The elimination of the ratchet effect

(1)

Amsterdam Business School

The elimination of the

ratchet effect

Name: Annemieke Kee Student number: 10336516

Thesis supervisor: dr.ir S. Morssinkhof. Date: 15 June 2016

Word count: 16820

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

1 Statement of Originality

This document is written by student Annemieke Kee who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

Abstract

This paper is about the mitigation of the ratchet effect. The ratchet effect is considered a problem that should be solved as it decreases the effort employees exert after reaching their target, which could lead to a decrease in profits. A possible solution might be to increase the strength of implicit promotion-based incentives for employees. Consequently, the research question asked in this paper is: “Does the strength of the implicit promotion-based incentives provided mitigate the effect that target ratcheting has on the effort exerted by employees?” The research question has been examined through the use of an experiment. This paper finds a significant difference of around 5 in the additional hours participants would be willing to work when comparing the weak and strong implicit promotion-based incentives groups. This result indicates that employees will exert more effort if the strength of those incentives is higher and that those incentives have a negative moderating influence on the effort reduction due to target ratcheting. Therefore it would be beneficial to companies that use target ratcheting to think about implementing strong implicit promotion-based incentives. The paper makes two contributions. First, it contributes to the academic literature on target ratcheting by answering to certain calls of researchers. Secondly, it has practical relevance for the management of companies as it provides a solution to the ratchet effect.

(4)

3

Acknowledgements

I would like to thank several people for their help that supported the completion of this thesis. First of all, I thank dr. ir. S. Morssinkhof whose feedback and advice was very helpful. Furthermore, I thank all students from the second and third year BSc Economie en bedrijfskunde study programme, the students from the second year BSc Fiscale Economie study programme, the premaster Fiscale Economie and finally the students from the Postmaster Accountancy all at the University of Amsterdam that have participated in my experiment. Finally, I want to thank drs. E.M. van der Veer and drs. B.J.M. van Dam for giving me the opportunity to conduct my experiment.

(5)

4

1 Introduction

Employees are an important part of any organization as they do the work that generates the revenues and consequently the profit. Therefore a lot of research has been centered on motivating employees to do their job, some authors that wrote articles on motivating employees are Eisenhardt (1989), Gibbs (1995) and Prendergast (1999). According to Eisenhardt (1989), the theory about the relationship between the employer and his employees is called the agency theory. She claims that agency theory is similar to the risk sharing problem, but with the inclusion of two considerations, namely the division of labor and having different objectives. Jensen and Meckling (1976) state that employees do not have the same interests as their employers. The owners of the company get the profit from its operations, but the employees do not, thus the owners have to provide incentives for the employees to motivate them to do their job. If employees are rewarded based on the company’s performance, their interests will be aligned with those of the owners of the company. One of the most commonly used tools to motivate employees is by providing a bonus when they reach a certain target (Prendergast, 1999).

According to Bouwens and Kroos (2011), the agency theory explains the findings of the target ratcheting literature. Target ratcheting is defined by the literature as the tendency of some companies to take past and current performance into account when setting the target for the next period (Indjejikian, Matějka, & Schloetzer, 2014). The literature states that target ratcheting has a negative influence on the effort exerted by employees after they have reached their target, which is called the ratchet effect (Bouwens & Kroos, 2011; Leone & Rock, 2002; Weitzman, 1980). Bouwens and Kroos (2011) claim that this happens because employees do not get any incentives to perform the additional work. They are actually punished if they work harder as they will have to work harder next year to get their bonus.

According to Bouwens and Kroos (2011), the negative influence created by target ratcheting decreases the effectiveness of target ratcheting, as it ensures that employees reduce their effort instead of increase their effort. Furthermore, it has been proven by them that in some cases there is a reduction in sales, due to the reduction in effort, which leads to a reduction in revenues and therefore production. These two reasons show that it would be beneficial to companies if the ratchet effect could be mitigated. A possible way of doing this is to provide the employee with implicit promotion-based incentives that are based on the same performance measure used to award the bonus. Implicit promotion-based incentives are defined as promotions that are offered to incentivize employees to work harder (Ederhof, 2011). According to Prendergast (1999), it is expected that employees increase their effort when they want to get promoted. If employees are

(7)

6 evaluated based on the same performance measure as on which target ratcheting is applied, then it is expected that employees will work harder either way in order to get the promotion. Prendergast (1999) states that in some cases there is a higher probability of promotion when more effort is provided. So, it is expected that employees will increase their effort in order to get promoted if this will give them a higher probability of promotion. However, this will only happen if employees would appreciate the higher probability more than they would dislike the increase in their target for next year if they do not get promoted (Bouwens & Kroos, 2011).

As was mentioned before, target ratcheting reduces the exerted effort and this could laed to reduced revenues and profits. Therefore it is interesting to examine if the negative influence could be mitigated so that the management technique can be used without loss of production. That’s why this paper examines the moderating effect of the strength of implicit incentives on the relationship between the use of target ratcheting in a company and the reduction of effort of the employees of this company. Hence, the research question examined in this paper is:

Does the strength of the implicit promotion-based incentives provided

mitigate the effect that target ratcheting has on the effort exerted by

employees?

The relationship and the effect of the moderating variable studied can be depicted as demonstrated in Figure 1.1 below. The moderating effect is studied in this paper through the use of a one way between subjects experiment. The participants in this experiment are 106 randomly chosen students from the University of Amsterdam.

?

(-)

Target ratcheting Effort

Strength of implicit promotion-based

incentives

(8)

7 Through analysis of the results of the experiment, this paper finds a significant difference of 5.027 between the mean additional hours worked by participants in the 0% and 33% strength of implicit promotion-based incentives groups. This result suggests that participants in the weak implicit promotion-based incentives group tend to work 5 hours less than the other participants. This provides evidence for the hypothesis that employees will exert more effort if the strength of implicit promotion-based incentives is higher. Furthermore, this implies that the strength of implicit promotion-based incentives has a negative moderating influence on the effort reduction due to target ratcheting. The implications of this result is that companies now have a solution to the negative effects caused by target ratcheting.

This thesis makes two contributions. Firstly, it contributes to the academic literature on target ratcheting and answers to calls from Indjejikian et al. (2014) for more research on the way target ratcheting affects the behavior of managers and about possible situations in which the ratchet effect is less pronounced. It does so by finding a situation in which the ratchet effect is mitigated and adds to the understanding of managerial behavior by looking at the behavior of the participants in this situation. To my knowledge there is no previous study in existence that has looked at the effect of the strength of implicit promotion-based incentives on the ratchet effect. The probable reason why this has never been studied before is that it is difficult to find a research site that is appropriate, which means that it is not easy to come by the required data. Indjejikian et al. (2014) recognize this as well.

Secondly, this thesis has practical relevance for the management of companies as it provides a solution for the ratchet effect, which should be solved as the reduction in effort leads to reduced production and therefore reduced revenues and profits. One other practical solution has been found by Indjejikian et al. (2014). They state that an organization should only ratchet poorly performing managers. However, their solution has one issue that is not shared by the solution provided by this thesis, namely that the company is unable to profit from the positive aspects for well performing managers and still has to deal with the negative effects for poorly performing managers.

The rest of this thesis is structured as follows, in the second paragraph the relevant literature is discussed and the hypothesis studied is developed. In the third paragraph, the research method used will be explained and the different versions of the case will be described. In the fourth paragraph, the results of the research will be displayed and finally in the last paragraph some conclusions will be drawn, some limitations of this study will be discussed and some suggestions for future research are provided.

(9)

8

2 Literature review

In this section, the theory on which the hypothesis is build will be discussed. First, a summary of the agency theory in general will be provided. Second, a short paragraph discusses the relevant parts of the literature on performance measures. Third, there will be a discussion of target ratcheting, the ratchet effect and the connection to agency theory in particular. Thereafter the implicit, promotion-based incentives concept will be considered. Finally, in the last sub-paragraph links will be made between the theories discussed and the hypothesis will be developed using those theories.

2.1 Agency theory

As mentioned before, the agency theory evolved from an early research stream assessing the risk-sharing problem and expanded it by adding the agency problem (Eisenhardt, 1989). From that point on researchers went in two different directions. According to her, they can either study Positivist agency theory or contribute knowledge to the Principal-Agent stream. She claims that the first theory is mostly concerned with finding a solution to the governance problems regarding principal-agent relationships between owners and managers of public companies.

Jensen and Meckling (1976) add to the stream of literature on Positivist agency theory for instance. Their definition of principal-agent relationships, which is used in this thesis, describes such a relationship as a relationship between two parties, the principal(s) and the agents in which the principal(s) convince the agent to do some work for them. According to the authors, the aforementioned work has to include a decision making component to be considered a principal-agent relationship. According to Eisenhardt (1989), the Principal-Agent stream tries to find a general agency theory that can be applied to all principal-agent relationships, for instance employer-employee relationships. As this relationship is usually examined in the literature researching target ratcheting (Bouwens & Kroos, 2011) and therefore is also researched in this thesis, the principal-agent research stream will be described more thoroughly below.

Eisenhardt (1989) states that in both lines of research the agency theory’s main aim is to solve problems, namely agency problems and risk-sharing problems, in order to get the optimal contract between the principal and the agent. The agency problems will be discussed further below, but the risk-sharing problems will not as they are outside the scope of this thesis. Principal-agent relationships are considered to be either formal or informal contracts by both Eisenhardt (1989) and Jensen and Meckling (1976) and therefore the contract is the unit of analysis of all agency theory research. The two problems will be discussed from the perspective of the Principal-Agent

(10)

9 stream as the relationship studied in that literature represents the one examined in the literature about target ratcheting (Bouwens & Kroos, 2011).

The two agency problems Eisenhardt (1989) considers, both stem from a misalignment in the interests of the principal and the agent and from the inability of the principal to determine what the agent does and what his or her abilities are. The aforementioned misalignment of interests will reveal itself when the agent has to make a decision between working for the principal and engaging in activities that are beneficial to him or her personally and the agent is self-serving. This last condition is a conventional assumption of the agency theory and is incorporated into this thesis as well. The inability of the principal to determine the ability and effort of the agent, causes the agent to have private information at his disposal, thus it implies that there is information asymmetry between the principal and the agent (Eisenhardt, 1989). According to her, information asymmetry exists in a situation in which one of the parties has more information than the other party and can use this to their advantage. In this case, the agent has more information than the principal, which she states can lead to less optimal outcomes for the principal. According to Merchant and Van der Stede (2012), the outcome that is the most important to companies is the firm value due to the fact that maximizing this outcome is the main goal of a company.

Eisenhardt (1989) claims that two types of agency problems can arise when these two conditions are met, namely moral hazard and adverse selection. She defines moral hazard as the lack of effort exerted by the agent. This lack of effort, she claims, stems from the divergence in interests and from the fact that the agent can act inappropriately, because it cannot be observed how he or she behaves. For example, an agent is working at his or her own house and that day a soccer match is on television. He or she now has to decide to either watch the game or work for the principal. Under the assumptions mentioned before, the agency theory predicts that the agent will watch the soccer match. In this example, the agent has private information as he knows what he does and the principal does not. It is also demonstrated how information asymmetry can lead to worse outcomes for the principal as the agent does not exert effort.

Eisenhardt (1989) also states that the agent is able to mislead the principal about his or her ability, this is called adverse selection. Assume, for instance that the principal has got some work for which the performer would need some specialist knowledge about a specific animal. The agent really needs some money and thinks that he or she will be able to do the work without the specialist knowledge. Therefore the agent states that he or she knows a lot about this animal and will be happy to do the work. The principal cannot know how much the agent actually knows about the animal however as there is no evidence. In this example, it is once again the agent that has private

(11)

10 information and therefore the principal is worse of. The private information in this situation is that the agent knows how much knowledge he has about the animal and the principal does not. The information asymmetry can lead to detrimental results for the principal as he or she might lose a client when accepting the agent’s offer, due to the bad work done by the agent.

In order to solve these agency problems so that the principal’s outcomes are improved, Jensen and Meckling (1976) suggest that the principal provides incentives to align the agent’s interests with those of the principal. Merchant and Van der Stede (2012) go even further to state that in an ideal situation an employee would receive incentives in such a manner that these would compensate the employee for their share in the firm value.An example of an incentive is given by Eisenhardt (1989), who states that the principal could offer the agent contingent compensation that will obviously only be awarded if certain performance measures are satisfactory, in other words the principal would give the agent a bonus when they reach a certain target. According to Eisenhardt (1989), the problem of unobservable behavior would then be solved as the agent has to exert effort in order to ensure that performance measures are satisfactory. Therefore, the probable bonus will lead to the employee in the example mentioned earlier choosing to perform work activities instead of watching the soccer match. The agent is now properly motivated by the incentive and it does not matter so much anymore that you cannot observe his actions. According to Bouwens and Kroos (2011), the aforementioned target can be set in different ways. In sub-paragraph 3 one of those ways, specifically target ratcheting, will be discussed, but first the topic performance measures is discussed.

2.2 Performance measures

As was mentioned before in order to incentivize an employee management might offer them rewards for reaching predetermined levels of certain performance measures (Eisenhardt, 1989). This incentive structure however leads to a tough decision to be made for companies, namely what performance measure would best reflect the individual influence employees have on firm value. According to Merchant and Van der Stede (2012), this causes issues for firms as there are almost no performance measures that directly measure the individual contribution of an employee. They state that organizations use proxies of the direct influence of employees, when determining which performance measures will be used for exactly that reason.

Wallace (1997) shows that if the right performance measure is used to compensate employees, this could lead to increased firm value. He examines a research setting in which the experimental group starts using residual income as a performance measure that employees are rewarded on and a control group that does not. He finds that in the experimental group managers

(12)

11 held on to the assets longer and made less new investments in assets. Moreover, he found that those companies used more share repurchases to pay dividends to shareholders and finally that the assets were used to a greater extent than the assets of the control group companies. These findings suggest that the proxies used to measure performance should be selected carefully.

Merchant and Van der Stede (2012) discuss the determinants of the effectiveness of the results control, which is their name for the system that rewards employees on their individual contribution to the firm value. They claim that the proxies for the contribution of a single employee to firm value that will be used as a performance measures, need to satisfy certain criteria. The two main criteria they identify is that the performance measure should be congruent and controllable. According to them, congruency is high if a measure induces an employee to work in such a way that the wanted results are obtained. They state that this criterion is so critical that a performance measure which has no measurement error, but that at the same time is not congruent should not even be considered for implementation.

Controllability is determined by Merchant and Van der Stede (2012) as the ability of an employee to influence the performance measure materially. According to them, this ability is diminished by the uncontrollable factors, such as market factors, that influence the performance measure as well. These factors are considered uncontrollable by the authors as the employee cannot influence them. According to Feltham and Xie (1994), noise is caused by these uncontrollable factors. Noise is defined by Sloof and Van Praag (2010) as the error that is included in the performance measure used when measuring actual performance. The reason controllability is one of the key criteria is that according to Merchant and Van der Stede (2012) and Ittner, Larcker and Rajan (1997) the informational value of the performance measure is determined by the amount of noise due to uncontrollable events. Merchant and Van der Stede (2012) state that if the informational value of a performance measure is low, it is hard to determine if an employee has behaved as desired because the performance measure provides little information on the decisions made by the employee and his or her actions.

Feltham and Xie (1994) also recognize these two criteria as key, as according to them the optimal compensation structure can only be obtained if there is perfect congruence and little noise. According to them, the reason therof is that incongruence leads to employees inadequately allocating their effort over their different responsibilities and that noise leads to employees exerting less effort than in the ideal situation that performance measures are a direct measure of the employee’s share in the entire increase in firm value.

(13)

12 Other criteria that Merchant and Van der Stede (2012) have distinguished are precision, objectivity, timeliness, understandability and cost efficiency. They define precision as how often the similar results are produced by the measurement in the case that the circumstances are approximately the same. If precision is high, they state, that the performance measure can be seen as a reliable measure. They claim that objectivity means that performance measures should not be affected by the evaluator’s personal feelings or interpretations as subjective measures lead to measurement error which leads to the measurement not being accurate. According to Merchant and Van der Stede (2012), objectivity and precision are related to each other as both are needed to produce a good measure.

Another criterion that Merchant and Van der Stede (2012) identify is the timeliness of a performance measure. According to them, timeliness relates to the period between the employee working for the company and receiving the results from the measurement of his performance. The more timely a performance measure, the quicker the results are received and the quicker rewards can be awarded and the quicker management can intervene when performance is below standards. They state that in order to fulfill the understandability criterion a performance measure must satisfy two conditions concerning the employee: it should be clear to employees how they can influence the measure and what the measure holds them responsible for. They state that this is important as employees will only be able to have an impact on the measure if they understand what needs to be done in order to affect it.Finally, the cost efficiency criterion, identified by Merchant and Van der Stede (2012), states that benefits of using a measure should exceed the costs of using that measure otherwise it is not worth using the measure even if it fulfills all other criteria. Obviously, when determining what performance measure to use in target setting, managers will have to consider many aspects. The integrity of the performance measure used in the experiment will be discussed later in the research method.

2.3 Target ratcheting

Weitzman (1980) introduces the structure of a “standard reward system”, which is an example of the solution offered by Eisenhardt (1989) to the agency problems discussed earlier. Weitzman (1980) claims that if in this reward system Y represents a performance measure for the company and Q serves as a company’s target, then the bonus of an employee is usually a function of the difference between Y and Q. According to him, however this reward system that solves the moral hazard problem causes two incentive problems, a static and a dynamic incentive problem.

Weitzman (1980) describes the static incentive problem as employees trying to change the planned target in their favor in the target setting stage. The employees can do this by lying about

(14)

13 their expectation of the height of Y. The dynamic incentive problem, that he introduces, is the aforementioned ratchet effect and will be discussed more thoroughly. The ratchet effect is described by the author as the propensity of target setters to take the current and past performance into account when setting the target for the next period. He states that the target setters use the performance as a point of departure for setting the next period’s target. In other words, they increase the performance with a certain amount or predetermined ratio.

When target ratcheting is adopted by target setters, Weitzman (1980) claims that agents that are subject to ratcheted targets have to consider trading off present rewards against future losses. According to him, these present rewards come from better performance at this moment and the future losses stem from getting higher targets in the next period. In his paper the concept of an agent is broadly defined and it can mean an employee, a department or a big sector, but in this thesis the term agent is used to mean an individual worker and therefore the word agent can be read as employee throughout this thesis.

Leone and Rock (2002) argue that ratcheting happens for two reasons. The first reason is senior managers’ intention to ensure profitable forecasting. The senior managers, and target setters alike, want to increase the performance of the company as their compensation is depended on that performance. According to the aforementioned authors, conservative financial reporting has led to senior managers introducing ratcheting in the target setting and budgeting process, because it encourages them to expect that increases in earnings, thus increases in the performance of the company as a whole, are more persistent than decreases in earnings.

The second reason, Leone and Rock (2002) offer, is that ratcheting is an incentive mechanism that rewards permanent increases in performance as proposed to increases that are temporary. For instance, an employee dealing with target ratcheting can choose to receive education that will improve his or her performance permanently which will then lead to a permanent increase in performance. Furthermore, an employee will also be able to increase his or her sales by giving huge discounts to convince customers to buy the products he or she sells, but this will only lead to a temporary increase in his or her performance as the employee cannot continue to give discounts.

Leone and Rock (2002) state that target ratcheting discourages agents from using these temporary increases in performance. According to them, these increases may lead to higher bonus pay-outs at the current moment, but they will lead to higher targets as well, which means that future expected bonuses are diminished. Here a moral hazard problem as discussed by Eisenhardt (1989) is once again created and leads to a lack of effort on behalf of the agent. The agent will no

(15)

14 longer exert any effort after reaching his target, because he or she is not offered any incentive for the additional effort. The agency theory, discussed by Eisenhardt (1989) then states that the employee will not work, because the agent wants something different from what the principal wants. The principal wants the company to perform as best as possible, as he or she gets the rewards. The agent does not get these rewards and thus agency theory predicts that he or she does not care about the company’s performance.

In fact there is a “punishment” as the additional effort in this period will generate a temporary increase in performance and thus will lead to lower expected future bonuses (Leone & Rock, 2002). This “punishment” leads to a bigger misalignment of interests between those of the principal and those of the agent and thus to a moral hazard problem. This remark is in accordance with the aforementioned statement by Weitzman (1980) that agents need to decide if they want to trade off present rewards against future losses. Bouwens and Kroos (2011) find that this even leads to agents reducing their effort after having reached their targets for the period. The authors state that agents do this if they are convinced that doing so will increase their total current and future benefits.

Bouwens and Kroos (2011) claim further that the decision to manipulate the performance revolves around the agent’s expectation of the possibility to reach future targets. This is related to the concept brought forward by Leone and Rock (2002) of temporary and permanent increases in performance mentioned before. According to Bouwens and Kroos (2011), this expectation is influenced by the persistence of the increase in performance. They claim that if the agents feel that they can sustain the performance levels necessary to reach their targets, this will likely lead to less manipulation as the incentives to do so are lower. According to them, this is the case when the increases in performance are permanent such as when an employee receives education to improve his or her performance. However they state that if an employee does not feel able to maintain the needed performance levels, because the increase in performance would be temporary, he or she will have more incentive to reduce his or her performance to lower the expected future performance levels.

Additionally, Bouwens and Kroos (2011) argue that there are two ways that an agent can manipulate its performance: the agent is able to use accounting techniques, for example manipulate discretionary accruals, or change the activities they perform while working, for instance reduce their current effort. They also state that when an agent has limited options to reduce his performance, because he is unable to affect accounting decisions, he or she will have even more incentive to reduce his or her current performance. They claim that the stronger incentives to

(16)

15 avoid target ratcheting are brought forward by the fact that the agents can only influence their actual performance, which is a temporary increase in performance and which therefore needs to be repeated every period in order to be able to reach the targets heightened due to better performance. This result implies that if the agent wants to reach his or her target, he or she will have to increase his or her effort in the following periods as well.

2.4 Implicit promotion-based incentives

According to Prendergast (1999), promotion-based incentives are widely used as a way of motivating people to exert effort or in other words do their job. It is so common that he describes the typical work environment of an employee as one where the employee does his job in order to get a promotion, a salary revision or a bonus. When promotions are awarded he states that there are, in most cases, other employees who are equally qualified for the job and that also want to be promoted. According to him, this naturally leads to a competition between the candidates. The author states that this competition over promotions is the subject of the tournament theory. As can be inferred from the name, the conditions in the tournament theory closely resemble those of an actual sports tournament. The author states that there is a fixed set of prices for which the agents compete, specified in advance, in this case the reward obtained from receiving the promotion. This reward is defined by Ederhof (2011) as the combination of the increase in the employee’s salary arising from the promotion and the value of having the possibility of being promoted further in the future and then receiving further rewards. Of course, there are the multiple employees competing for that set of prices and the winner is the one that has done the best in the competition relative to the others (Prendergast, 1999). According to Eisenhardt (1989), this means that the employee who either has exerted the most effort or has the highest ability is awarded the promotion, because both influence the performance that is measured by a performance measure. Gibbs (1995) states that if promotion decisions are based on a performance measure that is affected by effort that the possibility of promotion has an incentive effect, thereby satisfying the incentive role. Furthermore, the author states that when someone is promoted because their performance has exceeded a target this also satisfies the sorting role of promotions. Grabner and Moers (2013) discuss the two aforementioned roles that promotion-based incentives can take. The incentive role involves rewarding employees for their effort and the sorting role is concerned with ensuring that all employees are arranged in such a way that every position in the organization is filled by the best qualified employee when looking at his or her ability.

According to Grabner and Moers (2013), the relative importance of these roles is a subject on which the literature is divided. The authors state that accounting literature mainly studies the

(17)

16 incentive role, thereby implying that this is the most important role, the economic literature looks at the tradeoffs between the two roles, considering both of equal importance and they go even further and state that in their paper the sorting role is the most important. The relative importance of these roles is therefore not clear cut and it is assumed that this is also the case in practice. However they state that both of those roles could have an influence on the predictions of a paper. In order to ensure that there is no effect of these roles on this thesis the experiment is designed in such a way that both roles are fulfilled even though the incentive role alone is tested. This will be explained further in the paragraph concerning the research method.

Gibbs (1995) states that in an ideal situation the principal would separate the two roles of promotion decisions to increase their efficiency. However he claims that this is made impossible by the noise in the performance measure. This noise, he states, is caused by the inability of the principal to distinguish between an employee’s exerted effort and his or her ability. The principal is unable to do this as the employee has private information about both his or her ability and his or her effort (Eisenhardt, 1989). However, Ederhof (2011) states that this inability is also one of the reasons why implicit incentives arise in the form of a solution to the problem of unobservable behavior. According to Ederhof (2011), the promotion-based incentives, discussed in this thesis, are implicit incentives, which means that they are not written down in a contract, but are enforced by the employee himself in the hope of getting future payoffs (Pearce & Stachetti, 1998). Another reason identified by Ederhof (2011) is that an employee has the intention to change the principal’s perception of his or her ability by increasing his or her effort. This happens, for instance, when the employee’s ability is low and he works hard to make up for it. According to him, this aggravates the problem of unobservable behavior, which ensures that the principal uses implicit incentives to solve the adverse selection problem mentioned earlier.

The strength of implicit promotion-based incentives is an important determinant of the employee’s willingness to exert effort as proven by Ederhof (2011). According to him, this is due to the fact that the implicit incentive causes the employee to exert effort in order to get the incentive, for instance a promotion. He states that the strength of promotion-based implicit incentives is caused by two factors, namely the increase in the probability of promotion caused by an additional unit of effort and the aforementioned reward that is received when the promotion is awarded. It has been proven by Gibbs (1995) that the reward increases the willingness of the agent to exert effort as additional effort will increase his or her chances of receiving the promotion. He also suggests that the strength of implicit promotion-based incentives is related to the probability of promotion as perceived by the agent. This can be inferred from the fact that according to him

(18)

17 the strength of the implicit promotion-based incentives decreases as the perceived probability of promotion decreases.

Ederhof (2011) claims that the strength of the implicit promotion-based incentive is higher if the future reward is of greater importance to the employee, increasing the willingness to exert effort even further. Furthermore, he asserts that the position of the employee also influences the strength of the incentives as the higher an employee’s position on the organizational ladder, the less opportunities that employee has, to get promoted and thus the probability of promotion decreases. Additionally, the lower amount of opportunities causes the sum of the future rewards that determine the value of the reward to decrease. Thus the author concludes that when an employee has a high ranking position, the implicit incentives will be weaker and therefore the explicit incentives should be stronger. This negative relationship between the amount and size of explicit incentives and the strength of implicit incentives, which Ederhof (2011) researches and proves in his paper, shows that the two kinds of incentives are complements, as the two forms are used to get the agent to exert an optimal level of effort. This leads him to conclude that when an employee has weak implicit incentives the principal should provide strong explicit incentives in order for him to do his job. The fact that the two incentive types are complements, suggests that when the explicit incentives do not motivate employees to exert additional effort after having reached the target and therefore ensuring the bonus, implicit incentives can be used to motivate them to provide that additional effort.

2.5 Hypothesis development

As discussed before, the agency theory states that in order to solve moral hazard and adverse selection problems the principal should offer the agent conditional incentives based on a performance measure (Eisenhardt, 1989). According to her, this way the agent will not shirk work, but will exert effort in order to get the contingent compensation. She states that an example of this contingent compensation is receiving a bonus when reaching a predetermined target. However a special form of target setting, namely target ratcheting, has been proven, by Bouwens and Kroos (2011), to induce employees to reduce their effort after having reached their target. Weitzman (1980) states that target ratcheting is applied by a company when the target setters use the current or past performance as a starting point for setting the following period’s target. In other words, they increase this and former periods’ performance by a predetermined ratio or amount.

Bouwens and Kroos (2011) also state that, when employees are deciding whether to reduce their effort, they keep in mind the total amount of benefits that they will obtain and will lose if they go through with the manipulation of their effort. This implies that in some cases implicit

(19)

18 promotion-based incentives might influence their decision to reduce their effort. As mentioned before, Gibbs (1995) proves that the reward, obtained when an employee is promoted, increases the willingness of the employee to exert effort as this will increase his or her chances of receiving the promotion and thus the reward.

The amount of influence the implicit promotion-based incentive has on the employee’s decision depends on the strength of this incentive. In the same way that implicit incentives encourage an employee to exert any effort, Gibbs (1995) asserts that the greater the strength of the implicit incentives the more effort an employee is willing to exert. He also states that the strength of the implicit promotion-based incentives decreases as the probability of promotion decreases, which suggests that the strength of implicit promotion-based incentives is partly caused by the probability of promotion as perceived by the agent.

As mentioned before, it becomes clear from Ederhof’s (2011) research on implicit promotion-based incentives that implicit and explicit incentive mechanisms are complements. According to him, this means that in the situation that an employee has weak implicit incentives to motivate him or her, the principal needs to ensure that the explicit incentives create the motivation necessary to attain the optimal level of effort from the agent. As mentioned before, when an employee is not motivated to exert additional effort after having reached their target and is induced to reduce their effort, the implicit incentives can solve this problem as they can induce the employees to provide the additional effort even after they have reached their target.

Therefore, it is the expected that employees will not only stop reducing their effort, but will effectively increase their effort if the probability of promotion is high enough, thus if the implicit promotion-based incentives are strong enough, even if the company applies ratcheting to their targets. However, this hypothesis will of course only hold if the same performance measure is used for the promotion of employees and awarding of bonuses and if the performance measure satisfies the criteria identified by Merchant and Van der Stede (2012). In its entirety, the discussed theory leads to the development of the following hypothesis that will be studied in this thesis.

Hypothesis 1: employees whose targets are subject to ratcheting will increase their effort more if the strength of the implicit promotion-based incentives provided is higher.

(20)

19

Figure 2.1 – Expectation of the results

Figure 2.1 shows the relationship between the strength of the implicit promotion-based incentives and the additional effort that agents will exert that is expected to exist in a company in the case that this company applies target ratcheting. As can be seen, the additional effort that agents are willing to exert increases with the strength of the implicit promotion-based incentives.

0 (1/3) A d d iti o n al e ff o rt

The strenght of implicit promotion-based incentives

Figure 2.1 - Expectation of the results

(21)

20

3 The research method

In this section of the paper, the research method applied in this paper will be described carefully. The hypothesis developed earlier is researched by means of an experiment. Therefore the experimental design will be outlined first. Thereafter, the materials used for the experiment will be discussed extensively. Finally the research procedures, manipulation of variables and research sample will be explained.

3.1 Experimental design

The means of testing the hypothesis in this paper is a one way between subjects experiment. This means that there are two versions of a case with manipulation of one independent variable. The independent variable in the experiment is the strength of the implicit promotion-based incentives offered to the participant. This variable is manipulated by adjusting the strength of implicit based incentives in each case. In the experiment the strength of the implicit promotion-based incentives is measured by giving the probability of getting the promotion, which is used as a proxy by Gibbs (1995) as well. In case A the strength of the implicit promotion-based incentives is 33% and in case B it is 0%, as can be seen in appendix one and two. This implies that the variable has two levels and that it is quantitative and discrete. A dummy variable is used to check whether a respondent had a probability of 0 of getting promoted (coded 0) or a probability of 0.33 of getting a promotion (coded 1).

The dependent variable is the amount of effort a respondent is willing to exert after having done the regular work, because effort is hard to measure the number of hours is used to represent effort. It is measured by the number of additional hours, between 0 and 20, which a respondent states to be willing to work in order to increase the probability of promotion. This measure is adapted from a tool used by Natriello and McDill (1986), who also use hours to measure the effort of students, though they measure the effort exerted on homework and not the effort exerted in a work environment. The research sample, which will be described more thoroughly below, is comprised of students. Therefore it makes sense to adapt this measure and use it to estimate the additional effort exerted by participants as the question was designed to measure student’s effort and it has already been tested by the aforementioned authors. This variable is quantitative and continuous as it can embody all values between 0 and 20. The maximum restriction is incorporated in the experiment due to the inability to measure and process all possible values of additional hours and the minimum restriction ensures that a respondent is able to either fill out no additional hours

(22)

21 or some additional hours as it is not within the scope of this thesis to examine if target ratcheting decreases the effort exerted by employees.

Finally, there are two important conditions that are equal for both versions and therefore should have no influence on the results. The most important condition is the fact that the company described in the case applies target ratcheting when setting targets. This has been explained thoroughly in the case as it is assumed that most students do not have experience with the special form of target setting. This explanation involves an example over two periods of what happened with an employee’s target for the next period given that the employee’s performance in the first period was good, average or bad.

The second constant is that the company fulfills the sorting role. The sorting role is fulfilled by stating that everyone has the same chances of getting the promotion. This shows that the manager’s tasks are quite similar to the responsibilities of a regular employee, which is what the participant is in the situation described by the case. The incentive role is also satisfied as the promotion provides an incentive by giving the prospect of a better job and an assumed salary increase although this is not a constant, as mentioned before, the strength of the implicit promotion-based incentive is the manipulated variable.

3.2 Cases

The two versions of the case are the same in all respects except for the aforementioned manipulation of the independent variable. As mentioned earlier, there is either a 0% chance or a 33% chance that you get promoted. The case describes a situation where the participant is an employee of a restaurant that works in the kitchen. The participants are told that they have a target and are able to reach that target by just working their regular hours. If they reach this target they get a bonus. In order to go over their target they would have to work late, so spent additional hours working in the kitchen. This leads to two versions of the case. The English version of these two cases have been included in appendices one and two and the English version of the questionnaire that participants had to fill out was included in appendix three.

1. Case version A, with strong implicit promotion-based incentives. In this case there is a 33% chance of getting promoted.

2. Case version B, with weak implicit promotion-based incentives. In this case there is a 0% chance of getting promoted.

In both versions of the case the company that is described is a restaurant and the participant is a kitchen employee in charge of making sandwiches. This setting was chosen, because

(23)

22 it is common for students to work in a restaurant and this would make the answers given by the students more reliable as they will be better able to imagine working in a restaurant. According to the Centraal Bureau voor de Statistiek (CBS) (2014), 16 percent of the students work in a restaurant, which makes it the highest ranking work environment for students. Therefore, the research setting is a good representation of the students’ regular work environments. The CBS (2014) is an organization that keeps track of developments in the Dutch society, for instance developments in the composition of the workforce. The case states that the participant is obliged to work 20 hours a month and is allowed to work additional hours up to a maximum of 20 hours. According to the CBS, students work approximately 14 hours a week, which means that the regular hours in the case can be considered a little high. The decision to increase the regular number of hours by 6 hours was made to ensure that students realized that they would have to work more hours and thus that it would reduce their leisure time.

Furthermore, it is explained that an employee makes about 320 sandwiches a month, which is also the target, and that employees get a bonus when the target set for them has been reached. Moreover, it is stated specifically that the performance can only be influenced by working additional hours to ensure that the respondents know that no other circumstances affect the performance measure, which means that the performance measure is highly controllable. This is a reasonable assumption as, according to Ittner, Larcker and Rajan (1997), companies try not to use performance measures that are influenced greatly by noise, which is defined by Sloof and Van Praag (2010) as the error, with which actual performance is measured by the performance measure. Ittner, Larcker and Rajan (1997) claim that the reluctance to use performance measures that are noisy is due to the lack of informational value of those performance measures, because they do not measure the actual performance of the employee. Sloof and Van Praag (2010) state that another reason that a principal does not want to use noisy performance measures is that the risk premium is higher. The higher risk premium must be provided to insure the agent against bearing the risk that he will not receive his target due to noise (Sloof & Van Praag, 2010).

The other major criterion determined by Merchant and Van der Stede (2012) is that a performance measure should be congruent. In other words, a performance measure should be able to induce the desired behavior. The performance measure used in the experiment measures the exerted effort of an individual participant by looking at the number of sandwiches that are made by this participant. This measure is quite congruent as if people are rewarded on the number of sandwiches they make, they will try to make as much sandwiches as they can. This is the desired behavior as it increase the restaurant’s firm value.

(24)

23 Additionally, the performance measure used in the experiment also satisfies the other criteria identified by Merchant and Van der Stede (2012). Firstly, it is precise as it will give the same results for different participants under constant conditions. For instance, if two participants make the same amount of sandwiches, these can be counted, which will lead to the same result for both of the participants. Secondly, it is not subjective as the case makes clear that whoever makes the most sandwiches will get promoted. This means that it is just a comparison of the numbers of all participants and the evaluation is not affected by the boss’s feelings or interpretations. Furthermore, it is timely as it can be determined how many sandwiches an employee has made and therefore how much effort is exerted quickly after the employee has performed activities. Likewise, the measure is understandable for the participants as it is a simple measure, namely the number of sandwiches that they have made. This is due to the fact that it is clear for participants that they are held accountable for the amount of sandwiches they have made and that in order to influence this number they will just have make more sandwiches. Finally, it is cost-efficient as well, as the measurement of the number of sandwiches a participant has made is not complicated and does not require a lot of effort or money.

Thereafter, target ratcheting is introduced into the case. This special form of target setting is explained to students by stating that their actual performance is used to determine the target for the following period. Moreover, a table was added to the case that depicts the consequences for the targets set for three different employees, employees A, B, and C. These employees represent what happens to your target when you have a high (employee A), average (employee B) or low (employee C) performance level. It also specifically shows what the difference between each employee’s target for this period and that of the next period is and demonstrates that this difference is bigger for employee A than for both other employees and that the difference is bigger for employee B than for employee C. This is due to the higher performance level, which is explained in the case by discussing the consequences for employee A. This table tries to ensure that all participants understand the consequences and know what target ratcheting is.

At the end of each case the independent variable is incorporated. That section outlines that a management position has opened up and that one of three employees will be selected to fulfill that new position. In case A the respondent is included in those three employees, which leads to a probability of 0.33 of getting promoted and in case B he or she is not part of the three employees that have a shot at getting promoted, which leads to a probability of 0 of getting promoted.

Moreover it is stated that the decision will be made based on the performance of those three employees and that a higher performance leads to a higher probability of getting promoted,

(25)

24 that the additional hours worked will be taken into account when determining the target for the following period and that target ratcheting is introduced at the beginning of the month. The last statement was included to ensure that the students knew that the target of all three employees was at the same level at the beginning of this period. This assumption was necessary to provide the students with a probability without interference from previous periods. The period examined in the case was one month to ensure that it would be simpler to make a decision for the respondents. Actually, if more than one month was examined this would lead to a lot of uncertainty about the probability of getting promoted as the targets for the three employees would start to differ. To find the full version of both cases, see appendix A and B.

3.3 Measures

The question with which the hypothesis stated earlier is tested is the first question asked in the questionnaire: ‘How many additional hours, between 0 and 20, would you be willing to work?’ This question tests if the respondents would be willing to work additional hours given the situation as described by their version of the case. As mentioned before it is expected that respondents that receive case A will want to work more additional hours than respondents that receive case B. The expectations have been graphically depicted in Figure 2.1 in paragraph 2.

Questions two through eight will be discussed together as these questions combined measure one variable, namely how much a respondent would like to work for the specific company in the case. These questions are taken from a questionnaire developed by Maas (2015). This measure is used as a control variable as Dermody, Young and Taylor (2004) state that for employees working in a restaurant the work environment is an important determinant of job motivation. As according to Eisenhardt (1989) motivation is positively related to performance, the additional hours worked by the students in order to get the promotion could also be explained by how much the respondents liked the company and their willingness to work for that company. If the situation described in the case was not appealing to the participants, this could negatively influence the additional hours participants would want to exert in order to get the promotion. This can be explained by the fact that if participants would not want the promotion, because they do not want to work for the company, they probably do not want to work additional hours as they are not motivated by the prospect of the promotion.

Two of the questions, namely question six and eight, are negatively worded and the other questions are positively worded. The two negatively worded questions were added in order to try and ensure that the respondents would not just fill out all sevens to be done earlier with the questionnaire. The negatively worded questions were recoded to ensure that all those questions

(26)

25 could form one construct that measured the willingness to work in the same direction. Furthermore, the negatively worded questions can also provide an indication if the respondents have read the questions. This can be seen by checking if the negatively worded questions received a low score if the others received a high score and the other way around.

Finally, questions 9 through 15 are added to get an idea of the demographics of the research sample. Question 9, 10 and 11 ask about age, sex, and nationality respectively. Question 12 measures if a particular respondent has ever had a paid job and if this was a full time or part-time job or if the participant has had both a full time and a part-time job, question 13 measures if the respondent has worked in a restaurant and question 14 measures if a respondent has had to deal with a numbers target as used in the case in a work environment before. These three questions are also added to see how well the overall research sample was able to imagine working in the research setting. Finally, question 15 determines which study programme the participant is following at this moment.

3.4 Research sample

The research sample is comprised of 106 students taking one of two courses at the University of Amsterdam, from which 48 participants filled out case A and 58 participants filled out case B. This difference was caused by the need for randomization and the number of students either exceeding or being lower than the estimated number of twenty students that were expected to come to the tutorial. As twenty cases were prepared, from which ten were version A and the other ten were version B, sometimes cases were left unused and at other times more cases were needed and taken from another prepared set. Considering the percentages that can be found in Table 3.1 below, it was thought that this difference didn’t influence the data analysis significantly. A total of 108 students participated in the experiment, however not all questionnaires were filled out fully and in two of these cases the importance of the missing data was so severe that it was best to withdraw them from the research sample. 59 male students and 47 female students participated in the experiment, most of which were Dutch (98 percent) and were enrolled in the BSc Bedrijfskunde en Economie (81 percent). The ages of the students ranged from 19 to 31. All demographic attributes of the participants in the research sample are exhibited in Table 3.1 below.

The three variables measuring experience give an indication of how well the participating students would have been able to imagine working in the case setting. Most students have a job or have had one in the past and most of these students have or have had part-time jobs (98 percent). This indicates that most of the participants are familiar with having a part-time job and therefore have experience with a work environment. Some participants indicate that they have worked in a

(27)

26 restaurant (35 percent) and slightly less students have worked with a number target at one time or another (25 percent).

The two courses mentioned earlier were picked as they are both part of the curriculum of the Dutch programme Economie en Bedrijfskunde and therefore it is likely that these students will be the future managers and other employees working for businesses across the globe (Universiteit van Amsterdam, 2015). The two courses are Intermediate Financial Accounting 2 (IFA2), part of the 2nd_{year curriculum and Advanced Management Accounting (AMA), part of} the 3rd_{year curriculum. This makes the students that enrolled in these courses good subjects for} this experiment as the experiment will provide information about the behavior of the future managers and other employees. Moreover, it is interesting to research their behavior as it might lead to companies trying to mitigate the ratchet effect by providing strong implicit promotion-based incentives on the same performance measure. Furthermore, the students have taken enough management related courses to assume that they have enough knowledge to understand the case and make a good decision as to what they would do were they in the described situation.

As can be seen in Table 3.1 below there were also students from other programmes following the courses, namely from the study programmes: BSc Fiscale Economie, BSc rechten, the Premaster Fiscale Economie and the Postmaster Accountancy. Because the first three study programmes are also taught in Dutch, it is assumed that the students following these study programmes are sufficiently proficient in the Dutch language to understand the case and questions well enough to make a good decision. The Postmaster Accountancy is taught in English, but as the student followed the course IFA2 in Dutch it is assumed the student is proficient in Dutch otherwise the student would not have been allowed to be enroll in the course. The answers of the students not enrolled in the BSc Economie en Bedrijfskunde are still interesting as they might still become the future managers and other employees of a company that uses target ratcheting.

(28)

27

Table 3.1 Demographic data

Participants Number of participants Percentage of participants

Strong implicit promotion-based incentives 48 45 Weak implicit promotion-based incentives 58 55

Total 106 100* Age 19 11 10 20 21 20 21 24 23 22 20 19 23 9 9 24 10 9 25 2 2 26 4 4 27 1 1 28 2 2 30 1 1 31 1 1 Gender Male 59 56 Female 47 44 Nationality Dutch 104 98 Other: 2 2 German 1 1 Belgian 1 1 Study programme BSc Economie en bedrijfskunde 86 81 BSc Fiscale Economie 12 11 BSc Economie en bedrijfskunde en BSc rechten 1 1

Premaster Fiscale Economie 6 6

Postmaster Accountancy 1 1 Job experience Full time 6 6 Part-time 79 75 Both 18 17 No job 2 2 No response 1 1 Restaurant experience Worked in a restaurant 37 35

Did not work in a restaurant 69 65

Target experience

Has had a numbers target 26 25

Has not had a numbers target 80 75

*_{In some cases, the percentages do not add up to 100% due to rounding.}

(29)

28 3.5 Research procedures

When a study programme is taught in Dutch, the University of Amsterdam requires non-Dutch students to prove that they are accomplished in the Dutch language by taking an exam before possible international students can be accepted to take part in that study programme (University of Amsterdam, 2015). This leads to the assumption that all students enrolled in these two courses are proficient in the Dutch language as they are part of the curriculum of the BSc Economie en Bedrijfskunde. This means that the students in the research sample would better understand the cases and questions and therefore would make better decisions about their behavior if the experiment was conducted in Dutch. This would increase the reliability of the experiment and therefore the short pitch at the beginning of each tutorial was done in Dutch and both versions of the case and all questions were written in Dutch as well.

In every tutorial the same routine was followed to ensure that the research procedure was the same in all tutorials. First, there was a short pitch of about one to two minutes in which the students were asked to participate in the case. The text that was said is written down below: “Hello everybody, I’m Annemieke and I’m writing my master thesis at the moment. In order to do so I need to do research and I was hoping you would want to help me complete it. You can do so by filling in a case. This will only take five minutes, there are no wrong answers and you will get a small surprise when you’re done. Furthermore, I’ll thank the second or third year students in my thesis. I’ll be back at the end of class or during the break to hand out the cases, which means that if you want to participate you’ll just have to wait here. Are there any questions?”

However, in the tutorials of the course Advanced Management Accounting there was a sentence added making it clear that if there were any students who had already filled in the questionnaire in one of the IFA2 tutorials that they should not participate again. Therefore it is expected that there are not any participants that filled in the case twice. This statement is not expected to have an effect on the results as it is not related to the content of the cases. It might have led to a slight decrease in the participation grade in these courses, but a sufficient amount of students participated that both the second year and third year students were represented well.

The second step in the routine was to hand out the cases to the students who were willing to participate at the end of class for the IFA2 tutorial as there was no break and during the break for the AMA tutorial. This was done through separate consultation with both professors. In all tutorials from both courses there was a high participation rate and the average time it took to fill out the cases was approximately the same, which leads to the assumption that this difference didn’t have a big influence on the experiment. As can be understood from the presentation text,

(30)

29 chocolates were handed out after participating in the experiment to incentivize the students to participate. It was considered to be necessary to offer one piece of chocolate to each student in order to get the required amount of students to participate. Furthermore, the students were promised to be thanked in the foreword of my thesis as has been done. This was also offered to incentivize the students to participate.

(31)

30

4 Results

In this paragraph, the results of the analysis performed to examine if the strength of implicit incentives has a moderating effect on the relationship between target ratcheting and effort exerted by employees will be portrayed. Furthermore the performed analysis will be explained. First, the preliminary analyses will be performed, which are the description and descriptive statistics of the variables used in the analysis, the testing of the assumptions made about the data and the testing of the construct reliability and validity of the control variable. Finally, the hypothesis introduced in the literature review will be tested.

4.1 Variables and descriptive statistics

In this sub paragraph the variables used will be described and the descriptive statistics will be shown. Thereafter, the assumptions made about the data in order to be able to perform the analysis will be tested and finally the construct reliability and validity of the control variable, company, will be tested.

4.1.1 Measurement and descriptive statistics

The measurement and the descriptive statistics of the variables analyzed will be discussed in this section. The additional hours worked and the strength of the implicit promotion-based incentives variables have already been discussed extensively in paragraph 3. Therefore we will focus on the other variables. First of all, age, gender and nationality measure the participant’s age, sex and nationality respectively. Second, the study programme variable measures which study programme the participant is currently enrolled in. The Job experience, Restaurant experience and Target experience variables each measure if the participant has any experience with having a job, with having worked in a restaurant and with having worked with number targets.

All of the preceding variables were obtained from the filled out questionnaires. However the last variable, the control variable company, was constructed in order to analyze its effect on the dependent variable. It is a construct of seven separate questions measuring the job motivation or willingness to work for the restaurant, which according to Eisenhardt (1989) might have an influence on the results of the experiment. This can be explained by the fact that participants do not want to work additional hours when they do not want the promotion, because they do not want to work for the company. In Table 4.1 below an overview of the measurement of the various variables is given.

The elimination of the ratchet effect