The Role of Management Controls in the Higher Education Sector: An Investigation of Different Perceptions

Abstract

A variety of empirical studies have investigated the relationship between management controls (MCs) and the financial and nonfinancial performance of for-profit firms, but there is a dearth of research on MCs in the higher education (HE) sector. Our study contributes to MC research by investigating the mediating role of MCs in the relationship between structural autonomy and both research and teaching performance in HE institutions and by comparing the perceptions of two major actors: heads of administration and academic deans. We use survey data of two paired subgroups with 104 heads of administration and 104 academic deans and perform a multigroup analysis with structural equation modeling. We find significant differences between heads of administration and academic deans concerning the relationship between structural autonomy and MCs and between MCs and teaching and research performance for the diagnostic and interactive use of controls. For heads of administration, we find that only strategic boundaries and strategic beliefs are significantly associated with teaching performance, whereas for academic deans, more autonomy is associated with more emphasis on most MCs. Furthermore, for academic deans, more emphasis on diagnostic use, but less on interactive use, is associated with higher research performance.

Keywords:

• Management controls
• Higher education
• Professionalism theory
• Multigroup analysis

JEL Classifications:

• H83
• I23
• M10
• M40

Acknowledgments

This paper has greatly benefited from the constructive feedback and support of Sally Widener (associate editor) and two anonymous EAR reviewers. Earlier drafts of this paper were presented at the 8th EIASM workshop on Performance Measurement and Management Control in Nice 2015 (France), at the 78th Annual Conference of German Academic Association for Business Research in Munich 2016 (Germany), at the 41st European Accounting Association Conference in Milan 2018 (Italy) and at the 11th EIASM New Directions in Management Accounting Conference in Brussels 2018 (Belgium). We thank Eva Ström and all participants for their helpful comments and suggestions.

ORCID

Thomas W. Guenther http://orcid.org/0000-0002-9860-8668

Notes

1 Some authors (e.g., Abbott, 1988; Larson, 1977; Wilson et al., 2013) understand professionalism as a self-interested conception, in which the professional community is seen as operating like a cartel, promoting its own self-interests, protecting its occupational monopoly and safeguarding the power and status of its members.

2 In contrast with structural autonomy, autonomy can also capture work activities conducted in the absence of direct observation or engagement of management (e.g., Bedford & Malmi, 2015) based on Ito and Peterson (1986) and Kober, Ng, and Paul (2007).

3 Similar to Tessier and Otley (2012), we speak of ‘strategic beliefs’ and ‘strategic boundaries’ in the following.

4 We take the relationship between the MCs into account by adding paths between the four MCs in our alternative Model 2 (for further details, see section 5.3.).

5 These studies address two important variables, which are not included in our study. When decentralization and information asymmetry coincide, performance measures are used to monitor the use of decision-making rights of divisional managers (Abernethy et al., 2004). In addition, performance measures can be connected with incentive systems for managers to work towards achieving the organization's objectives (Bouwens & Van Lent, 2007; Widener et al., 2008).

6 We refer to the results of their third model and the constructs of academic management performance criteria, which is related to the diagnostic use of performance measures.

7 In three HE institutions, administrators had the title of professor but, no longer in academia, were in charge of the HoA function, which is not expected to bias our results.

8 If an HE institution had no school/department of humanities or social sciences, we contacted the dean of an alternative school/department. Whenever possible, we contacted two deans to improve the response rate.

9 157 / 521 HE institutions = 30.1 percent institution-based response rate; for each HE institution, by nature, only one HoA could be addressed.

10 281 / (521 × 2) = 27.0 percent of all addressed individual academic deans (two per HE institution) with 248 out of 521, i.e., 47.6 percent of HE institutions covered.

11 After pretests, we added one item for teaching performance measuring the internationalization of teaching and one item for interactive use of the control system addressing the discussion among academic staff. Furthermore, Abernethy and Brownell’s (1999, p. 202) item, ability to win resources, was divided into two items, cooperation with nonuniversity research facilities/companies and additional income (e.g., third-party funds, sponsoring) to measure research performance.

12 UniRank allows us to cover 99 of the 104 HE institutions of our sample. It is up to date, covers both research and teaching, and covers all three countries of our study, in contrast to other rankings, such as Shanghai, CHE, and FTSE.

13 Mardia’s (1970) test of multivariate kurtosis shows that the data are multivariate nonnormal. However, in contrast to covariance-based SEM techniques, PLS does not require multivariate normally distributed data.

14 Group comparisons in SEM can be misleading if the invariance of their measures (Henseler et al., 2016) is not assured. The comparison of path coefficients between the groups requires at least partial measurement invariance (Henseler et al., 2016; Steenkamp & Baumgartner, 1998). The results of the measurement of invariance validation are untabulated but are available from the authors upon request.

15 One item (SA1: Structural academic units can independently set their own goals) loaded (.689) marginally below the recommended threshold.

16 Only one item of the construct structural autonomy (SA1: Structural academic units can independently set their own goals) has a lower factor loading (0.579). However, according to Hulland (1999), this factor loading can be considered acceptable. Furthermore, based on theoretical considerations, we kept this item in our measurement.

17 Items means range between 4.17 and 5.36 for 7-point anchored Likert scales.

18 The Q²-values range between 0.303 and 0.574 for HoAs and between 0.239 and 0.571 for academic deans; thus, they also indicate prediction relevance for both subgroups.

19 The Excellence Initiative, starting in 2005, focused on graduate schools, clusters of excellence and institutional strategies and granted a total of €1.9 billion to selected German HE institutions.

20 Similar to the argumentation of institutional theory (e.g., DiMaggio & Powell, 1983).

21 With the exception of the subgroup of HoAs, the path coefficient of the control variable type of HE institution with research performance is not significant.

22 By considering mediator variables (MCs) in the SEM, it can even be shown that a variable X (structural autonomy) that does not have a significant total effect on Y (research or teaching performance) has significant indirect and direct effects on Y (performance), but which act in the opposite direction and therefore cancel each other out in the total effect. In our case, this concerns the variable interactive use for the aggregated sample and especially for the subgroup of academic deans.

23 For the alternative model 1 the associations remain robust with the exemption of the association between structural autonomy and strategic beliefs for the aggregated sample and for HoAs and between structural autonomy and strategic boundaries for HoAs which became positively significant. For the alternative model 2 only the association between structural autonomy and strategic beliefs became positively significant for HoA. For H4b, for both alternative models the path difference between strategic boundaries and teaching performance is no longer significant.

24 Available upon request from the authors.