Headquarters−Subsidiary Interdependencies and the Design of Performance Evaluation and Reward Systems in Multinational Enterprises

Abstract

This study investigates the impact of headquarters–subsidiary interdependencies on performance evaluation and reward systems in multinational enterprises. Headquarters–subsidiary interdependencies refer to the extent to which headquarters and subsidiaries depend on each other to accomplish their tasks. When headquarters–subsidiary interdependencies are present, it becomes more difficult to reward the performance of subsidiary managers because these interdependencies induce noise on subsidiary-level accounting performance measures, while at the same time high levels of goal alignment between headquarters and subsidiary managers are required. Based on survey data from 82 foreign subsidiaries operating in Belgium with headquarters in 14 different countries, our partial least squares path modelling results show that as headquarters–subsidiary interdependencies increase, headquarters use more participative performance evaluation and consider more the effects of uncontrollable factors on subsidiaries' performance when rewarding subsidiary managers. More importantly, while prior research suggests that interdependencies induce noise on unit-level accounting performance measures, our results indicate that participative performance evaluation may mitigate the noise so that headquarters still rely on subsidiary formula-based compensation using accounting measures to reward subsidiary managers.

Acknowledgements

We are grateful to Salvador Carmona (past editor) and two anonymous reviewers for their helpful comments and suggestions. The comments of Walter Aerts, Tom Groot, Marc Jegers, Neale G. O'Connor, Tatiana Sandino, Michael Shields, Alexandra Van den Abbeele, Wim Van der Stede, Laurence Van Lent, and Arjen Van Witteloostuijn are greatly appreciated. This paper has also benefited from presentations at the 2010 AAA Annual Meeting in San Francisco, the 2010 AAA Management Accounting Section Meeting in Seattle, the 2010 EAA Annual Meeting in Istanbul, the EAA 25th Doctoral Colloquium in Tampere, and the 5th Conference on Performance Measurement and Management Control in Nice. Financial support from the University of Antwerp is gratefully acknowledged.

Notes

One exception is the study of O'Donnell (2000), which empirically investigates headquarters–subsidiary interdependencies. The author finds that headquarters use several social control mechanisms in response to increasing headquarters–subsidiary interdependencies. Although O'Donnell (2000) provides valuable insights, she does not examine the impact of headquarters–subsidiary interdependencies on the formal parts of control systems in MNEs.

According to Thompson (1967), there are three types of interdependencies between business units within a firm: pooled, sequential, and reciprocal. Reciprocal interdependencies are the highest form of interdependencies, where intermediate output moves back and forth between business units several times before the final product is delivered. Similarly, in the context of headquarters–subsidiary interdependencies, reciprocal interdependencies describe a two-way relationship where key knowledge and resources move back and forth between headquarters and subsidiaries before final decisions are made.

Note that if subjective performance evaluation is used, headquarters may overweigh the importance of subjective performance evaluation to justify headquarters' favoritism (Prendergast and Topel, 1993; Lazear, 1999; Moers, 2005). However, the literature also shows that headquarters often make adjustments in performance evaluation to address fairness concerns and motivation (Bol et al., 2010; Bol, 2011; Bol and Smith, 2011), mainly because employee perceptions of unfairness in evaluation and compensation have a negative effect on satisfaction and motivation (Cohen-Charash and Spector, 2001; Colquitt et al., 2001). When headquarters–subsidiary interdependencies are high, subsidiaries have specific knowledge and play an important role in the MNE. Thus, the self-interested behaviour of subsidiary managers would bring huge losses to the MNE. Therefore, in our setting, headquarters are more likely to consider uncontrollable factors and make adjustments in performance evaluation to address fairness concerns and motivation.

We thank one of the anonymous reviewers for this constructive comment.

One can argue that subsidiary managers may not communicate with headquarters all their information or may not communicate honestly in the process of participative performance evaluation, which leads to a less efficient budget. This is less likely, because prior analytical studies show that even though the agent may misrepresent or suppress some of the private information, the resulting budget may be more efficient than the best one achievable without the agent's participation (Baiman and Evans, 1983; Penno, 1990; Dunk, 1993).

In addition to relative size, being a non-affiliated industry is another factor that induces information asymmetry between headquarters and subsidiary managers. If a business unit operates in a different industry from its headquarters, the business unit managers may obtain private information and specialised knowledge on the industry that is not available to the headquarters (Baiman et al., 1995; Bushman et al., 1995; Roth and O'Donnell, 1996; Bouwens and van Lent, 2007). Following Baiman et al. (1995) we measure non-affiliated industry as a dummy variable equal to one if the subsidiary's two-digit US SIC is different from that of its headquarters. However, this variable is not significantly related to any of the control systems. To maintain the power of the test, we delete this variable from the final model.

Additionally, prior research suggests that headquarters will delegate decision rights to business unit managers when information asymmetry is high (Jensen and Meckling, 1992; Brickley et al., 2004). Moreover, interdependencies can result in negative externalities for the firm if decision rights are decentralised (Christie et al., 2003; Abernethy et al., 2004). Finally, US MNEs often use more decentralisation in their foreign subsidiaries (Chow et al., 1999). Therefore, our model assumes relations from our independent variables (headquarters–subsidiary interdependency, inter-subsidiary interdependency, relative size, Anglo-Saxon) to decentralisation.

Ideally, we should also add a dummy variable to control for the influence of Japanese MNEs since prior research shows that Japanese MNEs use control mechanisms differently from other MNEs (Chang and Taylor, 1999). Our sample has only four Japanese subsidiaries, which does not allow us to estimate the influence of Japanese MNEs using PLS. However, when we delete the four Japanese subsidiaries and re-estimate the PLS model, the results are similar to those reported in the main analysis.

Smart PLS 2.0 is used for PLS path modelling. The data were standardised before conducting the PLS path modelling. Standardisation helps avoid computational errors by lowering the correlation between product indicators and their individual components (Chin, 1998). The PLS model does not constrain the relation between subjective performance evaluation and formula-based compensation to zero, since no theoretical or empirical evidence suggests that these are mutually exclusive. Thus, the model also links subjective performance evaluation to formula-based compensation.

Although excluding the two items is preferable, we also re-estimated the model, including the two items. The results are similar to those reported here.

We did not constrain the relation between inter-subsidiary interdependency and headquarters–subsidiary interdependency to be equal to zero. While inter-subsidiary and headquarters–subsidiary interdependencies are different, no theoretical or empirical evidence suggests that they are mutually exclusive. The correlation analysis above shows that, indeed, the two variables are significantly related. However, PLS is not able to model the inter-relations between latent variables. The results reported in the study reflect a relation from inter-subsidiary interdependency to headquarters–subsidiary interdependency. We also re-estimated the model assuming a relation from headquarters–subsidiary interdependency to inter-subsidiary interdependency. The results are similar to those reported in the main analysis.

Note that headquarters–subsidiary interdependency is not significantly associated with decentralisation. This may be due to the following two conflicting arguments. On the one hand, headquarters–subsidiary interdependencies are associated with the negative externalities of subsidiary managers, where maximisation of the subsidiary's profit does not necessarily lead to maximisation of the MNE's profit. Centralisation could reduce goal incongruence and avoid the negative externalities of subsidiary managers (Christie et al., 2003; Abernethy et al., 2004). On the other hand, headquarters–subsidiary interdependencies are associated with specific information and knowledge of subsidiary managers that can be utilised by other units of the MNE. Decentralisation would provide subsidiary managers with flexibility and motivation to use their knowledge to quickly respond to external uncertainty and complexity (Christie et al., 2003; Brickley et al., 2004).

The factor loading of the item measuring the extent to which the subsidiary depends on headquarters for managerial expertise and human resource management is lower than 0.50. We exclude this item in the analysis to ensure that the estimates of the parameters linking the constructs are not biased due to low loadings (Chin, 1998).

Additional information

Notes on contributors

Yan Du

Paper accepted by Salvador Carmona.

Marc Deloof

Paper accepted by Salvador Carmona.

Ann Jorissen

Paper accepted by Salvador Carmona.