Urban Hierarchy or Local Buzz? High-Order Producer Service and (or) Knowledge-Intensive Business Service Location in Canada, 1991–2001

Abstract

The location of high-order producer services has been extensively documented for the 1970s and 1980s when researchers turned their attention to the effects of tertiarization on regional development. In this article we propose, on the one hand, to update the spatial analysis of high-order producer services by investigating whether they have continued to diffuse away from the top of the urban hierarchy between 1991 and 2001. On the other hand, we also propose to incorporate certain hypotheses from the knowledge-intensive business services (KIBS) literature, in particular possible spillover effects, synergies between economic sectors and labor market effects, into the spatial analysis. Over and above city size, do these factors contribute to our understanding of the spatial dynamics of high-order producer services? Overall we find that these services reversed their diffusion process during the 1990s, decreasing their presence in smaller cities in peripheral regions. Labor market, synergies, and spillover effects contribute, to some extent, to understanding their overall spatial distribution and its evolution during the 1990s, but region and urban size remain their principal organizing factors across space.

La ubicación de los servicios de conocimientos complejos a la producción había sido extensamente documentada para las décadas de los 70 y de los 80, cuando los investigadores enfocaron su atención en los efectos de la terciarización en el desarrollo regional. En este artículo proponemos, por un lado, actualizar el análisis espacial de los servicios de conocimientos complejos a la producción identificando si su difusión continuó alejándose del nivel superior de la jerarquía urbana entre 1991 y 2001. Por otro lado, también proponemos incorporar en el análisis espacial ciertas hipótesis contenidas en la literatura sobre servicios comerciales de conocimientos extensos (knowledge-intensive business services, KIBS), en particular los posibles efectos del desbordamiento (spillover), las sinergias entre los sectores económicos y los efectos del mercado de trabajo. Cuando se excede el tamaño de la ciudad, contribuyen estos factores a nuestra comprensión de la dinámica espacial de los servicios de conocimientos complejos a la producción? En general encontramos que estos servicios revirtieron su proceso de difusión durante la década de los 90, reduciendo su presencia en ciudades de menor tamaño de las regiones periféricas. El mercado de trabajo, las sinergias y los efectos de desbordamiento contribuyen, hasta cierto grado, al entendimiento de su distribución espacial general y su evolución durante la década de los 90, pero región y tamaño urbano siguen siendo los principales factores de organización en el espacio.

Key Words:

• Canada
• high-order producer services
• KIBS
• location factors

关键词:

• 加拿大
• 高阶生产者服务业
• KIBS
• 位置因素

Palabras claves:

• Canadá
• servicios de conocimientos complejos a la producción
• KIBS
• factores locacionales

Acknowledgments

RICHARD SHEARMUR is Associate Professor of Economic Geography at the INRS, an urban studies research and teaching institute at the University of Quebec, 385 Sherbrooke East, Montréal, H2X 1E3, Canada. E-mail: [email protected]. He holds the Canada Chair in Spatial Statistics and Public Policy. His work covers both intrametropolitan and regional aspects of the spatial economy, with particular emphasis on high-order services and peripheral regions.

DAVID DOLOREUX is Associate Professor of Management at the Telfer School of Management, University of Ottawa, Vanier Hall, Room 257B, Ottawa, K1N 6N5, Canada. E-mail: [email protected]. He holds a University of Ottawa Chair in Entrepreneurship, Innovation and Regional Policy. His work focuses on the way in which firms interact and innovate, with particular attention given to the influence of, and effect on, their regional and institutional settings.

Notes

^aIn the local column the R ² increases attributable to variables are measured with formula (1), in the surround column with formula (2). Increases in R ² of over 4 percent are highlighted in bold.

^b n indicates the variable did not add at least 1 percent to the adjusted R ² of the basic model. Each model has been adjusted for outliers (Cook's D ≥ 0.1). In some cases the additional variable modifies the outliers. The increase inR ² is always relative to the same model (which may differ from the basic model).

^c n = 151.

^d n = 150.

^e n = 149.

^f n = 148.

^gThe independent variables for growth are 1991 variables. The independent variables for Factors 1 to 6 are 2001 variables. Multicollinearity (Variance Inflation Factor ≥ 10) has been indicated by ^M . Individual regression coefficients are not analyzed, so tests for heteroskedasticity are not reported.

*We would like to thank three anonymous referees and the editor for their comments and suggestions. We remain responsible for the contents of the article.

¹Of the 152 urban areas 140 are Census Agglomerations (CAs) and Census Metropolitan Areas (CMAs). These correspond to 1991 labor markets. The remaining twelve urban areas correspond to those municipalities, not included in a CA or a CMA, that had more than 10,000 inhabitants in 1991.

²The criterion for retention is variance explained greater than 1/17 before rotation.

³The network used consists of highways and regional roads. All cities are connected to this network. A few rural regions are not. In this case secondary roads are added to the network. Speeds of 100 km/h along highways and 60 km/h along regional roads are assumed.

⁴The centroid of each spatial unit has been adjusted using population counts at the block group level.

⁵ Table 4 uses a subsample of the sectors indicated in Table 1: The 2001 subsample gathers 85 percent of employment of the full KIBS selection.

⁶In an analysis of variance model of the type: LQ _KIBS = f(regions, urban), the regional classification is not significant in either 1991 or 2001. It has nevertheless been left in the model (results presented graphically in Figure 2) to ensure that the urban hierarchy results are not affected by regional variations. All values of r ² mentioned are adjusted r ².

⁷In 2001this selection of KIBS accounts for 85 percent of KIBS jobs that we identify using the finer NAICS data.

⁸This percentage indicated the variance explained by the component. The cutoff for selecting components is v ≥ 1/17; i.e., v ≥ 5.9 percent.

⁹To characterize each component, the NAICS codes for sectors with a component loading of 0.48 or higher are indicated. A cutoff of 0.48 is chosen because the next highest loading is 0.41, and a cutoff in the vicinity of 0.50 is sought. No sector is loaded (at the 0.48 level) with two components.

¹⁰Because heteroskedasticity and multicollinearity may be present, significance levels for individual coefficients are not reported or relied on in this analysis.