Amsterdam Business School
Master Thesis
Ex Post Goal Adjustments: Policies, Procedures and Fairness
Perceptions.
Name: N.A. (Sander) de Vries Student number: 10898565
Thesis supervisor: prof. dr. V.S. (Victor) Maas Date: 7 June 2016
Word count: 12858
MSc Accountancy & Control, specialization Accountancy and specialization Control Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Sander de Vries who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
A topic which has been getting more attention recently in research about incentives structures is the use of subjectivity. More specifically, the use of ex post goal adjustments as a tool to compensate employees for uncontrollable events that impaired the incentive payout. Recent studies showed that ex post goal adjustments only work when goals are moderate (Kelly, Webb, & Vance, 2014). I contribute to this stream of research by further examining the human behavior in response to ex post goal adjustments.
In order for a performance evaluation system to be viewed as fair, there needs to be a formal policy (Hartmann & Slapničar, 2009). In the literature about fairness and justice, there is a difference made between procedural and distributive fairness perceptions. Using the fair process theory by van den Bos (2001) I argue that a formal policy will lead not only to higher procedural fairness policies, but also to higher distributive fairness perceptions.
The results of my experiment revealed that using a formal policy rather than a non-formal policy increases the overall fairness perceptions of employees. Furthermore, the fair process effect is found by seeing an increase in procedural fairness perceptions as well as distributive fairness perceptions.
A supplemental analyses indicated that the procedural fairness perceptions are mediating the relationship effect of a formal policy on distributive fairness perceptions. The effect is not direct, the employee will first judge the procedure of a formal policy as fairer, and subsequently the distribution.
Contents
1 Introduction ... 5
2 Theory ... 6
2.1 Fairness Perception and Performance Measurement ... 6
2.2 Procedural fairness perceptions ... 7
2.3 Distributive fairness perceptions... 8
2.4 Performance Evaluation Policies and Procedures ... 9
2.5 Ex Post Goal adjustments ... 11
3 Hypothesis development ... 14 4 Research methodology ... 17 4.1 Experimental design ... 17 4.2 Questionnaire items ... 18 5 Results ... 19 5.1 Demographics ... 19 5.2 Independent variables ... 22 5.3 Dependent variables ... 22 5.3.1 Factor analysis ... 22 5.3.2 Reliability analysis ... 23 5.3.3 Normality assumption ... 24 5.3.4 Descriptive statistics ... 27 5.3.5 Correlation table ... 29 5.3.6 Hypothesis tests ... 30 5.3.7 Supplemental analysis ... 34
6 Discussion and Conclusion ... 35
References ... 37
Appendix 1: case and questions in English ... 42
Appendix 2: case and questions in Dutch ... 45
1 Introduction
In the first section of this thesis I will introduce the research question that I want to answer. First, I will give a short overview of the main concepts and theories that are led to my question. After reading this section, it should be clear how I want to contribute to existing literature and why my research question is interesting. In the next section I will elaborate the theories used in this project in greater detail.
A relatively new topic in the field of management control research is the use of subjectivity in incentive schemes. The first major study to attend this topic is about the determinants and effects of subjectivity (Gibbs, Merchant, Van der Stede, & Vargus, 2004). The use of subjectivity is driven by the fact that managers cannot foresee all events. To compensate managers for these uncontrollable events, firms can adjust goals afterwards (Gibbs et al., 2004).
The relation between targets and firm performance is not direct (Arnold & Artz, 2015). In order for targets to improve the firm performance, the targets have to flexible. This means that the firm should potentially adjust targets in the course of the target period.
A form of flexibility is downward adjusting the targets that were set at the beginning of the period (Gibbs et al., 2004). This is called ex post goal adjustments. This form of subjectivity is widely used in practice (Libby & Lindsay, 2010).
Very few studies have been done on the effects of ex post goal adjustments on employee behavior. A recent study about the influence ex post goal adjustments and goal difficulty on employee fairness perceptions showed that the positive effects only occur when goals are moderate (Kelly et al., 2014). When goals are difficult the effect does not hold.
I want to contribute to this new stream of literature by examining human behavior in response to ex post goal adjustments. The study of Kelly et al. (2014) suggests that further research is needed to assess under which conditions their results hold. Making sure that the employees have a formal policy available about the ex post goal adjustments could be an important factor. It could be the case that ex post goal adjustments only improve the fairness perception of employees when employees have a formal policy on how the performance evaluation process works.
In this project I address the question: How does a formal or non-formal policy on ex
2 Theory
This chapter provides an overview of the theoretical background of my research questions. First, there is explanation of how people decide on whether or not something is fair. This paragraph also includes an answer as to why companies would care about the fairness judgements of their employees. Further, I explain why policies and procedures are an important aspect of a performance evaluation system. Finally, I explain why subjectivity and more specifically ex post goal adjustments are important when employees assess the fairness of their benefits.
2.1 Fairness Perception and Performance Measurement
The relationship between an employee and an organization is defined by social exchange. The employee puts in effort to help the firm with achieving its objectives, while the firm rewards the employee for achieving these objectives (Cropanzano & Mitchell, 2005). This exchange is the basis for the relationship between the employee and the employer. This relationship changes over time, depending on the mutually rewarding transactions and relationships between the two parties. This means that the relationship between the employee and the employer is affected by the rewards that each party obtains. These rewards can be monetary or monetary. Examples of non-monetary rewards include knowledge and reputation.
Firm use performance measurement to help them achieve their goals (Merchant & Van der Stede, 2007). In order for a performance evaluation process to be successful, it must enhance the performance of a firm. Firms are using the performance evaluation process to reward and motivate their employees. The goal of this is to retain employees at the firm and be an attractive place to work for potential new hires (Folger & Cropanzano, 1998). So firms use their performance measurement system to assess the rewards earned, by having a social exchange with their employees. The objective from the perspective of the firm is to make sure is that it maximizes the benefits they are getting from the social exchange with their employees. This means that they want to let the employees work as efficient and hard as possible for the lowest amount of cost possible. So firms need to find a balance between the benefits that they offer their employees and the returns they expect in return.
Employees base their response to a performance evaluation system on the perceived fairness of the system (Latham, Almost, Mann, & Moore, 2005). Prior studies have indicated that employees who perceive their performance evaluation process as fairer are more satisfied and therefore more likely to improve their performance (Libby, 2001). This effect is explained by more effort and motivation of employees. Employees who perceive the performance evaluation process
as fair are more motivated and put more effort in their jobs. Organization want their employees to be more motivated and to put in more effort, because in most modern organizations the employees are the most important resource. If employees are treated unfair it is more likely that they leave or start behaving in a unsound way (Greenberg, 1993). This behavior can have negative consequences for the firm if this behavior spreads across the firm.
In order to understand what fairness perceptions are about, it is important to distinguish between distributive and procedural fairness perceptions (Greenberg, 1986). Procedural fairness is about the perceived fairness of the process that evaluates and rewards performances. Distributive fairness relates to the outcomes of a performance evaluation process compared to the work done by an employee (Colquitt, Scott, Judge, & Shaw, 2006). In the next paragraph there is a detailed explanation of the two types of fairness perceptions.
2.2 Procedural fairness perceptions
Procedural justice research focuses on the process in which decisions are made when two parties have conflicting interests (Thibaut & Walker, 1975). Based on this process, people will form a judgement about the fairness. This relates to the question whether a party thinks that is has been treated fairly by counterparty during the process. If employees think that they are treated unfairly this can lead, as explained earlier, to less motivation and effort, and unwanted behavior of employees.
There are two theoretical explanations for the psychological process that drives the decisions on procedural fairness. The first one is that people affected by a process want to have short term control over that process in order to improve the short term outcomes for themselves (Thibaut & Walker, 1975). Both the employee and the employer need to have sufficient information to make decisions. This information should be equal for both parties in order for the process to be viewed as fair. So both parties want to have information and influence on the process in order to achieve the best result in the own self-interest.
The second explanation is that parties want to have a long-term social relationship (Lind & Tyler, 1988). In order for this relationship to work, the underlying procedures are important. For example, the individuals' feelings of self-worth and group standing are influenced by the processes. This means that employees consider their relationship with other employees and superiors as complex. To assess whether or not the process is fair the employees will take into account the complexities of these relationships. An example of this is that a manager that had the duty to carry out the bonus policies was sick for a significant period of time. In the long term social relationship approach am employee will take this into account. In the short term self-interest approach, an
employee will only look at whether or not this impacted their control of the process in a negative way.
Both explanations are seen as accurate explanations of the psychology of procedural fairness (Lind & Tyler, 1988). The short term self-interest and the long term relationship approach both provide insights in how procedural fairness perceptions are formed.
2.3 Distributive fairness perceptions
Distributive fairness is about the outcomes of a performance evaluation process relative to the work done by someone (Colquitt et al., 2006). The first theory to provide an answer to the question of what outcomes employees think are fair, is given by the equity theory (Walster, Berscheid, & Walster, 1973). This theory argues that employees base their judgement about the fairness of an outcome on a comparison with what other employees get in terms of outcome. So the employees compare the bonus and salary they get with what others are getting. Employees will compare the outcome relative to the effort put in of themselves with the outcome-input ratio of their colleagues. If they feel they do not get enough in comparison with their peers, they will be unsatisfied. For example, someone will have lower fairness perceptions if a colleague with less work done gets the same bonus.
More recent scholars have argued that the equity theory is not sufficient in every case to explain the distributive fairness perceptions (van den Bos, 2001). There are many situations in practice possible where employees do not have information about the salaries and bonuses of co-workers to compare with their own. It could be that there is a lack of employees with the same function to compare your salary to. If, for example, there is only one controller in a small firm, it is impossible to compare your salary with other controllers within the firm. According to van den Bos (2011) in these situation employees will still be able to make an assessment on the questions if they feel the result is fair. However, these perceptions will not be based on the outcomes but on other information.
Several articles are written about the question whether or not it is important for organization to care about the distributive fairness perceptions of employees. It is important for firms to make sure their compensation system is fair, because the employees' evaluation of the fairness of a distribution will impact other opinions of employees (Alexander & Ruderman, 1987). The job satisfaction is lower for employees that feel the outcomes is unfair. Also, the evaluation of supervisors is lower in cases where employees feel unfairly treated. There is a higher turnover and there are more conflicts. Finally, the trust that employees have in management is lower.
2.4 Performance Evaluation Policies and Procedures
In order for performance evaluation processes to be successful, the organization has to be transparent towards its employees. Studies show that firms that are more openly communicating with its employees are more successful in integrating a complex system, for example (Chapman & Kihn, 2009).
The ideas of Adler and Borys (1996) are widely used in the field of management control research. There are two types of bureaucracy: enabling and coercive. The enabling use of performance measurement means that employees are helped in doing their work rather than restricted. In order to be successful in using performance measurement to enable employees, transparency is considered to be important (Wouters & Wilderom, 2008).
Adler and Borys (1996) identify two types of transparency that are important to the process of designing and implementing management control systems. Internal transparency refers to the understanding of the logic of a system’s internal function by the users of the system. Global transparency means that a management control system helps users understanding the broader context in which they do their work. The literature about organizational transparency shows that this is regarded to be an important factor in design and implementation of performance measurement systems. The effects on subjectivity in incentive schemes and more specifically ex post goal adjustments are not extensively studied.
Performance evaluation policies and procedures need to be adequate in order to be judged as fair by employees. Such policies and procedures are referred to as a due process (Folger, Konovsky, & Cropanzano, 1992; Taylor, Tracy, Renard, Harrison, & Carroll, 1995). This term is also used in the legal world where this has a more formal meaning of respecting a person’s legal rights during trial. In the context of a performance evaluation systems this refers to a system that treats the employees being evaluated fairly.
Folger et al. (1992) identity three characteristics that fair policies and procedures should have. The first one (1) is that there should be adequate notice. This means that the organizations publish, distribute and explain the content of the policies to employees. Also this, means that there should be enough feedback provided on the progress of the employees towards achieving the objectives.
The second (2) characteristic is that there should be a fair hearing. This refers to the formal meeting between superior and employee. In this meeting the employee should be informed about the superior’s assessment of performance. The employee should also have the opportunity to challenge the assessment of the manager by presenting evidence that opposes the evaluation.
The third (3) important aspect of a careful performance evaluation procedure is that managers should base their judgement on evidence. So the evaluation should be free of external pressures, personal corruption and external sources of bias. This also means that the evaluation standards are applied consistently between employees. There should be no favoritism, for example.
Locke and Latham (2004) developed an integrated model of work motivation. In this model they argue that the perceived fairness of the organizational policies and procedures affect the job satisfaction. Depending on how satisfied employees are they will be more or less committed to the goals of the organization. Both the satisfaction and the commitment will affect the actions of the employee. Forms of actions that are impacted include: job and work avoidance, protest, vengeance, defiance and adjustments. Below in figure 1 the relevant part of the model is included.
FIGURE 1
Work motivation and Justice Theories
From reviewing the literature about organizational policies and procedures the following can be concluded:
To enable employees in their work, it is important to have clear and transparent policies (Wouters & Wilderom, 2008).
Policies and procedures that are seen as adequate have the following characteristics: an adequate notice, fair hearing and careful procedure (Folger et al., 1992).
Organizational policies and procedures affect the satisfactions of employees. This will then translate into positive or negative subsequent actions by employees. (Locke & Latham, 2004)
2.5 Ex Post Goal adjustments
Subjectivity in an incentive contract means using not only quantitative performance measures, but also additional information relevant to fairness. This helps the firm by improving the alignment of the incentives and the objectives of the organization. The benefit for the employees is a reduced risk of uncontrollable events. The reason for this is that if such events occur, the evaluators have the opportunity to compensate the employees for these events (Baker, Jensen, & Murphy, 1988; Gibbs et al., 2004). Subjectivity only works if evaluators make fair and unbiased judgements. Also, the person being evaluated needs to accept the judgement and must no try to influence the evaluator in an inappropriate manner.
Traditionally, studies used the agency theory to explain why subjectivity in incentive schemes will improve firm performance. This theory states that if companies fails to compensate employees for uncontrollable risk, this will reduce effort (Prendergast, 1999). Because of this reduction of effort, the owners (principals) have to compensate the managers (agents) for this additional risk. The aim of this is to explain how subjectivity targets as a part of incentives contracts help improving firm performance.
Recent studies about subjectivity mainly focused on the determinants of subjectivity (Gibbs et al., 2004). So the question is: which firms are expected to use it and which are not. Also the effects were examined by Gibbs et al. (2004). Firms can use different ways of subjective performance evaluations. (Murphy & Cleveland, 1995; Prendergast, 1999). For example; (1) the complete bonus is based on subjective judgements, (2) the weight of the quantitative measures are determined subjectively or (3) a threshold is used to determine the bonus when a certain level is exceeded. Most firms use a combination of the above forms.
In most CEO compensation contracts there is some form of subjectivity present. The two most used form are: (1) the possibility to override the bonus formula ex post and (2) the ex-ante
absence of any formula in a bonus contract (Höppe & Moers, 2011). Both forms are used with different objectives. The ex post goal adjustments are used to decrease the risk associated with uncertainties. The absence of formulas is most often done to increase the congruence between the objectives of the firm and the incentives of the manager.
A recent survey by (Libby & Lindsay, 2010) showed that 80% of the companies that are using budget targets are using subjectivity to adjust goals after the target period. These ex post goal adjustment are becoming the subject of recent studies (Arnold & Artz, 2015; Burt, Libby, & Presslee, 2015; Kelly et al., 2014). The focus of these studies is on the employee behavior rather than the on the traditional agency theory in explaining how ex post goal adjustments can improve firm performance. Managers in practice take into account the negative consequences of unforeseen events in their ex-post performance evaluations (Burt et al., 2015). A specific form of doing such a correction for the unforeseen events is adjusting the goals downward after the period. The result is that employees do not have to increase their performance to make up for the uncontrollable event in order to earn a reward. The aim is of this is to make sure employees stay motivated.
In a recent survey by Arnold and Artz (2015) it is concluded that target adjustments are negatively associated with firm performance. The idea is that the employees see the signals that a target adjustment is likely, and they will as a consequence of that do less effort. The reason for this is that the employees already expect that the target will be adjusted and therefore are less motivated to achieve the initial higher target. So they will anticipate on the adjustment and change their behavior in a negative manner upfront.
An archival study by Burt et al. (2015) further explored the question whether ex post goal adjustments lead to lower firm performance. Their results indicate that the availability of target revisions can indeed result in lower individual performance. However, they conclude that subordinates will have increased fairness perceptions and will view their superior’s actions as more positive.
The study of Kelly et al. (2014) shows that ex post goal adjustments only help firms performance when goals are moderate and that it does not work when goals are difficult. This raises the question whether other conditions have to be met in order to successfully adjust goals afterwards. The authors raise the question whether the ex post goal adjustments only work if the policies process that led to the adjustments were clear and transparent.
From reviewing the literature about ex post goal adjustments the following can be concluded:
Firms use subjectivity with the goal to compensate employees for uncertainties (Gibbs et al., 2004).
Ex post goal adjustments is an often used form of subjectivity in practice (Libby & Lindsay, 2010).
Contrary to traditional agency theory, there are indications that ex post goal adjustment do not always increase firm performance (Arnold & Artz, 2015).
Recent research indicated that ex post goal adjustments only improve firm performance when targets are moderate (Kelly et al., 2014).
3 Hypothesis development
In this chapter is I describe the hypotheses that I have developed using the theory described in the previous chapter. After reading this chapter it should be clear what I want to study and how I aim to add to the existing literature.
Procedural and distributive fairness are two concepts that are closely related. People will sometimes use the same kind of information for their judgement about both types of fairness. This subsequently leads to the same judgement on both the procedural and the distributive fairness. This, for example, can be the case in situation where there is no outcome related information present to compare your own outcome with. According to the fairness heuristic theory proposed by van den Bos (2001), employees will then turn to information that relates to the performance evaluation process. So employees will judge the distributive and the procedural fairness using information about the process rather than also outcome related information. This can happen in practice, for example, when in an organizational culture employees are not talking and comparing their bonuses. This will leave employees only with process information to decide whether or not they think the bonus awarded is fair (Van den Bos, Vermunt, & Wilke, 1997; van den Bos, 2001). (Törnblom & Vermunt, 1999) argue that distributive and procedural fairness are two integrated concepts. Like Van de Bos et al. (2001) they argue that if information is lacking to make a judgement about the fairness, people will actively search for additional information or use information about the other component of fairness. If there is no information present about the process, people will look at the outcomes instead. If the outcome seems fair relative to the outcomes of colleagues, people will evaluate the process as more fair.
For this thesis I focus on the fairness perception of employees. The main reason for this is that it is difficult to measure a performance effect. As concluded in the archival study by Burt et al. (2015), there is doubt that ex post goal adjustments will increase firm performance. However, it is concluded that ex post goal adjustments will increase the fairness perceptions of employees.
The integrated model of work motivation by Locke and Latham (2004) also includes a theory on what factors influence fairness perceptions and satisfaction of employees. The model proposes that policies and procedures are an important design choice of the performance evaluation system.
My expectation is that if there are is a policy that is more formal on ex post goal adjustments, that people will see it as more fair. Folger et al. (1992) argue that the policies and
procedures need to include an adequate notice to employees, a formal fair performance review hearing and a careful process.
So I expect that a formal policy and procedure on ex post goal adjustments that includes these three characteristics, is considered to be more fair. My hypothesis is:
H1: Fairness perceptions of employees are higher when there is a formal policy about
ex-post goal adjustments rather than there is no such policy.
In the theory chapter I explained that fairness perceptions can be divided into two different theoretical concepts: procedural and distributive fairness. Procedural fairness is about the perceived fairness of the process that evaluates and rewards performance (Greenberg, 1986). A formal policy leads to a higher procedural fairness perception by employees (Giles, Findley, & Feild, 1997). This is because employees will have more trust in the management if there is a formal policy (Hartmann & Slapničar, 2009).
When there is a formal policy on ex post goal adjustments, I therefore expect the procedural fairness perceptions to be higher than when there is no such policy. The sub hypothesis I formulated is the following:
H1a: Procedural Fairness perceptions of employees are higher when there is a formal
policy about ex-post goal adjustments rather than there is no such policy.
There are situations thinkable where there is no information present to compare your own bonus with others. When there is no relevant outcome related information Van de Bos (2001) argues that employees will look for other information as a substitute. In that case, the only information left then would be information about the process. This is called the fair process effect by Van den Bos (2001). Subordinates will be more positive towards their outcomes following a fair rather than an unfair process. This theory leads to my hypothesis that when there is a formal policy that also the distributive fairness perceptions of employees will change. I expect higher distributive fairness perceptions for employees that are provided with a formal policy during the performance evaluation process. Employees will use the information about the process as a heuristic to judge the distribution when there is no information on the outcome of comparable other persons.
The idea that processes also influence the perception on distribution fairness by employees is challenged by an equity theory perspective (Adams, 1965). Employees will compare their effort put into the job with outcomes the employee gets. As a result of this theory, the expectation is that it does not matter whether there is a formal policy. The reason for this is that employees will only care about the outcome in comparison to their own effort. In case the outcomes are held constant,
this means that a change in distributive fairness perceptions is not expected, whether there is a formal policy or not.
In case of a formal policy versus no formal policy on ex post goal adjustments, I expect that the fair process effect of Van den Bos (2001) will be observable. The reason for this is that the subjective nature of the ex post goal adjustment will make employees more suspicious towards the management. This will make employees more curious towards the process where a formal policy could be a part of. As studied by Hartmann and Slapničar (2009) a formal policy leads to more trust in the management. Because of this enhanced trust, there will be less suspicion towards the process of the Ex Post Goal Adjustments. When there is no information present to compare your outcomes with other employees, the only thing left to look at will be the process.
My second sub hypothesis is therefore:
H1b: Distributive Fairness perceptions of employees are higher when there is a formal
policy about ex-post goal adjustments rather than there is no such policy.
FIGURE 2
Hypothesis overview
Formal policy versus non-formal policy H1: Overall Fairness perceptions H1a: Procedural fairness perceptions H1b: Distributive fairness perceptions
4 Research methodology
In order to test whether my hypotheses are confirmed in practice, I have performed an experiment. In this chapter I explain what I will explain my research methods. First, I explain the design of my experimental case. Second, I will explain the variables that I plan on using.
4.1 Experimental design
The most appropriate method to use for my thesis project would be an experiment. It would be very difficult to use actual data because all others factors that can impact the employee fairness perceptions would have to be held constant. The benefit of using experiments is that a high internal validity can be achieved. (Schulz, 1999; Smith, 2014). For experiments in management accounting it is feasible to use students (Liyanarachchi, 2007). A total of 48 respondents should be enough, because of the high internal validity of experimental research. I got the respondents by asking fellow students, business contacts and friends and family to participate. I sent everybody a personal message asking them if they want to help with my thesis by filling in a questionnaire about performance evaluation. All participants completed the questionnaire digitally (smartphone or computer). I have translated the survey into Dutch to make is easier for Dutch speakers to understand the questionnaire and case.
The case is about a manager of a bicycle store in the center of Amsterdam. The shop is part of a larger chain of bicycle stores. The general manager carries a lot of responsibility, but he has to perform very well in order to get his bonus from the top management of the chain. The chain is called “Pedaalgigant” and there is also a picture included to bring the case more to life (see appendix 1 for the picture). The case also mentions that the top management sets the targets and that these targets are challenging but achievable.
I plan on using two groups that will each get a different version of an experimental case. The first group gets a case where there is no formal policy on the ex post goal adjustments. In the second group, there will be a formal policy and process on the ex post goal adjustments explained in the case.
TABLE 1
Experimental groups
Case I: Case II:
No formal policy on Ex Post Goal Adjustments
A formal policy on Ex Post Goal Adjustments
In order to operationalize the formal policy versus no formal policy, I manipulated a couple of sentences in the case. In the situation with no formal policy, there is an explanation that instead of a formal policy management will use their own judgement and experience to decide on the bonus.
Conditions with NO formal Policy on Ex Post Goal Adjustments:
Conditions with A Formal Policy on Ex Post Goal Adjustments:
4.2 Questionnaire items
In this paragraph I will explain how I will operationalize the variables used in the experiment. In appendix 1 the questionnaire items are included. In this paragraph I refer to the authors that have developed these scales. I plan on using on existing scales, because this is more reliable than creating new items.
In order to measure the dependent variable fairness perception, I have used an existing scale consisting of 7 items (Kelly et al., 2014). This scale measures the fairness perceptions of employees by making a distinction between procedural and distributive justice.
I asked participants whether or not they think there was a formal policy on Ex Post Goal Adjustments. This is done to check if the participant paid attention when reading the case.
There are also questions asked about the individual respondent. This includes questions about age, gender, work experience, performance evaluation experience and education level.
There is no formal policy that describes in which situation and under which conditions the shop managers will be compensated for uncontrollable events. Instead of a formal policy, the top managers will use their own experience and judgment to decide on the bonus of Justin if these events occur.
There is a formal policy that describes in which situation and under which conditions the shop managers will be compensated for uncontrollable events. This policy contains a detailed description of uncontrollable events that qualify for target adjustment. The top managers will use this as a basis for their decision on the bonus of Justin if these events occur.
5 Results
In this chapter there is an overview of the results of the experiment. First, there is analysis of the demographics of the participants. Furthermore, the independent and dependent variables are described and analyzed. Finally, the hypotheses are tested and there is a supplemental analysis. Below there is an overview table of names used for the variables.
TABLE 2
Formalpolicy Nominal variable: 0 = no formal policy, 1 = formal policy
Education Ordinal variable: 1= University, 2= Professional, 3= Intermediate, 4=High school.
Responsetime Nominal variable: 0 = first day, 1 = later Language Nominal variable: 0 = Dutch, 1= English Age Continuous variable: number of years Gender Nominal variable: 0 = men, 1 = woman
Wexperience Continuous variable: number of years of work experience
PExperience Continuous variable: number of years of experience being the subject of a formal performance evaluation
ProceduralFairness Continuous variable: measured on a 7-point Likert scale DistributiveFairness Continuous variable: measured on a 7-point Likert scale CombinedFairness Continuous variable: measured on a 7-point Likert scale
5.1 Demographics
To test whether or not there was a difference between late and early response a dummy variable was created. There were 8 participants that filled the questionnaire in on the first day it was available. The other participants filled in the survey later (n = 40). An Independent Samples Test was performed with Formalpolicy as the dependent variable. This found t-statistic of 0.254 and a p-value of 0.801 (Spearman correlation = -0.037, p-value 0.801). So there was no difference in the randomization of the cases the between the group that received the questionnaire early and the group that received is later. So it was not the case that the early groups for example only
received the formal case and not the informal case. This was equally distributed among early and late respondents.
In order to make the survey better to understand for Dutch participants, there were two version of the questionnaire. A total of 32 participants completed the questionnaire in English and 16 in Dutch. An Independent Samples T-Test was performed with Formalpolicy, as the dependent variable. This found t-statistic of -1.651 and a value of 0.106 (Spearman correlation = 0.237, p-value 0.106). This means that there were no significant differences in the distribution of the cases between the two languages of the experiment.
In table 3 there is an overview of the demographics between the experimental groups. In the group with the formal policy the average age was 22.82 and in the group with a non-formal policy the average age was 22.82. The total of 48 valid responses included 31 men (=65%) and 17 women (=35%). The average number of years work experience was 4.67 across groups. On average participants had 2.88 years’ experience being the subject of a performance evaluation.
TABLE 3
Descriptive statistics of the Experimental Groups Demographics Formal Policy No (n=26) Yes (n=22) Age Mean 24,31 22,82 Std. Deviation 6,17 3,71 Std. Error Mean 1,21 0,79 Gender Number of men 16 15 Number of woman 10 7 Mean 1,38 1,32 Std. Deviation 0,50 0,48 Std. Error Mean 0,10 0,10 Wexperience Mean 5,04 4,23 Std. Deviation 4,85 2,84 Std. Error Mean 0,95 0,61 Pexperience Mean 3,23 2,45 Std. Deviation 2,84 2,24 Std. Error Mean 0,84 0,48
The questionnaire was mostly distributed among students. Therefore, the respondents had mostly a professional or a university education. Under 20% of the participants had a lower education. Table 4 gives an overview of the education level of respondents. I think this is a proper reflection of the population, because the employees that deal with incentive schemes in practice are also mostly higher educated.
TABLE 4
Highest Completed Education of Valid Respondents
Formalpolicy = no Formalpolicy = yes
Frequency Percentage Frequency Percentage
University (Bachelor or Master) 8 31% 8 36%
Professional Education (HBO) 14 54% 10 45%
Intermediate Education (MBO) 1 4% 1 5%
High School 3 12% 3 14%
Total 26 100% 22 100%
In table 5 there is an overview of the results of an independent T-test with the variable Formalpolicy as the independent variable. This is done to check if there are signals that indicate the demographics differ between the two experimental groups. The Levene’s test was significant (p > 0.05) for all the demographic variables. This means that there is homogeneity of variances within the sample for these variables. The education level of the respondent is not significantly different between the two experimental groups (t = 0.025, p=0.980). There is no significant difference in age between the groups (t = 0.327, p=0.990). Furthermore, there were also no differences found between males and females between the two cases (t = 0.471, p=0.640). Fourth, there was no (significant) difference in the participants' work experience between the two cases (t = 0.690, p=0.493). Finally, the experience with being evaluated on performance was not different between the two experimental groups.
TABLE 5
Independent Samples T-Test (Independent variable = Formalpolicy)
Education Age Gender Wexperience Pexperience
Levene's test 0,613 0,522 0,350 0,839 0,140 T statistic 0,025 0,990 0,471 0,690 0,768 Degrees of freedom 46 46 46 46 46 Significance (2-tailed) 0,980 0,327 0,640 0,493 0,446 Mean Difference 0,007 1,490 0,066 0,811 0,776 Std. Error Difference 0,277 1,505 0,141 1,175 1,011 Lower: 95% Confidence -0,550 -1,539 -0,218 -1,554 -1,258 Upper: 95% Confidence 0,564 4,518 0,351 3,176 2,811 5.2 Independent variables
A total of 51 participants completed the questionnaire. After reading the case, the participants were first asked a manipulation check question. This question was whether or not there was a formal policy. From the total of 51 participants there were 48 participants who answered this questions in accordance with the case read. The three responses that did not answer the manipulation check correctly were removed from the sample. A nominal variable was created that has the value 0 if there was no formal policy and 1 if there was a formal policy.
5.3 Dependent variables 5.3.1 Factor analysis
To measure the dependent variable I used a scale developed in prior research (Colquitt et al., 2006; Thibaut & Walker, 1975) and recently confirmed by Kelly et al. (2014). The scale makes a distinction between procedural and distributive fairness.
The items included in both scales are shown below in table 6. This table gives a factor analysis using a varimax rotation with Kaiser normalization. The factor loadings for the three procedural fairness items are 0.66 or higher. On the distributive fairness factor the four items have a loading of 0.84 or higher. This is very similar to the loadings observed by Kelly et al. (2014). In their factor analysis they found a procedural fairness loading of the items of 0.68 or higher, while those for the distributive fairness were 0.80 or higher. This means that I can conclude that the
TABLE 6
Factor Analysis of Procedural Fairness and Distributive Fairness
Factor Loadings Item (n = 48 for each) Procedural Fairness Distributive Fairness
1. I think the process that the top management uses to
determine incentive payouts is applied consistently to Justin. 0,88 0,18
2. I think the process that the top management uses to
determine incentive payout has been fair to Justin. 0,87 0,21
3. I think the process that the top management uses to
determine incentive payout to Justin has been bias free. 0,66 0,31
4. The amount of incentive payout received reflects effort that
has been put into the job by Justin. 0,19 0,90
5. Incentive payout is appropriate for the effort put into the job
by Justin. 0,35 0,84
6. The amount of incentive payout reflects the contribution to
the job of Justin. 0,30 0,87
7. The amount of payout is justified given the performance of
Justin. 0,20 0,92
Eigenvalue 2,26 3,30
Percent of variance explained 32,27 47,07
Method: Varimax rotation with Kaiser Normalization
5.3.2 Reliability analysis
To assess the reliability of the scales a reliability analysis is included in table 7. In this table on the first row Cronbach’s alpha of both the procedural fairness and the distributive fairness scale are included. Also, the combined scale of all the seven items is assessed in the last column. The Cronbach’s alpha of the procedural fairness scale of three items is 0.79. The distributive fairness scale of four items has a Cronbach’s Alpha of 0.94. The combined scale of all seven items has a Cronbach’s Alpha of 0.89. All three constructs have a Cronbach’s Alpha well above the desired threshold of 0.7.
In order to analyze whether it is necessary to exclude items from the scales, I have included the Cronbach’s Alpha’s when an item is deleted. The only noticeable improvement is that the procedural fairness scale could be improved from 0.79 to 0.84 by deleting the third items of the scale. Because the Cronbach Alpha of the procedural scale is already above 0.7 and the improvement by removing the third item is not very large, I have chosen to not remove this item. In order to measure the dependent variable procedural fairness, I will use the mean of the first three items, while I use the mean of the last four items to measure the distributive fairness.
TABLE 7
Reliability Analysis of Procedural Fairness and Distributive Fairness
Cronbach's Alpha if item deleted Item (n = 48 for each) Procedural Fairness Distributive Fairness Combined Fairness
Cronbach's Alpha 0,79 0,94 0,89
1. I think the process that the top management uses to determine incentive payouts is applied
consistently to Justin. 0,64 0,89
2. I think the process that the top management uses to determine incentive payout has been fair
to Justin. 0,64 0,88
3. I think the process that the top management uses to determine incentive payout to Justin has
been bias free. 0,84 0,89
4. The amount of incentive payout received reflects effort that has been put into the job by
Justin. 0,92 0,87
5. Incentive payout is appropriate for the effort
put into the job by Justin. 0,93 0,86
6. The amount of incentive payout reflects the
contribution to the job of Justin. 0,92 0,86
7. The amount of payout is justified given the
performance of Justin. 0,91 0,86
5.3.3 Normality assumption
The independent samples T-Tests that I used in the analysis is based on the assumption that the data follows a normal distribution. In this paragraph I will address this assumption by looking and the data.
In order to review the symmetry of the skewness is calculated. The skewness of the three dependent variables differs between -0,62 and – 0,12. This means that there is negative skewness. So the long tail should be at the right of the distribution. This is also visible in table 8.
In order to assess the significance, the Z-score of the skewness has been calculated. A z-score for skewness between -2 and +2 is considered acceptable for a normal distribution (George & Mallery, 2010). The DistributiveFairness variable in the condition with a formal policy has the lowest value with -1.26. This means that there is no indication that the data has unacceptable skewness.
To assess the tailedness of the distribution the kurtosis has been analyzed. Just as with the skewness, the z-score has been calculated. A z-score for kurtosis between -2 and +2 is considered acceptable for a normal distribution (George & Mallery, 2010). There were no values found outside this acceptable range. So there is no indication that kurtosis is a problem.
TABLE 8
Skewness and Kurtosis
ProcessFairness DistributiveFairness Combined Fairness Variable Policy No (n = 26) Policy Yes (n = 22) Policy No (n = 26) Policy Yes (n = 22) Policy No (n = 26) Policy Yes (n = 22) Skewness -0,28 -0,14 -0,12 -0,62 -0,43 -0,45 Std. err. Skewness 0,46 0,49 0,46 0,49 0,46 0,49
Z-score skew. (a) -0,61 -0,29 -0,27 -1,26 -0,95 -0,92
Kurtosis -0,87 -0,86 -1,46 -0,05 -1,13 0,31
Std. err. Kurtosis 0,89 0,95 0,89 0,95 0,89 0,95
Z-score kurtos. (b) -0,98 -0,90 -1,65 -0,05 -1,27 0,32
(a). Z-score skewness = skewness/std. error skewness (b) Z-Score kurtosis = kurtosis/std. error kurtosis
Besides assessing the numbers, it is also useful to look visually at the data. In order to see whether or not the data looks normally distributed, I have included the histograms and PP-Plots of the variables in figure 3. The data seems to roughly follow a normal distribution. Although there is some skewness to the right visible. The PP-Plot shows that the points are not very deviated from the diagonal. By looking at the data I conclude that the data is showing a roughly normal pattern.
FIGURE 3
Histograms and P-P Plots (Formal Policy = No)
ProcessFairness Distributive Fairness Combined Fairness
Histograms and P-P Plots (Formal Policy = Yes)
To further test the data for normality, I tabulated the results of the Kolmogorov-Smirnova and Shapiro-Wilk test. When the sample size is below 50, the Shapiro-Wilk test is considered to be the most accurate (Field, 2009). The Shapiro-Wilk test showed that there are normality problems with the variable ProcessFairness in the condition with a formal policy (p < 0.05). The other variables had no significant problems with the Shapiro-Wilk test (p > 0.05).
TABLE 9
Normality test of dependent variables
Kolmogorov-Smirnova Shapiro-Wilk
Policy No
(n = 26) Policy Yes (n = 22) Policy No (n = 26) Policy Yes (n = 22) variable Statistic sig. Statistic sig. Statistic sig. Statistic sig.
Process Fairness 0,17 0,05 0,19 0,05 0,95 0,20 0,90 0,03
Distributive Fairness 0,14 0,19 0,11 0,20* 0,90 0,02 0,96 0,45
Combined Fairness 0,19 0,01 0,13 0,20* 0,91 0,03 0,96 0,44
*. This is a lower bound of the true significance. a. Lilliefors Significance Correction
Overall, the data looks normally distributed. However, because of one of the normality tests showed that there are problems with data normality, I cannot assume normality fully. In order to address this issue, I will include the nonparametric tests in my analysis. These tests do not have the assumption that data is normal distributed.
5.3.4 Descriptive statistics
In table 10 there is an overview of the descriptive statistics three dependent variables that were computed earlier. The mean of the combined fairness perceptions is 4.29. On the seven-point scale that was used to measure the dependent variables, this indicates that on average participants thought that the case was more fair than unfair.
TABLE 10
Descriptive Statistics
N Minimum Maximum Mean Std. Deviation
ProcessFairness 48 1,00 6,33 4,22 1,39
DistributiveFairness 48 1,75 6,75 4,34 1,42
CombinedFairness 48 1,57 6,43 4,29 1,23
In the experiment there were two groups. The descriptive statistics between the groups are included in table 12. The first group read about a policy that was not formal. The case of the second groups was about a formal policy. A total of 26 (=54%) respondents read about a non-formal policy and 22 (=46%) had a non-formal policy. The group with the non-formal policy showed higher
means on all three dependent variables. The group with a non-formal policy has a mean of the procedural fairness of 3,37, for distributive fairness 3,87 and for the combined fairness 3,65. The group with a formal policy has a mean of 5,21 for procedural fairness, for distributive fairness 4,91 and 5,04 for the combined fairness.
In figure 4 the boxplots between the experimental groups are included. Again, it is visible that in the groups with a formal policy the mean on both dependent variables is higher. Also, there were no outliers found. This is an indication that there are no extreme values that are influencing the result in a disproportionate manner. Also, the minimum value (Lower Whisker) for both the dependent variables is higher. The maximum value (Upper Whisker) of both dependent variables is also higher in the groups with a formal policy.
By reviewing the descriptive statistics, it looks as if there are clear differences on the dependent variables between the two experimental groups.
FIGURE 4
5.3.5 Correlation table
In order to review whether there is unexpected correlation between variables, I included a correlation table. Because of the concerns with normality, I have included the non-parametric Spearman correlation above the diagonal. I found no unexplainable correlations. Age seems to correlate with education, work experience and the experience with performance evaluations. This is logical, because older people have more work experience and are higher educated.
TABLE 11
Correlation table (p values between brackets)
Age Gender Education WExp PExp PF DF CF
Age -0,056 -0,668** 0,552** 0,384** 0,134 -0,131 -0,064 (0,707) (0,000) (0,000) (0,007) (0,365) (0,375) (0,665) Gender -0,157 0,134 -0,295* -0,021 -0,068 0,039 -0,028 (0,284) (0,364) (0,042) (0,888) (0,646) (0,790) (0,848) Education -0,494** 0,126 -0,358* -0,175 -0,182 0,032 -0,054 (0,000) (0,393) (0,013) (0,234) (0,215) (0,828) (0,717) Wexperience 0,830** -0,255 -0,367* 0,629** 0,054 0,017 0,015 (0,000) (0,081) (0,010) (0,000) (0,715) (0,908) (0,921) Pexperience 0,729** -0,049 -0,235 0,806** 0,011 0,155 0,101 (0,000) (0,740) (0,108) (0,000) (0,940) (0,291) (0,497) ProcessFairness 0,168 -0,042 -0,134 0,112 0,032 0,458** 0,779** (0,253) (0,776) (0,364) (0,448) (0,828) (0,001) (0,000) DistributiveFairness 0,048 -0,003 0,059 0,077 0,139 0,532** 0,883** (0,744) (0,984) (0,692) (0,601) (0,347) (0,000) (0,000) CombinedFairness 0,113 -0,022 -0,026 0,105 0,107 0,831** 0,913** (0,445) (0,881) (0,861) (0,478) (0,471) (0,000) (0,000)
Pearson correlations appear below the diagonal, non-parametric Spearman correlations appear above the diagonal
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
5.3.6 Hypothesis tests
In this paragraph I will perform statistical test to check whether the effects that I expected were empirically found.
H1: Fairness perceptions of employees are higher when there is a formal policy about
ex-post goal adjustments rather than there is no such policy.
The main hypothesis is that fairness perceptions in general were higher in the group with a formal policy. This is tested using an independent samples T-test with as the dependent variable the combined fairness perceptions. The Levene’s test for equality of variances was significant p = 0.022. This means that the assumption of the homogeneity of variance has been broken. The standard deviation in the group with no formal policy was 1.16 and in the group with a formal policy the SD was 0.84. Because of this difference in SD the homogeneity of variances assumption was broken.
On average, participants given a formal policy had higher fairness perceptions (M = 5.04, SE = 0.18), than those with no formal policy (M = 3.65, SE = 0.23). The tstatistic of this test is -4.782. This difference was highly significant p = 0.000. Because of the problem normality concerns with the data, I have also performed a bootstrapped independent samples t-test with a 1000 samples. The p-value of this test was 0.001. The bootstrapped confidence interval is [-1.95, -0.82]. According to Field (2009), it is best for bootstrapping to use the confidence interval to interpret the result. In this case zero is clearly outside of the confidence interval, therefore I conclude that there is a significant difference between the fairness perceptions between experimental groups.
Furthermore, I did the nonparametric Mann-Whitney U test. This was also had a highly significant result with a p-value of 0.000 (U = 86).
Overall, my conclusion is that there are significant differences of the fairness perceptions between the group with a formal policy and the group with a non-formal policy. This finding is robust, because the results of the bootstrapped t-test and the nonparametric test also found highly significant results.
H1a: Procedural Fairness perceptions of employees are higher when there is a formal
policy about ex-post goal adjustments rather than there is no such policy.
The first sub hypothesis is that procedural fairness perceptions were higher when there is a formal policy. I also used an independent samples T-test for this hypothesis. The dependent
questionnaire items about procedural fairness. The Levene’s test for equality of variances was significant p = 0.006. This means that the assumption of the homogeneity of variance has been broken. The standard deviation in the group with no formal policy was 1.25 and in the group with a formal policy the SD was 0.72. Because of this difference in SD the homogeneity of variances assumption was broken.
On average, participants given a formal policy had higher procedural fairness perceptions (M = 5.21, SE = 0.15) than those with no formal policy (M = 3.37, SE = 0.25). The independent sample t-test found that this difference was highly significant with a p-value of 0.000 (t = -6.36). A bootstrapped independent samples t-test found a p-value of 0.001 (1000 bootstrap samples). The bootstrapped confidence interval is [-2.42, -1.23]. So zero is clearly outside the interval. The nonparametric Mann-Whitney U Test gave also a highly significant p-value of 0.000 (U = 62).
Figure 5 is a scatterplot of with a green dot for participants that got a formal policy and a red dot for participants with a non-formal policy. On the y-axis the score on the variable ProcessFairness is plotted. It is clearly visible that almost all green dots are higher on the y-axis than the red dots. So this means that almost all participants with a formal policy gave higher procedural fairness ratings than the participants with a non-formal policy.
My conclusion is that there are significant differences of the procedural fairness perceptions between the group with a formal policy and the group with a non-formal policy. This finding is robust because a bootstrapped t-test and a non-parametric gave high significance levels.
FIGURE 5
H1b: Distributive Fairness perceptions of employees are higher when there is a formal
policy about ex-post goal adjustments rather than there is no such policy.
The second sub hypothesis is that distributive fairness perceptions were higher when there is a formal policy. To test this, I used an independent samples T-test. The dependent variable used is DistributiveFairness. This variable was calculated using the mean of the three questionnaire items about procedural fairness. The Levene’s test for equality of variances was not significant p = 0.202. This means I can assume the homogeneity of variances. The standard deviation in the group with no formal policy was 1.42 and in the group with a formal policy the SD was 1.21. Because of this difference in SD, the homogeneity of variances assumption was broken.
On average, participants given a formal policy had higher distributive fairness perceptions (M = 4.91, SE = 0.26) than those with no formal policy (M = 3.87, SE = 0.28). This difference is highly significant; the p-value of the independent samples T-test was 0.009 (t= -2.71). The bootstrapped (1000 samples) independent samples T-test had a p-value of 0.008. The nonparametric Mann-Whitney U Test has a p-value of 0.014 (U = 167).
In figure 5 the scores on the distributive fairness are on the x-as of the scatterplot. In this plot it is visible that the green dots (formal policy) are more on the right side of the graph than the red dots (non-formal policy). This difference is however less evident than the difference on the y-axis concerning the procedural fairness. Overall, I conclude that there are significant differences of the distributive fairness perceptions between the group with a formal policy and the group with a non-formal policy. This finding is robust, because a bootstrapped t-test and a non-parametric gave high significance levels.
TABLE 12
Tests and descriptive statistics of the Experimental Groups
Formal Policy No (n=26) Yes (n=22) ProcessFairness Mean Δ 1,84 3,37 5,21 Std. Deviation Δ -0,53 1,25 0,72 Std. Error Mean Δ -0,10 0,25 0,15 Levene's test 0,006* T statistic -6,364 Degrees of freedom 40,826 Significance (2-tailed) 0,000***
Bootstrapped sig. (1000 samples) 0,001***
Mann-Whitney U test 0,000*** Lower: 95% Confidence -2,424 Upper: 95% Confidence -1,256 DistributiveFairness Mean Δ 1,04 3,87 4,91 Std. Deviation Δ -0,21 1,42 1,21 Std. Error Mean Δ -0,02 0,28 0,26 Levene's test 0,202 T statistic -2,711 Degrees of freedom 46,000 Significance (2-tailed) 0,009**
Bootstrapped sig. (1000 samples) 0,008**
Mann-Whitney U test 0,014** Lower: 95% Confidence -1,819 Upper: 95% Confidence -0,269 CombinedFairness Mean Δ 1,39 3,65 5,04 Std. Deviation Δ -0,32 1,16 0,84 Std. Error Mean Δ -0,05 0,23 0,18 Levene's test 0,022* T statistic -4,782 Degrees of freedom 45,004 Significance (2-tailed) 0,000***
Bootstrapped sig. (1000 samples) 0.001***
Mann-Whitney U test 0.000***
Lower: 95% Confidence -1,969
5.3.7 Supplemental analysis
To check whether there was any mediation in between the variables, I used the PROCESS procedure (Hayes, 2013). The output of the macro is included in appendix 3. I found a significant indirect effect of the formal policy on the distributive fairness perceptions. The direct effect was not significant with a p-value of 0.87. The indirect effect was assessed using bootstrap confidence intervals (Field, 2009). The indirect is considered significant if zero is well outside the confidence interval. In this case the confidence interval is [0.27, 1.77]. This means that the indirect effect is clearly significant in this case. Consistent with our initial analysis the relationship between a formal policy and the procedural fairness was significant (b = 1.84, p < 0.01). Also the link between the procedural fairness variable and the distributive fairness was significant (b = 0.52, p < 0.01).
The relationship between a formal policy and distributive fairness is mediated by the perceptions on the process fairness of an employee. In the experiment participants were first asked to give their opinion on the procedural fairness. After that they were presented with the information that the bonus was awarded. Then, the questions about the distributive fairness were asked.
So it was not the formal policy or the non-formal policy as such that influenced the fairness perceptions of employees. It rather was the opinion that was formed about that policy. The thinking about and judging the process mediated the subsequent judgement about the outcome/distribution.
FIGURE 6
Mediation model
The confidence inverval for the indirect effect is a BCa bootstrapped CI based on 1000 Formal Policy b = 1.84, p < 0.01 b = 0.52, p < 0.01 Direct effect, b = 0.08, p = 0.87 Indirect effect, b = 0.96, 95% CI [0.27, 1.77] ProcessFairness DistributiveFairness
6 Discussion and Conclusion
My main hypothesis was that if there is a formal policy about ex-post goal adjustments, that employees will have higher fairness perceptions about their compensation. I found significant evidence that there were higher fairness perceptions when there was a formal policy rather than an informal policy. This is consistent with the findings of Hartmann and Slapničar (2009) a formal policy leads to higher trust in managers, which in turn leads to higher fairness perceptions.
I formulated two sub hypotheses. The first sub hypothesis is that not only the fairness perceptions in general are higher, but also more specifically the procedural fairness perceptions. Consistent with this theory, I found that the fairness perceptions are significantly higher when there is a formal policy compared to when there is a non-formal policy (Giles et al., 1997). In my view, the reason for this is that the ex post goal adjustments are subjective. The nature of subjectivity is that it is uncertain for employees what the outcomes will be. A formal policy can offer something to hold on to for the employee evaluated. When a manager uses only judgement instead of a policy an employee can feel, that it is only the goodwill of the manager that decides.
The second sub hypothesis was that also the distributive fairness perception of employees is higher when there is a formal policy about ex post goal adjustments. With this hypothesis, I test the fair process theory of Van de Bos (2001) that argues that the process is also relevant for distributive/outcome fairness perceptions. I found conformation of this fair process effect, because the distributive fairness perceptions were higher when there was a formal policy compared with to there was no such policy. This finding contradicts the equity theory that states that employees will compare their own effort with the bonus that they receive (Walster et al., 1973). This theory argues that the relationship between an employer and employee is about effort and financial compensation. This contradicts the more recent social exchange theory (Cropanzano & Mitchell, 2005). This theory states that parties will compare the costs in term of effort and money, and the rewards in term of acceptance, support and money. My findings are consistent with this theory. I found that employees will also care about the performance evaluation process irrespective of the outcome in terms of money.
The supplemental analysis on mediation revealed additional insight in the forming of the opinions of participants about the fairness perceptions. When an employee has no relevant information to compare their own outcomes with, the employee will use the opinion that was formed about the process during the period. This can happen in practice when an employee has a very unique position where there are no other employees with similar positions to compare their own bonus with. An employee will then still have an opinion on the fairness of the bonus using
the opinion that was formed during the year about the process. So if an employee feels treated fairly during performance evaluation process, this will impact the extent to which an employee will be happy with his or her final bonus.
A problem with the internal validity is that the manipulation made could be too obvious. This is an inherent limitation with experimental research, because it is important that participants notice the manipulation. In reality the difference is not that obvious. Companies might have a formal policy that is in place but is not often used, or one that is not clear. In that case there is a formal policy, but this will likely not yield the same results as in this experiment.
Furthermore, there are concerns in regard to the external validity of the experiment. The fact that I used mainly students for this experiment could lead to difficulties when generalizing the results. Students might have less financial responsibilities in term of having to pay off a mortgage, for example. Because of this, students might be less focused on getting the highest bonus and are more focused on being treated fairly.
I operationalized this by holding the outcomes constant in both experimental groups. In both groups the bonus was awarded. The questions remain how relevant the process would be when the outcome will be unfair in comparison with colleagues. It might be the case that employees will not care about the process anymore if the outcome impacts them negatively.
Another suggestion for further research is to test whether the fair process applies to other situations then ex post goal adjustments. One would an expect a smaller effect in a formula bonus only system because this is set up front and then the process does less matter. There is also less opportunity for the manager to use discretion to influence to outcomes.
References
References
Adams, J. S. (1965). Inequity in social exchange. Advances in Experimental Social Psychology, 2(267-299)
Adler, P. S., & Borys, B. (1996). Two types of bureaucracy: Enabling and coercive. Administrative
Science Quarterly, , 61-89.
Alexander, S., & Ruderman, M. (1987). The role of procedural and distributive justice in organizational behavior. Social Justice Research, 1(2), 177-198.
Arnold, M. C., & Artz, M. (2015). Target difficulty, target flexibility, and firm performance: Evidence from business units’ targets. Accounting, Organizations and Society, 40, 61-77.
Baker, G. P., Jensen, M. C., & Murphy, K. J. (1988). Compensation and incentives: Practice vs. theory. The Journal of Finance, 43(3), 593-616.
Burt, I., Libby, T., & Presslee, A. (2015). The effects of ex post goal adjustment and social identity with a superior on subordinates’ performance. Canadian Academic Accounting
Association (CAAA) Annual Conference, doi:http://dx.doi.org/10.2139/ssrn.2375183
Chapman, C. S., & Kihn, L. (2009). Information system integration, enabling control and performance. Accounting, Organizations and Society, 34(2), 151-169.
Colquitt, J. A., Scott, B. A., Judge, T. A., & Shaw, J. C. (2006). Justice and personality: Using integrative theories to derive moderators of justice effects. Organizational Behavior and Human
