Amsterdam Business School
Does employee’s involvement influence likelihood of using
performance measures by managers?
Name: Yue Liu
Student number: 10985867 Date: 17.6.2016
Word count: 13,233
Supervisor: Sjors van der Heide
MSc Accountancy & Control, specialization Control
Statement of Originality
This document is written by student [Yue Liu] 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
This master thesis aims to figure out the relationship between involvement of employees and likelihood of using performance measures by managers, with a mediation factor of quality of performance measures, under the assumption of the existence of information asymmetry in organizations. This thesis firstly assumes that employee’s involvement in developing performance measures can improve quality of those performance measures, and then make managers are more likely use the improved performance measures with a purpose of evaluating and motivating their employees. This thesis uses the method of Baron and Kenny (1986) with a statistic analysis tool of SPSS to run the data obtained. The two relationships abovementioned are confirmed with statistic analysis results. However, the direct relationship between involvement of employees and likelihood of using performance measures by managers is not confirmed. With all the results, this master thesis makes several contributions to both literature and practical aspects. For literature aspect, the results of this thesis supplement what has been missed by previous literature. For practical aspect, this thesis provides several alternatives to improve the performance measurement system of an organization.
Key words: Performance measures, information asymmetry, involvement of employees, management accounting.
Content
Chapter 1 Introduction ... 5
Chapter 2 Literature Review and Hypothesis Development ... 8
2.1 Literature Review ... 8
2.2 Theory and Hypothesis Development ... 10
2.3 Terminology ... 14
2.4 Control Variables ... 17
Chapter 3 Research Methodology ... 19
3.1 Research Design ... 19
3.2 Sample Selection ... 21
3.3 Variable Operationalization ... 22
3.4 Responses Analysis ... 25
3.5 Statistical Analysis ... 27
Chapter 4 Data Analysis ... 29
4.1 Reliability Analysis ... 29
4.2 Convergent and Discriminant Validity Testing ... 30
4.3 Preliminary Analysis ... 34 4.4 Hypotheses Testing ... 37 4.5 Additional Checks ... 41 Chapter 5 Discussion ... 43 5.1 Explanations of Hypotheses ... 43 5.2 Contribution ... 45
5.3 Limitations and Suggestions ... 48
Appendix A. Items Used in the Scales ... 52
Appendix B. The Distribution of Employee’s Age ... 53
Chapter 1 Introduction
In management accounting field, the performance measurement system is an important research topic, due to the agency problems caused by the delegation of responsibility to the lower lever by top management within a company (Eisenhardt, 1989; Lambert, 2007). During the period of the last decades, some previous researches have found that involvement of employees when designing the performance measurement system has a positive influence on the implementation of the performance measurement system as well as the performance of the company (Abernethy and Bouwens, 2005; Groen, 2012; Moers, 2006; Wouters and Wilderom, 2008). However, not many studies have taken a mediation variable, which may significantly influence the relationship between involvement of employees and implementation of the performance measurement system, into consideration. This master thesis mainly studies the relationship between involvement of employees and likelihood of using performance measures by managers, under the assumption that lower level employees have a good understanding of their job (Atkinson, 2012). The research study is designed to regard performance measures from various industries as well as countries as the integrated inputs for profit organizations. Besides, based on previous studies, quality of performance measures is important and is likely to influence the relationship abovementioned significantly (Groen, 2012; Moers, 2006). Therefore, in this thesis, I introduce quality of performance measures as an important mediation variable to differentiate this thesis from most of previous researches, and contribute empirical knowledge in the field of management accounting.
The research question of this thesis is to study the relationship between the employee’s involvement in performance measures and the likelihood of using these performance measures by managers.Three questions can be asked to show the three stages of analyzing the research question. Does employee’s involvement positively influence the likelihood of using performance measures by top managers? Does employee’s participation in the development of performance measures improve the quality of performance
measurement? And does the improved quality then increase likelihood of using performance measures by managers for evaluation and incentive purposes?
The idea of the research question comes from the previous study conducted by Groen in 2012, which studied the enabling effects rather than the control functions of the performance measurement system under the circumstances of employee’s participation (Groen, 2012). Groen’s study showed employees were motivated to perform better when they developed the performance measures together with their managers (Groen, 2012). However, her study paid less attention to the influence on the likelihood of using those performance measures by managers for evaluation and incentive purposes. Therefore, the study in this thesis does not focus on the enabling effects of employee’s involvement of performance measures but focuses more on how the direct managers react to the involvement when using those performance measures for evaluation or motivation. When it comes to control functions of the performance measurement system, managers who use performance measures for evaluation or motivation tend to consider the quality of performance measures more important to see whether the performance measures can provide information related to decision making (Moers, 2006). However, previous literature only showed that enabling the lower level to participate in the design of the performance measurement system is likely to reduce measurement mistakes caused by agency problems (Abernethy and Bouwens, 2005), and to bring more relevant information to the higher level who makes decisions about performance evaluation or motivation (Hofstede, 1967). It did not consider the potential influence which can be brought about by the quality of performance measures. Therefore, this thesis considers quality of performance measures as a mediation variable to show the influence on the control functions of the performance measurement system. Furthermore, most of the previous researches only took the lower level in the company as research sample (Abernethy and Bouwens, 2005), rather than the non-managerial employees, such as operational employees in the production department. Also, they all had problems with industry limitation or geography limitation, since they only analyzed the relationship within one specific industry
in one country. However, my research is designed to focus on the involvement of the lowest level employees, as well as to involve different industries and countries to mitigate the limitations. Therefore, in conclusion, the aim of research in this thesis is to help profit organizations among different industries all over the world to solve agency problems between the lowest level employees and their direct managers when involving them into the development of performance measurement system with the consideration of quality of performance measures as an important mediation variable, and to mitigate limitations of previous studies abovementioned.
The research question model of this thesis contributes to both literature aspect and practical aspect. For literature aspect, the statistic analysis results of the hypothetical model contribute solid evidence to the relationships studied in previous literature in the field of management accounting, especially the study of performance measurement system. The second contribution of this thesis to literature aspect is that it mitigates geography limitation by analyzing performance measures of organizations in different countries. It also eliminates industry limitation by providing a big picture of performance measures among various industries, rather than only analyze performance measures in one specific industry. For practical aspect, the relationships studied in this thesis explain how employee’s involvement in performance measures helps to provide alternatives for managers to improve the quality of performance measures. The second contribution to practical aspect is that it indicates managers should improve their understanding of performance measures to narrow the gap of information asymmetry by comprehensive communication with their employees.
The remainder of this master thesis is as followed. In Chapter 2, previous literature, theory analysis, hypothesis development and terminology explanation are presented. In Chapter 3, research design, sample selection, and variable operationalization are presented. Response collection and data input selection are also included in Chapter 3. The statistic analysis results are showed in Chapter 4. And finally, discussion about the contributions, limitations, and future suggestions of this thesis is presented in Chapter 5.
Chapter 2 Literature Review and Hypothesis Development
2.1 Literature Review
Except in some highly automated industries, lower level employees still conduct the majority task of their job and obtain better information and understanding of how the work and process are accomplished (Atkinson, 2012). Therefore, a well-designed management accounting and control system (MACS), such as a performance measurement system, needs to involve employee’s participation in its design progress (Atkinson, 2012). Some researches were conducted among organizations to prove the influence of employee’s involvement in the performance measurement system. In 2005, Abernethy and Bouwens (2005) found when a company introduced a new management accounting system, the system usually failed during implementation. A research was then set up to study this phenomenon. Researchers collected data from a random sample of lower lever managers, such as production managers of a manufacturing company, in Australia (Abernethy and Bouwens, 2005). Questionnaires were sent to those selected production managers, and researchers believed responses received were reliable to do the study (Abernethy and Bouwens, 2005). The study used path analysis to estimate research question model with techniques of partial least squares (PLS) and ordinary least squares (OLS). The results showed when production managers involved in the design and implementation of a new management accounting system, the failure rate of system implementation was mitigated (Abernethy and Bouwens, 2005). Another research in 2004 focused more on the relationship between participation in the design of performance management system and performance. The research studied the relationship in the service department of a Dutch photocopiers suppliers, using a quasi-experimental design with participations in a ProMES performance management system (Kleingeld, Van Tuijl, and Algera, 2004). The results showed that involving employees in the design of performance management system can lead to a significantly increase in the performance (Kleingeld, Van Tuijl, and Algera, 2004).
However, limitations still exist among those early researches. For example, the first limitation is the geography problem. According to the context, the first study mentioned before only collected data in manufacturing industry in Australia (Abernethy and Bouwens, 2005), and the second only studied service industries in Netherlands (Kleingeld, Van Tuijl, and Algera, 2004), which means the results may be less convincing in other countries. It is noticed that both of these two studies only take one type of industries into consideration. The second limitation is that previous studies seldom regard the lowest level employees in companies as their primary sample. Studies conducted before focused more on the lower level managers rather than operational employees. In 2012, a study was conducted to shed lights on the enabling function of performance measures when the lowest level employees are involved in the development of performance measures (Groen, 2012). The research conducted by Groen in 2012 took work floor employees and their direct supervisors from all industries in Netherlands as a sample studied enabling effects of performance measures used by managers. Therefore, Groen’s study in 2012 provided information which considering the lowest level employees rather than lower level managers as the research sample. However, it still did not solve the geography problem since the research only collected data in Netherlands. This master thesis uses two existing data sets collected by Groen in 2016, where the data comes from both employees and managers all over the world, to mitigate the geography problem. It also focuses more on the lowest level of employees in a company, for example, the operational employees, and their direct supervisors, which can obtain more relevant data to the research questions in this thesis and provide the empirical contribution to the field of management accounting study.
2.2 Theory and Hypothesis Development
Figure 2.1 The Hypothetical Model
As we know from the research questions of this thesis, in which the study sheds lights on the relationship between employee’s involvement in performance measures and the likelihood of using these performance measures by managers, a delegation of rights to the lowest level employees in performance measures development is found within the relationship stated before. That is to say, an agency relationship exists in the research questions.
2.2.1 Agency Theory
Agency theory is the main theory presented in this master thesis. As the most important theory in the field of management accounting (Lambert, 2007), agency theory presents an agency problem between a principle and an agent delegated with specific rights (Eisenhardt, 1989). Principles are usually the people owning possessions (Eisenhardt, 1989), for instance, shareholders of a company. They tend to delegate other people working for them. Those people are agents (Eisenhardt, 1989). An example of an agent can be the board of directors. In this thesis, I study the influence to the likelihood of using performance measures by managers, caused by involvement of employees. Therefore, agents here are the lowest level employees and principles are their direct managers.
Problems usually come out when principles delegate rights to agents. The most common problems are regarded as two basic assumptions of agency theory: the first one is information asymmetry (Abernethy, Bouwens, and Van Lent, 2004); and the second one is the existence of a conflict of interests between principles and agents (Kunz and Pfaff, 2002). Information asymmetry indicates an information gap between principles and agents (Shields and Shields, 1998). In this situation, the lowest level employees tend to obtain more detailed knowledge and information related to their job than their direct supervisors (Shields and Shields, 1998). As for the existence of conflicts of interests, agents may have their own interests which are not beneficial to principles (Kunz and Pfaff, 2002). In this case, the lowest level employees prefer to devote less time and efforts to the work than the expectation of their direct managers (Kunz and Pfaff, 2002). Therefore, to deal with problems caused by agency theory, especially to mitigate the existence of conflicts of interests between managers and employees, managers need to implement management control systems, such as a performance measurement system, to control behaviors (Sprinkle, 2003), evaluate the performance of employees (Stajkovic and Luthans, 2003), and make decisions about rewards, bonus and promotions to them (Sprinkle, 2003).
2.2.2 Involvement of Employees and Quality of Performance Measures
When implementing a performance measurement system, managers tend to use performance measures which can faithfully reflect the true performance of employees to evaluate and motivate them (Moers, 2006). Performance measures which faithfully reflect employee’s behavior and/or actions at work, precisely measures and verify relevant perspectives of employee’s job performance are regarded as the high-quality measurement for evaluation or motivation, in the eye of managers (Moers, 2006). However, in a real world situation, performance measures developed only by managers cannot reflect all the aspects of work done by employees, because of information asymmetry caused by agency relationship (Groen, 2012). Therefore, to decrease the influence of information asymmetry, it is very important for managers to use performance measures with high quality (Moers,
2006). One way to mitigate information asymmetry is that employees should share private information to managers. Under the circumstance of performance measurement, high quality performance measures are built with employee’s involvement in development process, because of sharing specific job-relevant information (Shields and Shields, 1998). Other literature also presented pieces of evidence that employee’s involvement is most likely to increase the quality of performance measures, or at least it never has a negative effect (Baiman and Evans, 1983). The involvement of employees in developing performance measures mitigates certain information gap between managers and them (Groen, 2012). Therefore, the first hypothesis is formed as followed:
H1 Employee’s involvement improves the quality of performance measures.
2.2.3 Quality of Performance Measures and Likelihood of Using Performance Measures by Managers
When it comes to managers, they usually conduct control systems with both performance measures and compensation system to evaluate and motivate their employees (Jensen and Meckling, 1992). On the one hand, performance measures set orientations and objectives to employees; on the other hand, compensation system motivates and encourage employees to reach the objectives (Groen, 2012). Studied were conducted to shed lights on the relationship between quality of performance measures and likelihood of using them to evaluate and motivate employees by managers. The results showed that the quality of performance measures is positively related to using performance measures for evaluation, monetary compensation and nonmonetary rewards (Moers, 2006). Therefore, the second hypothesis is formed as:
H2 The improved quality of performance measures increase the likelihood of using them to evaluate and motivate employees by managers.
2.2.4 Involvement of Employees and Likelihood of Using Performance Measures by Managers
The third hypothesis should be developed to show the direct relationship between involvement of employees and likelihood of using performance measures by managers, as well as to enhance the model that quality of performance measures is considered as an important mediation variable. Previous literature seldom studied the direct relationship between involvement of employees and likelihood of using performance measures but introduce a moderate or mediate variable in those models to study the relationship in a more comprehensive way (Moers, 2006; Groen, 2012). However, the final results of those researches did show the positive influence on likelihood of using performance measures by managers made by involvement of employees (Groen, 2012). For example, the study conducted by Groen in 2012 showed the co-developing performance measures, which takes impacts of involvement of employees into the development of performance measures into consideration, have a positive relation to using those performance measures for incentive purpose by managers (Groen, 2012). Therefore, based on the abovementioned literature, the third hypothesis is formed as:
H3 Involvement of employees has a positive relationship to likelihood of using performance measures by managers.
The hypothetical model with the three hypotheses is showed at the beginning of this part.
2.3 Terminology
Table 2.1 The Definition of Variables
Variables Definition
Involvement of Employees The substantial influence made by the lowest level employees on the developmental phases of performance measures which are used to evaluate job performance and motivate employees (Groen, 2012).
Quality of Performance Measures
The extent to which the lowest level employees find performance measures faithfully reflect their behavior and/or actions at work, precisely measure and verify relevant perspectives of their job performance (Moers, 2006).
Likelihood of Using Performance Measures by Managers
The extent to which managers find the performance measures important and reliable for evaluation or motivation purposes (Moers, 2006).
2.3.1 Performance measures
According to previous literature, performance measures are traditionally defined as a method to qualify the efficiency and effectiveness of people’s behaviors and actions (Neely, Richard, Mills, Platts, and Bourne, 1997). In management accounting, performance measures are used to measure the performance of employee’s job (Groen, 2012). In an organization, managers use performance measures for different purposes. For example, they use performance measures to measure the job performance of both individual and group (Mendibil, and MacBryde, 2005). They also use performance measures to formally evaluate employee’s job performance and set reasonable reward system to reward them (Stajkovic and Luthans, 2003). With the reward system, which is built based on performance measurement system, decisions of managers about how to evaluate, motivate
and control employees can be influenced, as well as the behaviors and actions of employees (Sprinkle, 2003). Furthermore, if managers use performance measures in a proper way, it will help to enable employees to better performance, rather than control them to do their jobs in a right way (Wouters, and Wilderom, 2008).
Therefore, based on the literature stated above, in this thesis, the performance measures are regarded as a method used by managers to evaluate the job performance (Stajkovic and Luthans, 2003), decide bonus, and decide promotion opportunity of the lowest level employees in the organization (Sprinkle, 2003).
2.3.2 Involvement of employees
The independent variable of the research question in this thesis is the involvement of employee in development processes of performance measures. In this thesis, involvement of employee is defined as the substantial influence made by the lowest level employees on the developmental phases of performance measures used to measure job performance and motivate staff (Groen, 2012). According to the framework developed by Ferreira and Otley (2009), involvement of employees includes tasks of both managers and the lowest level employees, which corporate together to design, implement, and maintain a new or modified performance measurement system. However, this thesis only focuses on the influence made by employees, rather than influences made by managers or by employees and managers together. Therefore, involvement of employees here only reflects significant impacts of the lowest level employees on every potential process when developing a performance measure.
When designing a performance measurement system for an organization, the lowest level employees tend to know more about operational business processes and the data required by development and implementation of performance measures than managers (Masquefa, 2008). Their specific knowledge about operational business processes and required data helps to make performance measures better (Abernethy and Bouwens, 2005). To explain involvement of employees in more details, Abernethy and Bouwens introduce
“influence on system design” scale in their research in 2005, to define involvement of employees as the extent of substantial influence the lowest level employees feel that they have had on the developmental phases of performance measures (Abernethy and Bouwens, 2005). “Influence on the system design” scale indicates five perspectives that can be significantly impacted by involvement of the lowest level employees, which are: 1) designing the concept, definition, and content of performance measures; 2) identifying required data as input into performance measures; 3) modifying the design of performance measures in ongoing period; 4) implementing performance measures; 5) and maintaining performance measures (Abernethy and Bouwens, 2005).
2.3.3 Quality of performance measures
The mediation variable of the research question model is quality of performance measures. Based on previous literature, quality of performance measures is defined as the extent to which the lowest level employees find performance measures faithfully reflect their behaviors and/or actions at work, precisely measure and verify relevant perspectives of their job performance (Moers, 2006). Quality of performance measures is important for both the lowest level employees and their direct supervisors. As for the lowest level employees, performance measures with high quality can motivate and enable them to better job performance results (Wouters and Wilderom, 2008). And when it comes to managers, high quality performance measures can help them evaluate their employees precisely and correctly, and make better decisions for incentive purposes (Moers, 2006), such as increasing bonus, or offering a promotion to a higher position in the company.
The high quality of performance measures can be indicated in the following three perspectives. Firstly, performance measures with high quality should faithfully reflect actions of the lowest level employees (Moers, 2006), which means that employee’s efforts on work can be recognized by managers according to these high quality performance measures (Groen 2012). Secondly, high quality performance measures should be precise, which means the performance measures can precisely and correctly reflect employee’s job
performance (Keeping and Levy, 2000). Finally, high quality performance measures should be verifiable in the eye of the managers (Hall, 2008).
2.3.4 Likelihood of using performance measures by managers
The dependent variable of the research question in this thesis is likelihood of using
performance measures by managers. The previous research defined using performance measures for incentives purposes by managers as the extent to which managers find the performance measures
important and reliable for evaluation and incentive purposes, like both monetary and nonmonetary rewards (Moers, 2006). In my thesis, I defined likelihood of using performance measures by managers based on the definition of using performance measures for incentive purposes by managers in Moers research in 2006. Likelihood of using
performance measures by managers is defined as the likelihood that managers believe
performance measures can faithfully reflect employee’s efforts on work (Moers, 2006), and use them to evaluate employee’s job performance as well as motivate employee to perform better (Sprinkle, 2003). To be more specific, the motivation using purposes by managers can be divided into two parts: monetary motivation, such as providing salary increase, bonuses or extras, and nonmonetary motivation, such as offering an opportunity for promotion to higher position in the company (Moers, 2006).
A short definition for each variable involved in the hypothetical development model is showed in the table abovementioned.
2.4 Control Variables
In this thesis, I consider the gender and age of employees who participated in the survey as control variables. Because these two factors are likely to influence the relationship we showed in the hypothetical model, especially influence the behavior when employees participate in the development process of the performance measures (Groen, 2012). The previous literature states that gender and age of an employee have a positive influence on their performance as well as the evaluation of their performance (Ali and Davies, 2003).
Therefore, to mitigate the influence, I control these two factors to be the same in further data analysis. A correlation analysis which includes these two control variables, is conducted before the verification of the hypothetical model, to show whether these two control variables really have significant influence on the relationship abovementioned in the hypothetical model. More details of how to conduct statistic analysis with control variables are presented in Chapter 4.
Chapter 3 Research Methodology
3.1 Research Design
In order to prove my three hypotheses, I participated in a master thesis project conducted by Groen in the University of Amsterdam. The project aimed to get data to study performance measurement systems used in organizations, especially the enabling effect of co-developing performance measurement systems (Groen 2012). The reasons why this project can offer help to what I study are explained later in this chapter.
The project asked students to accomplish involving at least six pairs of participants into the research survey. After accomplishing the requirement of the project, I am capable to get access to two data sets from Groen, which are about the opinions from both employees and managers towards performance measures used in companies, to conduct quantitative research and analyze my hypothetical model abovementioned. As I mentioned in Chapter 2, the terminology part, opinions from both employees and managers on three variables are needed. The three aspects are whether the pairs think the lowest level employees involve in the development of performance measures to a large extent, whether the quality of performance measures is improved due to the involvement, and whether they think the managers are more likely to use performance measures to evaluate and motivate employees.
3.1.1 Project Processes
Groen used the questionnaire to gain opinions from employees and managers and then integrated the results into two separate data sets. The original version of the questionnaire is in Dutch. It has already been translated into English for international participants. However, the survey also involved some Chinese participants. To help them get the better understanding about what the project is about, with other five Chinese students, I translated the questionnaire into Chinese before sending the first invitations.
Due to technical problems, the system of University of Amsterdam cannot send invitations to those Chinese participants. Therefore, I also directly contacted the six pairs I found and sent them emails with the questionnaire. By accomplishing gathering six pairs of responses, Groen sent me two integrated data sets with all the responses she got at the end of February in 2016.
3.1.2 Questionnaire Design
The first reason that the project can provide data I need to analyze my own hypothetical model is that the content of the questionnaire includes the information I need in my study.
The questionnaire was designed and offered by Groen’s project. The questionnaire is separately designed both the lowest level employees and their direct managers are used to gain relevant data and information (Groen et al., 2015). But the content of questionnaire for both employees and managers is almost the same. Both questionnaires for both employees and managers include general information of performance measures used by a company (Moers 2006), opinions about influence on performance measures made by people’s behavior (Abernethy and Bouwens 2005), opinions about using performance measures by managers (Moers 2006), and other general personal information about participants, such as gender, age and nationality (Groen et al., 2015). Although the content in the questionnaire for employees and managers is almost the same, it shows a different point of view on the same question from different participants. Therefore, to make sure the responses are consistent, it is necessary to gain opinions from both two sides. Besides, each question in the questionnaire uses some similar statements to test whether people have similar judgments towards the same question (Groen et al., 2015). That people have similar answers to similar items can increase the reliability of data gained by questionnaire in a survey research (Groen et al., 2015).
3.2 Sample Selection
The second reason that the project helps to provide me with reliable data is that the sample selected with Grone’s requirements can support me to eliminate limitations of previous researches, such as industry limitation and geography limitation.
As for the project conducted by Groen, she considered pairs including the lowest level employees, such as operational workers, and their direct supervisors and managers as her research samples (Groen et al., 2015). Groen used pairs to collect more reliable responses because the responses gathered by participants were given in the eye of the relevant person (Groen et al., 2015). The sample pairs should come from the organizations which use performance measures to evaluate employee’s job performance (Groen et al., 2015). The lowest level employee in the pair should have at least one-year experience at his or her current work (Groen et al., 2015). In Groen’s research, she did not only gain data from only one country but also obtain data all over the world (Groen et al., 2015). Besides, she did not just look up into one specific industry but collected data among diversified industries with three categories of production, mass service, and professional service industries. During the project, Groen encouraged students to participate into obtaining high quality data by accomplishing getting at least six pairs of participants to answer the questionnaire. To gain trust and more reliable data and information from the participants, a contract was made by Groen that the researchers who conducted the survey research shall never release any information about the participants to any third party. Furthermore, she also made sure that students who participated in the project only have access to get the integrated data in the form of data sets, rather than the specific detailed individual responses to the questionnaire.
When the participants selected and contacted by students meet all the requirements of the research questions, advantages of the sample become obvious, compared to selecting and contacting potential participants randomly. There are four obvious advantages of the sample selected by students. Firstly, students who join the project come from various countries in the world, therefore, they select the sample from different
countries, which helps to eliminate geography limitation. Secondly, since students usually consider their acquaintances who also meet all the requirements as research sample, the response rate of the questionnaire tend to be higher, than sending the questionnaire to people we do not know randomly. It was easier for students to contact, communicate and remind participants to response the questionnaire. The response rate was very high according to the data we received. Although Groen did not provide a precise number of invitations of the questionnaire sent to the participants, the project still obtained 106 responses from the employees and 103 responses from the managers. Students who joined the project all achieved at least six pairs of responses. Thirdly, the project allowed students on behalf of the university. Although the participants were selected by students, the questionnaire was still sent officially in the name of University of Amsterdam. Questionnaire sent by official organizations tended to gain more attention from the perspective of participant’s side. They tended to consider the questionnaire more seriously and provide more useful responses than the questionnaire sent by students alone.
When it comes to my own research questions, the research sample is the same as Groen’s, Therefore, I directly use data sets provided by Groen’s project to conduct a statistic analysis. Research sample meeting requirements of Grone’s project was directly selected by students who participated in this master thesis project. Participants were invited officially in the name of University of Amsterdam to ensure the high quality of responses. Two data sets for both employees and managers were available at the end of February 2016.
3.3 Variable Operationalization
when it comes to variable operationalization, three variables, which are involvement of employees, quality of performance measures and likelihood of using performance measures by managers, and two control variables, which are gender and age of participants, are operationalized in this part for further data analysis in chapter 4. All the items are measured by participants by a seven-point fully anchored Likert scale: (1) totally disagree,
(2) disagree, (3) moderately disagree, (4) neutral, (5) moderately agree, (6) agree, and (7) totally agree (Groen et al., 2015).
3.3.1 Involvement of Employees (IOE)
The independent variable in my hypothesis development model is involvement of employees, which is denoted as IOE for further data analysis. In my thesis, involvement of employee is defined as the substantial influence made by the lowest level employees on the developmental phases of performance measures used to measure job performance and motivate staff (Groen, 2012). According to previous studies, five aspects show where involvement of employees can have effects on performance measures (Abernethy and Bouwens, 2005). The five aspects are: 1) designing the concept, definition, and content of performance measures; 2) identifying required data as input into performance measures; 3) modifying the design of performance measures in ongoing period; 4) implementing performance measures; 5) and maintaining performance measures (Abernethy and Bouwens, 2005). As is mentioned in the hypothetical model that involvement of employees can only be influenced by the lowest level employees in the organizations, I consider responses from employee’s dataset as the data input to the variable of involvement of employees.
Therefore, in my master thesis, Abernethy and Bouwens’s (2005) “influence on the system design” scale is used to operationalize involvement of employees with five items in questionnaire of employees, which are:
1) the employee had influence on how the performance measures are designed; 2) the employee had influence on the choice of which data are used as input into the performance measures;
3) the employee had influence on ongoing modifications to the design of the performance measures;
4) the employee had influence on the implementation of the performance measures;
5) and the employee had influence on the maintenance of the performance measures.
3.3.2 Quality of Performance Measures (QPM)
The mediation variable of my research question is quality of performance measures, which is defined as the extent to which the managers find performance measures faithfully reflect the lowest level employee’s behaviors and/or actions at work, precisely measure and verify relevant perspectives of the job performance (Moers, 2006). Quality of performance measures is denoted as QPM in data analysis. Literature shows quality of performance measures can be operationalized by whether the performance measure can faithfully reflect actions of the lowest level employees (Groen 2012), whether it can precisely and correctly reflect employee’s job performance (Keeping and Levy, 2000), and whether it can be verified according to the direct manager (Hall, 2008). According to the definition of quality of performance measure, managers are the people to give opinions on whether the performance measure is in good quality. Therefore, I operationalize quality of performance measures from the responses only from manager’s perspective by the following four items, which are:
1) the performance measures measure only what my employee can actually influence;
2) the performance measures express accurately whether my employee functions well or not;
3) if my employee performs well, it is directly reflected in the performance measures;
4) and the performance measures are objective and verifiable.
3.3.3 Likelihood of Using Performance Measures by Managers (LUPM)
The dependent variable of the research question in this thesis is likelihood of using performance measures by managers, defined as the extent to which managers find the
performance measures important and reliable for evaluations and motivations (Moers, 2006). Likelihood of using performance measures by managers is denoted as LUPM. Questionnaire based on Moers (2006) uses eight items to operationalize likelihood of using performance measures by managers to see the extent that the manager attaches high importance to the performance measures, which are:
1) the evaluation of the employee’s performance; 2) periodic performance reports;
3) officially rating the employee’s performance; 4) periodic discussions with the employee; 5) determining potential salary increases; 6) determining potential bonuses or extras;
7) increasing the employee’s chance of promotion;
8) and increasing the employee’s authority within the organization.
3.3.4 Control Variable Operationalization
The two control variables in my research are the gender and age of employees who participated in the research. According to literature, these two factors are likely to effect the relationship I study in the hypothetical model (Ali and Davies, 2003; Groen, 2012). As for the gender of the employees, it is operationalized to two categories, which are male and female. Male is denoted as 1, and female is denoted as 2 in the dataset. And for the age of the participants, an age distribution figure of valid data input is presented in the next chapter. Additional statistic analysis which takes control variables into consideration is conducted after proving the hypothetical model in Chapter 4.
3.4 Responses Analysis
The responses from participants are separated from employees and managers in two original data sets. In total, I got 106 responses from employee’s dataset and 103 responses from manager’s dataset. However, not all the responses meet the requirements of my
research sample. For example, for the variables studied in the hypothetical model, some values are missing. And not all the responses are collected in pairs. After deleting all responses which do not meet the requirements, I still have 93 valid pairs of responses. Within these 93 pairs of responses, 79 responses are from profit organizations, and 14 are from non-profit organizations. As it is mentioned in previous part of this thesis, the hypothetical model is formed under the circumstance of profit companies. Therefore, I only consider 79 pairs of responses from profit organizations as valid data input, which are enough for analyzing a master thesis research. The general information of responses gathered is showed in the table below.
Table 3.1 General Information of Responses
Responses Received Employees 106 Managers 103 Valid pairs 93 Profit organization 79 Non-profit organization 14
Valid Data Input 79
Before I start data analysis, I combine two original data sets of employee and manager into one integrate dataset. The new combined dataset includes all the variables demanded to test the hypothetical model. It only considers 79 valid pairs of profits organizations. The variables in the new combined dataset are involvement of employees (IOE), gender and age from employee’s dataset, quality of performance measures (QPM) and likelihood of using performance measures by managers (LUPM) from manager’s dataset. Variables selected to form the new database are based on statements from the variable operationalization part in Chapter 3. Table 3.2 also gives an overview of the characteristics of these 79 valid pairs of responses.
Table 3.2 Respondent Characteristics
Characteristic Employees Managers
Gender 53% male 73% male
47% female 27% female
Age Mean = 32.14
(SD = 9.493)
Mean = 38.34 (SD = 9.273) Industry type of the company 16.9% Production
24.7% Mass services 58.4% Professional services
Since employee’s age is regarded as an important control variable in this hypothetical model, I think it is necessary to present the distribution of employee’s age to find whether the data collected really includes a wide range of employee’s age. The distribution is showed in Appendix B. As showed in Appendix B, 45 employees come from the age range from 20 to 30 years old, 21 employees come from the range from 31 to 40, 11 employees from 41 to 60, and only one employee comes from below 20 and above 60, respectively. The dataset with wide range of employee’s age indeed helps me to test the hypothetical model with the control variable of employee’s age.
3.5 Statistical Analysis
As it can be seen from Figure 2.1, the hypothetical model describes a mediation relationship between involvement of employee and likelihood of using performance measures by managers. In this thesis, I use method put forward by Baron and Kenny to analyze the model. The mediation hypothesis model is defined as a model predicts that the effect of an independent variable, which is involvement of employee in this thesis, on a dependent variable, which is likelihood of using performance measures by managers, exists at least partly due to the independent variable firstly influents a mediate variable, which is quality of performance measures, and the mediate variable in turn influents the dependent variable (Baron and Kenny, 1986). To verify that quality of performance measures is the mediate variable in the hypothetical model developed before in Chapter 2, three conditions should be achieved according to Baron and Kenny method. Firstly, the independent variable must have an influence on the dependent variable (Baron and Kenny, 1986).
According to the analysis and literature stated before, Hypothesis 1 expects a positive influence on quality of performance measures by involvement of employees. Therefore, the first condition is accomplished. Secondly, the independent variable must be shown to affect dependent variable (Baron and Kenny, 1986). According to results of previous studies, although no researches studied the direct relationship between involvement of employees and likelihood of using performance measures by managers, the final results did show the positive influence between these two variables (Groen, 2012). So, I assume that the results of later statistic analysis support this relationship. Therefore, the second condition is accomplished. Lastly but not the least, the mediate variable must have an influence on dependent variable (Baron and Kenny, 1986). The literature analysis of developing Hypothesis 2 expects there is a positive relationship between quality of performance measures and likelihood of using performance measures by managers. Therefore, the third condition is accomplished. When theses three conditions are all achieved, the influence of involvement of employees on likelihood of using performance measures by managers should be less than the influence of quality of performance measures on likelihood of using performance measures by managers (Baron and Kenny, 1986). This statement should be supported with statistic analysis results, which are showed in the next chapter. After that, quality of performance measures can be confirmed to be the real mediate variable in the hypothetical model. The procedure of analyzing a mediation model with statistics is to run a series of linear regressions according to the statements of Baron and Kenny (Baron and Kenny, 1986). In this thesis, I use SPSS as statistic analysis tool to run all the linear regressions. The hypothetical model is tested with the new combined dataset which is abovementioned in previous paragraph.
Chapter 4 Data Analysis
In this Chapter, before testing the hypothetical model, I firstly conduct reliability analysis, convergent validity analysis and discriminant validity analysis to ensure the quality of the dataset is good. Then I conduct linear regressions to test the hypothetical model illustrated in Figure 2.1 based on the method of Baron and Kenny (1986). And finally, I conduct linear regressions with control variables to show whether they indeed influence the relationships showed in the hypothetical model.
4.1 Reliability Analysis
Firstly, I conducted a reliability analysis to make sure that all the items for each variable in the hypothetical model exactly measure the same thing. For example, the items mentioned in 3.3.1 Variable Operationalization of involvement of employees (IOE) are really used to reflect how the employees involve in the development of performance measures in the real world. I use Cronbach’s Alpha to test the reliability for these three variables introduced in this thesis, which are involvement of employees (IOE), quality of performance measures (QPM) and likelihood of using performance measures by managers (LUPM). A Cronbach’s Alpha exceeding 0.5 or preferably exceeding 0.7 shows a good reliability.
The results of the Cronbach’s Alpha for each variable in the hypothetical model are very high. The Cronbach’s Alpha of five items of IOE is 0.95, the Cronbach’s Alpha of four items of QPM is 0.79 and the Cronbach’s Alpha of eight items of LUPM is 0.902, which are all higher than 0.7 and show a good reliability. I also notice that the results show that when deleting any item, the Cronbach’s Alpha for both IOE and LUPM do not increase. Therefore, I do not delete any items of IOE and LUPM from the dataset at this moment and maintain the high reliability of these two variables. However, the results of QPM show that when deleting the first item of QPM, the Cronbach’s Alpha of QPM increases to 0.837, which is a little bit higher than its previous Cronbach’s Alpha (0.79)
with all four items. Compared with IOE that using five items to measure, and LUPM using eight items, QPM only has four items to measure. If the first item of QPM is deleted, only three items are left for measuring QPM. In this situation, I consider that three items for measuring QPM are not enough. Besides, considering that 0.79 is already a very high Cronbach’s Alpha and shows a good reliability, there is no need to delete the first item of QPM.
In one word, as showed in the results of reliability analysis, all variables used in the hypothetical model have a good reliability.
4.2 Convergent and Discriminant Validity Testing
4.2.1 Convergent Validity Testing
Then, to see whether there is an underlying factor that can effect the results of items used to measure three variables, I use the principal component analysis to conduct a convergent validity test.
The results of principal component analysis show that there is only one component for IOE, QPM and LUPM, respectively. In other words, for IOE, QPM and LUPM, no underlying factor exists to influence the measurement for these three variables by all the items included in the survey questionnaire. Furthermore, the cumulative percentages of both initial eigenvalues and extraction sums of squared loadings show the extent of the component can be used to explain the variable. for example, the cumulative percentage of Component 1 for IOE is 83.479%, which means the Component 1 can explain 83.479% of involvement of employees. The same goes for QPM and LUPM. For QPM, Component 1 explains 63.051%, and for LUPM, the extent of explanation is 59.883%. Although the cumulative percentages of the component for QPM and LUPM are not as high as that for IOE, they are all higher than 50%. Combined the results that these three variables only have one component after the principal component analysis, a conclusion can be drawn that no other underlying factor exists to influence the results of items used to measure
IOE, QPM, and LUPM. Therefore, the convergent validity for IOE, QPM, and LUPM is confirmed.
4.2.2 Discriminant Validity Testing
To make sure whether different items indeed reflect the different variables as they should, I also conduct a principal component analysis with all the items used to measure all different variables to test the discriminant validity. The results of discriminant validity test are showed below in the table of rotated component matrix.
Table 4.1 Rotated Component Matrix of all the Items Items Component 1 2 3 4 IOE_1 0.942 IOE_2 0.919 IOE_3 0.92 IOE_4 0.873 IOE_5 0.877 QPM_1 0.726 QPM_2 0.802 QPM_3 0.726 QPM_4 0.779 LUPM_1 0.762 LUPM_2 0.748 LUPM_3 0.723 LUPM_4 0.69 LUPM_5 0.535 LUPM_6 0.705 LUPM_7 0.877 LUPM_8 0.855
Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. Rotation converged in 21 iterations.
As showed in Table 4.1, four components are presented. The number in the table shows the “loadings” of each item to the variable it measures. The loadings should be at least higher than 0.5. The loadings below 0.5 are already deleted in Table 4.1. The results showed in Table 4.1 indicates that Component 1 reflects all the items used to measure the extend of employee’s involvement, and Component 4 shows all the items used to measure quality of performance measures. However, the results of items of likelihood of using performance measures by managers are not the same good as that of involvement of employees and quality of performance measures. Items of LUPM are in two components. The first five items of LUPM belong to Component 2, while the rest three items belong to Component 3. This result is not expected. The results expected should be that all eight items of LUPM are in one component or that the first four items, which measures manager’s usage of performance measures for evaluation purpose, belong to one component and the rest four items which shows manager’s motivation purpose belong to the other component. So I look back to the content of items, and find the last three items are more about how likely managers use performance measures as incentive methods. Item 5 and Item 6 of LUPM reflect how much important the managers attach to performance measures to make monetary incentive decisions. Item 6 and Item 7 of LUPM reflect how much important the managers attach to performance measures to make nonmonetary incentive decisions. This phenomenon indicates that the data of LUPM is possibly wrong because participants may not seriously answer this question. Besides, this thesis does not study further influence of manager’s motivation on employee’s job performance. Therefore, the last three items should be deleted from statistic analysis.
After factor analysis, I decide to use measure of contracts which include more rather than one item to measure the variables in the hypothetical model. I calculate the average score of all the five items of employee’s involvement as the contract of IOE, and that of all the four items of quality of performance measures as the contract of QPM. For LUPM, where the results of Table 4.1 show two components of this variable, I only calculate the average score of the first five items as the contract of LUPM, but delete the last three items.
The five items of LUPM can reflect how likely managers use performance measures for both evaluation or motivation purposes.
4.3 Preliminary Analysis
Before testing three hypotheses, I firstly get descriptive statistics of all the variables used in this research study and run a series of correlation analysis for these variables to obtain the first understanding of the data used in the research sample. The descriptive statistics of all variables are showed in Table 4.9 below.
Table 4.2 Descriptive Statistics of Variables
Variables Number of
Observations Min. Max. Mean
Std. Deviation Employee's gender 79 1 2 1.47 0.50 Employee's age 79 18 62 32.14 9.44 IOE 79 1 7 3.46 1.73 QPM 79 1 7 5.12 1.08 LUPM 79 2 7 5.55 0.98
The number of observations is 79, which shows that 79 valid pairs of data are all used to run statistic analysis tests. For employee’s gender, I denote male as 1 and female as 2. The mean of employee’s gender is 1.47 shows 47% of employees are male, and the rest 53% are female. It is possible for me to use this dataset to study the potential influence made by employee’s gender on the hypothetical model. As for employee’s age, the youngest employee is 18 years old, and the eldest is 62. The average age of employees participating in the survey research is 32.14. The standard deviation of employee’s age is 9.439, which is higher than the standard deviation of another variable, indicating that the measures of dispersion of employee’s age are high. Therefore, the data of employee’s age collected is suitable for testing the hypothetical model with a control variable of employee’s age.
The average scores for IOE, QPM and LUPM are 3.46, 5.12, and 5.55, respectively. And the standard deviations for IOE, QPM and LUPM are 1.73, 1.08 and 0.98,
respectively. The low standard deviation means the scores of IOE, QPM and LUPM are stable, which is good for hypotheses testing.
Then, I run both a Pearson and a non-parametric Spearman correlation for all the variables presented in Table 4.3, and the results are integrated into the table below.
Table 4.3 Correlation Results of all Variables
Employee's Age Employee's Gender IOE QPM LUPM
Employee's Age 1 Employee's Gender 1 IOE 1 QPM * * 1 LUPM * * * 1
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).
The correlation results of the dataset show the existence of significant correlation relationship between IOE and QPM with a Pearson correlation of 0.232 (p-value<0.05), and the relationship between QPM and LUPM with a Pearson correlation of 0.385 (p-value<0.01) and a non-parametric Spearman correlation of 0.47 (p-(p-value<0.01). However, the results do not show a significant correlation relationship between IOE and LUPM. These results give the first understanding of the relationships I study in the hypothetical model. As can be seen from the table, the most significant correlation exists between QPM and LUPM. However, the correlations between IOE and QPM is less significant. And for IOE and LUPM, there is not relationship between these two variables. When it comes to control variables in this thesis, both employee’s gender and employee’s age have no significant correlation with other variables. Therefore, in the next step, I test each hypothesis with the data of all 79 valid pairs, and then conduct additional checks to see whether these two control variables indeed have influence on the hypothetical model studied in this master thesis.
4.4 Hypotheses Testing
Although the previous results of correlation analysis provide some evidence to support the method of Baron and Kenny (1986), linear regression analysis is still needed to provide more concrete evidence. Therefore, I use the method of Baron and Kenny (1986) to test the hypothetical model. According to Baron and Kenny (1986), three steps should be taken to confirm that the QPM is the mediation variable of the relationship between IOE and LUPM. Baron and Kenny (1986) also states that the relationship between IOE and LUPM should be less than the relationship between QPM and LUPM. The three steps conducted according to Baron and Kenny (1986) are:
(1) running a linear regression of LUPM on IOE; (2) running a linear regression of QPM on IOE;
However, the method of Baron and Kenny (1986) only test whether quality of performance measure is indeed a mediation variable in the hypothetical model, but not test the relationship assumed in Hypothesis 2. Therefore, after conducting three linear regressions abovementioned, I also run a linear regression of LUPM on QPM to test Hypothesis 2. The results of linear regressions are showed in the table below.
Table 4.4 Linear Regression Results of Hypotheses Testing
1 2 3 4 5 6 7 8 9 10
Dependent variable is QPM Dependent variable is LUPM
Independent variables IOE () () ( )* * )* ) * - - - - QPM - ) ( -* ) ( *( * Control variables Employee's gender ( ) )) ( )) Employee's age ( ) * - ) R square )( )) )* ) (- ( ) )-Number of observations 79 79 79 79 79 79 79 79 79 79
***p<0.001; **p<0.01; *p<0.05; p-values are two tailed Standardized coefficients are under unstandardized coefficients
According to Table 4.4, Model 5, Model 1, and Model 6 show the linear regressions of LUPM on IOE, the regression of QPM on IOE, and the regression of LUPM on both IOE and QPM. These three regression model indicate three steps of Baron and Kenny (1986) which are to test the hypotheses. Model 7 shows the extra step I take to test the relationship between QPM and LUPM directly.
The result of Model 5 firstly shows that no significant relationship exists between IOE and LUPM, with a standardized coefficient of 0.182 (p-value>0.05). Therefore, the relationship I assume in Hypothesis 3 before does not exist. In other words, employee’s involvement in performance measures does not have a positive relationship to the likelihood of using those performance measures by managers. Then, the result of Model 1 shows that the regression model of QPM on IOE for Hypothesis 1 is significant with a standardized coefficient of 0.232 (p-value<0.05). Therefore, a positive relationship indeed exists between involvement of employees and quality of performance measures. The results of Model 6 show that for IOE, the unstandardized coefficient is 0.056 (p-value>0.05), however, for QPM, the unstandardized coefficient is 0.329 (p-value<0.001). The regression of LUPM on IOE and QPM is also significant (p-value<0.001), which indicates QPM is a partial mediation variable when both Model 5 and Model 1 are significant. However, in this case, Model 5 is not significant. Therefore, no conclusion can be drawn from the results analyzed above.
However, when it comes to the results of Model 7, which test the direct relationship between QPM and LUPM, the model shows a great significant results with a standardized coefficient of 0.385 (p-value<0.001). Combined with the results of Model 6, I come up with a conclusion that managers are more likely to use performance measures with high quality. Therefore, Hypothesis 2 is supported with concrete data analysis results.
In summary, the first two hypotheses in the hypothetical model are supported by the results of linear regressions, but the last hypothesis is not. Besides, the relationship with a mediation variable showed in the hypothetical model does not exist, either.
***p<0.001; **p<0.01; *p<0.05; p-values are one-tailed
Figure 4.1 The Standardized Coefficient Results of Testing Hypothetical Model 4.5 Additional Checks
In previous regression analysis, I do not include control variables when running the statistic. Therefore, in additional checks, I take the influence of two control variables on the hypothetical model into considerations. These two control variables are employee’s gender and employee’s age. Employee’s gender and age are only likely to influence the variable of involvement of employees. In other words, they can only influence the hypotheses studying IOE. In previous analysis, Model 1, Model 5 and Model 6 all involve IOE as independent variables into the linear regression. However, Model 5 is not significant, which is not taken into consideration when running additional checks. Therefore, in this part, I only run linear regressions with these two control variables to test the relationships studying in Model 1 and Model 6 again. The results are showed in Table 4.4 above.
Model 2, Model 3, and Model 4 test the regression of QPM on IOE with control variables of employee’s gender and age. The results of Model 3 show that when controlling the variable of employee’s age, the relationship existing between IOE and QPM is still
significant. However, the results of Model 2 and Model 4 show that the model is less significant when controlling employee’s gender or controlling both employee’s gender and age. Then I check the p-values for Model 2 and Model 4. As for Model 2, the p-value is 0.059, and for Model 4, the p-value is 0.056. Although the p-values for both Model 2 and Model 4 exceed 0.05, which indicate less significance of the model, these deviations can be tolerant because the p-values are very close to 0.05. Combined with the fact that the coefficients of the model also do not change in a large extend, I consider that the regression of QPM on IOE is significant, and that both employee’s gender and age do not influence the relationship between IOE and QPM. Therefore, Hypothesis 1 is still supported with the results of data analysis after additional checks.
Model 8, Model 9 and Model 10 test the regressions of LUPM on both IOE and QPM when controlling variables of employee’s gender and age. The results show that the significance of the regression of LUPM on IOE and QPM dose not change when controlling the variable of employee’s gender and age respectively, or controlling the variables of employee’s gender and age at the same time. The coefficients of the regressions for each independent variable also do not change to a large degree. Therefore, the conclusion can be drawn that employee’s gender and age do not influence the linear regression model of LUPM on IOE and QPM.
Chapter 5 Discussion
The purpose of this master thesis is to shed lights upon the relationship between involvement of the lowest level employee who carries out the operational activities in the organization to the development process of performance measures, and the likelihood of using these performance measures to evaluate and motivate the employee by the direct manager, with a mediation variable of quality of performance measures. Based on the results showed in Chapter 4, Hypothesis 1 which states that employee’s involvement in the design, implementation and maintenance processes of performance measures can improve the quality of those performance measures is proved. The statistic analysis results show a direct positive relation between involvement of employees in employee’s perspective and quality of performance measures in manager’s perspective. The second hypothesis is also supported by the statistic analysis, where the results find performance measures with improved quality increase the likelihood of using these performance metrics to evaluate employee’s performance during the work or motivate them compensations. In other words, there is also a direct positive relation between quality of performance measures and the likelihood of using these performance measures to judge employee’s grades by managers. The results also indicate that the relation of Hypothesis 2 is more significant than the relation of Hypothesis 1. However, contrary to my expectation, the results do not show a significant linear regression model of Hypothesis 3, which studies the direct relationship between involvement of employees and likelihood of using performance measures by managers. The results only show a positive model, but this model is not significant. Therefore, Hypothesis 3 is not confirmed with statistic analysis results, as well as the hypothetical model with a mediation variable of quality of performance measures.
5.1 Explanations of Hypotheses
The first two hypotheses in the hypothetical model are supported by the results of statistic analysis in Chapter 4, which indicates that the explanation in Chapter 2 when
