Amsterdam Business School
Research Seminar Accountancy & Control
Master Thesis
The effect of Control tightness on Job
Tension and Job Performance in
Professional Service Firms
Name: Hicham El Kaddouri
Student number: 10676473
Thesis supervisor: Prof. dr. D.M. Swagerman
Date: 19 June 2016
Word count: 16.002
MSc Accountancy & Control, specialization Control
2 Statement of Originality
This document is written by student Hicham El Kaddouri 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.
3 Acknowledgements
Firstly, I want to express my gratefulness to Allah. Then, I would like to thank my supervisor Prof. Dr. Dirk Swagerman for his time and energy during the difficult thesis period. I am also grateful to Mr. Sander van Triest for his valuable guidance on crucial moments and some practical help during the thesis procedure. I also would like to thank Ms. Helena Kloosterman MSc for giving me the opportunity to participate in the survey project and always finding time to answer my questions. I wish her all the luck with her PhD research. Last but not least, I want to thank my fellow students, especially Yakub, Mohamed, Enes, Luke, Selcuk, Idris and others for the joyful moments. I wish them all the luck in their future career and life.
4 Abstract
This master thesis examines the impact of Control Tightness on Job Tension and Job Performance in the field of Professional Service Firms (PSFs). The study was conducted through a survey on 312 professionals working in the broad area of PSF worldwide. The results reveal that the PSFs emphasize the use of MCS more on Cultural and Personnel Controls, rather than Behavior and Results Controls. The Tightness of Cultural and Personnel Controls is positively related with Job Performance. The study also reveals that the same Control Tightness has a negative impact on Job Tension. Finally, the study investigates the influence of contingency variables size and strategy on the extent of Control Tightness. The findings of the study provide empirical support for the stated prediction that Control Tightness in PSF increases Job Performance and decreases Job Tension. The study contributes to understanding of MCS in the PSF and provides ideas for further research in this relatively unexplored field.
5 Contents
1. Introduction ... 7 2. Theory ... 9 2.1. Literature Review ... 9
2.2. Professional Service Firms ... 9
2.3. Management Control System ... 10
2.4. Control Tightness ... 11 2.5. Job Tension ... 12 2.6. Job Performance ... 13 2.7. Research hypotheses ... 13 3. Research methodology ... 15 3.1. Survey project ... 15 3.2. Response Analysis ... 16 3.3. Data screening ... 17 3.4. Variable Management ... 17
3.5. Exploratory factor analysis (EFA) ... 18
3.6. Reliability Analysis ... 22
3.7. Control variables ... 23
4. Research results ... 25
4.1. Descriptive Statistics ... 25
4.2. Main Findings ... 28
4.3. Additional Analyses: control variables ... 32
4.4. Additional Analyses: independent sample t-tests ... 33
6
5.1. Conclusion ... 36
5.2. Limitations ... 37
5.3. Directions for further research ... 38
References ... 39
Appendix 1A: Respondents characteristics (gender and occupation) ... 43
Appendix 1B: Respondents characteristics (location and education level) ... 44
Appendix 1C: Respondents characteristics (work experience and qualification) ... 45
Appendix 2: Questionnaire - Survey Questions ... 46
Appendix 3A: Factor Analysis ‘Behavior Controls’ ... 59
Appendix 3B: Factor Analysis ‘Results Controls’ ... 60
Appendix 3C: Factor Analyses ‘Cultural Controls’ ... 61
Appendix 3D: Factor Analyses ‘Personnel Controls’ ... 62
Appendix 3E: Factor Analysis ‘Job Tension’ ... 63
Appendix 3F: Factor Analysis ‘Job Performance’ ... 64
Appendix 4A: Group statistics by ‘work experience’ and ‘commercialization’ ... 65
7
1. Introduction
Over the past decades we witnessed a remarkable change from an industry-based economy to service-based variant (Rogers et al, 1994; Greenwood, 2005; Buera et al, 2009). One of the key actors in this development, are the Professional Service Firms (PSFs). Examples of such PSFs are accountant firms, lawyers, engineering companies and actuarial services (Von Nordenflycht, 2007). Due to their increasing importance, it is important to obtain a better understanding in how these firms are organized and managed. Various literature and case studies have revealed the management control systems (MCS) in different types of firms (Chenhall, 2003; Anthony and Govindarajan, 2004; Simons, 1995; Fischer, 1995; Hared et al.; 2003). There are also various notions made about the effectiveness of Control Tightness (Merchant and van der Stede, 2003; Abernethy, 2004; Widener, 2007). However, research of MCSs in this specific profession field is still lacking (Von Nordenflycht, 2010).
This thesis aims to analyze whether there is a link between the level of Control Tightness on the individual performance of a professional. Previous studies revealed that high Control Tightness improves firm performance (Kober et al, 2007; Chow, 1983). Some studies found evidence that individual employee performance is also positively affected by tighter controls (Chow, 1983; Shields et al., 2000). However, it is not exposed whether these conclusions could also be applied on professionals. Existing literature considers professionals as individuals who highly value their autonomy and resist higher Control Tightness (Raelin, 1985; Ram, 1999). Other distinctive characteristic is that the professionals perform a high level of non-routine tasks, which are basically hard to measure for a non-professional (Shields et al, 2000; Raelin, 1985 Ram, 1999). In order to assess individual Job Performance, the managers prefer to tighten controls and ensure that professionals act in a desirable way (Raelin, 1985; Merchant and Van der Stede, 2003). On the other side, professionals experience tightening controls and procedures as a barrier to perform this job adequately (Shields et al, 2000; Raelin, 1985 Ram, 1999). Professionals tend to retrieve their satisfaction by focusing on high quality of their jobs, rather than complying to rules and procedures (Raelin, 1985). In this light a Management Control System, or at least a tight one, could decrease job satisfaction and thus lower Job Performance (Raelin, 1985; Ram, 1999). Low job satisfaction may also lead to increased Job Tension as experienced by a professional (Shields et al, 2000; Rogers et al, 1994; Judge et al, 2001). Shields et al. (2000) demonstrated that Job Tension has a negative relation on Job Performance. Contemplating the different views, this would imply that higher Control Tightness will lead to lower Job Performance and higher Job Tension.
8 Like stated before, early researches are conducted on Control Tightness, Job Tension and Job Performance. But from our best knowledge, no previous study sought to link all these aspects in the field of PSFs. Therefore, the thesis tries to answer the following research question:
What is the effect of Control Tightness on Job Tension and Individual Job Performance in the Professional Service Firms?
The remainder of this study is organized as follows. We next discuss the theoretical background of Control Tightness in the Professional Service Firms and conclude with the development of the research hypotheses. The third section discusses the research methodology. The fourth section reports our results and the fifth section concludes.
9
2. Theory
2.1.
Literature Review
This literature review aims to analyze the theoretical background of the research topic. Paragraph 2.2 discusses the Professional Service Firms, while paragraph 2.3 covers the theoretical background of Management Control Systems. Next in paragraph 2.4 and 2.5, Control Tightness, Job Tension and Job Performance will be addressed. The last paragraph of this section provides a wrap-up of literature review and concludes with the research hypotheses.
2.2.
Professional Service Firms
The thesis research will focus on the Professional Service Firms (hereafter: PSF). But what is a PSF? Despite growing economic importance of the service sector, no unambiguous definition was formed in literature (Von Nordenflycht, 2010). This led to cases of organizations considered by some researchers as service, while others saw this differently. Despite some difference in views, most authors state that the service firm is characterized by intangibility, inseparability, non-inventorability and variability (Reichheld and Sasser 1990). Intangibility means that products or outputs cannot be touched by customers, which makes measuring performance quite challenging. Inseparability means that products or outputs cannot be delivered or measured separate from each other, due to intangibility. Non-inventorability refers to the notion that service products cannot be stored. Variability means that results provided by service firms are mainly processed by non-routine activities, leading to low standardization of results. Løwendahl (2005) mentions five factors which characterize PSF in extreme extent: intangible outputs, invisible assets, strong interaction with clients, innovation and high information asymmetry. The aspect of innovation is entailed with low amount of routines and standard procedures, while interaction with clients requires a high level of professional judgement and individual autonomy to deliver tailor made solutions.
Considering the above mentioned characteristics, we can notice that measuring performance of this type is difficult, if not impossible. But to what extent do the Professional Service Firms differ from other service types? Buera et al (2012) point out the PSF is characterized by high skilled labor force. This point of view is also shared by other authors (Løwendahl, 2005). Greenwood et al. (2005) added unusual inputs and outputs as main
10 characteristic of PSF, although they focused on accounting firms. Von Nordenflycht (2010) on the other hand, developed a theoretical framework around the PSFs. This broad framework revealed that the PSFs shared the same characteristics, namely: Knowledge Intensity, Low Capital Intensity and Professionalized Workforce. The typical responses and challenges on these firms is very difficult retention of employees, high preference of autonomy and informality and heavy reliance on ethical ‘soft’ controls (Auzair and Langfield-Smith, 2005).
The study of Raelin (1985) steps one level further and considers that high preference of autonomy could have negative influence on performance and positive impact on Job Tension. However, the same study doesn’t provide empirical evidence on this matter.
2.3.
Management Control System
Management control system (hereafter: MCS) is a system which managers use to ensure that the behaviors and decisions of their employees are consistent with the organization’s objectives and strategies (Merchant and van der Stede, 2007). This system can differ from one firm to other, depending on such characteristics as strategy, size and environment (Chenhall, 2003; Yan, 2011). However, the types of control remain the same across all organizations. Merchant and Van der Stede (2007) indicate the following four different types of control: Action1 Controls, Results
Controls, Personnel Controls and Cultural Controls (see figure 1). This research builds on categorization provided by the Merchant & van der Stede-Framework.
11 Figure 1: Management Control System matrix (Merchant and van der Stede, 2012).
As we read in chapter 2.2, the ability to measure results in the PSF is very low. The same can be said about the knowledge which actions are desirable. This information asymmetry between management and professionals has impact on how to organize the MCS in the PSF. Results Controls try to influence the professional’s performance, through sanction and compensation system. Behavior Controls aim to ensure that professional behavior is in line with organization goals. Personnel Controls focus on aspects such as personnel career plan and professional skills. Cultural Controls focus on the organization culture, e.g. mission and vision. According to the Merchant & Van der Stede framework, the management control system will rely heavily more on culture and personnel controls, and less on Behavior and Results Controls. It is even the question whether Results Controls are effective in this setting, since measuring performance in the PSF is very hard, if not impossible. Effective management control on professionals is more challenging, due to Knowledge Intensity, Low Capital Intensity and Professionalized Workforce (Van Nordenflycht, 2010).
2.4.
Control Tightness
12 (2007) define Control Tightness as the level of adherence to rules, policies and plans. Merchant and Van der Stede (2007) define Control Tightness as the “degree of certainty that employees will act as the organization wishes’’. So, Control Tightness refers to both each specific control, as well as the composition of the control package (Widener, 2007). Tightness may increase benefits, but may also increase the cost of controls. A potential disadvantage is that it deters creativity and innovation, which can be crucial in a PSF (Raelin, 1985; Ram, 1999).
If we look to the characteristics of the PSF, it is interesting to see to what extent the tightness of the control system should reach. Control Tightness can be either Implicit or Explicit. Explicit tightness can be achieved by more controls and rules. Implicit Tightness can be achieved by reducing the level of tolerance for deviations from the MCS. Both implicit and explicit Control Tightness could be applied for all types of control (Results, Behavior, Personnel and Cultural). Prior researches have not addressed how the four control types effect the use of MCS. Do the four control types behave the same way or does the effect of a control type from the other? Perrow’s Model of Structure and Technology (Perrow, 1970) distinguishes between formal administrative mechanisms and informal co-ordination mechanisms. Formal administrative mechanisms can be seen as equivalent of results and behavior controls, while informal co-ordination mechanisms represent Personnel and Cultural Controls. Perrow (1970) suggests that these two types of mechanisms differ to some extent in certain situations. Abernethy and Brownell (1997) draw on that model and found out that in situation of low task analyzability and high task variety, personnel controls have a stronger effect than accounting or behavior controls2. On other situations, with higher task analyzability and/or lower task variety,
the results show the opposite. Their study suggests that Personnel and Cultural Controls react inversely to Behavior and Results Controls. Therefore, the Control Tightness will not be considers as one single package, but divided over Results/Behavior Control and Personnel/Cultural Controls both having an opposite effect.
2.5.
Job Tension
Job Tension is the extent to which employees are bothered by work-related matters (Shields et al, 2000; Rogers et al, 1994). Studies have related Job Tension to various serious dysfunctional consequences and lower job satisfaction. Shields et al (2012) found a positive and significant relation between Control Tightness and Job Tension. In the same study, they also found evidence
2
Abernethy and Brownell (1997) mention ‘personnel controls’, which is equivalent with Cultural and Personnel Controls in this thesis. Accounting controls correspond with results control.
13 between negative and significant effect of Job Tension on Job Performance. Other researchers also point in this direction, although no empirical evidence is provided (Rogers et al, 1994; Judge et al, 2001). All the prior studies refer regarding their Control Tightness to Behavior and/or are Results Control. So, we expect to find a positive relation between Results/Behavior Control Tightness and Job Tension. Like argued in chapter 2.4, Personnel/Cultural Control Tightness have contrawise negative relation on Job Tension. This is not addressed in other researches, but can be reasoned with the logic that professionals who identify their selves with the beliefs and values of the company are less subject to conflict situations and thus lower degree of Job Tension.
2.6.
Job Performance
Many researches have revealed that performance is positively related with Control Tightness (Chow, 1983; Shields et al; 2000). In contrast, Kenis (1979) argued the opposite point of view. Shields et al (2000) examined the relation between Control Tightness and Performance and found a positive relation, but not significant. However, their research focused on design engineers at a Japanese automotive company, raising the question whether these outcomes apply on the PSF. Moreover, individual performance of a professional is very hard to measure. The professionals value autonomy and informality, believing that less rules and controls lead to better performance (Raelin, 1985). Abernethy and Brownell (1997) found out that personnel control has a positive and significant effect on performance. This effect is significantly more positive than that of either accounting or behavior controls. Their study was conducted on senior research officers in the research and development divisions of a large Australian industrial company and a major US scientific organization, which is to some level comparable to professional service firms. While their study examined managerial performance instead of Job Performance, the research gives an indication that Personnel Control Tightness increases performance.
Considering all different views, we would state that Results/Behavior Control Tightness is likely to have a negative effect on Job Performance. Otherwise, Personnel/Cultural Control Tightness will have positive impact on Job Performance.
2.7.
Research hypotheses
14 H1a: Behavior and Results Controls are positively related with Job Tension.
H1b: Cultural and Personnel Controls are negatively related with Job Tension.
H2a: Behavior and Results Controls are negatively related with Job Performance. H2b: Cultural and Personnel Controls are positively related with Job Performance.
Figure 2: Theoretical model
Figure 2 displays the graphical illustration of the research hypotheses. The control variables in this research will be size and strategy. In paragraph 3.7 the operationalization of the control variables will be discussed, along with some theoretical underpinning. The theoretical model does not predict the impact of size and strategy. The regressions are estimated by the following formulas:
1) JOB_TENS = β0 + β1 * BEHAV_CTRL + β2 *RESLT_CTRL + β3 *CULT_CTRL + β4 *
PERS_CTRL + β5*SIZE + β6*STRATEGY + εt
2)
JOB_PERFRMANCE = β0 + β1 * BEHAV_CTRL + β2 * RESLT_CTRL + β3 *15
3. Research methodology
3.1.
Survey project
The Faculty of Economics and Business Studies at the University of Amsterdam initiated a project aiming to research management control systems in professional service firms. Students were able to enter the research project to write the master thesis. In order to join the project, they had to obtain a minimum of 7 completed questionnaires. The respondents must meet the following criteria:
● The respondent works in the Professional Service Firm (excl. non-profit or government); ● The respondent has worked in the field for at least 3 years;
● The respondent is not an owner/partner or board member of the organization. In other words, the respondent needs to be subject to the management accounting and control system rather than design it;
● The respondent works in a medium/large size organization (> 50 employees);
The deadline of submitting the questionnaires was on February 1st, 2016. At the due date,
an amount of 372 questionnaires3 are received. After confirming the project participation, the
students could start to work individually on their thesis. The questionnaire has many constructs4, so the project members were able to develop their original research question (table 1). This thesis examines the impact of Control Tightness on Job Tension and Job Performance, while organizational size and strategy will serve as control variable. The relatively high amount of data seems robust enough and provides high internal and external validity.
Table 1: the survey contains the following constructs. The independent variables are highlighted in bold, dependent variables in red and control variables are Italic.
• Human Capital Intensity • Task Complexity
• Customer Reliance • Environmental Uncertainty
• Compensation/Reward Structure • Strategy
• Professionalized Workforce • Organizational Structure
• Organizational Reputation • Organization Type
• Organization Size • Professional Tension
• Control Tightness • Performance
3 The expect amount at the start of the project was about 150-200 respondents. 4 The whole questionnaire can be found in Appendix 2
16 - Explicit Behavior Control Tightness
- Implicit Behavior Control Tightness - Explicit Results Control Tightness - Implicit Results Control Tightness - Explicit Personnel Control Tightness - Implicit Personnel Control Tightness - Explicit Cultural Control Tightness - Implicit Cultural Control Tightness
o Unit Performance o Individual Performance
Before distributing the surveys to respondents, two pre-tests were conducted to reach a high level of quality effectiveness. These pre-tests were done by the project supervision staff. While the first pre-test focused on the quality of the variable measurement, the second test gave more attention to the survey as a whole. The findings resulted in minor changes in the survey5.
The final version of the survey was ready on November 20th, allowing the students to distribute
this among the respondents.
3.2.
Response Analysis
The data results were collected from one source, namely survey. The response analysis aims to measure reliability of the response, since the data results and conclusions rely heavily on the quality of the received responses. Like described in the paragraph 3.1, the survey project was designed to send only the survey link to specific individuals (contacts of students/project members who met the project requirements). The aim was to prevent that surveys were send randomly, specifically to not eligible respondents. However, under this set-up it was not possible to measure the amount of respondents who received the survey invitation and did not respond. Therefore, this set-up didn’t leave room to analyze the non-response bias. The amount of participating students was limited and the delivered questionnaires were verified by the project supervision staff. During the data collection procedure no irregular occasion occurred. In the data screening process (paragraph 3.3) missing and invalid responses are analyzed. In the light of absence of an adequate response analysis, no explicit conclusions can be made about the
5
Fourteen professionals took part on the first pre-test. They were asked to identify statements which represented the best definition of control tightness. The results of the first pre-test revealed a maximum of six mismatches (implicit behavior control). The second pre-test led to minor changes in wording and addition of few multiple-choice questions.
17 respondent validity and the sample representativeness. However, there is no reason to assume that responses are not representative for the population.
3.3.
Data screening
A total amount of 372 responses were received at the end of the data collection period. A data screening procedure was undertaken to analyze the quality of the responses. A total of 46 (12.4%) respondents has not finished or submitted the questionnaires, and thus removed from the sample. The answers of the remaining 326 respondents where examined on missing values and low standard deviation. Considering the relative high amount of 167 variable items, it is not unthinkable that some items were missing. The threshold of missing values was determined on a maximum of 10% blank. Furthermore, a low standard deviation indicates that answers do not vary sufficiently. For example, a respondent fills the same answer for all questions. The answers might express the actual score, the results for a low variated questionnaire is not useful for the analysis. The threshold of standard deviation was determined on a maximum of 0.5 (on a scale of 5). These criteria are applied for items measuring Control Tightness (all the four types), Job Tension, Individual Performance (both general and in-role), Size, and Strategy. This step removed 14 (3.8%) respondents and further reduced the sample to 312 (83.9%) responses.
The total of 312 respondents contained also 89 (28.5%) respondents who did not meet some requirements6. These limitations however do not pose a serious violation to overall findings, since the requirements are based on insufficient theoretical support and thus guidelines than hard rules. Several independent sample tests are performed to check whether these respondents differ (see additional analysis; chapter 4.4), which will provide interesting insights on the PSF characteristics. Removing this large part of the sample (almost one-third) will also reduce the quality of the data. Therefore, it is decided to keep these respondents in the sample. Any potential deviations will be discussed during the additional analyses.
3.4.
Variable Management
The dependent variables in this research are Job Tension (Question 33) and Job Performance (Questions 35/36). The constructs to measure work tension are adapted from prior research by (Rizzo et al., 1970) and (Aranya and Ferris, 1984). The constructs to measure Job Performance
6
18 respondents had less than 3 years work experience, 13 respondents worked in a non-profit organization and 43 were employed in a firm with less than 50 employees.
18 are based on prior research by (Podsakoff and MacKenzie, 1989) and (Welboune, Johnson and Erez, 1998). Furthermore, the Control Tightness is used as the independent variable. In the survey this variable is not measured in one single item, but further allocated over the four control types: Behavior Control (Question 4), Results Controls (Questions 5/6/7), Personnel Controls (Question 3) and Cultural Controls (Question 10). The construct Behavior Control is adapted from the research of (Hage and Aiken, 1968), (Van den Ven and Ferry, 1980) and (Cunningham and Rivera, 2001), while Results Controls is based on (Hage and Aiken, 1968), (Simons, 1987) and (Van den Stede, 2001). Cultural Controls is in the prior research not excessively explored, however some guidelines are retrieved from O’Reilly and Chatman (1986). Personnel Controls is a new construct, which is not examined before as a stand-alone item. The effect of Control Tightness will be measured in two steps. The first method is to test the effect of the four control types individually on Job Tension and Job Performance. Then four control types will be tested together in one single regression model.
Finally, the items Strategy (Question 40) and Firm Size (Question 15) will serve as control variables. The construct of Strategy is derived of the case study of (Auzair and Langfield-Smith, 2005). The measurement of Organization Size is a new item, not specifically based on prior study.
3.5.
Exploratory factor analysis (EFA)
In the previous chapter (3.4) we identified the relevant variables for the research. Table 2 shows a summary of construct variables and corresponding question group numbers:
Table 2: Construct variables used in this research.
Construct Variable Item measurement Number of items Nature of variable
Job Tension Question 33 8 Dependent variable
Job Performance Questions 35/36 19 Dependent variable
Behavior Controls Question 4 13 Independent variable
Results Controls Questions 5/6/7 14 Independent variable
Cultural Controls Question 10 9 Independent variable
Personnel Controls Question 3 11 Independent variable
Strategy Question 40 11 Control variable
Organization Size Question 15 1 Control variable
19 The 85 items were grouped in 10 questions asking respondents’ opinions on the extent of agreement in terms of aspects of the Control Tightness (all the four control types), Strategy and Job Tension. The items were measured on a five-point scale ranging from “strongly disagree” (1) to “strongly agree” (5), except 12 items measuring Job Performance in-general. These items ranged from “Needs much improvement’’ (1) to “Excellent’’ (5). Questions 6 and 7 measure tightness of Results Controls, but unlike question 5 they differ in scaling and wording. Question 6 is about how many performance targets are used in the evaluation of Job Performance and ranges from “zero’’ (1) to “9 or more’’ (5). Question 7 focus on how often the respondent discuss the performance results with their supervisor and varies from “Daily’’ (1) to “Les often than annually’’ (6). The item about organization size asked “How many people are employed by your entire company?’’ and was answered by (1) “Less than 100”, (2) “More than 100 but less than 500”, (3) “More than 500 but less than 5000” or (4) “More than 5000”.
The exploratory factor analysis was conducted on the two Dependent variables and four Independent variables, containing 74 items. To optimize the quality of the factor analysis, it was essential to reverse code 12 items7 of the survey and to rescale 1 item8. The 74 items were not specifically categorized into a certain amount of control groupings, but was dependent on the outcomes of (rotated) component loadings. A set of principal component analyses reduced the sampling to 56 items (table 3). The extractions were based on loadings at least 0.400 and negative loading were removed, with varimax rotation was conducted to substantiate these groupings.
Table 3: Rotated Component Matrix
Rotated Component Matrix Component
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Q35_3 .830
(1) Job Performance as measured by supervisor
Q35_2 .828 Q35_5 .821 Q35_1 .818 Q35_6 .795 Q35_7 .722 Q35_4 .695
7 The following 12 questions are negatively formulated in wording: Q3_7, Q3_11, Q4_10, Q4_11, Q4_12, Q4_13,
Q5_9, Q5_10, Q5_11, Q7, Q10_5 and Q33_3.
8 Question 7 was measured on a 6-point scale. For the factor analysis ‘annually’ (5) and ‘less often then annually’ (6)
20 Q33_5 .772 (2) Job Tension Q33_8 .756 Q33_4 .723 Q33_7 .696 Q33_2 .660 Q33_1 .580 Q33_6 .520 Q36_10 .805
(3) Process improvement as performance indicator
Q36_9 .798
Q36_11 .781
Q36_12 .641
Q4_2 .783
(4) Degree of formalization rules/procedures
Q4_4 .718
Q4_1 .703
Q4_11 .616
Q4_8 .543
Q5_7 .767
(5) (Rigid) Strictness of Performance Measurement
Q5_6 .759 Q5_1 .660 Q5_4 .599 Q7R .547 Q5_12 .435 Q10_8 .769 (6) Socializing/Teambuilding Q10_4 .723 Q10_9 .716 Q10_2 .610 Q10_7 .526 Q36_8 .793
(7) Personal career as performance indicator
Q36_5 .766 Q36_7 .750 Q36_69 .592 Q4_12R .814 (8) Degree of compliance to Rules/procedures Q4_13R .774 Q4_10R .729 Q3_2 .840 (9) HR Policy (individual) Q3_1 .814 Q3_3 .687
9 Question 36_6 was both loaded on Factor 3 (0.412) and Factor 7 (0.592). Both factors represent Job Performance.
21
Q36_3
(10) Work Output as performance indicator
.669
Q36_2 .651
Q36_1 .494
Q610 .441
Q3_6
(11) HR policy (team formation) .830
Q3_8 .781
Q10_6
(12) Personal involvement/identification employee .696
Q10_3 .606
Q3_11R
(13) HR policy (team formation) .770
Q3_7R .699
Q5_9R
(14) Flexibility towards deviations from targets .805
Q5_11R .682
Extraction Method: Principal Component Analysis.
Rotation Method: Varimax with Kaiser Normalization. Rotation converged in 22 iterations.
Table 4 shows the final results of the factor analyses. 14 grouping factors were extracted, explaining variance of 64,809%. The Bartlett test of Sphericity was significant (χ
2 = 7085.77, p
< 0.001), and the Kaiser-Meyer-Okin measure of sampling adequacy was greater than 0.6 (KMO = 0.814), indicating that factor analysis on the sample was possible. Less than 1% percent of the correlations in the correlation matrix exceeded 0.300, indicating high discriminant validity. Convergent validity revealed 1 case of cross-loading (see footnote 10). Finally, face validity was undertaken to test whether factors are consistent with the construct variables as designed in the survey (table 4).
Table 4: Factors from EFA in relation with construct variables
Variable Items Factor loadings
Job Tension Question 33 Factor 2
Job Performance Questions 35/36 Factor 1, 3, 7 and 10
Behavior Controls Question 4 Factor 4 and 8
Results Controls Questions 5/6/7 Factor 5 and 14
Cultural Controls Question 10 Factor 6 and 10
Personnel Controls Question 3 Factor 9, 11 and 13
Considering the comparisons in table 4, the factor loading based on EFA do not violate the construct design of the variables. For instance, items about Job Tension are not grouped with
10 Question 6 measured intentionally Results Controls. However, the EFA grouped this item together with Job
Performance. The question formulation could both be interpreted both as performance measurement and control tightness.
22 Personnel Controls. However, the 6 variables are further divided over 14 factors. A reason could be that the survey distinguishes between implicit and explicit controls (for all control types). The survey also makes a distinction between Job Performance in general and in-role. Our thesis does not go into these detailed categorizations and treats the four controls and Job Performance as summarized in table 1.
3.6.
Reliability Analysis
The exploratory factor analysis showed that the construct design of the survey has a strong statistical background. Despite the fact that EFA identified 14 factors, the same analysis does not show any significant inconsistencies in variables constructs. Since the theoretical framework builds on the 2 dependent and 4 independent variables (table 2), we decided to continue with the item structure followed in the survey project. The variables, as labeled in table 2, were subjected to reliability analyses. The reliability is measured by the Cronbach’s alpha (1951), which has to exceed 0.60 to be considered as acceptable (Nunnally, 1978).
Table 5: Reliability analysis-Control Tightness, Job Tension and Job Performance
Items constructs Cronbach’s alpha
Job Tension .819 Job Performance .889 Behavior Controls .711 Results Controls .768 Cultural Controls .790 Personnel Controls .752
Table 5 summarizes the Cronbach’s Alpha values for the several test variables. The details of this factor analyses can be found in appendices 3 (a-f). The reliability tests aimed to exceed the minimum level of 0.700 for the six cases, which provides more robust background for creating the construct variables. Items deteriorating the Cronbach’s Alpha to lower than 0.70 were removed. In some cases, it was possible to get higher Cronbach’s Alpha then listed in table 5. However, the improvement is negligible and does not outweigh the information loss caused by
23 removing items. As last step, the construct variables are calculated as the mean of the underlying items (see appendix 3).
3.7.
Control variables
Finally, the variables size and strategy served as control variables. As mentioned in paragraph 3.4, strategy variable was retrieved from the study on the effect of business strategy on bureaucratic MCS in service organizations by Auzair and Langfield-Smith (2005). In that research, the strategy construct measured by 11 items on 7-point scale11. The authors separated between the first four and the other seven items, measuring respectively cost leadership and differentiation according to Porter’s (1980) competitive strategy framework. Unlike that research, this thesis operationalized the variable strategy as one construct. We opted to conduct exploratory factor analysis to discover whether the number of factors will match the two factors in the study of Auzair and Langfield-Smith (2005). Results of the factor analysis are summarized in Table 6. The EFA identified three factors in contrast with the study of Auzair and Langfield-Smith (2005), which may indicate that the sample does not base their strategic decision on cost leadership and differentiation. Chenhall (2003) examined the link between strategy and MCS in his contingency based research. The few studies which address the topic, are mainly built on frameworks including entrepreneurial-conservative (Miller & Friesen, 1982); prospecters-analysers-defenders (Miles & Snow, 1978); build-hold-harvest (Gupta & Govindarajan, 1984); and product differentiation-cost leadership (Porter, 1980). However, the item loading on table 6 are not sufficiently covered by these frameworks, since those do not focus on the PSF. Therefore, the strategic theoretical model of Løwendahl (2005) is used. Løwendahl (2005) contributed significantly on the understanding of the strategic positioning by the professional service firms. Løwendahl (2005) distinguished three strategies: A) Client relation based strategies, B) Solution or output based strategies and C) Problem solving or creativity based strategies.
Table 6: Factor analysis—Strategy
Items loading A. Client
relation based strategies. B. Solution or output based strategies. C. Problem solving or creativity based strategies.
1) Lower cost than competitors ,728
11 In the thesis research, the same questions are asked but measured on a five-point scale instead of seven-point
24
2) Cost efficient ,652
3) Improving coordination cost ,623
4) Improving utilization ,726
5) Introduce new service quickly ,657
6) Distinct service from
competitors ,580
7) Broader range of services ,489
8) Time to provide services ,705
9) High quality services ,521
10) Customizing services ,420
11) After-sales service ,431
Variance explained (%) 55,9 %
Cronbach´s Alpha ,735 ,673 ,566
Type A firm, or “‘individually based firm”, are often small partnerships. They lean more towards individual expertise, individual reputation, and individual client networks. Type B firms, the “professional bureaucracy”, are the more developed counterparts. They focus more on firm level expertise, firm reputation, routines and methodologies. The last category, type C firm or “expert firm”, can be considered as a compromise between the other extremes. This type of firms aims to deliver unique solutions to complex problems.
Concerning the size variable, question 15 asked “How many people are employed by your entire company?’’ and was answered by (1) “Less than 100”, (2) “More than 100 but less than 500”, (3) “More than 500 but less than 5000” or (4) “More than 5000’’. Form the literature study, is this item measurement new. Though it was possible to treat the size variable as (linear) scale, we chose to create 3 dummy variables with the small firms as reference group. The role of this group has received little attention in the contingency-based MCS literature (Chenhall, 2003).
25
4. Research results
This section has four parts. The first part reports descriptive statistics for the variables and the correlations, the second part describes the regression models and the main findings, and the other two parts present the additional analyses. The characteristics of the sample group are enclosed in Appendix 1 a-c.
4.1.
Descriptive Statistics
Table 7 presents the descriptive statistics for all variables and table 8 displays the variable means grouped by size. The variables can be mutually compared, due to the same measurement scale (five-point scale). The figures show that Cultural and Personnel Controls are higher scaled than Behavior and Results Controls. This is according to the rationale that PSF in general emphasize their MCS more on Cultural and Personnel, rather than Behavior and Results Controls (Merchant and Van der Stede, 2007). This pattern can be observed over all the size types (table 8). In general, the Control Tightness figures show an increasing pattern as the firm size becomes higher. Furthermore, the strategy variables show that STRA_CREATIVITY is higher than STRA_CLIENT and STRA_OUTPUT. It is likely that PSFs, characterized by professionalism (Løwendahl, 2005), are more willing to combine delivering unique solutions with intensive client relationship.
Table 7: Descriptive statistics
N Mean Std. Deviation Minimum Maximum
Theoretical Range JOB_TENSN 312 2.3444 .69578 1.00 4.38 1 - 5 JOB_PERFRMANCE 312 3.8706 .49571 2.58 5.00 1 - 5 BEHAV_CTRL 312 3.0353 .61591 1.00 4.88 1 - 5 RESLT_CTRL 312 2.8127 .68311 1.00 4.57 1 - 5 CULT_CTRL 312 3.5582 .71444 1.25 5.00 1 - 5 PERS_CTRL 312 3.3350 .81844 1.00 5.00 1 - 5 STRA_CLIENT 312 3.5468 .68743 1.60 5.00 1 - 5 STRA_OUTPUT 312 3.5267 .80493 1.00 5.00 1 - 5 STRA_CREATIVITY 312 3.9690 .68577 1.67 5.00 1 - 5
26
Table 8: Descriptive Statistics (grouped by size) Mean (Total) Mean (1) (N=43) Mean (2) (N=77) Mean (3) (N=76) Mean (4) (N=114) JOB_TENSN 2.3444 2.3721 2.2345 2.3683 2.3994 JOB_PERFRMANCE 3.8706 3.7882 3.8267 3.8574 3.9480 BEHAV_CTRL 3.0353 2.9157 2.9608 2.9236 3.2135 RESLT_CTRL 2.8127 2.9894 2.7403 2.6807 2.8947 CULT_CTRL 3.5582 3.3632 3.3976 3.6814 3.6624 PERS_CTRL 3.3350 3.0872 3.1364 3.5089 3.4460 STRA_CLIENT 3.5468 3.4837 3.5656 3.5138 3.5632 STRA_OUTPUT 3.5267 3.5620 3.5043 3.4956 3.5409 STRA_CREATIVITY 3.9690 4.1240 3.9697 3.8465 3.9912
(1) Firms with less than 500 employees
(2) Firms with more then 500, but less than 1500 employees (3) Firms with more then 1500, but less than 5000 employees (4) Firms with more than 5000 employees
Table 9 displays the correlation matrix. The correlation matrix indicates that there are no correlations greater than .70 between the variables, multicollinearity will not be problematic. The variable MCS_TGHTNESS_TOTAL is included to measure the impact of a total MCS package which the average of BEHAV_CTRL, RESLT_CTRL, PERS_CTRL and CULT_CTRL. Like prior research (Shields et al, 2000) have stated, the correlation between Job performance (JOB_PERFRMANCE) and Job Tension (JOB_TENSN) is confirmed to be significant negative (r = .252, p ≤.01). The Behavior Controls (BEHAV_CTRL) has no significant correlations with Job Tension and Performance. The Result Controls (RESLT_CTRL) has positive correlation with Job Tension (significant) and also with Performance (not-significant), confirming the study of Shields et al. (2000). Personnel (PERS_CTRL) and Cultural Controls (CULT_CTRL) are both significant correlated with Performance and Tension. However, the correlation with performance is positive and negative with tension. The dummy sizes variables have no significant correlations with performance or tension. Only large firms (SIZE_DUMMY_SUPER) are positive significant with performance, while other dummy variables negative. Medium-sized firms (SIZE_DUMMY_MEDIUM) seem to have the best fit with Job Tension. Concerning strategy, all three variables are positively correlated with performance. For Job Tension only (STRA_CREATIVITY) is significant, while other strategies are negative correlated. This may be explained by the notion professionals prefer high autonomy, which can be guaranteed the best by the ‘expert firm’-strategy (Løwendahl, 2005). Finally, the variable (MCS_TGHTNESS_TOTAL) is positive correlated with Job Performance and negative with Job Tension, underscoring that
27 T a b le 9 : C o rr el at io n m at ri x P ea rs o n C o rr el at io n (N = 31 2) (1 ) (2 ) (3 ) (4 ) (5 ) (6 ) (7 ) (8 ) (9 ) (1 0) (1 1) (1 2) (1 3) (1 ) JO B _ T E N S N (2 ) JO B _ P E R F R M A N C E -. 25 2 ** (3 ) B E H A V _ C T R L .0 42 -. 03 0 (4 ) R E S L T _ C T R L .0 95 * .0 45 .3 43 ** (5 ) P E R S _ C T R L -. 25 3 ** .2 22 ** .1 18 * .2 38 ** (6 ) C U L T _ C T R L -. 33 3 ** .2 55 ** .0 02 .1 51 ** .4 03 ** (7 ) S IZ E _ D U M M Y _ SM A L L .0 16 -. 06 7 -. 07 8 .1 04 * -. 12 1 * -. 10 9 * (8 ) S IZ E _ D U M M Y _ M E D IU M -. 09 1 -. 05 1 -. 06 9 -. 06 1 -. 13 9 ** -. 12 9 * -. 22 9* * (9 ) S IZ E _ D U M M Y _ L A R G E .0 20 -. 01 5 -. 10 3 * -. 11 0 * .1 21 * .0 98 * -. 22 7* * -. 32 5* * (1 0) S IZ E _ D U M M Y _ S U P E R .0 60 .1 19 * .2 20 ** .0 91 .1 03 * .1 11 * -. 30 3* * -. 43 4* * -. 43 1* * (1 1) S T R A _ C L IE N T -. 02 7 .1 32 ** .1 01 * .1 49 ** .0 82 .0 91 -. 03 7 .0 16 -. 02 7 .0 18 (1 2) S T R A _ O U T P U T -. 07 1 .1 72 ** -. 11 1 * .0 40 .0 97 * .0 86 .0 18 -. 01 6 -. 02 2 .0 13 .3 78 ** (1 3) S T R A _ C R E A T IV IT Y -. 17 9 ** .2 75 ** -. 07 6 .0 81 .1 02 * .1 78 ** .0 91 .0 01 -. 10 2 * .0 25 .3 64 ** .4 18 ** (1 4) M C S_ T G H T N E S S _ T O T A L -. 17 4 ** .1 89 ** .5 86 ** .6 82 ** .6 13 ** .6 53 ** -. 07 6 -. 15 3 ** -. 00 7 .2 09 ** .1 68 ** .0 38 .1 15 * ** C o rr el at io n i s si gn if ic an t at t h e 0. 01 l ev el ( 1-ta ile d ). * C o rr el at io n i s si gn if ic an t at t h e 0 .0 5 le ve l (1 -t ai le d ). N eg at iv e co rr el at io n s ar e sh o w n p in k, p o si ti ve c o rr el at io n s in g re en .
28 PSFs lean more towards Cultural and Personnel, rather than Behavior and Results Controls. In addition to correlations of Job Performance and Job Tension, other correlations are analyzed. Hereby no clear patterns are found between Size and Control Tightness, though the most correlations are significant. Large firms (more than 5000 employees) are positive correlated with Personnel and Cultural Controls. Medium and Small firms seem to put very low focus on this type of controls; small firms (SIZE_DUMMY_SMALL) focus more on Results Controls.
Personnel and Cultural Controls are strongly correlated with ‘expert-firm’-based and ‘professional bureaucracy’-based (STRA_OUTPUT) strategies. Behavior Controls is in contrast very negative with strategies (except STRA_CLIENT was positive). Results controls is only significant positive with ‘individually based PSF’ (STRA_CLIENT), while other are positive and non-significant. There seems a logical hierarchy to exist between the level of strategy and control type. Strategies (from type A to type C) and Control Tightness (from Behavior to Personnel Controls) seem to move parallel. Lastly, no significant correlations are found between Strategy and Size, and no clear patterns are discovered.
4.2.
Main Findings
After the descriptive statistics and correlations are analyzed, two multiple regression analyses are undertaken to test the hypotheses. The first multiple regression analysis treats the impact of Control Tightness on Job Tension, while the second regression models addresses the Job Performance. Both the analyses contain a model summary and coefficient modelling. The predicted effects of Control Tightness are summarized in table 10. More theoretical background concerning the predicted effects can be found on theory part (chapter 2). The effects of the control variables are not predicted.
Table 10: Hypothesized effect of Control Tightness on Job Tension and Job Performance
Hypothesis JOB_TENSN Hypothesis JOB_PERFRMANCE
BEHAV_CTRL 1A
+
2A
-
RESLT_CTRL 1A
+
2A
-
CULT_CTRL 1B
-
2B
+
PERS_CTRL 1B
-
2B
+
29 Tables 11 and 12 present the results of regression analysis on Job Tension. Six models are shown. Model 0 examines the base model containing only the control variables SIZE and STRATEGY. Only variable STRA_CREATIVITY was significant. In Model 1 the variable BEHAV_CTRL is added. The effect of Behavior Controls is not significant (P > 0.90) and the model does not improve the explanatory power (Sig. F Change > 0.85). Model 2 contains RESLT_CTRL, which coefficients turn out to be positive and significant at 0.10-level. The following models (3, 4 and 5) show a significant improvement of the Adjusted R Square (F <0.001), indicating that CULT_CTRL and PERS_CTRL have a negative impact on JOB_TENS. Model 5 includes all independent variables, to test the total effect of control tightness. This model confirms negative effect of Personnel and Cultural Control Tightness on Job Tension (P < 0.001). There is also a certain level of support for the positive relation of Results Control Tightness on Job Tension (P <0.100). However, no support is provided for eventual Behavior Control Tightness on Job Tension.
Table 11: Multiple Régression Coefficients – Job Tension
Model 0 Model 1 Model 2 Model 3 Model 4 Model 5
Independent Variables Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig.
BEHAV_CTRL .011 .857 -.037 .515 RESLT_CTRL .099 .083 * .204 .000 *** CULT_CTRL -.339 .000 *** -.296 .000 *** PERS_CTRL -.264 .000 *** -.204 .001 *** Control Variables Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig.
SIZE_DUMMY_MEDIUM -.093 .246 -.093 .246 -.080 .322 -.080 .293 -.086 .271 -.047 .526 SIZE_DUMMY_LARGE -.024 .768 -.024 .769 -.007 .929 .052 .496 .038 .634 .124 .106 SIZE_DUMMY_SUPER .013 .873 .011 .897 .018 .833 .088 .275 .069 .398 .139 .083 * STRA_CLIENT .048 .441 .046 .467 .033 .598 .056 .340 .057 .346 .038 .515 STRA_OUTPUT -.009 .885 -.008 .905 -.006 .924 -.009 .883 .004 .947 .003 .965 STRA_CREATIVITY -.196 .002 *** -.195 .003 *** -.198 .002 *** -.132 .032 ** -.173 .006 *** -.131 .030 **
30 The explanatory power of the models is measured by Adjusted R Square (table 11). The base model has an Adjusted R2 of 2.4%, which ascends to 17.9% in Model 5. This indicates that control tightness in general explains 15.5% of Job Tension, which is very evident. Model 3 has the highest Adjusted R2 on individual basis (R2 = 13%, F > 7.600, P < 0.001). Model 4 is also significant on 0.001-level. This is in accordance with the conclusions from the descriptive statistics and correlations that PSFs rely in their MCS tightness in significant level on Personnel and Cultural Controls. The level of model change is for all models (expect Model 1) significant, underlining the strong robustness of the regression models. Thus, we can conclude that regression models confirm that Personnel and Cultural Control Tightness are negatively related with Job Tension. The models also show that Results Control Tightness is positively associated with Job Tension. These results are in accordance with the hypotheses 1B and partly 1A (table 10). The models do not support the hypothesis concerning Behavior Controls Tightness.
Table 12: Regression Model Summary – Job Tension
Model R Adjusted R Square F Sig. Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 0 .208a .024 2.296 .035b .68724 .043 2.296 6 305 .035 1 .208b .021 1.966 .059c .68833 .000 .032 1 304 .857 2 .229b .031 2.413 .020c .68497 .009 3.021 1 304 .083* 3 .386b .130 7.616 .000c .64912 .106 37.869 1 304 .000*** 4 .330b .088 5.294 .000c .66441 .065 22.317 1 304 .000*** 5 .453b .179 7.772 .000c .63051 .162 15.338 4 301 .000***
a. Dependent Variable: Job Tension
The results of regression analysis on job performance are displayed in tables 13 and 14. The model comparisons are identical of those used for Job Tension. In the base model 0 only variable STRA_CREATIVITYwas significant (P < 0.001). Model 1 shows BEHAV_CTRL has a negative impact on job performance, though this effect is not significant (P > 0.500). In Model 2 is the RESLT_CTRL positively associated with JOB_PERFRMANCE, which contradicts the hypothesis 2A (table 10) and is consistent with the findings of prior research (Shields et al, 2000). However, this effect is not significant (P > 0.700). The next models (3 and 4) show a significant improvement of the Adjusted R Square (F <0.005), indicating that CULT_CTRL and PERS_CTRL have a positive impact on JOB_PERFRMANCE. The total model 5 confirms positive effect of Personnel and Cultural Control Tightness on job performance (P < 0.05). On
31 the other side, no evidence is provided for possible Behavior and Results Control Tightness on Job performance.
Table 13: Multiple Regression Coefficients – Job Performance
Model 0 Model 1 Model 2 Model 3 Model 4 Model 5
Independent Variables Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig.
BEHAV_CTRL -.034 .548 -.042 .482 RESLT_CTRL .021 .711 -.020 .742 CULT_CTRL .194 .001 *** .149 .013 ** PERS_CTRL .175 .002 *** .131 .030 ** Control Variables Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig. Beta Sig. SIZE_DUMMY_MEDIUM .073 .347 .074 .342 .076 .331 .066 .390 .068 .374 .062 .419 SIZE_DUMMY_LARGE .120 .126 .120 .127 .123 .119 .076 .328 .079 .312 .052 .508 SIZE_DUMMY_SUPER .195 .017 ** .203 .015 ** .196 .017 ** .152 .061 * .158 .053 * .143 .084 * STRA_CLIENT .016 .797 .022 .722 .013 .839 .011 .854 .010 .870 .018 .762 STRA_OUTPUT .062 .318 .057 .363 .063 .314 .062 .311 .053 .386 .049 .431 STRA_CREATIVITY .251 .000 *** .247 .000 *** .250 .000 *** .214 .001 *** .235 .000 *** .208 .001 *** *, ** and *** denotes 10%, 5%, and 1% significance levels (two-tailed), respectively.
The explanatory power of the base model amounts8.1%, which ascends to 12.0% in Model 5 (table 14). Thus, Control Tightness in general explains 3.9% of Job Performance, which is remarkably lower than the Job Tension models. The low R2 is however consistent with the finding of Shields et al. (2000) and Auzair and Langfield-Smith (2005). Like Job Tension, model 3 has the highest Adjusted R2 on individual basis (R2 = 11.4%, F > 6.700, P < 0.005). Model 4 is also significant on 0.005-level. This is in accordance with the conclusions from the descriptive statistics and correlations that PSFs rely in their MCS tightness in significant level Personnel and Cultural Controls. The level of model change is for all models (expect Model 1 and 2) significant, underlining the strong robustness of the regression models. Thus, we can conclude that
32 regression models confirm that Personnel and Cultural Control Tightness are positively related with Job Performance. These results are in accordance with the hypothesis 2B (table 10). The models do not support the hypothesis 2A concerning Behavior and Results Controls Tightness.
Table 14: Regression Model Summary – Job Performance
Model R Adjusted R Square F Sig. Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 0 .315a .081 5.596 .000b .47509 .099 5.596 6 305 .000 1 .317b .080 4.838 .000b .47559 .001 .362 1 304 .548 2 .316b .079 4.802 .000b .47576 .000 .137 1 304 .711 3 .366b .114 6.718 .000b .46659 .035 12.214 1 304 .001*** 4 .358b .108 6.375 .000b .46819 .029 10.054 1 304 .002*** 5 .385b .120 5.243 .000b .46500 .049 4.346 4 301 .002***
a. Dependent Variable: Job Performance
As last step in the regression model, multicollinearity tests are performed using variance inflation factors (VIF). The maximum VIF should not exceed the critical value of 10 (Neter et al. 1990). All VIF-values were lower than 2.00 for the independent variables BEHAV_CTRL, RESLT_CTRL, CULT_CTRL and PERS_CTRL. The VIF-values for the SIZE-variables were between 2.00 and 3.00. For the STRA-variables this was between 1.00 and 2.00. This analysis reveals no evidence of a collinearity problem.
4.3.
Additional Analyses: control variables
The paragraph 4.2 discussed the outputs of regression models on the hypotheses tests. This paragraph will broaden the scope to check whether other variables or factors than Control Tightness have influence on job performance and Job Tension. This is in particular relevant for job performance, since those regression models showed a low R2. The first part of the additional analyses focused on the control variables SIZE and STRATEGY. The control variables have more impact on job performance (R2 = 8.1%, F > 5.500, P < 0.001) than on Job Tension (R2 = 2.4%, F > 2.200, P < 0.050). The regression models (table 12 and 14) agree that PSF following the ‘expert-firm’ strategy (STRA_CREATIVITY) are associated with higher job performance and
lower Job Tension for the respondents. Contrary to the other strategies (STRA_CLIENTand STRA_OUTPUT), the variable STRA_CREATIVITY is significant at 0.01-level implying that PSF are willing to adopt this strategy. Furthermore, the results show that size does not have
33 significant impact on Job Tension. Larger firms allow their employees to perform better (B=0.195, P < 0.02). However, this impact weakens if the Control Tightness variables are included in the regression model. To gain better understanding about Job Performance in PSF, other control variables should be included in future regression models. Also the size measurement should be further development, to determine any possible impact.
4.4.
Additional Analyses: independent sample t-tests
The independent sample group testing formed the other part of additional analyses. Four independent sample tests were undertaken to measure whether certain characteristics have significant on Control Tightness, Job Performance and Job Tension. The Levene's Test for Equality of Variances tested whether equal variances are assumed (by P > 0.100).
Table 15: Independent Samples Test - grouped by profession Does the Respondent work in Accounting?
No (0) = 239 cases Yes (1) = 73 cases
Levene's Test for Equality of
Variances t-test for Equality of Means
F Sig.
Mean
Difference t Sig. (2-tailed) BEHAV_CTRL Equal variances not assumed 4.434 .036 -.31378 -4.262 .000*** RESLT_CTRL Equal variances not assumed 9.694 .002 -.08323 -1.066 .288
CULT_CTRL Equal variances assumed 1.883 .171 -.19767 -2.080 .038**
PERS_CTRL Equal variances assumed 1.760 .186 -.10362 -.947 .345
JOB_TENSN Equal variances assumed 1.335 .249 .01584 .170 .865
JOB_PERFRMANCE Equal variances assumed 1.638 .202 .07247 1.094 .275
In table 15 the respondents working in the accounting branch are compared with their counterparts working in other PSF branches. This was driven by the rationale that the accountants will be overrepresented, since the project members (mostly accountancy students) approached the respondents in their close network. The results of this test indicate that professionals working in accounting, experience higher Behavior Control Tightness (µdifference=
0.314, P < 0.001) and higher cultural Control Tightness (µdifference= 0.198, P < 0.001). The higher
Behavior Control Tightness could be explained by high degree of law and regulations by which the accountants are confronted with.
34
Table 16: Independent Samples Test - grouped by firm size Does the Respondent work in a firm >100 employee?
No (0) = 43 cases Yes (1) = 267 cases
Levene's Test for Equality of
Variances t-test for Equality of Means
F Sig.
Mean
Difference t Sig. (2-tailed)
BEHAV_CTRL Equal variances assumed .001 .971 -.14243 -1.408 .160
RESLT_CTRL Equal variances assumed .231 .631 .20010 1.802 .073*
CULT_CTRL Equal variances not assumed 8.822 .003 -.22826 -1.949 .052*
PERS_CTRL Equal variances assumed .548 .460 -.28740 -2.151 .032**
JOB_TENSN Equal variances assumed .026 .872 .02908 .254 .799
JOB_PERFRMANCE Equal variances not assumed 4.668 .032 -.09903 -1.043 .302
Table 16 compares the groups of professionals working in firms with less than 100 employees with other groups. One of the survey project requirements was that the respondents have to work in firms with more than 50 employees. However, there was no item in the questionnaire which measured this requirement. Question 15 grouped respondent by (1) “Less than 100”, (2) “More than 100 but less than 500”, (3) “More than 500 but less than 5000” or (4) “More than 5000’’. The sample test compared the group who answered (1) with other group sizes. There is a significant difference in Control Tightness between these two groups. The larger the company lead to lower Results Control Tightness (µdifference= 0.200, P < 0.100) and higher
Personnel (µdifference= 0.287, P < 0.050) and Cultural Control Tightness (µdifference= 0.228, P <
0.050). This suggests that the regression model were mainly driven by large firms (see also Auzair and Langfield-Smith, 2005). The MCS of small firms differs significantly from their large counterparts, justifying the requirement of excluding small firms in the survey project. Future research could focus on this sample group, to gain better understanding of MCS in small firms, since this group received little attention (Chenhall, 2003).
Table 17: Independent Samples Test - grouped by profit and non-profit Does the Respondent work in commercial organization?
No (0) = 35 cases Yes (1) = 277 cases
Levene's Test for Equality of
Variances t-test for Equality of Means
F Sig.
Mean
Difference t Sig. (2-tailed)
BEHAV_CTRL Equal variances assumed .411 .522 .02458 .222 .824
RESLT_CTRL Equal variances assumed .185 .667 -.34883 -2.880 .004***
CULT_CTRL Equal variances assumed .244 .621 -.34602 -2.728 .007***
35
JOB_TENSN Equal variances assumed .844 .359 .22723 1.827 .069*
JOB_PERFRMANCE Equal variances assumed .835 .361 .01830 .206 .837
Other requirement of the survey project was that the respondents have to work in for-profit organization. However, 35 respondents12 appeared to work for non-profit organization. Table 17 reveals surprisingly that professionals in non-for-profit organizations experience higher Results Control Tightness (µdifference= 0.348, P < 0.001) and Cultural Control Tightness (µdifference=
0.346, P < 0.001) than then their colleagues in the commercial firms. The difference in Results Controls cannot be explained by existing literature. The low number of the respondents could misrepresent the outputs. Future research on PSF should include this sample group, to gain better understanding of MCS in non-for-profit firms.
Table 18: Independent Samples Test - grouped by work experience Does the Respondent have at least 3 years work
experience? No (0) = 18 cases
Yes (1) = 281 cases
Levene's Test for Equality of
Variances t-test for Equality of Means
F Sig.
Mean
Difference t Sig. (2-tailed)
BEHAV_CTRL Equal variances assumed .067 .796 .03209 .214 .831
RESLT_CTRL Equal variances assumed .006 .941 .05648 .335 .738
CULT_CTRL Equal variances assumed .113 .737 -.15401 -.878 .381
PERS_CTRL Equal variances assumed 1.185 .277 -.13406 -.667 .505
JOB_TENSN Equal variances assumed .207 .649 .00871 .051 .959
JOB_PERFRMANCE Equal variances assumed .603 .438 -.20693 -1.729 .085*
The last sample test focused on work experience. This was also a survey project requirement, excluding respondents with less than three years work experience. Generally, no (significant) differences exist between the both groups (see table 18). Only JOB_PERFRMANCE differs significant at 0.100-level. The more experienced staff deliver higher performance than the junior colleagues (µdifference= 0.207, P < 0.010). However, considering the low number (18
respondents) and significance above the 0.050- level, no clear statements can be done on this matter.
12
36
5. Discussion and conclusion
5.1.
Conclusion
The aim of this research was to examine the impact of Control Tightness on Job Tension and Job Performance. The research was conducted under Professional service types, an area which received little attention (Van Nordenflycht, 2010). The theoretical framework of this study combined literature from MCS perspectives, especially Merchant and van der Stede (2007) including areas that have received little attention in prior MCS studies, namely, the impact of different control types on business strategy and firm size. To some extent, results of this study reconfirm some findings from similar prior studies. However, this study also offers several new insights into the MCS design in PSFs.
Figure 3 displays the research results. The test results indicate that there is no ‘single’ Control Tightness, but Control Tightness differs four each control type (i.e. Behavior, Results, Cultural and Personnel control). As predicted in the theoretical part of this thesis, PSF’s build their MCS on Cultural and Personnel Controls. The test results confirm that Cultural and Personnel Control Tightness have an adverse impact on Job Performance and Tension then Behavior and Results Controls. Hypotheses 1b and 2b were supported by regressions test, suggesting that Control Tightness has a positive impact on Job Performance and negative effect on Job Tension. Professionals are characterized by high knowledge and have a strong preference for autonomy, resisting higher control tightness. This poses a conflicting situation with the firms which have interests to high Control tightness, to achieve their organizational goals. Professionals are more benefited by Cultural and Personnel Controls, as it appear from the data results. PSFs seem to fit their MCS on the needs of professionals.
Hypotheses 1a and 2b are not supported, albeit the results confirm that tightness of Results Controls increases Job Tension. The robustness tests show that Control Tightness explains Job Tension to a large extent. The explanation rate for Job Performance is remarkably lower, but not negligible. This is consistent with the study of Shields et al. (2004), held under design engineers at a Japanese manufacturing firm. Furthermore, opting for ‘expert-firm’ service strategy has the same impact as Personnel and Cultural Control Tightness. Finally, firm size has little influence on Job Performance and none on Job Tension. This study has important implications for management practices in PSFs.
