Urban Clustering, Development Similarity, and Local Growth: A Case Study of Canada

ABSTRACT

The effect of urban clustering on the long-term patterns of urban growth is arguably two-fold: in sparsely populated areas, the presence of neighbouring towns increases the chances of facilitating local development due to inter-urban exchanges, while in more densely populated core areas, increasing clustering of the urban field reduces development rates due to inter-town competition for potential investors and migrants. In the present study, the effect of urban clustering on the patterns of urban growth is investigated for both centrally located and peripheral areas of Canada. Neighbouring towns in urban clusters of the country appear to exhibit similar levels of socio-economic development. However, when measured by different development indicators, inter-town development association differs in both nature and degree. In core areas, for instance, only population and housing variables appear to exhibit a strong spatial association, while that of employment-related variables—average income, and unemployment rate—is weaker. As suggested, this tendency reflects fundamental differences between the two groups of variables. While population and housing variables are associated with the clustering of residents in socially homogenous areas, inter-town development similarity in respect to employment-related variables is weaker, due to long-distance commuting.

Acknowledgements

This study was carried out in the framework of the Canadian Studies Programme, sponsored jointly by the Israel Association for Canadian Studies and the Department of Foreign Affairs of the Government of Canada. The present paper is based, in part, on B. A. Portnov & B. Wellar (2004), Development similarity based on proximity: A case study of urban clusters in Canada, Papers in Regional Science, 83(2), pp. 443–465. In a shorter form, the paper was presented at the 9th Jerusalem Conference in Canadian Studies, held 30 June–4 July 2002 at the Hebrew University of Jerusalem. The author thanks Dr Moshe Schwartz, Social Studies Unit, Jacob Blaustein, Institute for Desert Research, Ben-Gurion, University of the Negev, for his valuable help in preparing the paper for publication.

Notes

1. According to Porter's (2000, p. 16) definition, industrial clusters are “geographic concentrations of interconnected companies, specialized suppliers, service providers, firms in related industries, and associated institutions … in a particular field that compete but also cooperate”. This definition is essentially similar to that suggested by Portnov and Erell (2001) for UCs—“a group of towns located in close proximity to each other and connected by socio-economic and functional links”.

2. The definition of the urban field used in this study is different from that suggested by Friedmann and Miller (1965), who defined it as a predominantly rural hinterland surrounding a major population center and accessible by its residents. In contrast, in the present study, by “urban field” we understand a cluster of towns and cities of different sizes located close together, normally within a daily commuting range.

3. The size of the sample—four UCs—was restricted by a limited number of neighbouring urban localities of comparable size in peripheral areas of the country. Though two other “peripheral clusters”—Regina and Winnipeg—were considered as potential candidates, they were subsequently omitted from the analysis due to the small number of neighbouring urban localities of comparable size that they contain.

4. Global Moran's I used at this stage of the analysis is a commonly used measure of spatial association which helps to determine whether a parameter's values arranged in space in a systematic manner. The expected value of Moran's I is −1/(n − 1). If a calculated value of I is equal to the expected value (within the limits of statistical significance), the value of variable x in a locality i is statistically independent of the values of this variable in adjacent localities. If a calculated value of I exceeds the expected value, this indicates a positive spatial correlation. If a calculated value of I is below −1/(n − 1), negative spatial correlation occurs. Under this condition, neighbouring values are not independent but tend to be dissimilar (Cliff & Ord, 1981).

5. The values of the G_i (d) index are sensitive to how the spatial neighbourhood of a location (d _i or search radius) is defined. In the present analysis, different search radii (20 through 100 km) were tested. The results for the best performing spatial lag (d _i = 40 km for core clusters and d _i = 80 km for peripheral clusters) are the only ones reported.

6. The LISA statistic has four basic properties: (a) values that exceed the expected value (E[I _i ]) indicate positive spatial autocorrelation, in which similar (either high or low) values are clustered around point I; (b) values below the expected value indicate negative spatial autocorrelation; (c) the sum of LISAs for all observations is proportional to a global indicator of spatial association—Global Moran's I; (d) a normally distributed Z statistic (two-tailed) is calculated to determine significance of LISA (Anselin, 1994).

7. The Upas tree (or Antiaris toxicaria), an evergreen plant in the mulberry family, is believed to have an adverse effect on all animal life around it.