Introduction
Identity has been at the centre of recent accounts of contemporary politics in advanced democracies, with both academic and popular commentary highlighting the growing political mobilization of nationalist, racial, gender, sexual, religious and other identities by both the left and right (for discussions, see Alcoff et al., Reference Alcoff, Hames-Garca, Mohanty and Moya2006; Béland, Reference Béland2017; Bernstein, Reference Bernstein2005). These patterns of mobilization are also thought to be reflected, at least in part, in changing patterns of identity politicization among voters, with significant political implications. In the United States, the 2016 presidential election revealed an electorate deeply divided on racial and religious questions and sorted into Republican and Democratic parties along lines of race, sexuality, religion and other demographic identities (Egan, Reference Egan2020; Mason, Reference Mason2018). In Western Europe, scholars have similarly documented how the shifting politicization of group identities along education, geography and ethnicity lines has caused major disruptions in these countries’ electoral politics, undergirding a number of shifts such as the emergence of new left parties and the growth in support for the radical right (Bornschier et al., Reference Bornschier, Husermann, Zollinger and Colombo2021; Ford and Jennings, Reference Ford and Jennings2020).
These developments have led some popular and scholarly commentators to raise concern that the fragmentation of polities has entrenched an “us” versus “them” mentality that impedes citizens from working toward common political goals (for a discussion, see Abrams et al., Reference Abrams, Sides, Tesler, Vavreck, Richeson and Fukuyama2019). Of course, such arguments have also been met with criticism; among other critiques, scholars contend that identity has always been fundamental to politics and that it is too often only the powerful, who benefit from the status quo, that protest the efforts of previously marginalized groups to gain space in the political arena (Alcoff et al., Reference Alcoff, Hames-Garca, Mohanty and Moya2006; Béland, Reference Béland2017). Others also point out that people's attachments to groups with more narrowly defined memberships need not come at the expense of more universal attachments to each other (Abrams et al., Reference Abrams, Sides, Tesler, Vavreck, Richeson and Fukuyama2019).
In Canada, it is not evident how these patterns have developed. On contemporary issues that structure divisions in other countries’ electorates, such as immigration and multiculturalism, Canadians often appear to be less divided (Bloemraad, Reference Bloemraad2012; Triadafilopoulos and Taylor, Reference Triadafilopoulos, Taylor, Samy and Duncan2021). At the same time, the Canadian political landscape has arguably always been defined by a complex politics of identity. As a country that former Prime Minister Joe Clark referred to as a “community of communities” (Vipond, Reference Vipond1993), Canadian politics has long contended with an intricate array of salient group memberships, including French Canadian and Indigenous minority nations, multi-ethnic immigrant groups and strong regional attachments. Meanwhile, other identities that were elsewhere thought central to politics—most notably, class—have historically been considered to be more muted in the Canadian political arena (Pammett, Reference Pammett1987).
This article takes stock of the Canadian landscape of social identities and their consequence for politics with a specific focus on redistribution. A core worry for those expressing concern about identity's role in contemporary politics is that people's tendency to define themselves along narrower identity lines threatens the redistributive state, undermining the development of broad-based coalitions to advance redistribution and/or the cross-class social solidarity thought necessary to sustaining support for redistribution, even among those who may not directly benefit (e.g., Lilla, Reference Lilla2017; Fukuyama, Reference Fukuyama2018b). This concern connects to empirical work showing how shared commonality rooted in ethnicity, race, and region can influence redistributive preferences, engendering people to prefer to redistribute to members of their shared groups rather than those who most need it (Cavaillé and Trump, Reference Cavaillé and Trump2015; Finseraas, Reference Finseraas2012; Lupu and Pontusson, Reference Lupu and Pontusson2011; Luttmer, Reference Luttmer2001; Wong, Reference Wong2010).
Considering these claims, this study is therefore motivated by two exploratory research questions. First, which group memberships form the basis of Canadians’ perceptions of shared identity with each other? And second, do Canadians’ preferences for redistribution depend on these shared memberships? While the first question asks about the sources of Canadian social identity against the shifting backdrop of contemporary politics, the second question asks about discrimination in the allocation of resources as one potentially important political consequence of group divisions.
To examine these questions, this study draws on two conjoint experiments fielded as part of the 2019 Canadian Election Study (Stephenson et al., Reference Stephenson, Harell, Rubenson and Loewen2020). Both experiments present survey respondents with the profiles of two hypothetical Canadians who vary on a number of dimensions of identity and need, including income, education, wealth (homeownership), employment status, province and place type, gender, ascriptive identity (race/ethnicity, language, or religion) and several other characteristics. In Experiment 1, to assess perceptions of shared identity, respondents were asked to select the profile of the hypothetical Canadian with whom they feel they have more in common. While research often focuses only on single aspects of people's group memberships in isolation, overlooking that people's group identities are invariably plural, this design allows for the measurement of the relative strength of Canadians’ multiple group identities. In Experiment 2, examining redistribution, respondents were again presented with profiles of hypothetical Canadians who vary on the same dimensions of identity and need but were asked to select the Canadian to which they would prefer to redistribute a hypothetical expanded government tax benefit. This experiment allows for a comparison of how Canadians weigh their narrower group memberships against more universal concern for recipient need in making judgments about redistribution.
Results suggest that Canadians’ perceptions of shared identity are often more strongly rooted in indicators of social class position (income, education and wealth) than in regional or ascriptive identities. However, shared group memberships are overwhelmingly secondary to perceptions of need when Canadians make assessments about who should benefit from redistribution; across memberships, the effect of shared group belonging is marginal as compared to the effect of the recipient having a low income. While it is important to note that Experiment 2 does not directly test if perceptions of shared identity influence redistribution, findings that certain group memberships strongly predict Canadians’ perceptions of shared identity in Experiment 1, but not redistributive preferences in Experiment 2, suggest that the shared identities examined in this study are not a strong source of redistributive attitudes in Canada. Thus, while a number of different group memberships condition Canadians’ perceptions of shared identity, these narrower group memberships have not displaced more universal considerations in the political context of redistribution in Canada. This article ends with a discussion of the significance of these findings for understanding social identity and its political implications in Canada.
Social Identity and Redistribution
In political science, identity has often been invoked as a key explanatory variable to understand political behaviours ranging from voters’ issue positions to partisanship, and political participation (Achen and Bartels, Reference Achen and Bartels2016; Campbell et al., Reference Campbell, Converse, Miller and Stokes1960; Fowler and Kam, Reference Fowler and Kam2007; Mason, Reference Mason2018). Research in this vein has drawn on insights from social identity theory, which highlights how people satisfy the fundamental human desire for belonging through group attachment. According to social identity theory, this need to belong motivates people to attach emotional significance to their objective group memberships, in the process, incorporating them into their subjective sense of who they are—their social identities (Tajfel, Reference Tajfel1981). Social identity further fulfills the human need for self-esteem, which people derive from comparing the social standing of their groups with other groups and positively distinguishing their in-groups (Tajfel and Turner, Reference Tajfel, Turner, Austin and Worchel1979; Tajfel, Reference Tajfel1981).
Although identity has long been an important variable in political science, social identity-based explanations of politics have gained newfound traction with more recent political developments in established democracies, including the transformation of left parties and the upsurge of the populist radical right in Europe and the United States. Looking at the left, scholars point to the growing politicization of issues such as gender equality, lifestyle choice and migration (Bornschier et al., Reference Bornschier, Husermann, Zollinger and Colombo2021). The radical right, too, has, in a very different way, also extensively employed identity-based discourse by leaning on ethno-nationalist rhetoric (Bonikowski, Reference Bonikowski2017). Citizens themselves also appear to be increasingly sorted into different parties based on their social identities, with Democrats and Republicans in the U.S. and the new left and far right in Europe appearing increasingly distinct on sociodemographics, such as ethnicity and race, urban-rural geography and education. Importantly, research shows not just that people are divided along these lines, but more fundamentally, that these divides are often underpinned by people's more deeply held social identities (Bornschier et al., Reference Bornschier, Husermann, Zollinger and Colombo2021; Mason, Reference Mason2018).
For some academic and popular commentators, these political developments threaten cohesion and solidarity in established democracies. Fukuyama (Reference Fukuyama2018a, 93) argues that “democratic societies are fracturing into segments based on ever-narrower identities, threatening the possibility of deliberation and collective action by society as a whole.” In particular, he expresses concern about the effects of this fragmentation on redistribution. While it is important to challenge the normative conclusion that is sometimes drawn from such arguments that society is suffering because long-marginalized groups are gaining influence in the political process (Abrams et al., Reference Abrams, Sides, Tesler, Vavreck, Richeson and Fukuyama2019; Alcoff et al., Reference Alcoff, Hames-Garca, Mohanty and Moya2006; Béland, Reference Béland2017), the argument that certain social cleavages can undermine popular support for redistributive politics is not new. A large literature, most often focused on ethnic and racial cleavages, has shown that diversity can impede redistribution (although this conclusion is far from universal; for a discussion of this literature, see: Johnston et al., Reference Johnston, Banting, Kymlicka and Soroka2010).
Motivated by these observations about the shifting lines of identification and the potential consequences of these shifts for a core function of government—redistribution—this article asks two questions in the Canadian context. First, which group memberships are most important for Canadians’ perceptions of shared identity? Second, do these group memberships determine Canadians’ preferences for redistribution? These questions are discussed in turn in the next sections of this article.
Social Identity in Canada
The first question this article asks is: which group memberships matter most for Canadians’ perceptions of shared identity? The revived attention to social identity in comparative political science research has enhanced our understanding of contemporary politics, but patterns of identity division vary across countries, and we should anticipate divides in Canada to follow their own unique pattern. As a multinational and multicultural country, Canada is comprised of multiple distinct national and cultural groups, including English and French Canadians, Indigenous Peoples and longstanding and more recent immigrants, and the observed salience of many of these identities has led to the foregrounding of these groups and their dynamics in examinations of social identity in Canada (Lalonde et al., Reference Lalonde, Cila, Yampolsky, McKeown, Haji and Ferguson2016). Social identity theory suggests that the cohesion of these groups may be shaped, at least in part, by perceptions of threat to the in-group by out-groups (Brewer, Reference Brewer2007; Huddy, Reference Huddy, Huddy, Sears and Levy2013). For example, Québécois identity is heavily shaped by Francophone Quebecers’ perceptions of threat to their identity, attributed to forces such as the dominance of the English language, declining birthrates among French Canadians and immigration (Lalonde et al., Reference Lalonde, Cila, Yampolsky, McKeown, Haji and Ferguson2016; Turgeon and Bilodeau, Reference Turgeon and Bilodeau2014). Indigenous identity is similarly strengthened by threats to territorial rights from the dominant settler population (Lalonde et al., Reference Lalonde, Cila, Yampolsky, McKeown, Haji and Ferguson2016; Wilkes, Reference Wilkes2006). Finally, a prominent concern in the literature in the last decade is the extent to which Canada's changing ethnic composition, linked to changing immigrant patterns, might fuel in-group identification and intergroup hostility (Banting and Kymlicka, Reference Banting and Kymlicka2017; Hyman et al., Reference Hyman, Meinhard and Shields2011; Johnston et al., Reference Johnston, Banting, Kymlicka and Soroka2010).
Considering the intricate web of identities that coexist in Canada, it is thus not surprising that Canada has been argued to be a particularly “fertile ground” for social identity theory (Lalonde et al., Reference Lalonde, Cila, Yampolsky, McKeown, Haji and Ferguson2016). And while recent academic and public debates have renewed concern that narrower group identifications may impede citizens from sustaining more broad-based identities, Canada has long grappled with the political exigencies of reconciling and accommodating these multiple—at times conflicting—identities in the country (Kymlicka, Reference Kymlicka2003). On the one hand, Canada's multinational foundations have fostered perpetual anxieties about the trade-offs between narrower group belonging and broader community solidarity. On the other hand, as Harell et al. (Reference Harell, Banting, Kymlicka and Wallace2021, 3) argue, “precisely for that reason, the country has made concerted efforts over time to build a more multicultural conception of nationhood that accommodates multiple identities and diverse ways of being Canadian.”
Beyond multiculturalism and multinationalism, this study is more generally interested in identifying which social identities may lie behind observed political divisions in Canada. Another likely candidate for strong social identity in Canada is regional identity. Research on Canadian voting behaviour has long highlighted the persistent strength of the regional cleavage, particularly in contrast to the perceived weakness of the class cleavage (Kay and Perrella, Reference Kay, Perrella, Kanji, Bilodeau and Scotto2012). However, while regional divides are likely rooted at least in part in social identity, this possibility has more often been assumed than directly tested. One exception is work by Donnelly (Reference Donnelly2020), which measures Canadians’ perceptions of “linked fate” (the idea that one's life chances are tied to the outcomes of their groups); while he finds Canadians report high levels of linked fate with their regions, he also finds similar levels of linked fate when examining other identities, such as class.
More broadly speaking, social identity implies more than just objective group membership; it involves the integration of that group into one's identity or sense of self (Tajfel and Turner, Reference Tajfel, Turner, Austin and Worchel1979; Tajfel, Reference Tajfel1981), and stark political and policy divisions among groups may be present—based, for example, in material concerns—even where these groups do not constitute salient social identities for their members (Huddy, Reference Huddy, Huddy, Sears and Levy2013). Thus, while we have substantial evidence of group divisions in political behaviour in Canada—studies show Canadians are politically divided along a number of group lines, including province and region, urban-rural geography and education (these divisions are discussed in detail below)—we cannot deduce from the presence of these divides that they are rooted in a deeper sense of group identity. This is important because where political divides are in fact reinforced by social identity, they are more likely to incite behaviours such as discrimination in favour of one's in-group over out-groups and to become entrenched as political cleavages (Helbling and Jungkunz, Reference Helbling and Jungkunz2020; Mason, Reference Mason2018). The strength and persistence of some Canadian political divides suggest they may be fuelled by more deeply held identities, but existing studies generally have not tested if these divisions have social identity bases, or their strength relative to one another.
Redistribution
The second question this study asks is: to what extent do Canadians’ group memberships influence their preferences for redistribution? Those who express concern about growing identity divides often argue that citizens’ attachments to their narrower groups diminish support for redistributive politics. Such arguments draw on a large body of research connecting group identity to redistributive preferences through a variety of different mechanisms. Arguments can loosely be sorted into two camps: those who highlight perceptions of self-interest, and those who highlight altruist or solidarity mechanisms. For those who focus on perceptions of self-interest, one argument that is made is that identity divides fracture working-class coalitions that would otherwise mobilize to fight for redistribution. In this view, ethnic and other identity divides inhibit workers from developing class consciousness around their mutual interests (e.g., Alesina and Glaeser, Reference Alesina and Glaeser2004).
Others focus less on class struggle, arguing instead that identity divides can diminish a sense of shared social solidarity. Solidarity can operate at different subgroup levels, but research has often focused on solidarity at the macro level (i.e., the nation), arguing that diversity can erode national identity or societal bonds, which are thought to foster a shared sense of obligation or concern for community that inclines citizens to support redistribution even where they are not direct beneficiaries (for a discussion of this argument, see: Banting and Kymlicka, Reference Banting and Kymlicka2017). Studies often conceive of such solidarity as being altruistic, although it may also be rooted in self-interested motivations. Studies that emphasize altruistic behaviour suggest that people will feel greater empathy or affinity—and by extension, greater generosity—toward people whom they perceive to be more like them. This idea is sometimes expressed in the concept of “parochial altruism,” or altruism that is extended only within the borders of shared group membership and common experience (see also Fowler and Kam, Reference Fowler and Kam2007; Lupu and Pontusson, Reference Lupu and Pontusson2011). Alternatively, other scholars have proposed that people connect their own self-interest to the interests of their groups (Finseraas, Reference Finseraas2012; Donnelly, Reference Donnelly2020, Reference Donnelly2021). Although this study's design does not discriminate between these mechanisms, it is important to recognize that there is likely considerable overlap between the two motivations; as Lupu and Pontusson (Reference Lupu and Pontusson2011, 318) contend, “social solidarity may become an operative behavioural norm when individuals have some rational reason to suppose it might serve their own interests over the long run.”Footnote 1
Debates about mechanisms aside, it is now well-established in the empirical literature that group membership can influence redistribution. A number of studies show that people often prefer to redistribute toward their racial and ethnic in-groups (e.g., Cavaillé and Trump, Reference Cavaillé and Trump2015; Finseraas, Reference Finseraas2012; Luttmer, Reference Luttmer2001). Scholars have further examined how support for redistribution can be affected by perceptions of other shared group memberships, such as shared geographic region and religion (Davidson et al., Reference Davidson, Lesch, Hroux-Legault, Whyte, Asaf, Czuba and Porisky2017; Donnelly, Reference Donnelly2020; Wong, Reference Wong2010). Compellingly, Lupu and Pontusson (Reference Lupu and Pontusson2011) suggest that shared social class membership (as measured using low-, middle-, and high-income groups) can also engender perceptions of similarity that shape redistributive preferences, even if class is often thought to be less of a social identity than other identities, such as religion or race (see also Shayo, Reference Shayo2009).
At the same time, it does not necessarily follow from people's proclivity to identify with their social groups (and particularly, their more narrowly defined groups) that these identities will influence their redistributive preferences, for several reasons. For one, social identity need not generate political identity in the sense that social identity need not take on political relevance (Huddy, Reference Huddy, Huddy, Sears and Levy2013, 739). Indeed, a good number of people's social groups and identities—we can think, for example, of identifying with one's pick-up volleyball team—are never reflected in their political behaviour or attitudes (Titelman, Reference Titelman2023). Moreover, even if we accept that people's group identities sometimes lead them to discriminate in their redistributive preferences, people's identities are plural, and people's multiple identities have the potential to build bridges across more narrow identity groups in support of redistribution. For example, a Catholic mother working as a nurse and living in a rural community in Ontario may support redistribution on the basis of her social class identity, religion, place type, province or family role. Yet, while research has highlighted how redistributive attitudes can depend on the distribution of multiple identity groups in populations—for example, whether ethno-racial or ascriptive group membership is concentrated in certain income groups (reinforcing) or distributed across them (cross-cutting) (e.g., Finseraas, Reference Finseraas2012; Lupu and Pontusson, Reference Lupu and Pontusson2011; Yakter, Reference Yakter2019)—we have only limited evidence of which of Canadians’ multiple identities will matter for their redistributive preferences and when.Footnote 2
The work that comes closest to this study's interest in how shared identity influences redistribution is Bridgman et al. (Reference Bridgman, Merkley, Loewen, Owen and Ruths2021), which uses survey experiments to examine redistributive preferences in the context of the global COVID-19 pandemic. However, their approach to measuring perceptions of similarity is significantly different from the conceptual approach used in this study, as they measure contributor-recipient similarity by focusing primarily on characteristics related to recipient need and “control” over their circumstances.Footnote 3 In contrast, this article examines a broader range of identities that have been theorized as social identities and as dividing cleavages, both in Canada and comparatively, including social class (income, education, wealth), place and community type, and ascriptive identity. It should also be noted that Bridgman et al.'s (Reference Bridgman, Merkley, Loewen, Owen and Ruths2021) study was fielded in November 2020 during the first year of the global COVID-19 pandemic; as such, factors associated with the pre-vaccination days of the pandemic, such as the unprecedented scale of the Canadian government's redistributive efforts, may have more ephemerally influenced people's redistributive preferences. Because the experiments reported here were fielded just prior to the pandemic, they contribute to Bridgman et al.'s findings by providing an assessment of identity versus need considerations that are not shaded by pandemic-specific considerations.
Group Identities in Canada
Taken together, then, this study seeks to answer two questions. First, with which of their group memberships do Canadians most identify? Second, how do these group memberships map onto redistributive preferences—does shared group membership trump more universal measures of need? To examine group membership and shared identity, this study draws on existing research on political and policy divides in Canadian and comparative politics to identify four broad group memberships that have the potential to be of particular importance to Canadians: ascriptive identity, place (or geography), social class and gender (see Table 1 for these identities, and the indicators used to capture them).
Potentially Salient Identities and Indicators Used
The first group that this study investigates is ascriptive identity, which is defined here to variously include ethnic/racial, linguistic, or religious membership, depending on the group (see Yakter, Reference Yakter2019 for a discussion of ascriptive identity and redistribution). This approach recognizes that cleavages in the same country may form along various ascriptive identity lines—that is, for some groups ethnicity may be more significant; for other groups, language may be what defines them. In this study, the ascriptive identities analyzed are meant to capture what are often perceived to be salient ascriptive dividing lines and include: English Canadian, French Canadian, Chinese, Indigenous, and Muslim. The first group, English Canadian, represents Canada's historically largest ethno-cultural group. The French Canadian group reflects the country's linguistic duality stemming back to its colonization by the French and English, while the Indigenous group reflects the land's first inhabitants. Chinese Canadians were, until more recently, the largest visible minority population in Canada;Footnote 4 they are included in this study as a group whose dynamics are expected to be informed by experiences of racism not encountered by other white ethnic groups of European origin in Canada (Kymlicka, Reference Kymlicka, Guo and Wong2015).Footnote 5 Finally, this study also includes profiles of hypothetical Muslim Canadians,Footnote 6 given that this group has strong shared experiences of discrimination in Canada (Wilkins-Laflamme, Reference Wilkins-Laflamme2018).
For place—or geography-based—identification, this study focuses on two group memberships: province and community size. Provincial identities are widely considered to be paramount to Canadian politics, and there is a wealth of research linking province to differences in policy preferences, political attitudes, and party support (e.g., Simeon and Blake, Reference Simeon, Blake, Simeon and Elkins1980; Henderson, Reference Henderson2010). This study further identifies community size as a potentially salient place-based identity. Scholarship has found that place-based attachments—in the sense of affective attachment to a geographic area's people and institutions—are present at other levels of geography beyond province (Borwein and Lucas, Reference Borwein and Lucas2021), and the urban-rural (or urban-suburban-rural) divide is often highlighted as an important, and growing, cleavage in Canadian politics (Armstrong et al., Reference Armstrong, Lucas and Taylor2022; Roy et al., Reference Roy, Perrella and Borden2015; Walks, Reference Walks2004).
Three indicators of social class are also identified as potentially salient social and political identities: income, education and wealth (measured using homeownership). Canada presents a hard test case for studying class, elsewhere thought fundamental to identity and politics. Class voting has historically often been regarded to be marginal or missing in Canadian federal politics, overshadowed by more salient linguistic, regional and religious divides (Alford, Reference Alford1963). While some scholars attribute this pattern to Canadian political parties actively avoiding class appeals (Brodie and Jenson, Reference Brodie and Jenson1988), it is also thought to reflect lower levels of class consciousness in Canada (Pammett, Reference Pammett1987; Gidengil, Reference Gidengil2022). At the same time, other work has disagreed with the conclusion that class is irrelevant to Canadian politics (Kay and Perrella, Reference Kay, Perrella, Kanji, Bilodeau and Scotto2012) or has suggested that class's effects on voting must be understood in interaction with the effects of region (Gidengil, Reference Gidengil1989). More recent work by Polacko et al. (Reference Polacko, Kiss and Graefe2022), observing that contemporary economic transformations such as globalization and deindustrialization may have altered class's political salience in Canada, further supports the existence of a weak, though not entirely absent, class cleavage in Canada.
To explore class identities, this study looks first at income, which is among the bluntest and most commonly used measures of social class (alongside occupation) (Ansell, Reference Ansell2014). It should be noted that while income is a clear signifier of redistributive need, evidence also suggests that income position can form the basis for perceptions of commonality (Lupu and Pontusson, Reference Lupu and Pontusson2011). Thus, this study examines income in two ways; first, how low income (denoting need) influences redistributive preferences, and second, how shared income group (denoting shared group membership) influences preferences. Second, this study examines education, which is anticipated to be of growing salience to individuals’ perceptions of their own social class. Education has come to more clearly demarcate economic winners and losers in recent years, and some studies show it has in tandem become an increasingly central aspect of social identity (Kuppens et al., Reference Kuppens, Spears, Manstead, Spruyt and Easterbrook2018). Finally, this study examines wealth, operationalized using homeownership. Recent research has drawn attention to the centrality of wealth in determining people's economic circumstances and their policy preferences (Ansell, Reference Ansell2014). For most people, their wealth is predominantly dictated by the value of their home, and homeownership, moreover, carries a social significance for people that often extends beyond just its value as an asset (Rohe and Watson, Reference Rohe and Watson2007).
The final identity examined here is gender. On the one hand, gender is an important identity in that it is central to how many people think about themselves, and gender-based inequalities continue to be substantial in society. However, as compared to the politicization of other identities such as race, research suggests people are significantly less likely to identify with their gender in politics (Burns and Kinder, Reference Burns, Kinder and Berinsky2012).
Data and Methods
Survey Experiments
This article draws on two conjoint experiments embedded in the Campaign Period Survey of the Canadian Elections Study (CES), conducted in the Fall of 2019 (Stephenson et al., Reference Stephenson, Harell, Rubenson and Loewen2020). In total, 1,075 Canadian citizens aged 18 and over, living in one of Canada's ten provinces, participated in the two experiments.Footnote 7 Conjoint analysis is now widespread in political science, and has been used to study a range of topics, from candidate choice, to preferences over taxation and housing (for an overview of recent studies using the method, see Leeper et al., Reference Leeper, Hobolt and Tilley2020). In these conjoint designs, respondents are normally presented with two hypothetical options with multiple randomly varying attributes, and then asked to make a selection. The attributes presented to respondents are expected to be important for respondents’ choices; through their random variation, researchers can estimate the relative importance of each attribute for the choices respondents make (Hainmueller et al., Reference Hainmueller, Hopkins and Yamamoto2014).
This article uses two conjoint experiments to examine, first, with which groups Canadians are most likely to perceive commonality; and second, which if any of these group memberships are most influential for their redistributive preferences. In both experiments, respondents were presented with two profiles describing two hypothetical Canadians. In Experiment 1, aimed at assessing respondents’ most salient group memberships, respondents were prompted with the following question: “Read the descriptions of the two people below. Based on what you know, would you say you have more in common with [Person A] or [Person B]?” Table 2 provides an example of one set of conjoint profiles to which respondents could be exposed in the first experiment. Experiment 2 followed the first experiment but was designed to examine how respondents’ identities relate to their redistributive preferences, and asked the following question: “The government is planning to expand the GST/HST credit. If it were up to you, would [Person A] or [Person B] receive the increased credit?” Table 3 provides an example for Experiment 2.
Experiment 1: Example Conjoint Measuring Perceived Commonality
Read the descriptions of the two people below. Based on what you know, would you say you have more in common with Michelle Huang or Christopher Morrison?
Experiment 2: Example Conjoint Measuring Redistributive Preferences
The government is planning to expand the GST/HST credit. If it were up to you, would Fatima Hassan or Jason Perry receive the increased credit?
For both Experiment 1 and Experiment 2, names were drawn without replacement from a list and inserted into the question, such that no name appears more than once. Both experiments followed a forced choice design, with respondents asked to select one of the two profiles. Respondents completed each experiment three times. To guard against order effects, each respondent was randomly assigned to see attributes as listed in Table 2 or listed in the reverse order. To limit cognitive burden, the order of attributes was kept constant within respondents (Clayton et al., Reference Clayton, Ferwerda and Horiuchi2019).Footnote 8
For each profile, the following attributes were randomly varied: ascriptive group, gender, income, education, homeownership, province, community size, job security, favourite sports team and preferred hobby. As discussed above, the focus in this study is on group memberships that are often salient as social identities—ascriptive group, gender, social class (income, education, homeownership) and place (province and community size)—as well as indicators of need—job security and income. Two attributes, hobby and sports team, were further included to add definition to the lives of the hypothetical Canadians. Table 4 shows a list of attributes, the indicator group to which it belongs (ascriptive identity, gender, social class, geography, need) and the attribute levels. Because of the repetitive nature of the tasks in these experiments and to increase the realism of the profiles, multiple values were sometimes used to measure each level; the second column of Table 4 shows the levels that are of interest in the study, while the third column shows the values used to measure each level in the survey. All attribute levels were randomized without constraints, but to make the experiments more realistic, some levels were randomized to appear with greater frequency.Footnote 9 These weights are highlighted in subsequent sections.
Conjoint Attributes and Levels
a Note: For the full list of names used in the conjoints, see Appendix A.1.
The first indicators outlined in Table 4 are for ascriptive group and gender. Both attributes were signalled to the respondent using the first and last name of the hypothetical Canadian. Ascriptive identity includes the following levels: French Canadian, English Canadian, Indigenous, Chinese, and Muslim. Column 3 shows examples of names used to signify these levels; for example, “Elaine Yang” was one name used to signify a Chinese woman, while “Jean Desjardins” was used to signify a French Canadian man (see Appendix A.1 for the full list of names used in this study). Profiles were equally likely to denote men and women. To reflect that most provinces have primarily English Canadian populations, respondents were more likely to see an English Canadian name than all other groups (3/7th of all names were English Canadian, while 1/7th of the names were each of French Canadian, Chinese, Indigenous, and Muslim).
Table 4 next outlines attributes used as social class indicators, followed by indicators of redistributive need. Education has three levels, each seen with equal probability—high school or less, some university or college, and a university degree. Homeownership has two levels, renter and homeowner, with half of the profiles denoting a renter or a renter hoping to buy and half denoting a homeowner. Income, which can be considered both an indicator of social class and need, has three levels—low, middle and high—each of which respondents saw with equal probability.Footnote 10 The second indicator of need, employment security, includes “recently promoted” (drawn with 2/9th probability), “recently laid off” (drawn with 2/9th probability), and currently employed (drawn with 5/9th probability).
Under place indicators, Table 4 shows that two geographic identities—community size and province—are measured in a single item. First, to signify community size, the hypothetical Canadian was described as living in a province's largest city, a medium-sized city, or a small community (of population between approximately 3,000 to 20,000). For example, profiles for hypothetical Ontarian residents included, “Toronto, Ont.,” “Sarnia, Ont.,” or “Port Colborne, Ont.” Each community size appeared with equal frequency. Second, in order to test the importance of province as an identity, respondents were exposed with equal weight to profiles for which the hypothetical person lives in their home province, and profiles for which the hypothetical person lives in any of the other nine provinces. Given the large number of provinces, province identity is then coded as “in-province” and “out-province,” with respondents who report living in the same province as the hypothetical Canadian being coded as “in-province,” and vice versa.
Because the focus of this study is on the extent to which people perceive commonality with other Canadians who they perceive to be “like them,” this study also identifies respondents’ own group memberships. Conjoint attributes were matched as closely as possible with respondents’ self-reported demographic information in the CES, but not all matches are exact—Appendix A.2 outlines group categorizations, and reports the basis on which matches were made.
Methods
Experiment 1: Perceived Commonality
This study begins by examining perceptions of shared commonality in the first conjoint experiment. To examine the relative influence of each group membership in determining to which hypothetical Canadian respondents feel most similar, this study begins by regressing an indicator for whether the respondent chose a given hypothetical Canadian's conjoint profile on a set of indicator variables, indicating whether or not the respondent shares each of the hypothetical Canadians' identities in the profile. For example, a “match” for gender implies that both the respondent and the hypothetical Canadian in the profile share a gender (Gender Match=1) and so on for other identities in the conjoint. The equation further includes indicators for preferred sports team and preferred hobby of the hypothetical Canadian. Following common practice in conjoint analysis, analyses use linear regression and report both average marginal component effects (AMCEs) and marginal means. The AMCEs show how the presence of a given shared identity changes overall support for the profile relative to not sharing that identity (averaging over all other shared identities and profile attributes), while marginal means report the level of favourability toward profiles containing a given shared identity or attribute level (that is, the probability with which an attribute level is selected) (Leeper et al., Reference Leeper, Hobolt and Tilley2020). Standard errors are clustered at the respondent level to capture non-independence in respondents’ choices (Hainmueller et al., Reference Hainmueller, Hopkins and Yamamoto2014).
After determining which shared group memberships have the most influence on Canadians’ perceptions of commonality, this study then conducts more exploratory analysis examining subgroup variation. This analysis allows for a more in-depth examination of how the different categories of each identity (for example, being low, middle, or high income) relate to perceptions of commonality, but should be interpreted with some caution, given the diminished sample size when examining subgroups. For this analysis, a second equation is estimated, with each conjoint level of interest entering into the equation as dummy variables (for example, for income group, the equation includes a dummy variable for middle and high income; low income is the reference category). Because the focus here is on preferences among subgroups of respondents, these analyses report marginal means, which Leeper et al. (Reference Leeper, Hobolt and Tilley2020) show are most appropriate for subgroup analysis.Footnote 11
Experiment 2: Redistributive Preferences
For the second experiment on redistributive preferences, this study begins by presenting AMCEs and marginal means for all feature levels in the conjoint. In doing so, this analysis provides an initial assessment of how responsive Canadians are overall to more universal indicators of need (having a low income or having recently been laid off) as compared to other identity signifiers. To then examine how Canadians weigh need against shared identity, the final set of exploratory analyses in this article includes subgroup analysis of respondent support for profiles that share their group memberships on each attribute.
Results
Experiment 1: Perceived Commonality
Figure 1 reports results for the first model which regresses perceived commonality on shared group membership. Each coefficient shows the extent to which sharing an identity with the hypothetical Canadian in the conjoint influences the selection of that profile (see Appendix A.3 for results in table form). Figure 1 suggests that group memberships related to social class—income group, education and homeownership—are most associated with people's perceptions of shared commonality. Although not commonly studied as a social identity, shared employment status is also associated with perceptions of commonality. Place indicators—shared province and shared community size—further influence perceptions, although to a lesser extent. Neither ascriptive identity nor gender appears to inform respondents’ perceptions as to which hypothetical Canadian they feel most similar.
Perceptions of Group Commonality
Note: Figure shows AMCEs (left) and marginal means (right) for the relationship between shared identity and perceptions of commonality, with 95% confidence intervals. Models also include hobby and sports team attributes, not shown.
Having determined from Figure 1 that both social class indicators (income group, education and homeownership) and place ones (province and community size) are associated with perceptions of commonality, this article now turns to examining in more detail who within each group perceives this similarity. Beginning with social class indicators, Figure 2 shows marginal means for the measures examined in this study (see Appendix A.4 for all results in table form).
Social Class Attributes and Perceptions of Group Commonality
Note: Figure shows marginal means for subgroup analyses examining the relationship between the hypothetical Canadian's income (top left), education (top right), homeownership (bottom left) and perceptions of commonality, broken down by the respondent's own group membership, with 95% confidence intervals.
Looking first at income, the top left panel in Figure 2 shows how perceptions of commonality with low-, middle- and high-income earners vary across respondents who themselves have similar incomes. Figure 2 suggests that income is an important source of commonality across income groups; low-, middle- and high-income respondents all perceive greater commonality with hypothetical Canadians who have incomes similar to their own. The difference is particularly pronounced for high-income earners. Among the high-income group, perceived commonality with a hypothetical Canadian who is likewise a high-income earner is 27 percentage points higher than for a hypothetical low-income earner.
The top right panel of Figure 2 examines education. It shows that respondents with a university education perceive greater commonality with university-educated Canadians as compared to high-school-educated Canadians, by just under 25 percentage points. Both high-school and college-educated groups also perceive greater commonality with their own education groups, although in-group versus out-group differences are less pronounced for these groups as compared to the highest education group. Finally, for homeownership, the bottom left panel of Figure 2 suggests that both homeowners and non-homeowners perceive somewhat more commonality with Canadians who are homeowners and renters, respectively, as compared to the opposite.
Turning next to perceptions of commonality based in geography, Figure 3 reports marginal means for province and community size. Province is visualized here for completeness, but as discussed above, due to the number of provinces, sample sizes are too small to analyze each province as a separate subgroup. For this reason, conjoints were designed for province to be analyzed only as in-province and out-province, so respondents saw a location within their province and a location outside of their province with equal probability. The left panel of Figure 3 shows that the marginal mean for respondents selecting a profile outside of their province is approximately 0.47, and 0.53 for an in-province profile. This constitutes a 6 percentage point gap in perceived commonality for in-province as compared to out-province, a difference that is significantly smaller than many of the differences observed among subgroups for education and income.
Place Attributes and Perceptions of Group Commonality
Note: Figure shows marginal means for subgroup analyses examining shared provincial membership and perceptions of commonality (left) and the hypothetical Canadian's community size and perceptions of commonality, broken down by respondent community size (right) with 95% confidence intervals.
The right panel of Figure 3 examines the extent to which respondents identify with their community size. Findings suggest that residents from smaller areas, in particular, perceive greater commonality with Canadians who similarly live in small towns as compared to medium-sized cities. The marginal mean for respondents from small towns selecting as similar the profile of a hypothetical Canadian who likewise lives in a small town is 0.58, as compared to 0.44 for selecting profiles of hypothetical Canadians from medium-sized communities.Footnote 12
Experiment 2: Redistributive Preferences
Having established that aspects of both social class and geography constitute important sources of perceived commonality for Canadians, the next question is: do these group memberships also relate to redistributive preferences? Examining Figure 4, which reports how the main conjoint attributes in this study shape redistributive preferences across all respondents, it is evident that respondents heavily favour redistributing toward less well-off Canadians (as measured by income) (see Appendix A.6 for results in table form). Put differently, Canadians are by far most in favour of targeting redistribution toward those in economic need. Employment status (having recently been laid off) is also related to redistributive generosity, albeit much less substantially.
Attributes of the Hypothetical Canadian and Redistributive Preferences
Note: Figure shows AMCEs (left) and marginal means (right) for the relationship between the hypothetical Canadian's attributes and redistributive preferences with 95% confidence intervals. Models also include hobby and sports team attributes, not shown.
Because Figure 4 reports redistributive preferences across all Canadians, one concern is that it may mask differences across subgroups. In fact, even when breaking down analyses by subgroups, we see that the patterns of shared commonality observed in Experiment 1 are not also observed in Experiment 2. Figures 5 and 6 show these analyses for social class indicators (income, education, homeownership) and place indicators (province, community size), respectively (see Appendix A.7 for results in table form). There are only a few instances where respondents show any affinity at all for redistributing toward hypothetical Canadians who share their group memberships, but these effects are small; for example, respondents are more supportive of redistributing toward Canadians living in their own province than other provinces, by about 3 percentage points, and high-school respondents are slightly more supportive of redistributing toward other Canadians with high-school education as compared to university education. Again, however, these findings are eclipsed by people's strong preferences for redistributing toward those who most need it. Respondents from low, middle, and high-income groups are all more supportive of redistributing toward Canadians with low over high incomes, by between approximately 30 to 40 percentage points. In general, perceptions of need are much more important for redistributive preferences than shared group memberships, whether based in shared income, homeownership, education, province or community size.
Social Class Attributes and Redistribution Preferences
Note: Figure shows marginal means for subgroup analyses examining the relationship between the hypothetical Canadian's income (top left), education (top right), homeownership (bottom left) and redistributive preferences, broken down by the respondent's own group membership with 95% confidence intervals.
Place Attributes and Redistributive Preferences
Note: Figure shows marginal means for subgroup analyses examining shared provincial membership and redistributive preferences (left) and the hypothetical Canadian's community size and redistributive preferences (right) with 95% confidence intervals.
Design Limitations
This study's design has several important limitations. First, because of the study's interest in shared identity, it focused on subgroup analysis, matching attributes of hypothetical Canadians presented in the conjoint to characteristics of real Canadians responding to the survey. However, since both introducing more levels to conjoints and analyzing subgroups diminish statistical power, shared identity could only be studied within relatively aggregated groups, which may mask finer-grained variation. For example, among university-educated Canadians, it may be the case that the most educated Canadians (with graduate and professional degrees) identify more strongly with their education group than Canadians with only a bachelor's degree (or vice versa), but such granular group analyses would depend on a significantly larger sample size. Second, this study's design prioritized realism in profiles, which introduces some uncertainty about how attributes may be interpreted. For example, to analyze identification with community size, respondents read about hypothetical Canadians living in real Canadian places ranging in size. Canadians’ views of these places may depend not only on their size but also on other features unique to these places. The same may be true for other attributes as well.
Conclusion and Next Steps
This study examined the strength of Canadian social identities and their potential consequences for redistributive preferences. It asked two questions: with which groups do Canadians most identify, and to what extent do these group memberships influence their redistributive preferences relative to needs-based considerations? Despite the longstanding emphasis on certain identities in Canadian politics—particularly those relating to province and ascriptive identity—findings suggest that indicators of social class are as or more important for Canadians’ perceptions of shared commonality. Even while examining Canada, a country sometimes argued to have low class consciousness (Pammett, Reference Pammett1987), findings suggest that Canadians, for the most part, perceive that they have more in common with others who make similar incomes, are similarly educated, and are similarly located in the housing market, as compared to Canadians who live in the same provinces or similar sized communities, or who belong to the same ascriptive groups.
This is not to say that the latter identities are not important. For ascriptive groups, it is important to recognize that results presented here only examined overall perceptions of similarity based in specific indicators of shared race/ethnicity, religion or language, which because of the predominance of English (and to a lesser extent French) Canadians in the survey sample, is in practice largely a test of these groups’ perceptions. However, differences may exist among subgroups; some groups are likely to be more meaningful sources of similarity for respondents than others. As previously mentioned, people may also identify with more narrow groups than the ones used here. For example, names used to signify Indigenous group membership combine different groups under one umbrella group, which may influence identification. Finally, names unavoidably often convey multiple pieces of information about a person—for example, names used here to signify Muslim Canadians also carry potential ethnic and cultural markers with them—and respondents may identify with the hypothetical Canadian in the experiment on one dimension but not the other.
In terms of redistributive preferences, this article finds that all else equal, Canadians strongly prioritize redistributing toward those Canadians most in economic need. Identity considerations are, at best, only secondary. Even when examining a broader range of salient social identities in this study, and doing so in the pre-pandemic context, this finding largely accords with Bridgman et al.'s (Reference Bridgman, Merkley, Loewen, Owen and Ruths2021) finding that needs-based considerations primarily shaped Canadians’ redistributive preferences in the pandemic period. There are only several instances where Canadians’ group memberships appear to match their redistributive preferences at all, but given how small the differences are and the smaller sample size when analyzing subgroups, we should be cautious in interpreting any of these findings as evidence of meaningful group differences in support for redistribution. Differences among these groups are also often not evident in the smaller sample replication survey presented in Appendix A.5, further suggesting caution in interpreting these findings is warranted.
Future work should explore several avenues. First, this article has focused on the relative strength of people's identities, but identity is intersectional (Crenshaw, Reference Crenshaw1989; Dubrow, Reference Dubrow2008). To more fully capture identity and its implications for redistribution, future work should consider how people's multiple identities intersect to create unique experiences of advantage and disadvantage, in the process, shaping their political attitudes and preferences. To give just one example, future research could build on insights from earlier research on class voting, which shows that class voting is prevalent only in Canadian economic centres (Gidengil, Reference Gidengil1989), to examine how social class and place-based disadvantage may combine to shape identity and attitudes.
Second, although this study examined a number of social identities thought to be relevant to voting behaviour, it does not study vote choice as its dependent variable. An important next step would be to examine how the strength of Canadians’ social identities predicts affiliation with different political parties. Given evidence that political cleavages that are rooted in social identity are often particularly intractable, where Canadians’ perceptions of commonality with certain groups are highly predictive of their partisanship, this may point to a particularly durable political divide. Yet while we know that a number of different group memberships predict differences in Canadian party support, we know less about how entrenched these memberships are for Canadians as social identities. This is a task for future research.
Supplementary Material
The supplementary material for this article can be found at https://doi.org/10.1017/S0008423924000131.
Acknowledgements
For helpful advice on this article, I am grateful to Michael J. Donnelly, Jack Lucas, Ben Allen Stevens, John McAndrews, Olivier Jacques, Zain Asaf, Minh Do, Heather Millar, Beatrice Magistro, and members of the Policy, Elections and Representation Lab at the University of Toronto.
Competing interests
The author declares none.
Open access
