Testing for control system interdependence with structural equation modeling: Conceptual developments and evidence on the levers of control framework. Issue 1 (9th March 2018)

• Record Type: Journal Article
• Authors: Johansson, Tobias
• Is Part Of: Journal of accounting literature. Volume 41:Issue 1(2018)
• Page Start: 47
• Page End: 62
• Publication Date: 2018-03-09
• Subjects: Structural equation modeling -- Control practice interdependence -- Seemingly unrelated regression -- Non-recursive path models -- Control systems -- Management control -- Diagnostic control; Accounting literature -- Periodicals; Accounting -- Periodicals; Accountancy; Comptabilité -- Documentation -- Périodiques; Comptabilité -- Périodiques; Accounting; Accounting literature; Periodicals; 657.05
• DOI: 10.1016/j.acclit.2018.02.002
• Languages: English
• ISSNs: 0737-4607

Abstract

Abstract : This article deals with how to test for and evaluate interdependence among control practices in a management control system using structural equation modeling. Empirical research on the levers of control (LOC) framework is used as an example. In LOC research, a path model approach to interdependence has been developed. The appropriateness of this approach is evaluated, developed, and compared with the correlation of residuals approach (seemingly unrelated regression) implemented in the wider complementarity literature. Empirical examples of the different models are shown and compared by using a data set on LOC of 120 SBUs in Sweden. The empirical results show that modeling interdependence among control practices in a management control system as non-recursive (bi-directional) paths or as residual correlations evidently affects the conclusions drawn about interdependence in terms of both presence and magnitude. The two models imply different views on how to conceptualize interdependence and are not statistically and empirically comparable. If using non-recursive path models, several model specification issues appear. To be able to identify such models, this needs to be carefully considered in the theory and research design prior to data collection.