Professional Service Firms: a study into the effect of
professionalism and the moderating effect of strategy on control
tightness
Name: Jordy Bakker Student number: 11110961
Thesis supervisor: H. Kloosterman, MSc Date: June 26, 2017
Word count: 15211
MSc Accountancy & Control, specialization Control
Statement of Originality
This document is written by student Jordy Bakker 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 paper investigates what the effect is of contingency variables professionalism and strategy on the design of management control systems (MCS) in professional service firms (PSFs). This paper uses a recently developed framework which recognizes several types of PSFs and conceptualizes MCS as having tight versus loose control. Historically, researchers and theorists have studied and tried to explain the characteristic of PSFs. However, due to the nature of PSFs, limited empirical research has been conducted. Recently, interest from researchers into PSFs has re-emerged, noted was the lack of insight how the contingency variable strategy affects PSFs. Using the framework, I try to empirically test whether professionalism leads to a loose MCS and I try to gain insight how strategy affects the MCS design for PSFs. Data was collected from surveys conducted with employees working on operational and low-management level at PSFs. I conducted t-tests and used General Linear Modelling (GLM) analysis to test my hypotheses. I found (1) no significant evidence that professionalism leads to a loose MCS, (2) evidence that strategy significantly impacts the design of MCS for PSFs, and (3) indication that PSFs are to some extend heterogenous and are to some extend homogenous. This paper concludes with areas for further research.
Keywords: professional service firm; management control system; professionalism; control tightness; strategy; contingency.
Contents 1 Introduction ... 7 1.1 Research objective...7 1.2 Background...7 1.3 Research question ...8 1.4 Motivation...9 1.5 Outline ...9
2 Theory and Hypothesis... 10
2.1 Professional Service Firms ... 10
2.2.1 Characteristics ... 11
2.2.2 Challenges and opportunities ... 13
2.2.3 Organizational responses... 15
2.3 Contingency theory: strategy ... 16
3 Research methodology ... 19 3.1 Research design ... 19 3.2 Sample ... 19 3.3 Variable measurement ... 22 3.4 Analysis ... 33 4 Results ... 34 4.1 Descriptive statistics ... 34 4.2 Analysis – H1 ... 34 4.3 Analysis – H2 ... 35
4.4 Analysis – H3a and H3b... 35
4.4.1 Normality ... 35
4.4.2 Descriptive statistics ... 36
4.4.4 Post hoc ... 39
4.5 Additional analysis – H3a and H3b, Model 1... 40
4.6 Additional analysis – H3a and H3b, Model 3... 40
5 Conclusion ... 42
5.1 Discussion ... 42
5.2 Limitations ... 43
5.3 Further research... 44
References ... 45
Appendix A. Survey questionnaire ... 49
Appendix B. Respondent requirements... 61
Appendix C. Descriptive statistics respondents... 62
Appendix D. Control tightness variables ... 64
Appendix E. A taxonomy and theory of knowledge-intensive firms ... 66
List of tables Table 2.1 - Taxonomy PSFs ... 10
Table 3.1 - Sample selection ... 20
Table 3.2 - Descriptive statistics... 21
Table 3.3 - Gender distribution ... 21
Table 3.4 - Respondents specified per occupation ... 23
Table 3.5 - Respondents specified per PSF type ... 24
Table 3.6 - Factor analysis for control tightness ... 26
Table 3.7 - Factor analysis for explicit control tightness ... 27
Table 3.8 - Factor analysis for cost leadership and differentiation strategies ... 29
Table 3.9 - Factor analysis for cost leadership and differentiation strategies ... 30
Table 4.1 - Descriptive statistics... 34
Table 4.2 - Testing of H1... 35
Table 4.3 - Testing of H2... 35
Table 4.4 - Normality test ... 36
Table 4.6 - Outcome GLM ... 37
Table 4.7 - Outcome GLM, including control variables ... 38
Table 4.8 - Post hoc pairwise comparison ... 39
Table 4.9 - Outcome GLM, including control variables ... 40
Table 4.10 - Outcome GLM, including control variables ... 41
List of figures Figure 1. Control tightness, professionalization and MCS fit ... 13
Figure 2. Control tightness, strategy and MCS fit ... 17
Figure 3. Overview of hypotheses ... 18
1 Introduction
1.1 Research objective
Management controls are mechanisms designed to direct employees in such a way so that organizational objectives are achieved (Ouchi, 1979). Management control systems (MCS) are described by Simons (1994) as the formal, routines and procedures based system which are used by managers to maintain or alter organizational activities. Since the ’70, there has been many studies done in which the MCS has been studied. However, researchers often described the MCS characteristics in their own way (Auzair and Langfield-Smith, 2005). This can be confusing, therefore, Auzair and Langfield-Smith (2005) summarize control along a continuum for each MCS dimension. On one end of the control continuum is tight control, also referred to as more bureaucratic, action, formal, restricted and impersonal controls. While, on the other end of the control continuum is loose control, also referred to as less bureaucratic, results, informal, loose, flexible, and interpersonal controls. In this thesis, these definitions are used interchangeably.
An important aspect of MCS, is the contingency-based research to determine the fit of the MCS with contingency factors (Chenhall, 2003). In this paper, I study MCS for Professional Service Firms (PSFs). More specifically, I study what the effect is of the contingency factors professionalism and strategy on the MCS for PSFs.
1.2 Background
Economies in developed countries have a large professional service industry. In the Netherlands, a statistics report by the Dutch Consulting branch (2014), shows that the Dutch professional service industry envelops 21% of the Dutch economy. Despite this large impact PSFs have on the economy, there hasn’t been much research into PSFs. PSFs mostly have been studied by theoretical sociologist during the ’60s and ’70s. Only at the start of this century, PSFs caught increasing interest of organizational theorists into the study of professional service firms (Von Nordenflycht, 2010). He develops a taxonomy in order to distinguish the different types of PSFs, this taxonomy distinguishes four types of PSFs: Technology developers professionals), Neo-PSFs (quasi-professionals), Professional campuses (professionals) and Classical PSFs (professionals). For my research, I link my variables to these four different types of PSFs.
Amongst social theorists, there is a consensus that professionalization leads to looser MCSs (Mintzberg, 1978; Raelin, 1985; Von Nordenflycht, 2010, Wallace, 1995). They suggest that a professionalized workforce prefers autonomy and that the MCSs therefore is looser. However, to
my knowledge, there is no quantitative study done to support this consensus. I examine the effect of professionalization on the use of the MCSs. More specifically, I study if MCSs are looser when professionalization is higher. I operationalize professionalization as professionals (high professionalized) and quasi-professionals (low professionalized).
Furthermore, I examine the moderating effect of the contingency variable strategy on control tightness. There are ample studies done which examine the effect of strategy on MCSs (as is summarized by Auzair and Langfield-Smith, 2005, and Chenhall, 2003), findings from these studies suggest that strategy is an important contingency variable for MCS. However, these studies traditionally focus on manufacturing companies. Thus, it is unknown to what extent these findings are generalizable for PSFs. Therefore, Greenwood, Li, Prakash and Deephouse (2005) and Von Nordenflycht (2010) call for more (theoretical) attention towards strategy for PSFs.
1.3 Research question
In order to test the moderating effect of strategy on control tightness, I first test whether professionalization indeed leads to looser MCS. This is relevant to test, because this is the first study that has data of classical professionals as well as data from quasi-professionals. Other studies, such as Auzair and Langfield-Smith (2005) compare PSFs with mass service firms.
Building on this relationship between professionalization and MCS tightness, I test whether professionalization is moderated by strategy. Based on the contingency theory for strategy it is expected that a differentiation strategy leads to a looser MCS and a cost-leadership strategy leads to a tighter MCS, however, this may not be generalizable to PSFs. PSFs operate in a unique environment, therefore, the MCSs for PSFs should differ compared to other firms (Von Nordenflycht, 2010). He further argues that professionalization in PSFs leads to muted competition and hypothesizes that, professionalization leads to inefficiency. Therefore, I developed the following research question (RQ):
RQ: To what extent does professionalization lead to an effective management control system and how does strategy moderate this relationship?
I consider the MCS to be effective when the contingency factors fit the MCS. Based on theory and with the contingency factors as specified in figure 1, figure 2 and figure 3, I expect that PSFs that match the characteristics of their activities perform effectively compared with those firms that fail to achieve such fit (adopted from Speckle and Verbeeten, 2014).
1.4 Motivation
Answering my research question adds to the existing knowledge into professionalism, as it quantitatively shows the effect of professionalization on the MCS of PSFs. Further, it gives more insight into the moderating effect of strategy into MCS tightness, thereby answering the call of Greenwood et al. (2005) and Von Nordenflycht (2010) to address the lack of (theoretical) attention towards strategy. Moreover, this research question is a response to the call of Ditillo (2004) for more research specifically focused on the management control issues of PSFs. My research question is also interesting from a societal point of view, lately multiple classical PSFs in the Netherlands, such as law-firms (Boekel) or accounting-firms (KPMG) have reorganized their business. Answering my research question gives more insight in the control mechanisms for PSFs, therefore it might help PSFs when they will restructure their firms in the future. To my knowledge, this study is the first study that has data into the MCS of both the classical professionals as from quasi-professionals. Having this data allows me to test for heterogeneity between PSFs. Researchers often assume that PSFs are homogenous (Greenwood et al. 1990; 2002; 2003) (for an overview see Malhotra and Morris, 2009). However, this might not be the case, especially when considering strategy as moderator. The presence of heterogeneity between PSFs is therefore relevant, because it adds to the theory of PSFs and it limits the generalizability of studies into PSFs that consider PSFs as homogenous. Thus, this dataset gives me the opportunity to empirically test the effect of professionalization and the moderating effects of strategy on the use of MCSs and find empirical evidence whether professionalism indeed leads to a looser MCS.
To my knowledge, only Goodale, Kuratko and Hornsby (2008) and Saber (2016) tried to quantitatively study the effects of professionalization. However, Goodale et al. (2008) only measured what the effect is of professionalization on the compensation scheme. Further, they operationalized professionalization as the extent to which professionals have to subscribe to industry journals, newsletter and attend industry conferences. Saber (2016) operationalized professionalism using an untested construct, whereas I operationalize professionalism according to Von Nordenflycht’s (2010) taxonomy.
1.5 Outline
The remainder of this paper is organized as follows. Chapter two reviews the prior literature about professionalism, PSFs and strategy, based on this I develop my hypotheses. In chapter three I describe my research method and my variables. In chapter four I perform statistical tests and I analyze the results. Last is chapter five, in this chapter I discuss my findings and I suggest areas for further research.
2 Theory and Hypothesis
2.1 Professional Service Firms
Even though PSFs are such an important part of our society, there is some ambiguity how to classify PSFs. Ambiguity occurs because research into PSFs often does not define PSFs or they define a PSF indirectly by listing examples of PSFs (Von Nordenflycht, 2010). Furthermore, to identify a profession, researchers have used overlapping criteria (Sorensen and Sorensen, 1974). To avoid confusion, I use the taxonomy developed by Von Nordenflycht (2010), he distinguishes PSFs based on three distinctive characteristics: (1) knowledge intensity, (2) low capital intensity and (3) professionalized workforce. Knowledge intensity indicates the degree of output which relies on the complex knowledge of employees. Low capital intensity entails the degree to which output is dependent on employee knowledge and not on the output by machines. Professionalized workforce (professionalization) is specified by the knowledge base of its employees and how this is regulated and controlled. In this thesis, a high professionalized workforce is referred to as professionals, while a low professionalized workforce is referred to as quasi-professionals. Based on these characteristics he developed the following taxonomy, where he distinguishes four types of PSFs:
Table 2.1 - Taxonomy PSFs
Category (with examples) Knowledge intensity Capital intensity Professionalized workforce
Technology Developers High High Low
Biotech R&D labs
Neo-PSFs High Low Low
Consulting Advertising
Professionalized Campuses High High High
Hospitals
Classic PSFs High Low High
Accounting Law
Architecture
(Von Nordenflycht, 2010)
Why this framework? Research into PSFs has two streams, the first stream focusses on the processing of information of knowledge-intensive firms and seeks to understand how this information is used (Greenwood et al., 2005). The second stream focusses on developing the theory of professional service firms (Greenwood et al., 2005). This paper focusses on developing
the theory of professional service firms, and uses the taxonomy developed by Von Nordenflycht (2010) because it recognizes a wide range of PSFs and it gives a clear structure how to differentiate between the different types of PSFs.
2.2.1 Characteristics
Von Nordenflycht (2010) specifies professionalism based on two criteria: (1) regulation and control and (2) ideology1. Regulation and control is based on whether a profession has a monopoly
on its knowledge, if it is monopolized by an autonomous organization independent of the state and that this regulation excludes nonprofessionals and also mitigates competition among professionals (Von Nordenflycht, 2010). Ideology refers to the professional codes of ethics, as well as less explicit norms that define appropriate behavior for professionals (Von Nordenflycht, 2010).
A consequence of regulation and control is that the professional association excludes the unqualified and provides its members with the legal right to practice (Hall, 1968). Thereby giving its members a work setting where they are assumed to operate autonomously. The professional ideology further affects the mindset of the professional. First, professionals feel entitled to their autonomy and therefore, they wish to make decisions without external pressure from those who are not a member of their profession (Hall, 1968; Raelin, 1985). Second, professionals overspecialize in technical skill, resulting in feeling superior over others (Hall, 1968; Raelin, 1985). Thereby, they create a setting where the feel only the professional organization and its members can judge the professional in his/her work. Third, professionals desire self-regulation, this is the belief that evaluation should be based on professional standards and the best way to do so is by collegial control as opposed to evaluation by their managers, which is based on compliance to the rules and regulations of the organization (Hall, 1968; Raelin, 1985). Fourth, professional education is theoretical and differs from real-work practice. Talented young-professionals develop a sense of solidarity with their field and they become less motivated with monetary rewards (Hall, 1968). This theoretical focus of professionals’ conflicts with organizational goals, therefore professionals experience organizational requirements as intrusive. Fifth, professional’s belief that their service and their profession is indispensable to the public and that their work benefits both client and practitioner (Hall, 1968; Raelin, 1985). Last, organization strategy focuses on profits and efficiency, this is achieved through routinization. Professionals are opposed to routinization when it
1 This definition differs from others because it does not entail technical knowledge as is used in other definitions such
as Wilensky’s (1964). Von Nordenflycht (2010) excludes technical knowledge, because his taxonomy (knowledge intensity) already addresses this characteristic.
constrains their practice (Raelin, 1985). Professionals deal with unique client problems and they view their skills as necessary to truly help their clients and therefore they resist to a routinized ‘best practice’ treatment (Brivot, 2011). Hence, professionals resist to obey the rules and regulation that is enforced upon them. Due to these factors, professionals prefer to operate autonomously and are resistant to managerial control.
Organizations, on the other hand, are reluctant to provide autonomy to its employees, because when employees are given more autonomy, they can opt to act in a certain manner, which can conflict with the goals and procedures of the organization (Raelin, 1985). This leads to a strain between management and professional to find the level of control versus autonomy which satisfies both parties. If management implements too much control, it risks that professionals experience cognitive dissonance (Raelin, 1984a). When dissonance occurs, professionals feel tension. To reduce tension, professionals may engage in adaptive behavior detrimental to the organizational objectives. Thus, to mitigate cognitive dissonance, management use a less bureaucratic MCS for professionals.
Before becoming professionals, aspiring professionals must undergo extensive training in order to obtain a license from the industries association (Wilensky, 1964). By obtaining this license, professionals display that they have the necessary technical skills. During their training, aspiring professionals and their co-students, are submerged in the professional codes of ethics. After a professional obtains his/her license, the professional then needs to comply with the code of conduct as set by the industries association, thereby creating a sense of professionalism and collegiality. The sense of professionalism and collegiality develops further, when they work under the tutelage of a senior partner (Batt and Katz, 2004). During their working career, professionals mostly work independently, but share cooperative working relationships with other professionals, thereby creating a sense of community (Wallace and Kay, 2008). Therefore, it is important for professionals to understand and adhere to the social norms. Thus, professionals are continuously monitored by peers from the association if they comply with the professional code of conduct and adhere to the social norms of their society. These social norms therefore perform as external control mechanisms (normative controls) and allow professionals to operate more autonomously, as opposed to quasi-professionals (Montagna, 1968). Quasi-professionals do not have a strong association body, therefore PSFs with quasi-professionals must rely on internal and more formal control mechanisms, such as formalized rules and procedures. Hence, Nelson (1998) sees normative controls as central for an effective MCSs for highly professionalized PSFs, more so than formalized rules and procedures. Consequently, it is expected that quasi-professionals have a tighter MCS compared to professionals.
Hence, professionals resist to managerial control, therefore making bureaucratic control counterproductive to the organizational goals. This notion is supported by numerous researcher, for instance, Mills, Hall, Leidecker and Margulies (1983) argue that the nature of professional services, and the complex transaction between professional and client makes it difficult for managers to employ an efficient control system. Therefore, they see self-supervision (normative controls) as the most efficient management control system for PSFs. Thus, my starting hypothesis is as follows:
H1: PSFs with a low professionalized workforce put more emphasis on a tight MCS compared to PSFs with a high professionalized workforce, who emphasize a loose MCS.
Figure 1 summarizes hypothesis H1.
Figure 1. Control tightness, professionalization and MCS fit 2.2.2 Challenges and opportunities
The characteristics of PSFs lead to some unique challenges and opportunities for PSFs. These challenges and opportunities consists of cat herding, opaque quality, no investor protection, trusteeship norm, and muted competition.
The first challenge for PSFs is cat herding. Cat herding is the recruitment, retention and directing of employees, it is linked with having a highly skilled workforce. Highly skilled employees are scarce, more so than quasi-professionals, this is due to the high standards set by professional associations, and often professionals can easily use their skillset at other employers (Hitt, Bierman, Shimizu and Kochhar. 2001; Von Nordenflycht, 2010). Further, they have a close relationship with clients, who often follow professionals if they change firms (Greenwood et al., 2005). Therefore, highly skilled employees have a strong bargaining position. Professionals prefer to operate autonomously and dislike strict control mechanisms. Having a strong bargaining position, professionals can demand for less control, if their demands are not met, they can easily change employer. Thus, I expect that firms, to mitigate cat herding, subject professionals to less control compared to quasi-professionals.
The second challenge and opportunity for PSFs is opaque quality. Opaque quality arises when it is difficult to evaluate the quality of the output. Opaque quality arises within PSFs because of the information asymmetry between the professional and the client, thereby making it impossible for the client to assess the quality of the service (Greenwood et al., 2005). Thus, firms have to signal their competence to their clients. Professional firms can implement a code of conduct, a normative control, to credibly signal their competence (Von Nordenflycht, 2010). Quasi- professionals do not have a strong professional association, a code of conduct will therefore be a less credible signal for quasi-professionals. Quasi-professionals can opt to implement guidelines from the International Organization for Standardization (ISO) or they can solely rely on their reputation to signal quality. ISO guidelines are more formative compared to a code of conduct which characterizes a tight MCS, while relying on reputation signals a loose MCS. Therefore, I expect that the effect of mitigating opaque quality leads to the same MCS tightness for professionals as for quasi-professionals.
The third challenge and opportunity for PSFs is that there is little investor protection. PSFs have a high knowledge intensity, as a result, most PSFs distinguish themselves based on their ‘intangible assets’ (Pennings, Johannes, Lee, Kyungmook and van Witteloostuijn, 1998). Therefore, employee retention is essential for PSFs. Low capital intensity increases the bargaining position of employees, because they can easily leave and start their own firm, leaving investors with worthless shares (Von Nordenflycht, 2010). However, if the nature of work (e.g. hospitals) requires substantial investments, employee retention becomes easier, thus high capital intensity acts as investor protection and mitigates the risks of cat herding. Therefore, it is expected that with higher capital intensity, there is more MCSs tightness.
The fourth challenge and opportunity for PSFs relates to the trusteeship norm. The trusteeship norm is closely related to professionalism, it is the adherence of professionals to the professional code of ethics and professional norms and is led by their belief that professionals have the responsibility to protect the interest of their clients and the public in general (Edwards and Wigger, 2015). This professional responsibility is often in conflict with the commercial values of firms and clients, who are in constant pursuit of their self-interest (Von Nordenflycht, 2010). This professional responsibility is an intrinsic function that serves as a normative control measure. Furthermore, it adds to the belief of professional autonomy, as nonprofessionals introduce pressure to compromise the interest of clients. Quasi-professionals do not feel this professional responsibility, thus the need for autonomy is less and quasi-professionals therefore resist less to control. Hence, the trusteeship norm adds to the theory of professionalism and helps to explain why the MCS is less tight for professionals compared to quasi-professionals.
Finally, it is hypothesized by Von Nordenflycht (2010), that PSFs have muted competition. Muted competition is caused by two phenomena which are unique to PSFs with a highly-professionalized workforce. The first phenomenon is that of certification. Certain professions require a certification to perform, to become lawyer one must first get a diploma from the association body. Thus, certification acts as an entry barrier and this limits the competition. The second phenomenon is self-regulation, free market competition is seen as an influence that deteriorates the quality of the service provided, thus professional codes tend to prohibit commercial behavior (Torres, 1991).
2.2.3 Organizational responses
To deal with these challenges and opportunities, to mitigate the risks from the challenges and to try to profit from the opportunities, the PSFs respond to these challenges and opportunities (Von Nordenflycht, 2010). Organizational responses are the use of alternative compensation, providing employees with more autonomy and information, inside ownership and organizational slack.
Per Von Nordenflycht (2010), alternative compensation is used to incentivize professionals to address cat herding and to handle situations where quality is opaque. An example which is traditionally associated with classical PSFs, is the up-or-out incentive structure, which functions in accordance with the tournament theory (Edwards and Wigger, 2015). By implementing an up-or-out incentive structure, organizations increase employee loyalty (Greenwood and Empson, 2003). It thereby gives firms an alternative motivational instrument besides the overrepresented intrinsic motivation that is used for professionals (Suddaby, Greenwood and Wilderom, 2008). Furthermore, it acts as an apprenticeship system, where professionals learn from and are directed by experienced partners (Greenwood and Empson, 2003). This apprenticeship system with increased employee loyalty functions act as a normative control mechanism. Thus, I expect that the MSC is less tight if an up-or-out incentive structure is present.
Inside ownership occurs when the professionals are exclusive owners of the firm. Inside ownership is associated with more focus on normative rather than bureaucratic controls, which is regarded as a more effective form of control for professionals (Greenwood et al., 2005). Inside ownership influence is interpreted by organizational economists as a response to cat herding, more specifically employee retention (Greenwood et al., 2005; Von Nordenflycht, 2010). Greenwood and Empson (2003) argue that with a partnership model (inside ownership) professionals have the prospect to become partner, this motivates professionals to remain loyal to the organization. Furthermore, inside ownership controls for the trusteeship norm. First of all, professionals work in organizations were there is a high degree of receptiveness to professional expertise (Raelin,
1984a). Secondly, the services of professional organizations are in high demand (Raelin, 1984a). Finally, administrators are knowledgeable about professional concerns, they might even have a professional background (Raelin, 1984a). Additionally, quasi-professionals and professionals who work as a bureaucrat become less directly dependent on the professional community for his/her career advancement and is more dependent on the organization, therefore social control from the professional community is less effective for this group (Goode, 1957; Raelin, 1984a). Therefore, the fit between ownership structure and professionalization-level of the workforce is important, because it minimizes agency costs and job dissatisfaction (Raelin, 1984a; Greenwood and Empson, 2003). For pure professionals, job satisfaction might manifest itself as a lack of autonomy to pursue their work, while for quasi-professionals, it may result from lack of power and participation in organizational affairs. This is confirmed by Hall (1968), who empirically finds that bureaucratization threatens the professional autonomy, which can cause conflict between the professional and the organization. Hence, Greenwood and Empson (2003) conclude that inside ownership is an effective organizational model for professionals, because it provides an organizational context based on collegial controls, which provides professionals more autonomy, enabling them to perform effectively. Hence, I expect the MCS to be less tight if the firm has an inside ownership structure.
2.3 Contingency theory: strategy
PSFs differ distinctly from other organizations (Greenwood et al., 2005). Despite this, PSF research has had a lack of attention towards strategy. Perhaps, this is related to the fact PSFs have a less tight MCSs/a more normative MCSs. In general, there is a strong linkage between MCSs and organizational goals (Chenhall, 2003). However, if social theorists are to be believed, professionalization leads to a looser MCS. Therefore, the linkage between the MCS and strategy is less clear. This makes it more difficult to study strategy for PSFs. Hence, it is unsure to what extent the contingency research into strategy can be extended towards PSFs, since contingency research focuses on large manufacturing organizations, which differ significantly from PSFs (Suddaby et al., 2008).
Strategy is a relative new contextual variable for the research into the design of MCSs, however it is seen as an important variable for the design of MCSs (Chenhall, 2003). According to contingency theory, the MCS should be designed in a way to support the business strategy of the firm (Simons, 1987; Langfield-Smith, 1997). Strategy is an instrument that enables managers to influence the nature of the external environment, the technologies of the organization, the structural arrangements and the control culture and the MCS (Chenhall, 2003). Thus, strategy gives
managers the opportunity to position their organization in a particular environment. In this paper, I focus on the effect of strategy on the control culture.
There are multiple taxonomies for strategy (for an overview see Chenhall, 2003). An often-used taxonomy, is the taxonomy developed by Porter (1980), which is often-used in this paper. Porter’s strategy typology is used by a large body of empirical studies to examine the relationship between the MCSs and strategy using the contingency approach (Auzair and Langfield-Smith, 2005). Porter (1980) identifies two opposite strategies: cost-leadership and differentiation strategy. Govindarajan (1988) hypothesizes that the strategic choice relates to uncertainty, where differentiation is associated with high uncertainty because of its focus on product innovation, while cost leadership is associated with low uncertainty because of its focus on cost reduction. The greater the uncertainty, the more difficult it is for superiors to accurately control the performance of employees and for employees to be evaluated based on this highly uncertain environment. Therefore, differentiation strategy is associated with a looser MCS compared to cost leadership strategy. Evidence suggests that cost-leadership strategies are associated with centralized control systems and where work is more formalized, while differentiation strategies are associated with decentralized control system, where there is a lack of standardized procedures and where work is more flexible (Chenhall, 2003). Therefore, I hypothesize the following:
H2: PSFs with a cost-leadership strategy put more emphasis on a tight MCS compared to PSFs with a differentiation strategy, who emphasize a loose MCS.
Figure 2 summarizes hypothesis H2.
Figure 2. Control tightness, strategy and MCS fit
Contingency research towards strategy suggest that differentiation strategy is associated with loose controls, while cost-leadership is associated with tight controls (Auzair and Langfield-Smith, 2005). Extending this well develop contingency research towards PSFs, I expect the same relationship for low professionalized PSFs. Therefore, I hypothesize the following:
H3a: PSFs with a low professionalized workforce who pursue a cost-leadership strategy put more emphasis on a tight MCS than PSFs with a low professionalized workforce who pursue a
differentiation strategy.
As previously mentioned, contingency theory predicts that differentiation strategy is associated with loose controls, while cost-leadership strategy is associated with tight controls (Auzair and Langfield-Smith, 2005). I therefore hypothesize the following:
H3b: PSFs with a high professionalized workforce who pursue a cost-leadership strategy put more emphasis on MCS tightness/looseness compared with PSFs with a high professionalized workforce who pursue a differentiation strategy.
Figure 3 provides an overview of my hypotheses and shows the relationships between my variables.
3 Research methodology
3.1 Research design
This section describes the research method and it explains how I empirically test my hypotheses. Data was collected through a pre-developed online questionnaire (survey). The questionnaire consists of questions related to control variables, MCS variables, organizational variables and outcome variables. In this questionnaire, most constructs are measured based on a 5-point Likert scale. See appendix B for the criteria respondents had to meet.
This paper uses data collected during the period September 2015 – January 2017. Data was collected by students who joined the PSF Professional Service Firm Thesis Survey Project from the University of Amsterdam. Students administered the questionnaire to mid-level employees from midsize to large PSFs. To control for robots, respondents provided an e-mail with their function and corporate signature, confirming that they completed they survey. The set-up of this project did not allow to measure the response rate.
3.2 Sample
A total of 612 responses were received, from this a total 239 (39,1%) responses were usable. Of these 612 responses, 89 did not finish or submitted the survey. Additionally, the survey contained a control question, which asked if respondents read the above statement and remarks, 17 respondents indicated that they did not read above statement and remarks and were thus removed. Furthermore, 2 respondents did not specify their occupation. Since, I use occupation to operationalize professional workforce, I removed these respondents. Additionally, 72 respondents indicated that they have less than 3 years of work experience or did not specify how much work experience they had. Professional competence consists of theoretical knowledge as well as practical experience, novices do not have practical experience and therefore lack insight into concrete settings (Bromme and Tillema, 1995). Hence, I removed respondents with less than 3 years’ work experience, since they lack in practical experience. Moreover, I removed 185 respondents whose highest level of education was a bachelor degree or lower, since a prerequisite of professionalism is a high level of education. Finally, I removed executives from my sample, since I study professionals who are affected by the MCS, imposed by management. Although executives are also subjected to the MCS, they participate in designing the MCS and thereby they can influence the MCS of the organization. Hence, executives are deleted from my sample.
Table 3.1 - Sample selection
Criteria # Frequency % Percentage
Completed PSF-questionnaires 2015 to 2017 612 100,0%
Unfinished questionnaires -89 14,5%
523
Have you read the above statements and remarks? -17 2,8%
506
Missing values (occupation, ownership and organizational unit) -4 0,6% 502
Less than 3 years’ work experience or work experience not
indicated -72 11,8%
430
Bachelor degree or lower -185 30,2%
245
Executives -8 1,3%
Total respondents 237 38,8%
After data screening a total of 237 respondents are left over. Generally, a good sample size has a minimum of 300 samples, while a poor sample size contains less than 100 samples (Field, 2013). My sample size contains 237 samples and lies between these thresholds, it can therefore be considered adequate. Table 3.2 contains an overview of the respondents’ characteristics.
As can be expected from PSFs, due to the up-or-out incentive structure and the tournament theory, the employee participation is highest for younger employees. In my sample, the employee work experience distribution was indeed higher for employees with less work experience. The firms’ workforce consisted for 18.6% of employees with 3 years of work experience, while the workforce consisted for 2.8 % of employees with 9 years of work experience. Appendix C.2 consist with more information about the age distribution.
Table 3.2 - Descriptive statistics
% Percentage Mean # Number Std. Deviation
Gender Male 65.6% Female 34.4% Education Master degree 73.7% Doctorate degree 26.3%
Age (in years) 36 8.52
Youngest 22 Oldest 64 Work experience 3 years 18.6% 4 years 14.2% 5 years 9.3% 6 years 10.1% 7 years 6.9% 8 years 6.9% 9 years 2.8% 10 years or more 31.2% Average 7.6 2.76
The gender distribution in the Netherlands is 38.2% female and 61.8% male, to test normality I run a Chi-Square Goodness-of-fit test. This test whether the sample is significantly different from the population. The female participation in PSFs in the Netherlands in 2015 (“Financiële en zakelijke diensten”) is approximately 44.9% (CBS statline, 2017). I test normality of the gender distribution in the Netherlands, because Dutch participants form by far the most substantial group in my sample (see appendix C for an overview of nationality). Below in table 3.3 is an overview of the observed values in my sample and the expected values based on the population.
Table 3.3 - Gender distribution
Observed # Expected # Residual
Male 94 84 11
Female 58 68 -11
Total 152 152 0
The outcome of the test is not statistically significant, with χ2(2) = 2.661 and p = .103. Therefore, I can reject the alternative hypothesis, meaning the gender distribution does not statistically differ from the population.
3.3 Variable measurement
3.3.1 Independent variable – professionalized workforce
Professionalized workforce is operationalized using Von Nordenflycht’s (2010) taxonomy. He distinguishes the following types of PSFs: Technology developers, Neo-PSFs, Professional campuses and Classical PSFs. Technology developers and Neo-PSFs are considered having a low professionalized workforce and Professional campuses and Classical PSFs are considered having a high professionalized workforce. Thus, professionalized workforce is operationalized as a categorical variable (dichotomous variable). Table 3.4 gives more insight how the occupations are distributed.
Table 3.4 - Respondents specified per occupation
Occupation # Frequency % Percent Model 1 Model 2 Model 3
Accounting 37 15.5% 1 1 1 Advertising 1 0.4% 3 3 3 Architecture 2 0.8% 1 1 1 Biotechnology 6 2.5% 4 4 4 Consulting Engineering 1 0.4% 5 1 Consulting IT 14 5.9% 3 3 3 Consulting HR 3 1.3% 3 3 3 Consulting Management/Strategic 23 9.6% 3 3 3 Consulting Technology 1 0.4% 3 3 3 Engineering 11 4.6% 5 1 Financial advising 5 2.1% 5 3 Investment banking 2 0.8% 5 1
Investment management (hedge
funds, VC,mutual funds) 1 0.4% 5 1
Law 6 2.5% 1 1 1
Marketing/public relations 3 1.3% 5 3
Media production (film, TV, music) 1 0.4% 5 3
Medicine/Physician practices 31 13.4% 2 2 2 Pharmaceutical 6 2.5% 5 4 Project management 7 3.3% 5 3 Real estate 3 1.3% 5 3 Recruiting - executive 6 2.5% 5 3 Research/R&D 20 8.4% 4 4 4
Risk management services 9 3.8% 5 3
Software development 2 0.8% 5 3
Talent management/agency 1 0.4% 5 3
Other 35 14.6% 5 5
Total 237 100%
1 = Classic PSFs Professionals
2 = Professional Campuses Professionals
3 = Neo-PSFs Quasi-professionals
4 = Technology Developers Quasi-professionals
5 = Other Quasi-professionals
To operationalize professional workforce, I assigned the different occupations to Von Nordenflycht’s (2010) taxonomy. Von Nordenflycht (2010) gives a few examples per PSF type. The different Models specify how the occupations are distributed amongst the different PSFs. Under model 1, the occupations are distributed amongst the PSF types, in accordance with Von Nordenflychts’ specifications. Model 2 is the same as model 1, except that I included the
unspecified occupations, these are labelled as ‘Miscellaneous’. Under model 3, the occupations are distributed amongst the different PSF by interpreting Von Nordenflycht’s (2010) framework. Table 3.5 displays the number of respondents per PSF type per model.
Table 3.5 - Respondents specified per PSF type
PSF type # Frequency model 1 Percent % # Frequency model 2 Percent % # Frequency model 3 Percent %
Classical PSFs 45 19.0% 45 19.0% 60 25.3% Professional Campuses 31 13.1% 31 13.1% 31 13.1% Neo-PSFs 42 17.7% 42 17.7% 79 33.3% Technology Developers 26 11.0% 26 11.0% 32 13.5% Other 0 0.0% 93 39.2% 35 14.8% Total 144 61% 237 100% 237 100%
PSF type # Frequency model 1 Percent % # Frequency model 2 Percent % # Frequency model 3 Percent %
Professionals 76 32.1% 76 32.1% 91 38.4%
Quasi-professionals 68 28.7% 161 67.9% 146 61.6%
Total 144 61% 237 100% 237 100%
3.3.2 Dependent variable – control tightness
Research towards MCSs uses different characteristic, such as action control vs result control, formal control vs informal control (see Auzair, Langfield-Smith, 2005, for a more detailed overview). I examine the extent to which MCSs are considered having tight control versus loose control. Control tightness is the degree to which controls are integrated into the organization, this can be explicit and implicit. On one end of the spectrum is explicit loose control, this occurs when the work is uninspected and unevaluated and on the other end of the spectrum is tight explicit control, this occurs when the work is constantly monitored and evaluated (Weick, 1976). Implicit control is based upon how the rules are defined and how flexible the rules are enforced. Clear defined rules with no deviation possible are considered tight implicit controls, while flexible rules that function as guidance are considered loose implicit controls (Butler, Price, Coates and Pike, 1998). I model control tightness as a formative construct that is defined by two indicators, i.e. the explicit control tightness (the degree to which the amount and scope of controls are used in the organization) and the implicit control tightness (the degree to which controls and its compensation scheme are uphold). These two indicators are formative constructs, which, in turn, consists of four different measures: action control, result control, personnel control and cultural control. These measures are reflective constructs, they are measured using a variety of questions. Most of these
questions are adopted from previous literature, while others are new (see Appendix – Table D for an overview).
To compute the reflective constructs, factor analysis (principal component analysis) was conducted. Based on previous literature, I expect to extract eight factors, therefore I extracted a fixed number of factors, namely eight factors. Further, I selected Varimax rotation and suppressed loadings below 0.4 to improve readability. The reliability test, showed that a total of twelve respondents (2.8%) had some missing values, generally the threshold for missing values for a specific variable is 10% (Field, 2013), with 2.8% missing values, my data is well within the threshold, therefore I selected replace missing values with mean.
Table 3.6 shows the result of the factor analysis, the lowest item in the factor analysis has a loading of .467. The Kaiser-Meyer-Okin measure was .779, this is seen as middling, and it indicates that the factor analysis has an acceptable correlation (Field 2013). Furthermore, the Bartlett test of Sphericity was significant (χ2 = 3289.12, p-value = 0.000), which means there is
correlation between the variables (Field, 2013). Initially, no items were deleted because of correlation, each item had at least one correlation above .3 and the highest correlation was .511, therefore multicollinearity is not an issue, this is confirmed by the determinant of the R-matrix which is .099, this is well above the minimum requirement of .00001 (Field, 2013). Finally, all items had an eigenvalue above 1, thereby adding explanatory power to the model. The total variance explained by the factors explicit personnel control tightness (15.59%), explicit result control tightness (10.18%), explicit action control tightness (7.31%), explicit cultural control tightness (5.93%), implicit action control tightness (5.27%), implicit cultural control tightness (4.93%), implicit result control tightness (4.34%) and implicit personnel control tightness (3.64%) is 57.19%.
Table 3.6 - Factor analysis for control tightness Items Explicit personnel control tightness Explicit result control tightness Explicit action control tightness Explicit cultural control tightness Implicit action control tightness Implicit cultural control tightness Implicit result control tightness Implicit personnel control tightness Q5_6 .781 Q5_1 .729 Q5_7 .719 Q5_4 .713 Q5_12*** .610 Q10_4 .810 Q10_8 .794 Q10_9 .765 Q10_7 .577 Q4_1 .760 Q4_4 .696 Q4_2 .690 Q4_11** -.634 Q4_8** .406 .549 Q3_2 .837 Q3_1 .820 Q3_3 .669 Q3_5** .467 Q4_13 .813 Q4_10 .774 Q4_12 .768 Q3_6 .787 Q3_8 .756 Q3_11 .592 Q3_7 .495 Q10_5 .781 Q10_2 .706 Q10_6 .472 Q10_3 .469 Q5_10 .658 Q5_9 .621 Q5_11 .602 Variance explained 15,59% 10,18% 7,31% 5,93% 5,27% 4,93% 4,34% 3,64% Cronbach' s alpha .805 .795 .445 .728 .751 .613 .613 .437 Cronbach' s alpha* .797 .738 .747
* Cronbach's alpha after deleting items ** Deleting this item improves reliability
The Cronbach’s alpha of personnel control is far below the acceptable level for reliability. Therefore, implicit personnel control tightness cannot be used. This increases the difficulty to compare the constructs explicit control tightness and implicit control tightness. Furthermore, the Cronbach’s alpha for implicit cultural control and implicit action control are just barely acceptable. This indicates that for implicit tightness it is difficult to obtain a passable level of reliability. Hence, I only use explicit control tightness.
Thus, I conducted factor analysis to measure the four different reflective constructs that form the formative construct explicit control tightness. Table 3.7 shows the result of the factor analysis, the lowest item in the factor analysis has a loading of .674. The Kaiser-Meyer-Okin measure was .779, this is seen as middling, and it indicates that the factor analysis has an acceptable correlation (Field, 2013). Furthermore, the Bartlett test of Sphericity was significant (χ2 = 3289.12,
p = 0.000). The total variance explained by the factors explicit personnel control tightness (27.52%), explicit result control tightness (16.41%), explicit action control tightness (10.60%) and explicit cultural control tightness (10.27%) is 64.80%.
Table 3.7 - Factor analysis for explicit control tightness
Items Explicit result control tightness Explicit personnel control tightness Explicit action control tightness Explicit cultural control tightness
Q10_4 .845 Q10_8 .824 Q10_9 .740 Q10_7 .674 Q5_6 .827 Q5_1 .801 Q5_7 .726 Q5_4 .710 Q4_1 .829 Q4_4 .813 Q4_2 .718 Q3_2 .846 Q3_1 .831 Q3_3* .708 Variance explained 27.52% 16.41% 10.60% 10.27% Cronbach's alpha .795 .797 .738 .747
* Cronbach alpha improves upon deleting, to 0.789
Based on the factor analysis results, I compute the explicit result control tightness, explicit personnel control tightness, explicit action control tightness and explicit cultural control tightness by averaging the relevant item scores. I then compute the formative construct explicit control
tightness by summating the scores of explicit result control tightness, explicit personnel control tightness, explicit action control tightness and explicit cultural control tightness
3.3.3 Moderating variable - strategy
Strategy is measured using different questions rated based on a 5-point Likert-scale, where 1 means placing a little emphasis on the respective construct and 5 means placing a great deal of emphasis on the respective construct. In accordance with Auzair and Langfield-Smith (2005), both cost leadership strategy and differentiation strategy were measured as separate scales, thereby recognizing that firms may pursue both strategies in varying degrees. A higher score on the respective strategy, indicated a higher emphasis on the respective strategy. Therefore, strategy is operationalized as an ordinal variable. In addition, factor analyses are used, were factors re be loaded for differentiation strategy and cost leadership strategy.
Cost leadership strategy is measured, as the degree of emphasis on the following activities (Auzair, Langfield-Smith, 2005):
1. Achieving lower cost of services than competitors 2. Making services/procedures more cost efficient
3. Improving the cost required for coordination of various services 4. Improving the utilization of available equipment, services and facilities
Differentiation strategy is measured, as the degree of emphasis on the following activities (Auzair, Langfield-Smith, 2005):
1. Introducing new services/procedures quickly
2. Providing services that are distinct from that of competitors 3. Offering a broader range of services than the competitors 4. Improving the time it takes to provide services to customers 5. Providing high quality services
6. Customizing services to customers need 7. Providing after-sale service and support
Examination of cost leadership (4 items) and differentiation (7 items) on reliability (Cronbach’s alpha) reveals that they were adequate for factor analysis. Cost leadership has a reliability score of .72 and differentiation has a score of .75, so both items score above .7, which is considered the acceptable level of reliability (Field, 2013). The reliability cannot be improved by deleting items. Since the reliability is acceptable, both cost leadership and differentiation can be further analyzed using factor analysis.
Factor analysis (principal component analysis) was conducted. Based on previous literature, I expect to extract two factors, therefore I extracted a fixed number of factors, namely two factors. Further, I selected Varimax rotation and suppressed loadings below 0.4 to improve readability. The reliability test, showed that a total of twelve respondents (2.8%) had some missing values, generally the threshold for missing values for a specific variable is 10% (Field, 2013), with 2.8% missing values, my data is well within the threshold, therefore I selected replace missing values with mean.
Table 3.8 shows the result of the factor analysis, all items in the factor analysis have a minimum loading of .524. The Kaiser-Meyer-Okin measure was .827, this is seen as meritorious, and it indicates that the factor analysis should have reliable factors (Field, 2013). Furthermore, the Bartlett test of Sphericity was significant (χ2 = 1140.95, p = 0.000). No items were deleted because
of correlation, each item had at least one correlation above .3 and the highest correlation was .511, therefore multicollinearity is not an issue, this is confirmed by the determinant of the R-matrix which is .099, this is well above the minimum requirement of .00001 (Field, 2013). Finally, all items had an eigenvalue above 1, thereby adding explanatory power to the model. The total variance explained by the factors cost leadership (12.38%) and differentiation (34.52%) is 46.9%.
Table 3.8 - Factor analysis for cost leadership and differentiation strategies
Items Cost Leadership loadings Differentiation loadings
Achieving lower cost of services than competitors .700 Making services/procedures more cost efficient .791 Improving coordination cost of various services .770
Improving utilization .524
Introduce new services/procedure quickly .569
Provide distinct services from competitors .727
Offer a broader range of services than competitors .683
Improve the time to provide services to customers .596
Provide high quality services .549
Customizing services to customers need .563
Providing after-sale service and support .654
Variance explained 12,38% 34,52%
Cronbach's alpha .75 .73
In table 3.8 the item “Improving the time to provide services to customers” loaded high for the factor cost leadership, however based on previous literature, I would expect this item to load high on differentiation instead. In contrast to Auzair and Langfield-Smith (2005), I specifically focus on PSFs and not all service firms, and PSFs often provide services that are deemed unique
due to client dependency (Greenwood et al., 2005). Improving the time to provide services to customers can be achieved by standardizing operations. Thus, improving the time to provide services to customers indicates a well-organized and efficient organization, which acts more cost efficient. Organizations that have a differentiation strategy might have more unique services that are more difficult to standardize and the characteristics of these services might make it impossible to distinct itself with the time to provide services to customers. Therefore, improving the time to providing services to customers for PSFs is not a differentiation strategy, instead it is a cost leadership strategy.
Despite this, I conducted a factor analysis where I removed the item “improving the time to provide services to customers”. Cost leadership has a reliability score of .72 and differentiation has a score of .73, so both items score above .7, which is considered the acceptable level of reliability (Field, 2013).
Table 3.9 shows the result of the factor analysis, all items in the factor analysis have a minimum loading of .534. The Kaiser-Meyer-Okin measure was .804, this is seen as meritorious, and it indicates that the factor analysis should have reliable factors (Field, 2013). Furthermore, the Bartlett test of Sphericity was significant (χ2 = 973.46, p = 0.000). No items were deleted because
of correlation, each item had at least one correlation above .3 and the highest correlation was .511, therefore multicollinearity is not an issue. The total variance explained by the factors cost leadership (13.37%) and differentiation (34.59%) is 47.96%. Thus, by deleting the item “improving the time to provide services to customers”, more variance is explained.
Table 3.9 - Factor analysis for cost leadership and differentiation strategies
Items Cost Leadership loadings Differentiation loadings
Achieving lower cost of services than competitors .738 Making services/procedures more cost efficient .785 Improving coordination cost of various services .779
Improving utilization .534
Introduce new services/procedure quickly .576
Provide distinct services from competitors .717
Offer a broader range of services than competitors .669
Provide high quality services .570
Customizing services to customers need .582
Providing after-sale service and support .657
Variance explained 13,37% 34,59%
The results shown in table 3.9 show that the items are adequate for factor analysis. I summarized the items for both cost-leadership and differentiation for further analysis.
For analysis of hypothesis 1, I classified strategy consistent with Auzair and Langfield-Smith (2005) and I used an untested classification for strategy. Strategy, which consist of the items cost-leadership and differentiation, is measured on two separate scales of cost-leadership and differentiation. In accordance with Auzair and Langfield-Smith (2005), I undertook a median split to determine which firms uses a cost-leadership strategy and which firms uses a differentiation strategy. I define cost-leaders as those firms placing high emphasis on a cost-leadership strategy (median ≥ 5.25) but low emphasis on differentiation (median < 5.43), and I define differentiators as those firms that place a high emphasis on a differentiation strategy (median ≥ 5.43) but a low emphasis on cost-leadership (median < 5.25) (Auzair and Langfield-Smith, 2005).
For further analysis, I used an untested classification for strategy. I recoded strategy as a dichotomous variable. To determine how to classify firm strategy, I compared the mean of cost-leadership with the mean of differentiation. Firms were classified as cost-leaders when (cost-leadersip >= differentiation) and firms were classified as differentiators when (cost-leadership < differentiation). On account of differentiation having a higher mean compared to cost-leadership, I classified firms also as cost-leaders when their focus on cost-leadership was equal to their focus on differentiation. I use this classification for strategy because with this method I retain more participants than with the classification of Auzair and Langfield-Smith (2005), this is due to the cut-off method used.
3.3.4 Control variables
The first control variable is alternative compensation, more specifically up-or-out promotion. Some PSFs choose to implement an up-or-out incentive structure to mitigate the challenges of cat herding and opaque quality, which are unique to PSFs. The up-or-out incentive structure acts as an apprenticeship system and it increase employee loyalty, thus presence of an up-or-out incentive structure leads to a less tight MCS (Greenwood and Empson, 2003). I therefore control for an up-or-out incentive structure, since I expect it significantly affect my dependent variable (MCS tightness) and otherwise it distorts my findings. Respondents were asked what was included as part of their compensation/reward/promotion system. I created a dummy variable for up-or-out promotion structure present compared to no up-or-out promotion structure present.
The second control variable is inside ownership, inside ownership controls for capital intensity and employee retention (Greenwood and Empson, 2003; Greenwood et al., 2005; Von Nordenflycht, 2010). I control for inside ownership since capital intensity is associated with a
tighter MCS, however this is only possible if the firm mitigates the risk for investor protection. Hence, I expect that inside ownership leads to a less tight MCS and therefore I control for this variable. Respondents are asked about the primary ownership type of the company they work in. They can give the following responses:
- People who work in the organization (i.e. partnership).
- People employed outside of the organization (i.e. shareholders, investors). - No one. I work in a non-profit/public organization.
I created a dummy variable where I compare inside ownership with other types of ownership. This variable is not used in prior research.
The third control variable is size of the firm. Chenhall (2003) argues that size is an important contingency variable because large organizations must incorporate additional controls due to an increase in information. Additionally, large organizations often operate globally, this adds more complexity to the organization which is mitigated through additional administrative controls. Therefore, organizations adopt a different MCS based on size. This notion is supported by Auzair and Langfield (2005) who finds that firms size significantly influences the MCS design. To test for firm size, respondents were asked how many people are employed by their entire company, they could give the following responses:
- Less than 100 employees
- More than 100 but less than 500 employees - More than 500 but less than 5000 employees - More than 5000 employees
Consistent with Auzair and Langfield (2005), I regarded firms as small when they have less than 100 employees and I regarded firms as large when they have 100 employees or more.
The fourth control variable is size of the work unit. Montagna (1968) studied the eight largest accounting firms in the US and found that all firms were highly professionalized and highly centralized, with most administrative components spread throughout the organization. Normally, when the size of the organization increases, the autonomy of professionals’ conflicts with additional bureaucratic control mechanisms. However, this was not the case for the accounting firms, this was due to the fact that, since administrative components were performed throughout the organization and accountants work in small teams, there was no increase needed in bureaucratic control mechanisms. Therefore, I control for size of the work unit. To test for work unit size, respondents were asked how many people work in their organizational unit, they could give the following responses:
- Less than 10 employees
- More than 10 but less than 50 employees - More than 50 but less than 100 employees - More than 1000 employees
Montagne (1968) argues that small groups of less than 10 persons is associated with less administrative constrains and thus less control. I therefore coded work size unit as a dummy variable, with either a small work size unit (less than 10 employees) or a large work size unit (more than 10 employees).
3.4 Analysis
This study tests the theoretically assumed relationship between the independent variable professionalized workforce (PW), the dependent variable control tightness (CT), the dependent variable strategy (STR) and the interaction between the dependent variables.
First, I conduct two independent samples t-test to test hypothesis 1 and hypotheses 2. Hypothesis 1 assumes that the CT is significantly different between professionals and quasi-professionals. Hypothesis 2 assumes that the CT is significantly different between a cost-leadership strategy and a differentiation strategy. I test this hypothesis with an independent-samples t-test.
Finally, I test hypothesis 3a and hypothesis 3b. Hypothesis 3 assumes that the CT is significantly different between; a) professionals with cost-leadership strategy and professionals with differentiation strategy and b) professionals with cost-leadership strategy and quasi-professionals with differentiation strategy. In order to test hypothesis 3, I perform a General Linear Modelling (GLM) analysis. Both the independent variable (PW; two categories) as the moderating variable (STR; two categories) are categorical (dichotomous). The dependent variable (CT) is ordinal (continuous). The quantification of my variables makes it appropriate to test my hypotheses with GLM analysis. The moderating effect of my variable STR, as assumed in H3, is tested with a newly created variable Interaction (STR*PW).
Model:
𝐶𝑇 = 𝛽0 + 𝛽1 𝑃𝑊𝑖+ 𝛽2 𝑆𝑇𝑅𝑖 + 𝛽3 𝐼𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑖(𝑃𝑊 ∗ 𝑆𝑇𝑅) + 𝜀
𝐶𝑇 = 𝛽0 + 𝛽1 𝑃𝑊𝑖+ 𝛽2 𝑆𝑇𝑅𝑖 + 𝛽3 𝐼𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑖 (𝑃𝑊 ∗ 𝑆𝑇𝑅) + 𝛽4 𝑆𝑖𝑧𝑒𝑖
+ 𝛽5 𝑆𝑖𝑧𝑒 𝑤𝑜𝑟𝑘 − 𝑢𝑛𝑖𝑡𝑖+ 𝛽6 𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝑖 + 𝛽7 𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒 𝑖𝑛𝑐𝑒𝑛𝑡𝑖𝑣𝑒𝑖
4 Results
4.1 Descriptive statistics
This section first presents the descriptives statistics for all my variables, this is shown in table 4.1. CT has a mean of 12.9, this lies slightly above the average (12.0) of the actual range. For the independent variables, all means are slightly above the average of the actual range. Differentation strategy has a minimum of 1.5, indicating that all organizations have incorporated at least some level of differentiation into their strategy.
Table 4.1 - Descriptive statistics
Variable Actual range Mean Median Min Max Std. dev
Control tightness 4-20 12.90 12.92 8.25 18.75 2.25 Independent variables Strategy 1-2 1.54 2.00 1.00 2.00 0.50 Cost-leadership 1-5 3.45 3.50 1.00 5.00 0.77 Differentiation 1-5 3.71 3.67 1.50 5.00 0.64 Professionalized workforce 1-2 1.68 2.00 1.00 2.00 0.47 4.2 Analysis – H1
Hypothesis 1 predicted that professionals have a significantly lower CT compared to quasi-professionals. An independent sample t-test was conducted to establish whether there is a significant difference between the CT for these groups.
Table 4.2 shows the result of the independent samples t-test using Model 1 for professional workforce. First, I produced boxplots to inspect for outliers, hereby, I found four outliers. I modified the outliers’ values with the next, respectively, largest and smallest value. My sample size is not high; therefore, I test for normality using the Wilk test. Assessment of the Shapiro-Wilk’s test (p-value professionals = 0.085 and p-value quasi-professionals = 0.097) showed CT scores to be normally distributed for each level of professionalization and for each level of strategy. The Levene’s test indicates whether homogeneity of variance is present between groups, the p-value = 0.323 and is therefore not significant. The results show, that the CT is lower for professionals compared to quasi-professionals, as was hypothesized. However, the differences between the means is statistically insignificant. Therefore, hypothesis 1 is not supported. Additionally, the same test was performed using Model 2 and Model 3, the results for these models were also statistically insignificant.
Table 4.2 - Testing of H1
Variables N SD Mean for control tightness Test for normality
Professionalized workforce
Professionals 76 2.48 12.84 0.085
Quasi-professionals 161 2.13 12.90 0.097
t-value -0.14
significance 0.89
* Equal variances are assumed (p-value = 0.323) 4.3 Analysis – H2
Hypothesis 2 predicted that a cost-leadership have a significantly lower CT compared to a differentiation strategy. An independent sample t-test was conducted to establish whether there is a significant difference between the CT for these groups.
Table 4.3 shows the results of the independent sample t-test. Assessment of the Shapiro-Wilk’s test (p-value cost-leadership = 0.432 and p-value differentiation = 0.748) showed CT scores to be normally distributed for each level of strategy and the Levene’s test is not significant (p-value = 0.266). The results show that the CT is lower for a differentiation strategy compared to a cost-leadership strategy. Nonetheless, the differences between the means are statistically insignificant. Therefore, hypothesis 2 is rejected. Additionally, the same test was performed using another cut-off for strategy, this method was described in section 3.3.3. The result for this method is also statistically insignificant (p-value = 0.189).
Table 4.3 - Testing of H2
Variables N SD Mean for control tightness Test for normality
Strategy
Cost-leadership 57 1.85 13.04 0.432
Differentiation 39 2.17 12.63 0.748
t-value 0.912
significance 0.364
* Equal variances are assumed (p-value = 0,266) 4.4 Analysis – H3a and H3b
4.4.1 Normality
As specified before, I conduct independent-samples t-tests and GLM analysis to test my hypotheses. In the analysis below, I have used Model 2 to specify professional workforce, additionally, I conducted the same analysis using Model 1 and Model 3 to specify professional workforce, section 4.5 contains this analysis.
