Building on the seminal paper of Adler and Borys (1996), this paper contributes to the management control literature by investigating the relationship between formal controls and employee performance. Using a survey among 517 employee-manager pairs, we find that enabling control as experienced by the employee is positively related to the employee’s performance as assessed by their manager, while a high level of control extensiveness is negatively related. The relation between enabling control and employee performance is more positive for uncertain tasks, but a preference for enabling control does not moderate the relation between perceived enabling control and employee performance. Furthermore, the relation between the extensive use of controls and employee performance is less negative when employee tasks are uncertain, and when employees prefer more extensive controls.
The positive relation between perceived enabling control and employee performance is in line with the theory of Adler and Borys (1996), who suggest that enabling control will generally have positive effects. This finding resembles prior research findings at the managerial level focusing on how enabling performance measures and budgets supported managerial performance. We believe that finding a positive effect at the employee level is relevant in itself due to the differences in employee versus managerial tasks. Additionally, since we assess the extent of enabling control at the operational level, we can evaluate whether this relationship is different under contingencies with respect to the task and the employee.
Specifically, the positive relation between enabling control perceptions and employee performance is stronger for employees with uncertain tasks, which explains earlier positive findings among managers (e. g., Chapman and Kihn, 2009; Mahama and Cheng, 2013) who typically have more uncertain tasks (e.g., Hall, 2010; Mintzberg, 1971). The discretion resulting from enabling control allows employees to respond more effectively to any contingencies that are not covered by the controls.
This interpretation is further supported by the results of the additional analysis using the analyzability and repetitiveness components of task uncertainty: enabling control has a more positive impact when the analyzability of the task is low. In this case, it is more difficult to design good controls, and dealing with the contingencies is more likely to require deviating from the controls. Task repetitiveness does not have a moderating effect, which fits with the interpretation that variety as such does not impact the possibility of designing good controls.
At the same time, the type of controls employees prefer does not result in a different relationship between enabling control and performance. This suggests that using controls in an enabling way has limited downsides, in that it does not result in lower performance for employees who are less keen on the freedom and initiative involved in enabling control. There may be other moderators at the personal level that may affect the relationship between enabling control and work outcomes. For example, Burney et al. (2017) find that the relationship between enabling control and counterproductive work behavior is negative, which is a positive outcome for the organization, but for employees who perceive their co-workers to be less committed, the relationship is less negative. This suggests that the relationship with team members or co-workers in general may be important.
We find that control extensiveness is negatively related to employee performance. This differs from Adler and Borys’s (1996) theorization of this relationship. They expect that a negative relationship between control extensiveness and performance occurs because control extensiveness often coincides with a coercive approach to controls. Consequently, when statistically controlling for the control type, the negative effects should disappear. However, the bivariate correlation between type and extensiveness of control is low (see Table 4), and the regression results show that control extensiveness has a negative relationship with performance also when control type is included in the regression. This suggests that the increased control burden and the reduced autonomy and meaningfulness prevail over any positive effects from reduced ambiguity and increased goal clarity.
Our moderation analyses provide more insight into this result. The relation between extensiveness and performance is more negative for routine tasks. When tasks are more routine, controls are less necessary and less helpful, while still requiring time and attention that cannot be dedicated to the task at hand. Furthermore, the restrictions in autonomy from more extensive controls likely result in a downward pressure on motivation. For uncertain tasks, this is compensated by the extra guidance given by the controls, but for routine tasks this guidance is not needed, so it does not offset the reduction in autonomy. Thus, for routine tasks, the motivational channel seems to outweigh the operational one. 
We find support for this interpretation when we split up task uncertainty into its components analyzability and repetitiveness. The relationship between control extensiveness and performance is more negative when task repetitiveness is high, but there is no moderating effect of task analyzability. When the routineness is mostly the result of high repetition, employees gain experience with their tasks in their daily activities. Thus, the supportive role of controls is less needed, while an increase in control extensiveness will reduce the room for employees to create some variety in their daily tasks, leading to feelings of reduced autonomy. When the routineness stems from analyzability, this means that the activities are well understood and the controls are more likely to be well designed. If this is the case, an increase in control extensiveness will then match the tasks at hand, and consequently is less likely to be experienced as reducing autonomy. The moderating effect of preference for extensive control further supports the importance of the motivational aspect of control: the relation between control extensiveness and performance is less negative for employees who prefer a high level of control extensiveness (Fig. 4). An important avenue for future research here is to investigate whether the negative relationship of control extensiveness with performance is affected by the extent to which the controls fit the ‘technically required’ formalization (Adler and Borys, 1996, p. 77).
We use a survey to collect data, because this allows us to investigate contingencies that may moderate the distinctive relations of perceptions of control type and extensiveness with employee performance in a real life setting, with a large sample of various types of operational employees. As such, it enables us to contribute to the limited body of empirical results regarding the relationship between enabling control and performance. Furthermore, the survey method makes it possible to explore the relation between control extensiveness and employee performance, while controlling for the control type. A downside of the survey method is that it cannot be used to investigate causality (Chapman and Kihn, 2009; Hall, 2008): it may be the case that managers choose to use controls in a more enabling way when they feel employees are better at their jobs, or that managers use more extensive controls when employees are worse performers. However, reverse causality does not affect the theorizing behind the moderating effects. For example, the relationship between enabling control and performance is more positive for uncertain tasks, because good performers can benefit more from the freedom offered by enabling control in such tasks. For routine tasks, providing more freedom has a less positive effect, because for those tasks there are fewer exceptions or contingencies where good performers can make a difference.
We strengthened our research design by surveying dyads of employees and their managers. This allowed us to measure all variables at the relevant source, and to rule out common source bias by measuring the independent and dependent variables at different sources. To obtain respondents, we used a non-random sample generated from networking via students who participated in this project to finish their master’s thesis. According to the meta-analytic study of Derfuss (2009), the results of studies with such a non-random sample are comparable to studies with random samples. A downside of using dyads is that responses cannot be fully anonymous, since otherwise the responses cannot be linked. This may bias the sample toward better performing employees.10 However, when potential respondents signed up with the students to receive the survey, they had no knowledge of the survey itself beyond a general description. Given the high response rate among the potential respondents, we believe the selection bias is limited. Furthermore, we investigated within-sample patterns, rather than absolute levels of sample characteristics.
To be able to study control perceptions of operational employees, we developed a new survey instrument. Thus far, most accounting survey studies on this topic focused on enabling results controls (Chapman and Kihn, 2009; Hartmann and Maas, 2011; Mahama and Cheng, 2013; Wouters and Wilderom, 2008). To make the instrument more applicable to a diverse sample of operational employees and to better align with Adler and Borys (1996), our instrument measures both action and results controls. While reliability and validity analyses show acceptable to good outcomes, especially for the combination of action and results controls, the measurement of enabling results controls in particular shows room for improvement. This may be due to the fact that rules and procedures involved in action controls are likely intuitively understandable for most employees, while results controls may be more ambiguous. They cannot always be easily matched to individual employees when outputs are created through teamwork. It is also more difficult to observe results control enforcement than action control enforcement, especially when the results are assessed over longer periods (monthly or yearly).
Finally, our theory and instrument are focused on formal controls and do not include informal controls such as cultural and personnel controls (Merchant and Van der Stede, 2017; Pfister and Lukka, 2019). For this study, we chose to focus on the total of formal controls, as we were interested in testing the theorizing of Adler and Borys (1996). Including informal controls offers an interesting avenue for future research, for example because the strength of the relation between the different types of formal controls and performance has been shown to be dependent on the use of informal controls (Tiwana, 2010).
