Across the developed world, governments and political parties are debating the extent to which they should accommodate ethnic and religious minorities. Politicians and public figures have become more outspoken about their worries surrounding both increasing ethnic diversity and the integration policies meant to manage it. For instance, German Chancellor Angela Merkel claimed that multiculturalism is dead (Vertovec and Wessendorf, Reference Vertovec and Wessendorf2010). Disputes in Quebec over religious accommodation also fueled public debate during the last Canadian and Quebec elections. Skeptics of multiculturalism argue that it creates segregation instead of integration and fosters stereotyping and prejudice instead of tolerance. Joppke (Reference Joppke2004: 1), notably, suggests that the retreat of multiculturalism policy in Western Europe is linked to a “chronic lack of public support,” as well as its alleged inherent defects and failures.
The death of multiculturalism may be greatly exaggerated. While some radical changes in integration policies are discussed in party manifestos, only a few have been implemented, mainly in countries where governments have relied on radical right parties for votes in parliament. In practice, many of the policies associated with multicultural policy have been left intact (Korteweg and Triadafilopoulos, Reference Korteweg and Triadafilopoulos2012). And recent efforts at building indices of multiculturalism policy suggest extensions, not contractions (see, for example, the Banting and Kymlicka Reference Banting and Will Kymlickan.d.). Suffice it to say that the debate about the appropriate nature and scope of multiculturalism policy is far from over.
While policy trends have been a regular focus of academic work, relatively little is known about public attitudes towards (and support for) specific multiculturalism policies. Work on attitudes toward immigration, diversity and social tolerance has often been mistaken as interrogating support for multicultural policy. This research addresses multiculturalism as a demographic phenomenon, perhaps, and it often captures attitudes that are likely related to multiculturalism support. But this work does not directly capture public attitudes towards multicultural policies. We seek to fill this gap below, at least in part, by relying on a unique online survey experiment within the 2011 Canadian Election Study (CES) asking about multicultural policies, alongside an online survey experiment probing the conditions under which citizens support various types of multicultural policies in Canada. In particular, we examine two types of benefits, financial grants for activities of an ethno-religious group and access to public space for a meeting of this group, and, in each case, whether this support is contingent on, or at least affected by, the ethno-religious groups to which those policies apply. This is measured both by varying the ethno-religious background of the organization (Portuguese Catholic versus Turkish Muslim) and a visible indicator of religiosity, the Muslim headscarf (presence or absence of the hijab). We also explore whether citizens’ general attitudes toward assimilation moderate this effect.
This article focuses on the Canadian case. In 1971, Canada was the first country to announce an official policy of multiculturalism, later enshrining multiculturalism in the Charter of Rights and Freedoms in 1982. The Canadian Multiculturalism Act was then passed into legislation in 1988. Indeed, some argue that the idea of multiculturalism has actually become a key component of contemporary Canadian identity (Kymlicka, Reference Kymlicka1998; Mackey, Reference Mackey2002). Multiculturalism has certainly not suffered nearly the same level of public disavowal in Canada as it has in some European countries.
Canada provides an interesting case for this analysis because of—rather than in spite of—what appear to be long-standing and high levels of public support of diversity, immigration and multiculturalism. This is true in part because, contrary to what many suggest, Canadians are by no means universally supportive of multiculturalism policies. There is, after all, an ongoing debate about the effects, both positive and negative, of multiculturalism in Canada.Footnote 1 Canada is also a country in which survey respondents will be comparatively well-acquainted with a range of multicultural policies. It is this variance in what we suspect is comparatively stable support for multicultural policy that we seek to examine here. Results not only provide descriptive information about the state of Canadian public opinion on multiculturalism, but more general evidence about the nature, authenticity and limits of public support for multicultural policy.
Attitudes towards Multiculturalism
Multiculturalism is a concept with multiple meanings. It can refer to the presence of multiple ethnocultural communities, as well as the ways in which these communities can and should interact. In other words, multiculturalism can be used to describe a society,Footnote 2 as an ideological position,Footnote 3 and as a set of policies.Footnote 4
From a public opinion perspective, most work has focused on people's attitudes toward multiculturalism as an indicator of ethnic diversity (Berry and Kalin, Reference Berry and Kalin1995, Reference Berry and Kalin1996) or multiculturalism as an ideological position (Berry, Reference Berry1997; Berry et al., Reference Berry, Kalin and Taylor1977; Breugelmans and van de Vijver, Reference Breugelmans and Van De Vijver2004; Dandy and Pe-Pua, Reference Dandy and Pe-Pua2010; van de Vijver et al., Reference Van de Vijver, Breugelmans and Schalk-Soekar2008; Verkuyten and Martinovic, Reference Verkuyten and Martinovic2006). Less is known about how people feel about multiculturalism as a policy, especially the sources and structure of citizens’ (individual-level) attitudes toward concrete multicultural policies (MCPs). Measures of support for “concrete” multicultural policies are important; these are the attitudes most relevant to policy change, after all. But asking questions about multicultural policy is not easy. As Banting and Kymlicka (Reference Banting and Kymlicka2010) note, the term “multicultural policy” is ambiguous and overlaps many policy areas. Broadly speaking, multicultural policies “impose on public institutions an obligation to reduce barriers to immigrant participation and more accurately reflect the diversity of the population” (Kymlicka, Reference Kymlicka2003: 202). Thinking more narrowly, multicultural policies concern a complex range of discrete policies, such as government regulations that support cultural heterogeneity in public institutions’ grants for cultural activities and other provisions that help to maintain cultural traditions and particularities (see Banting and Kymlicka, Reference Banting and Kymlicka2006). Complexity has meant that measures of public opinion on multicultural policy are relatively scarce. Thus, little if any work explores how individual citizens feel about specific multicultural policies, the conditions under which citizens are more or less inclined to support them, or how this support relates to other political attitudes or experiences.Footnote 5
There are some attempts to get at policy attitudes that are closely related to multiculturalism, to be sure. Schildkraut (Reference Schildkraut2011) examines support for boundary politics, operationalized as more restrictive language policies as well as internment policies. The latter are not typically regarded as multicultural policies in Canada, but they nevertheless offer valuable insights on the predictors of policy support. For example, those who believe that it is important to feel American and be an American citizen show higher support for restrictive language policies. Conversely, people who believe that “true Americans” should be informed and involved in politics and volunteering indicate dislike for English-only ballots and tend to support bilingual education.
There also exist aggregate-level studies exploring Canadians’ support for multicultural policy, broadly defined. We know that in the early 1980s, the majority of Canadians were completely unaware that multicultural policy existed and that this changed gradually throughout the following years, such that by the turn of the century about 80 per cent of the population knew about “multicultural policies” (Dasko, Reference Dasko2003). We also know that, overall, Canadians show rather high levels of support for diversity, immigration and the general idea of multiculturalism (Soroka and Roberton, Reference Soroka and Roberton2010). As of 2007, 84 per cent of Canadians agreed that “Canada's multicultural makeup is one of the best things about this country” (Ipsos-Reid, 2007). Canadians also tend to favour immigration more than people in other countries and are also less likely to adjust their support based on the ethnic background of immigrants themselves (Harell et al., Reference Harell, Soroka, Iyengar and Valentino2012; Iyengar et al., Reference Iyengar, Jackman, Messing, Valentino, Aalberg, Duch, Soroka, Harell and Kobayashi2013). When asked about Canada's multicultural policy, three out of four young Canadians and far more than half of those over 25 years old have positive views of Canadian multicultural policy (Association for Canadian Studies, 2012).
Survey work on multiculturalism nevertheless tends to be relatively limited. There is little work gauging support for specific aspects of multicultural policy (see, for example, Guimond et al., Reference Guimond, De Oliveira, Kamiejski, Kteily, Kuepper, Lalonde, Levin, Pratto, Tougas, Sidanius and Zick2013), such as grants for maintaining and practising specific cultural traditions of immigrant communities or government support for ethnic groups’ meetings and gatherings. Developing measures of specific policy support is thus an important contribution of this article.
What Drives Support for Multiculturalism Policy?
We focus here on three factors in particular: the type of policy under consideration, the ethnic and religious background of policy beneficiaries and individual attitudes towards assimilation.
Policy type
We expect that support varies across specific multicultural policies. Banting and Kymlicka (Reference Banting and Kymlicka2013) distinguish three broad types of multicultural policies, namely, recognition (including symbolic statements about the nature of the community and representation), accommodation (including adaptation of institutions or policies to accommodate difference) and financial support (including grants to ethnic organizations, heritage language education, affirmative action). We need not view these categories as mutually exclusive; the adaptation of institutions can be financially costly, of course, just as financial support can be highly symbolic. The idea that multicultural policy varies on these dimensions should play an important role in understanding public support for specific policies.
We focus on two specific benefits directed at a hypothetical ethnic organization: access to public space and financial support for group activities. The former allows access to a public good so can be seen as an in-kind financial benefit but also reflects accommodation in so far as it provides a minority group with access to a public space that belongs to the entire community. Importantly, this policy has a proximity component to it that the financial grant does not, as it specifies physical proximity of the two groups (majority and ethnic/religious minority). The latter policy is a purely financial benefit. Financial benefits for ethnic groups have been reduced over the last years in Canada (Canadian Press, 2013); however, the multiculturalism budget is often used for such activities and is commonly associated with multiculturalism policy in Canada.
These two types of benefits, then, could be seen as instances of accommodation versus financial support, using Banting and Kymlicka's terms—though access to public space is not quite the archetypal accommodation-oriented policy, which allows minorities to continue to practice their cultural or religious values and habits.Footnote 6 Our question wording (discussed below) suggests accommodating ethnic beneficiaries within a majority institution, though, and has the advantage of suggesting physical proximity between minorities and the majority, something that the financial support does not.
Hypothesis 1:
We expect that more financially costly programmes will receive lower overall support.
Relatedly, we also suspect that the effect of sharing public space will be more sensitive to recipients’ background than the financial benefit due to its explicit proximity to respondents; we discuss this further below.
Ethnicity of recipient
Not all groups are viewed equally by the general public (Berry and Kalin, Reference Berry and Kalin1995) and so support for multicultural policies is likely to be affected by who the recipients of these policies are. We know from extensive research in social psychology that people are prone to categorizing out-group members in negative ways, particularly when they are perceived as more distant from the in-group (Allport, Reference Allport1954; Dovidio et al., Reference Dovidio, Gaertner and Kawakami2003; Tajfel and Turner, Reference Tajfel, Turner, Austin and Worchel1986). Berry and Kalin (Reference Berry and Kalin1995) find that immigrant groups of non-European background are less accepted by Canadians than those of European origin, and later public opinion research suggests that more than half of Canadians tend to have moderately or very unfavourable opinion towards Islam. Indeed, Islam had the highest proportion of unfavorable views in Canada compared with other religions (Angus Reid, 2009).Footnote 7 Anti-Muslim feelings have been rising and might be particularly influential for determining policy support (Yavçan, Reference Yavçan2013). And political science research shows that prejudicial attitudes do have an effect on policy support, especially when policies are known (or seen) to benefit specific groups (Bobo and Kluegel, Reference Bobo and Kluegel1993; Federico and Holmes, Reference Federico and Holmes2005; Gilens, Reference Gilens1995; Harell et al., Reference Harell, Soroka and Ladner2014; Hurwitz and Peffley, Reference Hurwitz and Peffley1997, Reference Hurwitz and Peffley2002; Sniderman et al., Reference Sniderman, Peri, Figueiredo and Piazza2000). For example, feelings towards Muslims should influence how people feel specifically about the Charter of Values debate in Quebec, which proposed to restrict public sector employees from wearing ostentatious religious symbols, because the policy was tied so heavily in the public debate to Muslim women's veils.
We therefore expect that group distinctions will matter. When multiculturalism policies are linked with specific ethnic minorities, support should be affected by group-based considerations. This expectation is drawn specifically from the cultural threat hypothesis, which posits that the more culturally distant an out-group, the more threatening it appears to be (Stephan et al., Reference Stephan, Ybarra, Martnez, Schwarzwald and Tur-Kaspa1998). While some would argue that anti-Muslim sentiment is driven more by perceived value differences between Christians and Muslims (Biernat et al., Reference Biernat, Vescio, Theno, Crandall, Seligman, Olson, Zanna, Seligman, Olson and Zanna1996; McConahay and Hough, Reference McConahay and Hough1973), our argument is essentially:
Hypothesis 2:
Cultural distance, whether based in lifestyle choices, ethno-racial differences or values, should drive support (from White, Christian majorities) downwards. Footnote 8
This distance might, in some cases, be underlined by different looks or dress codes in daily life, different lifestyles, traditions or customs.
We operationalize increasing degrees of cultural distance by contrasting a Portuguese Catholic immigrant with a Turkish Muslim immigrant both without, and with, a headscarf.
Attitudes about diversity and integration
Finally, we expect that more general attitudes toward assimilation will affect support for MCPs. Note that the degree to which people reject assimilation is often used as the measure of support for multiculturalism. Here we have separate measures of support for actual multicultural policies and use broader attitudes about cultural assimilation as a predictor rather than an outcome.
The tension between promoting cultural diversity versus assimilation has been a regular theme in policy debates surrounding multiculturalism. Canadian multiculturalism policy has always reflected an interest in integration within a larger nation-building project (Abu-Laban and Gabriel, Reference Abu-Laban and Gabriel2002). After all, two of the four components of Pierre Trudeau's original policy included the removal of cultural barriers to allow for the full participation of ethnic groups in Canadian society as well as training in the two official languages.Footnote 9 At the same time, the policy was put in place in contrast to more assimilationist approaches to integration that valued the shedding of distinct cultural identities in favour of the dominant cultural identity of the majority. Multicultural policy then sought to ensure that the terms on which integration took place recognized the cultural diversity of society, rather than focusing solely on assimilation into the cultural majority. Multicultural policies have thus maintained a delicate balance between promoting integration without effacing cultural diversity.
This tension is also reflected in public opinion. Schildkraut (Reference Schildkraut2011) identifies a tension in what she terms “incorporationist norms” in the US context: on the one hand, Americans value individuals carrying on the cultural traditions of their ancestors; on the other hand, they prefer that immigrants blend into the larger society. Similarly, while Canadians support the idea of a society where everyone is accepted, they are simultaneously concerned about immigrants not wanting to adopt Canadian values and expect immigrants to make an observable effort to become citizens and to internalize the national narrative (for example, Banting and Kymlicka, Reference Banting, Kymlicka, Johnston, Soroka, Banting and Kymlicka2006).
The question for us is whether citizens who tend to favour cultural assimilation react differently to specific multicultural policies and intended recipients. The policies we examine below focus, we believe, more on the accommodative aspects of MCP. There is no doubt that respondents who value cultural assimilation will be less supportive of these policies. More importantly, these attitudes may also have a moderating effect (see Guimond et al., Reference Guimond, De Oliveira, Kamiejski, Kteily, Kuepper, Lalonde, Levin, Pratto, Tougas, Sidanius and Zick2013); the impact of cultural distance on policy support may be greater for respondents who value cultural assimilation, since distance implies a greater perceived need for (and perhaps also a lower likelihood of) integration into the cultural majority. Put differently, support for cultural assimilation likely conditions the impact that the ethnicity of recipients has on policy support.
Hypothesis 3:
People with higher levels of assimilationist attitudes should be more prone to make distinctions between the Muslim and Catholic out-groups, whereas such distinctions should matter less to people with lower levels of such attitudes.
Method and Design
We rely below on the 2011 Canadian Election Study (CES). Our principal findings are based on a new online survey-experimental measure of support for multiculturalism policy, namely, support for a specific ethno-religious organization receiving (a) funding and (b) access to public space.Footnote 10
Respondents are randomly assigned to one of three vignettes. Each uses an image of the same woman, which minimizes the possibility that recipients’ appearance influences responses. The vignettes themselves are listed in full in Appendix I.Footnote 11 The critical part is that in the first treatment the woman is presented as Helena, the president of the Canadian Portuguese Catholic Action Network; in the second vignette, using exactly the same photograph, the woman is presented as Fatma, the president of the Canadian Turkish Muslim Action Network; and in the third vignette, the text is identical to the second vignette, but this time the photo shows Fatma wearing a headscarf.
The three manipulations thus allow us to examine how citizens respond to cultural proximate versus cultural distant groups. Portuguese Catholic immigrants have been established in Canada for over 50 years and share a common Christian heritage with the cultural majority; Turkish Muslim immigration is more recent and in our scenario characterized by religious difference. Moreover, within the Turkish Muslim vignettes, we vary the evidence of this religious difference by portraying the recipient with and without the hijab. This not only cues more clearly the religious distinction but may also be seen as a symbol of non-integration.Footnote 12
After reading a randomly assigned version of the vignette, respondents are asked two questions: first, they are asked whether they support or oppose the group receiving an $80,000 grant from Canada's Multiculturalism Grants and Contributions program to fund an outreach project to raise awareness of the group's contribution to Canada's culture; second, they are asked whether they support or oppose the municipality providing space in a local community centre for their project. There are four possible responses to each question: strongly support, somewhat support, somewhat oppose and strongly oppose. Respondents are not given the option to answer “don't know” or “neither support nor oppose.” (Again, the exact question wording of these vignettes can be found in Appendix I.)
Both dependent variables were originally asked on a four-point scale ranging from strongly oppose to strongly support. The two variables are highly correlated, at .57 (p < .001); still, the correlation coefficient makes clear that there is real variance across responses as well, in spite of the fact that the two questions are asked one after the other. The two questions appear to capture different dimensions of support for multiculturalism; they also have, as we shall see, slightly different predictors. For the analysis, we use a dichotomous variable of oppose (0) versus support (1).Footnote 13
Attitudes towards assimilation are measured using three questions from the mail-back portion of the CES (the third survey administered in the CES). The three questions are: (1) Too many recent immigrants just don't want to fit in; (2) Recent immigrants should set aside their cultural background and blend into Canada; and (3) Speaking English or French should be a requirement for immigration to Canada. Responses are four-point agree/disagree scales. The three items are used to create an additive scale ranging from 0 to 1 (Cronbach's alpha = .65),Footnote 14 where scores closer to 0 represent disapproval of assimilation and scores closer to 1 indicate a desire for assimilation. The average score on this index is 0.41 with a standard deviation of 0.22. We have no specific expectations regarding the functional form of this variable; to allow for nonlinearity and also to facilitate a slightly easier interpretation of the moderating effect of assimilationist attitudes on our experimental treatments, we use a categorical version of assimilationist attitudes below. The specification is based on the distribution of the index in our web-wave respondents, where the categories are the bottom 25 per cent of respondents (0), those 50 per cent in the interquartile range (1) and those in the top 25 per cent of assimilationist attitudes (2). Recall that we expect a preference in favour of assimilation to drive down the magnitude of support for multicultural policies, especially for culturally distant beneficiaries.
Examining Support for Multiculturalism Policy
We first explore support for each policy based on the ethnic/religious background of the policy recipient and then examine how attitudes regarding assimilation affect support for specific policies. Figure 1 presents the basic experimental results for average support for each policy type by treatment group. The results show that support is higher for the less costly community centre access, compared to the government grant (Hypothesis 1). Where a score of .5 represents evenly divided sentiment, Figure 1 makes clear that community centre access, on average, is supported by a majority of respondents, whereas the $80,000 grant is consistently, on average, opposed. Thus, when confronted with concrete policies and real benefits, the average Canadian seems to be relatively hesitant to support financial grants, but more willing to make public space available. Our suspicion, though we cannot explore this directly in these data, is that this difference in support is due to the latter type of accommodation being less costly.
Figure 1 Manipulation Effects on Support for Two Multicultural Policies
Recall that Hypothesis 2 suggests that cultural distance should influence levels of support. Figure 1 makes clear that the type of benefit is more important than who is receiving it. There is in fact no significant effect of cultural distance on access to community space, at least in this uncontrolled comparison. There is in contrast a slight increase in support for the $80,000 grant as we move toward the more culturally distant group (Muslim); and the hijab does not appear to exacerbate this distinction.Footnote 15 This is contrary to our hypothesis, yet as we will see, this trend hides a more complex underlying relationship.
This underlying relationship, rooted in the way in which assimilationist attitudes moderate the impact of our experimental treatments, is examined in Table 1. The table presents three binary logistic regression models for each of our dependent variables: the first shows treatment effects alongside the direct effects of attitudes toward assimilation; the second adds an interaction between treatment effects and attitudes towards assimilation; the third adds a battery of control variables, including gender, age education, income, immigrant status, French as a first language and so forth. This last model in part allows us to minimize the effects that might be due to inequality across treatment groups that were a by-product of the sampling (see Appendix II for a discussion).
Table 1 Support for Specific MCPs (Binary Logit)
* p < 0.10, ** p < 0.05, ***p < 0.01.
Note: Table presents coefficients from a binary logistic regression, with standard errors in parentheses. Canadian Election Survey, 2011.
These models clearly show that attitudes toward assimilation have a significant impact on support for specific MCPs. Those who score higher on assimilation are less likely to support multiculturalism policies (p < .01) in both policy cases. This is true in the base (uncontrolled) models (1 and 4), and remains true even as we add interactions and controls. More importantly, these attitudes moderate the experimental treatments. Coefficients in models 2/3 and 5/6 suggest that those with more assimilationist attitudes react more strongly to the target group cues in the experiment.Footnote 16
We also note some important results amongst our control variables in models 3 and 6, for instance: education is positively associated with support, age is negatively associated with support, those with higher incomes are more supportive of community centre access and French- speaking respondents give generally less support to the community centre access. The significance of these findings lies in the moderating impact of assimilationist attitudes, however. These are difficult to discern from the raw coefficients in Table 1, but Figures 2 and 3 illustrate these critical findings.
Figure 2 Support for Community Centre Access by Treatment and Prior Attitudes toward Assimilation
Figure 3 Support for Financial Grant by Treatment and Prior Attitudes toward Assimilation
These figures show the estimated probability of supporting access to the community centre and funding, across the three assimilationist-attitudes groups, based on models 3 and 6 in Table 1. In so doing, they address Hypothesis 3: people with higher levels of assimilationist attitudes should be more prone to make distinctions between the Muslim and Catholic out-goups. Figure 2 shows community centre support. Note that treatment effects of high assimilationist-attitudes respondents are roughly in line with Hypothesis 2: support for community centre access decreases (albeit marginally) as we present more culturally distant groups. The same is true for those in the middle of the assimilationist scale; but the effect is nearly the opposite for low assimilationists. Cultural distance clearly matters differently across respondents with high versus low assimilationist attitudes.
The dynamic for funding is similar, though it appears to be driven more by low assimilationists rather than high assimilationists. Increasing cultural distance leads to a very slight decrease in support among high assimilationist respondents. The main story in this instance is that those with medium and especially low assimilationist attitudes show increasing levels of support for spending on multiculturalism for more culturally distant groups. The division amongst low and high assimilationists for the Turkish woman with a hijab is a function of both downward movement among high assimilationists and upward movement among low assimilationists.
Hypothesis 3 is clearly borne out in Figures 2 and 3: assimilationist attitudes moderate our treatment effects, and allowing for this interaction provides a much better picture of the structure of MCP support. MCP policies are more divisive (that is, generate a wider range in support/no-support) when they are directed at Muslims rather than Catholics. This is true for both MCP policies. The hijab itself emphasizes the difference between people with different levels of assimilationist values. Put differently: community centre access support for the Portuguese Catholic woman produces majority support across all three assimilationist groups, ranging from an average estimated level of support of roughly .58 to .75. When confronted with a Turkish Muslim woman wearing a hijab, the difference in support is more polarized, with those high on assimilationist values dropping below majority support (where the range in support across groups is roughly.47 to .85.) As we have seen, a similar trend is evident for financial support, although in this case the only instance of majority support is for the Turkish woman with the hijab, and only among those with low assimilationist attitudes. Clearly, support for multiculturalism policy is heavily affected by financial costs.
Conclusion
This article is one of the first to measure specific support for multiculturalism policies for two types of benefits. While Canadians are often portrayed as overwhelmingly supportive of multiculturalism, this study demonstrates that this support is more restrained. Support depends on the type of program benefits under consideration as well as the groups that benefit from such programmes. Most importantly, it also varies considerably across people's predispositions toward assimilation. When confronted with more cultural distant recipients, such predispositions play a particularly important role.
Our findings suggest that Canadians are more willing to consider offering access to public space to minority groups than to support an $80,000 grant for an ethnic program. While symbolic and accommodative policies tend to be contentious in public debates, in this case we deal with a relatively easy and low-cost accommodation within a local public space and find more support for it compared to a relatively costly financial program. This points to the importance of financial considerations in citizens’ support for multiculturalism policy. More expensive forms of support are likely to receive more resistance.
Moreover, our results suggest that support for multiculturalism policies is conditioned by a combination of the ethnicity of recipients and attitudes about assimilation. Canadians who believe more strongly in assimilation are markedly less supportive of MCPs overall and tend to penalize more culturally distant groups; though even the most assimilationist groups still support inexpensive multiculturalism policies that cater to Christian immigrants from Europe. These same assimilationists are less likely to favour more expensive policies that offer direct cash grants to ethnic organizations and are more likely to oppose multiculturalism policies of all kinds when they cater to Muslim immigrants. Conversely, those who strongly reject assimilationist values tend to be more supportive overall and particularly for groups perceived as more culturally distant. The implication for Canada is that rising diversity with a higher share of more distant minority groups might make multiculturalism policy more polarized. For example, immigration flows from crisis regions in the Middle East such as Syria, Iraq and Afghanistan might cause substantial policy disagreement on the use of multicultural and accommodation policies overall. This disagreement might also strengthen when it comes to the discussion of integration requirements that are loosely related to multicultural policy, though not tested here. Thus, while multiculturalism policies tend to emerge in diverse societies, they also appear to be divisive within the population, particularly based on people's pre-existing attitudes toward integration.
Note that our sample of Canadians is not perfectly representative of the general population. As discussed above, the participants of the fourth wave are more educated and richer than other survey participants who are in all likelihood already selected and of slightly higher social status than Canadians overall. Future studies can explore if these relationships hold in groups with lower socio-economic status.
Even so, we believe that these results are informative about support of multiculturalism in Canadian society. Clearly, support for multiculturalism, although relatively high when measured with less concrete questions, is not independent from the desire for integration and even assimilation. While attitudes toward assimilation and support for multiculturalism are negatively related, the median Canadian tends to have a moderate score on both attitudes toward assimilation as well as moderate support for multiculturalism policies. Thus we cannot characterize Canadians as people who unconditionally support multicultural policies. Even in this more highly educated and politically involved sample, Canadian support for MCPs is embedded in the understanding that these policies integrate and, to a certain degree assimilate, minority groups. This differentiated view of Canadians will shape the future of multicultural and integration policy in the broadest sense and potentially cause debates on the universal extension of these policies across all new incoming groups.
We suggest several extensions of our current work. One is a shift in context: future research should examine whether this type of support also exists in countries with smaller immigration flows and/or with less developed multiculturalism policies. Another relates to changes in our treatments; for instance, it is unclear whether the same results hold when an immigrant organization seeks prayer space or a more concrete religious demand. It might also be true that the headscarf might not be a strong symbol of cultural distinctiveness in Canada, at least for some, and thus does not trigger the same reactions as a niqab potentially would—as became clear in the current Canadian election campaign of 2015 (see also O'Neill et al., Reference O'Neill, Gidengil, Côté and Young2015).
More broadly, we believe that future work should develop concrete questions about support for multicultural policies in order to further explore the multiple facets of public perceptions and support for multiculturalism as a policy. Support for multiculturalism should be measured across a number of different types of policies and different levels of symbolic policies, accommodation and financial support. More concrete scenarios varying the dollar amounts and types of accommodation might provide a better understanding of support for these policies and the limits of this support. And, armed with these more concrete measures of support, future work should set out to more thoroughly explain policy attitudes. Other out-group attitudes and experiences with out-groups, for example, may be important for understanding varying levels of support. And a cross-national comparison of such policy support and its predictors could provide richer insight into the dynamics between different types of levels of diversity, out-group attitudes, overall experiences with multiculturalism and policy support.
There is little doubt that the accommodation of diversity is one of the central political issues of the early twenty-first century. It is incumbent on social scientists to improve our understanding not just of the political psychology of attitudes towards immigration and diversity, but also the scope and nature of support for actual multicultural policies.
Appendix I: Vignette Text
Now we would like to know what you think about multiculturalism programs in Canada. For example, please consider the following story:
Appendix II: Descriptive Statistics of the Sample in Experimental Vignettes
Appendix II illustrates the descriptive statistics of the sample on our experimental vignette capturing specific multicultural policy support in Canada (again using the web wave). Comparing the three experimental groups (see Appendix I, group 1 = Portuguese-Catholic, group 2 = Turkish-Muslim and group 3 = Turkish-Muslim with a hijab) there are minor potential threats to internal validity due to uneven distributions between the three groups even though they were randomly assigned. Looking at age, group 1 had 10 per cent more participants in the 35–54 age range than group 2, and 13 per cent more than group 3. In the 55+ age category, group 1 had 13 per cent less than group 2 and 11 per cent less than group 3. Looking at age as a continuous variable, the average age for group 1 is lowest at 54.2, group 2 is highest at 56.6 and group 3 has an average age of 55.6. A two-tailed t-test reveals that the average age for group 1 is significantly different than the other two groups at p < 0.1 (p = 0.09). There are also 7 per cent fewer females in group 2. This difference is significant from the other two groups at p < 0.01 (p = 0.09). At the very least, we thus need to include these socio-demographic variables as controls in our robustness checks.
It should also be noted that participants in the fourth wave web experiment differ from those in the first wave campaign survey for reasons of attrition. Two-tailed t-tests indicate that participants in the web experiment are significantly older by around two years at p < 0.01, wealthier at p < 0.01, more likely to be male at p < 0.05 and more educated at p < 0.01. Thus the sample for the web experiment may yield conservative estimates, since individuals who are more educated and have higher incomes tend to feel less threatened by diversity and immigrants (Citrin and Sides, 2008). Respondents of the web wave live also in slightly more diverse census tracts than the respondents of the campaign wave (11.9 per cent visible minority versus 10.6 per cent of the entire sample). Again, we will control for these differences across our manipulation groups in some of the analyses below.
Descriptive Statistics of Web Treatment
