen (1985) in which an investment project is considered by a superior. Both subordinate and superior know the revenue and expected cost of the project once implemented, but only the subordinate has the private information about the actual cost. The superior attempts to get this information by letting the subordinate submit a cost budget proposal. In practice, most firms include final budget authority in their

participative budgeting-setting (Douthit & Stevens, 2015). In these settings superiors have the authority to reject or approve the budget that is being proposed by the subordinate. A budget-context in which subordinates are able to approve and reject budgets are less realistic in that managers are generally not allowed to set their own budgets unilaterally (Hannan et al., 2006). However, in for example investment centers, the responsible manager is authorized to make investment decisions up to a specified point.

When the superior is able to reject budgets, subordinates’ intrinsic motivation to behave honestly will be crowded out by the external control of final budget authority, limiting their ethical concerns (Rankin et al., 2008). The subordinate adopts a business frame in which a strategic reasoning process is activated where a trade-off is made between self-interested

(15)

wealth maximization and the interest of others (Haesebrouck, 2016; Rankin, et al., 2008; Hobson, 2007). Because, the threat of rejection of the budget motivates managers only to appear honest. Therefore, they will build slack at a level where there is an equal distribution between the preferences of the superior and self-serving benefits. However, when the subordinate has final authority over the budget, they may frame the budgeting task as an ethical dilemma which is associated with budgetary slack.

Both studies of Rankin et al. (2008) and Fisher et al. (2000) find evidence that less slack is built when the superior has final budget authority than when subordinates have this authority. Subordinates were more concerned about the fairness of the proposal in the superior authority case, indicating the fear for rejection of the budget (Rankin, et al., 2008). This causes cost-benefit thinking which is evident from the level of slack being built. The amount of slack that was built by the subordinates when reporting the budgets of Rankin’s study in case the superior had final authority was around $0,5 on a range of $0 to $1 which is thus an equal split of the mean maximum possible slack. This is consistent with studies investigating ultimatum games where proposers offer a surplus and responders can accept or reject these. In these games, the offered amount is also generally around 30-50% of the maximum possible surplus (Stanca, 2010; Charness & Gneezy, 2008; Croson, 1996). Rankin et al. (2008) and Fisher et al. (2000) also show that when subordinates have final authority, more slack is built into the budget. This is because subordinates may adopt an ethical frame. Following this explanation, Hobson et al. (2007) and Schatzberg and Stevens (2008) also found out that individuals who judge slack as unethical, build high levels of budgetary slack. They explain this as individuals create social norms for such unethical behavior. However, this is

inconsistent with studies of Evans et al. (2001) and Stevens (2002). They found that ethical concerns restrain budgetary slack as individuals sacrifice wealth in exchange of honest reporting. Therefore I propose the following non-directional hypothesis:

(16)

2.5.2 Mode of reporting

Besides formal authority I also investigate whether the mode of reporting has an effect on honest budget reporting. The task of reporting budgets can be fulfilled using explicit(with factual assertions) or implicit(with no factual assertions) reporting. Recall the agency model of Antle and Eppen (1985). This budget-context requires an assertion of fact as the

subordinate is obligated to explicitly report the cost of the project to the superior. When he has to report not the cost but the profit of the project, than no factual assertion of costs is asked for and the task of reporting is dependent on the interpretation of the individual.

When building slack requires more explicit lying because of factual assertions, ethical concerns become more prominent as the moral awareness of the individual is triggered. Also, the ethical decision frame is adopted when creating slack involves explicit reporting. When no factual assertion has to be made, strategic concerns become prevalent as budget reporting requires implicit lying. In this case, a business frame is adopted where a trade-off is being made between self-interested behavior and honest reporting.

This is illustrated with studies that investigate forms of implicit and explicit communication. As Lundquist et al. (2009) provided evidence that individuals who explicitly have to formulate their lie while communicating private information were more concerned about honesty than individuals who confirmed to false pre-fabricated messages. Also both studies of Rankin et al. (2008) and Douhit and Stevens (2015) demonstrate that when factual assertions have to be made, individuals are more influenced by the importance of honesty than when it does not require factual assertions leading to less budgetary slack. Haesebrouck’s (2016) study shows that this is caused by the fact that these individuals were more affected by ethical concerns than individuals under the no factual condition. As reporting cost that results in a higher payoff for the subordinate than for the superior was seen as more unethical than when this is achieved by reporting profit, indicating that factual assertions offer ethical dilemmas. Therefore I propose:

H2: When budget reporting requires a factual assertion, this is negatively associated with budgetary slack.

(17)

3. Methodology

3.1 Experimental design

I test whether superior final budget authority and the mode of reporting have an effect on the slack creating behavior of the subordinate. To test this, I conducted a web-based

2*2-between- subject experiment. An overview of the four conditions is presented in Table 1. The experiment is based on a typical budget-context that originates from Antle and Eppen (1985). Just like this budget-context, respondents were given an experimental case in which they performed the role of manager. This manager receives private information about an investment project after which the manager is asked to communicate this to the owner. The manager has a monetary incentive to misreport this information as this increases their pay-off at the expense of the owner’s pay-off3. Two manipulations are included in the experiment. The mode of reporting is manipulated by requiring factual or no factual

assertions. Factual assertions are made when managers are asked to report cost to the owner. In the other experimental case, no factual assertions have to be made as the managers are asked to report a profit that is calculated with revenue and cost of the project. Besides the mode of reporting the final budget authority is manipulated by allocating this authority to either the owner or the manager. In table 1 an overview is given of the four manipulations and experiment cases.

Table 1 : Experimental Design

Budget Authority Superior Authority (SUP-A) Subordinate Authority (SUB-A) Mode of reporting

No Factual Assertion (NFA) NFA/SUP-A NFA/SUB-A

Factual Assertion (FA) FA/SUP-A FA/SUB-A

3.1.1 Independent variables

For the experiment I use a 2*2 experimental design as the mode of reporting and the final budget authority are manipulated at two levels.

3

In this experiment, the terms ‘manager’ and ‘owner’ are used instead of subordinate and superior. This is because more people are familiar with these definitions. For consistency purposes these terms are continued to be used in the remainder of this chapter.

(18)

The mode of reporting is manipulated according to the study of Rankin et al. (2008). The budget reporting is either explicit through factual assertions or implicit by

requiring no factual assertions. This is explained in the case as follows. In the factual assertion treatments, the managers observed the actual cost of EUR 90 and are then asked to give an indication of these actual cost. As the total revenue of the project equals EUR 200,

respondents were only able to answer within a range of 90 and 200. Answers had to be typed into a blank field in which a restriction was made of values lower than 90 and higher than 200. In the no factual assertion treatment, the manager was asked to indicate what the profit of the project will be, which is calculated as revenue minus actual cost. Since the actual profit is EUR 110, the managers had a range between 110 and 0 to report. A restriction of values below 0 and higher than 110 was made here also. The amounts of cost of 90, revenue of 200 and the fixed payment of 75 are consistent with Haesebrouck (2016). So the first independent variable consist of the budget report: either a cost report (factual assertion) or a profit report (no factual assertion).

The second independent variable is the manipulation of the final budget authority. This manipulation is consistent with the studies of Evans et al. (2001) and Rankin et al. (2008). In the superior authority treatment the owner is allowed to decide whether or not to implement the project. The manager decides to start the project in the subordinate authority treatment. In both cases however, the owner will take care of the funding once the choice is made to start the project.

3.1.2 Dependent variable

The dependent variable is the amount of slack that was created by reporting a cost or profit. Slack is calculated in two ways for the different budget reporting treatments. In the factual assertion scenario, slack is calculated as reported cost minus 90 (actual cost). Without factual assertions, slack is calculated as 110 (actual profit) minus reported profit. In this way, the maximum amount of slack is in both cases 110. To make certain that respondents would give a budget report between the desired restrictions, this was mentioned in the questions.

3.1.3 Control Variables

In the post-experimental questionnaire, respondents were asked questions known to be of influence on the independent and dependent variables. Examples are gender, work experience and age. The latter two variables are indicated in years. Specifically for this budget-context I created two new variables: ‘Ethicalpos’ and ‘Importance’. Ethicalpos is based on the study of

(19)

Haesebrouck (2016) and shows the rating of how honest and ethical the respondents think they are. This variable is created from the means of the two questions: ‘How honest do you

think you are?’ and ‘How ethical do you think you are?’ both rated on a Likert scale from 1

(strongly disagree) to 7 (strongly agree). The Cronbach’s alpha is 0,88 and given the threshold of 0.7, this new variable can be seen as reliable. Besides this, also the variable ‘Importance’ is created. It is based on the question ‘How much importance do you give to your task of

reporting cost to the owner?’. Its purpose is to find evidence for the crowding-effect as

studies that are discussed in paragraph 2.4 explain that individuals who are influenced by this effect, give less importance to a certain task than when this effect did not occur. It is measured on a scale of 1 (a great deal) to 5 (none at all) and recoded as seen in table 2. It will only be used to explain the underlying motivations of the participants in the supplemental analysis. For an overview of all the variables see table 2.

Table 2: Variables used in analysis

Variable Label Measure

ModeOfReporting The case that was presented: factual or no factual

assertions case

0 = No factual assertions 1 = Factual assertions

FinalBudgetAuthority The case that was presented: final superior authority or

subordinate authority

0 = Final superior authority 1 = Final subordinate authority

Slack Slack created by respondents. Calculated as:

• Reported Cost -/- 110 • 110 -/- Reported Profit

Continuous (from 0 to 110)

Gender ‘What is your gender?’ 0 = Male

1 = Female

Age ‘What is your age in years?’ Continuous (from 15 to 55)

Workexperience ‘How much years of work experience do you have

(part-time and fulltime)?’

Continuous

Education ‘What is your highest completed level of education?’ 1 = VMBO-TL

2= Havo 3= VWO 5= MBO 5 = HBO 6= WO bachelor 7= WO master 8 = other

EthicalPos ‘How honest do you think you are?’

‘How ethical do you think you are?’

Likert scale from 0 to 7: 1 = Very dishonest/unethical 7 = Very honest/ethical

(20)

Table 2: Variables used in analysis (continued)

Ethical ‘To report cost that leads to a significant higher payoff

for the manager than for the owner is unethical’.

Likert scale from 1 to 7: 1 = Strongly disagree 7 = Strongly agree

Fairtreat ‘I was concerned that the owner desired to be treated

fairly’.

Fairconcerns ‘I was concerned that the owner desired to be treated

fairly’.

Importance ‘How much importance do you give to your task of

reporting cost to the owner?’.

Likert scale from 1 to 5: 1 = None at all

5 = A great deal

3.2 Participants

A total of 238 respondents participated in the experiment. The participants that took part in the experiment mainly consist of students. Students are often used as subrogates for managers in experiments (O’Fallon & Butterfield, 2005). The choice to use students to simulate

managerial decisions is based on the following two studies. Trottier and Gordon (2001) conducted an experiment where they tested whether students would make the same decisions as managers. They found that the differences between the two groups mainly depend on work experience and education, both included as control variables in my paper. Cohen et al. (2001) investigated the differences in ethical decision-making between business students and

accounting professionals. They found only a few differences in ethics between the groups and most differences were found in gender. So in conclusion, using students is a valid method for measuring managerial (ethical) decision-making.

Of the 238 respondents that participated a total of 176 completed the experiment. This is because respondents stopped during the experiment or due to the fact that they were inactive for quite a long time. In order to make sure respondents understand the cases, four control questions were asked to assess whether the manipulations were correctly interpreted. 71 of the 176 respondents answered all four control questions right. The rest of the

respondents are excluded from analysis. With four manipulations each manipulation is required a minimum of fifteen respondents. As seen in table 3, all of the four manipulations meet this criteria. Of the 71 respondents included in the analysis, the minimum age is 19, the maximum age is 52 and the mean age is 24 years old. 68% of the respondents is male and

(21)

most of the respondents have an academic degree or higher (75%). At last, half of all respondents have work experience between two and six years.

Table 3: participant descriptive per scenario

Superior Authority Subordinate Authority

Factual assertion

N = 18

Average age = 24 Gender = 58.5% male Average Education = HBO Average work experience = 3.67

N = 18

Average age = 25 Gender = 66,5 % male

Average Education HBO-WO Bachelor Average work experience = 4.42 No Factual

Assertion

N = 18

Average age = 23 Gender = 75% male Average Education = HBO Average work experience = 3.36

N = 17

Average age = 24 Gender = 64,5% male

Average Education = HBO-WO Bachelor Average work experience = 6.76

3.3 Procedures and task

To conduct the research, a web-based experiment was distributed among students in English to increase the potential sample. Respondents were approached in several ways. Some of the students received an e-mail with the link leading to the experiment, others were approached via LinkedIn and Facebook. Besides this, I visited two universities in the Netherlands and asked students to take part in my research. They all received a link that directed them to the experiment. They could complete it by using their mobile phone, tablet or laptop. A brief introduction was displayed in which respondents were thanked for participating and asked to read the following case very carefully. After they read the brief instructions they were randomly assigned to one of the four experiment cases.

In these cases they were given a situation in which they performed the role as manager of a privately held Dutch firm where the CEO/owner holds 100% of the shares and that is active in a couple of unrelated industries. Of those industries, the sportswear division was the industry where the manager was responsible for. The manager is asked by the owner of the company to run an important advertising project to boost sales. Both manager and the owner know the expected revenue of EUR 200, but only the manager knows that the actual cost for implementation are EUR 90. Based on this information, the manager is asked to report a budget to the owner that will determine the payoffs of them both, but the owner will not know the pay-off of the manager. In case a profit had to be reported, the manager receives the difference between reported and actual profit. When the cost of the project had to be

(22)

reported, the manager received the difference between the reported cost and actual cost. Final budget authority was granted to the owner when the case explained that the owner will

evaluate the profitability of the project and decides whether or not to fund it. When the manager held budget authority, the case described that the manager is able to decide whether or not to start the project and the owner will fund it. In both cases however, managers

received EUR 75 as a fixed payment regardless of whether the project was implemented or not.

After they read the experiment, respondents were asked to answer four control questions to ensure they understood their task and pay-off. These questions are described in section 3.5. When they continued the experiment, they had to fill in the post-experimental questionnaire. Here six statements were presented and had to be ranked on a scale from 1 to 7 from ‘I totally disagree’ to ‘I totally agree’. These statements indicated how honest they wanted to be, how much fairness concerns they had about the owner and how much they cared about their self-importance. To measure the potential effect of crowding-out, they were also asked to rank how much importance they felt about their task to report cost/profit. This was done on a reversed scale to prevent acquiescence bias4. At the last page of the web-based experiment, respondents were asked to provide their age, education, gender and work

experience.

3.4 Validation

In order to ensure the validation of the results, some techniques were used. In case individuals at the universities were asked to fill in the experiment and when people were included in the same mailing-list, I explicitly told them that they could not talk to each other and may not look at each other’s cases while filling it in. This is to prevent people from being affected in their decisions and knowing what variables are being manipulated. Another one of these techniques made sure that respondents have to answer every question on their screen before getting to the next page. When participants missed one or more questions, a red text appears that directed them to the missing answer. This makes sure that data is not missing and therefore cannot be used for further analysis. Also, when closing the experiment all

respondents that opened but not completed the experiment were automatically deleted from the data. This also prevents data from missing. To ensure respondents answered the control questions based on what they just read, respondents weren’t able to go back to the previous page. Another technique ensures that the reported cost is never lower than the actual cost or

4

Acquiescence bias can be described as the tendency to fully agree to all of the questions.

(23)

higher than the revenue, and reported profit is never negative or higher than the actual profit5, by putting restrictions on the answering field. At last, only one IP address could complete the experiment once, making it impossible for respondents to fill in the experiment twice.

3.5 Manipulation checks

To ensure respondents understand the case scenario and that the manipulations are effective a total of four true or false-questions were asked. To ensure that they understand that the owner will not know that the manager is misreporting, the following statement is asked: ‘Only you

know the actual cost’. When the owner knows the pay-off of the manager, this would indicate

that he knows the manager is misreporting as well. Therefore the second statement that is asked exist of: ‘The owner will not know the payoffs’. The third question was asked to make sure that respondents know what their own pay-off will be. In case of factual assertions:

‘You'll receive the difference between the reported cost and the actual cost’ and in case of no

factual assertions: ‘You'll receive the difference between the reported profit and

the actual profit’. At last, to ensure that the manipulation of the final budget authority is

understood, the following question was asked in case of superior budget authority:

‘The owner will evaluate the profitability of the project and will then decide whether or not to fund the project.’ And in case of subordinate budget authority: ‘You will decide whether you want to start the project or not.’ For all the 71 respondents these worked effectively as they

all answered all four questions with ‘’true’.

5

As reporting a higher profit than the actual profit is not in the interest of the manager, it would indicate that respondents that report a higher profit would not understand the case scenario.

(24)

4. Results

4.1 Descriptive statistics

In table 4 the descriptive statistics of the amount of slack across the different manipulations are presented. A total of 17 respondents were classified in the No factual/Subordinate authority treatment and 18 in the other treatments. The mean amount of slack created by all respondents is 26,68 (with a Std. Dev. of 33,516) on a maximum of 110. With respect to hypothesis 1, superior authority is associated with slack as the mean amount of slack in the superior authority case is lower than the mean amount in the subordinate authority treatment. This could indicate that respondents felt a fear for rejection of their budget proposal.

Regarding hypothesis 2, the mean amount of slack in the factual assertion treatment is lower than the amount of slack created in the no factual assertion situation. This could indicate that when budget communication requires factual assertions, this could offer ethical dilemmas. To statistically test the hypotheses and gain additional insight into the motivations, I used a two-way between-subjects ANOVA and conducted a supplemental analysis in section 4.3.

Table 4: Descriptive statistics

Superior Authority Subordinate Authority Total Factual assertion N = 18 Mean = 28,56 Std. Dev.= 32,475 % of Slack = 25,96% N = 18 Mean = 22,50 Std. Dev.= 31,958 % of Slack = 20.45% N = 36 Mean = 25,53 Std. Dev.= 31,902 % of Slack = 23,21% No factual assertion N = 18 Mean = 21,94 Std. Dev.= 31,349 % of Slack = 19,95% N = 17 Mean = 34,12 Std. Dev.= 39,459 % of Slack = 31,02% N = 35 Mean = 27,86 Std. Dev.= 35,527 % of Slack = 25,32% Total N = 36 Mean = 25,25 Std. Dev.= 31,636 % of Slack = 22,95% N = 35 Mean = 28,14 Std. Dev.= 35,750 % of Slack = 25,58%

4.1.1 Normality and homogeneity

In order to use a two-way between groups ANOVA, there are a few assumptions that needs to be applicable to the data. The first assumption requires two categorical independent variables, in this case final budget authority and the mode of reporting and one continuous dependent variable, in this case slack. Besides this, the experimental groups need to consist of different subjects. Also is assumed that the population from wich the sample is taken is normaly distributed and homogenity of the variances exist. To test if the population is normaly

(25)

distributed a Kolmogorov-Smirnov test and the Shapiro-Wilk test is executed for each of the four experimental cases. In Table 5 is shown that the null-hypothesis of normality is rejected as all coefficients are below the applicable treshold of p <0.05. This is propably because the sample is relatively small. Besides a normal distribution, the four different scenarios should have the same variances. The technique that tests the equality of variance is the Levene’s test. The null hypothesis states that the variance in slack is the same for each of the four cases. The levene’s test gives a statistic of 1.135 with a significance level of 0.341. This means that the assumption of homogenity of variance is not violated.

Table 5: normality test

Kolmogorov-Smirnov Shapiro-Wilk

Case Statistic df Sig. Statistic df Sig.

FA/SUP-A 0.271 18 0.001 0.829 18 0.004

FA/SUB-A 0.263 18 0.002 0.757 18 0.000

NFA/SUP-A 0.314 18 0.000 0.734 18 0.000

NFA-SUB-A 0.259 17 0.004 0.795 17 0.002

FA= factual assertion NFA= no factual assertion SUP-A = superior authority SUB-A = subordinate authority 4.2 Hypotheses test

To test my hypotheses I conduct a two-way between subjects ANOVA. Note that the sample is not normally distributed, but that the variance in slack seem to be the same for all four conditions. Therefore an ANOVA can be conducted.

4.2.1 Hypothesis 1

Hypothesis 1 states that there is a relationship between superior authority and budgetary slack. This means that when the superior holds final authority over the budget, there will be

significantly more or less slack created than when the subordinate has final budget authority. Therefore, the underlying hypotheses that I test can be depicted as follows6:

H0: μ slack (SUP-A) = μ slack (SUB-A) H1: μ slack (SUP-A) ≠ μ slack (SUB-A)

6

Where SUP-A is superior authority and SUB-A is subordinate authority

(26)

Table 4 already showed that the mean slack in case of final superior authority is equal to 25,25 while for final subordinate authority the mean is higher (mean = 28,14). But as the results of the two-way ANOVA in table 6 shows, the relationship between final budget authority and slack is not significant (F = 0.145, p=0.71). This means that the amount of slack is independent of the hierarchical allocation of final budget authority.

Table 6 Results Two-way ANOVA

Dependent variable = slack Source Type III sum

of Squares

DF Mean Square F Sig.

Corrected Model 1721.896a 3 573.965 0.500 0.684

Intercept 50885.540 1 50885.540 44.329 0.000 Mode of reporting 111.160 1 111.160 0.097 0.757 Final budget authority 165.974 1 165.974 0.145 0.705 MOR * FBA 1473.623 1 1473.623 1.284 0.261 Error 66468.961 67 1147.905 Total 129156.000 71 Corrected Total 78631.549 70

a. R Squared = 0,155 (Adjusted R Squared = 0.061)

MOR = Mode of reporting FBA = Final budget authority 4.2.2 Hypothesis 2

The second hypothesis describes that when budget reporting requires a factual assertion, less budgetary slack is created. This means that the mean slack for the factual assertion case should be lower than the mean slack for the no factual assertion case. So the formulas for the aforementioned hypotheses can be formulated as follows7:

H0: μ slack (FA) = mean slack (NFA) H1: μ slack (FA) < mean slack (NFA)

As from table 4 could be concluded that the mean slack for factual assertion of 25.53 is lower than the mean slack without factual assertions of 27.86. But as table 6 shows, the relationship between the mode of reporting and slack is not significant with an F-statistic of 0.097

(p=0.76). This means that it doesn’t matter for the amount of slack created whether the communication requires factual or no factual assertions.

7

Where FA is factual assertions and NFA is no factual assertions.

(27)

4.3 Supplemental analysis 4.3.1 Correlations

In order to make sure that the control variables as explained in chapter 3.1.3 do not

significantly affect the dependent variable slack, I test the correlations of each variable. When one variable correlates too much with slack and randomization of the experimental groups doesn’t correct for this, I may have to control for it in my analysis as this variable explains variance that is not related to my hypotheses. For six variables, the correlations with each other are calculated using the Pearson correlation coefficient and the Spearman rank-order correlations as depicted in table 7. This table presents four significant correlations. The first two significant correlations are between education and age with r = 0.353 (P <0.01) and ρ = 0.407 (P<0.01) and work experience and age r = 0.658 (P<0.01) and ρ = 0.286 (P <0.05)8. More importantly, the third is the negative correlation between slack and gender with r = -0.351 (p<0.01) and ρ = -0.360 (p<0.01). This indicates that women build less slack than men. This is consistent with studies from Tse & Au (1997), Eynon et al. (1997), Keyton & Rhodes (1997) and Mason & Mudrack (1996) where women are found to be higher on an ethical scale. The last significant correlation could explain this: when an individual is more ethical they build less budgetary slack. Because the ethical position9 of the respondent is negatively associated with building slack with an r = -0.330 (p<0.01) and ρ = -0.478 (p<0.01). These results are consistent with Evans et al. (2001) and Stevens (2002) who also found that individuals with more ethical concerns build smaller amounts of slack.

Since ‘ethicalpos’ is a variable with only two items and measured after the experiment was done, it isn’t a very reliable measure. Therefore I do not control for this variable. Differences in gender however may not be adequately controlled for through randominzation. Therefore I aim to repeat the prior analyses while controlling for gender. It could be regarded as a covariate in an ANCOVA. But to use a variable as a covariate, it has to be measured on a continuous scale. Since gender is a dichotomous variable it cannot be used as such. Another way to control for it is using a multiple hierarchical regression which will be explained in the next chapter.

8_{Where ‘r’ indicates the Pearson’s correlation coefficient and ‘ρ’ indicates the Spearman’s rank-order coefficient} 9

Note that the variable ‘EthicalPos’ is a construct in which the respondents ranked how ethical and honest they think they are.

(28)

Above the diagonal: Pearson’s Correlation coefficient Under the diagonal: Spearman’s rank-order coefficient. **.Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed). 4.3.2 Hierchical multiple regression

In order to make a prediction of a continuous outcome variable using multiple independent variables while controlling for cofounding variables, a hierarchical multiple regression can be conducted. To use a hierarchical multiple regression a few assumptions have to be met. First, there has to be no multicolinearity. This exists when the independent variables are highly correlated with each other. This is indicated by the variance inflation factor (VIF). As can seen in table 8, the VIF for both variables is slightly higher than 1 (the applicable treshold) which assumes that there is some multicolinearity. Also, the same assumptions with the ANOVA of normality and homogenity of variance are applicable. I already concluded from table 5 that the variance of each experiment case is the same, but the population is not

normally distributed. This can also be concluded from testing for kurtosis and skewness of the independent variables in table 8. Using a treshold between 1 and -1, the population is not normaly distributed. But because these tresholds are only slightly violated, I will conduct a hierarchical multiple regression analysis. This provides the opportunity to control for gender, education, experience and age. The equation for this regression can be therefore depicted as follows10:

Ŷ = 110.82 -/- 4.21 (MOR) -/- 4.38 (FBA) -/- 26.12 (GENDER) -/- 1.39 (EDUCATION) + 0.53 (EXPERIENCE) -/- 1.68 (AGE).

10_{Where Ŷ is the predicted of the expected slack. An overview of the numbers mentioned in this equation are} given in appendix 1.

Table 7: Correlation coefficients

Variable Slack Gender Age Education Work experience Ethical Pos Slack 1 -0.351** (0.003) -0.157 (0.191) -0.162 (0.178) -0.037 (0.762) -0.330** 0.005 Gender -0.360** (0.002) 1 -0.093 (0.440) 0.123 (0.306) -0.081 (0.502) 0.152 (0.206) Age -0.177 (0.140) -0.106 0.381 1 0.353** (0.003) 0.658** (0.000) -0.005 (0.678) Education -0.132 0.273 0.130 (0.281) 0.407** 0.000 1 0.208 (0.081) 0.178 (0.139) Work experience -0.015 (0.899) -0.021 (0.859) 0.286* (0.016) 0.145 (0.228) 1 0.039 (0.749) Ethical Pos -0.478** 0.000 0.159 (0.185) 0.050 (0.678) 0.117 (0.329) 0.059 (0.624) 1

(29)

Table 8 Normality and multicolinearity Kurtosis Skewness VIF Factual assertion 0.815 1.277 1,080 No Factual assertion -0.727 0.904 Superior Authority 0.063 1.108 1,114 Subordinate Authority -0.281 -0.281

I included the control variables in model 1 and the predictor variables (mode of reporting and final budget authority) in model 2 as shown in the summary in table 9. From this table can be concluded that the control variables alone explain only 0.7% more of the variance in slack than the original model, as the R square increases to 0.176. The additional variables therefore don’t significantly contribute in predicting the outcome of slack (p=0.756). However, the ANOVA-test (included in appendix 2) shows that the model as a whole is a significant predictor of slack (p= 0.047).

Table 9 Model summary Change Statistics

Model R R Square Adjusted R Square Std. Error of the Estimate R Square Change F Change Df1 Df2 Sig F Change 1 0.411 0.169 0.188 31.473 0.169 3.345 4 66 0.015 2 0.419 0.176 0.099 31.821 0.007 0.282 2 64 0.756

To find out if the mode of reporting and budget authority are significant predictors of slack after controlling for gender, education, workexperience and age, the t-statistic is calculated using the hierarchical multiple regression. The effect of final budget authority (t= -/-0.558,

p=0.579) and the mode of reporting (t=-/-0.528, p=0.599) remain insignificant and each of

the variables only contribute for 6.6% and respectively 6.3% to the overal model in predicting slack. These statistics are found in table 10.

4.3.3 Motivation

In the post-experimental questionnaire I asked some statements about the motivations underlying the participant’s behavior. The answer on the statements ranged from 1 (strongly disagree) to 7 (strongly agree). Table 11 presents the six statements that were asked and the mean and standard deviations for all four cases. To see whether these statements have a

Table 10 Coefficients B Std. Error Beta t Sig.

Final budget authority -4.381 7.851 -.066 -.558 .579

(30)

significant influence on the budgetary slack that is created, I ran an ANOVA for each of the statements with the statements and the final budget authority/mode of reporting as

independent variables and budgetary slack as dependent variable. The significance of the contribution of each statement is measured with the interaction coefficient of the statement and independent variable (final budget authority or mode of reporting). First, focus on budget authority. The mean response to the statement ‘I wanted to treat the owner fairly’ is 5,6 with subordinate authority and 5,17 with superior authority. The concerns about fairness was significantly higher in the superior authority treatment than in the subordinate authority treatment (t= 2.997, p= 0.02). Also the statement ‘’I was concerned that the owner desired to

be treated fairly’’ is significantly higher rated in the superior authority case with a mean of

4,69 and 4,14 in the subordinate treatment (t= 2.428 p=0.04). This suggest that respondents were concerned that their proposal was not fair enough and decided to build less slack while the superior had final authority. Secondly, I examine the mode of reporting. But for none of the statements I can find any significant relations.

Also I asked the respondents whether they think that reporting cost/profit that results in a higher pay-off for managers than for the owners is unethical. Like the

aforementioned questions they could answer on a Likert scale from 1 (strongly disagree) to 7 (strongly agree). Although the results are not significant, the budget reporting that requires factual assertions was ranked higher (mean = 5,47) than budget reporting that did not require factual assertions (mean = 5,29). This suggests that budget communication that requires making factual assertions raised more ethical concerns than when no factual assertions were required. This might indicate that an ethical frame is adopted when the budget communication is more explicit. As we saw in table 7, EthicalPos seem to have a significant impact on slack. But when this variable is treated as an independent variable in a one-way ANOVA, the relationship is insignificant (F= 1.360 and p=0,221).

(31)

Table 11: post-experimental questionnaire : means and (Std. Dev)

Statements/ Case Subordinate Authority/No Factual assertion Subordinate Authority/ Factual assertion Superior Authority/No Factual assertion Superior Authority/ Factual assertion I wanted to be honest.* 5.59 (1.417) 5.83 (1.654) 5.83 (1.654) 4.78 (2.074) I wanted to treat the owner fairly.* 5.59

(1.543) 5.61 (1.819) 5.61 (1.819) 4.72 (2.191) I was concerned that the owner desired to be treated fairly.* 4.71

(1.611) 3.61 (1.819) 4.83 (1.249) 4.56 (1.688) I was concerned that the owner would reject my offer.* 2.82

(1.629) 2.61 (1.378) 3.61 (1.577) 4.06 (1.924)

I wanted to maximize my earnings.* 4.29

(1.724) 3.33 (1.970) 3.78 (2.184) 4.22 (2.415)

I wanted to maximize profits.* 5.12

(1.616) 4.78 (1.734) 5.06 (0.998) 5.06 (1.662) *Statements are ranked on a Likert scale from 1 to 7.

Besides using the framing-theory, I also tried to explain why participants in the superior budget treatment created more slack than in the subordinate treatment by using the theory of crowding-out. As explained in chapter 2.4 motivational crowding out occurs when the

intrinsic motivation of performing a task is crowded out by the extrinsic motivation of, in this case, extrinsic control. An indication that this effect occurs is when individuals give less importance to a given task, as they are more concerned about the extrinsic motivation.

Therefore I ran an ANOVA with ‘Importance’ and ‘FinalBudgetAuthority’ as the independent variable and ‘Slack’ as the dependent variable. The interaction coefficient with an F-statistic of 0,394 (p=.758) appears not significant. However, comparing the mean score for

‘Importance’ shows that respondents gave more importance to their task of reporting

cost/profit when they hold final budget authority (with a mean of 3.8) than when the superior holds final budget authority (3.56). This indicates some evidence that extrinsic control is able to influence the individual’s affection for their task incentivizing them to misreport but probably due to the small sample sizes this effect is insignificant.

(32)

5. Conclusion

This paper investigates whether final budget authority and the mode of reporting have an effect on the tendency to create slack by managers. An experiment was conducted in which participants in the role of manager had to report a budget regarding the implementation of a project to the owner of the firm. Previous literature on budgeting suggest that the allocation of final budget authority to the superior crowds out the intrinsic motivation of behaving

honestly, but is quite inconclusive about the amount of slack created when either the

subordinate or superior holds final authority. Therefore is expected that superior authority is associated with budgetary slack. Also, predictions were made concerning budget reporting that required factual assertions and reporting that does not. As a budget-context in which managers base their report on factual assertions is more often seen as an ethical dilemma, budgetary slack will be less apparent.

Although results of this paper are not significant, it provides some evidence of the aforementioned projections. The amount of slack created when superiors have final budget authority is less than when subordinates are allowed to decide themselves if the project will start. From the supplemental analysis, significant findings were found that explain this behavior. Subordinates build less slack because they were concerned about the fairness of their proposal and they feared rejection of this proposal. When budgeting required factual assertions, less slack was built. From the post-experimental questionnaire can be concluded that situations where factual assertions have to be made is more often seen as an ethical dilemma.

This study contributes to previous literature regarding budgetary slack in budgeting. This literature is quite inconclusive about the relationship between final budget authority and budgetary slack. By re-examining this relationship this study tries to give some conclusive evidence. As it is often argued that participative budgeting can increase costly side-effect such as gamesmanship, this study tries to find evidence that the allocation of budget authority and the mode of budget communication can influence these effects. At last it also contributes to societal literature as it tries to deepen our understanding of the underlying motivations regarding slack creating behavior.

This study is subject to the following limitations. Firstly, because more than 60% of all respondents that participated in the experiment did not answer all the manipulation checks right, only a total of 71 respondents were taken into analysis. This makes each sample size quite small influencing the significance of the results. Secondly, although the

(33)

experimental cases are mainly based upon existing experimental settings, the case itself is partly self-developed. This could have some important implications for the quality and understanding of the experimental case.

These limitations provides opportunities for future research. As the case is partly self-developed it also created opportunities to measure and test new settings and variables. One of which ‘importance’ measures how participants rate the importance of their task to report a budget. Although it was not significant, it provided evidence that respondents in the superior authority treatment gave less importance than the participants in the subordinate authority treatment. This could indicate that the intrinsic motivation is crowded-out by the extrinsic motivation of budget authority. Taking this measure into new studies with more respondents could provide more evidence regarding this theory.

How budget authority and the mode of reporting are able to alter decision frames