The relationship between lack of controllability and proactive work behaviour: an empirical analysis of competing theoretical explanations

Abstract

The controllability principle suggests evaluating managers solely based on performance measures they can control. In practice, however, companies often disregard this principle. Therefore, our study addresses organisational benefits linked to the lack of controllability in measures used for managers’ performance evaluations. We draw on important case-based findings to establish a positive ‘base relationship’ between lack of controllability and proactive work behaviour. We test this base relationship with a large-scale sample and find that companies encourage higher levels of proactive work behaviour when they rely on less controllable performance measures. Drawing on recent developments in role theory, we advance previous research and extend the base model by including the theoretical construct of flexible role orientation. We examine different mechanisms through which flexible role orientation potentially impacts the base model. Using survey responses from 432 managers, we find evidence for a mediation model as opposed to an interaction model. Specifically, we find that lack of controllability enhances role conflict, which in turn induces more flexible role orientations ultimately resulting in higher levels of proactive work behaviour.

Keywords:

• controllability principle
• management control systems
• role theory
• role conflict
• flexible role orientation
• proactive work behaviour

Acknowledgements

The paper has benefited from comments made by Rainer Lueg, Martin Messner, and Felicitas Morhart. We especially thank the associate editor Sven Modell and two anonymous reviewers for their excellent comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See also Parker and Collins (2010), Wu and Parker (2011), and Strauss and Parker (2015).

2. The results in this article are based on the same dataset as the results in Burkert et al. (2011).

3. Our study also responds to criticism that behavioural accounting is ‘made up of many small, one shot-research projects’ (Burgstahler and Sundem 1989, p. 86), endangering the field to accept results from a single study or a limited number of observations as generalisable findings (Nelder 1986).

4. McNally (1980) points out differences between formal control and informal control. Formal control reflects formal authority and typically supports a binary division of control (i.e. either no control or full control). Informal control reflects a manager’s ability to influence outcomes independent of formal authority. McNally (1980, p. 168) critically elaborates on a binary division of control and points out that control should be understood ‘being a point on a continuum from full control to no control’. Our understanding of control is in line with McNally’s (1980) understanding and reflects more recent definitions of control in the accounting literature (e.g. Merchant and Otley 2006).

5. See Giraud et al. (2008), Burkert et al. (2011), and Merchant and Van der Stede (2012, chapter 9) for detailed literature reviews.

6. Our line of reasoning does not change depending on whether the use of less controllable performance measures is a clearly deliberate MCS design choice or an allegedly inevitable choice due to a lack of better measures. We argue that organisations have the option to completely abstain from holding managers accountable for any performance measure. In this regard, the use of performance measures is always a deliberate MCS design choice.

7. On the other hand, theory and emerging empirical evidence document that disregarding the controllability principle has also dysfunctional effects. Specifically, Burkert et al. (2011) find that disregarding the controllability principle evokes role ambiguity, which in turn results in impaired managerial performance. By focusing on the beneficial effects of disregarding the controllability principle, our study is in line with Burkert et al.’s (2011) conclusion that companies should use personnel controls to cope with the dysfunctional effects rather than obeying the principle.

8. Bettencourt et al. (2005) refer to similar behaviours as ‘internal influence behaviour’. Bateman and Organ (1983) have coined the related term ‘organisational citizenship behaviours’ (OCB). The operationalisation of OCB in the literature has focused on items that transport ideas of general compliance and (pure) altruism (Williams and Anderson 1991). We depart from OCB, as its operationalisation is unrelated to the behaviours that we are interested in (i.e. enhanced communication and cooperation with others to solve organisational issues). Complying with a rule or norm is unlikely to reflect proactive behaviour, but captures a more passive behaviour constraining rather than enhancing proactive communication and cooperation. In addition, the idea of altruism in OCB focuses on helping others for the sake of good rather than for the sake of performance. The communication and cooperation that we refer to, however, is directed towards an optimisation of the organisation and allows for self-interested motives. Moreover, our construct also captures the idea of encouraging others to engage in similar behaviours, while OCB does not reflect such a proactive stimulation. Therefore, we differentiate our operationalisation of proactive work behaviour from other constructs in the literature.

9. Accordingly, we regard proactive work behaviour as a functional and desirable outcome.

10. There is little large-scale empirical research providing evidence ‘whether’ formal MCS can elicit proactive work behaviour (Burney et al. 2009).

11. Recent examples for its use in management accounting research are Hall (2008), Maas and Matejka (2009), Marginson and Bui (2009), and Marginson et al. (2014).

12. Dorenbosch et al. (2005, p. 129) label this cognitive state as ‘feelings of ownership for work-related issues’.

13. Similar to (flexible) role orientation and the notion of responsibility for a company’s goal achievement, the literature has investigated concepts such as ‘psychological empowerment’ (e.g. Spreitzer 1995, 1996) and ‘structural empowerment’ (e.g. Ahearne et al. 2005). Psychological empowerment focuses on individuals’ perceptions about (a) the importance of their jobs for themselves (meaning), (b) their capabilities to succeed in their jobs (competence), (c) their autonomy in deciding how to do their jobs (self-determination), and (d) the wider impact of their jobs (Spreitzer 1995). Structural empowerment, on the other hand, focuses on the extent to which managers help their employees to feel psychologically empowered within their job (Ahearne et al. 2005). Although these concepts address individuals’ responsibility for overall goal achievement, they do not capture individuals’ responsibility for factors outside their designated area of control. As we are interested in factors, over which managers have only little control, we concentrate on the concept of ‘flexible role orientation’ stressing individuals’ perceived responsibility for factors outside their designated area of responsibility.

14. Although Burkert et al. (2011) find that a lack of controllability is associated with role ambiguity and role conflict, they emphasise that companies may accept these consequences in order to encourage managers to address organisational issues outside their designated area of responsibility.

15. We test in Section 5 whether the effect between the use of less controllable performance measures and flexible role orientation is direct or indirect in nature (i.e. additionally mediated by role stress).

16. Prior research indicates that senior managers have a higher span of accountability (Simons 2010) causing higher perceived responsibility for the overall organisation. Therefore, they can be expected to exhibit more proactive work behaviour than lower-level managers. RBSE is the extent to which individuals perceive their own personal capabilities to be high (Parker 1998). In other words, managers who are convinced about their own capabilities can be expected to exhibit more proactive work behaviour (see Parker 2007). In addition, to these reported control variables, we also conducted further tests including control variables such as (i) gender, (ii) company size, (iii) age, (iv) tenure, (v) distributional justice, (vi) job satisfaction, and (vii) trust in superior that we do not include in the analysis reported in Section 5. The inclusion of these control variables neither changes the direction nor the significance of any of the reported effects.

17. Psychological empowerment tries to identify individuals’ perceptions about job meaning for the self and impact within their own work area (Spreitzer 1995). Moreover, ‘psychological empowerment’ captures individuals’ perceptions about their skills to succeed in their jobs and the extent to which they are autonomous in deciding how to do their jobs. These aspects of the construct refer to personal capabilities and the degree to which people are able to determine their work procedures rather than a feeling of responsibility. ‘Structural empowerment’ focuses on the extent to which managers help their employees to feel psychologically empowered within their job (Ahearne et al. 2005).

18. In particular, OCB has been operationalised with two constructs (Williams and Anderson 1991): OCB that benefits specific individuals (OCBI) and OCB that benefits the organisation (OCBO). Unlike OCBI, our construct of proactive work behaviour not only targets specific individuals in an organisation, but the organisation as a whole. While OCBO is targeted at the whole organisation, it mainly focuses on rule compliance, including aspects such as absenteeism or break-taking, and (pure) altruism, not necessarily accounting for self-interested motives.

19. Investigating self-efficacy in more detail, we find that dropping one item (i.e. item 3) would improve all levels of fit from acceptable to good. However, we rely on the originally developed construct and verify that our results (mediation versus interaction) do not change when dropping this item. In order to account for potential problems with the self-efficacy measure in SEM, we remove two restrictions from the measure in the CFA and in subsequent SEM. In particular, we allow the error variances of two pairs of self-efficacy items to vary instead of holding them fixed at zero (Kline 2005). We do this for correlations between the error variances of items 1 and 2 as well as for those of items 5 and 6. The correlation between error variances of the first (last) two items is likely to be influenced by an individual’s general skills to perform a job (belief about an employment change). Removing these restrictions allows for the possibility that correlations between these terms are also caused by other influences. Our results (mediation and interaction) do not change when restricting these correlations to zero.

20. Arguably, explaining 50% of variance of many indicators is more difficult than explaining 50% of variance of few indicators. In particular, deleting two items of flexible role orientation would have allowed us to obtain an AVE above 50%. However, deleting these items leads to the same results for the main models tested (mediation and interaction). Similarly, our main results remain robust when eliminating one or more items of RBSE (or even the whole construct). Importantly, we employ self-efficacy as a control in our models. Thus, potential problems with the measure (e.g. variance extracted due to measurement error) likely translate into increased validity of our main results (mediation and interaction) and reduce noise in these estimates.

21. We also conduct Harman’s (1967) single-factor test including only those items that belong to the three main constructs of interest, that is, controllability, flexible role orientation, and proactive work behaviour. Again, we obtain exactly three factors with eigenvalues larger than 1.0 and do not find evidence for common method bias, as only 46.76% of overall variance is explained by the first factor. This first factor mainly captures flexible role orientation, which consists of seven items as opposed to four items for each of the other two constructs.

22. All significance tests in our study are two-tailed.

23. We include the construct flexible role orientation and the effects of self-efficacy and hierarchical level for two reasons. First, we want to compare the base model with the mediation model requiring the two models to be nested. Second, best practices testing mediation in OLS regression include the same control variables for the mediating as well as for the dependent variables. Although this practice impairs fit indices in SEM, it does not change our main results (mediation and interaction).

24. In order to test whether the mediation model better fits the data than the base model, we perform a chi-square difference test between the two models. The chi-square difference between the models amounts to 54.461, with two additional degrees of freedom in the base model. Thus, the chi-square statistic (Δχ²)/df amounts to 27.301, indicating that the mediation model fits the data significantly (p < .001, one-tailed) better than the base model. Using the chi-square statistic, we also test whether single steps of model building, such as reducing restrictions one by one from our base model, indicate significant improved fit to our data. We find that single steps of model building towards the mediation model are also significant at the 1% level (independent of their order).

25. A potential threat to the proposed mediation model is an endogeneity concern with regard to the relationship between controllability and flexible role orientation. Potentially, causality could run from flexible role orientation to controllability. In other words, companies could adapt performance evaluations to their managers’ role perceptions, expecting this to result in more proactive work behaviour. We also test such an alternative mediation model. Appendix 3, column 3 shows the results of this test. We find that there is a significant effect of flexible role orientation on controllability (b = −0.147***, t = −2.79, p < .01). However, there is no significant effect of controllability on proactive work behaviour (b = −0.035, t = −0.64, p = .52). In addition, the direct effect of flexible role orientation on proactive work behaviour remains highly significant (b = 0.302***, t = 5.64, p < .01), barely reducing the initial significant direct effect of flexible role orientation on proactive work behaviour (b = 0.310***, t = 6.03, p < .01, not tabulated). This result confirms that flexible role orientation is impacting proactive work behaviour. Thus, we do not find support for the alternative mediation model.

26. Specifically, we use four items from the Rizzo et al. (1970) scale to depict role ambiguity and six items – also from the Rizzo et al. (1970) scale – to depict role conflict. Both constructs show high levels of internal validity. Moreover, including the items of these two additional constructs in our EFA, we obtain – as expected – eight factors with eigenvalues larger than 1.0. Finally, our results remain robust (in direction and significance) when we add various additional control measures, such as (i) gender, (ii) company size, (iii) age, (iv) tenure, (v) distributional justice, (vi) job satisfaction, and (vii) trust in superior, to the two controls already employed.

27. In order to facilitate the interpretation of our results, we mean-centred the two interacting variables and calculated the interaction term as the product term of these two mean-centred variables.

28. Appendix 3 shows results of two additional interaction models using different indicators of flexible role orientation to create the interaction term assuring that our results are not driven by this choice.

Additional information

Funding

Karl Schuhmacher was supported by the Swiss National Science Foundation through an Early Postdoc.Mobility fellowship (P2LAP1-155482). Franz Fischer is grateful to the Hanns-Seidel Stiftung e.V. for the obtained PhD scholarship.