Does consistency in management control systems design choices influence firm performance? An empirical analysis

Abstract

This study examines the impact on firm performance of theoretically consistent relationships between three management control systems (MCS) design choices – delegation, performance measurement, and incentive compensation. Based upon the ‘three-legged stool’ model and agency theory, the hypothesis is that theoretically consistent MCS design choices are associated with enhanced firm performance. Using survey data from large Australian firms, the findings support the hypothesis, suggesting that an appropriate MCS design is a determinant of firm performance. The study contributes to the literature by moving beyond a focus on the antecedents of the three key MCS design choices to the consequence of alignment of those choices.

Keywords:

• delegation
• incentive compensation
• performance measurement
• MCS design
• survey

Notes

1. Considering that large companies may have multiple business units and small companies may themselves be a single business unit, the research instrument stated that questions referred to a typical business unit in the company and that if no business units existed, respondents should respond to the questions in relation to the company as a whole.

2. The 1476 firms comprised 34 with more than 5000 employees, 46 with 2500–5000 employees, 108 with 1000–2500 employees, 164 with 500–1000 employees, 331 with 250–500 employees, and 793 with 100–250 employees. We randomly selected 67.8% (i.e. 1000/1478) of the firms in each of the above strata.

3. The 23 cases removed include: (a) three responses from not-for-profit/public sector organisations; (b) 11 responses from firms that did not provide details of the weighting of financial/NFMs on the basis that incentive-based pay did not exist; (c) eight responses from firms that did not provide incentive-based pay, but for which respondents indicated the weighting of financial/NFMs, and; (c) one case of an outlier firm. Since the question regarding weighting of financial/NFMs specifically asked weightings in the context of incentive compensation (cf. Appendix), we judged it safer to eliminate the eight cases where weightings were provided but no incentive-based pay. These cases may signal misinterpretation of the question or respondent's unwillingness to provide sensitive information (weighting of incentive-based pay). However, we conducted a separate analysis with these eight cases in dataset (n = 137; not tabulated) and found that the substance of the findings is not sensitive to their exclusion. The outlier was found in the incentive-based pay variable, with the respondent indicating an incentive of 300% of the base pay, well above the upper threshold of 104% (i.e. mean plus three standard deviations (Hair et al. 1998)).

4. We thank an anonymous reviewer for the recommendation to use Cluster Analysis.

5. Since PLS makes no distributional assumptions, we use bootstrapping to evaluate the statistical significance of path coefficients (Chin 1998). We followed a 1000 subsamples bootstrapping procedure.

6. PLS Graph Version 3.0 was used in this study.

7. We have also considered segmenting the sample into three sub-samples to provide a sharper contrast between high performing firms and low performing firms. However, the limited number of data points available made this unviable for the purpose of the PLS analysis.

8. The Two-Step Cluster Analysis procedure is an exploratory approach that seeks to reveal the natural groupings (i.e. clusters) within a dataset. It follows a log-likelihood measure to calculate the distance between clusters and uses Schwarz's Bayesian criterion to determine the number of clusters. No noise handling was deemed necessary because variables had been previously screened for outliers. IBM^® SPSS^® Statistics version 20 was used.

9. We also note that the cross-loading (shown in Table 2) was low for all constructs, providing additional reassurance regarding internal and external validity.

10. Incentive-based pay ranged between 3% and 100% in our sample. The average incentive compensation is approximately three times higher than that reported by Nagar (2002).

11. We chose No. of employees because it is likely to have a more direct impact on organisational structure decisions and possibly MCS design choices. However, the high correlation between the two options (>0.80) suggests no material difference.

12. The PLS path model does not provide a measure of overall goodness-of-fit, only R² for each dependant variable (reported in Table 4). We note that R² is especially low in the whole sample analysis, reflecting opposing forces at play between the high and low performing firms.

13. An additional analysis (not tabulated) indicates that the correlation between cluster assignment and firm performance shows minor changes (in the 0.02–0.03 range) when we control for business strategy, suggesting differences in performance are not driven by strategy.