Introduction
Contrary to expectations that globalization and integration would diminish the significance of place, geography is today as pivotal as ever for Canadian politics. Researchers continue to document clear geographic divisions in Canadians’ ideological orientations, policy preferences and vote choice (Anderson, Reference Anderson2010; Armstrong et al., Reference Armstrong, Lucas and Taylor2022; Cochrane and Perrella, Reference Cochrane and Perrella2012). Less settled is the question of whether these divisions are better understood as contextual or compositional features of place—that is, the consequence of features of the places themselves, or a reflection of the people who inhabit them (Maxwell, Reference Maxwell2019). This fundamental question of whether places make people or people make places raises the larger question of whether geographic variation itself has any causal explanatory power, or whether documenting differences among populations in different places is primarily a descriptive exercise (Cochrane and Perrella, Reference Cochrane and Perrella2012). The first of these possibilities—referred to as “contextual effects”—implies that features of people's environments directly shape their perspectives and actions. The second proposes that differences observed among people living in different environments primarily stem from distinct characteristics of the populations in these places—or what are often called “compositional effects.”
In this research note, we propose a new typology of place type for use in advancing the study of contextual and compositional effects in Canadian politics. Any meaningful attempt to disentangle contextual and compositional effects depends on making decisions about which geographies are most consequential for people, yet there is considerable disagreement in the literature as to how to delineate Canadian political geography. At present, three approaches are most prevalent. The first and most long-standing historical tradition in Canadian political science demarcates geographic dividing lines using provincial or regional (transprovincial) boundaries (Gibbins, Reference Gibbins1980; Gidengil et al., Reference Gidengil, Blais, Nadeau and Nevitte1999; Simeon and Elkins, Reference Simeon and Elkins1974; Wiseman, Reference Wiseman2007). A second and more recent tradition has highlighted the urban-rural (or urban-suburban or urban-suburban-rural) divide as a salient political division crosscutting traditional regional boundaries (Armstrong et al., Reference Armstrong, Lucas and Taylor2022; Cutler and Jenkins, Reference Cutler, Jenkins, Telford and Lazar2002; McGrane et al., Reference McGrane, Berdahl and Bell2017; Walks, Reference Walks2004, Reference Walks2005), with some variation in how these places are classified.Footnote 1 A third approach has attempted to more inductively identify geographic divides by aggregating and examining various levels of geography in different configurations (Cochrane and Perrella, Reference Cochrane and Perrella2012; Gidengil, Reference Gidengil1989; Henderson, Reference Henderson2004).Footnote 2
Each of these approaches has its benefits, depending on the question at hand, but each also has limitations. The first two approaches both rely on aggregations (province, region, or urban/rural) that risk masking more fine-grained variation. For example, a measure that relies on provincial boundaries implies that a farm owner in British Columbia is more similar to an urban Vancouverite than to a farm owner just across the provincial border in Alberta, while a measure that relies on urban-rural distinctions implies that a resident of one of British Columbia's more remote rural Northern communities is highly similar to a farm owner in the province's south.
Research using the third inductive approach attempts to capture greater variation in geographic differences, often by leveraging data at the level of federal electoral districts (Cochrane and Perrella, Reference Cochrane and Perrella2012; Henderson, Reference Henderson2004). While electoral districts are a crucial geographic setting for political geographers, given their central role in elections, approaches centred on electoral districts also have limitations. Importantly, where federal electoral districts are used as the unit of analysis, the resulting measure of place is to some extent predetermined by politics.Footnote 3 More significantly, this measurement approach can be considered endogenous to politics because it derives geographic groupings from political indicators, instead of first demarcating salient geographic differences and then connecting those geographic boundaries to political outcomes. For example, Cochrane and Perrella (Reference Cochrane and Perrella2012) delineate how—and at what level of aggregation—place shapes politics by mapping out geographic differences in people's attitudes toward government intervention using contextual measures at both the constituency and provincial levels. Henderson (Reference Henderson2004: 603), for her part, delineates distinct Canadian political cultures by aggregating federal electoral districts based on indicators that capture the country's initial fragment cultures, as well as indicators that she determines “currently influence political attitude variation.” However, if the researcher's goal is to understand how place shapes politics, then an ideal measure of geographic differences should—where possible—capture variation that exists independent of politics, not potentially because of it.
Responding to these limitations, in this research note we propose a new approach to typologizing place types for use in the study of political behaviour. Our approach is similar in spirit to the inductive approaches used in Gidengil (Reference Gidengil1989) and Henderson (Reference Henderson2004) but draws on a more fine-grained unit of geography only more recently made available by Statistics Canada, called the Aggregate Dissemination Area (ADA). This approach has several advantages as compared to earlier work. First, ADAs are substantially smaller than the federal electoral districts and census districts used in earlier studies (typically comprising between 5,000 and 15,000 residents), allowing for a much more fine-grained analysis of potential differences across geographic place types. Second, unlike census tracts—previously the best available measure of local geography—ADAs are inclusive of all Canadian residents. Third, Statistics Canada makes available a substantial amount of information characterizing each place type, including information on ADA residents’ average sociodemographics, housing, employment and commutes. This is important not only because it provides us with a rich dataset from which to delineate places, but also because it enables us to include a wide variety of indicators that may differentiate places, rather than just those previously linked to political variation. Finally, ADAs are not delineated along politically meaningful boundaries. Instead, they are assembled from less politically salient geographic units, such as dissemination areas, census tracts and census subdivisions.
Leveraging these features of ADAs, combined with techniques in hieararchical cluster analysis (HCA), we classify Canadian geographies into a specified number of place “types” based on similarities in the aforementioned characteristics. We describe our inductive approach as working “from the ground up,” in that our approach starts by defining Canadians based on belonging to very small geographic areas, only later aggregating places together to the extent that they share characteristics.
Using our approach, we obtain seven distinct place types in Canada. While our typology shows clear evidence that urban-suburban-rural distinctions capture important place variation, it also reveals substantial heterogeneity within these categories—highlighting the added value of our approach. For example, we uncover three types of rural places, which vary substantially in economic prosperity and include one rural group that primarily comprises Indigenous communities. We further uncover two types of suburban areas, which share common features such as moderate density, age and education rates, but also diverge considerably on levels of diversity. Finally, we identify two types of urban places, both with high levels of racial diversity, density and public transportation use, one of which represents the more historical working-class urban communities, and the other the more recent concentration of highly educated professions in cities.
Our approach facilitates a more nuanced delineation of place types than has been possible with previous approaches. This nuance is valuable on its own; however, in the final section of this research note, we also show how our place typology can be useful for assessing the role of context and composition and revealing meaningful variation in political outcomes. Examining party vote shares by place type in the 2021 Canadian federal election (interpolated from polling station returns), we show that Canadian parties’ performance varies considerably not just across urban and rural places, but also across our more defined seven place types. We also show that our place types interact in compelling ways with Canada's historic and ongoing regional electoral divides, and are particularly predictive of voting patterns in Canada's most electorally competitive provinces. Our findings suggest that compositional features of place, although important, are not the only way place shapes electoral politics.
While election results are one outcome we think can be better understood using our place typology, analyses drawing on our place types can shed light on the effects of place for a number of other political outcomes. Such analyses could include an examination of political ideology, or attitudes toward diversity and inequality, to name just a few. To this end, we make publicly available the necessary files for researchers to merge our place types with their own survey data using respondent postal code. It is our hope that researchers will use our place delineation to contribute a richer and more nuanced accounting of place's effects on politics to the already vibrant field of Canadian political geography.
Data and Methods
Our place typology relies on a relatively new set of geographic boundaries called ADAs, first released by Statistics Canada for the 2016 census. As previously highlighted, these boundaries are valuable for several reasons. First, unlike many other fine-grained census geographies, such as the Census Tract, ADAs cover the whole of Canada, which is critical for any attempt to develop a pan-Canadian place typology. Second, except in very low-population regions, ADA populations are very consistent, ranging between 5,000 and 15,000. This consistency makes ADAs more attractive than other possible pan-Canadian geographies, such as municipalities (which are extremely variable in population) or federal electoral districts (which are somewhat consistent in population, but too large to capture finer-grained variation across place types). Finally, Statistics Canada makes a great deal of census data available at the ADA scale, allowing us to build a place typology using a large and diverse set of demographic and economic indicators.Footnote 4
To select our census indicator variables, we opted for an inclusive and inductive approach. We began by gathering all available ADA-level data from Statistics Canada, converting indicators into percentages from raw values whenever appropriate (such as the percentage of the ADA whose mother tongue is French, or the percentage of the ADA who are first-generation immigrants).Footnote 5 All told, we collected 143 indicator variables covering information on ADA population size and density, age distributions, dwelling types, marriage rates, income distributions and averages, language, racial identity, immigration, housing construction, religion, education, labour and employment, commuting patterns and geographic mobility. We provide a complete list of these 143 indicator variables in the supplementary material (Appendix A.1).
With more than 5,400 ADAs and 143 indicator variables, any patterns of similarity or difference that may exist across ADAs in Canada are hidden in the raw data beneath incomprehensible, multidimensional complexity. The task of our cluster analysis procedure was therefore to inductively identify interesting patterns beneath this complexity. To carry out this task, we use HCA, a straightforward, readily interpretable and widely employed algorithm for creating clusters from continuous indicator variables. We preprocessed our 143 indicator variables by dividing each indicator variable into 15 quantiles, carried out principal components analysis on the quantile-transformed indicators and then employed HCA on the first 50 principal components from the principal components analysis.Footnote 6
The HCA algorithm creates cluster solutions ranging from the original number of observations (in which each observation is its own cluster) to one (in which all observations are in a single cluster). After the HCA procedure, the researcher's next task is therefore to select the appropriate number of clusters to retain. This decision involves a mix of theoretical, practical and statistical considerations. Statistically, post-clustering tests allowed us to narrow down to a set of candidate solutions, all of which provided particularly large improvements in fit; identifying these candidates involves tests akin to the scree plots used to select factors after factor analysis. Our tests indicated that five-cluster and seven-cluster solutions performed especially well, and further inspection suggested that the seven-cluster solution offered substantive improvements over the five-cluster solution without adding excessive complexity for interpretation and data visualization. Thus, for both statistical and substantive reasons, we settled on a seven-cluster solution. We provide additional detail on the statistical tests we employed to make this decision in the supplementary material (Appendix A.2).
Because most political science research on elections and voting behaviour relies on public opinion survey data, we wanted to enable researchers to easily connect our place typology to their survey data. Unfortunately, ADAs are not yet included in the Government of Canada's postal code conversion files. We thus used GIS data from the postal code conversion file (longitude and latitude points) to map each of Canada's 876,000 postal codes into their corresponding ADAs. While we consider this conversion file to be a temporary measure, pending an official government-issued conversion file, we make our conversion table available for other researchers in our data repository. This makes it easy for researchers with six-digit postal codes to attach their survey respondents to ADAs in general and to our place types in particular.
Canadian Place Types: An Overview
While HCA is valuable for identifying a relatively small number of internally similar latent place “types” from our census data, it does not interpret these latent types for us—much like factors in factor analysis, researchers must interpret the latent clusters and provide each cluster with a name that captures its salient characteristics. To interpret our seven place types, we begin with two figures. First, Figure 1 summarizes each place type on a subset of the census indicators that we used in our cluster analysis. For each indicator, the shaded grid displays whether the place type (the seven columns) tends to be low (the brown shades), moderate (white), or high (the green shades) on a given census indicator. In the first row of the Figure, for instance, notice that the first place type tends to have moderately high levels of vehicle-based commuting, and the second place type is even higher on this indicator. In contrast, the third, sixth, and seventh place types are characterized by very low rates of commuting by car or truck. Working through each of the census indicators and comparing across place types helps to clarify what it is that makes each place type distinct from the others.
Figure 1. Summary of Place Type Characteristics. Columns are place types; rows are illustrative census characteristics. Brown colours indicate that the place type is below the median for a particular characteristic; green colours indicate that the place type is above the median for the characteristic.
Given that our clusters represent geographic place types, a map is also a helpful way to interpret the meaning of the seven clusters. In Figure 2, we summarize the distribution of place types across Canada, with each of the seven types shaded in a distinct colour. To aid in interpretation, we also provide inset maps of eight metropolitan areas across Canada at the bottom of the map. Readers should note that the place type colours in the map in Figure 2 correspond to the colour labels at the bottom of Figure 1, as well as the place type labels we use in Figure 4 below.
Figure 2. Map of Place Types.
Taken together, Figures 1 and 2 allow us to make considerable progress in interpreting the meaning of our seven clusters. To begin, we have labelled the first three columns in Figure 1 as “rural”—each of these clusters has very low population density, and all three are clearly most prominent in the rural and remote portions of the map in Figure 2. However, while all three clusters can be described as rural, there are also important differences across the three place types.
The first two clusters, “rural 1” and “rural 2,” tend to have high rates of car and truck commuting, low levels of racial diversity and a high proportion of Christian and white residents. Both are also characterized by single-detached housing, agriculture and resource-sector employment and somewhat older populations. What distinguishes the two clusters, however, is income: rural 1 places tend to contain a large number of people in the lowest income decile and very few in the highest decile, as well as very low overall employment rates, whereas rural 2 places are wealthier and enjoy high employment rates. We might therefore describe the first place type as lower-income rural and the second place type as higher-income rural. In comparison to both of these types, our third category, rural 3, involves low rates of commuting, a very high proportion of Indigenous identifiers and is demographically much younger than the other rural types. Inspection of the map in Figure 2 in combination with these demographic characteristics makes it clear that most of the “rural 3” places are First Nations reserves.
Moving further across the columns in Figure 1, the next two place types involve recent housing construction (1981–1990, 1991–2000 and after 2000), moderate levels of population density, moderate to high rates of education and moderate median ages. These demographic characteristics, combined with the geographic distribution of these types, suggest that these columns capture what most would consider “suburban” parts of Canada, with purple “suburban 1” place types capturing more distant suburbs, and yellow “suburban 2” place types capturing inner suburbs. Once again, however, there are very important differences between these two types as well. While both place types are relatively wealthy, “suburban 2” places tend to be considerably more racially diverse, less white and are home to far more immigrants than “suburban 1” places. Unsurprisingly, given their geographic locations, these places also tend to involve higher rates of transit-based commuting. When we inspect the map, we can see that the cluster analysis also labels some obviously non-suburban places into what we have called the “suburban 1” category, including portions of the northern Prairies, the Yukon Territory and portions of Alberta along the Canadian Rockies. This speaks not only to the need for caution in relying too heavily on the labels we apply to each type but also illustrates the deeper value of the inductive cluster-analytic approach, revealing similarities between places that might otherwise go unnoticed. In this case, some remote and northern regions of Canada resemble suburban parts of Canada in shared education levels, common patterns of employment in the health, education and government sectors, and an increased tendency, when compared to the “suburban 2” places, to have residents who work in natural resources and agriculture. It might therefore be reasonable to treat the “suburban 1” category as belonging somewhere between places we would typically consider rural and those we would consider suburban.
The two remaining types, in the final columns of Figure 1, are the ADAs in the heart of Canada's most densely populated metropolitan areas. These are recognizably urban places and involve above-average rates of transit commuting, high levels of racial diversity, large quantities of pre-war housing, youthful populations and very high population density. Here, too, there are important differences as well, most notably in the areas of income and employment. While both urban place types contain high numbers of low-income residents, the “urban 2” places also contain a great many high-income residents as well. In these “urban 2” places, in other words, income inequality is very high, with large numbers of low-income Canadians living alongside large numbers of high-income Canadians. The two urban place types are also very different in terms of education and employment sector; those who live in the green “urban 1” places tend to work in service and manufacturing, while those in pink “urban 2” places are concentrated in professional occupations, academic and scientific research, government employment and arts and culture. In general, then, the yellow regions in the map capture urban residents who are employed in the service and manufacturing sectors, while the pink regions capture Canada's “urban knowledge workers”—highly educated and diverse young professionals.
We draw two main lessons from this interpretive exercise. First, the results suggest that our inductive, socio-demographic approach to place type clustering, using Canada's new ADA geographies as our unit of analysis, yields latent place types with strong face validity, while also revealing interesting patterns of similarity across places and a high degree of geographic “resolution” to distinguish among place types even within specific cities, federal electoral districts or geographic regions.Footnote 7 Second, our results indicate that while higher level distinctions among urban, suburban and rural Canadians are indeed appropriate for some purposes, each of these categories contains important internal variation and benefits from finer-grained analysis as well. Distinguishing among these place types allows us to better understand how the socio-economic composition of the Canadian public varies not only across, but also within, our urban, suburban and rural communities.
Place Types and Their Politics
We now turn to the political implications of these place types. To what extent do the place types we have identified correspond to recognizable patterns of party support or electoral competition? Are some place types more politically competitive than others? And how consistent are these patterns across Canadian regions?
Figure 3 offers the beginning of an answer to these questions. To construct the Figure—and all of the analysis we carry out in this section—we began by using a geographic estimation procedure called areal weighted interpolation to estimate party vote share within each ADA in Canada, using polling-station-level election returns from the 2021 Canadian federal election.Footnote 8 This provides us with an estimate of 2021 party support aggregated to the boundaries of each ADA. We can then use these interpolated estimates to calculate the distribution of party vote shares (along with other quantities) by place type. In Figure 3, we summarize this distribution along with each party's average vote share in each place type. These averages are labelled within each panel and marked with a vertical black line.
Figure 3. Distribution of 2021 Vote Share by Party and Place Type. Each panel summarizes the distribution of the party's estimated vote share at the ADA scale by place type. Vertical lines and percentages are average vote shares.
The distributions in Figure 3 suggest that Canadian parties’ performance does indeed vary considerably not only across “macro” place types, such as urban and rural, but also across our seven more specific place types. The Liberal Party, for example, performs especially well in two place types: suburban 2 (the more diverse, immigrant-heavy inner suburbs) and urban 2 (professionalized knowledge-worker places). The Liberal Party performs moderately well in lower-income, service-oriented urban places and is also competitive in some Indigenous rural (“rural 3”) ADAs. It performs poorly in both rural place types and in the less diverse outer suburbs.
The Conservative Party, while contrasting starkly with the Liberals, is by no means a simple mirror image of the Liberal Party. The Conservative Party performs best in the wealthier rural areas (orange, or “rural 2” in the map in Figure 2) that stretch across the Prairies, southwestern Ontario and portions of Eastern Canada, as well as the distant suburbs (purple in the map, or “suburban 1”) that surround Canada's metropolitan regions. The Conservatives perform moderately well in lower-income rural places and some of the more diverse inner suburbs. Only in Indigenous rural areas (“rural 3”) and in the heart of major cities do the Conservatives perform especially poorly.
Finally, the distributions for the smaller parties—the NDP and the Bloc Québécois—support recent findings on place-based electoral divides in Canada. Aside from Indigenous rural areas, where the NDP performs well, the party's performance is remarkably unvarying across place types. This consistency speaks to the “double disadvantage” faced by the NDP under Canada's current electoral system, with a relatively even concentration of vote share not only at the regional level but also across place types within regions (Armstrong et al., Reference Armstrong, Lucas and Taylor2022). The Bloc Québécois, in contrast, enjoys its strongest support in rural Quebec and the more distant suburbs; interestingly, unlike the Liberals and Conservatives, the Bloc's performance is not especially variable across the two main rural place types, indicating that the compositional differences that distinguish these place types are less important as predictors of vote choice for the Bloc Québécois than for the two major parties.
Of course, Canadian political parties’ electoral performance varies dramatically across Canadian regions, and the pooled results in Figure 3 hide this variation. Figure 4 therefore provides a breakdown of party performance by place type and Canadian region. It shows the calculated percentage of ADAs for each place type “won” by each party in each region (based on receiving the highest interpolated vote share in an ADA) in the 2021 Federal election.
Figure 4. Probability of Winning ADA by Region and Place Type. Each bar is the calculated percentage of ADAs “won” by each party in the 2021 Federal election, using interpolated vote shares. Columns (and bar colours) are distinct parties; rows are distinct place types.
The results in Figure 4 reinforce the impression from Figure 3 that party support does vary substantially across place types in Canada, but the Figure also reveals that the relationship between place type and party support is also powerfully shaped by region. In both “rural 1” and “rural 2” places, for example, the Conservative Party dominates, except in Atlantic Canada, where the Liberal Party performs well. Relatedly, both the Liberals and the NDP perform quite poorly in the first two place types, except in British Columbia, where the NDP performs quite well in some parts of the rural interior. Political parties’ rural performance, in short, is partly conditional on region.
Figure 4 also adds nuance to our interpretation of each party's suburban and urban performance. As we noted earlier, the Conservative Party is stronger in the less diverse outer suburbs (suburban 1) than in the more diverse inner suburbs (suburban 2). Yet the Conservative Party in fact does extremely well in the inner suburbs on the Prairies and is quite competitive in some inner suburbs in British Columbia as well. In urban areas, the Liberal Party completely dominates both “urban 1” and “urban 2” places in Atlantic Canada, Quebec and Ontario but performs much less well in both place types on the Prairies. In British Columbia, lower-income “urban 1” ADAs are a battle between the Conservatives and the NDP, while knowledge-worker “urban 2” ADAs involve competition between the NDP and the Liberals.
These results suggest that the place types we have identified are very important for understanding Canadian voting patterns and election outcomes, but especially when understood alongside Canada's persistently regionalized party competition. In regions that have recently been dominated by a single party—such as Atlantic Canada for the Liberals and the Prairie Provinces for the Conservatives—place types are less important for explaining election outcomes. In more competitive regions, such as Ontario and British Columbia, variation across place types is much more strongly predictive of voting patterns.
This region-specific place-type variation extends to electoral competition as well. In Figure 5, we conclude our results by summarizing median margins of victory, at the ADA level, across place types and regions—in other words, the Figure summarizes the typical level of competitiveness to be found in each place type, based on our interpolated vote share results from the 2021 election. We plot pooled results across all regions in the top panel, followed by region-specific results. The place types are organized from least competitive (that is, highest margin of victory) to most competitive. Overall, we find that “rural 3” (Indigenous rural) places tend to be least competitive, followed by “rural 2” places (orange in the map). Both suburban place types are at the bottom of the Figure, confirming the widespread view that suburban Canada represents the electoral battleground in recent elections.
Figure 5. Margins of Victory in ADAs by Place Type and Region. Estimated margin of victory by ADA place type overall (top panel) and by region (remaining panels). Higher values indicate that elections in the place type tend to be less competitive.
Once again, however, we find interesting variation across regions. In Quebec, for example, the diverse inner suburbs are dominated by the Liberals and are not competitive, but outer suburbs remain a major battleground. On the Prairies, margins of victory tend to be higher across the board, reflecting the Conservative Party's dominance in that region, but the most competitive places are in fact in the heart of the region's major cities, such as Calgary, Edmonton and Winnipeg. Thus, while our suburban place types tend to be the most competitive across regions, we also see substantial variation in electoral competition both across urban, suburban and rural places and, in some regions, within the place types.
Taken together, the results in Figures 3, 4, and 5 reinforce the value of digging beneath the coarse-grained categories, such as “urban” and “rural,” to explore place-based politics in Canada. They also suggest that the importance of these finer-grained distinctions itself varies across Canadian regions. In Atlantic Canada, Quebec and Ontario, for example, we would lose very little by collapsing the “urban 1” and “urban 2” types into a more general “urban” place type, at least from the perspective of party support and competition, because the patterns appear to be consistent in those regions across the two urban place types. In British Columbia, however, political competition is dramatically different in “urban 1” and “urban 2” places. Much the same is true of our distinction between “suburban 1” and “suburban 2” types, which are especially important for understanding variation in party support in Quebec, Ontario and British Columbia, and less critical in Atlantic Canada and the Prairie provinces. One important conclusion of our regional analysis is thus not only that the relationship between place type and party support does vary across regions, but also that the value of the more detailed place typology we employ itself varies depending on the region whose party support and election outcomes we aim to explain.
Discussion and Conclusion
In this research note, we leveraged Statistics Canada's newly demarcated unit of census geography, the ADA, to develop a new classification of Canadian geography for use in the study of political behaviour. Using ADAs as our initial unit of geography allowed us to go beyond traditional urban-rural (or urban-suburban-rural) dichotomies, to uncover important—and to date under-explored—sources of variation that further differentiate urban, rural, and suburban place types. Our analysis produced seven distinct place types, including three unique types of rural places, and two of each suburban and urban place types.
This variation should be of inherent interest to political geographers, but we also show that it can be exploited to gain insight into key questions in Canadian politics. As one example, our analysis provides insight into the long-standing question of whether compositional or contextual place effects exert a greater influence over politics. On the one hand, our analysis shows that places that share compositional similarities on dimensions such as average education, employment type and racial diversity share similar voting behaviour. However, compositional explanations also have their limits, as place types that are strongly predictive of voting for a given party in one region of Canada do not always predict voting for that party in another region. These findings suggest that compositional explanations of election outcomes in Canada may be more powerful for explaining within-region variation than between-region variation, as regional identities, historical trajectories and specific kinds of place identities matter differently across regions. In other words, our analysis supports the conclusion that living in a particular place with a particular set of socio-demographic characteristics increases the probability of supporting one or another party, but that this is also likely refracted through history and social identity.
As a second example, insights obtained from our typology can help us better understand why place-based political appeals have arguably been more muted in Canada than other contexts such as the United States. We find that although the Conservatives generally do well in rural areas, and the Liberals in urban ones, both parties have significant support in their “out-places” in other parts of the country—the Conservatives in urban areas in the Prairie provinces and parts of British Columbia, and the Liberals in rural areas in Atlantic Canada. In the context of centralized elections, this may reduce the incentive for both major parties to pursue strong place-based appeals.
Many other questions of similar importance to the study of political geography can likewise be better understood using our typology of place types. Such questions include how divided are voters in the different place types on specific policy issues, how issue divides based in place type compare to divides based in region and where across the spectrum from rural 1 through to urban 2 we see the sharpest punctuation on different issue questions. In constructing a new characterization of Canadian place types, it is our aspiration that our typology will be used by researchers to answer questions such as these, in the process deepening our understanding of place's impact on Canadian politics.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0008423924000842
Data and Replication Files
Data and replication files to reproduce the analysis in this article are available at https://doi.org/10.7910/DVN/XNAIUM. A complete database of ADA place type codes, along with a file to match postal codes to our ADA place types, is available at https://doi.org/10.5683/SP3/9C5TU0.
Acknowledgements
This project was supported by SSHRC (PDG 890-2018-0019). We are grateful to Dave Armstrong for helpful advice in the early stages of this project.
Competing interests
The authors declare none.
Open access
