Place, Space and Distance: Towards a Geography of Knowledge‐Intensive Business Services Innovation

Abstract

Much has been written about the link between local networks and institutions, about place and territory, and the capacity to innovate. In this paper we set out to answer two questions, based upon a survey of 1,122 knowledge‐intensive business service (KIBS) firms in the province of Quebec, Canada. First, do KIBS firms in different regions display different propensities to innovate? If so, this will be taken as prima facie evidence that there is some connection between local context and innovation. Second, can any regional level explanatory variables be found to explain the different levels of regional innovation? We find evidence that geographic patterns of innovation exist amongst KIBS firms in Quebec, although they are not those expected if there were a connection between local territory and innovation. We find that innovation first decreases with distance from the core of metropolitan areas, then, after 30–50 km, begins to increase again, though this pattern is not the same for all sub‐sectors. This pattern is in keeping with recent theoretically derived expectations relating to the geography of innovation.

Keywords:

• Knowledge‐intensive business services (KIBS)
• innovation
• distance
• territories

Acknowledgements

This paper has been funded by a Canadian Social Sciences and Humanities Research Council ordinary research grant. The authors would like to acknowledge the contribution of Mark Freel, Réjean Landry and Nabil Amara to the elaboration of the survey, and of the Innovation Systems Research Network—in particular David Wolfe and Meric Gertler—for providing a forum in which some of these ideas have been developed. The authors would also like to thank three anonymous referees for their constructive comments.

Notes

1. The Canadian Survey on Innovation only includes establishments with over $250,000 in revenues and 15 employees.

2. Insufficient numbers of establishments have introduced zero, one, seven and eight innovations.

3. Restricted maximum likelihood estimations are made. No corrections are made for over‐dispersion given that we have restricted our analysis to innov2–innov6 and innovR2–innovR6. Otherwise, default settings have been used. HLM 6.2 is the name of a particular software package that performs hierarchical linear modelling.

4. This amounts to modelling the regional percentage of innovators across regions. The advantage of a two‐level model is the possibility of entering explanatory variables at establishment and regional levels. In this exploratory paper we do not fully exploit this possibility.

5. Results are adjusted for default SAS proc logistic settings (which models the likelihood of “0” events). SAS is the name of a particular statistical software package.

6. The distance from three smaller cities does not enter the models significantly.

7. In most cases at least 80 per cent of employment in a labour market (aggregation of municipalities by intensity of commuter flows) is included in one MRC.

8. Only one innovation variable is presented for each of the three sectors, and only analyses at the two‐digit level are presented.

9. Available upon request.

10. There are only 82 observations within the 120/150 km radius: this is sufficient for a logistic regression with three explanatory variables.

11. Sectors were aggregated to provide a sufficient number of observations for tests: sectors 5411, 5412 and 5419 (n = 139); 5417 and 5418 (n = 91).