The tendency of advanced services to co-locate and the implications for regional government policy

Abstract

If services have a tendency to agglomerate in some places and not in others, does it make sense to attempt to divert them to areas where agglomeration benefits are weak or not present at all? Using a new spatial data set for Canada applied to a stochastic location model [Ellison, G., & Glaeser, E.L. (1997). Geographic concentration in U.S. manufacturing industries: A dartboard approach. Journal of Political Economy, 105, 889–927], we find that services co-locate with other services and with other industries across space. Agglomeration economies are pronounced, and general rather than industry specific. Services appear no less prone to co-agglomeration than manufacturing industries. The implications for government policy incentives are mixed, suggesting caution.

Keywords:

• services
• co-location
• agglomeration
• spatial
• policy

Acknowledgements

This research was supported by a grant from the Leslie Harris Centre for Regional Policy and Development, Memorial University of Newfoundland. This financial support is gratefully acknowledged. Excellent programming and research assistance was provided by Awais Chaudhry and Jared Tobin. Thanks are due Lanfranco Senn for helpful comments on an earlier draft.

Notes

For example, ocean technology, life sciences, environmental and chemical engineering, (naval) architecture, ICT, legal, financial and advertising services, and myriad professional consultancies.

Maurel and Sèdillot (1999) propose an alternative measure of raw geographic concentration to be used in the above model, namely . Their associated measure of localisation, γ _ms, is obtained by substituting G _ms for G in Equation (2). Industries appear to be somewhat more geographically concentrated on γ _ms.

The trigger values suggested by EG and Maurel and Sèdillot are as follows: γ < 0.02, 0.02 < γ < 0.05, γ > 0.05 imply, respectively, low, medium, and high concentration (localisation).

See Endnote 1.

Statistics Canada – CANSIM II Table No. 2820068, total employed by census metropolitan area and census agglomeration and by non-census metropolitan area and non-census agglomeration.

Statistics Canada – Cat. No. 72-002-XIB.

Statistics Canada – Cat. No. 15-201.

Statistics Canada – Cat. No. 61F0040XCB.

It should be noted that the way CD boundaries are drawn, it is not necessarily the case that no core CDs are more distant from core areas (and therefore more rural in character) than the areas we classify as non-core.

Contiguous spatial units may influence one another in a non-random fashion. Analysing the data at the CD level using CSD data aggregated in the manner described minimises but does not eliminate the way in which arbitrary spatial boundaries can bisect agglomerations of establishments, ignoring the presence of establishments that lie beyond the boundary even if they are located near it, thereby generating imprecise estimates of the true levels of localisation.

See Wernerheim (2007) for detailed results, including those for the Maurel–Sèdillot estimator.

The ratio of the number of establishments with up to 100 employees to the total number of establishments.

Commodity #39 at the S-level aggregation of the national input–output tables.

See Wernerheim (2007).