The effects of management control system design on performance

Amsterdam Business School

The effects of management control system design on performance

Name: Rutger Booij Student number: 11163496

Thesis supervisor: Sander van Triest Date: 14 August 2016

Word count: 16669, 0

MSc Accountancy & Control, specialization Control

Statement of Originality

This document is written by student Rutger Booij who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

This study investigates which control measures and corresponding degree of tightness, should be included in the management control system (MCS) to maximize employee and task performance in professional service firms (PSF). In addition a distinction is made between high and low complexity tasks. A survey based database is used with 265 participants. We make use of the categorization of control measures by Merchant and van der Stede (2012), identifying behavioral, results, personnel and cultural control measures. We found statistical significant results for a negative relationship between implicit behavioral control tightness and performance. For the cultural controls we identified a positive relation between tightness and performance.

Furthermore we found that the employee reputation, as to be considered as experts, had a positive effect on performance.

Contents

1 Introduction ... 6

2 Literature and hypothesis ... 8

2.1 Literature Review ... 8

2.1.1 Agency theory ... 8

2.1.2 Management Control System ... 9

2.1.3 Performance measures ... 10

2.1.4 Contingency theory ... 10

2.1.5 Management control system tightness ... 12

2.1.6 Professional service firms ... 13

2.1.7 Task complexity ... 14

2.1.8 External factors... 15

2.2 Hypotheses ... 16

3 Methodology ... 18

3.1 Main dependent variables ... 18

3.1.1 Employee performance ... 18

3.1.2 Task Performance ... 19

3.2 Independent Variables ... 19

3.2.1 Explicit and implicit control tightness ... 20

3.2.2 Control Measures ... 20 3.3 Control Variables ... 22 3.3.1 Size ... 22 3.3.2 Experience ... 22 3.3.3 Market Competition ... 22 3.3.4 Market Uncertainty ... 23

3.3.5 Relative company performance ... 23

3.4 Task Complexity ... 24

3.5 Schematic overview ... 25

3.6 Regression models ... 25

4 Data ... 28

4.1 Part of the research project ... 28

4.2 The questionnaire ... 29

4.3 Data Summary ... 30

5 Results ... 32

5.1 Effects on own performance ... 32

5.1.1 Univariate analysis ... 32

5.1.2 Multivariate analysis ... 33

5.2 Effects on supervisor performance... 35

5.2.1 Univariate analysis ... 35

5.2.2 Multivariate analysis ... 36

5.3 Role of task complexity ... 38

5.3.1 Univariate analysis Employee Performance ... 38

5.3.2 Multivariate analysis Employee Performance ... 39

5.3.3 Univariate analysis Task Performance ... 41

5.3.4 Multivariate analysis Task Performance ... 42

6 Discussion ... 44

6.1 Effects on own performance ... 44

6.2 Effects on task performance ... 45

6.3 Role of task complexity ... 45

6.4 Hypotheses evaluation ... 46

6.5 Limitations ... 47

7 Conclusion ... 49

1 Introduction

Having a good functioning management control system (MCS) in place is of vital importance for the survival of a company. A company that is out of control could suffer severe losses or even bankruptcy (Merchant & van der Stede, 2012). There are numerous examples in the recent past of companies running out of control leading to reputation damage, recovery costs and bankruptcy. There was fraud discovered at the German car manufacturer Volkswagen regarding emission values, a case that is still pending (Wikipedia 1). The reputation damage has already been done and a large fine and additional costs are expected to come. Another example is the financial fraud at Imtech, a technical service provider. In 2013 the fraud was discover and the company went bankrupt in 2015 (Wikipedia 2).

One of the underlying problems lies in the fact that people are self-interested and pursue their own goals. A problem arises when there is a difference in hierarchy, where a manager is asked to operate on behave of the owner for example (Ross, 1973). This problem is known as the agency theory. This problem can be mitigated using a well functioning MCS (Merchant and van der Stede, 2012). The problem however is that there is not a single best system of configuration of a MCS. A system has to be adapted to a specific company or a specific environment (Fisher, 1998). This is captured in the contingency theory, where a configuration has to be found to fit the situation. Choices then have to be made which control measures to include into the MCS and the degree of tightness has to be determined.

In this study we make use of the categorization of control measures as developed by Merchant and van der Stede (2012). Four measures are identified; Behavioral controls which include rules, guidelines and best practices. The second are the result controls such as goals and targets. The third category are the personnel controls entailing amongst others selection and training. The last category are the cultural controls including norms, values and code of conduct.

Next to the choice in control measures the degree of tightness has to be determined for each measure. There is distinction between explicit and implicit tightness. The first defines the scope and magnitude of the measures. The latter is a measure on how strict the explicit measures need to be applied or lived up to.

The role of the MCS is both coercive and enabling (Adler and Borys, 1996). Coercive to force actions in the desired directions but also enabling to ensure that the manager or employee is able to maximize its performance.

This study focuses on the professional service firms (PSF). A large share of the current studies in the control research are devoted to manufacturing firms, however only a small part is devoted to the service firms (Auzair, 2005). This is a bit strange as the percentage contribution to

the global gross domestic product (GDP) in 2014 was 68.5% by the service industry compared to only 14.7% by the manufacturing industry (Worldbank). Two distinct attributes of the PSF is that its main assets are the employees and the product delivered is an intangible perishable, coproduced service (Greenwood et. al, 2005).

Since the literature is still lacking extensive research in this area, there is a lot of room for contributing to the current knowledge. With this study we try to find which control measures, and corresponding degree of tightness, should be included in the MCS to optimize employee performance in professional service firms. Employee performance is measures in two ways. First there is the own employee performance measured on quality and quantity of work. Second is the task performance, which is measured as to what extent they feel they fulfill the required the required tasks and duties in their job. Effects of the afore mentioned four categories are tested one by one on their effect, if any, on the two dimensions of performance. Finally we make a distinction between high and low complexity tasks and examine if there exists a difference between them.

This study is part of the PSF Thesis Survey Project 2015 - 2016, a project initiated by the University of Amsterdam. Data is gathered using a survey which had already been developed in the program. Together with a limited amount of other students we made a contribution to the database and gained access to this complete dataset. This study is able to make use of the unique dataset and therefore produce unique results. In addition little research is devoted to the PSF so this study can add to the limited amount of available existing data.

The results of this study show that the use of implicit behavioral control results in a negative relation between tightness and performance. In addition we found a positive relation between tightness and performance for both forms of the cultural controls. To conclude we found that the relative employee reputation, defined as being the best in the field and an expert on the job, had a large effect on the individual employee performance.

In the coming sections there is first a literature overview and hypotheses, after that we present our methodology. This is followed by the data description and results. We end this study with a discussion of the results and a conclusion.

2 Literature and hypothesis

To provide a basis for this research we will give a summary of the prevailing which is relevant for this study. This is done in the first paragraph. In the second paragraph the hypotheses are stated and explained. The hypotheses are based on the knowledge acquired in the literature section.

2.1 Literature Review

The first paragraph is devoted to the literature review. In the following eight sub paragraphs we will discuss the agency and contingency theory which are important reasons to have a proper designed MCS in place. This is followed by the aspects of a MCS, tightness, control measures and task complexity. In addition we outline the special characteristics of a PSF and external factors.

2.1.1 Agency theory

The agency theory relates to a phenomena whereby difficulties can arise when two parties cooperate. These difficulties can arise if the two cooperating parties have different goals or interests because individuals tend to act in their self-interest. This is most predominant when there is a difference in hierarchy between the two parties (Ross, 1973). The problem arises when a party (the principal) requests work to be carried out for him by another party (the agent) whereby their goals are not aligned. In addition the other part of the agency theory describes the problems to monitor the agent to ensure the goals are compatible with the goals of the principal (Jensen 1976). These problems can be addressed using control measures, that provide guidance to the agent and enables monitoring by the principal.

An aspect of the agency theory is the information asymmetry, where one party knows more than the other party. Typically the agent knows more about his qualifications and the actual tasks carried out, along with the quality of it. The principal, on the other hand knows better what is expected on how the task is carried out or what the (quality of) the outcome should be (Pratt & Zeckhauser 1985). Another problem is the opportunism, an agent acting in its own interest. Elements of opportunism are described by Jap and Anderson (2003) as “distortion of information”, which includes cheating, stealing and lying but also the selective disclosure of information. But also failing to meet predefined commitments by exhibiting shirking behavior and by not living up to promises and obligations.

Ekanayake (2004) sums up the agency problems in two categories. The first is the monitoring problem as the principal cannot easily determine the quality of work and the efforts

of the agent. The second is the risk-sharing problem when the agent and principal have a different level of risk attitude. The research conducted on agency theory has two main areas; positivist and the more general principal-agent research. The positivist research is mostly concerned with the alignment of goals between the agent and the principal and the mechanisms to achieve this. Positivist research focuses mainly on the relation between the owner and managers. The other area is the more general principal-agent research whereby the role of both the principal as well as the agent can be interpreted on many different levels besides the regular owner – manager relation, such as employer – employee or buyer – supplier (Harris & Raviv, 1978). The two research areas are complementary where the positivist research identifies different contract alternatives and the principle-agent research determines which is the most effective given the level of risk aversion, information and uncertainty (Eisenhardt, 1989).

2.1.2 Management Control System

It is of vital importance to be in control of your organization. Failure of proper control by management, being out of control, may lead to financial losses, reputation damage or even bankruptcy. (Merchant & van der Stede p3, 2012)

The most simple form of MCS is the cybernetic form, whereby outcomes are compared with predefined standards. If there are any deviations management can take corrective actions to get the organization back on track towards the predefined standards of values. (Merchant & van der Stede p5, 2012)

This cybernetic form of MCSs is a reactive system, it acts on results or outcomes produced and measured. However such a system can also be proactive whereby it prevents the organization of going out of control rather than correcting an out of control state. Proactive MCS serve to encourage or enable employees to certain actions or prevent unwanted actions. This can be achieved by for example direct supervision, employee selection, segregation of duties and codes of conduct.

So a well-designed MCS is able to influence the behavior of employees and thereby aligning their goals with the goals of the organization. A proper functioning MCS increase the probability that an organization will achieve its goals. (Merchant & van der Stede, 2012)

According to Ansari (1977) researchers have two perspectives when focusing on the MCS design, the structural view and behavioral view. Structural view is the cybernetic approach where the emphasis lays on the information and communication role of the control system. The behavioral view has a focus on the human and social aspects of the control system, getting employees to align their goals with the firm’s goals and achieve performance goals. Often research is done in either of the two views, Ansari combines the two views to assess their interaction.

Two key roles of the MCS are identified in the literature, the decision-facilitating

role and the decision-influencing role. The facilitating role enables managers to improve their operations by providing them with information about current performance. The influencing role makes sure that both the employees and the managers act in the best interest of the company by aligning their goals with those of the company (van Veen-Dirks, 2010).

The same concept in described by Adler and Borys (1996) as the enabling and coercive role of the MCS. This is based on the degree and purpose of formalization and bureaucracy, the amount of written rules and procedures. The coercive view implies strict rules for employees how to perform certain tasks and is found to be related to a lower level of creativity, satisfaction and motivation. However the enabling view points out a underlying basis from which the employees can work on. In provides structure to the employees to do their tasks and appoint responsibilities and is related to a reduction of stress and increased feeling of effectiveness. These bureaucracies are the necessary to get the routine work of society done (Clawson, 1980), and are not just the most efficient way but the only efficient way to do this.

2.1.3 Performance measures

Merchant and van der Stede (2012) have made a categorization on the control measures, identifying 4 categories. The first one are the behavioral controls. The controls include guidelines, best practices and standardized processes. The second category are the result controls, such as goals or targets. The third category are the behavioral controls, which are amongst others prevailing norms and values as well as a code of conduct. The last category is the personnel controls, which is the selection of employees as well as training and education.

Budget control are a form of result controls as well as behavioral controls and appear to be used in knowledge intensive firms which are relative larger. Larger firms are often more decentralized which makes it suitable for the use of budget controls. For smaller firms however it is found that social controls lead to better results. In smaller centralized firms the social controls are more prevalent. Such social controls are amongst others employee selection, direct supervision, high degree of peer interaction and professionalized profession (Rockness and Shields, 1988)

2.1.4 Contingency theory

Contingency theory is all about 'fit', it departs from the thought that one size does not fit all. A definition of contingency is given by Morton & Hu (2008) as “any variable that moderates

the effect of an organizational characteristic on organizational performance”. So for every situation an individual solution has to be created. It is an answer to the prior believes that there is only one optimum in management control configuration, the universalistic view ( Hambrick & Lei, 1985). The opposite of the universalistic view is the situation-specific view which identifies every firm as a unique situation whereby an individual solution set has to be created and standards and models can’t be applied. The contingency theory is a compromise between the universalistic view and the situation-specific view, with tailored solutions and predefined packages. This implies that for every control problem in an organization a specific set of controls has to be selected to address this problem (Fisher 1998). By doing this the degree of fit increases which results in an increase in firm performance (King and Clarkson, 2015). King et al. (2015) found in their study that rather than a specific situation or setting, the variations in the use of formal control measures that matches to the specific situation to achieve 'fit' leads to an increase in firm performance. There are two different approaches to achieve 'fit' as explained by van de Ven and Drazin (1985). The first is the interaction approach which looks at how all the different contingencies interact with each other and assesses their effect on the other contingencies. The second is the system approach which takes the contingencies for granted as a whole, and examines the combined effect they have on a certain outcome like performance.

Donaldson (2006) stipulates that fit is a dynamic process, whereas prior research has formulated the contingency theory as a static process (Galunic and Eisenhardt, 1994; Woodward, 1965) where a company tries to reach an equilibrium to optimize performance. Donaldson however developed the SARFIT concept, Structural Adaption to Regain Fit, and is seen as a disequilibrium theory. SARFIT is illustrated in figure 1. The underlying idea is that ‘fit’ is only a temporarily status. As soon as fit is (nearly) achieved, performance will increase and along with this increase other organizational factors will change such as size or diversification (Hamilton and Shergill, 1992). The organizational changes in turn reduces the level of fit. So there should be a constant cycle of adaption to achieve fit, increased performance, changed company attributes and again adaption to regain fit.

2.1.5 Management control system tightness

According to Abernethy and Brownell (1997) tight controls are suitable for low complexity tasks. This finding is supported by the study of Daniel and Reitsperger (1991), where they found that in a production firm employees perform better if tight controls are in place. For the standard task such as production runs of uniform products or production with a distinct relation between input and output, main stream accounting techniques such as standard costing can be applied. In addition the control measures to coordinate the operations can be extensive to properly support the business. This can be in the form of procedure guidelines, job codification and tight regulations on reporting on accountability and behavior of employees. Such behavioral controls are preferred to accounting controls as the can be used for live-monitoring instead of backward looking evaluation (Merchant, 1985).

However such tight controls are less suitable for more complex tasks which require somewhat more room to deviate from the guidelines if necessary. As pointed out by both Abernethy and Brownell (1997) as well as Merchant (1985), control has to be achieved by other means. Using more strict selection criteria regarding education and training prior to hiring someone, can provide a company with the opportunity to use a more loose control system and rely on the knowledge and skill of the individual employee or individual team. They conclude from their analysis that the use of personnel control has a positive effect on employee performance, and a larger compared to behavior or accounting controls. More complex tasks, tasks with a large number of exceptions, are better controllable using behavior controls rather than accounting controls.

According to the agency theory described in paragraph 1.1.1 employees need some sort of rules and guidance to make sure that the goals of the firm as well as those of the employee are aligned. However mid and lower level employees often have local knowledge that the management does not have. Therefore it is better to give a certain amount of room for the lower levels in the organizations to operate to avoid the risk of missing out on critical and timely information. Even if a well functioning system is set up to avoid this problem, this introduces the next problem, namely the cost of collecting all information and presenting this in a proper way to top management (Nagar, 2002). So there is a tradeoff between a tight structure to maximize goal congruence and a more loose structure to fully benefit lower level knowledge. One of the possibilities to achieve this is described by Campbell, Epstein and Martinez-Jerez (2011), as the exception approach. It is a widely used technique in the design of management control. Using the exception approach a lower level employee is allowed to operate autonomous within certain limits, if the limits are exceeded the manager needs to review and approve certain actions. An example could be a bank employee who is granted permission to issue a loan up to a certain amount, exceeding this amount requires approval from the supervisor.

It is confirmed by Ho, Huang and Wu (2011) that tight controls relate to a higher chance an employee acts in the best interest of the company. Loose controls on the other hand limits the ability to proper monitor an employee and therefore lower the chance of goal congruence. The downside of tight controls is that is costly to implement and to maintain. For action controls to be assessed as tight controls they have to entail frequent en thorough evaluations, direct monitoring as well as significant rewards or punishments. For result controls, tight is defined as a direct link between achieved results and goals and again significant rewards. Tight controls such as frequent monitoring are found to be related to less inappropriate behavior by employees (Hunton, Mauldin & Wheeler, 2008). However the downside of the tight monitoring appeared to be that employees under direct supervision tend to be more risk averse. Especially when the incentive horizon is short employees tend to stay away from risky, but proven viable, investments.

2.1.6 Professional service firms

Two key aspects of a professional service firm is that its primary assets are the professional workforce and where the output, or product, are intangible services which are produced based on complex knowledge (Greenwood et al., 2005).

The PSF can be seen as “knowledge engines for business” (Lorsch and Tierney 2002), where for example the accounting firms serve to protect the credibility of the financial markets ( Levitt, 2000) and the law firms give guidance to complex business arrangements (Sharma, 1997). The PSFs account for the majority of the GDP in most economies in the world (World Bank). Even though this sector contributes a large share to the economy, only a small share of academic research is devoted to this sector, especially concerning the management control system designs in those organizations (Auzair, 2005). The authors from this study further stipulate four important differences between the more often studied manufacturing sector and the service sector: The core product, the service, is an intangible asset. Second the production and consumption occur both at the same time and the customer is hereby involved in the production process. Third the fact that a service cannot be stored or saved for later, it is perishable and lastly the service is an heterogeneous products, which differs between the persons supplying it and the customers consuming it.

As most of the research on management control systems in service firms is conducted in non-profit firms (Sharma, 2002) this study contributes to the literature as it focuses purely on the for-profit firms and excludes the non-profit firms.

2.1.7 Task complexity

Before we can assess the role of task complexity in the design of the MCS we must first define what task complexity is. Bonner (1994) describes this as either task difficulty, which points out the degree of mental processing needed for the task, or task structure, whether a task is well defined. So there is a difference in what makes a task to be classified as complex. The first suggests that the task to be performed itself is difficult and needs to be carried out by an individual or group of individuals with a high level of intelligence. The second suggests that it is not clear how the task should be carried out or what the outcome ought to be. In this case it may not be necessary for the person performing the task to have a high level of intelligence but for example needs to build on experiences from prior tasks.

A similar kind of distinction is made by Campbell (1988). In this review the distinction is made between three categories. First an isolated perceived complexity, or subjective complexity, from the task-doer point of view. There is however little research on this part, the results most related to this are found in the task design literature, where factors such as motivation or challenge come into play. But even that research is not purely based on subjective experiences of the task-doer. The second category is the person-task interaction complexity, a more widely researched area. Tasks are defined as complex relative to the capabilities of the individual performing the task. And the last category is the objective task complexity. There are four

attributes identified on which the degree of complexity can be measured; the situation in which there are more than one way to reach the targeted outcome, the possibility where there are multiple outcomes possible but it is not clear which is the best, conflicting interdependence between the multiple possible ways a variety of results and lastly if the connection between the different ways and results are not clear.

On the implications of the degree of task complexity it is found that a more complex task decreases the performance but that can be compensated by skill and experience (Bonner, 1994). Paquette (1988) found that a change in the degree of complexity should be accompanied by a change in strategy to maintain performance. In this study the results show that when tasks and decisions become more complex, shifting strategies to solve the tasks eliminates the drop in performance or accuracy. In other research the same routine of shifting strategies is found, when information density, in other words complexity, increases people's tendency to shift strategies in order to compensate for the increased time needed to process all information (Payne, 1976).

In the study of Abernethy and Brownell (1997) another key feature of a non-manufacturing organization is highlighted regarding the task complexity. They refer to the model of Perrow (1970), as displayed in figure 1, who distinguishes two dimensions of routineness. Whether there are predefined methods of performing tasks (analyzability) and the variations in the tasks (exceptions). These two dimensions are part of the task uncertainty and according to prior literature this task uncertainty makes accounting controls less usable.

2.1.8 External factors

There are several factors of influence on the design of a MCS. One of these is the organizational context. Another factor is the size of the company or department. Larger firms

tend to rely more or formal performance measures to overcome the more complex communication and coordination (Rockness and Shields, 1988).

2.2 Hypotheses

In this section we state the hypotheses. The hypotheses for this study are based on the prevailing literature. The following hypotheses are formulated for this study:

Hypothesis 1 (H1): A tighter MCS increases the employee performance and task performance

Employees are self-interested and want to pursue their own goals. To assure the employee operates in the best interest of the company and that they perform and fulfill the required tasks and duties a tighter MCS is required.

Hypothesis 2 (H2): A tighter MCS decreases the employee performance and task performance

The production of services by a professional service firm is a co-production with a client. Since every client is unique, so will be the tasks that need to be carried out. In addition there is often a professionalized workforce from which a company can benefit if it makes use of the unique knowledge and skills of the employees.

Hypothesis 3 (H3): Implicit tightness and explicit tightness are substitutes

As explained at hypothesis two the production of services requires some degree of freedom to adapt to each unique situation. Therefore there either have to more rules and procedures in place which are flexible or less rules and procedures which then can be more strict.

Hypothesis 4 (H4): There is a positive relationship between performance and tightness for low complexity tasks

Low complexity tasks are defined as task with little exceptions and a clear input-output relationship. This is relative easy to capture in rules and guidelines so processes can be optimized and efficiency increased.

Hypothesis 5 (H5): There is a negative relationship between performance and tightness for high complexity tasks

High complexity tasks have either a certain amount of exceptions, lacking a clear input-output relation or both. This makes it difficult to be captured in best practices and requires room to operate to adapt to unique situations.

3 Methodology

With this study we try to find answers to the research questions on the effects of control tightness on performance. To find these answers we use OLS regression in Stata. In this chapter we explain and define the variables and regression models used in this study. First the main dependent variables are explained followed by the independent and the control variables. Next a schematic overview is given outlining the composition of the model. We end this chapter by providing the regression models we use to find the results.

3.1 Main dependent variables

In this paragraph the main dependent variables are explained. The two dependent variables used are Employee performance and Task performance. In the upcoming two sub paragraphs the creation and composition of the variables is explained.

3.1.1 Employee performance

The first measure of performance used is the employee performance. Employee performance in this study is measured as the quality, quantity and accuracy of the job performance as well as the customer service provided. Data is gathered using the following questions:

 How would you rate your own job performance in term of the following:  Quality of work

 Quantity of work  Accuracy of work

 Customer service provided (in- and external)

Participants are asked to answer the questions based on a five point scale, ranging from 'needs much improvement' to 'excellent'. The average of the answers to those four questions is calculated and this forms the variable 'Employee Performance'.

These questions are selected to give an impression of the work performance of an employee. With this variable we measure the outcome of work that is done by the employees. This should give us an indication whether an employee benefits from more structure and guidelines, so a tighter system, to improve quality and efficiency. Or the opposite may be the case that because of the complex nature of the work some degree of freedom is required, so a looser system, to adapt to the circumstances to maximize output.

3.1.2 Task Performance

The second dependent variable is the task performance. With this variable we can measure to which extent an employee carries out the tasks in such a way as they believe they ought to do. The variable is based on the results of six statements:

 Imagine that you are in the role of your supervisor. How do you think your supervisor would rate:

 This employee always performs all tasks that are expected of him/her.  This employee always performs all essential duties.

 This employee always fulfills all responsibilities required by his/her job.  This employee always meets all formal performance requirements of the job.  This employee always completes all duties specified in his/her job description.  This employee never neglects aspects of the job that he/she is obligated to perform.

Again the participant are asked to provide an answer to these statements on a five point scale ranging from 'strongly disagree' to 'strongly agree'. The variable 'Task Performance' is computed by calculating the average of the answers to these questions.

The value we obtain by applying the afore mentioned method gives us an impression of whether their goals and work method is in line with what they believe is expected of them. We assume that the expectations are properly communicated to the employee via a written or oral job description and business meetings. Using the data from this variable we are able to investigate if there are specific parts of the control system that needs a certain degree of tight construction to assure proper guidance and supervision and thereby enabling the aligning of goals and priorities of the company with those of the employee

3.2 Independent Variables

In the previous paragraph we explained how and why we have constructed our dependent variables. In this section we will describe our independent variables from which we investigate if they have any effect on our dependent variables. In the first sub paragraph we describe the distinction that is made between implicit and explicit tightness in the control measures. In the second sub paragraph we describe how we have constructed our behavioral control, result control, personnel control and cultural control variables respectively. The statements are shown per category in table 2.

3.2.1 Explicit and implicit control tightness

As mentioned previously we have four main control categories which we will use as our independent variables. Each control category consists of two parts, namely the explicit tightness and the implicit tightness. Explicit tightness provides a measure of the coverage of the control system. The coverage is defined in term of number of measures in place and the frequency of evaluation on those measure. A high degree of explicit tightness results for example in a large number of guidelines, targets, social events and selection procedures for hiring new employees. Implicit tightness on the other hand gives a measure of how strict the controls in place should be applied or met. So it provides us information whether an employee has the opportunity to deviate from the pre established goals or rules if he/she thinks that would fit the situation better.

The degree of tightness in terms of explicit and implicit do not have to move together in the same direction. It may be the case that there are a lot of guidelines, so a high degree of explicit tightness, but that an employee may take its own course if he feels that the situation asks for it, so a low degree of implicit tightness. The other way around is also possible where there are only a few controls in place, for example only a profit target. This indicates a low degree of explicit tightness, but when this profit under no exception have to be met there is a high degree of implicit tightness.

A rapidly changing environment makes it perhaps difficult to use standardized procedures but the a company may very well demand to market at least 3 new products. So high implicit and low explicit tightness. Or it may be the case that there are standardized processes for production but the company faces big fluctuations in raw materials, so profit deviations are allowed. This indicates a high level of explicit tightness but low level of implicit tightness.

3.2.2 Control Measures

As we have already mentioned for each control category there are both explicit and implicit control tightness measures. Al of the statements that are used are shown in table 1. The statements are grouped into behavioral, result, personnel and cultural control measures. And each measures is again separated into explicit and implicit control tightness. The choice is made to include both the explicit as well as the implicit tightness of each measure separately. When both categories were taken together they seem to level each other out and thereby dampening the effects on our dependent variable. As the prevailing literature could not give a conclusive answer to whether use either of the two we decided to include both the implicit and the explicit measures.

Tabel 1 Statements per control tightness variable Explicit Behavioral Control Tightness

Whatever situation arises, we have existing processes, procedures or rules to follow in dealing with it. Established processes, procedures and rules cover all of my job tasks.

In my organization, we have rules for everything.

My supervisor frequently monitors the extent to which I follow established process, procedures and rules.

Implicit Behavioral Control Tightness

My job allows me to decide how to adjust rules to best perform my job tasks.

The organization I work in primarily uses established processes, procedures and rules to give broad guidelines as to how activities are to be performed.

Employees in my organization are encouraged to use procedures flexibly.

Employees in my organization are encouraged to adjust procedures to suit the situation.

Explicit Result Control Tightness

In my job, there is a performance measure for everything.

My organization sets a large number of performance goals/targets that I am expected to meet. Employee attainment of goals/targets is checked constantly.

My supervisor frequently checks to make sure that I am meeting my performance targets.

Implicit Result Control Tightness

In our organization, goals/targets are essentially a guideline rather than a true commitment.

My supervisor is very considerate of my explanations of deviations from pre-established goals/targets.

Responding to new, unforeseen opportunities is considered more important by my supervisor than achieving pre-established goals/targets. In my organization, employees are expected to meet pre-established goals/targets with no exceptions.

Explicit Personnel Control Tightness

The hiring process to become employed at my firm is extensive. You have to go through many steps in order to be hired at this firm.

I interviewed with several people in my organization before being offered a position.

The hiring process at my organization evaluates the knowledge, skills, abilities, values and motives of prospective employees.

Implicit Personnel Control Tightness

Before being hired, most of my colleagues and I acquired the same kind of job experience. There seems to be little consistency in the type of professional that gets hired for my job.

Before being hired, most of my colleagues and I followed the same type of education and training. The competence of employees within my job title varies greatly

Explicit Cultural Control Tightness

I socialize with my colleagues outside of work.

Since starting this job, my personal values and those of this organization have become more similar. I am not friends with any of my colleagues. REV

I feel a sense of “ownership” for this organization rather than just being an employee.

Implicit Personnel Control Tightness

My organization regularly hosts social events for employees. My organization communicates its core values to employees. My organization plans team-building events for employees.

My organization creates company sponsored teams for sporting events/fundraisers/volunteer events.

For each of the four control categories there are eight questions or statements which are evenly divided between explicit and implicit tightness. For each statement the participant was asked to give an answer on a five point scale ranging from 'strongly disagree' to 'strongly agree'. We then took the average of the four statements per measure to create the corresponding variable. Most statements are put in such a way that a higher score implies a higher degree of tightness. For the statements that are put in the opposite way, a higher score implies lower tightness, the results are reversed coded before taking the average of the four statements.

3.3 Control Variables

In this paragraph we discuss the creation of the control variables as well as the motivation to include them into our regression models. We use six control variables which are explained in the following corresponding sub paragraphs

3.3.1 Size

It became clear from the current knowledge in the management control field that the size of the company has an impact on the control measures that are selected. Larger companies tend to be more decentralized and have a more extensive control system with more formal controls compared to the relative smaller companies. Therefore we include this variable in our regression. The data is obtained from the participant by asking the following question:

 How many people are employed by your entire company?

The answers resulted in four categories with a range of the number of employees 3.3.2 Experience

The experience variable asks for the years of experience in the field of the participant and is included for two reasons. The first is that we have found that experience may compensate loopholes in the control system or it affects the choice of control tightness. A more experienced worker is expected to make the right choices and may therefore be under a less tight control system. The second is that it takes some time to fully understand how an organization works and how management is imposed on an employee. To prevent the effect that an inexperienced employee may not fully grasp which control measures are in place and provide (partly) incorrect data we include the experience as a control variable. The question asked to obtain the data:  How many years have you worked in the current field?

Answers were given as integers.

3.3.3 Market Competition

This variable includes the environment which the company is in and the forces which it is exposed to. If market competition is high a company may need to acts fast in certain cases or move beyond the limits of discounts to attain new clients. This could indicate that such a company would choose for a lower degree of either implicit or explicit tightness. The question asked was how fierce the competition is regarding:

 Price competition

Answers were given on a five point scale ranging from 'of negligible intensity' to 'extremely intense'. The average of both items was taken to obtain the competition variable.

3.3.4 Market Uncertainty

The market uncertainty assesses the predictability of the industry. If the industry is very predictable standardized processes and best practices may dominate the control system as a company have learned from past experiences. To control for this factor we include this as our control variable. The question is asked how predictable is your industry?

 How would you classify the market activities of other firms in the industry?  How would you describe the tastes and preferences of your clients?

Answers were provided on a five point scale scoring from 'very predictable' to 'very unpredictable'. The average of the answers to point items gives the uncertainty variable.

3.3.5 Relative company performance

We want to investigate the effects of the different degrees of control tightness on employee and task performance. We have seen that company characteristics may alter the decisions made in the design of the MCS and the degree of tightness. To control for this fact we include the relative company performance. It is a broad measure, where participant are ask to rank their firm's performance relative to competitors. We created an average score of relative performance to correct for any these differences in company attributes.

Participant were asked how their company relates to competitors regarding five attributes.  More competitive

 Greater market share  Growing faster  More profitable  More innovative

Answers were given on a five point scale ranging from 'strongly disagree' to 'strongly agree'. The average of the five items is calculated to form the relative performance variable.

3.3.6 Employee reputation

The last control variable is the employee reputation. Participants were asked to give a rating on a five point scale to what extent the following items are applicable to their workforce:  Our employees are highly skilled.

 Our employees are widely considered the best in our industry.  Our employees are creative and bright.

 Our employees are experts in their particular jobs and functions.  Our employees develop new ideas and knowledge.

The average of these five aspects formed our employee reputation control variable. If employees are considered to be the best in class and experts in their fields, control tightness is likely to be lower as managers rely on their experience and knowledge. The same holds for employees who are creates and develop new ideas. Less (strict) procedures and guidelines are expected to be in place to allow employees to fully utilize their unique capabilities. We control for this aspect by including the employee variable.

3.4 Task Complexity

In this paragraph the construction and use of the task complexity in this study is explained. We describe on which data the variable is based, the distinction which is made between high and low complexity and how this variable is used in the models.

The data is gathered by posing statements about two aspects of the complexity of tasks. The first aspect is the predictability of the tasks that are carried out. The following statement were asked to be rated on a five point scale:

 I can easily determine whether I have performed my work correctly.  Most of the time, I know what to outcome of my work efforts will be.  I would describe my work as routine.

The data gathered from the answers given to the statements has been reversed coded so that a higher score implies a higher degree of task complexity.

The second aspect of task complexity is about the exceptions encountered in performing tasks. Again three statements were opposed and requested to rate these on a five point scale.

 I often encounter problems in my work for which there are no immediate or apparent solutions.

 In my work, I spend a lot of time solving difficult problems with no immediate solutions.  The situations, problems and issues that I encounter in performing my major tasks are usually

the same.

The answers to the last statement are reversed coded so that again for these statements a higher score implies a larger degree of task complexity.

The scores of all six statements are averaged and this formed the variable task complexity.

This variable will not be used directly into the regression model, but instead it is used to divide the sample into two groups. The mean of the variable is computed and taken as a cut of point for

the degree of complexity in the tasks. Every participant rating its tasks at a value below the mean of the sample is assigned to the low complexity task group. Ratings of the mean value or above are assigned to the high complexity task group. This gave us an almost even distribution of participants between the two groups

Regressions are then executed using either the data from the low complexity group or from the high complexity group. We use the results that we obtain by using this method to compare the two groups and investigate if there is any noticeable difference between these groups regarding the use of different control tightness measures and their effects on employee and task performance.

3.5 Schematic overview

We have now explained all the variables that are included in the regression models. In this section we give a small recap of the aim of the study. Figure 2 gives an overview of how the variables relate to each other and where the hypothesis as stated in chapter 2.2 are situated.

Figuur 2 Overview study

3.6 Regression models

In this paragraph we outline the regression models we will use in this research to find the answers to our research question. We start off with a univariate analysis where we assess the direct effects of the independent and control variables on employee performance:

Model 1 : Univariate Regression

(1) (2)

In this first section we insert the eight control tightness variables into the regression (1) and also the control variables (2).

We then move to the multivariate regression where we include all the control variables into one regression model and test the effects of every control tightness measures separately on the employee performance. We use the following model (3).

Model 2 : Regression 1 - 8

(3)

For Control Measure we use the independent variables as described in paragraph 2.2 of this chapter. This model is used for regression 1 to 8.

In the second part we investigate the effects of the control measures on employee task performance. We start with a univariate analysis using to model below.

Model 3 : Univariate Regression

(4) (5)

In the first regression (4) we insert the eight different control tightness measures. In the second (5) regression we include the six control variables.

In the next part we use the multivariate regression and include, next to the separate independent variables all six control variables. We use model 4 to get the results

Model 4 : Regression 9 - 16

(6)

As with model 2 we include the control tightness measures one by one. For regression 9 - 12 we include the explicit tightness measures and the implicit tightness measures are included in regression 13 - 16.

In the last part of our results we redo the regressions, so we apply the same six equations as described above. Only now we first separate our sample into the low and high task complexity group and then rerun the regressions.

For all four models, so the six equations, we expect the error term to be normally distributed with a mean of zero and a standard deviation of one.

4 Data

4.1 Part of the research project

The data which is used for this research is generated using a survey. As part of the PSF Thesis Survey Project 2015 - 2016, a project initiated by the University of Amsterdam, this study is able to use the unique data. Several students joint this project and each student contributed to the joint database. Participating students used their network to acquire responses and to further contribute to the database which had been set up by the project leader. Since this data is not publicly available the results will be unique and a contribution to the current prevailing literature. Even though the data is shared with a limited number of students, everyone focuses on its own area and thereby prevent results being similar or deriving the same conclusions.

The research project is part of the Accounting section of the University of Amsterdam. The project focuses on the design of the management accounting and control systems and their effects on the employees. As mentioned before many studies have looked into this part of an organization within manufacturing companies, however little research is devoted to this system within the professional service firms.

To be eligible for filling in the questionnaire several requirements are set up to ensure the quality of the data. The first and most important requirement is that the participant must work within a service firm, where the core business is the "production" of services. Organizations operating in the non-profit sector, public (government) organizations and NGOs are excluded from this research as their corporate structure and goals are assumed to be organized in a different way. The project aims at medium to large firms, so the organization where the participant is employed should have at least 50 employees. This limit is set to ensure that there is some sort of MCS in place. In smaller organizations the need for a MCS may be absent and guidance is provided for example in a more informal way between managers and employees. In addition the participant should be in a mid-level function in the organization, where it is subject to a MCS but does not control or define the system. The last requirement for inclusion in the population of interest the participant has been with the firm for at least three years. This is to make sure that on the one hand the participant understands the questionnaire and on the other hand is completely familiar with its function in the field. Starters in the field of knowledge intensive jobs require a lot of guidance in their work and therefore are imposed to a tighter control system compared to move experienced employees.

4.2 The questionnaire

As mentioned the dataset is composed by participant filling in a questionnaire. It is an extensive list of questions and statements about their firm, performance, the organization of management and control and the role and experiences of the participant within the firm. The questionnaire is composed by the researcher leading the project and the structure is partly based on classifications used in prior literature and partly on own developed ideas and categories derived from previous research. Participants are asked to place them, their performance or attributes of their firm into a certain range or pick attributes that are applicable, or provide answers to statements on a 5 point scale to which extent they agree on a statement.

There are six main areas identified in the questionnaire. The first area are the antecedents. This consists of five subcategories. The first subcategory is the customer reliance, describing the degree of customer involvement in the production of the service. The next two subcategories are the (human) capital intensity, the amount of physical capital required to carry out the tasks. And the amount of human capital, so education and experience of the employee, necessary for providing the services. The fourth subcategory is the professionalized workforce, going into the organization of the profession as a whole. The last subcategory is task complexity, the participant is asked to describe predictability and variety in their tasks.

The second area is the control system as described by Merchant and van der Stede (1982). Four control categories are identified in this model. Behavioral control which entails the standardized processes and procedures as well as rules and routines. The second is the result controls, to which degree the task are measured and incentivized best on the results of the tasks and the achievement of predefined goals. The third category is personnel control focusing on employee selection. The last category is the cultural controls, which includes amongst others the prevailing norms and values the employee is subjected to and the rituals how employees are evaluated regarding the alignment of personal and organizational goals.

Next is the area of control tightness. This gives information on the flexibility in general of the MCS in place. It also focuses on the explicit tightness, so whether it is an extensive or limited control system. And lastly the implicit tightness which indicates if deviations from the predefined controls is allowed.

The fourth area combines the previous two areas where each of the four control categories from Merchant and van der Stede (1982) are split into explicit and implicit attributes.

The fifth area provides information on the consequences of the control system in place, described in the first four areas. Participants are asked to give information on the level of tension and stress they experience as a result of the MCS in place. They provide information on whether

they feel the control system enables them in performing well or if it limits them from functioning properly. The other aspect is performance, both the individual performance as well as the performance of the organizational unit as a whole.

The last area provides the dataset with control variables. The variables used are the uncertainty in the industry in which the firm operates, the strategy of the firm ( orientated on cost or differentiation). The reputation and size of the firm, and lastly the compensation and organizational structure.

4.3 Data Summary

After the closing date of the survey the dataset included 372 entries. After examining the data 107 entries appeared to be not completed, (partly) empty or corrupt. After deleting those entries 265 entries are available for our analysis. There are 159 questions answered or statements ranked. The survey is reasonably large which has the advantage that it provides a lot of data of different areas of management control. There is however also a downside of this aspect as we have received comments after completion that is maybe was a bit too long. Since the survey is quite extensive there is a risk that a participant answers some parts with less attention and give a range of statements the same ranking. This could lead to false data. To prevent this some statement are randomly phrased backwards. These answers were reversed coded after completion before using this in our regression models.

Of all participant in this survey 197 are male and 127 are female. The youngest participant is 19 and the oldest participant is 63 years old. We have 27 nationalities in our sample operating in 21 different countries.

Data on the dependent and independent variable is summarized in table 1.

Table 2 Summary statistics main variables

Variable # of observations Mean SD Min Max

Behavior explicit 265 3.09 0.839 1 5 Behavior implicit 265 2.64 0.689 1 5 Results explicit 265 2.86 0.845 1 5 Results implicit 265 2.79 0.546 1.25 4.25 Personnel explicit 265 3.31 0.82 1 5 Personnel implicit 265 3.06 0.72 1 5 Culture explicit 265 3.56 0.90 1 5 Culture implicit 265 3.52 0.76 1.25 5

Employee performance 265 3.96 0.56 2 5

Task performance 265 4.16 0.64 1 5

Task complexity 265 2.93 0.61 1 4.8

What we see is that none of the participants ranked its own performance with a 1, results range from 2 to 5. Also the mean of both performance measures is for both variables almost 1 point above the average score of 3 that could be given. The mean on the performance measure tightness are all around the expected average of 3, only the cultural performance measures tend to be on average tighter.

The variables are the average values for the corresponding area of interest, that explains why there are decimal score in the min and max value.

5 Results

In chapter four we gave a description of our data sample, and in chapter three we explained how we were going to use this data to obtain the results. In this chapter the results are shown from the regressions that are performed. The first paragraph examines the effects of the control measures on employee performance. Followed by the paragraph describing the results of regression of the control measures on task performance. The last paragraph reruns the regressions from paragraph 5.1 and 5.1, only after dividing the sample into two groups, based on the degree of task complexity encountered.

5.1 Effects on own performance

In this paragraph the effects of the different control measures on task performance are shown. The first sub paragraph shows the univariate analysis, followed by a multivariate analysis. 5.1.1 Univariate analysis

We start the analysis with an univariate regression to evaluate the effects of the different control measures as well as the control variables on employee performance. The results are shown in table 3

Tabel 3 Effects on employee performance

Variable Coefficient t-value

Behavior explicit 0.061 1.54 Behavior implicit -0.251 -5.38*** Results explicit 0.009 0.23 Results implicit -0.120 1.95* Personnel explicit 0.103 2.54** Personnel implicit -0.006 -0.13 Culture explicit 0.150 4.12*** Culture implicit 0.179 4.15*** Control variable Company size 0.060 1.94* Experience 0.016 1.51 Market competition 0.042 1.21 Market predictability -0.026 -0.61 Competitiveness 0.047 1.08 Employee reputation 0.222 4.83***

*=significant at 10% level **Significant at 5% level ***=Significant at 1% level

Based on the results from the univariate analysis there are some significant effects of the degree of control tightness on employee performance. The explicit behavioral and result controls as well as the implicit personnel controls appear to be insignificant in this regression.

The implicit behavioral and results controls however have significant effect (respectively at the 1% and 10% level) on employee performance. For both control measures we find a negative coefficient so an increase in tightness would result in a decrease in performance. This would indicate that to achieve a higher level of employee performance a looser set of controls is required. The opposite is the case for the explicit personnel controls and both of the cultural controls. We find significant results for the explicit personnel controls (at 5% level) and for both cultural control categories (at the 1% level). For these three control categories a more tight system is expected to deliver a higher level of performance.

Regarding the effects of the control variables; four out of the six are insignificant in this regression. The two variables that are statistical relevant are company size and employee reputation. An increase in company size is expected to result in a higher employee performance. Looking at the employee reputation we find a highly significant result, where an increase in reputation leads to an increase in employee performance.

5.1.2 Multivariate analysis

In the univariate analysis we have seen that there are some control measures that deliver a significant contribution to the performance of the employee. In this section we will include the control variables into the regression to assess if the previous found results will still hold.

Furthermore we will examine if the insignificance of the explicit behavior and result controls, and implicit personnel controls are changed when inserting the control variables into the regression. The results of the multivariate analysis is shown in table 4. The first four regressions examines the effects of the explicit form of the four control categories. Regressions 5 to 8 looks at the effect of the implicit form of controls.

When we look at the results of the explicit form of controls we find no significant results in this multivariate regression. This means that nothing has changed for the behavioral and results control measures. The significant levels of the personnel and culture controls in the univariate analysis were substantial but their effect disappears when the control variables are inserted. We now see that both control measures become insignificant. This indicates that the previously found results for these two measures were caused by omitted variables and not by the control measures themselves.

A different situation arises when we look at the implicit measures of the four control categories. Here we find that two of the four categories exhibit a significant effect on employee performance. The result controls had a weak significant effect in the univariate analysis but this disappears when the control variables are inserted. We find no change regarding the personnel controls.

Tabel 4 Results of multivariate analysis on employee peformance

Dependent variable: Employee performance

Regression 1 2 3 4 5 6 7 8

Independent variable Explicit Implicit

Behavioral controls -0.043 (1.06) -0.214 (-4.47)*** Result controls -0.002 (-0.05) -0.053 (-0.87) Personnel controls 0.020 (0.44) -0.072 (-1.54) Cultural controls 0.065 (1.53) 0.124 (2.71)*** Control variables Company size 0.366 (1.18) 0.040 (1.28) 0.038 (1.22) 0.032 (1.04) 0.054 (1.83)* 0.040 (1.31) 0.044 (1.42) 0.033 (1.09) Experience 0.155 (1.42) 0.015 (1.32) 0.015 (1.38) 0.016 (1.48) 0.011 (1.09) 0.014 (1.30) 0.166 (1.52) 0.016 (1.46) Market competition 0.037 (1.05) 0.037 (1.04) 0.037 (1.05) 0.029 (0.81) 0.044 (1.30) 0.036 (1.01) 0.033 (0.93) 0.025 (0.72) Market predictability -0.038 (-0.90) -0.042 (-1.00) -0.042 (-0.99) -0.036 (-0.84) -0.049 (-1.19) -0.040 (-0.94) -0.043 (-1.02) -0.031 (-0.73) Competitiveness -0.016 (-0.35) -0.008 (-0.18) -0.010 (-0.22) -0.013 (-0.30) -0.015 (-0.35) -0.006 (-0.12) -0.004 (-0.08) -0.013 (-0.30) Employee reputation 0.206 (4.24)* ** 0.212 (4.34)*** 0.202 (3.81)*** 0.178 (3.35)*** 0.153 (3.15)*** 0.204 (4.16)*** 0.226 (4.60)*** 0.173 (3.47)*** Statistics # Observations 265 265 265 265 265 265 265 265 R2 _{0.101 0.098 0.098 0.106 0.163 0.100 0.106 0.123} *=significant at 10% level **Significant at 5% level ***=Significant at 1% level

What we do find however are significant effects of the behavioral as well as the cultural control measures. Both were highly significant in the univariate analysis and after adding the control variables into the regression we find that these results still hold. For the behavioral control measures we find a negative coefficient of 0.214 (1% level). This points out that an increase in implicit control tightness results in an expected decrease of employee performance. The opposite is the case for the cultural controls. Here we find a positive coefficient of 0.124 (1% level). This coefficient points at an expected average performance increase of 0.124 point on a 5 point scale if the average implicit cultural control tightness increases with one point on a 5 point scale.

When looking into the control variables we still whiteness that four of the six variables are insignificant in all regressions. The company size has only a significant effect in regression 5.