Disentangling the Role of Management Control Systems for Product and Process Innovation in Different Contexts

ABSTRACT

This paper studies the relationship between the use of management control systems and different types of innovation (product and process), taking into account innovation contexts (high-tech versus low-tech sectors). We develop and test our hypotheses based on a large sample of Spanish manufacturing firms and find that input controls have a positive association only with process innovation in both innovation contexts. Behavior controls have similar effects on both types of innovation outcomes for high-tech firms, while they have stronger positive associations with process than with product innovation for low-tech firms. Output controls are equally relevant for product and process innovation in both contexts.

Keywords:

• Management control systems
• Product innovation
• Process innovation
• Sector characteristics
• Panel data

Acknowledgements

We greatly appreciate comments from the participants at the XVI Workshop on Accounting and Management Control at IE Business School, the Management Control Association conference 2016 at the University of Antwerp, the 10th Conference on New Directions in Management Accounting at the Metropole Hotel in Brussels, the IV Research Forum on Challenges in Management Accounting and Control at Universidad Pablo de Olavide and the invited Seminar at Universitat Autònoma de Barcelona. Finally, the comments and suggestions provided by Josep Bisbe, Beatriz Garcia Osma, Manuel Nuñez Nickel, the Associate Editor and the two anonymous referees have greatly contributed to the improvement of the manuscript.

ORCID

Bing Guo http://orcid.org/0000-0002-6127-8335

Evita Paraskevopoulou http://orcid.org/0000-0001-7820-524X

Luis Santamaría Sánchez http://orcid.org/0000-0002-3356-4342

Notes

1 We follow the OECD’s classification of manufacturing industries. According to OECD (2005, 2011), two indicators of technology intensity are used in the classification: i) R&D expenditures divided by value added; and ii) R&D expenditures divided by production. Then, the division of manufacturing industries into high-technology, medium-high-technology, medium-low-technology and low-technology groups was made after ranking the industries according to their average over 1991–1999 against aggregate 12 OECD countries’ R&D intensities. As in previous research (Heidenreich, 2009; Santamaría et al., 2009; among several others), we combine low-tech and medium-low-tech industries under the label of low-tech industries; and we join high-tech and medium-high-tech industries under the label of high-tech industries. Usually, low-tech industries are those with an average R&D intensity lower than 0.9% (textiles, food products, tobacco and wood, among others) and those (medium-low-tech) with an average R&D intensity of between 0.9% and 3% (rubber and plastic products, coke, refined petroleum products, basic metals, among others). High-tech industries are those (medium-high-tech) with an average R&D intensity between 3% and 5% (electrical machinery, motor vehicles, railroad and transport equipment, machinery, among others) and those (high-tech) with an average R&D intensity higher than 5% (aircraft and spacecraft, pharmaceuticals, office machinery, radio, TV, medical, precision and optical instruments, among others).

2 Many researchers have used the same database to study innovation, finance, human resources, etc. Their works have been published in top journals: https://www.fundacionsepi.es/investigacion/esee/en/sesee_articulos.asp.

3 The survey’s detailed questionnaire is available at http://www.fundacionsepi.es/investigacion/esee/en/svariables/indice2.asp.

4 We construct indexes for different types of MCS, rather than use individual proxies, as the main independent variables. According to Henseler (2017), if the goodness of fit of the model is not significantly worse, the composite indicators should be preferred over individual indicators based on the criterion of parsimony. Comparing the statistics of the estimated models of Table 4 (index effects) and that of Table 7 (individual indicator effects), we find that it is suitable to use composite indicators in our study. However, some authors raise concerns about the use of composite indicators due to the potential loss of information and suggest the use of individual indicators (Howell, Breivik, & Wilcox, 2007).

5 The Input control index can be negative because of the standardization of each proxy.

6 Our database allows us to capture the target setting and evaluation functions of output controls, however it remains constrained in offering valid proxies for rewarding mechanisms. Hence, the latter remain beyond the scope of the current analysis.

7 We also scale Leverage, Market share, Client concentration and Supplier concentration by taking natural logarithms. The regression results remain the same.

8 Our sample is a panel dataset. Hence, we also use a random-effect panel probit model, by controlling for firm characteristics and year effect, as a robustness check of our results. The results remain the same. They are available upon request.

9 We show the summary statistics of the full sample in Table 1 to provide a general pattern of our sample, even though the main estimations are based on the subsamples of different innovation contexts (high-tech vs. low-tech). The summary statistics of the two subsamples are untabulated. For high-tech (low-tech) firms, the mean of input control index, the behavior control index, and the output control index is 0.349 (−0.330), 1.560 (0.983), and 0.846 (0.534), respectively.

10 One concern of using composite indicators is high multicollinearity among the proxies. Following Bedford and Speklé (2018) and Henseler, Hubona, and Ray (2016), we examine the multicollinearity of the composite proxies for each type of MCS. In untabulated results, the individual VIF values of proxies for the three types of MCS in our samples are lower than 3 and the average VIF values are lower than 1.9. Hence, the multicollinearity of composed indicators is not a concern of this study. These results also indicate that each proxy with at least 70% of its total variance is not shared with the other proxies of the same type of MCS, which further confirms the use of composite indicators (Bedford & Speklé, 2018).

11 This is the average marginal effect of Input control index on innovation outcomes. To save space, the average marginal effects of all independent variables for all regressions are untabulated but they are available upon request. For the discussion of the remainder of the paper, we use the average marginal effects of independent variables.

Additional information

Funding

This work was supported by the Spanish Ministry of Economics and Competitiveness [grant number ECO2013-45864-P]; the Community of Madrid and the European Social Fund [grant number S2015/HUM-3417] and [INNCOMCON-CM]; and Ramon Areces Foundation [grant number What Triggers Innovation?]; FEDER [grant number UNC315-EE-3636].