Introduction

Urban–rural divides are large and growing in elections across North America and Europe. Since the 1990s, urban and rural voters in Canada and the United States have increasingly sorted into right and centre‐left parties (Armstrong et al., Reference Armstrong, Lucas and Taylor2022; Gimpel et al., Reference Gimpel, Lovin, Moy and Reeves2020; Mettler & Brown, Reference Mettler and Brown2022; Rodden, Reference Rodden2019; Roy et al., Reference Roy, Perrella and Borden2015). In many European countries, support for radical right parties has proliferated in rural and peripheral areas, while socially progressive and green parties have seen significant growth in cities (Ford & Jennings, Reference Ford and Jennings2020; Huijsmans & Rodden, Reference Huijsmans and Rodden2024).

This growing divide is thought to have intersecting economic, social and political causes. Economically, transformations to advanced industrial economies in recent decades, especially relating to globalization and the advantages of agglomeration in the knowledge economy, have made urban and rural places increasingly distinct demographically, while also creating politically consequential spatial patterns of ‘have’ and ‘have‐not’ places (Ford & Jennings, Reference Ford and Jennings2020; Rodriguez‐Pose, Reference Rodriguez‐Pose2018). Socially, urban–rural divisions in many countries appear to be reinforced by a strong sense of place attachment and a heightened sense of place‐based resentment, particularly among rural voters who perceive that their communities are deprived or overlooked by urban elites (Borwein & Lucas, Reference Borwein and Lucas2023; Cramer, Reference Cramer2016; de Lange et al., Reference Lange, Brug and Harteveld2022; Munis, Reference Munis2020; Trujillo & Crowley, Reference Trujillo and Crowley2022). Finally, in political terms, strategic appeals by elites to urban and rural issues and identities may have also heightened and activated urban–rural political cleavages (Armstrong et al., Reference Armstrong, Lucas and Taylor2022; Ogorzalek, Reference Ogorzalek2018; Taylor et al., Reference Taylor, Lucas, Armstrong and Bakker2024).

While these economic, social and political developments are now receiving sustained attention from political scientists, we know much less about the policy issue bases of the widening urban–rural political divide. Urban–rural differences in voting patterns are often assumed to have attitudinal foundations, but very little research has examined patterns of urban–rural issue disagreement at the individual level over time. With few exceptions, most studies of urban–rural attitudinal divides have focused on a select set of policy issues at a single point in time or have pooled data across multiple years (Kenny & Luca, Reference Kenny and Luca2021; Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023; McGrane et al., Reference McGrane, Berdahl and Bell2017; Pinggera, Reference Pinggera2023; Scala & Johnson, Reference Scala and Johnson2017; Thompson, Reference Thompson2023, but see Huijsmans et al. Reference Huijsmans, Harteveld, Brug and Lancee2021). Although these studies have documented substantial place‐based attitudinal divides on issues such as immigration, gender politics and family values, they cannot tell us if changes in urban–rural policy preferences are associated with growing urban–rural electoral divides. This omission is significant because policy disagreement may play little role in the widening urban–rural political divide, which could instead originate in other factors, such as deepening affective polarization among partisans concentrated in urban and rural locations, better sorting of urban and rural voters into parties that align with their ideological preferences, polarizing place‐based appeals by elites or changes in the electoral salience of specific issues (e.g., Fiorina et al., Reference Fiorina, Abrams and Pope2004; D. A. Hopkins, Reference Hopkins2017; Kenny & Luca, Reference Kenny and Luca2021; Mutz, Reference Mutz2018). To understand if issue disagreement is driving the growing urban–rural divide in support for political parties, we must document not only that urban and rural residents are divided on important policy issues, but also that this disagreement has increased over time.

In this paper, we present the first systematic analysis of urban–rural policy disagreement that spans both multiple election years and issues. We use every policy issue question available in the Canadian Election Study (CES) from the 1990s to 2021 – a period during which Canada, like many other countries, experienced a dramatic widening of the urban–rural electoral divide (Taylor et al., Reference Taylor, Lucas, Armstrong and Bakker2024). In total, we examined 456 unique issue questions (973 total instances) across 10 election studies, covering a very wide range of policy domains, from health and welfare to the environment, gun control, foreign trade and immigration.

Our analysis offers several important insights into the role of policy issue disagreement in fuelling the urban–rural divide in advanced democracies. First, we show that urban and rural residents are indeed strongly divided on a number of policy issues, particularly related to immigration, gun control and policies concerning Indigenous peoples, but they are substantially less divided on employment and welfare state policies. Across nearly all issue domains, we also show that policy divides in Canada reflect the ‘progressive cities’ hypothesis (Douglass et al., Reference Douglass, Garbaye and Ho2019; Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023); urban voters tend to hold more left‐wing or progressive views on issues, while rural voters' views tend to be more right‐wing or conservative.

More significantly, however, we find no evidence that urban–rural issue disagreement has grown over time, even as the urban–rural divide in voting patterns has widened dramatically. We examine ‘macro’ change across all policy issues as well as changes within repeated issue questions and issue domains, consistently finding that urban–rural issue divides, though large, have not grown over the 10 general elections held between 1993 and 2021. Taken together, our findings suggest that the urban–rural electoral divide has an important basis in very real policy issue disagreement, but this disagreement cannot explain why the divide has sharply intensified in recent years. We conclude with a discussion of alternative explanations for this intensifying urban–rural divide and strategies for further investigation.

Explaining the contemporary urban–rural political divide

In their classic study of the development of political cleavage structures and party systems in Europe, Lipset and Rokkan (Reference Lipset, Rokkan, Lipset and Rokkan1967) identify a cultural–territorial cleavage dividing an urban metropole from an agrarian hinterland as foundational cleavage of modern democratic politics. This political divide pitted the interests of rural inhabitants, engaged in agricultural production, against those of urban dwellers and tied to the bourgeoning manufacturing base of cities. Over time, the significance of this urban–rural divide appeared to recede relative to other axes of electoral competition: class and state intervention (Doggan, Reference Doggan, Karvonen and Kuhnle2001; Kenny & Luca, Reference Kenny and Luca2021). Nevertheless, the urban–rural divide, similar to other cleavages believed to have faded amidst the economic modernization of the postwar era, including region and religion, has in recent years reasserted itself with increasing intensity in electoral politics (Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023; Mettler & Brown, Reference Mettler and Brown2022; Rodden, Reference Rodden2019; Rodriguez‐Pose, Reference Rodriguez‐Pose2018; Taylor et al., Reference Taylor, Lucas, Armstrong and Bakker2024).

Scholars have advanced multiple explanations for this resurgence of the urban–rural political cleavage. We organize these explanations into three, overlapping, categories: (1) economic explanations, focused on the diverging economic prospects of urban and rural places over the last three decades; (2) socio‐cultural explanations, focused on identity‐based differences between place types; and (3) political explanations, which include factors such as growing partisan polarization in advanced industrial economies, the strategic activation of place‐based identities and issues by political elites and changes in the electoral salience of material and post‐material (often dichotomized as economic and cultural) issues.

Research connecting economic change to the urban–rural divide has highlighted how economic transformations in the last three decades have concentrated economic opportunity in cities (Glaeser, Reference Glaeser2011; Moretti, Reference Moretti2012; Sassen, Reference Sassen2001; Scott, Reference Scott2008). Cities have become hubs for skilled workers in knowledge‐based industries such as technology, finance and professional services, while rural areas remain more reliant on lower skilled employment in agriculture, resource extraction and manufacturing (Florida, Reference Florida2002; Rodriguez‐Pose, Reference Rodriguez‐Pose2018). Consequently, new immigrants, young people and highly educated workers are increasingly found in cities, while rural places are increasingly home to older and less‐educated populations working in manual labour jobs (Glaeser, Reference Glaeser2011; Sassen, Reference Sassen2001).

The diverging economic prospects of urban and rural areas are thought to shape political preferences both through ‘compositional’ and ‘contextual’ avenues. Arguments about the ‘compositional’ effects of place types capture how differences in the composition of urban and rural populations can produce differences in aggregate political behaviour. These arguments emphasize how, as a result of the growing economic spatial divergence between urban and rural areas, people in the two place types have become more socio‐demographically distinct on characteristics such as age, education and ethnic diversity (Ford & Jennings, Reference Ford and Jennings2020). Because these characteristics are associated with different political preferences, the clustering of these distinct socio‐demographic groups into urban and rural places can generate political division between urban and rural areas (Maxwell, Reference Maxwell2019).

In contrast, research focused on the ‘contextual’ effects of place types elaborates how characteristics of places themselves may contribute to political divisions between different types of places (Cantoni & Pons, Reference Cantoni and Pons2022; Cutler, Reference Cutler2007; Johnston & Pattie, Reference Johnston and Pattie1987; Martin & Webster, Reference Martin and Webster2020). From the contextual perspective, features of urban and rural places, often – but not exclusively – stemming from their economic organization, have an independent effect on political preferences. In this argument, factors such as the relative diversity of a place, the nature of its local economy, opportunities for interpersonal interaction and access to amenities all influence individuals' political attitudes and preferences, thereby generating aggregate differences in urban and rural political behaviour (Cramer, Reference Cramer2016; Gimpel et al., Reference Gimpel, Lovin, Moy and Reeves2020; Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023; Rodden, Reference Rodden2019; Salomo, Reference Salomo2019; Scala & Johnson, Reference Scala and Johnson2017). Thus, two people with otherwise identical demographic characteristics may exhibit different political attitudes and behaviours depending on where they live (Gimpel et al., Reference Gimpel, Lovin, Moy and Reeves2020; Johnston & Pattie, Reference Johnston and Pattie2006).

A second set of explanations, while also situated within the broader contextual effects tradition of research on urban–rural politics, has focused more on the social identity underpinnings of urban–rural divides. This body of research emphasizes that ‘urban’ and ‘rural’ are meaningful social identities for many people. Urban and especially rural residents often express strong attachments to their communities, which means that cultural or economic changes that reshape the relative prestige or power of urban and rural places can fuel a politics of political resentment as rural residents perceive that their communities are economically under‐resourced and culturally disregarded (Arzheimer & Bernemann, Reference Arzheimer and Bernemann2024; Borwein & Lucas, Reference Borwein and Lucas2023; Cramer, Reference Cramer2016; Munis, Reference Munis2021; Rodriguez‐Pose, Reference Rodriguez‐Pose2018; Trujillo & Crowley, Reference Trujillo and Crowley2022). In this research, urban and rural voters are understood to be socially polarized – divided as much, if not more, by their social identities and negative affect towards place‐based out‐groups than by fundamental differences in policy attitudes. The significance of such place identities is evident in studies of voting behaviour in the United States and support for Brexit in the United Kingdom (Cramer, Reference Cramer2016; D. J. Hopkins, Reference Hopkins2010; Munis, Reference Munis2021; Rodriguez‐Pose, Reference Rodriguez‐Pose2018; Sobolewska & Ford, Reference Sobolewska and Ford2020).

A final set of studies has highlighted the role of political elites and political transformations in recent decades. The first of these transformations is growing partisan polarization, which is thought to be occurring for several reasons. One source of partisan polarization that may be fuelling the urban–rural electoral divide is the rise in affective polarization across political systems (Wagner, Reference Wagner2021; Westwood et al., Reference Westwood, Iyengar, Walgrave, Leonisio, Miller and Strijbis2018). Research increasingly highlights the social identity bases of partisanship, pointing to a growing affective (emotional) divide between political groups, with partisans expressing a stronger affinity for members of their in‐group party, and animosity towards out parties (Iyengar et al., Reference Iyengar, Sood and Lelkes2012; Mason, Reference Mason2018). Contemporary political polarization is thought to be particularly potent, in part, because it is reinforced by – and in turn, reinforces – other social identities (Iyengar et al., Reference Iyengar, Sood and Lelkes2012; Mason, Reference Mason2018). One such identity is place identity. In the U.S. context, for example, Jacobs and Munis (Reference Jacobs and Munis2023) show that rural identity predicts a stronger in‐group affect towards the Republican Party, suggesting that the two emotional attachments – place and partisanship – may mutually reinforce one another.

Closely related, political polarization can also occur when voters' issue preferences and partisanship become more correlated (Abramowitz & Saunders, Reference Abramowitz and Saunders2008; Abramowitz, Reference Abramowitz2010; Merkley, Reference Merkley2023). Thus, another explanation for the growing urban–rural electoral divide is that urban and rural residents may increasingly be sorting into progressive or conservative political parties based on their bundling of issue attitudes, even if their disagreement with one another about specific policy issues may not actually be increasing. The different processes underlying partisan and place polarization may further be self‐reinforcing; as voters become more aligned on the basis of their partisan and/or place identities and their ideological preferences, this alignment may further cue other voters to sort into political parties based on their place of residence.

Similarly, political elites have a role in amplifying urban–rural issues and identity differences. In the contemporary context, a number of studies highlight the Republican Party's extensive use of campaign rhetoric that appeals to rural identity and resentment (Cramer, Reference Cramer2016; Jacobs & Munis, Reference Jacobs and Munis2023; Shea & Jacobs, Reference Shea and Jacobs2023). Research has also documented the crucial role of elites in building political coalitions and strategically activating issue and identity concerns among urban and rural voters at pivotal historical moments (Armstrong et al., Reference Armstrong, Lucas and Taylor2022; Ogorzalek, Reference Ogorzalek2018; Rodden, Reference Rodden2019). While these explanations do not deny the underlying compositional and contextual effects that enable urban–rural political divides, they suggest that strategic appeals by political elites can help explain why the strength of the urban–rural cleavage has grown and receded – sometimes very rapidly – at distinct historical junctures. Recent work examining differences in the timing and development of the urban–rural cleavage in Canada, the United States and the United Kingdom shows that political actors have used different issues at strategic moments to gain an advantage among rural voters (Taylor et al., Reference Taylor, Lucas, Armstrong and Bakker2024). In the United States, for example, Republican politicians Goldwater and Nixon initiated a pattern of inciting rural grievance and racial resentment to capture the white South from the Democratic Party. In Canada, conservative political actors since Prime Minister Diefenbaker in the 1960s have also deliberately inflamed anti‐urban resentment while largely avoiding race‐based appeals.

A final potential explanation for the urban–rural political divide is the changing salience of policy issues over time. This explanation builds on research suggesting that growing affluence in Western countries has led voters – particularly those with greater economic security – to de‐prioritize material (economic) issues in favour of post‐material issues, such as those related to the environment, self‐expression, immigration and other ‘cultural’ issues (De Vries et al., Reference De Vries, Hakhverdian and Lancee2013; Gelman, Reference Gelman2009; Inglehart, Reference Inglehart2018; Kriesi et al., Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2008). The increasing salience of these issues has resulted in significant shifts in parties' voter coalitions, as working‐class voters have shifted right on the basis of their greater cultural conservatism (Marble, Reference Marble2024). To the extent that the gap in issue preferences between urban and rural voters is larger for cultural than economic issues – whether for compositional reasons (e.g., rural residents tend to be lower in education and older) or contextual reasons (e.g., rural contexts provide residents with fewer opportunities for interaction with people with diverse backgrounds and lifestyles) – the increasing prominence of cultural issues in politics provides another explanation for the widening political divide between urban and rural voters (Huijsmans & Rodden, Reference Huijsmans and Rodden2024).

Urban–rural divides on public policy attitudes

These economic, social and political explanations for the urban–rural political cleavage, while emphasizing somewhat distinct causal processes, are all underpinned, to varying degrees, by the observation that policy issue attitudes differ between urban and rural places. Reflecting this contention, Figure 1 illustrates these divides for a sample of issue questions from Canada's 2019 federal election.Footnote ¹ The figure illustrates that issue preferences vary across geographic space and that the size of these issue divides is itself quite variable, from issues involving very little geographic disagreement (e.g., assisted dying) to issues that are clearly geographically divisive (e.g., environment vs. jobs).

Figure 1. Vote share and issue divides. District‐level maps of Conservative Party 2019 vote share (top left) and support for five issue statements in the 2019 CES.

In short, accumulating evidence indicates that urban and rural voters in advanced industrial countries hold different preferences on a number of policy issues (Cutler & Jenkins, Reference Cutler, Jenkins, Telford and Lazar2002; Kenny & Luca, Reference Kenny and Luca2021; Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023; Maxwell, Reference Maxwell2019; McGrane et al., Reference McGrane, Berdahl and Bell2017; Scala & Johnson, Reference Scala and Johnson2017). In most of these studies, however, researchers have focused on geographic polarization on a single policy issue, such as immigration (e.g., Maxwell, Reference Maxwell2019), or a carefully selected set of interrelated questions regarding, for example, moral issues or values linked to tolerance and choice (e.g., Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023; Scala & Johnson, Reference Scala and Johnson2017). While valuable, this work paints an incomplete picture of the strength and character of urban–rural policy disagreement across various issue domains and across time.

Our analysis enables longitudinal exploration of the existence and magnitude of urban–rural policy divides across many policy domains. Reflecting theories about the growing electoral importance of ‘post‐material’ concerns, the most consistent focus in studies to date has been on these cultural issues, with an emphasis on broad societal values rather than preferences related to more specific government policies (Kenny & Luca, Reference Kenny and Luca2021; Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023; McGrane et al., Reference McGrane, Berdahl and Bell2017). Luca et al. (Reference Luca, Terrero‐Davila, Stein and Lee2023) provide the most systematic analysis of urban–rural values divides. Across 66 countries, they find substantial urban–rural polarization on questions of family values, gender equality and immigration, especially in high‐income countries. In Canada, McGrane et al. (Reference McGrane, Berdahl and Bell2017) document similarly pronounced differences in values and attitudes between urban and other Canadians on questions pertaining to gender, family, the environment and discrimination against visible minorities. Several studies have also examined urban–rural divisions on discrete questions related to the welfare state and tax preferences; for the most part, they find divides are substantially smaller than for post‐material issues (Kenny & Luca, Reference Kenny and Luca2021; McGrane et al., Reference McGrane, Berdahl and Bell2017). Importantly, for all issues examined to date, to the extent that divides exist, studies consistently find evidence in support of the ‘progressive cities’ hypothesis, whereby urban residents hold more liberal values and positions on policy questions (Douglass et al., Reference Douglass, Garbaye and Ho2019; Luca et al., Reference Luca, Terrero‐Davila, Stein and Lee2023). Building on these findings, we begin with two hypotheses about the substance, magnitude and ideological direction of urban–rural issue divides:

1. 1 Hypothesis Urban and rural residents are divided on policy issues. Divides are more pronounced on cultural issues than on economic issues.

2. 2 Hypothesis Urban residents are more progressive (left wing) on policy issues. Rural residents are more conservative (right wing).

A second, arguably more significant, weakness in existing research on the urban–rural issue divide is the general lack of attention to how these divides have changed over time. The notable exception is the work by Huijsmans et al. (Reference Huijsmans, Harteveld, Brug and Lancee2021), which analyses longitudinal change in attitudes towards immigration, multiculturalism and European integration in the Netherlands, showing that voters in large cities have leaned increasingly to the left. Several studies have also examined shifts over time in attitudes adjacent to policy preferences, such as political trust and satisfaction with democracy (Luca & Kenny, Reference Luca and Kenny2024; Mitsch et al., Reference Mitsch, Lee and Morrow2021; Vigna, Reference Vigna2024).Footnote ² However, as noted earlier, a more systematic analysis of over‐time change in urban and rural policy issue preferences is crucial for understanding how evolving individual‐level divisions on policy issues accumulate into broader urban–rural electoral cleavages.

Hypothesizing about these changes requires attention to how the economic, political and social drivers of the urban–rural divide in vote choice, discussed in the previous section, may reshape individuals' policy preferences. The most straightforward possibility is that urban and rural residents have simply drifted apart on policy questions, with urban residents becoming increasingly progressive and/or rural residents becoming more conservative. This growing attitudinal divergence could stem from any of the compositional or contextual processes described earlier. For example, urban populations may be becoming increasingly liberal as young people, skilled workers, the highly educated and new immigrants with more progressive policy preferences increasingly sort into these places (Ford & Jennings, Reference Ford and Jennings2020). Alternatively, urban contexts may shape individual preferences. The growing diversity of cities may therefore be self‐reinforcing; as cities become more diverse, urban residents are increasingly interacting with ‘different’ others, which itself might foster more liberal attitudes, particularly on questions of diversity and equal rights (Inglehart, Reference Inglehart2018; Vertovec & Cohen, Reference Vertovec and Cohen2002; Wessendorf, Reference Wessendorf2014). If this is the case, then we would expect to see ever‐widening urban–rural divides in policy preferences:

1. 3a Hypothesis The urban–rural divide on policy issues has grown in recent decades.

It is possible, however, that observed urban–rural electoral divides are widening even in the absence of increasing urban–rural issue divides. Partisan‐ideological sorting, discussed in the previous section, can widen political polarization even where voters' actual issue preferences are not diverging. Moreover, if indeed post‐material issues have become more salient than material ones in electoral politics in recent decades, and if urban and rural voters are increasingly voting on the basis of the former rather than the latter, then the growing electoral divide could reflect that voters in the two places are simply more divided on post‐material as compared to material issues (and have been for a long time). Finally, the growing urban–rural electoral divide could be less related to widening issue divides than to increasing affective polarization between the two groups, rooted in feelings of place identity and resentment and fuelled by the alignment of place identity with partisan identity. Such affective polarization could be particularly potent for electoral divides if successfully activated by strategic political elites, regardless of any change in urban and rural voters' issue positions.

Several recent studies, although focused on broader political attitudes rather than specific policy issues, support the possibility that issue divides need not underpin growing urban–rural electoral cleavages. Luca and Kenny (Reference Luca and Kenny2024) and Vigna (Reference Vigna2024) examine changes in Europeans' political trust and satisfaction with democracy over the last two decades. Despite the fact that urban–rural electoral divides in many European countries have grown in recent decades (Huijsmans & Rodden, Reference Huijsmans and Rodden2024), they find only limited evidence of divergence in the attitudes examined between urban and rural Europe. Thus, as an alternative hypothesis, we might expect not to find evidence of change over time:

1. 3b Hypothesis The urban–rural divide on policy issues has not changed in recent decades.

Data and methods

Measuring urban–rural issue divides

To measure urban–rural policy disagreement, we rely on data from 10 CES spanning the 1993–2021 federal general elections. We began by identifying every policy issue question in each CES dataset; following past studies in other countries (Caughey & Warshaw, Reference Caughey and Warshaw2018; Caughey et al., Reference Caughey, O'Grady and Warshaw2019), we interpret policy issue questions broadly, identifying every question where citizens are asked for their opinion on a topic about which the government could conceivably act. Using this criterion, we identified 456 unique questions in the survey data, which were collectively asked a total of 973 times across the 10 election studies. We summarize sample sizes and unique issue questions available in each CES dataset in Table 1.

Table 1. Sample size and number of unique issue questions by year of CES

To organize these 456 questions into policy domains, one team member coded all questions by major topic using the Canadian codebook for the Comparative Policy Agendas Project (for a discussion, see Gauvin & Montpetit, Reference Gauvin, Montpetit, Baumgartner, Breunig and Grossman2019), with all ambiguous coding choices discussed and validated by a second team member. Unsurprisingly, some policy domains that are very common in standard Comparative Policy Agendas documents are much less common in public opinion surveys (e.g., details of government operations), and other domains that may rarely occur in formal government documents tend to arise frequently in surveys (e.g., attitudes on the abolition of the Senate). Therefore, we consolidated several policy domains into combined codes and created more detailed codes for especially popular CES issue topics such as democratic reform and immigration policy. Moreover, because past research in Canada has linked gun control policy to the urban–rural cleavage (Heinmiller & Hennigar, Reference Heinmiller and Hennigar2022), we also created a distinct policy domain code for gun‐related issue questions. This procedure ultimately yielded 20 distinct topic codes; we summarize these codes, along with the number of questions available for each topic and election study, in Table 2.

Table 2. Number of issue questions by topic and year of CES

To test our first hypothesis, we also organized our policy issue questions into a simpler economic and cultural distinction. We categorize each policy question individually, rather than by issue domain, meaning that two questions pertaining to the same domain (e.g., health or immigration) can have different economic or cultural classifications. Although we acknowledge that distinguishing between cultural and economic issues can create something of a false dichotomy among policy issues that are, in fact, highly intertwined, we nevertheless see value in presenting our findings under these headings, as it allows us to situate our findings within existing research that documents larger issue divides for cultural attitudes than economic ones (Kenny & Luca, Reference Kenny and Luca2021; McGrane et al., Reference McGrane, Berdahl and Bell2017). Following past studies (Danieli et al., Reference Danieli, Gidron, Kikuchi and Levy2022; McGrane et al., Reference McGrane, Berdahl and Bell2017), we defined economic items as any questions about government spending, macroeconomic or tax policy, redistribution and social policy. We defined cultural items as any questions about citizenship and immigration, nationalism and regionalism, traditional morality and family values, and law and order. We added a third residual category for questions that did not fit either category, such as questions about Internet voting. Two team members independently coded the 456 questions and then discussed and resolved any differing codes. Our coding decisions for each of the 456 questions are available in a complete table in the online Supporting Information (SM7).

Because we are interested in changes to urban–rural issue divides over time, CES questions that are repeated in multiple surveys are especially valuable. To identify repeated questions, we inspected the policy topics, question wording and response options for each of the 973 questions in the CES data, creating unique variable names for each unique question. We defined a question as identical only when there were the same response options and identical or nearly identical question wording. We then standardized the variable coding for each of the questions and combined the CES datasets into a single comprehensive data file. This process yielded a dataset with 5.3 million rows: one observation for each response to each issue question in each election study.

Urban–rural issue divides: An overview

We begin with Figure 2, which provides a big‐picture summary of the urban–rural issue divides in our full dataset. Each open circle captures the absolute value of the urban–rural divide for a single issue question in a single CES survey; larger values thus indicate more substantial disagreement between urban and rural Canadians. In other words, each circle in Figure 2 represents the estimated relationship, drawn from the multilevel model described above, between issue position and district urbanity for a single issue question in a single election. Because both our urbanity variable and our issue position variables are scaled to range between zero and one, these values can be interpreted as the absolute difference in issue position when comparing respondents in the least urban and most urban districts in each election. As a general interpretive aid, we consider absolute values above 0.05 to represent substantively large differences in policy attitudes.Footnote ⁵ We organize the coefficients by policy topic, with the solid pink squares capturing the average size of the issue divide in each topic. The figure is organized from policy topics with the largest average urban–rural divides to those with the smallest average divides.

Figure 2. Urban–rural issue divides, by policy topic. Absolute value of coefficient estimates for the relationship between district urbanity and issue position. Each grey circle is drawn from a distinct Bayesian multilevel linear model; pink squares are average (absolute) coefficient sizes for each issue type (panel A) or issue domain (panel B).

The results in Figure 2 suggest that urban and rural Canadians are indeed divided in their attitudes on many policy issues. Overall, we find that more than half of the 973 issue questions feature urban–rural divides that are statistically distinguishable from zero.Footnote ⁶ As expected, we also find that these issue divides differ substantially by issue type, with the largest average urban–rural issue divides on cultural issues, followed by economic issues and finally by other issues.Footnote ⁷

The second panel in Figure 2 shows that urban–rural issue divides vary profoundly by policy topic. In the topics located at the top of the figure – guns and gun control, Indigenous affairs and immigration – urban–rural issue divides are consistently very large and distinguishable from zero in nearly every case.Footnote ⁸ At the bottom of the figure, in contrast, urban–rural divides are substantively small in magnitude and often statistically indistinguishable from zero.Footnote ⁹

While the results in Figure 2 provide strong evidence that urban–rural divides are highly variable across policy issues, the substantive importance of these divides is difficult to discern from Figure 2 alone. To gain more purchase on this question, Figure 3 summarizes the distribution of urban–rural policy issue divides compared to other relevant demographic divides. For each variable, we estimated bivariate models, regressing issue attitudes on the relevant independent variable (education, gender and union status) for each of the 973 available issue questions. We recorded the absolute value of the relevant coefficients in each of these models and plotted their distribution for all 973 questions in Figure 3. The figure thus provides additional information on the size of urban–rural issue divides relative to other politically salient divides. In each case, we mark the average absolute value of the relevant issue divide with a vertical line within the distribution.

Figure 3. Distribution of absolute marginal effects. Distributions summarize the estimated (absolute) marginal effects of each variable (urbanity, education, gender, union membership) on issue attitudes across each of the 973 issue questions available. Vertical black lines mark the average absolute value of the marginal effects.

Figure 3 reveals that urban–rural issue divides are not only common but are also substantively important relative to other politically important demographic cleavages in Canadian politics. It is important to note that Figure 3 does not capture the independent effect of each of these cleavages, and there is likely substantial overlap, particularly in the case of urbanity and education. Regardless, Figure 3 shows that urban–rural divides are consistently substantial and larger than divides between Canadians with and without university degrees, men and women and union versus non‐union households.

To complete the first stage of our analysis, Figure 4 summarizes the results of our test of the ‘progressive cities’ hypothesis across policy domains. To carry out these tests, we first regressed each of the 973 issues on citizens' ideological self‐placement scores. This provided us with an empirical measure of which position on each issue question could be characterized as the ‘left‐wing’ or ‘right‐wing’ view; in other words, it provided us with information about which side of the issue question tended to be taken by self‐identified left‐leaning respondents versus right‐leaning respondents. We then compared these ideological divides to our urban–rural divides, recording, for each issue, whether the issue position associated with the ideological left was also the issue position associated with more urban residents in our urban–rural analysis. When these two divides aligned in the expected direction – left/urban versus right/rural – we recorded the issue as supporting the progressive cities hypothesis.

Figure 4. Progressive cities hypothesis. Proportion of issue questions in which the relationship between urbanity and issue attitudes supports the ‘progressive cities hypothesis’, with more urban districts supporting the progressive pole on the issue.

In general, our results very strongly support the progressive cities hypothesis: urban Canadians are, on average, to the left of rural Canadians in 75 per cent of our issue questions. Of course, Canadians are not necessarily ideologically divided for all of the issue questions in our dataset; when we focus on issue questions with a statistically significant ideological divide, we find that the progressive cities hypothesis holds for 90 per cent of these questions. Figure 4 indicates this pattern holds for nearly all policy domains. Foreign trade is the only exception; in this topic area, the majority of issue questions do not align with expectations of the progressive cities hypothesis. This outlier reflects the distinctive character of trade attitudes in contemporary Canadian politics, with higher levels of support for free trade (a traditionally right‐wing position) among the highly educated professional knowledge workers and immigrants who tend to live in cities and more scepticism about free trade (a traditionally left‐wing stance) in rural Canada. With this important exception, we find strong support for the progressive cities hypothesis in our data.

Are urban–rural issue divides growing?

We have established that urban–rural issue divides are substantively large, highly variable across policy domains and generally conform to the progressive cities hypothesis. We now investigate whether these issue divides have changed over time. Figure 5 summarizes our analysis; in the main panel, we plot the absolute value of urban–rural issue divides in each study, with each issue question's urban–rural divide on the vertical axis and election years on the horizontal axis. The blue line summarizes the overall trend in these divides from 1993 to 2021. In the small panels on the right, we provide results for more formal tests of a relationship between urban–rural issue divides and time: the bivariate relationship between issue divides and time (top), the same relationship with fixed effects for each policy topic (middle) and the same relationship with fixed effects for each unique issue question (bottom).Footnote ¹⁰ This third and final model is especially valuable because it focuses exclusively on repeated questions over time, isolating within‐question changes, if any, in urban–rural issue divides. In all three models, we rescaled the ‘year’ variable to range between zero and one; each coefficient in the bottom panels thus captures the expected change in the size of urban–rural divides when we compare our earliest available year (1993) to our most recent available year (2021). Full tables for these models are available in the Supporting Information (SM4).

Figure 5. Change in urban–rural issue divides over time. Each circle summarizes the relationship between district urbanity and issue position for a single issue; each circle is drawn from a distinct multilevel model. The blue line summarizes the overall time trend. Panels summarize the overall relationship between urban–rural issue divides and time in a simple bivariate model (top panel), a model with issue domain fixed effects (middle panel) and a model restricted to repeated issue questions with issue‐specific fixed effects (bottom panel).

All of the results in Figure 5 tell a consistent and surprising story: we find no evidence that urban–rural issue divides have increased in Canada since the 1990s. In the top panel, the blue line is almost completely flat, with no obvious linear (or nonlinear) trend. In the right‐hand panels, each of the coefficients reinforces this message; not only are all three coefficients statistically indistinguishable from zero, but all are very precise null estimates. Even on a maximally generous interpretation, our data are only consistent with an increase of less than 0.01 points in the absolute value of the urban–rural marginal effects – a substantively minuscule effect, considering that our dependent variables (issue positions) range between zero and one. We thus find no evidence of a substantively (or statistically) significant increase in the size of urban–rural issue divides in Canada since 1993.

Robustness tests. Given the importance of our change‐over‐time analysis for our hypotheses in this paper, we carried out several additional tests of the robustness of our null finding. First, we tested for heterogeneity across issue topics by repeating our analysis within issue topics; our results, while necessarily less precise than the main pooled models, are substantively small and statistically indistinguishable from zero for each of the twenty issue topics, indicating that urban–rural issue divides have not substantially increased in any of our issue topics (see SM5 in the Supporting Information). We also assessed pairwise changes for each of the individual cases of repeated questions in our data (each pair in this analysis is a single issue question asked in two different surveys). Across all repeated questions, the average change in the size of the urban–rural issue divide is a tiny 0.0004.Footnote ¹¹

Discussion

In this paper, we examined the extent to which the growing urban–rural electoral divide reflects conflict over policy issues among urban and rural voters. Using 456 unique policy issue attitude questions asked in the past 30 years of CES, we offer what is, to our knowledge, a first‐ever analysis of urban–rural issue divides across a very wide range of policy issue questions rather than a select few issues or domains. This approach also allows us to mobilize repeated questions across multiple surveys to rigorously measure how urban–rural divides on the same issue questions have changed, if at all, over the past three decades.

Our analysis reveals that urban and rural residents do hold distinct preferences on a wide range of policy issues, especially in the general area of cultural rather than economic policy, with urban residents consistently holding positions that are, on average, to the ideological left of their rural counterparts. However, we find little evidence that these urban–rural divides in policy issue preferences have grown substantially in recent decades, even as the urban–rural divide in vote choice has widened significantly. Thus, increasing disagreement over policy direction cannot explain why urban and rural voters have grown apart so dramatically in recent years.

Even so, much remains to be understood about the role of issue disagreement in the long‐term trajectory of urban–rural electoral divides in advanced democracies. While our findings show that issue disagreement cannot account for Canada's widening urban–rural electoral divide, careful case studies and historical analyses have drawn attention to the role of specific policy issues in fostering division in Canada and other national contexts (D. A. Hopkins, Reference Hopkins2017; Taylor et al., Reference Taylor, Lucas, Armstrong and Bakker2024). Interestingly, while the urban–rural divide has widened at similar rates in many countries, the flashpoint issues have often differed – for example, abortion is a much more divisive issue among urban and rural voters in the United States than in Canada (Speer et al., Reference Speer, Loewen and Bertolo2021). These patterns may be the consequence of elite calculation, as political elites in different national contexts identify opportunities for electoral gain by highlighting particular policy issues that divide urban and rural residents.

Nevertheless, some issues, namely, cultural rather than economic policy debates – appear to be associated with much larger urban–rural divides. It may be that our findings reflect that divisive cultural issues (or other issues) have become more electorally salient for urban and rural residents over time, even as issue attitudes themselves have not changed. Political elites, again, likely play a crucial role in determining the extent to which cultural issues have gained salience over economic ones in different contexts. This underlying model of politics, whereby political parties gain traction by successfully appealing to pre‐existing voter attitudes at key moments of social and economic change, is similar to leading explanations for the success of radical right parties in recent decades (Bonikowski, Reference Bonikowski2017; Danieli et al., Reference Danieli, Gidron, Kikuchi and Levy2022). Thus, long‐term analyses exploring when and how cultural issues became particularly important for urban–rural voting patterns and how more specific issues are strategically activated (or deliberately dampened) by political elites to generate electoral support in urban or rural contexts should be a priority for future research.

One puzzle from our findings is why urban–rural issue divides have not widened solely on the basis that residents of urban and rural voters are now more sorted on education, arguably one of the most significant sources of demographic polarization in the 21st century (Grossmann & Hopkins, Reference Grossmann and Hopkins2024; Marble, Reference Marble2024). While answering this question is beyond the scope of this paper, one possibility may be that other patterns of demographic change have, in some cases, had countervailing effects. We speculate that the rapid growth of immigrant populations of non‐European origin in Canadian urban areas, who often possess more conservative views on cultural issues in particular (Besco, Reference Bescon.d.), may be one factor acting as a countervailing force to the growing progressivism of educated voters in urban areas.

Given our findings, future research should also consider other non‐issue bases of the widening urban–rural political divide. If, as our results show, growing issue disagreement is not the basis for the widened urban–rural divide, then we need to explore other possible mechanisms. One possibility is that as conservative and progressive parties have become increasingly uncompetitive in urban and rural places (respectively), they have reduced their organizational capacity and ceased to run their best candidates in these places, hastening their decline in these districts (Armstrong et al., Reference Armstrong, Lucas and Taylor2022; Rodden, Reference Rodden2019). Alternatively, rather than being rooted in attitudinal divides, widening urban–rural division may, in fact, result from deepening affective dislike between urban and rural residents. While recent work has shown that urban and rural residents do harbour substantial affective dislike towards one another in many countries (Haffert et al., Reference Haffert, Palmtag and Schraff2024; Hegewald & Schraff, Reference Hegewald and Schraff2022), it remains unknown how much this affective place polarization has deepened over time. Studies that explore urban–rural divides in affective polarization would be a valuable place to begin this work.

Another possible non‐issue basis for the widening urban–rural divide could simply be clearer ‘sorting’ among urban and rural voters. If urban and rural voters today are more likely than in the past to vote for the party that is closest to their ideological preferences, this could produce widening urban–rural electoral divides even if the magnitude of urban–rural policy disagreement has not changed. Studies of ideological‐partisan sorting in urban and rural places using national election study datasets would offer a valuable first test of this alternative possibility.