THE EFFECTS OF JUSTICE ON PAY SATISFACTION AND
JOB STRESS
Wouter Koning (0012521)
Amsterdam Business School University of Amsterdam
Master of Science: Accountancy and Control August 2007
Supervisor: Dr. V.S. Maas
Second Supervisor: Dr. Ir. S. van Triest
Abstract: This study examines the relations between justice, pay satisfaction and job stress in a reward system context. An additive model was used to look at these relations. It was hypothesized that feelings of justice will positively affect pay satisfaction and negatively affect job stress. The hypotheses are developed based on several theories like the equity theory, referent cognitions theory, and role dynamics theory, but also on existing literature. The survey was online distributed to 62 employees at a particular business unit at ABN AMRO Bank N.V.. The data and the hypotheses were researched through a regression analysis in terms of explanatory power and slope. Data analysis provided evidence for a positive relation between distributive justice and pay satisfaction. Furthermore a negative relation was found between procedural justice and job stress. No relation was found between procedural justice and pay satisfaction, and distributive justice and job stress however. Future research should investigate this peculiarity.
Table of Contents
Table of Contents
2
Chapter 1
Introduction
4
1.1 Research Objective & Contribution
5
1.2 Research Context & Method
6
1.3 Paper Outline
6Chapter 2
Literature Review
7
2.1 Reward System – the theory 7
2.2 Distributive Justice
10
2.2.1 Introduction & Definition
10
2.2.2 Theory
11
2.3 Procedural Justice
11
2.3.1 Introduction & Definition
11
2.3.2 Theory
12
2.4 Pay Satisfaction
13
2.4.1 Introduction & Definition
13
2.4.2 Theory
14
2.5 Job Stress
15
2.5.1 Introduction & Definition
15
2.5.2 Theory
16
Chapter 3
Hypotheses Development
18
3.1 Justice and Satisfaction
18
3.2 Justice and Stress
23
3.3 Hypotheses Summarized
28
4.1 Organizational Context
29
4.2 Sample
30
4.3 Reward system – the practice 31
4.4 Method
32
4.5 Measures
35
Chapter 5
Data Analysis
37
5.1 Construct Validity
37
5.2 Cronbach Alpha & Correlations
40
5.3 Descriptive Statistics
43
5.4 Assumption Testing
44
5.4.1 Normality 44 5.4.2 Linearity 45 5.4.3 Multicollinearity 475.5 Hypotheses Testing
49
5.5.1 Justice and Pay Satisfaction 49
5.5.2 Justice and Job Stress 52
Chapter 6
Conclusion & Discussion
56
Chapter 7 Implications & Recommendations
59
Chapter 8
References
61
1.
Introduction
In recent years there have been a lot of changes in the way reward systems are structured. Terms like the balanced scorecard, EVA, and stock option plans have become well known among the managers in the financial sector. The main goal of a reward system is to maximize the value of the company by aligning the interests of the employees with the goals of the company through rewards. One could imagine that when a company succeeds in rewarding their employees based on their expectations and needs; this will lead to a motivated workforce that is willing to accomplish the targets set by the management. On the other hand when the company fails to keep their employees satisfied with their rewards this will lead to demotivation and even disruptive behavior.
Keeping your workforce satisfied is not an easy job. And in the capitalistic environment of Western-Europe, keeping your workforce satisfied with their pay is even harder. The key to pay satisfaction can be found in the corner of justice theories. The justice theories help us explain how “employee reactions are affected by both the perceived fairness of the rewards they receive as well as the procedures used to determine the reward level”. (Sweeney and McFarlin, 1993) The prevailing justice theories that predict how employee reactions can be affected by a reward system are the distributive justice theories and the procedural justice theories. (Sweeney and McFarlin, 1993)
A company that designs its reward system has to be aware of the success factors and pitfalls that lie on its way. By paying attention to the justice theories a company’s management should be able to keep their workforce satisfied with their pay and thus motivated to perform, and prevent disruptive behavior like the reduction of effort, and job stress.
1.1
Research Objective & Contribution
The objective of this thesis is to be a descriptive as well as exploratory research about the effects of distributive and procedural justice on employees’ reactions in the context of a reward system. To be more specific this thesis will try to highlight how these two forms of justice can have a positive effect on employee reactions by influencing pay satisfaction but also how a lack of justice can lead to negative reactions in the form of job stress.
The goal is to find that distributive as well as procedural justice positively affects pay satisfaction where distributive justice plays a more important role in doing so than procedural justice. Furthermore the objective is to find that a feeling of injustice about rewards will lead to added job stress. By doing so the results of this thesis could guide a company’s management to redesign their reward system in a way that will satisfy and thus motivate their employees, and that will reduce job stress and therefore increase performance.
The above showed the objective of this thesis and the practical contribution the results could have on the proper design of a reward system for a company. In addition to the practical contribution this thesis also tries to contribute something to science. For instance the literature has shown us that perceptions of injustice can play an important role as a cause of added stress in an already stressful situation. However this research was done among nurses in a hospital and did not distinguish between distributive and procedural justice. (Zohar, 1995) This thesis will distinguish between the two forms of justice and uses another type of sample and environment. Research by Sweeney and McFarlin (1993) has shown that distributive justice is a good predictor of personal outcomes like pay satisfaction whereas procedural justice is a better predictor of organizational outcomes like commitment. However their research lacks an explanation about the theoretical prove behind these results. This thesis will therefore use a theoretical framework to found the hypotheses.
1.2
Research context & method
The research method is a questionnaire survey among the employees of a business unit called Fixed Income Capital Markets (FICM hereafter) at the head office of the ABN AMRO Bank N.V.. Based on an ‘Employee Engagement Survey’ done by ISR (2006), there is reason to believe that the employees in this particular business unit are disappointed with the reward system. According to the ‘Employee Engagement Survey’ the workers’ situation at this business unit is already pretty stressful because of the task difficulty and work overload, and therefore provides a fruitful environment for my research. The data was analyzed by means of a multiple linear regression and looked for correlations, slopes and the explanatory power between the variables.
1.3
Paper Outline
This thesis starts of with a chapter about the essence and nature of a reward system and continues with a description of the variables that found the hypotheses. The description will be given based on existing literature about: reward systems, distributive justice, procedural justice, pay satisfaction, and job stress, and based on theories like the expectancy theory, the equity theory, referent cognitions theory, and role dynamics theory.
The development of the hypotheses and an analysis of the theoretical framework will follow after these descriptive chapters. Then the thesis continues with a chapter about the research context and method. After that the construct and reliability of the measures are discussed. Hereafter follows the data analysis by testing the assumptions and running a linear regression. And finally there will be a conclusion, discussion, and a chapter about the implications of this research.
2. Literature Review
This chapter describes the theories concerning the variables that this research will use to develop the hypotheses. Every paragraph begins with a description of the variable and the history surrounding it. The theories and variables will be described in the following order: reward systems, distributive justice, procedural justice, pay satisfaction, and job stress.
2.1
Reward system - the theory
As previously mentioned, a reward system should be designed to align the interests of the employees with the goals of the employer, in order to maximize the value of the firm while keeping the workforce satisfied. You can look at this ‘incentive system’ from an economical and psychological/sociological point of view. This thesis will focus on the psychological/sociological point of view. One of the leading theories in this field is the expectancy theory by Victor Vroom (1964).
When an employer designs his reward system he tries to align his goals with the interests of his employees. He can do so by making formal and informal agreements about the rewards one will receive in exchange of effort or performance. The art of making the right agreements and thus motivate your workforce to exert effort can be explained by the expectancy theory. This theory predicts that employees will be motivated when they believe that: (Vroom, 1964)
• More effort will lead to more performance
• More performance will lead to higher organizational outcomes (e.g. salary or benefits) • The outcomes are valued by the employee
Furthermore this theory can help a company to maintain its workforce motivation by defining the incentive system as a product of expectancy, instrumentality and valence. The expectancy model of motivation by Atkinson (1997) contains the following main elements:
1 Effort. Can be defined as an employee’s contribution at his job in the form of
knowledge, ability, time and intensity.
2 Expectancy. Refers to the employee’s expectation and confidence about what is
achievable based on the amount of effort exerted.
3 Performance. Refers to how a performance measurement system judges the
contribution of the employee to the organization. The accuracy of this system determines whether an employee feels he is in control of his performance.
4 Instrumentality. Refers to the mechanism that converts ones performance to outcomes.
A good instrument that motivates its employees should be consistent, observable, and accountable.
5 Outcome. Is the total rewards package one receives based on his performance. In this
thesis the package consists of salary, benefits and bonus.
6 Valence. Is the attitude of the employee towards certain rewards. “The intensity of
desire for extrinsic (money, promotion, benefits, raises) and intrinsic (self-esteem, personal standards) rewards determines motivation.” (Wikipedia: Expectancy theory) This thesis will merely discuss extrinsic rewards.
Based on these seven elements you can form the following model:
Model 1 Expectancy model of Motivation
This model explains how reward systems can motivate and satisfy employees. Employees expect that with a certain amount of effort they can perform at a certain level. Then they assume that the company’s performance measurement system is able to record their performance and translate it with an instrument into an outcome. The employee validates what this reward is worth to him which determines his satisfaction.
An off balance in an element of this model can lead to a negative chain reaction in the rest of the model and thus dissatisfaction and maybe even stress among the employees. The balance of the elements depends on the perceptions of justice about the elements from the perspective of the concerning employee.
This thesis continues with a few theoretical chapters about the two main forms of justice that influence human behavior and could distort the chain of the model of motivation.
Effort Performance Satisfaction
Expectancy Valence
Outcome
2.2
Distributive Justice
2.2.1 Introduction & Definition
Traditionally the perception of justice and its effects on human behavior was being regarded as an outcome-oriented perception. Until the late eighties a lot of theoretical conceptualizations focused extensively on distributive justice as a cause of change in job satisfaction and performance.
Keep in mind that this thesis focuses merely on justice in a reward system and therefore does not include feelings of justice that are unrelated to the employees’ rewards. Shed in that light one could say that distributive justice focuses on “what the rewards are” whereas procedural justice focuses on “how the rewards are given”. “Distributive justice can be characterized as a person’s view of whether they received the outcome that was deserved, and is often judged by comparing the outcome to some referent”. (Huffman and Cain, 2001) The referents that determine someone’s feeling of justice can be interpersonal or intrapersonal. Interpersonal justice is the comparison of you to an almost identical employee which is explained by the equity theory (see below). The more you are treated equally the higher the feeling of justice. Huffman and Cain describe intrapersonal referents as follows: “when adjustments are made for task difficulty in sales, the unadjusted measure may be used to create an intrapersonal referent (i.e., the performance rating that would have resulted had no adjustment been made), thus, influencing assessments of distributive justice and satisfaction with the outcome”. (Huffman and Cain, 2001)
2.2.2 Theory
One of the most notable theories that focus on distributive justice is ‘the equity theory’ by Adams (1965). The equity theory states that “employees seek to maintain equity between the inputs they bring to a job and the outcomes they receive from it against the perceived inputs and outcomes of others”. (Adams, 1965) The inputs employees can bring to a job are for instance: qualifications, skills, experience and ambition. The outputs they can receive include: money, power, status, job variety and flexible working arrangements. As long as employees feel they are being treated fairly – especially compared to similar others – they will be motivated to perform. Therefore the ratio of an employee’s own outcomes and inputs should equal the ratio between a comparable employee’s outcomes and inputs. Furthermore the equity theory says that if an employee gets the feeling that there is an off-balance between the effort he puts in his work and the rewards he gets to do so, then he will likely react in a negative way. Some examples of the resulting behavior that originates from an off-balance are: demotivation to perform, reduction of effort, inward disgruntlement (like stress), recalcitrant and disruptive behavior.
The next variable to be described is procedural justice. Again we will discuss its definition and theory.
2.3 Procedural Justice
2.3.1 Introduction & Definition
The research area surrounding procedural justice has been expanding for the last two decades. Before that, organizational sciences about justice and the perception of fairness were
dominated by the equity theory. However since people started criticizing the equity theory and its lack of attention to the way rewards are distributed, new theories have gained ground.
The referent cognitions theory by Folger (1987) helped placing the causes of unfairness and injustice in a context broader than the equity theory, namely, “the role that decision-making procedures play in shaping perceptions of unfair treatment” was added to the existing theories. This gave rise to the variable procedural justice. “Whereas distributive justice focuses on the fairness of the distribution of resources, procedural justice focuses on the fairness of the procedures used to make those distributive decisions.” (Greenberg, 1987)
Procedural justice is characterized by resource distributions that are: “consistent across persons and over time, free from bias, based on accurate information, correctable, representative of all recipients’ concerns, and based on prevailing moral and ethical standards”. (Greenberg, 1987) If the allocation procedures fail to do so this will lead to extreme dissatisfaction, a feeling of unfairness, and adverse reactions.
An example of the need for procedural justice as a measure is the situation where two people respond differently to the inequity of the outcome/input ratio if they feel differently about how that inequity was created. A person who believes that the inequity was created through fair procedures will feel less dissatisfied than a person who believes the procedures were unfair. (Cropanzano, Folger, 1989)
2.3.2 Theory
The referent cognitions theory provides evidence for the use of procedural justice as a predictor of pay satisfaction. As stated above it adds to the equity theory that decision making procedures also play an important role in the perception of unfair treatment whereas equity theory mainly focus on their input/output ratio compared to similar others. Folger (1989)
captures the essence of referent cognitions theory in the following “would/should phenomenological account”. If you consider a situation where a decision maker has to allocate outcomes, then you can expect people to express the highest resentment and dissatisfaction when people believe they would have gotten better outcomes if the decision maker had used procedures that should have been used.
Since this thesis only discusses the two forms of justice in a reward system and since the rewards we discuss are merely extrinsic (salary, benefits, bonus), this thesis continues with a discussion about the definition and theory of the variable pay satisfaction instead of the comprehensive variable satisfaction .
2.4 Pay Satisfaction
2.4.1 Introduction & Definition
The personal outcome ‘pay satisfaction’ has been mentioned a few times in the past chapters. According to research done by for instance Sweeney and McFarlin (1993) pay satisfaction relates to distributive justice and to procedural justice. Furthermore this thesis regards pay satisfaction as the final element in the expectancy model of motivation, see figure 1. Therefore we can say that pay satisfaction relates to the expectancy theory, the equity theory and the referent cognitions theory. To show how these theories and variables interact I will give a short summary of the previous chapters.
According to equity theory of Adams (1965) people express dissatisfaction and react negatively when they perceive that the outcomes (rewards) of their effort are not meeting the
norms of justice that are largely based on referent others. The referent cognitions theory adds to this theory that people also evaluate their received rewards based on past events, implicit and explicit promises, and legitimacy of someone’s conduct. (Folger, 1987, 1993) The norms of justice that exist in past events, referent others and implicit/explicit promises can be divided into distributive and procedural justice. The expectancy theory shows that pay satisfaction is determined by the sequential and mediating interaction of six elements namely: effort, expectancy, performance, instrumentality, outcome, and valence. These theories combined show that pay satisfaction is largely determined by the interaction of six justifiable elements, where justice is based on norms that are determined by the equity theory and referent cognitions theory.
2.4.2 Theory
In his search to perfectly measure pay satisfaction, Heneman (1985) tried to build a pay satisfaction questionnaire based on his concept of pay satisfaction. His original concept of pay satisfaction argued that “pay in organizations exists in four relatively independent dimensions: level, benefits, raises, and structure”. Note that the first two dimensions are related to the distributive justice variable, whereas raises and structure are related to the process leading to rewards, or procedural justice. In his paper he studied the importance of these four dimensions of pay satisfaction and confirmed his hypothesis that “employees perceive, and experience differential satisfaction with the four dimensions: level, benefits, raises, and structure”. (Heneman, 1985) This is congruent with the existence of the element ‘valence’ in the expectancy theory. Every employee has its own preference for a kind of reward whether it is salary, benefits or bonus. Even more so every employee has its own preference for the way the reward system is structured whether he wants more subjective
measures, more growth opportunities or more performance feedback. Heneman found that the dimensions level, and benefits were highly and independently related to pay satisfaction.
In his paper, Heneman (1985) calls for “more research regarding likely relations between pay satisfaction and various dependent variables”. According to him there may be a possibility that a combination of dimensions of pay satisfaction may influence certain variables more than a single dimension does.
As mentioned before the expectancy theory helps explain how employees can get motivated. If one of the elements fails this could lead to dissatisfaction and maybe even stress. According to the equity theory employees compare their rewards to a similar employee and past events. If there is an inequity this can lead to demotivation to perform, reduction of effort, inward disgruntlement (like stress), recalcitrant and disruptive behavior. Therefore it is assumable that job stress will be an important aspect in the expectancy theory and equity theory.
The next chapter will introduce job stress and explain the theory behind it.
2.5 Job Stress
2.5.1 Introduction & Definition
In the fast pace environment of the last few decades job stress is a common problem. To attack this problem by its core it is necessary to understand the roots of job stress.
The variable job stress can be defined as “the harmful physical and emotional response that occurs when the requirements of the job do not match the capabilities,
resources, or needs of the worker”. (www.wfmh.org/wmhday/sec3_pt3_4_glossary.html) It can lead to a variety of subjective, behavioral, cognitive, physiological, organizational and health-related problems like depression, anxiety, excessive eating, hyper-sensitivity to criticism, increased heart rate, absenteeism and insomnia. (Weick, 1983)
2.5.2 Theory
The ‘controllability principle’ states that a person should be held accountable and should be rewarded for the things that he can control. When a person feels that he is rewarded for things that were out of his control then he will consider this to be unjustifiable or unfair. These ‘out of control situations’ can partly be described by the role dynamics theory.
The role dynamic theory explains why employees experience role stress. The role a person plays is determined by the expectations of others about appropriate behavior in a particular situation. (Kahn, Wolfe, Quinn, Snoek, Rosenthal, 1964) A work-family conflict is an example where the two roles a person is expected to play, conflict with each other because they are incompatible. (Katz, Kahn, 1978) These role conflicts can be time based, strain based or behavior based. (Greenhaus, Beutell, 1985) An example of a time based role-conflict in a work-family situation occurs when a father cannot pick up his child from daycare because he has to work late. Strain based conflicts occur for example when a father is anxious with his wife’s pregnancy and therefore cannot function well at his job. An example of a behavior based role conflict occurs when you are expected to be aggressive at your work but at the same time cannot show that behavior at home because that would result in conflicts with your family. Role conflict also occurs when your individual needs conflict with role demands. (Vasu, Stewart, Garson, 1990)
The second situation that influences roles stress according to the role dynamics theory is role ambiguity. Role ambiguity occurs due to a lack of clarity regarding duties, responsibilities, and/or authority. But it can also occur due to the complexity of the job. Usually role ambiguity is exhibited by less capable or less confident employees. The solution to this ambiguity lies in the simplification and/or more precise description of job demands.
Work-overload is probably the most well-known cause of role stress. Mikhail (1981) defines this situation as “a state which arises from an actual or perceived demand-capability imbalance in the organism’s vital adjustment actions and which is partially manifested by a nonspecific response”. This means that job stress originates from a situation of demand overload or underload and the adjustments a person can make to that situation. Demand overload or underload refers to a situation where quantity or quality demands by a superior do not match with what the subordinate thinks is achievable.
Low decision latitude is another stressor related to roles. An example of low decision latitude occurs for instance when a superior assigns tasks to a subordinate; tasks for which the subordinate does not have the authority to influence the outcome and therefore cannot meet the superior’s expectations, which subsequently leads to high stress levels. (Zohar, 1995)
Role conflict, ambiguity, work-overload, and low decision latitude are well known work-related stressors. There are more important work related stressors, as poor work relations, fear of losing your job due to reorganizations and so on. However after a discussion with a manager at FICM (the business unit researched) and after looking at the situation the ABN AMRO is in (possible take-over, fear of resentment), the stressors discussed above were considered to be the most important and less biased variables considering the relations to be investigated.
3 Hypotheses Development
In this paragraph the hypotheses will be developed based on the described theories and literature. Furthermore a practical example of the assumed relations is given in order to enhance the clearness of the hypotheses.
3.1 Justice and Satisfaction
The last few chapters discussed about the essence of a reward system and the definition of the variables within this setting that will be researched further on. By reading about these theories and literature it should have become clearer how these variables behave and possibly interact. This chapter starts off with some existing literature about distributive justice and further on develops the hypothesis.
A lot of research on equity theory was usually conducted through experiments. It was during these experiments that researchers showed the negative impact of inequity on satisfaction. (Adams & Rosenbaum, 1962; Garland, 1973) If people were aware of the fact that they were either underpaid or overpaid (compared to their co-workers) this would result in dissatisfaction, either through angriness or through guilt. During the states of inequity underpaid workers were found to be less productive and overpaid workers were more productive. However future research would show that pay satisfaction would not be influenced by interpersonal comparison alone.
Some theories of distributive justice ignore the interpersonal comparisons claimed to be so important by the equity theory. Jasso’s (1980) theory of distributive justice for instance focuses on “justice in terms of the comparisons people make between their actual share of goods and their beliefs about a “justifiable share””. When people feel they are underpaid
based on their capabilities, effort, and economical standards this will lead to negative reactions.
Sweeney and McFarlin (1993) tested the viability of four different models that each predicted different relations between the justice variables and personal and organizational outcomes. They showed that the so called two factor model had the best fit with the data. This model assumes that distributive justice is a more important predictor of the variance in personal outcomes like pay satisfaction, whereas procedural justice should be used to explain the variance in organizational outcomes like organizational commitment. Nonetheless they comment that different models might be viable under different conditions.
This means that future research should not focus on the assumption that one form of justice can only predict one form of outcome. The procedural primacy model for instance predicts a stronger relation between procedural justice and personal outcomes like pay satisfaction. And according to MacCallum (1986) ‘the additive model’ - that assumes that distributive and procedural justice can have an independent additive influence on personal outcomes – deserves further examination.
The above shows that distributive justice plays an important role as a predictor of various personal outcomes. In the context of a reward system, pay satisfaction is probably the most prominent analyzed personal outcome. Based on the past research, it would be interesting to see what direct influence distributive justice would have on pay satisfaction in a different work-environment than has the ones that have been researched before. Furthermore it will be interesting to see how distributive justice behaves in the context of an additive model. (Sweeney and McFarlin, 1993)
Based on the research above it is easy to assume a relation between distributive justice and pay satisfaction. It is more difficult however to theoretically found this assumption.
The theory about a reward system explains how a manager tries to reward his employees in a way that they become motivated to perform in line with the company’s interest. The psychological system of motivation by a reward system can be described by the expectancy theory. This theory supposes that through the sequential and mediating interaction of six elements, satisfaction originates (see figure 1). An off balance in these elements could lead to dissatisfaction. A manager should therefore keep a constant eye on all of these aspects. The off balance can be created by feelings of distributive and procedural injustice. Distributive justice is often described as to whether an employee feels he gets what he deserved. In our case we speak of justice in terms of pay (salary, benefits, and bonus).
There are several distributive norms that determine an employee’s pay satisfaction. These norms are based on inputs like experience, effort, stress, responsibilities and good performance. Notice that these inputs are all somehow part of the chain in the expectancy model of motivation; they represent for instance the elements ‘effort’, ‘expectancy’, and ‘performance’. When an employee thinks he does not get paid enough based on the inputs he brings to his job then we can speak of distributive injustice, which distorts the expectancy model of motivation, and therefore leads to pay dissatisfaction. This leads to the following hypothesis:
H 1.1 Distributive justice positively affects Pay Satisfaction.
In the previous chapter it was mentioned that distributive justice has been a good predictor of various variables like job satisfaction and performance. But research by Alexander, Ruderman (1987) has shown that procedural justice accounts for more variance in job satisfaction and organizational commitment than distributive justice. Even more so they
showed that the process of determining an employee’s salary accounts for more variance in job satisfaction than the level of salary.
Procedural justice is – compared to distributive justice - a relatively new perception of justice in the organizational sciences. The added positive influence of procedural justice on pay satisfaction has been extensively researched and confirmed. Important research done by Sweeney and McFarlin (1992) has shown that this influence is not as strong as between distributive justice and pay satisfaction.
When the two justice variables are being researched at the same time as independent variables we have to be mindful of possible interactive and correlative effects. This interaction can alter presumed independent relations that justice variables could have on a dependent variable. Therefore Greenberg (1987) and others have researched the interaction and correlation between distributive and procedural justice. For instance, a problem arising from procedural justice is that it can be a necessary precondition for distributive justice when outcomes (rewards) are low. If outcomes are medium or high then people will perceive the distribution to be fair regardless of the procedures used. But when the outcome is low people will start to complain if they feel that the allocation procedures were unfair. It is claimed that distributive justice can also influence the perception of procedural justice by the so called ‘egocentric bias’. This bias means that higher outcomes always positively influence procedural justice whereas lower outcomes do the opposite. Evidence for this relation between procedural and distributive justice is mixed.
Even though there are a few very good papers about the relation between distributive justice, procedural justice, pay satisfaction and other variables, many papers also conclude with a remark about the need for further investigation in the area of justice, pay satisfaction and other personal outcomes. (Greenberg, 1987)
The effects of procedural justice on pay satisfaction can again be explained by the expectancy model of motivation (see figure 1). If a company’s performance measurement system fails to correctly record its employees’ performance, or if the employees feel they are treated unfairly by the instrument that converts performance into outcome this will probably lead to dissatisfaction. Especially people in a prestigious environment feel the need to develop themselves and climb on the company’s ladder. That is why a lot of employees at dynamical financial company will probably feel the need for an objective and flawless promotion - , raise - , and performance feedback system. When managers use other procedures than they should have, employees will experience procedural injustice and dissatisfaction. Said from a reward system point of view, when wrong procedures lead to lower rewards (lower than the employees expected based on their inputs) this will lead to pay dissatisfaction. The literature and theory above leads to the following hypothesis:
H 1.2 Procedural justice positively affects Pay Satisfaction.
The two hypotheses above suggest that both distributive as well as procedural justice positively affects pay satisfaction. A manager should thus pay attention to both of those elements in order to keep the balance in expectancy model of motivation.
For a manager it would be interesting to know which of the justice variables contributes more to pay (dis)satisfaction so he could focus on the main cause of (de)motivation and (dis)satisfaction. From the expectancy model of motivation it is difficult to derive which justice variable has a greater impact on pay satisfaction. There are a few researchers though that investigated the differences in impact that both of the justice variables have on pay satisfaction. According to a paper written by Folger and Konovsky (1989):
“distributive justice accounted for more unique variance in pay satisfaction than procedural justice”. This in line with research by Sweeney and McFarlin (1993) which shows that although both the justice variables affect personal outcomes, procedural justice is a better predictor of organizational outcomes like organizational commitment and a worse predictor of personal outcomes like pay satisfaction.
This phenomenon can be explained by the fact that a high enough paycheck stands on top of the priority list for most employees and a fair procedural way to get to that paycheck comes second. Only when employees are dissatisfied with their level of pay, good procedures can temper this feeling of dissatisfaction. The findings above lead to the following hypothesis:
H 1.3 Distributive Justice affects Pay Satisfaction more than Procedural Justice
3.2
Justice and Stress
The previous chapter showed that feelings of justice (in relation to extrinsic rewards) will positively affect pay satisfaction. This could be explained by existing literature and through the expectancy model of motivation. A good manager can motivate its employees by paying attention to the elements that construct this model. If the model works fine people will feel pay satisfaction and motivation to perform. Can we assume the opposite when the model is disturbed by feelings of injustice? We already assumed that injustice leads to pay dissatisfaction but is it safe to go even further and conclude that injustice leads to other negative personal outcomes like job stress?
To give the reader some insight in the written literature about the causes of job stress a short literature summary of the papers that contribute something to the assumptions that will be made further on about the relation between justice and stress is included below.
Narayanan (1999) researched the possible relations between job stress and gender and occupations. In his research the variable ‘control’ is considered to be an important cause of stress. If we talk about a ‘lack of control’ you can imagine that this can be caused by one of the earlier mentioned work related stressors. For instance low decision latitude gives an employee not enough saying power to influence his own behavior in the way he would have liked, which results in stress. Narayanan (1999) finds that employees in the lower level jobs suffer from more stress due to a lack of control than the higher level academics. Regarding gender he finds that in a gender integrated occupation there is less overall stress due to a lack of control. This means that women will feel less stressed in an environment where they are regarded ‘equal’ if there is a lack of control. Another interesting point is that all the three occupational groups examined namely, clerical, academic, and sales, reported interpersonal conflict, time wasters, and work overload as sources of stress, whereas role ambiguity and role conflict were very rarely mentioned.
Schaubroeck and Merrit (1997) showed that the level of self-efficacy, or ‘the way you find yourself capable of doing something’, influences the relation between control and job demands. People with demanding jobs and that are confident with their abilities will feel less stress due to an increase in control. For people with a low level of self efficacy a combination of high demands and high control will lead to more stress.
In chapter 3.1 about procedural justice and pay satisfaction it was mentioned that according to an article written by Shields, Deng and Kato (2000) an increase in standard based incentives will lead to a reduced stress level due to the increase in arousal and
performance capabilities. Furthermore they found a positive relation between standard tightness and stress, and a negative relation between participative standard setting and stress. When standards are set too tight, employees will find it more difficult to live up to the expectations of his superior. This of course will lead to an increase in job stress due to work overload if the superior will not adjust the standards. If employees are not participating during the budget making process, they will feel less in control of the standards that they have to live up to. This lack of control due to low decision latitude will lead to an increase in job stress. It is likely that employees who were not involved in the budget setting process will feel procedural injustice when they will be negatively evaluated by year end, which will lead to added stress.
As you can see a lot of the causes of job stress can be explained by the employee’s ability to control his own performance. The known work related stressors like role conflict & ambiguity, work-overload, and decision latitude can lead to feelings of distributive injustice because employee can lose control of their performance and rewards. These stressors can bring each of the first three elements of the expectancy model of motivation off-balance. This off-balance will lead to uncontrollable performance and outcomes. The fact that the employees do not control their rewards can lead to distributive injustice seen in the perspective of the equity theory (see chapter 2.1.2) The off-balance in the expectancy model of motivation by the loss of control and distributive injustice will most certainly lead to stress. Below a few examples will be described that will lead to a better understanding of the interactive relation between work related stressors, a loss of control, distributive injustice and job stress.
In a situation of role conflict you can imagine that someone that had just gotten his first kid needs to be able to pick them up from daycare once in a while. If the employer does
not recognize this situation and keeps comparing him with the rest of his colleagues then the equity ratio will be off-balance because he did not get the flexible working arrangements that comparable colleagues with kids had gotten. This feeling of distributive injustice resulting from role conflict leads to added stress because on top of his role conflict situation he now has lost control of his rewards.
When a superior does not consider or recognizes that he is distributing outcomes to an employee facing role ambiguity then he is likely to compare him with an employee with the same tasks that does understand his role. The employee facing role ambiguity will be unfairly rewarded for tasks that are out of his control and understanding while he could have received the same rewards as his comparable employee if the employer had just recognized the situation of role ambiguity. This will result in an off-balanced equity ratio and thus a feeling of distributive injustice. Again a feeling of distributive injustice, but now resulting from role ambiguity, will lead to added stress.
When an employee faces work-overload he hopes that his superior recognizes this situation and will reduce the work-demand. If the superior does not reduce the work-demand he will continue his neglective path of destruction by negatively evaluating the subordinate’s performance. (Shields, Deng, Kato, 2000) This out of control situation will be unjustifiable for the employee. The feelings of injustice will lead to resentment and added stress.
These examples and the theory described in chapter 2.1 and chapter 2.4 lead to the following hypothesis:
Looked upon from the context of the expectancy model of motivation one could say that the same story goes for procedural justice as was the case for distributive justice in relation to the ‘controllability principle’ and job stress. The only difference is that the loss of control caused by the work related stressors and sequentially by procedural injustice can now be explained by an off-balance in the third and especially the fourth element in the expectancy model of motivation. If the procedures or the instruments leading to the outcomes of an employee fail or become unjustifiable this will most likely lead to an uncontrollable situation for the employee which will lead to increased job stress.
Again a few examples will be given of how procedural injustice could lead to job stress. Imagine a same situation of role conflict where an employee - in order to become eligible for promotion - is required to work at least eight hours a day and everyday between nine and five. This would mean that when an employee has to get his child from daycare during the day hours he cannot become eligible for promotion, even if this employee catches up with his lost hours by working at the office till late.
In the case of role ambiguity an employer should consider the fact that an employee that works at a new department does not fully understand his task description because he needs some adaptation time and needs more explanation by the managers. If the employer ignores this situation and evaluates his total performance based on cost savings while the employee might think that he had to get high sales, then it will lead to a feeling of injustice.
These examples and the theory about procedural justice and job stress lead to the following hypothesis:
3.3
Hypotheses summarized
To give the reader an overall view of the theoretical framework and the hypotheses that are to be researched, the following figure was constructed:
+ + _ _
Figure 1 Theoretical Framework
This model is based on the additive model as described in the paper of Sweeney and McFarlin (1993). In contrast to their model however I did not include a path between the two dependent variables. This is because of a lack of causal prove and because an evaluation of this unlikely relation goes beyond the scope of this paper.
There has been support for the use of an additive model by different authors. (Lind, Kurtz, Mustante, Walker, & Thibaut, 1980; Sheppard & Lewicki, 1987) Although Sweeney and McFarlin found that this model did not result in a better fit with the data compared to the so called two factor model, I believe that in the context where we have two personal outcomes as dependent variables (Folger and Konovsky 1989, found that distributive and procedural justice both independently add to the prediction of the personal outcome pay satisfaction), and
Procedural Justice
Distributive Justice Pay Satisfaction
where the literature provides answers to a lot of our questions regarding the assumed relations, the additive model is the appropriate one to use.
4
Research context & method
The next chapter describes the organizational context of my research. The setting is a particular business unit, called FICM, at ABN AMRO N.V.. Then the thesis continues with a description of the sample and the reward system. After having done so an outline will be given of the research method. Finally this chapter ends with an elaborate explanation of the constructed measures.
4.1
Organizational context
The organizational context for this study is ABN AMRO Bank N.V. with her roots in Europe and a history going back to 1824. “It has over 4500 branches in almost 56 countries and territories, a staff of about 107.500 full-time equivalents worldwide and total assets of EUR 987 billion as of year-end 2006”. (Annual Report, 2006) Furthermore they are listed on the Euronext and the New York Stock Exchange. The bank focuses on their consumer and corporate clients in Europe, North-America, Latin-America and Asia.
The growth-strategy of ABN AMRO N.V. for 2007 is to expand their strong position in the middle segment of the market. They will try to do so by offering high quality and innovative bank products and services. The recent acquisition and integration of Antonveneta, an Italian bank, has lead to a desire and expectation to penetrate the promising Italian market.
During the development phase of this thesis, The ABN AMRO Bank N.V. has been in the midst of takeover attempts by Barclays and a consortium existing of Banking Santander, Fortis and Royal Bank of Scotland.
4.2
Sample
The respondents are drawn from a business unit known as Fixed Income Capital Markets. This business unit specializes in offering financing solutions with capital market products, like bonds, private placements, securitizations and tax-efficient structures. It is a dynamic work environment, where words like leadership, innovation, international, focus and ‘going the extra mile’ prevail. There are currently 401 employees working at FICM. FICM can be divided by a few other business units. I have excluded the international business units from my sample since the variable performance contingent compensation is likely to be influenced by variables like culture and environment. I did send the remaining (62) Dutch employees an online questionnaire.
These employees receive bonuses, long term incentives and base salary. According to the statements about the reward package of the Managing Board and according to an interview with a Managing Director at the Business Unit Fixed Income Capital Markets (FICM hereafter), the main purposes of the reward package is to motivate the employees to exert effort and to retain the highest quality employees.
The average level of job stress at FICM is high compared to other Business Units at the bank. (Survey, 2006) This is largely due to the dynamic and demanding work environment and maybe due to the way performance is linked to compensation.
4.3 Reward system – the practice
The expectancy model of motivation explains how people can be motivated by a good working reward system. An oversight of the reward system at FICM will be given, to create a clear picture of the status quo of our research area.
The employees at FICM receive extrinsic rewards in the form of salary, bonus, and benefits. The salary level of an employee at the ABN AMRO N.V. is determined by the salary scale he is in. The progression an employee makes on a scale or from one scale to another is determined by working years, responsibilities, experience, and performance. An example of an such a salary model is given below.
Performance Rating Scale 1 Promotion Scale 2
# Y E A R S
Model 2 Salary and Promotion model
The salary levels in this model are not representable for FICM. An employee in this model starts his job with a market conform salary of 30k. At year end his manager evaluates his performance based on financial and non-financial targets. During this evaluation conversation the manager gives a rating for the overall performance of the employee. These ratings range from 1 (terrible) to 5 (excellent). The employee in this example received a ‘3’ performance rating after 1 year of employment. He will therefore advance one level downwards on salary
1 2 3 4 5 1 30k 30k 30k 30k 30k 2 31k 32k 33k 34k 35k … 9 39k 40k 41k 42k 43k 1 2 3 4 5 1 43k 43k 43k 43k 43k 2 44k 45k 46k 47k 48k … 9 51k 52k 53k 54k 55k
scale 1 and receive 33k in year 2 (performance rating = horizontal axis, # years of employment = vertical axis). When this employee has outgrown his function in terms of experience, responsibilities and other several personal characteristics, he will become qualified for promotion. The opportunity to promote is usually given during the evaluation conversation at year end. A promotion will put the employee in the next salary scale with new opportunities to grow.
The employees at FICM earn two kinds of bonuses. First of all they receive a normal bonus on top of their salary ranging from 5 to 15 percent of their base salary. The amount of this bonus is determined based on the predominantly objective performance rating an employee receives at year end. On top of this bonus the employee at FICM can also earn a ‘discretionary’ bonus which can vary from 1 to 40 months salary. The bonus pool of these ‘discretionary’ bonuses largely depends on the firm’s and department’s performance and is distributed to the employees based on a more subjective performance evaluation by their managers.
The sort and amount of benefits that employees receive at FICM is largely determined by the salary scale they are in. These privileges range from a larger lease car to flying business class instead of economy during business trips.
4.4
Method
From the sixty-two online surveys send to the employees of FICM, forty-one were fully filled in. This is a response rate of sixty-six percent. These surveys were gathered in a period of three weeks, after sending two email reminders and finally after a personal reminder by a manager at FICM.
The data received from the surveys will be analyzed in SPSS 15. This data does not include information about demographics. The main reason for the exclusion of demographics is that it would increase the amount of survey questions by such a number it would become too big a burden on the respondents.
The analysis of the data will start off with taking a closer look at the construct validity. This is a method to check whether a construct used to measure a variable actually measures the concept to be researched. The validity can be checked by looking at the convergent validity and the discriminant validity of the items that belong to a construct. Both validity checks are done by a factor analysis. According to Victor Maas (2007) factor analysis comes down to: “assessing the communality in item variance from the viewpoint that this variance is in part caused by an unmeasured variable that has an effect on all the items”. The communality measures the percent of variance in a given item explained by all the factors combined and can be interpreted as the validity of the item. Both the communality and factor loading of an item should be at least > 0.5 to be considered practically significant. (Hair, 1998) If all the items of a construct measure one factor with a loading of > 0.5, then there is convergent validity. Discriminant validity is measured by comparing all the items of all the constructs to see whether items of a certain construct belong to that construct alone. An item of distributive justice that would load high on the factor that represents job stress probably fits in better with the job stress construct.
An important step in determining the reliability of the construct of the variables can be taken by measuring the so called ‘Cronbach Alpha’. This is a well known reliability test to see whether the items or questions in the survey measure the same concept. For instance when combining five questions about distributive justice into a Distributive Justice variable it is important to know that all the items in fact measure distributive justice and not procedural
justice. As a rule of thumb, a Cronbach Alpha should be 0.5 or higher in order for a construct to be valid, and higher than 0.7 is preferable. If the Alpha of a variable proves to be below 0.7, measures will be taken by removing items that do not represent a concept well enough. By doing so, the constructs will end up being as close to 0.7 as possible. A reliability analysis in SPSS shows which questions you can remove from the summated scale in order to improve the Cronbach Alpha. In the same chapter the descriptives of the variables will be analyzed by looking at the mean, maximum and minimum, standard deviation, and correlation between them.
The third step in the data analysis will be a test of the assumptions. This test can be seen as a preparation for the linear regression and tells you beforehand what problems you could expect. This is done by checking for normality, checking for linearity, and finally a check for multicollinearity between the independent variables.
The final step in the research is testing the hypotheses. This is done by a multiple linear regression. An analysis of the coefficient of determination and the slopes of the regression models should provide enough evidence to prove the significance of this thesis’ hypotheses. The coefficient of determination or ‘R-squared’ represents how well the regression model approximates the data points that were gathered from the sample. A value of 1 indicates a perfect regression line whereas values below 0.1 are seen as insignificant values indicating a weak model. The slopes of the regression model determine the direction of the linear relations.
4.5
Measures
All the items used in the questionnaire survey are on a 5-point Likert response ranging from 1 (fully disagree) to 5 (fully agree).
The distributive justice index from Price and Mueller (1986) is used to measure the Distributive Justice variable. The original index consists of six items that all deal with the question if employees have been fairly rewarded given five general factors: experience, effort, stress, responsibilities and good performance. The term rewards in this index refers to money, praise, recognition and resources. Just as Sweeney and McFarlin (1993) did, the statements for this index were reduced to five. An example of a statement is: “Given the responsibilities that I have now, I have been rewarded fairly” Whereas the individual answers will be treated as ordinal values, the five statements will be combined to a summated scale which will be called the “Distributive Justice Variable” (DJ hereafter) with a range between 1 and 5.
Just as with the distributive justice index, the same procedural justice index was used as Sweeney and McFarlin (1993) did. These statements include asking about the fairness of four general procedures: communication of performance feedback, determination of pay raises, evaluation of performance, and determination of promotion. An example of a statement is: “the procedures used to evaluate my performance are fair”. The four ordinal questions will be combined to the “Procedural Justice Variable” (PJ hereafter) ranging from 1 to 5.
Pay satisfaction will be determined by making three almost identical statements about rewards. These rewards are: benefits, salary, and bonus. Each statement will ask the respondent about his satisfaction with a certain reward. An example of a statement is: “I am satisfied with my current bonus” The total “Pay Satisfaction Variable” (SF hereafter) will range from 1 to 5.
The variable Job Stress will be divided into Role Conflict, Role Ambiguity, Work-Overload, and Low Decision Latitude. These variables will be measured almost the same as Zohar (1995) did. This means that Role Conflict and Ambiguity will be measured by a selection of the loading items on the respective factors in the Rizzo, House and Lirtzman (1970) scales. In consultation with a manager at the FICM department, those items from the Rizzo, House and Lirtzman (1970) scales were picked that have the highest factor loadings and are most representative for the work-environment at FICM, (Factor Analysis helps clustering a group of statements that measure the same concept into a factor. If a statement has a high factor loading it means it is suitable to be used for that particular factor.) These handpicked items were combined into four statements about Role Conflict and Ambiguity
Work-overload and Decision Latitude will be measured by combining several items used by Karasek (1979). Karasek uses eight statements to measure Decision Latitude and seven to measure Work-Overload. Four of the Decision Latitude items used by Karasek were combined into two statements and five of the Work-Overload items into three statements. The selection and combination of these items were also done in consensus with the manager at FICM.
The four variables namely “Role Conflict & Ambiguity”, “Decision Latitude”, and “Work Overload” will again be used to create a summated scale. This is a scale that is formed by combining several variables into a single measure. This summated scale - divided by the number of statements used - represents the “Job Stress Variable” (JS hereafter) which ranges from 1 to 5.
5.
Data Analysis
The analysis starts off with a chapter about the construct validity. The convergent and discriminant validity will be assessed by a factor analysis. Chapter 5.2 will discuss the Cronbach Alpha, and the correlation between the variables. The next chapter will discuss the descriptives of the variables. After the assessment of the descriptives, chapter 5.4 continues with a test of the assumptions by looking at normality, linearity, and multicollinearity. And finally chapter 5.5 will test the hypotheses by running a multiple linear regression.
5.1
Construct Validity
The content validity of constructs deals with the question whether a measure represents all the aspects of a concept. In this thesis for instance job stress is constructed by four different variables namely, role conflict, role ambiguity, decision latitude, and work overload. The constructs used in this thesis do not comply with content validity rules. In order to have content validity a panel of experts has to judge whether an aspect is essential/useful and if important aspects were left out. The process of assigning experts to evaluate the content validity of the constructs goes beyond the scope of the thesis and abilities of the writer. However in a certain way there is some degree of content validity since a manager/expert of FICM has been consulted during the development phase of the constructs. The process of determining content validity in this thesis is more alike a subjective assessment based on theory and therefore the constructs do comply with ‘face validity’.
Convergent validity assesses if a construct is related to what it theoretically should be. A principal factor analysis will determine whether all the items used to construct a measure are helpful in making the variable theory proof. As said in chapter 4.3, a factor loading of > 0.5 makes an item statistically significant for that factor. Furthermore a single factor is
preferable; this would mean that a construct measures one concept. An analysis of the principal factors shows that each of the variables DJ, PJ, and SF consist of one factor, and all the items have loadings higher than 0.5. These variables are therefore convergent validated. Each variable measures the concept it theoretically should do and all the items included are important. The table below shows that JS or Job Stress is constructed by four factors and that one item - namely ‘JS – Hard/Fast’ - loads lower than 0.5 on any of the factors. Every factor represents each partial variable that was used to construct JS. Factor 1 represents Role Conflict, factor 2 –Work Overload, factor 3 – Role Ambiguity, and factor 4 – Decision Latitude. The item JS – Hard/Fast does not belong to any of these factors and therefore should be excluded from the construct. It could be argued that each of these factors and its assumed relations with the independent variables should now be individually researched instead of combining the factors into one construct. However Kim and Mueller (1978) state that a factor should at least consist of three to four items in order to be meaningful. This is not the case here and therefore the original construct is maintained throughout this chapter.
,81 9 ,12 7 ,16 2 -,16 3 ,77 0 -,09 6 ,04 0 ,08 4 ,31 5 ,02 8 ,74 8 ,22 6 ,10 3 ,39 0 ,72 4 -,25 6 -,12 1 ,12 2 -,06 7 ,77 8 ,07 2 -,08 6 ,18 4 ,77 8 ,38 2 ,37 2 -,59 2 -,26 0 -,26 1 ,81 3 ,19 1 -,13 0 ,31 3 ,75 2 -,06 2 ,29 3 JS-Reso urces JS-Rule Polic JS-JobDescrip JS-Diffe rently JS-Crea tivity JS-Decisions JS-Hard /Fast JS-Work JS-TaskConfl ict 1 2 3 4 Co mpon ent
Discriminant validity measures whether items belong to one construct alone and not interfere with other constructs. A construct should correlate low with other constructs in order to measure theoretically different concepts. Table 2 shows that this was not the case for DJ and SF. An analysis of the factor loadings indicate that the DJ construct and SF construct are much alike. However as a rule of thumb the loadings of constructs should be below 0.85 in order to be acceptable. Component 1 shows loadings higher than 0.5 for DJ and SF, but lower than 0.85 for SF, indicating discriminant validity and thus an acceptable level of intercorrelation between constructs.
,79 0 ,26 7 -,1 85 ,02 9 ,01 2 -,0 37 ,80 4 ,27 0 -,2 59 -,0 53 -,0 29 ,15 7 ,79 8 ,03 7 ,16 2 -,0 75 ,27 6 ,07 6 ,87 5 ,06 6 -,0 07 -,1 48 ,01 1 -,0 01 ,79 0 ,06 0 ,05 5 -,0 07 ,02 0 -,3 37 ,06 4 ,72 3 -,3 04 -,0 50 ,35 5 -,1 38 ,18 5 ,84 8 -,0 85 -,2 77 -,0 06 -,0 04 ,28 3 ,80 1 -,1 31 ,07 8 ,06 6 -,0 12 ,25 9 ,70 1 ,37 1 ,01 6 ,33 5 -,2 25 ,66 3 ,10 7 -,1 37 -,3 08 ,30 7 -,2 86 ,72 9 ,23 2 ,07 1 -,0 67 -,0 80 ,02 5 ,64 3 ,08 7 ,11 9 -,1 31 ,54 5 -,1 23 -,1 28 -,3 79 ,02 1 ,65 8 -,0 55 ,20 1 -,1 31 ,07 2 -,0 25 ,92 6 -,0 78 ,00 0 -,0 38 -,1 23 ,08 6 ,11 1 -,0 81 ,93 4 -,1 16 -,1 65 ,79 0 ,15 9 ,10 2 ,39 1 ,03 7 -,1 02 ,74 4 -,1 76 -,4 06 -,1 67 -,0 59 -,2 64 ,16 2 ,10 4 -,8 18 ,05 5 DJ -Resp onsb DJ -Expe rienc e DJ -Stress DJ -Effort DJ -Prod uctiv PJ -PayRaise PJ -PerfE valu a PJ -Prom otio n PJ -Feed back SF -Ben efits SF -Bon us SF -Sala ry JS -Reso urce s JS -Rule Polic JS -JobDescri p JS -Diffe rentl y JS -Work JS -T askConf lict 1 2 3 4 5 6 Co mpon ent
Table 2 Rotated Component Matrix All Variables
To summarize, the analysis of the constructs shows some minor validation issues. The variables were constructed by means of face validity. Convergent validity proved to be a problem because JS is constructed by four different concepts. However the original construct
for JS will be maintained due to the extraction of too many insignificant factors. Discriminant validity exists because factor loadings between constructs did not prove to be problematic.
5.2
Cronbach Alpha & Correlations
The Cronbach Alpha tells us whether the items used to construct a variable in fact measure the same concept. If not this could mean we have to either adjust or discard an item After checking the variable reliability this chapter continues with the descriptive statistics of the variables which will give us a better picture of the mean, standard deviation and maximum and minimum of each variable. And finally the correlation coefficient indicates the strength and direction of a linear relationship between two random variables. The results of our tests are shown in table 3, table 4, and table 5.
An analysis of the Cronbach Alpha will show us if we are using the right statements to measure a concept. Alpha’s higher than 0.7 are preferable and higher than 0.5 necessary in order to have a minimum level of reliability in the multiple linear regression. The diagonal of table 3 shows the Cronbach Alpha’s of the constructs. The high Alpha’s of DJ and PJ, 0.8955 and 0.8475 respectively, indicate that the statements that were used to construct these variables are indeed very valid and internal consistent. Although SF’s Alpha (0.7669) is considerably lower it is still more than acceptable. The Alpha of JS (0.4992) however is well below the preference level of 0.7. This can partially be explained by the fact that JS was constructed by four different elements of job stress. However Zohar (1995) did not encounter construct reliability problems in his research about job stress. Even though he used an almost similar construct for JS showing Alpha’s of 0.8 and higher for all four factors of JS.
,89 55 0 ,4116 ** 0,79 60** -0 ,3107 * 0 ,8475 0,45 75** -0 ,5425 ** 0,76 69 -0 ,4676 ** 0 ,4992 Distrib J ustice Pro ce Ju stice Pa y Sat isfact Job Stre ss
Distrib J ustice Pro ce Ju stice Pa y Sat isfact Job Stre ss
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
Table 3 Correlations and Cronbach Alpha (found on the diagonal)
The lower threshold of 0.5 for the Cronbach Alpha needs to be met or exceeded in order for the JS variable to be reliable enough to use in the data analysis. The Cronbach Alpha can be improved by leaving out certain items, however “just throwing out items until you have a maximum alpha is not considered good practice”. (Maas, 2007) Table 4 on the next page indicates which items to discard in order to reach a maximum Alpha. The item ‘JS-Hard/Fast’ was discarded first, leading to an Alpha of 0.53. Recall that ‘JS-‘JS-Hard/Fast’ did not load high enough on any of the four factors in table 1, and therefore construct validity will not be harmed by the exclusion of this item. A further increase of Alpha was reached by the exclusion of two items representing the Decision Latitude part of job stress. No validity issues are expected by the exclusion of these two items since they represent a factor on its own.
A new Correlations and Cronbach Alpha table is presented on the next page with the adjusted variable Job Stress after deduction of three items. The Alpha of Job Stress has increased from 0.499 to 0.588, which is still far below the 0.7 preference level but well above the lower threshold. Therefore the construct is reliable enough to measure job stress.
20, 3171 12, 522 ,35 0 ,42 1 21, 1220 13, 210 ,23 8 ,46 2 21, 0732 13, 020 ,33 8 ,43 1 20, 3415 12, 530 ,27 9 ,44 5 21, 1707 14, 845 ,05 1 ,52 1 20, 9512 14, 298 ,09 1 ,51 3 18, 9512 15, 348 ,00 1 ,53 0 19, 5854 14, 049 ,15 5 ,49 0 20, 4878 11, 456 ,44 0 ,37 8 JS-Reso urces JS-Rule Polic JS-JobDescrip JS-Diffe rently JS-Crea tivity JS-Decisions JS-Hard /Fast JS-Work JS-TaskConfl ict Sca le M ean i f Item De leted Sca le V arian ce if It em Delete d Co rrecte d Item -To tal Co rrelat ion
Cro nbac h's A lpha if It em Delete d
Table 4 Cronbach’s Alpha if item deleted
The correlation coefficient indicates the strength and direction of a linear relationship between two variables. It can range between -1 and 1, where -1 stands for a perfect negative correlation. In psychological research correlations between 0.1 and 0.29 are considered small, between 0.3 and 0.49 medium, and between 0.5 and 1 large. Table 5 indicates that both the justice variables have a positive and strong linear relationship with SF. The negative correlation between the justice variables and JS is considerably weaker.
,89 55 0 ,4116 ** 0,79 60** -0 ,2930 * 0 ,8475 0,45 75** -0 ,4634 ** 0,76 69 -0 ,3942 ** 0 ,5880 Distrib J ustice Pro ce Ju stice Pa y Sat isfact Job Stre ss
Distrib J ustice Pro ce Ju stice Pa y Sat isfact Job Stre ss
** Correlation is significant at the 0.01 level (2-tailed).
* Correlation is significant at the 0.05 level (2-tailed).
