The effect of New Public Management on
management controls and employee
motivation in Dutch municipalities
by
Damiro de Jongh
University of Groningen
Faculty of Economics and Business Master of Science Business AdministrationOrganizational & Management Control
Date: June 22, 2014
Author:
Damiro de Jongh Ganzevoortsingel 9a 9711 AH Groningen E-mail: Damirodejongh@gmail.com Student number: s1909177Supervisor:
MSc Berend van der KolkSecond assessor:
Prof. dr. ir. Paula van Veen-DirksAbstract
Municipalities are facing major changes in the last decades, amongst others due the financial crisis. This led to an increasing focus of the municipalities to use management controls, and specifically results controls. New Public Management (NPM) reforms encourage implementation of private-sector and business-like practices in municipalities. This way it is encouraged to increase the search for working more efficiently and to find possibilities for cost reductions. According to the literature, the aforementioned management controls can be used as a tool to influence the behavior and performance of employees through motivation. However, there is currently little to no literature available that analyses the relationship between NPM and management controls and the relationship between management controls and the motivation of employees. Therefore, this thesis will focus on these relationships. To analyze the relationship between management controls and employee motivation, the management control theory of Merchant and van der Stede (2007) in combination with the goal-setting theory of Locke (1988) will be used. By using these theories in combination with the self-determination theory of Deci & Ryan (1985) it is envisaged to define a relationship of management controls with intrinsic and extrinsic motivation. In addition, differences between departments are analyzed by combining the literature of the framework of organizational control of Ouchi (1979) and the management controls by Merchant and van der Stede (2007). The framework of organizational control contains two scales that determine which control structure is appropriate for a certain department. These scales are measured and thereafter used to analyze if the management controls are assigned to the departments as is theorized by the organizational control framework. These expectations of NPM, management controls and motivation developed by consulting the literature are reflected in eight hypotheses. To test the hypothesized relationships a survey is filled in by 150 employees of multiple Dutch municipalities. The analysis of the responses resulted in the rejection of all but one hypothesis. Hence, the findings in this thesis are not in line with the hypotheses developed by consulting the literature. Moreover, the analyses resulted in reversed relationships in which intrinsic motivation is enhanced by setting performance targets and providing external rewards, and where no relationship is found between extrinsic motivation and these elements.
Table of contents
1. INTRODUCTION ... 5
2. LITERATURE REVIEW... 8
2.1NEW PUBLIC MANAGEMENT ... 8
2.2MANAGEMENT CONTROLS ... 9
2.2.1 Management Control Packages ... 9
2.2.2 The framework of Ouchi ... 10
2.2.3 Employee perception of controls ... 11
2.3MOTIVATION ... 12
2.3.1 Management Controls on Motivation ... 13
2.3.1.1 Setting targets ... 13
2.3.1.2 Providing rewards ... 13
2.4SUMMARY AND HYPOTHESES DEVELOPMENT ... 14
3. METHODOLOGY ... 17 3.1RESEARCH METHOD ... 17 3.2DATA COLLECTION ... 18 3.3MEASUREMENTS... 18 3.3.1 Dependent variables ... 19 3.3.2 Independent variables ... 19 3.3.3 Control variables ... 19 3.4STATISTICAL ANALYSIS ... 20 3.4.1. Assumptions ... 20 3.4.2 Analysis ... 21 4. RESULTS ... 22
4.1OUTLIERS AND ASSUMPTIONS ... 22
4.2DESCRIPTIVE STATISTICS... 23 4.2.1 Main variables ... 23 4.2.2 Departments ... 24 4.3CORRELATION ANALYSIS ... 25 4.4HYPOTHESES TESTING ... 26 4.4.1 Comparing means ... 26 4.4.2 Regression analysis ... 27 4.5SUMMARY OF RESULTS ... 29 5. DISCUSSION ... 31 5.1CONTROL VARIABLES ... 31
5.2DEPENDENT AND INDEPENDENT VARIABLES ... 32
APPENDIX B ... 47
APPENDIX C ... 48
APPENDIX D ... 49
APPENDIX E ... 50
1. Introduction
The developments in the society require the municipalities to control more and more on policy implementation, goals and results. Important conditions are personal responsibility and authority. The ‘goal setting theory’ suggests that setting goals allows people to move in the right direction
(steering). In addition, setting goals influences motivation (the engine).
Source: Training and education: ‘Target and result driven’ on the website of the municipality of Amsterdam - 27th of January 2014.
The importance of this thesis is reflected in a paper published by the United Nations Research Institute for Social Development (Larbi, 1999). In the mentioned paper an overview is provided of the evolution of the New Public Management (hereinafter: “NPM”) approach, the combination of factors driving the change towards NPM, and the potentials and limitations of NPM. Not uncommon for countries implementing NPM reforms is that those countries have been hit by an economic crisis, which triggered the search for efficiency and possibilities for cost reductions. The role and institutional character of the public sector have been under pressure to be more market-and private sector-oriented. NPM is not an isolated concept and the implementation thereof will lead to changes in management controls and the way how employees will be motivated. This relationship is still indistinct in the literature and is therefore a value adding research topic.
NPM reforms lead to governmental organizations implementing multiple private-sector and business-like practices (Ferlie et al., 1996) and to an increase in using output-related performance measures and rewards (Perry et al., 2009). Consequently, governmental organizations have enhanced their focus on controlling and managing the results of their employees (ter Bogt & van Helden, 2000; ter Bogt, 2008; Groot, 1999; Groot & Budding, 2008).
Since the 1980s, the Dutch government initiated a movement towards more efficiency and output control. To that extent, the Netherlands implemented multiple business-like management control elements (Groot, 1999; ter Bogt, 2008; Hengel et al., 2014). Control elements can be used as a tool to align the interest of the managers and employees to achieve goal congruence (Flamholtz et al., 1985), and can be identified as follows: “a system, rule, practice, value or other activity that is in place in order to direct employee behavior”(Malmi & Brown, 2008, p. 290).
In the psychology literature, the definition of motivation contains three determinants: (1) influencing the direction, (2) intensity and (3) persistence of behavior. This definition gives an indication that motivation might play an important part in what Merchant and van der Stede (2007) describe as the aim of management controls. This thesis will focus on intrinsic and extrinsic motivation (self-determination theory; Deci & Ryan, 1985) and their expected response to changes inducted by NPM.
Besides the relation of NPM with management controls and motivation, the impact of differences between departments of the municipalities on using the various management controls is measured. To measure this relationship, the organizational framework of control of Ouchi (1979) is used in combination with the theory of management controls of Merchant and van der Stede (2007). The theory of employees’ perception of control (Tessier & Otley, 2012) is used to enable to view each response as an individual response.
To investigate the assumptions of NPM about management controls and motivation and its implementation in governmental organizations this thesis combines multiple theories of different disciplines. This thesis will focus on defining and exploring the relationship of NPM with management controls, and indirectly with motivation. The research question for this thesis is as follows:
“What are NPM’s assumptions with respect to the motivation of employees, and do these assumptions hold in municipal organizations?”
To find an answer to this question, this research will contain a theoretical and empirical part. The following questions will be answered theoretically in the literature review:
1. What is NPM and what does existing literature conclude with respect to the implementation of NPM in municipalities?
2. What are management controls and management control packages?
3. What is the influence of differences in the department of municipalities on management controls and how does this reflect on employees’ perception of controls?
4. What does the literature say about the implementation of management controls in municipalities and its influence on the motivation of employees?
The following questions will be answered by conducting empirical research: 5. What is the influence of NPM on management controls?
6. What is the main object of control in the management control package in a department? 7. What is the influence of management controls on motivation?
8. What is the influence of differences in departments on the implementation of management controls?
used or highest perceived. Moreover, assumptions regarding output controls positively influencing extrinsic motivation and decreasing intrinsic motivation are also rejected.
2. Literature review
This section answers the theoretical questions as outlined in the introduction, starting with covering the literature of NPM. Thereafter, the management controls are discussed using the theory of Merchant and van der Stede (2007). The framework of Ouchi (1979) is used to identify possible differences between departments of the municipalities, and the influence of these differences. Moreover, the employee perceptions of management controls (Tessier & Otley, 2012) are explained. Furthermore, the concept of motivation will be explained according to the self-determination theory of Deci and Ryan (1985), which is followed by a discussion of the influence of management controls on motivation. The last paragraph will summarize the literature review and connects the different theories. This will lead to the development of several hypotheses.
2.1 New Public Management
Since the 1980s, NPM reforms have been implemented in governmental organizations across multiple countries, including the Netherlands. Traditionally, the Dutch government is highly input-and process-oriented, which means that the organizations are being controlled based on financial budgets and administrative procedures (ter Bogt et al., 2011). The NPM reforms resulted in a shift towards output-oriented planning and control with private sector management approaches. NPM aims to improve the effectiveness and efficiency of the local government (Hood, 1995; Dixit, 1997), which in turn explains the strong emphasis on output related measurements of performance (Frey et al., 2013).
The theoretical basis of NPM follows the principal-agent theory (Jensen & Meckling, 1976; Kaboolian, 1998; Arellano-Gault, 2000), which argues that monetary incentives can be used to align the interest of the agent and the principal, whereby in this situation the agent is the employee and the principal the manager (Jensen & Murphy, 1990).From Frey et al. (2013) one could conclude that the agency theory views employees as economically rational: “Following the theoretical price effect, the higher the price the more effort is exerted. According to this view extrinsic incentives in the form of monetary rewards motivate individuals’ additional or marginal efforts.” This view has some empirical support in the private sector (Lazear, 2000). However, Frey et al. (2013) show that there are three reasons why the application of the principal-agent theory within public service might be misleading: there is a difference between work of public sector employees and working at an assembly line.
adequate motivation within public sector organizations” (LeGrand, 2003; Moynihan, 2010; Arellano-Gault & Lepore, 2011). With this in mind it is not so clear what the effect of NPM in public organizations will be.
In the Netherlands municipalities are able to choose their own management system. However, central government stimulates to increase the use of results-oriented controls (Hengel et al., 2014). With the budget cuts for municipalities and the ongoing depressing economic situation, one could argue that the use of output controls has increased in the last years.
2.2 Management Controls
The paper of Malmi and Brown (2008) contains a number of different definitions and descriptions of management control systems (hereinafter: “MCS”) and management controls. There are multiple ways of interpreting management control systems, management control packages and management control elements. In this paper, the main focus is the management control package (hereinafter: “MCP”), and in particular its various elements. First, the theory of MCPs and its various elements will be explained by using the theory of management controls of Merchant and van der Stede (2007). Second, the framework of Ouchi (1979) will be applied to select differences between the governmental departments, which may lead to the application of different management controls in those departments.
2.2.1 Management Control Packages
In previous research, authors were mostly focused on a single Management Control Element (hereinafter: “MCE”), which Malmi and Brown (2008, p. 290) define as “a system, rule, practice, value or other activity that is in place in order to direct employee behavior””. Malmi and Brown (2008) refer to a MCS when these MCEs are used as complete systems. Accordingly, the sum of these MCEs and MCSs are called a MCP. The focus in this thesis will be on the MCPs, which are defined by using the framework of Merchant and Van der Stede (2007). According to Sandelin (2008), one of the strengths of this framework is that it is broad enough to include all different management controls while also providing sufficient rigidity. Moreover, while some frameworks lack the vision and mission of the organization (Otley, 1999), Merchant and van der Stede (2007) include this in their framework. Simons (1995) argues that mission and vision of an organization can play a key part in management control. As a point of critique, Sandelin (2008, p. 326) mentions that the framework of Merchant and van der Stede (2007) “lacks specificity in explaining the couplings between control elements”. Since this is not the focus of this thesis, this is not seen as an irreconcilable argument to not use this framework.
able to determine what results are desired in the controlled area. Second, the employees should have significant influence on the results they are held accountable for. Third, organizations need to be able to measure the results effectively. The results controls clarify what is expected from employees and the rewards are directly coupled with performance, which should work motivating for employees. A possible indirect effect of results control can be seen in the retaining and attraction of the best employees.
The second type of management controls are action controls, which are defined as: “the most direct form of management control because they involve taking steps to ensure that employees act in the organization’s best interest by making their actions themselves the focus of control” (Merchant & van der Stede, 2007, p. 76). These steps can be rules or procedures which guide the actions of employees such that it leads to the desired results. Sometimes rules or procedures are in place to ensure that the employees do not perform actions that can be harmful for the organization. These controls are effective when the organization can determine what actions are beneficial or harmful, and that these actions can be influenced by the organization. Merchant and van der Stede (2007) define four forms of action controls, namely behavioral constrains, pre-action reviews, action accountability and redundancy.
The third type of management controls are personnel controls, which Merchant and van der Stede (2007, p. 83) define as: “Personnel controls build on employees’ natural tendencies to control and/or motivate themselves”. They serve three purposes; (1) to make sure the employees understand what the organization wants, (2) to ensure that each employee is able to successfully perform their job, and (3) to increase the likelihood that employees engage in self-monitoring. There are three methods to implement these controls. The first method is to devote considerable time and effort to employee selection and placement. The second method is to increase the likelihood of an employee to perform a job successfully, e.g. by providing training. The last method is to make sure that the job is designed such that motivated and qualified employees achieve a high probability of success.
Where personnel controls aim at encouraging employee self-monitoring, cultural controls encourage mutual monitoring. This is a form of group pressure that decreases the likelihood of individuals to deviate from group norms and values. Merchant and van der Stede (2007, p. 85) describe cultures and cultural norms as follows: “Cultures are built on shared traditions, norms, beliefs, values, ideologies, attitudes and ways of behaving. The cultural norms are written and unwritten rules that govern employees’ behavior”. There are ways of shaping a culture, namely codes of conduct, group rewards, intra-organizational transfers, and setting the tone at the top. Two of them will be briefly explained, to provide an overview about what their effect can be. Group rewards provide rewards or incentives on collective achievement. They encourage teamwork and stimulate peer pressure among groups. An intra-organizational transfer, or employee rotation, improves the socialization of employees to transmit culture in the organization.
2.2.2 The framework of Ouchi
knowledge of the process which transforms inputs into outputs is of great relevance. Therefore, these two determinants are the two scales of the framework, on which an individual can score high/low and perfect/imperfect respectively.
In short, market mechanisms are control mechanisms directed towards efficiency, where prices convey all of the information necessary for efficient decision making. Market controls reward individual contribution based on the information of the market. The availability of this information is one of the requirements of market mechanisms. The social requirement is minimal and exists of a norm of reciprocity to keep the market prices correct. The norm of reciprocity implies that when someone cheats, he will be punished by the whole social system. Concluding, market controls are able to measure the individual contribution, which makes it possible to reward accordingly. These controls can be compared with the results controls of Merchant and van der Stede (2007). They both use output control, and reward according to the measured output. Moreover, both controls have large information prerequisites to be able to measure all that is needed to control appropriately.
Bureaucratic controls can be used when market controls fail. These controls include a legitimate
authority and employees receive a payment in exchange for a decrease in autonomy. A
bureaucratic mechanism is directed towards explicit routines of monitoring and the guidance of subordinates by superiors. It involves close personal surveillance of employees. Rules contain the information necessary for successful task completion. Bureaucratic controls can be compared to the action controls of Merchant and van der Stede (2007), due to the similarity in the level of social requirements and the focus on beneficial actions to the organization.
Clan controls use an informal social structure and socialization processes. Next to reciprocity and a legitimate authority, these controls require a social agreement on personal values and beliefs. This mechanism can only rely on selection and screening to a limited extent. Therefore, the
organization might rely on training to get the desired skills and values. The clan mechanisms can
be compared with cultural controls and to a lesser extent also to personnel controls. Personnel controls are similar in the way that it also views intrinsic motivation as main motivator of employees, and that one of the major methods of implementation is through training. Cultural controls are similar in a way that they both rely on shared values and beliefs in the society to monitor employees, which is also the primary focus of clan mechanisms.
This thesis assumes, based on Ouchi (1979), that when the measurability of the outputs is high and the knowledge of the transformation process is perfect, market and bureaucratic controls (results controls and action controls) are used more intensively, and when the measurability is low and the knowledge is imperfect, clan controls (cultural controls and personnel controls) are used more intensively.
2.2.3 Employee perception of controls
theory of Simon’s framework. Important for this thesis is the ambiguity between the dual role of controls and perception of controls that Tessier and Otley (2012) identify. This deals with the difference between the intentions that managers have for controls (enabling/coercive) and employees’ perceptions of these controls. The perception of controls implies that controls may be interpreted on a different way by employees (Scott, 2001) and that therefore employees in the same department possibly perceive the controls differently.
2.3 Motivation
Motivation of employees is a key topic in psychology and management accounting literature. The literature of motivation is diverse and contains many different theories, such as contingent theories, the cognitive dissonance theory, the organizational justice theory and content theories. This thesis focuses on if management controls influence the way employees are motivated. The self-determination theory (Deci & Ryan, 1985) is concerned with the motivation behind the decisions and choices that people make. Furthermore, this theory divides motivation in intrinsic motivation and extrinsic motivation. This breakdown of motivation enables the ability to measure the difference in the degree to which an employee is self-motivated and self-determined when they are affected by different levels of management controls.
“To be motivated means to be moved to do something” (Ryan & Deci, 2000, p. 54). Motivation is a psychological force that enables action. People differ in the levels of their motivation and have different kinds of orientation of motivation. Porter and Lawler (1968) build on Vroom’s (1964) expectancy–valence theory of motivation and proposed a model of intrinsic and extrinsic work motivation. Gagné and Deci (2005, p. 331) use the following definitions: “Intrinsic motivation involves people doing an activity because they find it interesting and derive spontaneous satisfaction from the activity itself. Extrinsic motivation, in contrast, requires an instrumentality between the activity and some separable consequences such as tangible or verbal rewards, so satisfaction comes not from the activity itself but rather from the extrinsic consequences to which the activity leads” (Porter and Lawler, 1968). Hence, intrinsic motivation comes from within; it is based on the satisfaction that an individual can derive from an activity without receiving external rewards. Extrinsic motivation refers to doing something that is value-adding for someone else, which may result in recognition or rewards for the executor. Research has shown that the quality of experience and performance can differ when one is behaving according to intrinsic versus extrinsic motives (Ryan & Deci, 2000). The theory behind NPM, i.e. the agency theory, mainly bases its assumptions on extrinsic motivation.
motivation, leading to a shift in attention from the task-or normative goals to the reward. The most cited theoretical underpinning of this effect is the self-determination theory from Deci and Ryan.
2.3.1 Management Controls on Motivation
Managers use management controls to influence their employees’ behavior and performance. Motivation can be used as a mediating factor for management controls to influence behavior and performance. Frey et al. (2013, p. 950) notes for example that that one of the major characteristics of the reforms of NPM is to raise public servants’ motivation. The common conclusion about the relationship between management controls and motivation is that salary increase and other extrinsic rewards invoke extrinsic motivation. The link of motivation with management controls is briefly made by Merchant and van der Stede (2007) who state that step three and four (of the four steps) in implementing results controls influences motivation (Merchant & van der Stede, 2007, p. 25, 28, 29). Step three is about setting performance targets, which may improve motivation by providing clear goals for employees to strive for and allow employees to assess their performance. The fourth step is about providing external rewards to encourage behavior that leads to desired results.
2.3.1.1 Setting targets
According to Merchant and van der Stede (2007), results controls allow for setting performance targets. Setting performance targets influences behavior in two ways; first it enhances motivation by providing clear goals for employees, and second it allows employees to assess their performance. The goal-setting theory of Locke (1988) is very similar to the theory behind setting targets. Locke argues that motivation increases when targets are set. He defines goals as immediate regulators of human action, which when set results in clear actions for employees. Accordingly, goal setting can be used as a motivational tool. It is an external way of providing guidance to human action and enhances the extrinsic motivation of employees. Deci et al. (1999) tie previous research together in their meta-analytic review and amongst other they conclude that events such as deadlines, evaluations and imposed goals have been found to undermine intrinsic motivation. Other events such as the provisions of choice and the acknowledgement of feelings have been found to enhance intrinsic motivation (Smith, 1975; Amabile et al., 1976; Zuckerman et al., 1978; Mossholder, 1980; Koestner et al., 1984). Furthermore, Miao (2007) argued that autonomy is reduced by output control and setting performance measures, which has consequences for intrinsic motivation. This is confirmed by Cassel and Jain (2012), who conclude that a reduction of autonomy leads to a reduction of intrinsic motivation. Hence, setting targets has a decreasing effect on intrinsic motivation, but increases extrinsic motivation.
2.3.1.2 Providing rewards
2.4 Summary and hypotheses development
In this paragraph, the literature and treated theories discussed in this section will be summarized, and assumptions will be formulated into multiple hypotheses. This thesis investigates the influence of NPM on management controls, and management controls on motivation in municipalities. Multiple theories and assumptions are used and expressed to analyze this relationship, and to display this in a clear manner the hypotheses are explained step by step, starting with NPM, followed by management controls and motivation.
The literature review starts with briefly explaining the theory of NPM and its relation to management controls. It is expected that NPM has increased the use of results controls relative to the other controls of Merchant and van der Stede (2007).
H1: The use of results controls is higher than the use of action controls, cultural controls or personnel controls.
Thereafter, management Controls (Merchant and van der Stede, 2007) and motivation (self-determination theory; Deci & Ryan, 1985) are explained. The self-(self-determination theory splits motivation in two parts, namely intrinsic motivation and extrinsic motivation. Intrinsic motivation is based on the satisfaction an individual can derive from an activity, without receiving external rewards. Extrinsic motivation refers to doing something that is value-adding for someone else, which will possibly result in recognition or rewards for the executor. The theoretical basis of NPM is the principal-agent view, which assumes that individuals are economically rational. In this view, it is assumed that extrinsic incentives motivate individuals’ additional or marginal efforts. The following hypothesis can be formulated to test if this assumption of NPM holds.
H2: Employees of municipalities are more extrinsically motivated than intrinsically motivated.
Furthermore, Merchant and van der Stede (2007) distinguish four types of management controls, namely results controls, action controls, personnel controls and cultural controls. Results controls are the primary focus in their framework and of this thesis. There are four steps identified in implementing result controls, where the last two steps (setting targets and providing rewards) are assumed to have an influence on motivation. It is expected that setting targets has an increasing effect on extrinsic motivation, but decreases intrinsic motivation.
H3: Setting performance targets increases the level of extrinsic motivation of municipality employees. H4: Setting performance targets decreases the level of intrinsic motivation of municipality employees. Moreover, providing external rewards has, just like setting targets, an enhancing effect on extrinsic motivation and a reducing effect on the intrinsic motivation of employees.
These six hypotheses summarize the main relationship of this thesis. This is graphically displayed in the conceptual model in figure 1.
>
NPM Results
Controls
Setting targets Providing rewards Intrinsic motivation Extrinsic motivation
+
H1 H4+
H3 H5+
-H6 H2-Figure 1: Conceptual model of the main relationship
H7: The respondents scoring high/perfect on Ouchi’s scale perceive themselves relatively more results- and action controlled than respondents scoring low/imperfect on Ouchi’s scale.
3. Methodology
In this section the research method, the data collection, the actions preceding the statistical analysis and the statistical analysis itself will be described. According to Yin (1994) and Swanborn (1996) the most important quality criteria for research are controllability, reliability and validity, where controllability is a prerequisite for the evaluation of validity and reliability of a study (van Aken et al., 2012). Furthermore, Gill & Johnson (2002) and Saunders et al. (2009) note that an important issue to ensure reliability is to use a structured methodology to facilitate replication. The methodology section starts with a discussion of the research method, where the concepts of validity and reliability are treated in more detail. This is followed by an explanation of the data collection and the sample population. Thereafter, the design of the survey of this thesis will be discussed by explaining the constructs used to measure the variables. This will be followed by a detailed description of the statistical analysis that will is performed.
3.1 Research method
Cunliffe (2011) argues that objectivist research tends to take a macro perspective and studies organizations at a structural, strategic and societal/environmental level, drawing generalizations about group and individual behavior. This thesis takes an objectivism approach to research. Hence, the objective is to generate findings that are generalizable within the context and theme of the research.
Generalizability is described as external validity in the book of Saunders et al. (2009), who describe validity as: “Validity is concerned with whether the findings are really about what they appear to be about”. Next to validity, there is the concept of reliability. Reliability is described as “the extent to which your data collection techniques or analysis procedures will yield consistent findings” (Saunders et al., 2009). Based on research from Robson (2002), Saunders et al. (2009) one could argue that there are four threats to reliability; (1) participant error, (2) participant bias, (3) observer error and (4) observer bias. These threats deal with times of conducting interviews, anonymity, interview structures and ways of interpreting replies. This thesis uses an internet survey to collect data and the respondents can decide themselves when to fill in the survey. The respondent is therefore anonymous, the structure is fixed and the questions are closed and thus not multi interpretable. Furthermore, there is no missing data since it is not possible to finish the survey before answering all the questions. Consequently, the threats to reliability do not apply in this thesis.
3.2 Data collection
Saunders et al. (2009) describe a survey strategy as a popular method to collect large amounts of data. A survey collects quantitative data that can be analyzed statistically and be used to suggest reasons for specific relationships between variables. The survey used for this thesis uses a 7-point Likert scale - strongly disagree to strongly agree - to test the hypotheses. This scale is developed by Likert (1932) and produces ordinal data. However, in this thesis it is used to produce interval data, which is done frequently in research (Pell, 2005; Cooper & Schindler, 2006). There is a difference between Likert-type data and Likert-scale data (Boone & Boone, 2012). Likert-type data are singular measurements and fall into the ordinal measurement scale, whereas Likert-scale data are created by calculating a composite score from four or more type Likert-type items and can therefore be analyzed as interval data. Likert scaling measures either positive or negative reactions to a statement; this is also called bipolar scaling. The respondent is asked to give a numerical score that reflects his or her feelings to the statement. These statements are based on constructs that are developed to measure the variables. In this thesis, IBM SPSS statistics version 22 is used to statistically analyze the data.
The targeted populations are employees of the Dutch municipalities Emmen, Groningen and Zwolle. Helden (1998) writes in his article that municipalities with less inhabitants than 15,000 employees, personal direct controls will be preferred to more formalized systems of planning and control. The sample in this thesis consists of three large municipalities with more than 100,000 inhabitants and therefore well in excess of the 15,000 mentioned. Specifications about the departments surveyed can be found in table 1 below. The response rate of the survey is 41,78% (150 respondents from a population of 359). Two departments have a very low response rate (<20%). However, this thesis focuses on individual responses and therefore these departments are still included in the sample. The data was collected in the period of 6 May 2014 up to and including 5 June 2014, which is approximately 4 weeks. Two reminders have been sent to increase the response rate.
Department
Respondents/employees Percentages
Emmen: Policy and control 37/66 56,06%
Emmen: Towns and districts 57/104 54,81%
Groningen: Public administration 10/57 17,54%
Groningen: Policy and Programming 24/53 45,28%
Zwolle: Construction and management 7/44 15,91%
Zwolle: Social development 15/35 42,86%
Total 150/359 41,78%
Table 1: Sample population department and survey responses
3.3 Measurements
3.3.1 Dependent variables
Motivation (intrinsic and extrinsic) is the dependent variable in this model. Constructs of Spreitzer (1995), Gould-Williams and Davies (2005) and Gagné et al. (2010) are used to measure intrinsic motivation (see Appendix A). Spreitzer (1995) developed a construct to measure the variable he described as “meaning”. Meaning is the value of a work goal or purpose judged in relation to an individual's own ideals or standards (Thomas & Velthouse, 1990) and results in high commitment and concentration of energy (Kanter, 1984). Gould-Williams and Davies (2005) study the exchange relationships between managers and public sector employees working in government departments. One of the outcomes of this exchange relationship is employee motivation, which is measured with two items generating a Cronbach’s alpha of 0,66. The Cronbach's alpha (α) is a coefficient of internal consistency (Cronbach, 1951) and can be seen as a measurement of reliability. The higher the alpha is, the more reliable the developed scale is. Some researchers use a requirement of an alpha above 0,70 as a rule of thumb (Nunnally, 1978). However, Nunnally also argued that an alpha of 0,6 is acceptable. In addition to the two mentioned constructs, the paper of Gagné et al. (2010) is used. Their Motivation at Work Scale (hereinafter: “MAWS”) – which is based on the framework of self-determination theory of Deci and Ryan (1985) – contains three questions for measuring intrinsic motivation (α = 0,89). De Jong and van der Kolk (2013) combine the two constructs of Spreitzer (1995) and Gagné et al. (2010) in a study of a public organization in the Netherlands, which resulted in a Cronbach’s alpha of 0,76 for intrinsic motivation.
To measure extrinsic motivation constructs of Hemnessey and Amabile (2005), Gagné et al. (2010) and von Bondsdorff (2011) are used (see Appendix A). De Jong and van der Kolk (2013) also use these three constructs for measuring extrinsic motivation at a governmental organization. The combination of these constructs resulted in a Cronbach’s alpha of 0,665. Hemnessey and Amabile (2005) provide definitions for intrinsic and extrinsic motivation and argue that there is experimental evidence that extrinsic motivators in the social environment can undermine intrinsic motivation. Moreover, they provide evidence that rewards can enhance intrinsic motivation through a process of motivational synergy. The MAWS in article of Gangé et al. (2010) - which is also used for intrinsic motivation - contains three questions for measuring extrinsic motivation (α = 0,69). Furthermore, the paper of von Bondsdorff (2011) investigates reward preferences of nurses of different ages and results shows that non-financial rewarding elements are highly appreciated.
3.3.2 Independent variables
The measurements of management controls are specifically developed for this thesis and context, except for the variable setting targets and providing rewards. Kohli et al. (1998) performed a study in a sales context about the moderation of salesperson experience on the influence of supervisors on the (goal) orientations of sales people. To be able to measure their research question, several questions are developed to measure output control and its components, including setting targets and rewarding performance. These questions are converted to the context of this thesis. The rest of the measurements are designed by using elements that Merchant and van der Stede (2007) use to describe the various elements of the MCPs.
3.3.3 Control variables
Age and years of service are continuous variables, while gender (female = 0, male = 1) and education (VMBO or lower =1, Havo/VWO = 2, MBO = 3, HBO = 4, WO =5) are categorical.
3.4 Statistical analysis
Figure 1 in paragraph 2.4 graphically shows that multiple relationships are being measured. The first relationship measures the effect of NPM on management controls, whether the results controls are used relatively more compared to the other management controls. The second relationship will focus on the effects of results controls on motivation. This last relationship contains two dependent variables; intrinsic motivation and extrinsic motivation. To be able to statistically test this relationship it is divided into two models: a model of intrinsic motivation (IM) and a model of extrinsic motivation (EM):
SETTARG stands for setting targets, PROVREW for providing rewards, AGE for age, EDU for education, GEN for gender and YOS for years of service.
3.4.1. Assumptions
A check for outliers is conducted before beginning with the analyses of the data. “An outlier is a score very different from the rest of the data. When we analyze data we have to be aware of such values because they bias the model we fit to the data” (Field, 2005). Outliers bias the mean and inflate the standard deviation, and should be transformed or removed.
The next step is to test several assumptions regarding parametric tests. “A parametric test is one that requires data from one of the large catalogue of distributions that statisticians have described and for data to be parametric certain assumptions must be true” (Field, 2005). Testing these assumptions is important, since doing the wrong statistical test can lead to inaccurate results. There are four assumptions of parametric tests according to Field (2005):
1. Normally distributed data: this rationale behind hypotheses testing relies on having normally distributed populations. The logic behind hypotheses testing is flawed if this assumption is not met.
2. Homogeneity of variance: this is also called homoscedasticity, which means that the variances should be the same throughout the data.
3. Interval data: the distance between points of your scale should be equal at all parts along the scale.
4. Independence: the sample consists of independent observations.
values of sknewss and kurtosis should be generated. The values of skewness and kurtosis should be around zero in a normal distribution. The skewness is a measure of the symmetry of a frequency distribution, while kurtosis measures the peakedness (leptokurtic) or flatness (platykurtic) in the tails of a frequency distribution.
The assumption of homogeneity of variance or homoscedasticity means that as you go through levels of one variable, the variance of the other should not change. This assumption can be checked by looking at the residual scatter plot. However, this will not test the homoscedasticity statistically. To test this assumption statistically, an equal variance test of Koenker and Basset (1982) is performed. Homoscedasticity is rejected if the variable turns out to be statistically significant. The aim is homoscedasticity and thus insignificance.
The third and fourth assumptions, interval data and independence, are according to Field (2005) only tested by common sense. However, it is also common to conduct a correlation analysis to check if the sample consists of independent observations. There is a possibility that multicollinearity is present, which exists when there is a strong correlation between two or more predictors in a regression model and leads to unreliable outcomes. When the correlations are above ρ=0.8, multicollinearity is a problem and the variable should be considered to be deleted from the model. Furthermore, when the correlations are between ρ=0.3 en ρ=0.8 additional checks are required. One of the possible collinearity diagnostics is the variance inflation factor (VIF). The VIF indicates if a predictor has a strong linear relationship with the other predictors. There is some disagreement in the literature about the values of the VIF. The most common is the boundary of 0,2 to 10 of the VIF (Myers, 1990; Menard, 1995; Pallant, 2010; Saunders et al., 2009). However, some authors argue that the value should not be above 5 (e.g. Rogerson, 2001), and sometimes not even above 4 (e.g. Pan & Jackson, 2008). Hence, there is not much consensus about the height of the boundaries. They should be interpreted as a warning. A high VIF indicates that the multiple correlation with other variables is high, which suggest the possibility of multicollinearity (Pallant, 2010).
3.4.2 Analysis
The analysis starts with generating descriptive statistics to describe the characteristics of the sample. These statistics contain items as the mean, minimum and maximum and the standard deviation to provide a first indication of the direction of the tested relationships. Second, a correlation analysis is conducted to test whether the relations are significant or not. The most common correlation analysis is the Pearson correlation measure, which measures the linear correlation between two variables and gives a value between +1 (total positive) and -1 (total negative), where 0 implies that there is no correlation. There are different interpretations to values between 0 and 1. Cohen (1988) suggests to interpret 0,1 to 0,29 as a small correlation, 0,3 to 0,49 as a medium correlation and 0,50 to 1 as a strong correlation.
4. Results
This section contains a check of outliers, multiple tests of the assumptions, descriptive statistics, correlation analysis and statistical analyses. Further analysis of the results can be found in the next chapter. First, a reliability statistic is generated to check whether the constructs are internally reliable or not. Table 2 below displays the Cronbach’s alpha and the number of Likert-items per construct. It can be concluded that the variables have an acceptable alpha according to the minimum of 0,6 (Nunnally, 1978). The alpha of the developed scale for intrinsic motivation, setting targets, action controls and cultural controls appear to be extremely reliable. The constructs are generated by calculating the average of the total sum of the items.
Variable name
Cronbach’s Alpha
Nr. of items
Dependent variable: Intrinsic Motivation ,907 8 Extrinsic Motivation ,657 5 Independent variable: Setting targets ,919 5 Providing rewards ,770 2 Action controls ,819 10 Cultural controls ,810 4 Personnel controls ,687 5 Results controls ,740 8
Table 2: Reliability statistics
4.1 Outliers and assumptions
Box plots are generated to check if the data contains extreme values. These box plots are included in Appendix B. First, the dependent variables are plotted separately. The boxplot of intrinsic motivation shows a few outliers, but more importantly also extreme values. Extreme values can make an incorrect drawing of the results and should be deleted or transformed. After investigating the response of the two cases in more detail, it is concluded that these responses should both be deleted. Consequently, from now on the number of respondents is 148. The boxplot of extrinsic motivation shows one outlier and no extreme values. The third graph displays the boxplots of the independent variables and these variables contain a few outliers and no any extreme values.
-0,5) and approximately symmetric for extrinsic motivation (between -0,5 and 0,5). All values are within the range of -1 and 1 and are not treated as abnormal. While the kurtosis value of extrinsic motivation is close to zero (-0,055) and not a danger for the distribution, the kurtosis value of intrinsic motivation is above 0,5. According to Tabachnick and Fidell (2012), kurtosis values above zero indicate a distribution that is too peaked with short, thick tails. However, in large samples (100 or more cases) the variance associated with positive kurtosis disappears.
To test the second assumption, homogeneity of variance, a residual scatter plot is generated of the models of IM and EM (Appendix D). By looking at these residual scatter plots it is possible to test homoscedasticity with a visualization approach. If the variance of the residuals is constant, a residual scatter plots shows a constant spread around the mean. From the scatterplots of IM and EM, it can be concluded that the data is spread around the mean. However, there is some variance. To decide if there is homoscedasticity or heteroscendasticity on a significant statistical way, a Koenker-Basset (1982) test is executed. The results are 10,788 (0,095) and 1,162 (0,979) for intrinsic motivation and extrinsic motivation respectively. This significance is of importance in this test, which is reported in the parentheses. It is preferred that the result is not significant (on a 95% significance level), to reject heteroscendasticity and accept homoscedasticity. The significance values of intrinsic motivation and extrinsic motivation are both above 0,05. Hence, homoscedasticity is not rejected.
The third and last test of assumptions is to detect whether the observations are independent. To detect this, a correlation analysis of all the variables can be found in table 5. As mentioned in paragraph 3.4.1, variables that have correlations between ρ=0.3 en ρ=0.8 and are used in one formula or analysis should be tested for the VIF. The VIF indicates whether a predictor has a strong linear relationship with the other predictors and its value should be above 0,1 and below 4 (Myers, 1990; Menard, 1995; Pallant, 2010). As can be seen in table 5, there are many significant and high correlating variables. However, the ones important for the main analysis of this thesis are the correlations between providing rewards and setting targets (ρ = 0,524), age and years of service (ρ = 0,344), and education and years of service (ρ = -0,388). Next to the variables of the main analysis, the variables of the scale Ouchi (1979) and the management controls correlate with each other. Therefore, all these variables are tested for the VIF. The results of the variables of the main relationship are displayed in Appendix E. According to the results of the VIFs it can be concluded that there is no multicollinearity present. All the values are below 1,6 and above 1, which is within the boundaries of 0,1 and 4. The VIF values of the other strong correlations (action controls, cultural controls, personnel controls, results controls, measurability and knowledge) are all below 3, which is also within the boundaries.
4.2 Descriptive statistics
4.2.1 Main variables
the sample consists has an average age of 48 years with a minimum age of 22 and maximum of 63. Furthermore, the average of years of service is around 10 years.
Table 4 contains the frequencies of the two categorical control variables. There are more male respondents than female respondents (71,8% and 28,2% respectively), and most of the respondents are educated with a HBO certificate or higher (59,2%).
N Minimum Maximum Mean Std. Deviation
Intrinsic motivation 148 2,9 7 5,47 0,89 Extrinsic motivation 148 1 6,4 3,64 1,04 Setting targets 148 1 7 3,45 1,43 Providing rewards 148 1 7 1,98 1,24 Action controls 148 1,6 6,3 4,01 0,98 Cultural controls 148 1 6,5 4,40 1,22 Personnel controls 148 1,4 6,8 4,52 0,90 Results controls 148 1,8 6,1 4,17 0,84 AGE 148 22 63 48,07 10,18 YOS 148 0 41 10,25 10,45
Table 3: Descriptive statistics numerical variables
Frequency Percent Cumulative Percent Gender* Female 43 29,1 29,1 Male 105 70,9 100 Total 148 100 Education* WO 37 25 25 HBO 50 33,8 58,8 MBO 47 31,7 90,5 Havo/VWO 9 6,1 96,6 VMBO or lower 5 3,4 100,0 Total 148 100
Table 4: Frequences of categorical variables
*In the variable GEN male=1, female=0. In the variable EDU WO=6, HBO=5, MBO=4, Havo/VWO=3, VMBO or lower=1
4.2.2 Departments
Appendix F contains two tables with the descriptives of all variables per department. Notable is that the Likert-constructs of the high/perfect departments within their own municipalities rate everything higher than low/imperfect departments except for extrinsic motivation, cultural controls in Groningen, and measurability in Emmen. The frequency table makes it possible to conclude that, for example, the department of Towns and districts in Emmen has relatively low educated employees.
the first table in appendix F. The high/perfect departments score higher on the Ouchi scale than the low/imperfect departments, except for the measurability of results in the department of Policy and control in Emmen. The binary scores make it possible to measure when a department is high/perfect (both positive scores) and low/imperfect (both negative scores). Respondents who rate one scale high and the other low are not considered. Moreover, respondents that have rated once a four (neutral) on the Likert-item are also not considered. Next, the two measurements are added together in one Ouchi variable, which is 0 when low/imperfect and 1 when high/perfect. This transformation has led to the number of cases going down from 148 to 75.
4.3 Correlation analysis
The correlation analysis is included in table 5 below. While some expected correlations are not significant there are some notable unexpected correlations. All measured variables are included in this analysis and therefore not all significant correlations are in the scope of this thesis. For example, intrinsic motivation, setting targets and providing rewards all have high significant correlations with all of the management controls. This is outside the scope of this thesis and is therefore not further analyzed.
First, the significant correlations of the main relationship are discussed. Intrinsic motivation has a significant and small correlation with setting targets (on the 99% significance level) and providing rewards (on the 95% significance level). However, the correlations are both positive while it is hypothesized that intrinsic motivation would reduce as setting targets and providing rewards increased. Extrinsic motivation does not have a significant correlation with setting targets and providing rewards. There is a small negative correlation of extrinsic motivation with age, which implies that the older an individual becomes the less extrinsically motivated they are. Setting targets has a strong positive correlation with providing rewards, which is according to the expectation as setting targets and providing rewards both are part of results controls (Merchant and van der Stede, 2007). Hence, the presence of target setting implies the presence of results controls, explaining the presence of the providing of rewards. Furthermore, there are two correlations of the control variables age and education with years of service. The positive correlation of age with years of service can be explained by considering the higher chance that an older employee has served the municipality longer. The negative correlation of education with years of service may be explained by the possibility that employees who have a lower education tend to stick longer at one department, or that higher educated employees tend to leave the departments or municipalities.
is according to the expectation that when employees score high/perfect on the Ouchi scale there are more action controls and results controls. However, the correlation of the knowledge of transformation process with cultural controls and personnel controls are also positive and of medium strength, while the measurability of outputs has a positive correlation with personnel controls. Concluding, the management controls are all perceived stronger by employees that perceive themselves in a high/perfect department compared to low/imperfect.
IM EM
SET TARG
PROV
REW AC CC PC RC MEAS KNOW AGE EDU GEN YO S IM 1 EM ,050 1 SETTARG ,283** -,043 1 PROVREW ,210* -,066 ,524** 1 AC ,421** ,019 ,569** ,429** 1 CC ,487** -,070 ,496** ,277** ,491** 1 PC ,595** -,08 ,469** ,388** ,657** ,519** 1 RC ,446** -,053 ,627** ,419** ,717** ,547** ,674** 1 MEAS ,092 ,098 ,090 ,064 ,345** -,038 ,265** ,270** 1 KNOWL ,254** ,041 ,478** ,222** ,398** ,361** ,384** ,460** ,461** 1 AGE -0,010 -,231** -,028 -,066 ,035 -,061 ,131 ,155 ,089 -,089 1 EDU ,127 -,029 ,097 ,130 -,128 ,241** ,011 -,008 -,296** ,010 -,188* 1 GEN ,012 ,189* -,135 -,047 -,058 ,004 -,042 -,074 -,012 -,005 ,209* -,125 1 YOS -,058 ,076 -,193* -,136 -,039 -,344** -,091 -,060 ,122 -,191* ,344** -,388** ,098 1 Table 5: Pearson correlation analysis
**, Correlation is significant at the 0.01 level (2-tailed). *, Correlation is significant at the 0.05 level (2-tailed).
In this table AC stands for action controls, CC stands for cultural controls, PC stands for personnel controls and RC stands for results controls. MEAS and KNOW are the scales of measurability of outputs and knowledge of transformation process of Ouchi’s framework.
4.4 Hypotheses testing
In this paragraph, the hypotheses are statistically analyzed. A regression analysis is performed for most of the hypotheses, except for hypothesis one and two. To analyze hypothesis one and two a paired-sample T test will be conducted. The analyses are based on 148 observations.
4.4.1 Comparing means
The first hypothesis is formulated as follows:
H1: The use of results controls is higher than the use of action controls, cultural controls or personnel controls.
is higher. The mean of the cultural controls and action controls significantly exceed the mean of results controls. Therefore hypothesis one, which hypothesized that the use of results controls exceeds the use of every other form of control, is rejected.
Mean Std. Deviation Std. Error Mean t Sig. (2-tailed) Pair 1: Results controls -
Action controls ,157 ,698 ,057 2,740 ,007
Pair 2: Results controls -
Cultural controls -,231 1,040 ,086 -2,698 ,008
Pair 3: Results controls -
Personnel controls -,346 ,703 ,058 -5,986 ,000
Table 6: Paired sample T-test hypothesis 1.
The second hypothesis is formulated as follows:
H2: Employees of municipalities are more extrinsically motivated than intrinsically motivated.
To test this hypothesis the means of extrinsic motivation and intrinsic motivation are statistically compared. The results of the test are displayed in table 7 below. The mean of intrinsic motivation exceeds the mean of extrinsic motivation by far, as is concluded from the descriptive statistics. The difference is significant, which means that hypothesis 2 is rejected.
Mean Std. Deviation Std. Error Mean t Sig. (2-tailed) Extrinsic motivation - Intrinsic motivation -1,827 1,345 ,112 16,354 ,000
Table 7: Paired sample T-test hypothesis 2.
4.4.2 Regression analysis
The following hypotheses test the relationship between management controls and motivation and in particular the relationship between results controls and motivation. The formulas of IM and EM as described in paragraph 3.4 are used. A regression analysis is conducted to test the control variables before testing the hypotheses. The results show a low R-squared for extrinsic motivation (14,4%) and intrinsic motivation (1,7%). The R-squared (R²) displays the amount of variance explained by the model. Age and gender have a significant effect on extrinsic motivation on the 99% level, where age has a small negative influence on extrinsic motivation and gender a strong positive influence. Years of service has a significant effect on extrinsic motivation on the 90% level and has a small positive influence. The effects of the control variables on the dependent variables are analyzed, which enables the regression analysis of the hypotheses. Hypothesis three and four test the enhancing effect of setting targets on extrinsic motivation and the reducing effect on intrinsic motivation. The hypotheses are formulated as follows:
The results of the regression analysis of hypothesis three and four are displayed in table 8 below. The R² is still low for both models. However, there is an increase for intrinsic motivation compared to the model with only control variables. It is hypothesized that setting performance targets increases the extrinsic motivation of the employees. Nevertheless, in the regression analysis for extrinsic motivation the variable SETTARG is not significant and the R² has not increased or decreased compared to the regression with only the control variables. Hence, hypothesis three is rejected. Furthermore, it is hypothesized that setting performance targets has a decreasing effect on intrinsic motivation. However, the variable SETTARG is positive and significant, which is the opposite from what is expected. This reversed linear relationship leads to the rejection of hypothesis 4.
Intrinsic motivation Extrinsic motivation
(Constant) 4,342*** (,516) 4,805*** (,590) SETTARG ,179*** (,051) ,009 (,058) AGE ,0 (,008) -,036*** (,009) EDU ,107 (,076) -,008 (,087) GEN ,123 (,161) ,572*** (,189) YOS ,003 (,008) ,017* (,009) R Squared ,096 ,144
Table 8: Regression analysis hypothesis 3 and 4.
Standard errors are reported in parentheses. *, **, *** indicates significance at the 90%, 95%, and 99% level, respectively.
The next two hypotheses formulate the expected relationship of providing rewards to employees with intrinsic motivation and extrinsic motivation.
H5: Providing rewards increases the level of extrinsic motivation of municipality employees. H6: Providing rewards decreases the level of intrinsic motivation of municipality employees.
A regression analysis is executed with providing rewards as dependent variable. The results are displayed in table 9 below. Again the R² is low for both regressions. The dependent variable, PROVREW, is significant in the model of intrinsic motivation. However, it is positive while it is hypothesized to be negative. Hence, there is also a reverse linear relationship existent for providing rewards and intrinsic motivation. In the model for extrinsic motivation the variable for providing rewards is not significant and the R² has barely changed. Hence, both hypotheses are rejected.
Intrinsic motivation Extrinsic motivation
(Constant) 4,699*** (,507) 4,920*** (,567) PROVREW ,143** (,060) -,046 (,066) AGE ,002 (,008) -,036*** (,009) EDU ,096 (,078) -,004 (,087) GEN ,061 (,164) ,563*** (,187) YOS ,0 (,008) ,016* (0,009) R Squared ,056 ,147
Table 9: Regression analysis hypothesis 5 and 6.
The preceding regressions tested the hypothesized relationships that are displayed in the conceptual model. In addition, this thesis analyzes the relationship of the framework of organizational control of Ouchi (1979) in combination with the management controls of Merchant and van der Stede (2007). Two hypotheses are developed to measure if these management controls can be assigned to one of the control mechanisms according to the framework of Ouchi (1979). The hypotheses are tested according to the theory of employee perception of Tessier and Otley (2012), as explained in paragraph 4.2.2. When a respondent feels that the result of his work is highly measurable and that his managers and colleagues have perfect knowledge of the transformation process of his work, then we assume that his perceived department is high/perfect and vice versa.
Hypothesis seven and eight are formulated as follows:
H7: The respondents scoring high/perfect on Ouchi’s scale perceive themselves relatively more results- and action controlled than respondents scoring low/imperfect on Ouchi’s scale.
H8: The respondents scoring low/imperfect on Ouchi’s scale perceive themselves relatively more cultural- and personnel controlled than respondents scoring high/perfect on Ouchi’s scale.
With the management controls as dependent variables and Ouchi as independent variable, the results of the regression analyses are displayed in table 10 below. The results indicate that the Ouchi variable has a significant positive effect on all the management controls. Consequently, hypothesis seven is accepted, but hypothesis eight is rejected. Employees perceiving that the output of their work is highly measurable and that colleagues and managers have perfect knowledge of the transformation process feel significantly higher controlled by all four management controls compared to employees who perceive the measurability as low and knowledge as imperfect. This effect is less strong for cultural controls and the strongest for action controls. The R² of results controls and action controls are the largest, while cultural controls has by far the lowest R².
Results controls Action controls Cultural controls Personnel controls
(Constant) 2,519*** (,546) 4,336*** (,684) 2,714*** (,961) 3,317*** (,643) OUCHI ,999*** (,157) 1,203*** (,196) ,739*** (,276) ,969*** (,185) AGE ,033*** (,009) ,003 (,012) ,014 (,016) ,019* (,011) EDU -,023 (,086) -,127 (,108) ,27* (,151) -,004 (0,101) GEND -,35* (,184) -,396* (,231) ,098 (,324) -,197 (,217) YOS -,011 (,009) -,014 (,011) -,027* (,016) -,011 (0,011) R Squared ,467 ,410 ,198 ,332
Table 10: Regression analysis hypothesis 7 and 8.
Standard errors are reported in parentheses. *, **, *** indicates significance at the 90%, 95%, and 99% level, respectively
4.5 Summary of results
performed and the assumptions regarding parametric tests are tested. The check for outliers resulted in deleting two cases with extreme values. Moreover, all of the assumptions are met which means that the data is normally distributed, there is homoscedasticity and the observations are independent. The descriptive statistics summarize the sample and provides a first indication of the direction of each variable. Results of the descriptive statistics show contradicting values that do not meet the hypothesized explanation of hypothesis 1 and 2. The last precaution before testing the hypotheses is conducting a correlation analysis. The results of this analysis show some unexpected relations, while some expected relations are not significant. Strong correlations are tested by the VIF, which did not reveal any concerning values.
The first two hypotheses express the assumptions of NPM regarding management controls and motivation. It is expected that the reforms of NPM have led to results controls being the most strongly perceived management controls compared to action controls, cultural controls and personnel controls. However, the results of the Paired sample T-test show that personnel controls are the most used management controls, followed by cultural controls. Hence, hypothesis 1 is rejected. The second hypothesis tests the assumption of NPM regarding economically rational agents. Hence, it is hypothesized that employees are more extrinsically motivated than intrinsically motivated. However, this hypothesis is rejected as the mean of intrinsic motivation is significantly larger than the mean of extrinsic motivation.
The following four hypotheses test the assumptions of Merchant and van der Stede (2007) regarding elements of results controls influencing motivation. These elements of results controls are setting performance targets and providing external rewards. It is hypothesized that setting targets and providing rewards positively influences extrinsic motivation and negatively influences intrinsic motivation. However, the results of the regression analyses reject these four hypotheses. Instead it is found that they both have a positive effect on intrinsic motivation, while there is no significant effect on extrinsic motivation.
These six hypotheses test the main relationship. An additional analysis tests the hypotheses regarding the combination of the organizational framework of Ouchi (1979) and the management controls of Merchant and van der Stede (2007). It is hypothesized that employees perceiving the measurability of the outputs and the knowledge of the transformation process of collegueas and managers as high/perfect or low/imperfect perceive themselves respectively more results-and action controlled or cultural-and personnel controlled. The results of the regression analysis show that the employees who score high/perfect on the scales of Ouchi perceive themselves relatively more controlled by all four management controls than employees who score low/imperfect on the scales.
