THE IMPACT OF MANAGEMENT CONTROL SYSTEMS ON EMPLOYEE MOTIVATION AND JOB PERFORMANCE

THE IMPACT OF MANAGEMENT CONTROL SYSTEMS

ON EMPLOYEE MOTIVATION AND JOB

PERFORMANCE

Management Control Systems: mechanisms towards autonomous

motivation

By:

ANGELA GARCIA GOMEZ

S3900495

A.M.Garcia.Gomez@student.rug.nl

MCs Business Administration

Management Accounting & Control

Supervisors: Paula van Veen-Dirks

Wilmar de Munnik

1

THE IMPACT OF MANAGEMENT CONTROL SYSTEMS ON

EMPLOYEE MOTIVATION AND JOB PERFORMANCE

ABSTRACT

2

1. Introduction

1.1 Problem statement

How do employees feel proactive, and engaged to perform well? Does a specific way to boost the inner motivation of employees, while keeping an eye on the firm strategy exist? It is highly relevant to analyse the behavioural factor that moves human beings (Ouchi, 1979; Abernethy and Chua, 1996; Chenhall, 2003), as Merchant and van der Stede (2007) stated, “it is people who make things happen”. To gain organizational benefits and motivated patterns of behaviour, it is required to consider how to manage and apply the firm resources (Chenhall, 2003; Otley, 1999). From a socio-logical perspective, Management Control Systems (MCS) are identified as active tools, directing individuals to accomplish their goals aligned with the firm strategy (Chenhall, 2003). Van den Broeck, et al. (2010) argue that the degree to which needs are satisfied depends on the work context, for instance experiencing internal or external contingencies such as bonus or guilt, or social support to comply a task. Considering that MCS are part of this work context, these mechanisms will influence the fulfilment of needs, developing a specific type of motivation with a level of job performance. Autonomous motivation tends to be associated with positive outcomes and well-being (e.g. academic performance among students) (Grolnick, Ryan, and Deci, 1991). However, different perceptions can arise with the intervention of external mechanisms such as MCS (Osterloh and Frey, 2000). MCS can function as an informative guidance to employees, improving their capabilities and self-development while making their own decisions (i.e. increasing autonomous motivation). Nevertheless, the mere monitoring and correction of these mechanisms leads to a perception of MCS as a constraining force becoming a threat for the employee (i.e. decreasing autonomous motivation) (Osterloh and Frey, 2000). There is a common understanding that certain forms of control (e.g. formal controls) can threaten the autonomous motivation of individuals reducing the likelihood of successfully performing tasks, mostly related to creativity and innovation (Davila, Foster, and Oyon, 2009; Grabner 2014; Speklé, van Elten and Widener, 2017; Gagné and Deci, 2005).

3

1.2 Purpose and Research Question

The aim of this study is to delve deeper into the challenge for organizations to foster employee autonomous motivation. Different forms of MCS are used to assess the impact of job performance. To that end, firstly we draw on the Levers of Controls of Simons’ Framework (1995), to examine diagnostic and interactive use of MCS, and the balance between them, also called dynamic tension. Secondly, the Self-Determination Theory developed by Edward L. Deci and Richard M. Ryan, is employed to study autonomous motivation. The research is developed in an academic context among employees from universities of applied sciences in the Netherlands. Gagné et al. (2010) claim that health/educational jobs increasingly develop more identified and intrinsic motivation (i.e. autonomous motivation) than technical/manual jobs, sales/service jobs, and managerial/professional jobs. Therefore, studying this segment of the population will help to assess the relationship between MCS, autonomous motivation and job performance.

The formulated research question is the following:

RQ1: How do diagnostic use of MCS, interactive use of MCS, and the combination of both uses

as dynamic tension affect autonomous motivation and consequently, job performance of employees?

1.3 Significance of the research

This study generates various insights that help to contribute to the managerial accounting and the psychological field. First of all, this research gives great importance to the effects of the product of diagnostic and interactive use of MCS, called dynamic tension. In the last few years, several studies have critically reviewed the notion of balance in the combination of the uses of MCS (Mundy, 2010; Kruis, Speklé and Widener, 2016). Others such as Widener (2007) have focused on the complementarity and the interdependencies between the different Levers of Controls, but not the dynamic tension itself. However, Henri (2006) gives some insights about the effects of dynamic tension, although from a resource-based perspective.

Secondly, this study will contribute to the Self-Determination Theory (SDT) with knowledge related to one additional factor included in the work environment (Van den Broeck, et al., 2010; Gagné and Deci, 2005), which is the Management Control System. Furthermore, Gagné and Deci (2005) claim that organizations achieve better outcomes via autonomous motivation than via controlled motivation, although literature falls short in empirical evidence. Additionally, this study will empirically examine the uses of MCS and their consequences for autonomous motivation in a context were tasks are characterized as heuristic, complex and based on quality results (Deci, Olafsen, and Ryan, 2017) like university employees.

4 favourable outcomes in terms of job performance. In contrast to prior research (Van der Kolk et al., 2018; Pfister and Lukka, 2019), this study focuses on the use of MCS based on Simons’ Framework (1995) instead of the design of them, relied on the management control framework of Merchant and van der Stede (2007). Ferreira (2002) argues the use of control systems is more relevant than the design of control systems from a managerial perspective. Ferreira and Otley (2009) indicate the use of Performance Measurement Systems (i.e. MCS) as “cornerstone”. Besides, it gives greater significance to the combination of two contrasting roles of MCS and its impact on motivation and performance, since empirical analysis of this dynamic tension is still quite scarce (Henri, 2006).

1.4 Remainder of the paper

In the first section, relevant literature to this research is reviewed, which leads to the development of hypotheses and the conceptual model. Subsequently, the methodology section includes the sample, procedures, and measures. The results section presents the findings and analysis of the study. Finally, this paper ends with a discussion and a conclusion.

2. Literature review

2.1. Management Control Systems

In 1965, management control was described as a process utilized by managers to achieve organizational targets, obtaining and applying resources in an effective and efficient way (Anthony, 1965). Anthony (1965) examines management control in terms of influencing employees to do certain activities and preventing others. Furthermore, Merchant and van der Stede (2007) claim that management control ensures that employee’s decisions and behaviours are aligned with the firm strategy. All these definitions have one common factor, the individual behaviour. Managers receive the information through management control to move towards the organization’s best interests. Depending on the design and application of the mechanisms that management control offers, managers can assess and communicate the information in order to make an impact on employees’ behaviour and consequently on their performance.

5 Furthermore, MCS are considered as the drivers that boost autonomy, creativity, adaptability and innovation (Simons, 1995; Adler and Borys, 1996; Davila, et al., 2009). However, the question in this context is how MCS are used to reach these outcomes. To illustrate this issue, the study focuses on the way of use of MCS (e.g. diagnostic use) rather than the type of MCS (e.g. results control). For example, Osterloh and Frey (2000) argue that the external intervention of MCS implicitly relies on two components: a controlling effect that the individual perceives as constraining from the outside; and an informing effect that boost the competence feelings and autonomy to make their own decisions. Therefore, if the characteristics of the use of MCS can enhance the personal motivation of employees, it is likely that their job performance improves as well.

2.1.1. Levers of Control: Simons’ Framework

The Levers of Control of Simons (1995) are remarkably valuable because of its comprehensive view that allows the exploration of the variety of controls, and their managerial practice (Kruis et al., 2016; Widener, 2007). Simons’ Framework outlines the uses of MCS performed by managers (Simons, 1995). He proposes two separate uses of MCS: the diagnostic use and the interactive use.

The diagnostic use of MCS represents a traditional feedback instrument, which is restricted to monitoring and rewarding the pre-established performance targets (Simons, 1995). This feedback tool provides the possibility to adjust and refine inputs, thus the outputs will result closer to the organizational standards. The most common practices are budgets, profit plans, goals setting systems, business plans, and standard cost accounting systems (Simons, 1995; p. 61). In addition, one aim of the diagnostic use is reducing the manager’s burden of permanent monitoring. Simons (1995) argues that once the performance measurements are set to the employees, managers can work in other issues, assuming that the employees will adequately perform to reach the goals. However, this can lead to potential control failures in the attainment of the firm strategy (Simons, 1995).

The interactive use of MCS is defined as the formal information system used by managers to improve the regular and personal involvement with the rest of employees to take part in their decision-making. This use of MCS tends to look ahead and seek opportunities that strategically place the organization in the market (Widener, 2007). This use of MCS is characterised by four aspects: (i) continue and recurrent information exchange, (ii) frequent attention in all the structure of the organization; (iii) analysis and discussion in personal meetings with all the employees; and (iv) continuous argumentation about the possible problematic data, presumptions and action plans (Simons, 1995).

6 combination of positive and negative forces generated by the uses of MCS creates a dynamic tension that helps to manage the so-called inherent organizational tension. It needs to be in balance in order to obtain an effective control of the firm strategy (Henri, 2006, Kruis et al., 2016). The diagnostic use represents the negative force because it is focused on correcting deviations and on the enforcement of performance measurements. (Simons, 1995; Abernethy and Brownell, 1997; Henri, 2006). In addition, this use can bring dysfunctional behaviours (e.g. illegal acts or gaming) because of the large and complicated channels of communication and the limited information flow (Simons, 1995; Henri, 2006). Alternatively, the interactive use provides positive energy because it continuously encourages the learning development and the opportunity-seeking (Simons, 1995; Henri, 2006). As Ferreira and Otley (2009) argue, the diagnostic use takes a mechanistic and repressive approach, while the interactive use embraces an organic and productive approach.

Both diagnostic and interactive use of MCS show interdependency and complementary, which means that the utilization of one control system emphasizes the benefits received from the other control system, and vice versa (Widener, 2007). We can associate the complementarity argument with the perspective of Argyris and Schön (1978) in organizational learning (Henri, 2006). The diagnostic use could be the single-loop learning and it could be the basic condition for the interactive use to conduct the potential double-loop process. Single-loop learning might imply just the information exposition of the outcomes resulted from the initial goals, and the deviation is associated with a “deficient operationalisation”. Hence, its characteristics resemble diagnostic use of MCS. By contrast, the double-loop learning entails a continuous examination of the pre-set strategies. Hence, its characteristics resemble interactive use of MCS (Argyris and Schön, 1978; Usher and Bryant, 1989).

In line with this statement, Henri (2006) finds a positive relationship between dynamic tension and organizational learning –supporting the idea of Argyrin and Schön (1978) – in high uncertain environments, and an even stronger positive relationship in contexts with flexible values as organizational culture. Furthermore, there is empirical evidence that this dynamic tension contributes positively to the organizational performance under high environmental uncertainty (Henri, 2006). Therefore, the key is to find the right balance between uses of MCS for the adequate dynamic tension, thus it would lead to obtain the desired employees’ behaviour without diminishing the creativity and innovation to success in business (Kruis et al., 2016).

2.2. Autonomous motivation

7

2.2.1. Self-Determination Theory

Motivational theories support organizations to understand the root of certain patterns of behaviour. They facilitate the understanding of which contingencies directly affect employee motivation within the work environment. For this study, we draw on the SDT since it offers a valuable framework for managers and organizations (Gagné, et al, 2015). The self-determination of the employees rests on the degree to which employees perceive internal or external locus of causality (PLOC) (Deci, 1971). This PLOC represents “the degree to which an action is initiated and endorsed from this phenomenal centre that describes the relative autonomy of an act” (Ryan and Connell, 1989). The internal PLOC occurs when the behaviour is self-emanated and volitional (i.e. the external motivators that drive the employee to act are internalised). The external PLOC is when the behaviour of the employee is caused by control or pressure from others, thus the external motivation is not internalised (Ryan and Connell, 1989; Ryan and Deci, 2000).

Furthermore, the SDT posits a controlled-to-autonomous continuum, which shows the degree to which the external regulation is internalised by the employee, i.e. the PLOC (Gagné and Deci, 2005). At the far left of the Figure 1 is amotivation, which refers to the lack of motives that moves employees to act. This situation appears when individuals place no value on tasks (Ryan, 1995), when they do not feel capable to perform that task (Bandura, 1986), or even when there are no expectations about the potential results of doing that task (Seligman, 1975).

Figure 1: Self-Determination Continuum based on Gagné and Deci (2005)

Subsequently, we find types of motivation generated with external intervention, generally called as extrinsic motivation. They are exposed from less internalised extrinsic motivation to

Amotivation Intrinsic Motivation Extrinsic Motivation External Regulation Identified Regulation Introjected Regulation Integrated Regulation

-

+

8 more internalised extrinsic motivation. The first one is external regulation which is typically compared with intrinsic motivation, as being completely the opposite in terms of motivation. The employees with external regulation does not perform their tasks unless the supervisor is looking at them (Ryan and Deci, 2000). The kind of external regulation that triggers the individual to work might be an authority, a punishment or rule compliance. The next form in this continuum is introjected regulation. This motivation is caused by an external control, although in this case the employee takes into account his or her own personal feelings and values to perform in certain way. Thus, the individual behaviour depends on the feelings of pressure, guilt or anxiety, or even on satisfying others’ expectations (Gagné and Deci, 2005). The employee performs a task because that makes that person feel worthier (Ryan and Deci, 2000). These last two motivations (external regulation and introjected regulation) are associated to controlled motivation.

In the right side of this continuum, we can find the autonomous motivation, which is the variable under study. Therefore, the next type of motivation is identified regulation. This motivation occurs when the individuals are aware of the importance of their behaviour (Ryan and Deci, 2000). In this way, they show more willingness to act because they have reached certain level of congruency between their personal goals and organizational goals (Gagné and Deci, 2005). For example, employees that have to deliver their products on time —which action could be more or less pleasant— recognize their behaviour aligned with their inherent values. Furthermore, the most autonomous type of extrinsic motivation of this continuum is the integrated motivation. In these cases, employees fully integrate the identified motivation into their own values and needs. These behaviours emanate as a part of the self, although it is still triggered by an external factor. For that reason, the integrated motivation coexists closely with the other type of autonomous motivation, the intrinsic motivation (Ryan and Deci, 2000; Gagné and Deci, 2005). Finally, in the right extreme of the continuum we find the last type of autonomous motivation: intrinsic motivation. This form of motivation simply occurs when the individual behaves because the task is interesting for that person (Gagné and Deci, 2005). Ryan and Deci (2000) describe this construct as “the principal source of enjoyment and vitality throughout life”.

In short, the identified and integrated motivation occur when employees recognize the relevance of the behaviour and they value the potential outcomes as their own goals. Intrinsic motivation has a natural assimilation and interest from the activity itself that is born in the personality of the individual (Gagné and Deci, 2005).

2.2.2. Psychological Basic Needs

9 development of the employees (Ryan, et al., 1996). These needs are the need for competence, autonomy and relatedness (Ryan and Deci, 2000).

The need for autonomy is conceptualized as the individuals’ innate feelings of psychological freedom to make their own choices, and to have opportunities for self-determination to perform tasks (Ryan and Deci, 2000). It is about experiencing complete conscious volition; even though you are supervised and have some influence from others. Because of that, autonomy has been the most controversial need in organizational psychology (Van den Broeck et al., 2016). According to SDT, the need for autonomy can be generated with the dependency of others or if behaving in some way results in the compliance of others’ demands. By contrast, Hackman and Oldham (1976) claimed that the term of autonomy represents “substantial freedom, independence and discretion to the individual in scheduling the work and in determining the procedures to be used in carrying it out” (Van den Broeck, et al., 2010, p. 258).

The need for competence is defined according to SDT as the need to feel effective in performing tasks and interacting with the environment. Furthermore, it is most frequently generated when the environment is examined by the individual concern of the employee, thus this leads to a need to master the tasks and further develop one’s skills (Gagné and Deci, 2005; Van den Broeck, et al, 2016). This conceptualization of competence is supported by White (1959) who outlined competence as the accumulation of effectance motivation derived from the satisfaction of producing effects interacting with the environment (Ryan and Deci, 2017). This author takes competence as innate and associates it with intrinsic satisfaction. This position contrasts with Vroom’s (1964) Expectancy-Value Theory and Bandura’s (1986) Self-Efficacy Theory, whose competence is associated with the extrinsic satisfaction obtained by those outcome expectancies and self-efficacy that the performance of a task might lead to. Despite this difference, self-efficacy is highly correlated with competence, even considered the elementary motivational principle (Gagné and Deci, 2005; Van den Broeck et al., 2016; Ryan and Deci, 2017).

The need for relatedness is the degree to which individuals have the inner tendency to feel connected to other people or their own environment, to feel cared by others or feel caring for others, thus they develop a sense of belonging to colleagues, organizations, and environments (Ryan and Deci, 2000). This need is satisfied when the individual has a feeling of intimacy, building rapport and developing close relationships with others (Van den Broeck et al., 2010). Sometimes, people tend to behave according to others’ expectancies to create a sense of connection between them. However, the need for relatedness is not fulfilled until the individual has internalized the behaviour as one innate or emanated from one’s self.

10 offering honest, non-judgmental and positive feedback; boosting the self-development and knowledge sharing; and reducing the enforcement of measures and controls (Stone, et al., 2009).

2.3. MCS and Motivational Hypotheses

2.3.1. Relationship between Diagnostic Control Systems and Autonomous Motivation

The diagnostic use of MCS focuses on monitoring and rewarding or penalizing the pre-set goals, depending on the performance results. According to Simons’ (1995) perspective, only top managers or supervisors are in charge of this feedback instrument, and they alone can make the pertinent decisions to direct employees’ behaviour. The rest of employees are restricted to compliance with the achievement of those standards established by managers. Thus, there is an autonomy limitation for the rest of employees since there is barely any sense of volition to choose between alternatives according to their own values and interests.

In the same vein, the root of this use of MCS was to reduce the manager’s burden of constant monitoring and intervening (Simons, 1995), leading to a formal feedback tool with a lack of interaction between managers and employees. This internal communication issue has a great influence in the well-being of the employees within the work environment. It is likely that these employees do not feel belonging to their jobs or even to the firm because of this lack of communication among them. Thus, the working environment will not have a cooperative atmosphere, where employees might support each other to develop their own skills. Then, the need for relatedness might not be fulfilled and this psychological need might be thwarted. Consequently, it is likely that the autonomous motivation is diminished.

Furthermore, managers can provoke feelings of guilt and anxiety with the diagnostic use of MCS because of the mere monitoring of negative deviations. Thus, it is highly likely that the feeling of competence that might enhance the autonomous motivation is constrained. These arguments are aligned with Henri’s (2006) study, in which he finds a negative relationship between the diagnostic use of MCS and capabilities such as organizational learning and innovativeness. This constraint might be associated with the reduction in the perceived level of competence, since they cannot further develop new insights and knowledge, as well as contribute with new ideas.

11 interdependent self-construal1_{” (i.e. autonomous motivation) (Adler and Chen, 2011, p.76).}

Nevertheless, this study assumes the concept of diagnostic use of MCS as a feedback tool of negative deviations (Simons, 1995). Therefore, the following hypothesis is:

H1: The diagnostic use of MCS has a negative relationship with autonomous motivation

2.3.2. Relationship between Interactive Control Systems and Autonomous Motivation

As we mentioned before, one of the main features of using an MCS in an interactive manner is the constant and personal involvement between employees (Simons, 1995). This frequent communication is translated into a powerful mechanism that might reduce the risks that a hierarchical structure can carry, because it is used throughout the structure of the organization (Mundy, 2010; Adler and Chen, 2011). Thus it is likely to strengthen the bonds between employees and even toward the organization. Therefore, the feeling of relatedness is boosted using MCS in an interactive way. The fulfilment of this need is crucial for the internalization of the extrinsic motivation, i.e. creation of autonomous motivation (Gagné and Deci, 2005).

With regard to this matter, some authors claim a potential threat to the need satisfaction for autonomy due to this continuous involvement of managers (Bisbe, Batista-Foguet, and Chenhall, 2010). When the interactive use is applied from a constraining view, decision-making might be centralized by managers, due to the rejection of those critical perceptions and dissenting opinions made by employees (Bisbe, et al., 2010). However, according to Simons (1995), the frequent communication among members of the organizations aims to empower individuals to “generate dialogue” (Simons, 1995). In addition, the interactive use of MCS offers alternatives for the self-direction, facilitating a sense of volition to make their own choices. This feature would satisfy the need for autonomy, beneficial for the employee autonomous motivation.

Those individuals with a certain freedom to perform tasks, tend to increase their confidence to believe in their own capabilities, exploring new situations and learning throughout the process. Since the interactive fashion of MCS is intended to search new opportunities in the market (Widener, 2007), and promotes the innate desires for creativity and innovation (Henri, 2006; Adler and Chen, 2010), the level of perceived competence will be considerably promoted, which will enhance the autonomous motivation (Ryan and Deci, 2000). Therefore, we hypothesize the following:

H2: The interactive use of MCS has a positive relationship with autonomous motivation

1 _{These authors draw into the perceived locus of causality (PLOC) and self-construal. The first concept outlines a range}

12

2.3.3. Relationship between Dynamic Tension of MCS and Autonomous Motivation.

The diagnostic use of MCS pre-establishes the targets and ensures employee compliance, while the interactive use of MCS facilitates the communication and pursuits new opportunities. Thus, diagnostic control results in negative forces that reduce the autonomous motivation, and the interactive control generates the positive energy that boosts the autonomous motivation. The combination of both uses of MCS creates a dynamic tension (Kruis et al., 2016). As Mundy (2010) argues, these two opposed levers of control need to find an equilibrium between performing according to the firm strategy and facilitating employees’ environment giving them enough autonomy to make decisions. Henri (2006) studies dynamic tension as an interaction between these two levers of control from a resource-based perspective. Even though he could not find a correlation between dynamic tension and capabilities, he did find a positive correlation with organizational performance in uncertain environments and contexts with flexible values. In addition, the diagnostic and interactive use of MCS has an interdependent character in nature, and the increase of one use of MCS can strengthen the benefits resulted by the application of the other use of MCS (Widener, 2007).

However, there are some cases in which the two uses of MCS are not balanced, and thus their effects are detrimental for the organization. For example, an excessive emphasis on an interactive use of MCS can bring a disorganization of information and firm priorities due to the continuous change, which will lead to lost employees and negative performance (Mundy, 2010). Thus, the need for competence will be declined as well as the autonomous motivation. On the other extreme, an excessive diagnostic use will greatly reduce the employees autonomy to self-determine actions and behaviours, thus their motivations will be limited to extrinsic regulators, such as rewards, punishments, bonuses, etc. (Henri, 2006; Gagné and Deci, 2005; Kruis et al. 2016).

13

H3: The dynamic tension generated from the combination of diagnostic and interactive

use of MCS has a positive relationship with autonomous motivation.

2.4. Motivational and performance-related Hypothesis

Previous research has proven the relevance of motivation in the study of performance (Sutton and Brown, 2016; Burton, Lydon, D’ Alessandro, and Koestner, 2006) and even more in job performance (Langfred and Moye, 2004; Kuvaas, 2006). One of the founders of the SDT claimed that constructive feedback enhances the intrinsic motivation of an individual performing a task; while tangible extrinsic rewards could be detrimental for the level of intrinsic motivation (Deci, 1971). In cases that those extrinsic rewards are not internalized, the balance falls in favour of controlled motivation instead of autonomous motivation (Ryan and Deci, 2000). As we mentioned before, autonomous motivation is fed by the fulfilment of the three basic psychological needs (autonomy, competence and relatedness). Baard, Deci and Ryan (2004) determined that the personal satisfaction of the needs for autonomy, competence, and relatedness can estimate their performance and well-being in the work environment.

Several researches support the potential positive effect of autonomous motivation on job performance. The meta-analysis of Cerasoli, Nicklin and Ford (2014) confirms that intrinsic motivation is a moderate or strong predictor of performance. In addition, Cerasoli and Ford (2014) aim to improve the comprehensibility of the positive causal relation between intrinsic motivation and performance. One of the reasons for this statement is the higher level of engagement at work by intrinsically motivated employees (Rich, 2006; Cerasoli and Ford, 2014). However, the reason behind intrinsically motivated individuals being more engaged with certain tasks is because they want to increase their level of competence, through the acknowledgement of new skills. Therefore, they will set goals for their own benefit, and thus improve their own performance (Cerasoli and Ford, 2014). This reason is extended to autonomously motivated people since they internalized the organizational goal as their own goal to improve their level of competence. Moreover, individuals are intrinsically motivated because they like the task or the job itself, thus they would put greater effort into it. Therefore, the performance of people that endeavour to accomplish a task because they like it, is higher than the resulted performance of people that are not interested and their effort is minimum (Ryan and Deci, 2000; Cerasoli and Ford, 2014). Hence, is expected a positive relationship between autonomous motivation and job performance. In sum, the following hypothesis has been developed:

H4: The development of autonomous motivation has a positive impact on job

14 H2: + H3: + H4: + H1: -

2.5.

Conceptual Model

The following figure represents the conceptual model of this research, containing four different hypotheses related to the key elements of MCS, motivation and performance.

Figure 2: Conceptual Model

3. Methodology

3.1.

Approach

This research follows a quantitative fashion since this study aims to contribute to the literature through testing hypotheses using an aggregation of numerical data (Blumberg, Cooper and Schindler, 2014). This approach is based on a confirmatory scientific method, stating a set of hypotheses grounded in existing theory, and afterwards, testing these hypotheses with quantitative data to confirm whether they are supported or not (Antwi and Hamza, 2015). This confirmatory scientific method is identified with theory testing and deductive reasoning, in which existing theory is used to ground certain pre-established hypotheses. The existing theory is placed also in previous empirical researches, books and theoretical articles that are able to give a better comprehensibility of a theory (Colquitt and Zapata-Phelan, 2007). The deductive research fashion aims to analyse certain theory in literature, assumes certain related hypotheses and, evaluates whether that theory applies under those circumstances (Beiske, 2007). That is why this deductive reasoning is a process that goes from general to specific considerations (Pelissier, 2008).

15

3.2.

Data collection and sample

To analyse the effects of the proposed relationships, I have used secondary data property of my supervisors at the University of Groningen. The data set is composed by 500 complete surveys conducted in four higher educational institutions in The Netherlands (i.e. universities of applied science). This data base is chosen due to its usefulness for this study, since it relies on the Levers of Controls (Simons, 1995) and the types of motivation according to the SDT. Only a part of this data set is facilitated and included in this research. The use of surveys is highly recommended by several authors. Most of the methodologies for management accounting and control research based on surveys are applied in theory testing (Van der Stede, Young, and Chen, 2005). The use of surveys allows obtaining honest perceptions, feelings and opinions from an individual level. In addition, they facilitate the comprehension of a complex situation, since they “occur in their natural setting”, and simultaneously, maintain the required level of standardization for this type of analysis (Speklé and Widener, 2008, p. 3).

16 Table 1: Descriptive statistics sample

Variable Frequency Percentage

Language (n=252) EN 23 9% NL 229 91% Gender Respondents (n=252) Female 127 50% Male 125 50% Age Respondents (m=252) 20-30 years 7 3% 31-40 years 51 20% 41-50 years 68 27% 51+ years 126 50% Organizational tenure (n=252) 0-10 years 134 53% 11-20 years 76 30% 21-30 years 30 12% 30+ years 12 5% Departmental tenure (n=252) 0-10 years 162 64% 11-20 years 64 25% 21-30 years 20 8% 30+ years 6 2% Educational Background (n=252) Primary education 0 0% Bachelor’s degree 64 25%

Master’s degree or higher 184 73%

Secondary education 0 0%

3.3. Data analysis procedure

After gathering all the responses from the survey, the sample was checked for missing data and outliers. Subsequently, the normality of the constructs was tested. Both univariate and multivariate kurtosis and skewness indicate whether the complete set of indicators departs from a normal distribution. According to Kline (2011), the threshold for univariate kurtosis remains between -7 and 7 and for univariate skewness is placed between -2 and 2. Instead, the multivariate kurtosis needs to be above 5 points (Byrne, 2009). Moreover, a multicollinearity test was conducted analysing the values for Variance Inflation Factor (VIF), which cannot exceed 10 points, otherwise the variable might be redundant (Kline, 2011).

17 be evaluated, in order to draw stronger conclusions about the potential results. Then, a Confirmatory Factor Analysis is performed to check the model fit and the reliability and validity of the constructs. It is necessary to assess the model fit with respect to the available dataset, i.e. whether the model can define the used data. We use the following thresholds for the most important measures: chi-square p-values > 0.05; CFI > 0.95 (Hu and Bentler, 1999); AGFI > 0.80 (Hu and Bentler, 1999); RMSEA < 0.1 (Browne and Cudeck, 1993) and PCLOSE > 0.05 (Hu and Bentler, 1999). Once I obtain an appropriate model fit, we can proceed to check the internal validity and reliability of the constructs, i.e. find a clean and consistent pattern of indicators that represent the construct reliably. Nunnally (1978) claimed that a construct is internally reliable to perform statistical analysis when its Cronbach’s alpha is higher than 0.70. In terms of the CFA, the reliability will be measured with the Composite Reliability (CR>0.7); the convergent validity is determined with the Average Variance Extracted (AVE>0.5); and the discriminant validity is observed with the Maximum Shared Variance (MSV<AVE) (Kline, 2011).

After obtaining internally validated and reliable constructs, the multi-items indicators are summed up and divided by the total amount of items in each factor. In the case of the dynamic tension construct, we follow the approach proposed by Marsh et al. (2004). For that purpose, it is recommended to standardize the indicators –diagnostic use of MCS and interactive use of MCS– in order to avoid potential problems of multicollinearity (Jöreskog and Yang, 1996). The option to compose the match-pair indicators of dynamic tension is to multiply the most internally valid item of diagnostic use of MCS with the most internally valid item of interactive use of MCS, and so on (Marsh et al., 2004). This methodology has been recommended and applied in several empirical studies (see e.g. Eisenberger, et al., 2010; Bellora-Bienengräber and Günther, 2012). Since the interactive use of MCS only has three items, the indicators will be restricted to just three match-pairs. Once we have the three match-pairs indicators of dynamic tension, we proceed to compute the construct as previously described.

Finally, the data set is ready to perform SEM in order to delve into the relationships between the uses of MCS, autonomous motivation and job performance.

3.4. Measurements

3.4.1. Diagnostic use of MCS

18

3.4.2. Interactive use of MCS

The interactive use of MCS is measured according to Bedford and Malmi (2015). They describe it as a set of frequent proactive involvements between managers and subordinates to enhance the internal communication, creativity and undertake strategic uncertainties. This variable is established according to the formative measurement model from Bisbe et al. (2007), which is based on five different features: (1) intensive use by top management, (2) intensive use by operating managers, (3) face-to-face challenge and debate, (4) focus on strategic uncertainties, and (5) non-invasive, facilitating and inspirational involvement. According to Widener (2007), Henri (2006), and Bisbe and Otley (2004), a survey with 4 items based on a 7-points Likert scale has been applied. The Cronbach’s alpha for this construct is 0.916, although the first item had to be deleted due to its low level of correlation with this construct in the factor analysis. Thus, interactive use of MCS is a validated variable for this study.

3.4.3. Dynamic tension

Dynamic tension refers to the combination of diagnostic use of MCS and interactive use of MCS (Simons, 1995). Traditionally, researchers have relied on regression techniques instead of using structural equation modelling (SEM) to analyse interaction effects. Thus, they have suffered from low power, since they are not capable to manage measurement errors (Steinmetz, et al., 2011). For that reason, latent interaction modelling was proposed as an alternative to reduce the bias. This study will follow the unconstrained approach elaborated by Marsh, Wen and Hau (2004). This procedure relies on matched-pair products and the information of each of the indicators for interactive and diagnostic use of MCS are recognised only once (i.e. if factor A contains two indicators a1 and a2, and factor B contains two indicators b1 and b2, a set of

matched-pair indicators would be a1b1 and a2b2). Moreover, the indicators are standardized before the

formulation of the interaction variable (Marsh et al., 2004). The reasons to choose this approach are its ease of use and applicability and the fact of not considering any constraints about the multivariate normality assumption of the variables (Marsh et al, 2004).

The results of the reliability analysis are quite favourable since the Cronbach’s alpha reaches the value of 0.900. Therefore, this non-observable construct is internally valid and reliable to continue with the analysis.

3.4.4. Autonomous motivation

19 separately or the aggregation into autonomous motivation and controlled motivation. According to the authors, the most convenient way to estimate autonomous motivation in the work environment was through the measurement of intrinsic motivation and identified regulation, since integrated regulation has been highly complicated to distinguish from identified regulation (Vallerand et al., 1992). Therefore, identified and intrinsic motivation constructs consist of three different items, each one measured through a 1-7 Likert scale; and autonomous motivation is measured through the complete set of items of these two types of motivation. The reliability analysis indicates that the three constructs are internally validated, since their Cronbach’s alpha are 0.868 for identified regulation; 0.897 for intrinsic motivation; and 0.848 for autonomous motivation. Therefore, the final values are adequate and acceptable to continue with the analysis.

3.4.5. Job performance

Job performance is a widely recognized instrument developed by van de Ven and Ferry (1980); whose core focus is on public organizations. Several dimensions were taken into account to develop this measure, such as the perceived productivity, quality of work performed, innovation, and reputation, attainment of goals, efficiency, and morale of the personnel. The participants were asked to indicate the perceived job performance through a 5-points Likert scale (1 is far below average, 5 is far above average), with respect to other jobs. In contrast with the previous variables, job performance does not obtain a similar high Cronbach’s alpha (0.723) even excluding the first, the second, and the third item due to low level of factor loading in factor analysis. Nevertheless, its value stays above the limit of 0.70 (Nunnally, 1989).

3.4.6. Validation variables

The inclusion of various contextual factors as validation variables might affect the proposed hypotheses and enhance robustness to the findings. Since the organizational context can influence the use of MCS (Chenhall, 2003), we incorporate organizational structure as a validation variable to increase the robustness of the results. The study from Bedford, Malmi, and Sandelin (2016) was used to develop the questionnaire. This variable is represented as a continuum from mechanistic to organic (Burns and Stalker, 1961). The formulation of the items was based on Chenhall and Morris (1995), Covin, Slevin and Heeley (2001) and Leifer and Huber (1977) and answered on a 7-point Likert scale. The factor loading in the factor analysis indicates that one of the five items of this construct does not relate to it, therefore it was deleted from the analysis. The final Cronbach’s alpha is 0.781, thus this construct is internally validated.

20 and Humphrey (2006). Among the variables within the WDQ, only knowledge characteristics are included in this study. This variable refers to the set of “knowledge, skills, and ability demands” that an employee possesses in order to perform his or her job (Morgeson and Humphrey, 2006). Knowledge characteristics consist of four constructs measured through a 5-points Likert scale and 4 items for each of them. The first item of job complexity had to be excluded since it was not related to this construct during the factor analysis, thus the final Cronbach’s alpha was 0.858. The reliability of information processing was acceptable (Cronbach’s alpha = 0.846) and all the former items were included. In terms of problem solving, there is a problem of discriminant validity, although this issue might be predictable since this construct might be correlated to other components of knowledge characteristics such as skill variety. Thus, the Cronbach alpha of problem solving is 0.842. With respect to skill variety, the Cronbach’s alpha shows acceptable values (0.864), reflecting a reliable construct. Finally, the Cronbach’s alpha of knowledge characteristics is 0.820.

3.4.7. Control variables

Job tenure is included as a control variable since there is empirical evidence of the correlation with employee job performance (Tsai, Chen, and Liu, 2007). This variable is provided as organizational tenure and departmental tenure. Other variables such as gender, age, and educational background are also considered.

The inclusion of any of the abovementioned variables are included in the hypothesis testing only if it has a significant correlation with any of the dependent variables, i.e. autonomous motivation and job performance. Otherwise, the variables are not taken into account and therefore eliminated from the study.

4. RESULTS

4.1. Raw Data

After observing one missing value, we obtain a final sample of 251 responses. To perform all the preliminary checks, and hypothesis testing, we use IBM SPSS 25 and IBM SPSS AMOS 25 software programmes.

21 variables, thus the potential issue of multicollinearity incurred by the interaction of both uses of MCS is correctly mitigated.

A Confirmatory Factor Analysis (CFA) is displayed to evaluate the model fit and to assess the internal validity and reliability of all the constructs. We first build a basic measurement model with all the variables of our conceptual model. The diagram includes standardized values of the factor loadings, and it is represented in Appendix A. The p-value of the chi-square shows a poor fit of the model (Chi-squared = 337.115; p-value = 0.000), although this value is generally common due to the sample size (Hair et al., 2006). In terms of others fit measurements, there is an acceptable model fit with the data (CMIN/DF=1.915; AGFI=0.853; CFI=0.961; RMSEA=0.061; PCLOSE=0.040), although these results are obtained by allowing covariance between the error terms of identified regulation. The values remain at the limit of the thresholds, which might generate potential issues in the hypotheses testing. In addition, the CFA facilitates the reliability, discriminant and convergent validity of the constructs (Table 2).

Table 2. Validity and Reliability

Model 1 Model 2

CONSTRUCT CR AVE MSV CR AVE MSV

(1) Diagnostic use of MCS .964 .843 .468 .964 .844 .468 (2) Interactive use of MCS .916 .785 .468 .917 .786 .468 (3) Dynamic tension .901 .753 .154 .901 .752 .154 (4) Autonomous motivation .813 .460 .141 .864 .801 .090 (5) Job performance .737 .416 .141 .737 .417 .090

22 This new model obtains a Chi-square of 379.437 and a null p-value, thus again representing a poor fit of the data. However, as we previously indicated, these indicators usually result in unfavourable fit due to the great sample size (Hair, et al., 2006). With respect to the other measurements, we obtain the following values: CMIN/DF=2.144; AGFI=0.836; CFI=0.951; RMSEA=0.068; PCLOSE=0.001). These results might represent slightly worse values than in model 1 with first order factors. However, we could state that the model fit is relatively acceptable according to the sample size of our data base. Consequently, it is necessary to evaluate the standardized factor loadings of the first order factors, (i.e. identified and intrinsic motivation’s indicators), whose values have improved substantially. These differ between 0.74 and 0.94, while the values of the former model ranged from 0.23 to 0.95. Therefore, it seems that the second model is an upgraded version. Then, we assess the standardized regression weights in order to examine whether the first order factors (i.e. identified and intrinsic motivation) are significant to the second order factor (i.e. autonomous motivation). Fortunately, the obtained estimates are both significant at 0.05 with a p-value of 0.014 (r = 0.382 for identified motivation and r = 1.207). The value for intrinsic motivation seems to be a Heywood case, since the standardized regression weight is higher than 1 (Hair, et al., 2006). This issue might be due to the limited number of latent variables that forms autonomous motivation (i.e. identified and intrinsic motivation), and the strong correlation between intrinsic motivation and autonomous motivation, which makes it to have greater representation of this construct than with identified motivation. This issue will be outlined in the limitations of the study.

Finally, the convergent validity issues of autonomous motivation have been successfully mitigated with a new AVE of 0.801, and the reliability of this construct has increased to 0.864, way above the limit of 0.70 (Hair, et al., 2006).

4.2. Correlations

23 intrinsic motivation (i.e. autonomous motivation), which might support the existing literature (Gagné and Deci, 2005; Widener, 2007; Ferreira and Otley, 2009).

The other validation variable for this study is knowledge characteristics. The correlations of this construct have been exposed in separate components and as an aggregation. Nevertheless, all the individual variables of knowledge characteristics –complexity, information processing, problem solving, and skill variety– have a significant positive relationship with autonomous motivation. These results are in accordance with literature, in which jobs with high level of knowledge characteristics (i.e. professional occupations such as educational jobs) might have an autonomy-supportive environment (Morgeson and Humphrey, 2006; Gagné and Deci, 2005) and consequently develop higher levels of autonomous motivation. The only remarkable observation is that complexity does not have a significant correlation with identified motivation. This inexistent relationship might be due to a higher sensitivity of complexity, and its need for further research (Gagné and Deci, 2005). In addition, only problem solving, skill variety in particular, and knowledge characteristics in general are positively correlated with job performance.

The only control variable that is significantly correlated with any of the main variables is gender. The direction of this relationship is towards the two uses of MCS, although it is slightly stronger in diagnostic use of MCS (r = .164) than in interactive use of MCS (r = .147). In addition, gender is also significantly correlated with identified regulation (r = .161), thus is expected to be differences between male and female in the development of identified regulation (i.e. autonomous motivation).

Regarding the uses of MCS, diagnostic and interactive use of MCS are correlated to each other, as Widener (2007) addressed in her empirical research. Therefore, both uses of MCS are complementary and interdependent. However, only diagnostic use of MCS is significantly correlated with dynamic tension (r = -.365). The lack of significant correlation between interactive use of MSC and dynamic tension may be due to an unbalanced combination of both uses in these universities. Moreover, there is a positive relationship between the uses of MCS and intrinsic motivation (r = .173 and r = .187), being stronger in the case of interactive use of MCS. These first findings are in line with the arguments built from management literature and SDT, even though the relationship between diagnostic use of MCS and intrinsic motivation bears a positive sign. Lastly, it is highly likely that we will find a positive relationship between autonomous motivation and job performance, since the correlations are positively significant.

4.3. Structural Equation Models

24 simultaneously. Furthermore, it can display unobservable or latent constructs –as dynamic tension or autonomous motivation– that are measured by different indicators or also referred as manifest variables (Hair et al., 2006), as well as it can outline the measurement of specific errors (Hair, et al., 2006). Another reason to apply this method is having one interaction term, which is dynamic tension. Its results tend to be more accurate than using traditional techniques such as regression analysis (March, et al., 2004). We will conduct an assessment of the direct effects of the independent variables on the dependent variable job performance in order to reinforce our results. Hence, SEM is the most appropriate technique for these situations.

Table 3 displays the results from three structural equation models with all the measurement indicators. In preliminary trials, gender as control variable was included, although the significant paths arisen without the control were disappeared, thus gender was finally excluded from following analysis and tests. Firstly, Model 1 includes the relationships of diagnostic use of MCS and interactive use of MCS with autonomous motivation, and the relationship between autonomous motivation and job performance. These tested relationships correspond to Hypothesis 1, 2 and 4 respectively. Secondly, Model 2 additionally tests the relationship between dynamic tension and autonomous motivation, related to Hypothesis 3. The direct effects between the uses of MCS and dynamic tension, and job performance are also analysed in model 1 and 2. Thirdly, Model 3 is trimmed since the direct effects of the independent variables on job performance are insignificant. All models will be assessed for goodness of fit (Table 4). Model 1 has a better fit with the data comparing with Model 2. The model fit of the third model is quite similar to model 2, thus we support this model since hypothesis 2 has different results. Therefore, even though the chi-square measure indicates a poor fit of the data, the CMIN/DF, CFI, and RMSEA are all above the limit, which leads us to obtain adequate and acceptable models that can be tested and interpreted in this study.

25

4.4.

Hypotheses Tests

The results of the hypotheses testing are displayed in Table 3. Based on the analysis presented in Model 1 and Model 2, we found a marginally positive relationship between interactive use of MCS and autonomous motivation (p<0.10). In addition, the expected positive relation of autonomous motivation and job performance is strongly supported (p<0.01). The direct effects between the uses of MCS and dynamic tension were not significant in this study. These non-hypothesized relationships are trimmed in consecutive analyses. Hence, both Model 1 and Model 2 conclude with a support of hypothesis 2 and 4. In contrast, Model 3 shows only a positive relationship between autonomous motivation and job performance (p<0.01). For that reason, interactive use of MCS should be further analysed since its findings are not fully confirmatory.

Table 3. Model 1, Model 2, and Model 3

4.5.

Additional results

Table 4 indicates the results of the hypotheses testing within two segments of the population in terms of organizational structure, which reflect robustness of our findings. The relationship between interactive use of MCS and autonomous motivation has been reinforced only in a situation where the organizational structure is mechanistic. Since the mechanistic approach is rather associated with diagnostic use of MCS (Ferreira and Otley, 2009), it was more likely to find a negative impact of interactive use of MCS on autonomous motivation in this context. In addition, the non-expected positive impact of interactive use is even higher than in the preliminary results of Table 3 (r=.483; p<0.05). Contrary to previous results, dynamic tension shows a significant coefficient in its relationship with autonomous motivation in a mechanistic and organic

Model 1 Model 2 Model 3 (trimmed) Relationship

Stand. Estimate (p) Stand. Estimate (p) Stand. Estimate (p) Diagnostic use  autonomous motivation .077 (.432) .094 (.402) .101 (.321) Interactive use  autonomous motivation .140 (.060*) .129 (.103*) .125 (.122) Dynamic tension  autonomous motivation - .023 (.886) .023 (.865) Autonomous motivation  Job performance .381 (.009***) .293 (.008***) .311 (.004***)

Diagnostic use  Job performance -.067 (.640) -.046 (.865) - Interactive use  Job performance .092 (.433) .077 (.541) -

Dynamic tension  Job performance - .029 (.689) -

Model fit X2 251.825 379.437 380.329 p-value .000 .000 .000 Df 127 177 180 CMIN/DF 1.983 2.144 2.113 CFI .964 .951 .951 RMSEA .063 .068 .067

26 approach. In the former approach, this relationship is negative (p<0.05) whereas in the latter approach, it is positive (p<0.05). In relation with hypothesis 4, the positive significant results depending on organizational structure indicate an increase in their robustness.

Table 4. Multi-group analysis

Organizational structure

Relationship Mechanistic Organic

Diagnostic use  autonomous motivation (-) -.429 .177 Interactive use  autonomous motivation (+) .488** .025 Dynamic tension  autonomous motivation (+) -.638** .154**

Autonomous motivation  Job performance (+) .456** .361**

Model fit X2 630.799 .000 360 1.752 .934 .056 p-value Df CMIN/DF CFI RMSEA X2 diff. test X2 difference (df) 20 p-value S (.035)

***, ** and * refers respectively to a correlation significant at the 0.01, 0.05 and 0.10 (2-tailed)

27 Table 5. Moderator between uses of MCS and autonomous motivation

Relationship Knowledge characteristics Complexity Information processing Problem solving Skill variety Diagnostic use  autonomous motivation (-) .140 .199** .133 .137 .117 Interactive use  autonomous motivation (+) -.059 -.061 -.038 -.067 -.056 Dynamic tension  autonomous motivation (+) .076 .113 .064 .095 .096 Autonomous motivation  Job performance (+) .329*** .329*** .329*** .329*** .329*** Diagnostic use × Moderator  autonomous

motivation

-.105 -.212* -.060 .000 .080

Interactive use × Moderator  autonomous motivation

.037 .096 .092 -.052 -022

Dynamic tension × Moderator  autonomous motivation

-.089 -.263** -.074 .025 .146 Moderator  autonomous motivation .419*** .245*** .321*** .370*** .359***

Model fit X2 6.236 4.861 8.909 5.985 6.521 p-value .512 .677 .259 .542 .480 Df 7 7 7 7 7 CMIN/DF .891 .694 1.273 .855 .932 CFI 1 1 .997 1 1 RMSEA .000 .000 .033 .000 .000 ***, ** and * refers respectively to a correlation significant at the 0.01, 0.05 and 0.10 (2-tailed)

The moderation effect over the second part of the conceptual model (i.e. between autonomous motivation and job performance) has a great lack of significant results (Table 6). First of all, the moderation effect is not significant in any of the potential moderator variables. However, we find significant coefficients in the relationship between diagnostic use of MCS and autonomous motivation in terms of complexity and problem solving. Both results have positive impacts in the relationship (r = 0.162 and r = 0.152 respectively), which seems to contradict the expected negative hypothesis based on literature. On the other hand, hypothesis 4 is again fully supported in all the possible cases in Table 6.

Table 6. Moderator between autonomous motivation and job performance

Relationship Knowledge characteristics Complexity Information processing Problem solving Skill variety Diagnostic use  autonomous motivation (-) .130 .162* .127 .152* .095 Interactive use  autonomous motivation (+) -.063 -.027 -.029 -.097 -.042 Dynamic tension  autonomous motivation (+) .066 .084 .054 .094 .065 Autonomous motivation  Job performance (+) .345*** .353*** .338*** .325*** .355*** Autonomous motivation × Moderator  Job

performance

28

5. DISCUSSION AND CONCLUSION

5.1. Summary of the results and discussion

Firstly, the results cannot support the negative impact of diagnostic use of MCS on autonomous motivation outlined in Hypothesis 1. This might be because Simons (1995) analyses the uses of MCS from a managerial perspective. He argues that managers should be responsible for the adequate exercise of the organizational control. Moreover, since diagnostic use of MCS focuses on reducing the manager’s burden of constant monitoring, it is likely that the communication between managers and subordinates is poor. Therefore, the subordinates might not acknowledge the way performance measures are used. On the other hand, additional results revealed a positive relationship between diagnostic use of MCS and autonomous motivation mitigated by a higher level of complexity. Adding complexity to the model gives significance to Hypothesis 1, although obtaining a non-expected positive impact towards autonomous motivation. According to Adler and Chen (2011), diagnostic control systems can be applied in an enabling way, which enhances transparency and coordination within the company, leading to a higher self-development of employees (i.e. increasing autonomous motivation). In addition, complex situations involve top management restructuring the strategy of the firm, bringing uncertainty and ambiguity along for the rest of employees (Abernethy and Brownell, 1999). Thus, the positive development of autonomous motivation through the diagnostic control system is weakened.

In terms of Hypothesis 2, model 1 and model 2 outline a significantly positive coefficient for the relationship between interactive use of MCS and autonomous motivation, which is in line with the initial expectations. However, this relationship is no longer significant in the trimmed model, hence it leads us to delve deeper into this relationship. Besides that, the interactive use of MCS has a positive influence on autonomous motivation under a mechanistic structure. Previous research (Beldford et al., 2016; Ferreira and Otley, 2009) associates mechanistic structure with diagnostic use of MCS and organic structure with interactive use of MCS. Therefore, we might expect a negative impact of interactive use of MCS on autonomous motivation under a mechanistic structure, since this context is restrictive and repressive (Adler and Chen, 2011; Beldford et al., 2016; Ferreira and Otley, 2009). However, the organizational structure construct in this study is based on a scale from mechanistic to organic approach, thus different levels of structure can be implemented (e.g. more or less mechanistic or more or less organic). In this case, there might be a low level of mechanistic organizational structure, which has helped to obtain a positive impact of the interactive use of MCS on autonomous motivation.

29 MCS, associated with the interactive use of MCS. This neutralization of the negative and positive forces has led to a lack of impact on autonomous motivation. However, ex post analyses reveal a negative impact of dynamic tension on autonomous motivation under a mechanistic structure, and a positive impact on autonomous motivation under an organic approach. These results are coherent with Simons (1995) and other authors (Adler and Chen, 2011; Beldford et al., 2016; Ferreira and Otley, 2009) since a mechanistic structure is identified with centralized and tight controls, whose flow of information is rather restrictive; and the organic approach is associated with more flexible and informal controls. Therefore, since dynamic tension tends to foster the dialogue due to the negative and positive forces and, it fulfils the three psychological needs, i.e. autonomous motivation, the outcomes of this relationship change depending on whether the structure of the company is tighter like the mechanistic approach, or either more flexible like the organic approach.

Finally, looking into Hypothesis 4, we have obtained extensive evidence to support the positive impact of autonomous motivation on job performance, which is in line with the propositions claimed in the SDT (Gagné and Deci, 2005). However, no significant results have been found about the direct effects of the uses of MCS on job performance.

5.2.

Contributions

As mentioned in the introduction, this study aims to contribute to both the managerial accounting and the psychological field. In the first place, it sheds some light on the effects of dynamic tension. This combination of diagnostic and interactive use of MCS shows significant impact on autonomous motivation under a mechanistic structure and under an organic structure. Dynamic tension tends to be theoretically evaluated (Mundy, 2010; Kruis et al., 2016), and this quantitative study provides new insights from a motivational perspective. Secondly, this study generates further knowledge within the Self-Determination Theory, since it is found that an additional element from the workplace, i.e. the use of MCS has an influence on the level of motivation (Gagné and Deci, 2005; Van den Broeck, et al., 2010). This way, interactive use of MCS has significant positive effects on autonomous motivation –specifically under a mechanistic structure–, and unexpected evidence of a positive relationship between diagnostic use of MCS and autonomous motivation is dampened by high level of complexity. Moreover, this study empirically supports the line of research developed within the SDT about the positive well-being for the employees such as job performance through the acquisition of autonomous motivation (Deci, Olafsen, and Ryan, 2017).

30 placed in the work environment (Chenhall, 2003), managers can combine them and make use of them in different fashions. This study found that the uses of MCS can have certain influences on autonomous motivation. Moreover, we found relevant results about the positive effects of the combination of different uses of MCS, such as the product of diagnostic and interactive use of MSC as dynamic tension, specifically under an organic structure. Accordingly, this tension shows its advantages to create autonomous motivation, which could foster dialogue from different perspectives (Chenhall, 2003), complete the double loop learning to acquire certain level of competence (Argyris and Schön, 1978), and finally communicate and empowering new insights with other employees (Henri, 2006).

5.3.

Limitations

In order to obtain the desired and final outcomes, this study addresses certain limitations. The first limitation is related to the generalizability of the results. This study scopes only to the Netherlands. Moreover, it is restricted to the educational sector, which is non-profit oriented, while the original study of Simons (1995) is based on the private sector. In the same vein, the segment of this study –employees without managerial duties– may be too restrictive to analyse the uses of MCS. According to Simons (1995), the MCS are used by managers and this study gathers the perceptions of employees who are not managers, and it may not reflect the complete representation of what the uses of MCS are. Therefore, caution is needed about the generalization of results due to country, sector, and type of employee limitation. Secondly, we incurred a Heywood case in the construct of intrinsic motivation at the time that the second order factor of autonomous motivation is in place. This Heywood case is noticeable in this study because of a standardized regression weight greater than 1, which was caused by the generation of negative variance estimates in the first proposed diagram. This issue occurs quite often during the factor analysis of structural equation models (Kolenikov, and Bollen, 2012). The potential causes that might influence the occurrence of a Heywood case are construct misspecification or highly strong correlation between intrinsic motivation and autonomous motivation comparing to identified motivation. Last but not least, this empirical study leave outside the measure of the three psychological needs; it just includes them theoretically. This may also be a potential explanation for the misspecification of some constructs such as autonomous motivation in our study.

5.4.

Future research

31 the three psychological needs in the hypothesized relationships might be seized to advance in further research. Therefore, future research could look into the relationship between the different uses of MCS and the three psychological basic needs and consequently the impact on controlled motivation and autonomous motivation. In that way, a clearer image of the MCS literature and the SDT can be better understood. Future research could also study MCS as packages based on the LOC, since they share complementarity and interdependency (Widener, 2007), and evaluate their influence on the different types of motivation. Finally, this study has risen high interest in the complexity variable, and how the relationship between the different uses of MCS and autonomous motivation is affected. This proposed future research could take place in a company with diversified levels of complexity in order to see the differences in several contexts. Therefore, further research could add to the literature based on Deci, Olafsen, and Ryan (2017) who claim that heuristic, complex and quality-based tasks or jobs have different results whether it is emerged through autonomous motivation or through controlled motivation.

5.5.

Conclusion

32

33

34

Appendix C – Correlation Table

MN= Means; SD=Standard Deviation

***, ** and * refers respectively to a correlation significant at the 0.001, 0.01 and 0.05 (2-tailed)