Exploring the Causes and Consequences of Regional Income Inequality in Canada

abstract

The recent surge in populist movements sweeping many countries has brought into focus the issue of regional inequality. In this article, we develop a panel data set for Canada that includes information on 284 regions observed at 5-year intervals (from 1981 to 2011) and estimate a series of spatial econometric models to study the causes and consequences of regional inequality. Our results draw attention to the fact that the rise in inequality at the national level has been accompanied by greater cross-regional inequality. Differences in the level of economic development, precariousness of labor market conditions, and socioeconomic factors are among the key drivers of these regional patterns of inequality. We also find that the industrial mix of a region plays an important role in shaping its distribution of income: regions with high concentrations of manufacturing activities typically have lower levels of inequality, whereas regions with high concentrations of tertiary services, arts, and entertainment, as well as knowledge-intensive business services tend to have higher levels of inequality. In terms of the consequences of inequality, the growth/equity trade-off across Canadian regions varies significantly over the short- vs. medium-term horizons. In the short run, our results suggest that inequality is positively related to regional economic growth. This response changes as we move to a medium-term horizon, which suggests that as inequality persists over longer periods of time, it has a negative and significant impact on regional growth trajectories. Panel vector autoregressive models are also used to further explore the direction of causality of the growth-inequality relationship.

Key words:

• regional inequality
• Canada
• causes
• consequences

JEL codes:

• R11
• D30
• I30
• O18

Acknowledgments

We are grateful to the editor and reviewers for their valuable comments on earlier versions of the paper. This work was supported by the Social Sciences and Humanities Research Council of Canada [435-2018-1439]. The analysis presented in this paper was conducted at the Quebec Interuniversity Centre for Social Statistics (QICSS) which is part of the Canadian Research Data Centre Network (CRDCN). The views expressed in this paper are those of the authors, and not necessarily those of the CRDCN or its partners.

Supplemental data

Supplemental data can be accessed here.

Notes

¹ It is important to note here that we are emphasizing studies on the distribution of incomes and not disparities in terms of means incomes or per capita gross domestic product (GDP). Considerable attention has been given to the latter convergence debate in Canada (e.g., Coulombe 1999; Breau and Saillant 2016), but that is not the center of our attention here.

² More details relating to the nature of the Census and NHS data are available in the Appendix, in the online materials.

³ Total income, which is the primary income concept used in this article, refers to employment earnings (i.e., wages and salaries) and self-employment income received by individuals as well as dividends and other investment incomes, retirement pensions, and government transfer payments (e.g., Old Age Security, Canada and Quebec Pension Plans, employment insurance benefits, child benefits). Furthermore, income is measured on a pretax basis, and for each Census, it is reported for the calendar year prior to the Census.

⁴ CDs are akin to counties in the US and UK. They are aggregates of adjacent CSDs (n = 5,253 CSDs in 2011), which are in turn the spatial units used to define the boundaries of census metropolitan areas (CMAs) and census agglomerations (CAs) (i.e., the country’s urban centers). While CDs are the preferred units of analysis for regional studies and regional policy development in Canada, they do not correspond perfectly with the definition of CMAs or CAs. For instance, Montreal as a CMA consists of twenty-seven different CSDs (with a total population of roughly 3.9 million in 2011). These twenty-seven CSDs are part of eleven different CDs (the core of which—roughly 60 percent of the population in 2011—consists of the Montreal and Laval CDs).

⁵ Using a global data set from the UN’s World Urbanization Prospects, Castells-Quintana (2018) has recently found evidence of a U-shaped relationship where inequality, after first declining, tends to increase with average city size. Whether this relationship holds within the Canadian urban hierarchy is a question to be further investigated in a future article.

⁶ We also calculated the Moran’s I independently for Gini coefficients in each census cycle, and all values were positive and significant (ranging from 0.487 to 0.595). Similar results were also found for the Theil index.

⁷ Both fixed effects (FE) spatial lag, and spatial error models (SEM) were initially estimated, and a robust Lagrange Multiplier test was applied to determine which type of spatial interaction effects should be accounted for (the diagnostic favoring the FE SEM modeling approach). Hausman’s specification test also reveals that a FE approach was favorable (see Anselin, Le Gallo, and Jayet 2008; Elhorst 2009).

⁸ We again followed the usual practice of beginning by estimating both a spatial lag and spatial error model before applying the Lagrange multiplier tests to discriminate between the two forms of spatial autocorrelation (Anselin and Florax 1995). In this case, the spatial lag approach is preferred to the spatial error models.