Innovation Strategies: Are Knowledge-Intensive Business Services Just Another Source of Information?

Abstract

The objective of this paper is to contribute to the empirical literature on innovation strategies and services, by analysing the use of knowledge-intensive services, and their impact on innovation, in manufacturing firms. The analysis is carried out at the firm level, on the basis of a survey covering 804 manufacturing establishments in the Province of Quebec (Canada). We investigate the extent to which existing internal capabilities and their interaction with external sources of knowledge, in particular the use of knowledge-intensive business services (KIBS), affect the level of innovativeness of manufacturing firms. Then we examine the extent to which different innovation strategies, and the way KIBS are integrated into these, are associated with innovation. We show that manufacturing firms adopt a variety of innovation strategies, none of which preclude innovation, even introverted strategies whereby firms interact little with outside agents. However, those strategies that incorporate KIBS have a considerably greater chance of leading to innovation.

Keywords::

• Innovation
• innovation activities
• sources of information
• knowledge
• KIBS
• extroverted and introverted innovators
• Quebec
• manufacturing

Acknowledgements

The authors acknowledge the financial support from the Social Science and Humanities Research Council in Canada (SSHRC 410-2011-0108) and Industry Canada. The authors would like to acknowledge the contribution of Réjean Landry to the elaboration of the survey. The usual disclaimers apply.

Notes

¹ The distinction between these systems relates not so much to the actors involved as to the role being played by the actors. Thus, a particular firm is part of the productive system (a knowledge user) in so far as it applies knowledge to its own internal process and participates in the value chain, and is part of the knowledge system (knowledge creator or synthesiser) in so far as it participates in the overall accumulation of knowledge in the industry, shared either formally through collaborations or informally through knowledge spillovers, labour mobility, patent applications and so on.

² These are defined by the service received by the user, and not by the sector to which the provider belongs. This approach is inspired by Landry and Amara (2010): some of their categories have been aggregated, and some benchmark services (such as accounting and legal advice) have been included because they are not expected to be strongly associated with innovation.

³ The sample is representative of the geography of Quebec's manufacturing establishments based on a regional classification that distinguishes between major metropolitan areas, minor cities close to metropolitan areas, rural areas close to metropolitan areas and outlying areas.

⁴ There exist a variety of clustering methods which differ along three main criteria. The first is whether they are hierarchical or iterative. Hierarchical methods combine observations one by one – after each combination each group of observations (i.e. a cluster) is treated as a new observation. If the data comprise ‘n’ observations, there are thus ‘n’ solutions going from no clustering – each observation is its own cluster – to 1 – all observations are clustered. Choice of the number of clusters can depend on a variety of criteria such as the proportion of the variance explained, the increase in variance explained between two clusters and the rough number of clusters that the researcher considers tractable. Iterative processes usually specify the number of desired clusters, and observations are assigned iteratively to the clusters until a information metric is maximised or minimised. The second criterion that is used to distinguish between clustering methods is the way in which clusters are distinguished from another. Ward's minimum variance criterion specifies that the method will attempt to minimise the within-cluster variance, i.e. the sum of distances between observations within each cluster (this is the information metric to be minimised): the method is associated with the person who devised it, Ward. The final criterion is the distance metric used as an input for the information metric. This distance can be measured directly, subtracting one observation's variable value from the other's or transformed in some way. The squared Euclidian distance means that the distances between observations are squared – thus amplifying the effect of larger differences and allowing clusters to be better distinguished. More information on clustering techniques can be found in Aldenderfer and Blashfield (1984), and discriminant analysis – a means of testing the robustness of the clusters – is described in Tabachnik and Fidell (2013).