Declining inner suburbs? A longitudinal-spatial analysis of large metropolitan regions in Canada

Abstract

Spatially referenced quantitative research that attempts systematically to assess inner suburban decline in Canadian metropolitan areas is almost completely missing from the literature. This paper aims to fill this gap and examine whether inner suburban decline is occurring in Canada. Aggregated census tract level data are used to assess all zones for decline based on relative prosperity changes in median household income, average dwelling values, and average gross rent. The results indicate that inner suburbs in Canadian cities experienced a decline in median household income, average dwelling value, and prosperity factors between 1986 and 2006. While a few possible explanations and policy approaches are offered, more research is necessary to explore the implications of these trends.

Keywords:

• inner suburb
• decline
• Canada
• factor analysis

Acknowledgments

The authors would like to thank Pierre Filion and four anonymous reviewers for their insightful and encouraging comments. We are grateful to Editor Elvin Wyly for his editorial and linguistic support.

Notes

^1. Noncontiguous and often dispersed census tracts that experienced redevelopment in the core or inner city tracts are included in the “core” and “inner city” areas. The 1971 census year choice partially addresses this concern.

^2. It is important to note that there is a significant variation in size between the largest and the smallest CMAs studied. Smaller CMAs have relatively few census tracts, precluding advanced data analysis that requires a large number of observations.

^3. Here a single prosperity statistic for each urban zone and for the entire CMA for each census year is created by aggregating the census tract value for the three variables involved by the appropriate weight. The urban zone statistic then is divided by the corresponding statistic for the entire CMA. Thus, prosperity is measured here by weighted average for average value of dwelling and average gross rent as well as weighted median for household income.

^4. The IC values greater than 1 indicate increasing prosperity while those less than 1 decreasing prosperity. The percentage change of the RV indicates difference in prosperity over the 5-year period.

^5. Kaiser-Meyer-Olkin (KMO) and Bartlett’s Test confirmed that the dataset was a good fit for factor score analysis. Consequently, three factor scores were extracted. Analysis of Variance (ANOVA) mean comparison tests were conducted for every variable separately. Factor scores may not be size dependent. KMO and Bartlett’s Test was utilized in order to assess the appropriateness of factor score analysis on the data set. Only factor scores that explain a significant amount of variance and broadly measure economic prosperity are retained. Change in the retained factor scores is then calculated for all urban zones between 1986 and 2006.

^6. The relatively small size of some of the original 15 CMAs results in statistical insignificance at the regression model analysis. In fact, it is arguable whether 9 CMAs are too many, due to sample size of the smaller CMAs included in the regression analysis. However, this does not mean that the basic descriptive portrait of the 15 CMAs is worthless: the consistent effects in household income shares in inner suburbs compared to other urban zones at the very least suggest that there is a systemic decline in inner suburbs in all 15 CMAs with very different compositions, geographies, policies, historical and current leadership, etc.

^7. Due to the longitudinal nature of the study, random-effect generalized least squares (GLS) regression and fixed-effect GLS regression models are the two most useful regression candidates. The fixed-effect model, however, omits coefficients of covariates that vary between clusters (Rabe-Hesketh & Skrondal, 2008).

^8. It should be noted that the measure of median household income advantages outer suburbs and fringe/exurbs where households are the largest and disadvantages the core and the inner city, and to a lesser extent the inner suburbs.

^9. Cautions should be taken regarding the limitation and the reliability of average dwelling values because census respondents are asked by Statistics Canada to estimate their housing values in census surveys.

^10. Inner cities are likely to be influenced by institutions that persist, while new fringe/exurban developments are influenced significantly by expensive new developments with small background noise of other variables but which over time become more diffuse.

^11. Apartment rental units are regulated in location through zoning. Older nonapartment single rental units are more likely to be rented than newer housing. They may also be affected by municipal policy such as rent control in order to be of use on their own as a measure of investment or disinvestment.

^12. Recently, federal government has a few targeted programs and partnerships that address urban core area decline, suburban infrastructure revitalization, and transit improvement in several cities.