Introduction
Relations between Canadian Muslim communities and the wider Canadian society have been a politically charged issue in recent years. One need only recall the 2013 Parti Québécois proposal for a Quebec Charter of Values widely seen as targeting Muslims, or the 2015 federal election and the proposals of the then-incumbent Conservative party government to ban the niqab from Canadian citizenship ceremonies, and implement a telephone hotline to report suspected cases of “barbaric cultural practices.” Neither election—which saw both the PQ and Conservatives lose power—put such issues to rest, however. Incidents of vandalism and arson at mosques across Canada have occurred sporadically, typically followed by denunciations by leaders from politics and civil society. There is also the tragic mass shooting at the Centre Culturel Islamique de Québec in January 2017, and in response the “rings of peace” formed by members of the public around mosques in different parts of Canada intended as displays of solidarity with Muslim Canadians. There have also been protests against parliamentary motions denouncing Islamophobia, protests against the accommodation of Muslim prayers in Ontario public schools, and counter-protests in response. Most recently, the government of Quebec has (again) attempted to legislate a ban on religious face coverings, widely seen as targeting Muslim women wearing the niqab.
These events underscore that majority-group Canadians’ attitudes toward Muslims range widely from acceptance to antipathy. With just over 1 million Muslims in Canada, nearly 300,000 of whom are Canadian-born (as of the 2011 National Household Survey), understanding majority-group Canadians’ attitudes toward Muslims in Canada is important, not only as it pertains to social cohesion, but negative attitudes toward Muslims are also linked to increased support for militarism and reduced support for civil liberties (Schildkraut, Reference Schildkraut2002; Sides and Gross, Reference Sides and Gross2013). Such issues are not, however, unique to Canada; rather, they are a salient political issue in many Western democracies (Savelkoul et al., Reference Savelkoul, Scheepers, der Veld and Hagendoorn2012; Wike and Grim, Reference Wike and Grim2010). Research focusing on the Canadian case, however, is limited (but see Harell et al., Reference Harell, Soroka, Iyengar and Valentino2012; O'Neill et al., Reference O'Neill, Gidengil, Côté and Young2015; Roach, Reference Roach and Sinno2009; Wright et al., Reference Wright, Johnston, Citrin and Soroka2017; Zafar and Ross, Reference Zafar and Ross2015). Further, I take the position that research exploring the dynamics of majority–Muslim relations in Canada—written for a Canadian audience, as well as those interested in Canada—is an important endeavour (Noël, Reference Noël2014). What, then, are the factors that shape Canadians’ attitudes toward Muslims? I pursue this question using Canadian Election Studies (CES) survey data from 2011 and 2015, and a theoretical framework that links outgroup perceptions to intergroup contact (friends), local demographic context at both the micro-level (neighbours) and meso-level (townspeople) and political factors (parties). My contention is that a complete explanation of Canadians’ attitudes toward Muslims requires reference to all four factors.
I structure the article as follows. First, I review existing research on attitudes toward Muslims in Canada and in other country contexts. Next, I advance my hypotheses grounded in intergroup contact theory (Allport, Reference Allport1954; Pettigrew, Reference Pettigrew1998), contextual theories of racial and ethnic attitudes (Glaser, Reference Glaser1994; Key, Reference Key1949), and party identification as a source of elite cues (Zaller, Reference Zaller1992). I then describe the CES survey data and my methods, including the appending of local-level measures of religious/demographic context, multiple imputation and regression modeling. I then discuss my results and their substantive implications before reflecting on areas for future research.
Attitudes toward Muslims in Canadian and Comparative Perspective
Whether one defines the object of inquiry as Islamophobia, anti-Muslim prejudice or anti-Muslim sentiment, it is important to note that negative attitudes toward Islam and/or Muslims are not a new phenomenon in Western democracies. Rather, they represent a continuation of long-standing negative stereotypes of Muslims (Said, Reference Said1979). Attitudes toward Muslims have been found to comprise facets pertaining to cultural outgroups as well as racial/ethnic outgroups (Kalkan et al., Reference Kalkan, Layman and Uslaner2009). Waves of migration from Muslim-majority countries in recent decades as well as terrorist attacks have renewed interest in mass public attitudes toward Muslims, in particular, the attitudes of publics in Western Europe (Clements, Reference Clements2012; Savelkoul et al., Reference Savelkoul, Scheepers, der Veld and Hagendoorn2012; Strabac and Listhaug, Reference Strabac and Listhaug2008) and the United States (Jung, Reference Jung2012; Kalkan et al., Reference Kalkan, Layman and Uslaner2009; Nisbet et al., Reference Nisbet, Ostman, Shanahan and Sinno2009; Sides and Gross, Reference Sides and Gross2013). Other research has sought to make cross-regional (transatlantic) comparisons in attitudes toward Muslims which include Canada (Savelkoul et al., Reference Savelkoul, Scheepers, der Veld and Hagendoorn2012; see also Strabac et al., Reference Strabac, Aalberg and Valenta2014; Wike and Grim, Reference Wike and Grim2010).
Still, findings relating specifically to Canada are relatively sparse. Strabac and colleagues (2014) find that anti-Muslim sentiment is relatively muted in Canada compared to Europe. Guimond and colleagues (Reference Guimond, Crisp, Oliveira, Kamiejski, Kteily, Kuepper, Lalonde, Levin, Pratto, Tougas, Sidanius and Zick2013) similarly find lower levels of anti-Muslim prejudice in Canada compared to Britain, Germany and the US. At the same time, other research (based on a student sample) has found that Muslims are rated more negatively than other religious groups in Canada (Zafar and Ross, Reference Zafar and Ross2015).
As for the factors shaping Western publics’ attitudes toward Muslims, recent research has pointed to intergroup contact, ethno-religious context and political factors, as well as individual-level demographic and religious characteristics.
Drawing on the long tradition of intergroup contact theory (Allport, Reference Allport1954; Pettigrew, Reference Pettigrew1998) first developed in the context of black–white relations in the US, recent work on majority group–Muslim relations has similarly hypothesized that contact between groups can, in the presence of supporting conditions, lead to more positive outgroup perceptions. Several studies have confirmed this expectation. Research on Britain (Hewstone and Schmid, Reference Hewstone and Schmid2014), the US (Jung, Reference Jung2012; Kalkan et al., Reference Kalkan, Layman and Uslaner2009; Merino, Reference Merino2010), and the Netherlands (Van Der Noll et al., Reference Van Der Noll, Poppe and Verkuyten2010; Velasco González et al., Reference Velasco González, Verkuyten, Weesie and Poppe2008) has found that sustained, positive contact with Muslims is associated with more positive perceptions of Muslims.
Attitudes toward Muslims has also been approached from the perspective of inter-religious relations, framing the issue in terms of relations between Muslims and Christians (or different Christian sects), Jews, and other religions—or for that matter, those with no faith at all. Results, however, vary widely. Some research finds a positive relationship between identifying with an organized religion and attitudes toward Muslims (Abu-Rayya and White, Reference Abu-Rayya and White2010; Fetzer and Soper, Reference Fetzer and Soper2003) while other studies find specific Christian sects (in particular, Evangelicals) hold more negative attitudes (Jung, Reference Jung2012; Merino, Reference Merino2010; Penning, Reference Penning2009). Still others find no clear relationship (Clements, Reference Clements2012; Savelkoul et al., Reference Savelkoul, Scheepers, Tolsma and Hagendoorn2011).
An alternative approach to studying attitudes toward Muslims emphasizes not intergroup or inter-religious contact, but rather demographic or cultural context. Some research defines context nationally, using the country-level proportion of Muslims as an explanatory variable, but with inconclusive results (Savelkoul et al., Reference Savelkoul, Scheepers, der Veld and Hagendoorn2012; Strabac and Listhaug, Reference Strabac and Listhaug2008). Other research builds on the tradition of local contextual effects on outgroup attitudes. Seminal work by Key (Reference Key1949) and Blalock (Reference Blalock1956, Reference Blalock1957) on the southern US identified a positive association between the local-area proportion of African Americans and the prevalence of racially conservative attitudes among whites (see also Enos, Reference Enos2017; Glaser, Reference Glaser1994). This work has been extended to examine links between local ethnic or immigrant context and immigration attitudes in Canada (Gravelle, Reference Gravelle2017), the US (Gravelle, Reference Gravelle2016; Hood and Morris, Reference Hood and Morris1997; Hopkins, Reference Hopkins2010) and Europe (Kaufmann, Reference Kaufmann2017; Kaufmann and Harris, Reference Kaufmann and Harris2015; Schneider, Reference Schneider2008; Weber, Reference Weber2015). In general, this research has found a link between higher local immigrant or ethnic minority concentration and more open immigration attitudes, effectively, the reverse of the type of dynamic previously found by Key and Blalock. A central question in this literature, though, is the extent to which local context serves as a mere proxy for intergroup contact, or whether other social processes (such as foreign language exposure or the content of local media) are at work (Gravelle, Reference Gravelle2016; Hainmueller and Hopkins, Reference Hainmueller and Hopkins2014; Newman et al., Reference Newman, Hartman and Taber2012; Stein et al., Reference Stein, Post and Rinden2000). In a similar vein, recent research has posited that a larger Muslim population at the local level serves to increase the likelihood of intergroup contact, which in turn predicts more positive attitudes toward Muslims (Hewstone and Schmid, Reference Hewstone and Schmid2014; Savelkoul et al., Reference Savelkoul, Scheepers, Tolsma and Hagendoorn2011). In this formulation, the effect of local context is mediated by contact.
Finally, attitudes toward Muslims (as with other outgroups) may bear the imprint of political party identification. In recent years, anti-immigrant and anti-Muslim rhetoric has been a regular feature in the rhetoric of far-right political figures in Europe. More racially conservative attitudes, restrictionist attitudes toward immigration, and ascriptive notions of belonging (all ideas linked to anti-Muslim sentiment) have also been prominent among the American right. Still, only a small number of studies have sought to test the links between party identification and attitudes toward Muslims at the level of the mass public. Examining the British case, Clements finds no differences between Conservative, Labour, and other party supporters in their attitudes toward Muslims. By contrast, studies of the American case by Jung (Reference Jung2012), Kalkan and colleagues (Reference Kalkan, Layman and Uslaner2009) and Lee and colleagues (Reference Lee, Gibbons, Thompson and Timani2009) all find more negative attitudes toward Muslims among Republican identifiers than among Democrats. Nisbet and colleagues (Reference Nisbet, Ostman, Shanahan and Sinno2009) and Penning (Reference Penning2009) also find that conservative-leaning Americans are more likely to hold negative perceptions of Muslims.
Theory and Hypotheses
Given existing explanations for attitudes toward Muslims, I focus on the effects of intergroup contact, local context (measured at different levels) and party identification in exploring Canadian attitudes toward Muslims. Other potential explanatory variables, such as demographics, region and religious denomination, are included as controls.
Following the existing literature on intergroup contact, I hold that having sustained personal contact with Muslims will be associated with more positive attitudes toward Muslims. It is important to stress that such contact must rise above “casual contact” to more meaningful, personal “acquaintance” (as per Allport's initial articulation (Reference Allport1954) of intergroup contact theory) in order to affect outgroup attitudes in a positive manner. Accordingly, this research expectation might be termed the friends hypothesis, and may be stated as follows.
H1:
Positive contact with Muslims is associated with more positive attitudes toward Muslims.
Following the discussion above, it is important to parse out the effects of interaction (meaning direct, personal contact), and proportions (of an outgroup within an area), since they may imply different social processes (Forbes, Reference Forbes1997). A focus on local (demographic, cultural or religious) context is important for the study of mass opinion, since individuals “develop subjective understandings of the places they live in based on objective local characteristics, particularly the social composition” (Cutler, Reference Cutler2007: 579). Building on recent research on Canada, the US and Europe which finds that higher proportions of minorities or immigrants at the local level are associated with more open attitudes toward immigrants (Gravelle, Reference Gravelle2016, Reference Gravelle2017; Hopkins, Reference Hopkins2010; Kaufmann, Reference Kaufmann2017; Kaufmann and Harris, Reference Kaufmann and Harris2015), one might extend this logic and expect that higher proportions of Muslims in the local area will be associated with more positive attitudes toward Muslims among majority-group Canadians. At the same time, possibilities exist for measuring (and modelling the effects of) local context at different levels of aggregation, that is, micro-level and meso-level contexts (Baybeck, Reference Baybeck2008; Weber, Reference Weber2015). To be sure, the ethnic or racial composition of one's immediate neighbourhood, or micro-level context, is one type of milieu to consider, given that within a neighbourhood, opportunities may exist for intergroup contact—or at least passive exposure to foreign languages, or seeing individuals different in their appearance on the street. For example, Stolle and Harell (Reference Stolle and Harell2013) find a positive effect of neighbourhood racial diversity on having inter-racial friendships among young Canadians. Still, individuals often travel outside their neighbourhoods (for example, for employment), moving through space, as part of their daily lives (see Moore and Reeves, Reference Moore, Reeves, Bachner, Ginsberg and Wagner Hill2017). Larger, more aggregated definitions of context also align better with the media environments individuals inhabit, and may point to variation in media content (Dunaway et al., Reference Dunaway, Branton and Abrajano2010). This suggests a broader—town- or city-level, or meso-level—measure of ethnic or racial context. Research on whites’ racial attitudes in the US suggests that different contextual dynamics exist at the neighbourhood and city levels (Baybeck, Reference Baybeck2008). Research on Europeans’ immigration attitudes similarly find dynamics that vary between metropolitan areas and larger sub-national regions (Weber, Reference Weber2015). At the very least, it is worthwhile to test for effects of Muslim concentration at different levels of aggregation.Footnote 1 This leads to the neighbours and townspeople hypotheses, respectively.
H2:
Increasing concentration of Muslims at the neighbourhood level is associated with more positive attitudes toward Muslims.
H3:
Increasing concentration of Muslims at the town/city level is associated with more positive attitudes toward Muslims.
Though previous studies in other country contexts examining the link between party identification and attitudes toward Muslims have produced mixed results, there are compelling reasons to expect that majority-group Canadians’ attitudes toward Muslims ought to bear the imprint of party identification. A repeated finding in research on mass public opinion is that public attitudes are shaped by political factors, including party identification. One of the primary explanations for this is that the mass public uses “elite cues” as a type of cognitive shortcut in attitude formation. When cleavages in attitudes are present at the level of political elites, these tend to be mirrored by the mass public, who take cues from those elites with whom they share a partisan or ideological affinity. Zaller (Reference Zaller1992), for example, provides an account of this process of elite opinion leadership in the context of the evolution of racial attitudes in the US. From the 1940s through the mid-1960s, the racial attitudes of Democrats and Republicans exhibited no appreciable differences. Beginning with the administration of Lyndon Johnson, the Democratic party emerged as the party of racial liberalism, while the Republican party adopted more racially conservative positions. At the level of mass opinion, aggregate opinion among identified Democrats similarly became more racially liberal, with opinions among identified Republicans became more racially conservative. While perhaps attributable to a partisan realignment based on racial attitudes, Zaller nevertheless concludes that the most plausible explanation for the evidence is that “elite cues can change racial opinions” among the mass public (Reference Zaller1992: 13). Indeed, such elite cues provide an explanation for outgroup attitudes generally (and attitudes toward Muslims in particular) when considering those segments of the mass public that have minimal (or no) contact with or exposure to outgroup members.
In the Canadian case, partisan cues relating to majority–Muslim relations have been especially prominent in recent years, particularly in actions taken by the previous Conservative government of Stephen Harper (2006–2015) seen as antagonistic to Muslim Canadians. One notable case is that of Abousfian Abdelrazik, a Sudanese-born Canadian citizen accused of terrorist activities by the US government and who had previously come of the attention of Canadian and American law enforcement. Having been imprisoned in Sudan, Abdelrazik sought to return to Canada in August–September 2008, but the Conservative government refused to issue him emergency travel documents on several occasions. Only in June 2009 did the government commit to Abdelrazik's repatriation, having been ordered to do so by the Federal Court of Canada. Throughout the affair, the opposition Liberals and NDP were critical of the Conservative government's conduct. Former NDP leader Jack Layton argued at the time that Abdelrazik should be returned to Canada and charged criminally, if necessary, while former interim Liberal leader Bob Rae noted the hypocrisy of the Harper government's actions in denying consular services to a Canadian citizen on putative national security grounds while simultaneously granting him refuge inside the Canadian embassy in Sudan.
Turning to the widely publicized niqab issue, the Conservative government had in 2011 implemented a policy banning face coverings during citizenship ceremonies, arguing that it was necessary to ensure that new Canadian citizens were in fact saying the oath of citizenship. The rule was challenged in 2014 by Zunera Ishaq, a Pakistan-born Toronto-area woman seeking to take the oath of citizenship while wearing a niqab. The rule was struck down by the Federal Court of Canada in March 2015, a decision subsequently upheld by the Federal Court of Appeal. The Conservative government then stated that it would appeal to the Supreme Court of Canada. In the context of the 2015 election, the Conservative party announced its plan to entrench the ban on face coverings during citizenship ceremonies in law, a proposal criticized by both the Liberals and NDP, both of which committed to dropping the appeal should they form the government. The sovereignist Bloc Québécois, on the other hand, supported the niqab ban proposal. The Bloc also ran explicit anti-niqab ads targeting the NDP in Quebec, seen as its main electoral competition, asking rhetorically, “faut-il se cacher le visage pour voter NPD?” while using the image of the floor of the House of Commons as if viewed through the eye slits of a niqab.
The Conservatives paired their niqab ban proposal with a plan for a government-run “barbaric cultural practices” hotline to allow members of the public to report suspected or potential cases of forced or polygamous marriage, or female genital mutilation. While the Liberals initially supported the substance of the act, they criticized the proposed hotline as unnecessary and divisive. The NDP, on the other hand, opposed the act in Parliament. On the niqab issue, both the Liberal party and NDP opposed the ban, accusing the Conservatives of engineering it as a wedge issue. The hypothesis linking party identification in the recent Canadian context to attitudes toward Muslims is therefore clear.
H4:
Conservative Party and Bloc Québécois identifiers will hold less positive attitudes toward Muslims than Liberal party identifiers.
Data and Methods
To test the hypotheses linking intergroup contact, micro-level and meso-level context, and party identification to attitudes toward Muslims in Canada, the CES data from 2011 and 2015 are a particularly rich source of data. The CES has historically been conducted as a nationally representative, probability-based telephone survey, with campaign period and post-election survey waves, plus a mailback survey. In recent elections, it has also introduced an online survey, either as a recontact survey (2011) or as a separate non-probability sample (2015). The key survey item capturing feelings toward Muslims living in Canada (using a 0–100 “thermometer” scale) appears in the post-election survey in both 2011 (n = 3,362) and 2015 (n = 2,988). While some operational measures of Islamophobia stress the emotional responses of fear, anxiety or aversion engendered by Muslims (Lee et al., Reference Lee, Reid, Short and Campbell2013), the CES item is best seen as a measure of positive (or conversely negative) affective feelings toward Muslims in Canada that is distinct from but related to such emotional responses (Altareb, Reference Altareb1997). This understanding of Islamophobia as generalized negative attitudes toward Muslims is closer to both everyday language and its typical operationalization in social scientific research (Clements, Reference Clements2012; Savelkoul et al., Reference Savelkoul, Scheepers, der Veld and Hagendoorn2012; Sides and Gross, Reference Sides and Gross2013; Velasco González et al., Reference Velasco González, Verkuyten, Weesie and Poppe2008). Such affective feelings can similarly be distinguished from more specific attitudes toward policy issues, such as restrictions on Muslim female dress (for example, the niqab), or religious accommodation in public institutions such as schools (Fetzer and Soper, Reference Fetzer and Soper2003; Gustavsson et al., Reference Gustavsson, der Noll and Sundberg2016; Helbling, Reference Helbling2014).
The CES surveys also contain measures of the relevant concepts pertaining to the hypotheses above. Party identification is measured in the post-election surveys, while self-reported socializing with Muslims (as an operational measure of intergroup contact) comes from the mailback surveys. Respondent-level demographics (age, sex, education and region) are captured in the campaign period survey. The CES campaign-period surveys also contain detailed self-reports of respondents’ religious, ethnic, linguistic and cultural backgrounds (descriptive statistics and full question wording appear in the online appendix). This allows me to align my analyses with the practice of other studies focusing on majority-group attitudes toward minority groups, for instance, American non-Hispanic whites’ attitudes toward African Americans and Hispanics (Glaser, Reference Glaser1994; Gravelle, Reference Gravelle2016), or American, European and Australian whites’ attitudes toward Muslims (Abu-Rayya and White, Reference Abu-Rayya and White2010; Sides and Gross, Reference Sides and Gross2013; Strabac and Listhaug, Reference Strabac and Listhaug2008). Adapted to the Canadian context, I limit my analyses to the sub-samples of Canadian-born non-Muslims who speak English and/or French as their mother tongue, and who are of European descent. This yields an effective sample size of 3,883 (1,804 for the 2011 CES, and 2,079 for the 2015 CES).
In examining the views of this subpopulation, views toward Muslims in Canada are, on average, relatively positive (see Roach, Reference Roach and Sinno2009), with a mean score of 70.7 (on a 0–100 scale). Attitudes nevertheless range widely, indicated by the standard deviation of 27.5. The mean thermometer score for Muslims is also somewhat lower than those for Aboriginals (77.7), unspecified “racial minorities” (78.8), and immigrants (78.8).Footnote 2 Median thermometer scores for all four groups are 80. Further, 3.2 per cent of respondents in this subpopulation gave Muslims a thermometer score of zero, while 19.3 per cent gave a score of 100. In cross-national perspective, Canadian attitudes toward Muslims remain comparatively positive: mean thermometer scores of 50 or below have been reported in the US (Panagopoulos, Reference Panagopoulos2006; Sides and Gross, Reference Sides and Gross2013), Britain (Clements, Reference Clements2012), and the Netherlands (Velasco González et al., Reference Velasco González, Verkuyten, Weesie and Poppe2008).Footnote 3

Figure 1 Canadian feelings toward Muslims (CES 2011-2015)
To test the hypotheses relating to neighbourhood-level and municipality-level demographic context, I use Statistics Canada dissemination areas (DAs) as my working definition of neighbourhood. Dissemination areas are small areas with a population typically between 400 and 700 persons and which cover all of Canada. I use Statistics Canada census subdivisions (CSDs) as my working definition of town or city context. Census subdivisions typically correspond to cities, towns, townships and municipalities as defined by provincial legislation. Their population totals therefore range widely. Though many more operational definitions of local context ranging from very localized to larger regions are possible, this measurement strategy assures that contextual “containers” are assigned in a consistent manner, and that contextual demographic data are readily available. To be sure, individuals residing in large cities (such as Toronto, Montreal and Vancouver) may reside in and only be exposed to parts of such cities, but mainstream media sources will nevertheless cover social issues across the city or metropolitan area. City-level context thus remains potentially relevant.
Appending such contextual data to the 2011 and 2015 CES datasets, however, presents a practical challenge in “data wrangling.” Specifically, the two survey datasets do not include a consistent set of census geographic indicators: 2011 census geographic products were not available at the time of the 2011 federal election, and the CES 2015 dataset was not appended with the most current (2011) DAs by the survey data collection agency.Footnote 4 Further, censes in different years redraw geographic boundaries to accommodate population growth and changes in municipal boundaries set by law. I solve these issues by using the provided respondent-level latitude–longitude co-ordinates (originating from then-current vintages of the Statistics Canada Postal Code Conversion File Plus) in performing point-in-polygon overlays using 2011 Census shapefiles to assign each survey respondent to their appropriate 2011 DA and CSD. This method places all 2011 and 2015 CES survey respondents in a consistent set of micro-level (DA) and meso-level (CSD) geographic units. I am then able to append DA-level and CSD-level estimates of the percentage of Muslims taken from the 2011 CensusPlus databases from Environics Analytics to the survey data (the CensusPlus data are small area estimates based on official Census data, and are widely used in marketing research).Footnote 5
Survey item non-response is yet another challenge, and one that calls for a rigorous solution given the potential for missing data to bias model parameter estimates. This problem is particularly acute in relation to the self-reported measure of socializing with Muslims from the CES mailback surveys: only 43 per cent of respondents to the 2011 and 2015 CES post-election surveys returned a mailback survey. This raises the prospects of bias as well as reduced statistical power through a naïve application of listwise deletion. I avoid these pitfalls by implementing techniques for the multiple imputation of missing data (Allison, Reference Allison2001; Little and Rubin, Reference Little and Rubin2002). In light of the proportion of missing data due to mailback wave non-response, I create 20 multiply imputed datasets—far more than the typical 3 to 5. Linear models are then fit to the multiply imputed data using complex sample procedures that account for the stratification and weighting of the CES data. These results are then combined using procedures presented by Rubin (Reference Rubin1987) to produce the final reported results.Footnote 6
The model specifications test separately the contextual hypotheses pertaining to effect at the neighbourhood/DA level (Model 1) and municipality/CSD (Model 2). These models do not control for the effect of intergroup contact, which is added in Models 3 and 4. This strategy allows me to test whether effects of local Muslim concentration are mediated by socializing with Muslims (Hewstone and Schmid, Reference Hewstone and Schmid2014). Party identification is included in Models 1–4. Given this setup, I discuss the results pertaining to the individual hypotheses in reverse order from the discussion above.
Results and Discussion
The regression analyses yield several notable findings (see Table 1). Examining first the effects of some of the control variables, males express less positive feelings toward Muslims than females, as do older Canadians compared to their younger counterparts. A large education gap in attitudes toward Muslims is also apparent, with increasing levels of educational attainment associated with more positive attitudes toward Muslims. Depending on the model specification, the gap between those with less than secondary education and those with a university education is roughly 15–16 points on the Muslim feelings thermometer. These results align with previous findings from Britain, the US and Europe (Clements, Reference Clements2012; Kalkan et al., Reference Kalkan, Layman and Uslaner2009; Nisbet et al., Reference Nisbet, Ostman, Shanahan and Sinno2009; Strabac et al., Reference Strabac, Aalberg and Valenta2014). Though previous research focusing on the American context has found differences between religious denominations in their attitudes toward Muslims (Jung, Reference Jung2012; Merino, Reference Merino2010; Penning, Reference Penning2009), in the Canadian case, no robust differences between faith groups emerge. There are no significant differences between Protestants, Catholics, Jews and members of other faiths. Thermometer scores are also systematically higher in 2015 compared to 2011 (by approximately 4 points), even after controlling for political, contextual and demographic factors. This result is potentially attributable to the greater salience of majority–Muslim relations and religious accommodation policies in the context of the 2015 election campaign, issues that, while politically relevant in the preceding years, did not factor prominently in the 2011 election.
Table 1 Explaining Canadian Attitudes toward Muslims

Notes: (1) Results are ordinary least squares regression estimates with standard errors adjusted for the complex sample design (stratification and weighting) and multiple imputation. Continuous independent variables (age, logged percentage of Muslim population at the DA level, and logged percentage of Muslim population at the CSD level) are mean-centred.
(2) Missing data are multiply imputed 20 times using all available data.
(3)† p ≤ 0.10;* p ≤ 0.05;** p ≤ 0.01;*** p ≤ 0.001
The results also point to regional and linguistic cleavages in attitudes toward Muslims. Feelings are most positive in British Columbia and the Atlantic region and most negative in Quebec (Ontario and the Prairie provinces occupy a middle position and are statistically indistinguishable from one another). These regional and provincial differences persist even after controlling for local-level concentrations of Muslims (measured at either the DA or CSD level), suggesting that attitudes toward Muslims bear the imprint of multiple geographic contexts at varying levels of resolution (see Cochrane and Perrella, Reference Cochrane and Perrella2012). Notably, there is a further marginally significant negative effect of being francophone, even after controlling for province (or region). The model results thus imply than some of the most negative attitudes toward Muslims are found among French-speaking residents of Quebec. These results bolster those of Wright and colleagues (Reference Wright, Johnston, Citrin and Soroka2017) showing lower support in Quebec for the wearing of Muslim religious garb in public spaces compared to the rest of Canada. Their results, and those here, deserve further reflection in light of the tragic shootings in Quebec City.
Turning to the effects of party identification, the results align with theoretical expectations. Majority-group Canadians who identify with the Conservative party hold significantly more negative attitudes toward Muslims than those who identify with the Liberal party—a difference of about 8 thermometer points. Bloc Québécois identifiers are similarly more negative in their attitudes toward Muslims by approximately 9 points. By contrast, NDP identifiers are not significantly different from those of Liberal identifiers (though no theoretical expectations were advanced as to differences between Liberal and NDP identifiers). In sum, the results provide solid confirmation of H4. They also align with previous findings from the US where (right-of-centre) Republican identifiers exhibit more negative attitudes toward Muslims, though they provide a contrast to findings from Britain showing no partisan cleavages in attitudes toward Muslims.
Examining the effects of Muslim concentration at the local level, the results from Model 1 indicate that increasing proportions of Muslims at the DA level correspond to a modest, marginally significant (p = 0.06) increase in positive feelings toward Muslims among majority-group Canadians. Specifically, increasing the percentage of Muslims in the DA from 0 to 20 percent (the 99th percentile) corresponds to an increase in the predicted thermometer score from 70.4 to 73.9. Similar patterns are seen in Model 2 which tests the effect of increasing proportions of Muslims at the CSD level. Here, the effect of the (logged) percentage of Muslims is again marginally significant (p = 0.10). These results indicate that increasing the percentage of Muslims in the CSD (town or city) from 0 to 10 percent (again, the 99th percentile) corresponds to an increase in the predicted thermometer score from 70.0 to 71.3. The results thus offer tentative (but not compelling) support for H2 and H3.
Models 1 and 2 thus indicate that the effects of local Muslim concentration—whether at the neighbourhood or municipal level—on majority-group Canadians’ attitudes toward Muslims are modest at best. Further, these effects are no longer significant in Models 3 and 4 when socializing with Muslims is included as a covariate, which yields large positive (and highly significant) effects. Respondents who report socializing with Muslims are roughly 10 points higher in their thermometer scores than those who do not. In terms of absolute magnitudes, the positive effect of contact is roughly the same size as the negative effects of residing in Quebec, or identifying with either the Conservative party or the Bloc Québécois. What is more, the results from Models 3 and 4 point toward the mediation of the effects of local demographic context (specifically, Muslim concentration) by intergroup contact, a dynamic previously found in Britain (Hewstone and Schmid, Reference Hewstone and Schmid2014). Thus, these results offer solid confirmation of H1 while eroding support for H2 and H3.Footnote 7
Conclusion
To recap, majority-group Canadians’ attitudes toward Muslims in Canada are complex and find their roots in multiple factors. I have grounded my exploration of Canadian attitudes toward Muslims in varied strands of empirical social science: intergroup contact theory, contextual theories of political behaviour and elite cue theories of mass public opinion. Taken together, I locate the sources of Canadians’ views of Muslims in who they count among their friends, who their neighbours and their fellow townspeople are, and which of the federal political parties they identify with.
Intergroup contact is shown to play a crucial role, with those reporting socializing with Muslims having significantly more positive views. The effects of local demographic context (Muslim concentration) are modest at best, and further are not significant once one accounts for such contact. This points to the mediation of context effects by actual intergroup contact, a dynamic that is quite different from what holds in the United States, where contextual effects in ethnic relations persist, and operate independently of intergroup contact (Enos, Reference Enos2017; Gravelle, Reference Gravelle2016; Hainmueller and Hopkins, Reference Hainmueller and Hopkins2014). In short, in the case of relations between majority-group and Muslim Canadians, it is ultimately less consequential if Muslims are one's neighbours or fellow townspeople; it is far more consequential if one counts Muslims among one's friends. Such a finding not only attests to the continued and broad relevance of intergroup contact theory, but it also attests to the value of community engagement and outreach by Muslim Canadians in combating prejudice.
In addition to such interpersonal and contextual dynamics, there remains a parallel political dynamic in which party identification shapes majority-group Canadians’ attitudes toward Muslims. Conservative party and Bloc Québécois identifiers stand apart from Liberal party and NDP identifiers in holding distinctly more negative attitudes toward Muslims. Recalling the Abdelrazik case and the high-profile campaign pledge by the then-Conservative government during the 2015 federal election campaign to ban the niqab at Canadian citizenship ceremonies, such results are not especially surprising, yet their absolute magnitude remains noteworthy. These findings thus confirm the power of political elites to shape mass attitudes toward outgroups—in both positive and negative directions—as previously seen in the American context.
While the article has deliberately focused on majority–group Canadians’ broad affective attitudes toward Muslims, it has set aside a range of corollary issues, including more specific attitudes toward the niqab and religious accommodation in public spaces. These issues have been explored in the European context (Fetzer and Soper, Reference Fetzer and Soper2003; Gustavsson et al., Reference Gustavsson, der Noll and Sundberg2016; Helbling, Reference Helbling2014), yet despite their salience in Canada, few studies of Canadian attitudes on these topics exist at present, though notable exceptions are those by O'Neill and colleagues (Reference O'Neill, Gidengil, Côté and Young2015) and Wright and colleagues (Reference Wright, Johnston, Citrin and Soroka2017). More research involving nationally representative samples of the Canadian public is needed on these topics, and based on the results presented here, such research ought to examine the roles of intergroup contact, local context, and party cues in shaping mass public attitudes.
The decision to focus on majority-group attitudes has also necessitated that I set aside an investigation of the attitudes of Muslim Canadians themselves, an area of inquiry no less important. Though a methodologically rigorous study of Muslim Canadians’ experiences with intergroup relations in Canada would involve complex, targeted sampling strategies that are quite different from those employed by general population surveys (such as the CES), recent research has endeavoured to meet this challenge (Leuprecht and Winn, Reference Leuprecht and Winn2011; Neuman, Reference Neuman2016). Work in this vein ought to continue so that we might develop a more complete understanding of both the views of and views about Muslim Canadians.
Supplementary materials
To view supplementary material for this article, please visit https://doi.org/10.1017/S0008423917001470