Amsterdam Business School
Decentralization, tight budget control
and explorative innovation
Name: Richard Molenkamp Student number: 10681132
Supervisor: mw. dr. ir. B.A.C. (Bianca) Groen Date: June 21, 2015
Word count: 12725, 0
MSc Accountancy & Control, variant Control
Statement of Originality
This document is written by student Richard Molenkamp
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
Management control systems are used in almost all organizations. A management control system should be designed to achieve organizational goals. In explorative innovative settings, use of formal management control system design is found to co-exist with, limit, excel or ignore the organizational goals with respect to explorative innovation. In this study I examine the effect of formal management control systems, in particular tight budget control on explorative innovation. In contrary to earlier studies, I examine this effect by considering the formal management control system a result of organizational design. A survey was completed by 112 managers of operational departments, mainly based in the Netherlands. The four hypotheses formulated are based on the agency theory and the self-determination theory. These hypotheses are: 1) more decentralization will lead to more explorative innovation. However, 2) more decentralization will also lead to more use of tight budget control. Which in return will 3) lead to less explorative innovation. By considering the management control system design a result of organizational design, the relation of management control systems with explorative innovation is considered a cum hoc ergo propter hoc fallacy. This could then possibly explain why the findings of various studies contradict each other. Overall, I expect 4) tight budget control suppresses the relation of decentralization with explorative innovation. I did find support for the first and third hypothesis. The null hypothesis with respect to the second and fourth hypothesis, could not be rejected. I discuss theoretical, and practical implications for the view on the use of formal management control systems and make suggestions for future research.
Contents
1 Introduction... 6
2 Theory ... 9
2.1 The variables and their relation... 9
2.2 Decentralization ... 10
2.3 Innovation... 10
2.4 Tight budget controls ... 11
2.5 The relationship of decentralization with explorative innovation ... 12
2.5.1 Self-determination theory ... 13
2.6 The relation of decentralization with tight budget controls ... 14
2.7 The relation between tight budget controls and explorative innovation ... 15
2.8 The indirect effect... 16
2.8.1 The indirect effect: a suppressor variable... 16
2.9 Control variables ... 17
3 Methodology... 18
3.1 Participants... 18
3.2 Constructs ... 19
3.2.1 The construct decentralization... 23
3.2.2 The construct tight budget control ... 23
3.2.3 The construct Innovation... 24
3.2.4 Control variables... 24
3.3 Statistical analyses... 26
3.3.1 Equations ... 27
4 Results... 29
4.1.1 The Sobel test... 31
5 Discussion and conclusion ... 33 5.1 Theoretical implications ... 33 5.2 Limitations... 34 5.3 Future research ... 35 5.4 Practical implications ... 35 Literature... 37 Appendix A... 40
1 Introduction
There has been an increased interest in the relation between innovation and the use of formal management control systems. The existing literature has produced contradictory findings. The findings varied from management control systems: ignoring-, being incompatible with-, coexisting next to- or stimulating- innovation (Bisbe & Otley, 2004, p. 710). The conclusions on these four main findings are summarized hereafter.
In their study, Dougherty and Hardy (1996) found that ”large, mature organizations could not achieve sustained innovation because innovators within them could not solve innovation-to-organization problems . . . it depended primarily on individuals rather on the innovation-to-organizational system” (p. 1146). More strikingly, the budget process merely seems to support day to day work. The orientation was not on innovation. Innovation was occasionally even removed from budgets (Dougherty & Hardy, 1996). Abernethy and Stoelwinder (1991) found: “a mismatch between task uncertainty and the use of budgets (high/high)” (p. 114), when studying the “fit” between budget use, task uncertainty, system goal orientation and subunit performance. A lower use of budget control will lead to a higher task uncertainty, thus creating more space for innovation. In the study of Bart (1991), a comparison was made between the use of formal and informal controls in an innovative setting, or rather the coexistence of the management control systems. With new products, the emphasis on informal control was greater and presidents of the firms in the study were more involved in the innovative process (1991). This was because: “it was necessary to take extra precautions, to make sure that everything was done right” (Bart, 1991, p. 11). Bart does mention that the formal management controls were applied more loose as this would enable innovation. All previous authors have been negative to neutral on the use of formal management control systems. The founder of the ‘levers of control’, Simons (1991), has another view on the use formal management control systems. “Top managers understand that formal process is often essential to foster dialog from which new ideas and action plans can emerge” (p. 61). This is a very positive view on the use of formal management control systems. The views of the authors are all very distinct as to how formal management control systems deal with innovation. This seems very logical when we look at Simons (1990) casual observation: “all large, complex organizations have similar types of management control systems. Short and long range plans, financial budgets, capital budgets, variance analyses and project reporting systems are common place tools in virtually every large, professionally managed corporation” ( p. 135). One would expect organizations to adopt a management control system that ‘fits’ the goals of the organisation. And the goals of all large, complex organizations are likely to differ. The addressed studies provided findings in either positive, neutral or negative directions. The existing literature has to my opinion not yet sought to
understand the effect of management control systems on innovation. If the effect of the formal management control systems on innovation seems difficult to establish despite the fact that all large and complex organizations have similar types of management control systems, the outcome of the various studies is most likely influenced by another independent variable. Dougherty and Hardy (1996) were primarily focused on the influence of the management on the innovation process (e.g. manager innovativeness). Abernethy and Stoelwinder (1991) looked at the relation of budget with performance measurement and Bart (1991) and Simons (1991) were linking strategy to the management control systems.
One could argue that management control systems do not directly attribute to the power of innovation. The relation between management control systems and innovation could be biased. This is the starting point in this study. This study considers the effect of management control systems on innovation acum hoc ergo propter hoc fallacy. Management control systems are only present due to the fact that another independent variable is correlated with management control systems. None of the studies addressed before have also looked at the influence of decentralization on innovation. However, the effect of decentralization on innovation is expected to have a positive effect, as will be discussed in the theory section. Therefore, in this study decentralization is the independent variable. Decentralization is the independent variable for innovation. Decentralization does have a relation with the use of formal management control systems and therefore mistakenly management control systems are held accountable for the change in innovative power. Management control systems are merely present or absent in the process of innovation and are not really correlated with this process, although I do expect I will find correlation because of the fact that a management control system is always present in organizations. Thus, because it is a cum hoc ergo propter hoc fallacy. This study contributes to the existing literature because it considers use of formal management control systems a consequence of organizational design. To show management control systems are present but cannot be hold accountable for everything that ‘is’, the formal management control system will be defined as a suppressor variable. “Unlike mediator variables whose effects are often presented as some form of intervention under the investigators control, suppressors are not typically theorized and discussed in the introduction section of articles” (Ludlow & Klein, 2014, p. 2). In this study, the suppressing effect is discussed in the theory section. To empirically prove the relation, a survey is used. This survey will specifically aim at the use of one of the most formal management control systems, the tightness of budget control. Bart (1991) defined “the tightness of operating budgets” and “the amount of detail in the operating product plans and budgets of subordinate managers that the president sees” (p. 7), as formal controls in his study.
Another distinctive notion should also be taken into account. The focus of Bart (1991), Simons (1990) and Abernethy and Stoelwinder (1991) has primarily been on new product innovation. However, innovation can come in two distinctive forms. It can either be explorative or exploitative, where explorative is new innovations and explorative is (some) improvement to existing products, clients or markets (Jansen, Van Den Bosch, and Volberda, 2006). The difference is further elaborated in the theory section. It is mentioned here because the main question will only focus on one of these two possibilities.
That main question is: will the effect of decentralization on explorative innovation be more positive if formal tight budget controls are considered a suppressor?. This main question will be divided into three sub questions: 1) does high decentralization lead to a high level of explorative innovation?, 2) does high decentralization lead to a high level of use of formal tight budget controls? and 3) does high use of formal tight budget controls lead to a lower level of explorative innovation? Again, the latter is of course an interesting notion with respect to the discussed literature in this introduction. If that sub question will be answered with yes, which is the objective of this study, then formal management control systems (or formal tight budget controls) do apparently have a negative effect on explorative innovation. However, this is only be caused by the fallacy. If organizational design would be different from decentralization, the effect of the formal tight budget control on explorative innovation could, possibly due to this difference, also be positive, neutral or co-existing.
In paragraph two decentralization, management control systems and tight budget control ,and innovation are discussed. Also, in this section the relation between those variables will be explained through the agency and self-determination theory, and the hypotheses based on the theory are developed. The research method, will be further explained in paragraph three. In paragraph four the findings of the survey will be discussed. The discussion and conclusion will be addressed in the fifth paragraph,
2 Theory
In this section, the hypotheses will be developed, based on existing literature. First, the variables and their relation is described. The variables are then further explored, followed by theorizing the relation between the variables. Based on the theory, the hypotheses will be developed.
2.1 The variables and their relation
This study will look at the effect of decentralization on explorative innovation on the one hand and the effect of decentralization on explorative innovation through the use of formal tight budget controls on the other. The first relation is the direct effect. The latter the indirect effect. Decentralization is expected to increase explorative innovation. Decentralization is the independent variable. Explorative innovation is the dependent variable. The use of formal tight budget controls is a dependent variable for decentralization. But in relation to the explorative innovation, it is an independent variable. Decentralization is expected to increase the use of formal tight budget controls. The latter is expected to have a negative effect on explorative innovation. The variables and their relations are demonstrated in figure 2.1.
In figure 2.1, four relations are demonstrated. For all four relations, hypotheses have been developed. In the next paragraphs, the variables and their relations are explained, based on existing literature. On the basis of this literature review, the four hypotheses which are needed to empirically prove this model, will be developed.
2.2 Decentralization
“In the decentralized solution, the organization determines the appropriate performance measure for the unit, and it is the person or unit's responsibility to figure out what the performance drivers are, and how they influence performance” (Jensen, 2002, p.250). Decentralization in essence is moving power from the higher level (owner, CEO) towards the lower level (division or BU management) of an organization and measuring if this power is used by lower level to achieve the organization’s objectives. “In decentralized firms, decisions are delegated to managers who may have information superior to that of the owners of the firm” (Rajan & Saouma, 2006, p. 677). Delegation of decision rights will under normal circumstances only happen if the knowledge transfer costs are higher than the costs of control. Which is clearly the case if “Some of this information superiority arises due to the manager’s expertise regarding local conditions, such as his knowledge of consumer preferences in specific markets or for specific product attributes” (Rajan & Saouma, 2006, p. 677). The trade-off between knowledge transfer cost and cost of control is best described by Bushman, Indjejikian, and Penno, (2000): “Determining the optimal decentralization rights and information in organizations requires balancing the cost of poor decisions due to lack of relevant information against those due to the divergent objectives of principal and agents” (p. 579).
2.3 Innovation
Innovation can be very divers. “Common types of innovation relate to new products, materials, new processes, new services, and new organizational forms (Ettlie and Reza, 1992). These different forms of innovation draw to varying extents on different teams, departments, and professional disciplines” (Bareghe, Rowley, & Sambrook, 2009, p. 1234). Since the definition of innovation is divers, Bareghe et al. (2009) conducted a study to a multidisciplinary definition of innovation. The definition Bareghe et al. (2009) found is: “Innovation is the multi-stage process whereby organizations transform ideas into new/improved products, service or processes, in order to advance, compete and differentiate themselves successfully in their marketplace” (p. 1334). Transforming ideas into ‘new/improved products’ raises queries with respect to the uniformity of this definition. New and/or improved, a distinction for these two categories is mandatory. Jansen
et al. (2006) distinct exploratory and exploitative innovation. Both these two types of innovation are explained by the Jansen et al. (2006) in multiple definitions. The most distinctive with respect to exploratory innovation are: “They offer new designs, create new markets, and develop new channels of distribution (Abernathy and Clark, 1985). Exploratory innovations require new knowledge or departure from existing knowledge (Benner and Tushman 2002, Levinthal and March 1993, McGrath 2001)” (Jansen et al., 2006, p. 1662). For exploitative innovation: “Conversely, exploitative innovations are incremental innovations, and are designed to meet the needs of existing customers or markets (Benner and Tushman 2003, p. 243; Danneels 2002)” (Jansen et al., 2006, p. 1662).
The distinction between exploratory and exploitative innovation is important for this study. Jansen et al. (2006) formulated amongst others the following hypotheses “Hypothesis 1. The higher a unit’s centralization of decision making, (a) the lower its level of exploratory innovation, and (b) the higher its level of exploitative innovation” (p. 1663). They found support for 1(a) but not for 1(b). So centralization does not lead to high levels of exploitative innovation. However, there are also studies that have proven that exploitative innovation is increasing in a more standardized environment, an environment where people work with strict procedures and guidelines. Benner and Tushman found that exploitative innovation under ISO 9000 conditions increased. This effect was stronger if the incremental innovations were lower. They described ISO 9000 as process intensity (2002). Thus a high level of process intensity will lead to high levels of exploitative innovation.
Since explorative and exploitative innovation are two totally different ways of innovating, with other drivers enabling this, in this study I have chosen to only study the effect of decentralization and tight budget control on explorative innovation.
2.4 Tight budget controls
Use of formal tight budget controls is one of the possible tools of the management control systems (Bart, 1991). “The term management control systems refers to the set of procedures and processes that managers and other organizational participants use in order to help to ensure the achievement of their goals and the goals of their organization (Otley & Berry, 1994)” (Bisbe & Otley, 2004, p. 710). These systems can be used diagnostic or interactively. If the formal management control systems are used interactively, managers “become personally and regularly involved in the decision activities of subordinates and that become the basis for continual exchange between top managers and lower level of management as well as between organizational members” (Bisbe & Otley, 2004, p. 717). There is a downside to the interactive use of management control systems, especially if all
systems are used interactive. “Organizations in transition where all systems are interactive report incredible stress as employees are pushed to their limits to respond to the short-term information and action demands of superiors. Top managers begin monitoring and asking” (Simons, 1991, p. 59). As mentioned earlier, Simons (1995) has composed the ‘levers of control’ which is often cited. These ‘levers of control’ consist of diagnostic control-, belief-, boundary- and interactive control systems. The latter is aimed at “focus on constantly changing information that senior managers consider potentially strategic” (Simons, 1995, p. 87). The ‘levers of control’ is about finding the right balance and set of controls to be used. But as Simons (1995) puts it: “managers gain still more control by using a variety of diagnostic controls- among them profit plans, budgets, and goals and objectives”(p. 88). Management control systems, and also tight budget controls can be used in an interactive or in a diagnostic way. This study will focus on the diagnostic use of these formal systems, to be more specific: the diagnostic use of formal tight budget controls. Simons (1991) identified amongst others profit planning systems as a type of control system. “These management control systems focus on individual business units and encompass annual profit plans or budgets, . . . these systems report planned and actual revenues and expenses for each major business by revenue and cost-category” (Simons, 1991, p. 53). As discussed in the introduction, Bart (1991) identified tight budget controls as one of the formal controls. According to Bart (1991), budget control tightness is “the degree to which a subordinate manager can exceed his formal budget allocations and/or the amount of “cushion” or “discretionary” funds a manager has in his budget” (p.7). Bart also mentions other actions with respect to the budget which might be considered part of the formal management control systems, like the involvement of the president with the budget (1991). This distinction is also visible in the construct used in this study to measure tight budget control. Based on a factor analysis of this construct, and on the existing literature, certain initial items have been deleted from the construct. This is further elaborated in the construct section.
2.5 The relationship of decentralization with explorative innovation
“Innovation is about organizations transforming ideas” (Bareghe, et al., 2009, p. 1334). In order to be creative and transform ideas in new products, employees need to be high skilled and have specific knowledge of the area in which they are innovating. This specific knowledge is costly to transfer. High knowledge transfer cost will lead to delegation of authority (Bushman et al., 2000). If a firm does not commit to innovation and give some authority to innovators, innovators are likely to not succeed in innovation. Dougherty and Hardy (1996) found that “where individual innovation projects were successful, they depended on the efforts of particular individuals to use their organizational positions (e.g., Brass, 1984) to further and protect innovation efforts” (p.
1133). So only when individuals have the power to decide and make resources available, innovation will be successful.
2.5.1 Self-determination theory
As shown in the introduction, people play a key role in innovative processes. People have to put an effort in the innovative process, otherwise the process will not be successful (Dougerthy & Hardy, 1996). People will only put an effort in this process if they feel a need to do so. Need is one of the key elements in the Self-determination theory. According to Deci and Ryan (2000), “needs specify innate psychological nutriments that are essential for ongoing psychological growth, integrity and well-being” (p. 229). Deci and Ryan (2000) identified three needs: needs for competence, relatedness, and autonomy.
These needs motivate people. “Autonomy is essential for the intrinsic motivation” (Deci & Ryan, 2000, p. 234). And intrinsic motivation is essential to be able to deploy innovative activities. “Deci (1975) suggested that intrinsically motivated behaviors represent the prototype of self-determined activities: They are activities that people do naturally and spontaneously when they feel free to follow their inner interest” (Deci & Ryan, 2000, p. 234). As such, this feeling is essential in the process of innovation. Deci and Ryan (2000) draw on a study of Reeve and Deci (1996) in which was shown that applying more controls will lead to less intrinsic motivation. If people experience less autonomy they will not be able to “regulate their own actions in accord with their full array of felt needs and available capacity, thus coordinating and prioritizing processes toward more effective self-maintenance” (Deci & Ryan, 2000, p. 254).
Den Hartog and Belschak (2012) found that “a more autonomous context allows those high on RBSE1more room to act, experiment and contribute than a low-autonomy context” (p.
200) and Benner and Tusman (2002) found that “Increased process management was associated with a significant decline in the number of patents that were based entirely on knowledge new to the firm” (p. 693). This effect shows that the more steering on the way employees work, the less explorative innovation is achieved. Process management itself is only beneficial for organizational effectiveness under stable conditions. It is not if the organization is a radical innovator (Benner & Tusman, 2003).
The first hypothesis with respect to the effect of decentralization on innovation is based on self-determination theory and bear in mind the findings of Jansen et al. (2006) that
1“Role breadth self-efficacy (RBSE) refers to employees' perceived capability of carrying out a broader and more proactive set of work tasks that extend beyond prescribed technical requirements” (Parker, 1986, p. 835)
centralization has a negative effect on explorative innovation. I expect that a greater autonomy will lead to more explorative innovation:
H1: More decentralization will lead to more explorative innovation
Decentralization is believed to have this effect only on exploratory innovation. Based on existing theory, it is not likely that the null hypothesis with respect to the effect of decentralization on exploitative innovation can be rejected. Neither way the hypothesis will be formulated, negative or positive. Therefore, exploitative innovation will in this study primarily be used as a control variable. Benner and Tusman (2003) found that exploitative innovation increased in a controlled environment. It is of course interesting to see if a more loose environment will lead to less exploitative innovation. But this is another supposed direction then this study is aimed at. However, I will briefly report on this effect in the additional testing since the data is available.
2.6 The relation of decentralization with tight budget controls
This relation is best understood through an agency perspective. For the agency perspective, I have adopted the definition of Jensen and Meckling (1976): “a contract under which one or more persons (the principal(s)) engage another person (the agent), to perform some service on their behalf which involves delegating some decision making authority to the agent” (p. 308). The principal and agent characters can be found throughout an organization. It can be a CEO and a direct subordinate. In this study, an agent is the participant in the survey project. A principal would then be the superior of that participant. Delegation of authority will have the risk of misuse of information asymmetry by the agent. The agent can have more information than the principal. When information asymmetry occurs, there is a problem in verifying the agents actions. Abernethy, Bouwens, and Van Lent (2004) provided empirical evidence that information asymmetry is an important driver to decentralization. “Divergence of interest, in combination with imperfect information give rise to the post-contractual agency problem known in the literature as moral hazard” (Østergren & Stensaker, 2011, p. 154). Moral hazard is a commonly accepted definition for the change of attitudes of parties regarding activities if they are not directly accountable for the outcome of these activities. By delegating authority to a lower level, the principal wants to be sure that the firms’ goals are met. He will therefore place adequate controls on the agent. These controls can come in many forms. They can be incentives and/or actual monitoring, designed to limit the change in attitude of the agent from organizational interest to self-interest (Jensen & Meckling, 1976). Østergren and Stensaker (2011) especially place emphasis on the use of budgets to limit the risk of the agent using private information for his on behalf. With budgets, the principal can control the activities of the agent, more specific, his use of resources.
The second hypothesis is based on the phenomenon of the agency theory. One would expect formal controls to increase if decentralization is higher. In order to gain some form of control, tight budget control, being one of the most common diagnostic controls, would give some assurance to the principal (Simons, 1995). Formal tight budget control will therefore increase in the case, if more rights are delegated.
H2: Decentralization will lead to more use of tight budget control 2.7 The relation between tight budget controls and explorative innovation
Østergren and Stensaker (2011) concluded that ‘beyond budgeting’ empowers and gives more freedom to employees to reach their goals. Consequently, tight budget controls are a limitation for people. These systems do not give ultimate freedom when employees want to reach their goals. Budgets are not known for their contribution to the explorative innovation. As Hansen, Otley and Van der Stede (2003) concluded with respect to budgets, based on prior literature: “budgets; are not responsive to rapidly changing environments; impose a vertical command-and-control structure; centralize decision making; and stifle initiative” (p. 110).
To get a full understanding of why a tight budget control is not compatible with explorative innovation, I will discuss the same theory as is used in the previous paragraph, self-determination theory. This helps explain the effect size and direction of the influence of tight budget controls on explorative innovation. Tight budget controls, like process management techniques as discussed by Benner and Tushman (2002), “focus on improving an organization’s efficiency through high-level coordination of an organization’s activities in a rationalized system of end-to-end processes” (p. 677).
This is contradictory to one of the three needs identified by Deci and Ryan and pertaining to the self-determination theory (2000). The need for autonomy is violated. As discussed in the theory section of the relation between decentralization and explorative innovation, autonomy is essential for the intrinsic motivation (Deci & Ryan, 2000). If employees need to be stimulated to feel free to innovate, autonomy is essential. Not feeling autonomous can have demotivating effects on the intrinsic motivation:
Subsequent studies indicated that events such as evaluations, rewards, and choice, which had been shown to affect intrinsic motivation in reliable ways, also had corresponding effects on creativity, cognitive flexibility, and conceptual learning. For example, rewards and evaluations were found to decrease creativity (Amabile, 1982), complex problem
solving (McGraw & McCullers, 1979), and deep conceptual processing of information (Grolnick & Ryan, 1987) (Deci & Ryan, 2000, p. 234).
Intrisic motivation is affected by emphasizing control. The study of Reeve and Deci (1996) has been addressed shortly in the paragraph 2.5.1. This experiment showed that “the element that undermined intrinsic motivation was the controlling interpersonal context that pressured participants to win” (Reeve & Deci, 1996, p. 31)
Tight budget controls affect the autonomy of employees. They are not free to make the choices with respect to available resources they seem needy. As such, the intrinsic motivation of employees will be lower, due to that. Above the self-determination theory, Benner and Tushman (2002) concluded that “increasing the use of process management activities tips the innovation balance toward exploitation at the expense of exploration” (p. 702). Subsequently the following hypothesis can be derived from the literature above:
H3: More tight budget control will lead to less explorative innovation 2.8 The indirect effect
The theorization of the indirect effect is a summary of the hypotheses one to three. The figure 2.1 showed the indirect effect consists of the sum of: decentralization on explorative innovation, decentralization on the use of formal tight budget controls and the use of formal tight budget controls on explorative innovation. A higher level of decentralization will lead to more explorative innovation, based on the self-determination theory. A downside to decentralizing decision making, is that the principal would like to exercise more control over the agent. In particular a tight budget control would expand the principal’s possibility to monitor actions of the agent. So, an increase in decentralization will lead to a higher use of formal budget control systems. And a higher use of the latter, will of course lead to a lower level of explorative innovativeness. Due to the latter hypothesis, it is expected that the direct relation between decentralization and explorative innovation is suppressed by tight budget controls.
2.8.1 The indirect effect: a suppressor variable
The suppressor variable has been discussed in many studies (Friedmann & Wall, 2005; Ludlow & Klein, 2014; Menderhausen, 1939). The definition of Menderhausen is “A clearing variate in the strict sense is a useful determining variate without causal connection with the dependent variate; its role in the set consists of clearing another determining (observational) variate of the effect of a disturbing basis variate” (1939, p. 99). It is obvious, based on the 3rd hypothesis, a causal
Friedmann and Wall (2005): “Horst (1941) . . . gave the name “suppressor variable” to an independent variable that (1) has no correlation with the outcome variable, but (2) is correlated with the other independent variable” (p. 127). This contradicts hypothesis three and I expect this will not hold when running a bivariate correlation on the data used in this study. These previous examples already show that there is no uniform definition for a suppressor. Friedmann and Wall even mentioned ”that we have all been taught that correlation does not imply causality” (p. 135). Therefore the definition for suppressor variable from Conger (1974) will be used in this study:
A suppressor variable is defined to be a variable which increases the predictive validity of another variable (or set of variables) by its inclusion in a regression equation. This variable is a suppressor only for those variables whose regression weights are increased (p. 36).
The overall hypothesis with respect to the model 2.1 is:
H4: tight budget control will suppress the relation between decentralization and innovation. 2.9 Control variables
In this study five control variables will be used. These control variables are: exploitative innovation, manager innovativeness, manager optimism, educational level and size of the department in number of fte. The characteristics of the manager were used as a control variable because a high score on manager innovativeness, manager optimism and education is expected to lead to more explorative innovation. This is because these three characteristics are also related to a high RBSE. A high RBSE is related to more self-esteem and proactivity. Next to that, it is also shaped by organizational experiences (Parker, 1998). Exploitative innovation is chosen as a control variable because it is related to explorative innovation. It evolves from explorative innovations and co-exists in innovation departments with the explorative innovation. According to Benner and Tushman (2002), organizations persist in the challenge of retaining explorative innovation as exploitative innovation increases.
3 Methodology
Empirical data for this study was gathered by conducting a survey. The survey that was conducted was part of the survey project 2014/2015 of the Faculty of Economics and Business (FEB) of the UvA. The aim is to answer the research question for a broad population. A survey method is a useful and generally accepted tool in achieving this.
3.1 Participants
The survey project was a project in which students of the FEB could participate to collectively gather a substantial number of surveys. All students had to acquire a minimum of 7 completed surveys. Because of this joint effort, the total number of respondents was 112. The study was conducted in The Netherlands. However, some of the participants worked in other countries in Europe. 2 respondents worked in Belgium, 1 in Germany and 25 did not answer the question where there department was located. Respondents had to meet four criteria: (1) being a business unit manager; (2) managing at least 20 fte with at least some decision rights, (3) managing an operational department, so not HR, finance, IT and (4) direct client/customer impact was considered important. One of the participants was identified as an outlier. On the section of exploitative in the constructinnovative performance, respondent 88 was identified as an outlier, when testing for normality. The kurtosis had a value of 1.091 and the standard error of the kurtosis was .453. The Z-score was 2.41, which means a P < .05. Since the total sample size is > 100, central limit theorem tells us we do not have to worry about the assumption of normality (Field, 2013). Additional testing did prove this central limit theorem. By eliminating respondent 88, on hypothesis H1b changes from .208 to .202 and significance was somewhat lower. Therefore the outlier was not deleted. The 112 respondents did answer 34 questions with respect to the variables used in this study. Out of all these questions, a total of 3,808, 18 were unanswered. Table 3.1 gives an overview of the characteristics of the respondents.
Tabel 3.1 respondent characteristics
Characteristic
Age Low/ high/ mean (SD) 23/ 65/ 48 (9.1)
Number of fte Low/ high/ mean (SD) 4/ 450/ 74 (85.9)
Business unit tenure Low/ high/ mean (SD) 0.1/ 27/ 6.1 (5.8)
position tenure2 Low/ high/ mean (SD) 0.5/ 42/ 9.2 (9)
Education % lower % higher % academic % phd 4 47 44 5 Industry % agriculture % industrial environment % commercial service % professional service % financial services % non-profit 0 13 17 22 4 44 3.2 Constructs
The questionnaire used for this study was a broad pre-developed questionnaire. The survey of the FEB consisted of in total fourteen proven constructs. For this study, three of these constructs have been used as main variables for answering the research question. A construct for the independent variable decentralization, one for the suppressor variable tight budget control and one for the dependent variable innovation. The latter construct consisted of nine items, of which the first four pertained to explorative innovation and the last five to exploitative innovation. This construct was for that reason split into two constructs. The items pertaining to exploitative
2For the position tenure, the question was: “for how many years have you been working in this position?”. The mean is higher than business unit tenure. The question was clearly interpreted as position being independent from business unit.
innovation were used as a control variable since this study, with respect to innovation only focuses on explorative innovation. Next to exploitative innovation, two additional constructs were used as control variables. These constructs were: manager innovativeness and manager optimism. The overall aim of this survey was to investigate the design and use of management accounting & control systems. The items used to measure the constructs which are specifically used in this study are shown in appendix A. For the construct decentralization, respondents had to rate the questions on a 5-point likert scale: (1) my superior has all the decision rights, (2) my superior has most of the decision rights, (3) Decision rights are equally distributed between me and my superior, (4) I have most of the decision rights, (5) I have all the decision rights. For the other five constructs used, respondents had to rate the questions on a 5-point likert scale: (1) strongly disagree, (2) disagree, (3) neither agree nor disagree, (4) agree, (5) strongly agree. Since the surveys were collected by multiple students, control for non-response bias was not possible.
On all constructs, a varimax principal rotated component analysis (PCA) was conducted. This PCA was conducted at three stages. The first stage was a PCA per construct. The second stage was a joint PCA3on all latent variables4. The number of factors in the second stage was
limited on six, identical to the number of constructs. The results are summarized in table 3.2. The third stage was a joint PCA on all latent variables without a maximum number of factors. This is summarized in table 3.3. Before conducting these analyses, all negative worded items were reversed. These were items: 57, 59, 60, 62, 64, 66 and 67. For a loading to be significant, the value should exceed .505. Also, in the first and third stage, only components with an eigenvalue >1 were
selected.
The analysis of PCA in all three stages is elaborated on in the next paragraphs, where all construct are further explained. Next to the PCA, the Cronbach alpha was determined to assess the reliability of the constructs. Every construct was expected to exist of one component. However, if a construct exists of multiple components, the Cronbach Alpha should be calculated for each subtest (Cronbach, 1951). The reliability results are also explained under the appropriate constructs in the next paragraphs.
3A joint factor analysis is performed to for the following purpose: “The joint factor analysis also allows us to assess the potential for common rater bias (Harman [1967])” (Bouwens & Van Lent, 2007, p. 677)
4Including the latent control variables
Table 3.2 Summary of the joint factor analysis of all latent variables (with a maximum of six factors, factor loadings < .50
suppressed)
Components
Construct Item 1 2 3 4 5 6
Decentralization 20. Strategic decisions .755
21. Investment decisions .721
22. Marketing decisions .823
23. Decisions regarding internal processes .579 24. Human resources decisions
Manager
innovativeness 55.56. I often surprise people with my novel ideasPeople often ask me for help in creative activities .650.628
57. I obtain more satisfaction from mastering a skill
...coming up with a new idea .548
58. I prefer work that requires original thinking .556 59. I usually continue doing a new job in exactly the way it
was taught to me
60. I like a job which demands skill and practice rather than
inventiveness .695
61. I like to experiment with various ways of doing the same
thing .555
62. I am not a very creative person .733
Manager
optimism 63.64. In uncertain times, I usually expect the bestIf something can go wrong for me, it will .681
65. I’m always optimistic about my future
66. I hardly ever expect things to go my way .729 67. I rarely count on good things happening to me .686 68. Overall, I expect more good things to happen to me
than bad
.569 Tight budget
control 79.80. Puts much emphasis on meeting the budgetDoes not easily accept budget revisions during the year .553.691 81. Has a detailed interest in specific budget line items .638 82. Does not lightly tolerate deviations from interim budget
targets
.776
83. Is intensively engaged in budget-related
communications
.652 Explorative
innovation 84.85. Our business unit accepts demands that go beyondexisting products and servicesOur business unit invents new products and services .564.764
86. Our business unit commercializes products and services
that are completely new to them
.799
87. Our business unit frequently utilizes new opportunities
in new markets
.812 Exploitative
innovation 88.89. Our business unit frequently refines the provision ofexisting products and servicesOur business unit regularly implements small .574 adaptations to existing products and services
.711
90. Our business unit improves the efficiency of products
and services
.829
91. Our business unit increases economies of scales in
existing markets
.730
92. Our business unit expands services for existing clients
Eigenvalues 4.70 3.00 2.64 2.33 2.17 1.82
Table 3.3 Summary of the joint factor analysis of all latent variables (factor loadings < .50 suppressed)
Component
Construct Item 1 2 3 4 5 6 7 8 9 10 11
Decentralization 20. Strategic decisions .716
21. Investment decisions .775
22. Marketing decisions .847
23. Decisions regarding internal processes -.582
24. Human resources decisions .754
Manager
innovativeness 55.56. I often surprise people with my novel ideasPeople often ask me for help in creative activities .592.601
57. I obtain more satisfaction from mastering a skill than coming
up with a new idea
.643
58. I prefer work that requires original thinking .637 59. I usually continue doing a new job in exactly the way it was
taught to me
60. I like a job which demands skill and practice rather than
inventiveness
.702
61. I like to experiment with various ways of doing the same
thing
.533
62. I am not a very creative person .696
Manager
optimism 63.64. In uncertain times, I usually expect the bestIf something can go wrong for me, it will .542 .723
65. I’m always optimistic about my future .804
66. I hardly ever expect things to go my way .820
67. I rarely count on good things happening to me .758
68. Overall, I expect more good things to happen to me than bad .723
Tight budget
control 79.80. Puts much emphasis on meeting the budgetDoes not easily accept budget revisions during the year .874
81. Has a detailed interest in specific budget line items 82. Does not lightly tolerate deviations from interim budget
targets .886
83. Is intensively engaged in budget-related communications .683
Explorative
innovation 84.85. Our business unit accepts demands that go beyond existingproducts and servicesOur business unit invents new products and services .500.819
86. Our business unit commercializes products and services that
are completely new to them .853
87. Our business unit frequently utilizes new opportunities in
new markets .800
Exploitative
innovation 88.89. Our business unit frequently refines the provision of existingproducts and servicesOur business unit regularly implements small adaptations to .821
existing products and services .704
90. Our business unit improves the efficiency of products and
services .756
91. Our business unit increases economies of scales in existing
markets .757
92. Our business unit expands services for existing clients .732
Eigenvalues 4.70 3.00 2.64 2.33 2.17 1.82 1.65 1.44 1.25 1.13 1.12
3.2.1 The construct decentralization
For measuring decentralization, the construct for measuring decentralization designed by Abernethy et al. (2004) is used. The authors have designed this construct, based on two other studies. One additional question regarding decision rights with respect to internal processes was added to this proven construct.
When conducting a PCA on this construct, the first three items of this construct: (20) strategic, (21) investment, and (22) marketing, loaded on one component, the last two: (23) internal processes, and (24) HR, on another component. When conducting a joint PCA with a maximum of six factors, item 24 has a score <.50. Based on the joint PCA with no maximum, item 20-22 load on a single component, internal processes (23) and HR (24) load on various components. The fact that the items load on various components, seems understandable when analyzing the items. The first three items have an external focus, where the last two have an internal focus. Based on the findings, internal processes and HR were deleted. By deleting these item, the items left all loaded on one component only. Cronbach’s Alpha (.77) of this construct excluding item 23 and 24 is above the threshold (.50) and the desired minimum value (.70).
3.2.2 The construct tight budget control
For measuring the concept of budget control, the construct for measuring budget control tightness by Van der Stede (2001) is used. Van der Stede was able to construct a measurement instrument for tight budget control, based on previous literature (2001). The construct consists of five items. Van der Stede (2001) notes that of these five items, detailed interest in specific budget line items (81), and engaged in budget-related communications (83) can be considered interactive controls. The other items can be considered diagnostic. When running a PCA on this construct, all items loaded on one component. When conducting a joint PCA with a maximum of six factors, all items loaded on the same component. However, when testing a joint PCA without a maximum, interest (81) did not load, and engagement (83) loaded on another component. Thus, the distinction recognized by Van der Stede did affect the discriminant validity. And as discussed in the theory section, the same distinction was made by Bart (1991). Therefor interest (81) and engagement (83) were removed from the construct. Also, because in this study I am not studying interactive use. The item: puts much emphasis on meeting the budget (79) did not have significant loadings on the tight budget control component. I have tested the reliability with and without item 79. With, Cronbach’s alpha was .72. Without, it improved to a very high score (.82). The fact that item 79 did not load in the joint PCA without a maximum factors, and the improvement of Cronbach’s alpha made me exclude item 79 as well.
3.2.3 The construct Innovation
For measuring the concept of innovation, the construct for measuring innovation by Jansen et al. (2006) is used. The authors have examined existing literature to develop measures for exploratory and exploitative innovation (2006). Exploratory innovation pertains to development of new products and services, exploitative innovation means improving products and services (Jansen et al., 2006) This distinction in innovation is recognized in their study. Items 84-87 pertain to explorative innovation, items 88-92 pertain to exploitative innovation. I have conducted a PCA on the entire construct of Jansen et al. (2006) with a maximum of two factors and the outcome indeed confirmed the distinction in innovation. Based on this, I have split the construct in two constructs and tested explorative innovation and exploitative innovation separately.
3.2.3.1 Explorative innovation
In the PCA on the construct, convergent validity, all items load on one component. The same counts for the discriminant validity, with or without a maximum of six factorss. So no item is questionable. Cronbach’s Alpha (.78) is very good.
3.2.4 Control variables
In this study exploitative innovation, manager innovativeness, manager optimism, eductional level and Log10 number of fte are used to control for confounding effects of these control variables on the weight of decentralization and innovation. I have tested the constructs used as control variables in the same way as I have done with the main variables.
3.2.4.1 Exploitative innovation
In the PCA for the construct, the items: refines provisions of existing products and services (88), and small adaptions to existing products (89) load on another component than items: improves efficiency of products and services (90), increases economies of scale in existing markets (91), and expands services for existing clients (92). When having a closer look on the items in this construct, this distinction seems logical. The first two items are pertaining to exploitative product innovation. The last three items, are measuring efficiency and scale. Since, according to table 3.3, none of the items load on a component which also loads items from other constructs, all questions can be uphold. The fact that item 92 was not loaded when conducting a joint PCA with a maximum, is ignored. However, a reliability test for both components is obligatory when using multiple components. The items 88 and 89 have a Cronbach’s alpha of .68. The other items have a Cronbach’s alpha of ,71. Although the score of .68 is below the desired minimum, it is above the threshold. Therefore all items will be used for measuring exploitative innovation.
3.2.4.2 Manager innovativeness
The construct for manager innovativeness was derived from a study of Mueller and Thomas (2001). This construct consists of 8 items. Four of these items have been recoded into new variables because the question was negatively worded (items 57, 59, 60, 62). In the construct PCA, all items load on one factor. Upon conducting a joint PCA with a maximum of six factors, the item: continue doing a job in the way it was taught to me (59) did not load. In the joint PCA with no maximum, the item: I like a job which demands skill and practice, rather than inventiveness (61) loaded on another component, together with an already deleted item from tight budget control. These items differ from the other items in the sense that they speak of varying with existing tasks. The other items are about new ideas. Items 59 and 61 were deleted. Cronbach’s alpha (.80) showed a high reliability on the remaining six items.
3.2.4.3 Manager optimism
Manager optimism was measured using a construct from a study of Graham, Harvey, and Rajgopal (2005). This construct consist of six items. Three of these items have been recoded into new variables because of the negative wording. Items 64, 66 and 67 were recoded. When testing a PCA on the construct, two factors were discovered. The former negative items loaded on one component, the positive items on another. When testing joint PCA with with a maximum of six, items 63 and 65 did not load. The other items loaded on the same component. In the last test, the joint PCA without a maximum, the recoded items happen to load on the biggest component for manager optimism. The items 65 and 68 load on another component but without an overlap with another item from a different construct, and item 63 loads on an entirely different component, together with an item from decentralization construct. The latter was therefore deleted from this construct. The biggest factor scored a Cronbach’s alpha of .58. The smaller scored .69 on the reliability test. Although Cronbach’s alpha was lower than the desired score, both scores were above the threshold, and therefore were held for representing the constructmanager optimism.
3.2.4.4 Education
The respondents were asked for their highest level of education. This was a nominal question. It was made ordinal by coding education as follows: Lower/ intermediate=1, higher=2, academic=3 and phd=4.
3.2.4.5 Number of fte
On the number of fte a log10 transformation was conducted to eliminate the positive skewness and to be able to analyse the correlations between the number of fte and the other variables.
3.3 Statistical analyses
Based on the previous paragraph, the variables which measured each construct at its best were computed. The first step in the analysis is the descriptives of the various variables. This is set out in table 3.4. Also the correlation between the variables is computed. These results are set out in table 3.5.
Table 3.4 descriptive statistics
N Min Max Mean SD
Decentralization 111 1 5 3.07 0.97
Tight budget control 110 1 5 2.88 0.95
Explorative innovation 111 1 5 3.28 0.81 Exploitative innovation 111 2.20 5 3.48 0.60 Manager innovativeness 111 1.83 4.33 3.29 0.41 Manager optimism 111 2.40 5 4.12 0.48 Education 110 1 4 2.50 0.66 Number of Fte 110 4 450 72 83.84
Log10 number of Fte 110 0.60 2.65 1.64 0.42
Table 3.5 correlation table
1 2 3 4 5 6 7 8 1. Decentralization 1 -.11 .27*** .14 .05 .06 -.01 -.19* 2. Tight budget control -.10 1 -.23** -.10 -.09 .03 -.08 .05 3. Explorative innovation .28*** -.26*** 1 .33*** .23** .03 .15 -.14 4. Exploitative innovation .13 -.12 .31*** 1 .06 .11 -.01 -.08 5. Manager innovativeness .06 -.13 .28*** .05 1 .056 .34*** .05 6. Manager optimism .09 -.02 .07 .17* .14 1 .09 .01 7. Education -.01 -.09 .12 .01 .30*** .11 1 .28*** 8. Log10 number of Fte -.14 .05 -.16 -.11 .04 -.05 .24** 1
Pearson correlations appear below the diagonal, Spearmans correlations appear above the diagonal. *correlation is significant at the 0.1 level (two-tailed)
**correlation is significant at the 0.05 level (two-tailed) ***correlation is significant at the 0.01 level (two-tailed)
Multicollinearity exists when the correlation table shows values of .80 and higher between the independent variables. This is not the case as can be concluded from the table 3.4. However, this approach is considered quick and dirty (Field, 2013). In chapter 4, multicollinearity will be tested with VIF and tolerance.
3.3.1 Equations
For testing the hypothesis of the suppression, a model should be used which can make the indirect effect visible. The model which will be used in this study is derived from Ludlow & Klein (2014). It is based on the standard regression equation:
Innovation (Y’) = (bY1) decentralization + control variables + error
A second predictor, tight budget control, will be introduced in the model. The new regression equation will then then be:
Innovation (Y’) = a + (bY1.2) decentralization + (bY2.1) tight budget controls + control variables
+ error
By adding the second predictor, the regression coefficient and the standard error of decentralization will change.
By the definition presented earlier, if “the strength of the relationship between the predictors and the outcome is reduced by adding the [second predictor]”(Field, 2013, p. 408), or in our terms, |bY1.2| < |bY1|, then X2is (or, conversely, acts as) a mediator.
“[When] the original relationship between two variables increases in magnitude when a third variable is adjusted for in an analysis” MachKinnon 2008, p. 7) or in our terms |bY1.2|
> |bY1|, then X2is (or, conversely, acts as) a suppressor variable (Ludlow & Klein, 2014, p.
8).
The adding of the second predictor should increase the coefficient to prove tight budget controls being a suppressing variable. After the change in coefficients is calculated, significance of this change should be determined. For determining the statistical significance, the Sobel test will be used. “Since the third variable estimation procedure is the same regardless of what the second predictor is called, the well-established tests of Sobel (1982) . . . for the statistical signifance of a
mediator’s effect also hold for testing a suppressor’s effect (Mackinnon et al. 2000, p. 176)” (Ludlow & Klein, 2014, p. 8).
4 Results
Table 4.1 presents the results of the hierarchical regression analysis for tight budget controls, and exploratory innovation. Models 1 and 2 pertain to the tight budget controls. Models 3, 4, 5, and 6 pertain to explorative innovation. The models 1 and 3 are the baseline models. These models contain the control variables exploitative innovation, manager innovativeness, manager optimism, education and number of fte.
Table 4.1: Results of Hierarchical Regression Analyses: Standardized effects on Explorative innovation
Tight budget controls
Explorative innovation Expectation
(model 1,2)
Model 1 Model 2 Expectation (model 3-6)
Model 3 Model 4 Model 5 Model 6 Exploitative innovation NP -.07 -.06 NP .26*** .25*** .24** .23** Manager innovativeness NP -.12 -.12 NP .19* .17* .18* .16* Manager optimism NP -.03 -.02 NP -.01 -.01 -.02 -.02 Education NP -.07 -.07 NP .08 .06 .08 .07 Log10 number of fte NP .07 .06 NP -.17* -.16* -.14 -.13 Independent variable: Decentralization + -.11 + .22** .21** Suppressor: Tight budget control - -.18** -.16*
Model fit indices
R² .03 .05 .15 .18 .20 .22
R² change .01 .03 .05 .0
F change .72 1.21 3.59*** 4,01** 5,99** 3,19*
* significant at the 0.1 level (two-tailed) ** significant at the 0.05 level (two-tailed) *** significant at the 0.01 level (two-tailed)
In the tight budget controls model 2, the independent variable decentralization is introduced. The coefficient of decentralization is not significant. Hypothesis 2, decentralisation will lead to more tight budget control, is not supported. I have run an additional robustness test. I have performed a bootstrap with 1.000 bootstrap samples on the dependent variable budget control tightness systems and independent variable decentralization, with control variables included Decentralization reported a BCa CI [-.291,.084]. This confirmed the finding with respect to hypothesis 2. Since hypothesis 2 is not supported, the model cannot reject the overall null hypothesis. That is the hypothesis of the indirect effect (hypothesis 4). For hypothesis 4 to be supported, the effect of decentralization on tight budget controls should have been significant and positive. This is because an indirect effect can only be measured if a mediator variable (or suppressor as in this study) is affected by the independent variable (Baron & Kenny, 1986). This is not the case since hypothesis 2 is not supported.
For hypothesis 1, more decentralization will lead to more exploratory innovation, model 5 gives an explanation. The coefficient is .222 with a p<.05. Here bootstrap test gives a BCa CI .033,.316]. As expected, I found support for hypothesis 1. For hypothesis 3, more tight budget controls will lead to less exploratory innovation, the coefficient is -.183 with a p<.05. Here bootstrap test gives a BCa CI -.301,-.001]. This hypothesis is supported as well. This is also as I expected.
At the beginning of this paragraph, I noted that hypothesis 2 could not be supported and that therefore hypothesis 4: use of formal tight budget controls will suppress the effect of decentralization on exploratory innovation, would also not be supported. For a proper test of hypothesis 4, the unstandardized regression coefficients should be used. In table 4.2 the unstandardized regression coefficients have been denoted. I have entered the unstandardized regression coefficients in the equations.
Standard regression equation:
(explorative) innovation = 0.866 + .18*decentralization (+ control variables + error)6
Adjusted regression equation:
(explorative) innovation = 1.475 + .16*decentralization -.13*tight budget control (+ control variables + error)
The unstandardized regression coefficient for decentralization is decreasing instead of increasing. The null hypothesis with respect to hypothesis 4 could therefore not be rejected.
Table 4.2: Results of Hierarchical Regression Analyses: Unstandardized effects on Exploratory Innovation
Model 1 Model 2 Model 3
Constant 1.291 .866 1.475
Exploitative innovation .33*** .30** .29**
Manager innovativeness .37* .36* .32*
Manager optimism -.02 -.03 -.03
Education .09 .09 .08
Log10 number of fte -.31* -.25 -.24
Independent variable: Decentralization .18** .16**
Suppressor: Tight budget control -.13*
* significant at the 0.1 level (two-tailed) ** significant at the 0.05 level (two-tailed) *** significant at the 0.01 level (two-tailed)
4.1.1 The Sobel test
The results of this study indicate no indirect- suppressor- effect by comparing the unstandardized regression coefficient before and after entering the suppressor variable in the model. I have tested the indirect effect of tight budget control on decentralization with respect to explorative innovation. The test was performed using Hayes Process tool in SPSS (Hayes, 2013). This is a Sobel test., which as Ludlow and Klein indicated could be used to calculate an indirect effect. As the results from the Sobel test confirm, the indirect effect is non-significant.
Table 4.3 Normal theory test for indirect effect (Sobel test)
Explorative innovation Decentralization Size of indirect effect
Standard error Z-score P-value .0138 .0163 .8441 .3986 Standardized effect .0187
The total sample size for measuring the indirect effect was N=108 and the df was 100. It should be noted that Field (2013) warns that Sobel can only be used in large samples. Sobel (1982) himself recognizes this but does not give guide to when a sample is large enough. Sobel (1982) only mentions that “these methods may be inappropriate in particularly small sample” (p. 308). As the results of the Sobel test indicated, the indirect effect was not present. This was the same finding when comparing the regression coefficients. But, in addition to the Sobel test, to avoid performing a wrong test, I have performed a bootstrap with 1.000 bootstrap samples on dependent variable explorative innovation with supressor tight budget control and independent variable decentralization. The indirect effect showed a BCa CI[-,0055,.0588]. Thus, this supports the conclusion; there is no indirect effect.
4.1.2 Additional testing
Also the model was tested on multicollinearity. All variables in the models 1 to 6 which were presented in table 4.1, had a VIF of just above 1 and a tolerance of just below 1. Since multicolliniearity is expected at a VIF>10) or tolerance <.2 (Field, 2013), multicolliniearity does not exist in the models used.
As discussed in the theory section, exploitative innovation is more positive in a more controlled environment (Benner & Tushman, 2002). Empirical evidence that decentralization is leading to less exploitative innovation, is to my knowledge not yet provided. Based on the correlation table 3.3, exploitative innovation is not significantly correlated with decentralization. When performing a bootstrap with 1.000 bootstrap samples on DV exploitative innovation with IV decentralization, the latter showed a BCa CI [-.049,.212]. Thus, confirming there is no effect.
5 Discussion and conclusion
As Friedmann and Wall (2005) concluded: “Correlations of independent variables with the criterion are important: correlations of independent variables with each other may flesh out the predictive network of the regression model” (p. 135). This observation was important for this study. I wanted to strengthen the regression model of decentralization with explorative innovation; to prove that the relationship of decentralization with explorative innovation would be magnified if we would include another variable, tight budget controls. This effect is known as a suppressor effect. This suppressor effect requires basically the same hypothesis as a mediator. first hypothesis is the direct relation, the independent variable and the dependent variable. In my study, this was: more decentralization will lead to more explorative innovation. The second hypothesis is the relation of the independent variable, with the suppressor, tight budget controls. The third relation would be a relation with the suppressor and the independent variable. Overall it was expected that including a suppressor would lead to a significant higher regression coefficient for the independent variable in relation to the dependent variable. I found support for hypothesis one and three. Unfortunately, hypothesis two was not supported. Based on this finding, already could be concluded that tight budget control would not act as a suppressor for decentralization. A relation between the independent variable and the suppressor was mandatory to prove the that tight budget control could act as a suppressor on decentralization (Baron & Kenny, 1986). I did run a test on hypothesis four, but as expected I found no significant indirect effect.
5.1 Theoretical implications
To my knowledge, suppressor effect has not been investigated with respect to management control systems. Also combining agency theory and self-determination theory, is not common practice in management control studies. These are the general theoretical implications for this study. I show it is possible to innovate in management control research. It is possible to combine mainstream and interpretive theory. And it is possible to use seldom used statistical concepts. The latter is not that innovative when we look at when it originated. Menderhausen (1939) already mentioned it 76 years ago.
The findings do not imply, diagnostic tight budget control being a suppressor. The assumption that tight budget control would be acum hoc ergo propter hoc fallacy was incorrect. The fact that the tight budget control was highly correlated with explorative innovation, and had no relation with decentralization, gives support for the general theory that use of formal management control systems do limit explorative innovation. This study does in that aspect contribute to earlier
findings (Abernethy & Stoelwinder, 1991; Miles, Snow, Meyer & Coleman, 1978). These studies did not investigate the relation of formal management control systems with innovation in a quantitative setting. Nor has prior research focused on diagnostic tight budget control in isolation. Since the research question was aimed at studying multiple relations, other hypotheses have been tested as well. As expected, I did find support for the effect of decentralization on explorative innovation. A more decentralized setting has a positive effect on explorative innovation. This is to my opinion a valuable contribution to the existing literature. So far, as to my knowledge, only the effect of centralization on explorative innovation has been thoroughly investigated. Centralization, or increased process management does lead to less exploratory innovation (Benner & Tushmann, 2003; Jansen et al., 2006). Until now, only intuitively, decentralization was expected to have a positive effect on decentralization. This study gives empirical evidence for this intuitive feeling.
The finding with respect to the relation of decentralization on tight budget control, being non-significant, is unexpected. I was expecting to find an increase in tight budget control because, if a company is highly decentralized, diagnostic surveillance is to my expectation one of the few possibilities to exercise some form of observation. In contrary to the finding of Abernethy et al. (2004): use of divisional summary performance measures increases in decentralized firms, in this study, the use of tight budget control does not increase. This is a theoretical implication. When investigating the use of formal management control systems, the researcher should test for isolated types of formal management control systems. Management control systems are not that straightforward. Like Bisbe and Otley (2004) concluded with respect to another type of management control systems, the interactive:
One plausible reason why the null hypothesis of no correlation between interactive use of formal MCS and product innovation cannot be rejected is that the relationship between interactivity of control systems and product innovation is more complex than expected according to the initial theoretical development (e.g. segmented or non-linear as opposed to initial simple linear models (p. 726).
5.2 Limitations
There are several limitations in this research. The first is simplicity of the model used. I have looked at only one possible component of the management control systems. I have used tight budget control because I had to make a choice in that sense, based on available data and theory. This way I have only covered a part of the management control system market. However, by isolating tight budget control from all the possibilities for management control systems, and using a proven
