Introduction

The influence of racial attitudes on a wide range of political evaluations is well established (e.g. Gilens, 2000; Bobo & Johnson, 2004; Valentino et al., 2020; for non-U.S. evidence, see, Ford, 2006; Harell et al., 2016). Past research, however, has generally been premised on the existence of a direct association between the target of the evaluation—such as a policy or political party—and the racial group in question. In the case of White Americans’ welfare attitudes, for example, the perception that African Americans are the principal beneficiaries of welfare programming serves to create a cognitive link between racial attitudes and opinion on welfare (Gilens, 2000). Recently, however, Tesler (2012, 2015, 2016) has proposed a less direct source of racial influence in political attitudes: the racial characteristics of prominent political figures associated with policies, parties or other objects of evaluation. The salient connection between, for instance, a politician and her policy proposals may suffice to generate an association between racial and policy attitudes via the politician’s perceived racial characteristics. In a key illustration of his “spillover of racialization” theory, Tesler (2012) shows that Barack Obama’s proposed healthcare reforms led to a linking of—previously largely unrelated—racial and healthcare attitudes. The upshot of such “racial spillover” is that opinion regarding policies and other political objects that in themselves have no manifest connection to race (real or imagined) becomes infused with racial attitudes.

While the growing body of findings on racial spillover is impressive, it is limited by a strong focus on a single, and rather exceptional, political figure: former U.S. President Barack Obama (Benegal, 2018; Luttig & Motta, 2017; Sheagley et al., 2017).Footnote 1 Tesler (2015, 2016) himself has pointed to features of the U.S. context and the “Obama case” that may elevate the potential for racial spillover, such as the salience of Obama’s policy positions, Obama’s historic status as the first Black president, the peerless significance of race as a social identity in American politics, and the centrality of the presidency in the American political system.

Accordingly, we extend racial-spillover theory to a new case: the attitudinal influence of New Democratic Party (NDP) leader Jagmeet Singh, the first non-White leader of a major federal political party in Canada. A longstanding center-left party in Canadian politics, the NDP was formed in 1961 out of the Cooperative Commonwealth Federation, which was founded in 1932. The NDP regularly serves in government at the provincial level and reached its electoral high point at the national level in 2011, when it captured the second-largest number of seats in parliament. Thus, Singh’s ascent to the leadership of the NDP, in October, 2017, is an important opportunity to expand the evidence base regarding racial spillover, given that major parties in established democracies in Europe and North America rarely select leaders from minority racial groups. Indeed, like the Obama case, Singh’s capturing of a major party leadership was, as noted, a historic first for Canada. Unlike the Obama case, however, Singh’s party does not dominate its competitors on issues of racial equality—the NDP, we argue, does not “own” the issue of combatting racial discrimination and inequality in the way the Democratic Party does in the United States. Also, as leader of an opposition party (albeit an enduring one), Singh does not occupy a position of high centrality in the Canadian political system. Finally, while levels of racism and racial discrimination in Canada are significant, race is not highly salient in Canadian electoral politics, which has historically been anchored in other social identities, such as language, religion and region.

Our analysis combines an experimental study of policy attitudes during the 2019 Canadian federal parliamentary election—Singh’s first as party leader—with an observational study of evaluations of the NDP over more than three decades. Both studies uncover clear evidence of racial spillover. Consistent with past research, in our experimental study, we find the largest racial-spillover effect involves opinion on a policy that lacks a manifest connection to a racial group. We also find suggestive evidence that effects differ across partisan groups in ways that render them electorally consequential. In the observational study we find a sudden, sharp increase in the correlation between racial attitudes and evaluations of the NDP in the 2019 election, a difference that is robust to extensive controls and two alternative measures of racial attitudes.

Taken together, our findings suggest that racial spillover does not depend on peculiar features of U.S. politics or of the U.S. presidency. Specifically, we argue that the centrality of race in American politics and of the presidency in the American political system are not necessary conditions for racial spillover. Likewise, our results suggest that the Democrats’ high level of dominance on racial justice issues is not a prerequisite for spillover of racialization to occur. Thus, to the extent these features of the Obama case are exceptional, our findings suggest the phenomenon may generalize to other racialized political leaders and to diverse political contexts.

The implications of racial spillover are potentially profound. As a source of racialization in politics, it is highly insidious. Whereas explicit “race-baiting” may violate norms of racial equality in public discourse (Mendelberg, 2001), the conditions for racial spillover are almost inevitable, given that politicians’ salient background characteristics will often be communicated automatically through their participation in political conflict (Tesler, 2012, p. 702). More broadly, the imprint of racial attitudes, owing to their strength and affective intensity, can impart a rigid and emotional quality to political discourse—a point emphasized by Tesler (2016). In this way, racial-spillover effects reinforce larger trends toward polarization, which are present in Canada (Kevins & Soroka, 2018) just as in the U.S. (Iyengar, 2019).

Racial Spillover Across Attitudes and in Context

Tesler (2012, 2015, 2016) proposes a number of conditions that may increase the likelihood that an elite source’s association with a given political object (e.g., a policy) will create (or strengthen) an association between attitudes toward the object and attitudes toward the source’s perceived racial group. Some of these conditions concern properties of the attitude in question, while others involve features of the political context that connect (a.) the political object to the source and (b.) the source to the racial group.

As regards properties of political attitudes, a critical proposed condition, or moderator, is ex ante attitude strength: consistent with standard findings on attitude change, Tesler proposes that racial spillover is more likely in relation to less developed, less stable political attitudes. Thus, he observes stronger racialization with respect to weakly crystallized health care attitudes than in relation to party identification, which is perhaps the most persistent of political attitudes (2015, p. 107). Relatedly, another important potential moderator of racial spillover is the baseline racialization of the attitude in question—that is, the impact of racial attitudes on policy views or other political evaluations prior to, or in the absence of, exposure to messages that associate the racialized elite source with the attitude-object in question. Tesler proposes “ceiling effects” on racial spillover, such that racialization is decreasing in the prior weight of racial attitudes. “We would not expect,” Tesler writes, “Obama’s support for immigration reform… to produce significant spillover effects… because racial and ethnocentric attitudes were already strong determinants of Americans’ immigration opinions” (2015, p. 104). For the same reason, racial spillover should be weaker in relation to novel issues with readily perceived racial implications.

Less familiar than the attitude-level moderators, and generally more relevant to the present analysis, are the features of political context Tesler identifies as important potential facilitating conditions for racial spillover. One such condition is the salience of the connection between the elite source and the political object in question (Tesler, 2016, pp. 33–34). For instance, a policy position that is clearly articulated by the source and reinforced by the media over time is more likely to engender racial spillover than one that is relatively obscure. In the case of Obama, because his pro-choice stance on abortion was relatively less salient than his position on health care, racial spillover was more likely in the latter case than in the former (p. 33).

Another set of contextual moderators concerns associations between the source and the racial group of which they are perceived to be a member. Importantly, group membership, as such, is not enough; rather, Tesler proposes that it is “a politician’s perceived solidarity with his or her social group” that matters (2016, p. 200). Simply put, the politician’s (or source’s) life experience, policy positions and political affiliations must be seen as in alignment with the perceived interests of the racial group in question. Tesler suggests, for instance, that Obama’s perceived “rootedness in African-American life as a community organizer in Chicago” may have helped establish that “he was not only black but also for African Americans” (p. 201). Similarly, a racialized politician’s affiliation with a political party that is perceived to favour the interests of their racial group is likely necessary to support a relationship of “perceived solidarity” between the politician and the group (p. 201).

The broader politics of race may also be important here. First, the novelty of racialized politicians in a given political context may engender stronger associations between the politician (qua elite source) and their perceived racial group. As Tesler writes, “[t]he next black president… would lack Obama’s historic position as the first”, which, he suggests, may mean the next Black president will not have the “same racializing effect” (2016, p. 202). Elsewhere, Tesler emphasizes the role of the “profound racial symbolism surrounding Obama’s position as the first black president” in racial spillover involving the former president (p. 692). Arguably, Tesler’s suggestion is that, against a backdrop of historical and continuing racial oppression, racialized politicians in novel political settings are inevitably seen as identified with the perceived interests of their racial groups (all other things being equal). Secondly, the pervasiveness of race and racial attitudes in political life generally may increase racial spillover with regard to particular issues. One important feature of the U.S. case may be that, as Tesler writes, “race has historically been more salient in American political life than other social identities” (p. 201). As a result, citizens may more readily consider—and by extension, identify—links between a racialized politician and their perceived racial group.

A final important feature of political context may be the elite source’s position in the political system. As U.S. president, Obama was “the center of the political universe” (Tesler, 2016, p. 201), which may have facilitated racial spillover with respect to evaluations of his policies and closely associated political actors (e.g., the Democratic Party). A straightforward possibility is that the high level of attention paid to President Obama by the media meant that information about his policies and affiliations, and about his connection to his racial group, were very widely disseminated and internalized. At the same time, the president’s, or any senior political office-holder’s, status as a political leader may be important in itself. The policy positions of prominent political leaders, as opposed to other politicians, are important political facts to which citizens may ascribe greater significance and pay more attention, inasmuch as prominent leaders’ positions have greater potential to shape political outcomes. The result, in turn, may be a higher likelihood that attitudes toward the leader, as an elite source, may become linked with (that is, spillover to) policy attitudes and other political evaluations.

To summarize, Tesler’s analysis suggests that racial spillover is most likely (1.) in relation to weak attitudes that are not already racialized, (2.) when the association between the source and object of evaluation is relatively salient, (3.) when a relationship of “perceived solidarity” exists between the source and their perceived racial group, and (4.) when the broader political context encourages attention to the source’s perceived racial characteristics, policy positions and political affiliations.

Racial Spillover, Canadian Racial Politics, and Jagmeet Singh

How might we expect racial spillover to manifest in Jagmeet Singh’s influence on Canadian political attitudes? Certainly, as in the U.S., we expect racial spillover to be moderated by attitudinal features—that is, we expect less spillover in relation to strong attitudes and those that are already highly racialized. Likewise, we anticipate that a salient connection between Singh and the political object in question will be a necessary condition for racial spillover.

As regards associations with his perceived racial group, like Obama, Singh had established a record of advocacy on issues of broad concern to racial minorities during his time as deputy leader of the Ontario NDP, prior to his entering federal politics in 2017 (Zimonjic, 2017). In one telling episode, Singh advocated for a ban on the practice of “carding”, which involves police officers stopping and questioning individuals without making an arrest. As he remarked at the time, “people that are black and brown or racialized people are stopped far more often than other folks” (CBC, 2015). Perceptions of Singh’s solidarity with racial minorities as a group were likely solidified during the 2019 election campaign, given the role he played in the fallout from the so-called “blackface controversy”. The controversy involved the publication, during the campaign, of photos and a video of prime minister Justin Trudeau wearing blackface in high school, at a summer camp in the 1990s, and while a teacher at a private school in 2001 (CBC, 2019b). While the photos themselves attracted extraordinary attention, for our purposes, the significant feature of the controversy was Trudeau’s request, in the aftermath of the photos’ publication, to speak with Singh in order to apologize “to him personally as a racialized Canadian” (CBC, 2019a). The episode effectively cast Singh in the role of confessor for Trudeau’s racist sins, and Singh’s astute handling of the controversy was widely credited with reviving the NDP’s campaign fortunes (Scherer, 2019). The details of Singh’s background and the campaign aside, perceptions of Singh’s solidarity with racial minorities may also have been enhanced by the fact that, like Obama, his political ascent reflected an important first: he was the first non-White leader of a major federal party.

One important difference between Singh and Obama concerns the wider context of the party system. To be sure, like Obama, Singh’s partisan affiliations are compatible with the perception that he is an advocate for the interests of racial minorities, as those interests are typically understood in Canada. While the party commonly emphasizes its social democratic credentials, the NDP has for decades clearly placed itself on the side of racial equality. Emblematic of this commitment, Bird (2005) notes that “[t]he NDP is one of very few parties in the world that have formal guidelines for recruiting more visible minority candidates” (p. 453). At the same time, a key contrast with Obama’s partisan affiliations is that, unlike the Democratic Party, the NDP does not dominate its opponents on such issues. On the contrary, it is the Liberal Party that, historically, has been seen as the natural home for the votes of racial and ethnic minorities, and until recently, the party has attracted an outsize share of the votes of racialized Canadians (Blais, 2005). More generally, Canada’s major parties are much less divided than Republicans and Democrats on the benefits of racial and ethnic diversity (Banting & Thompson, 2016); indeed, as Ambrose and Mudde (2015) observe, “multiculturalism has remained the official government policy, irrespective of the party composition of the government” (p. 228).Footnote 2 In short, whereas Obama’s Democratic Party might be portrayed as alone in defending the principle of racial equality, Singh’s NDP is better seen as embracing a widely shared value of racial equality that is rarely questioned openly in national politics. As such, Singh’s partisan affiliations may not engender, to the same degree as Obama’s, an automatic identification with the interests of racial minorities.

A second obvious difference between Singh and Obama is that, whereas Obama was President of the United States, Singh was leader of an opposition party with the third-largest parliamentary delegation at the time of the election. Indeed, the party held fewer than half as many seats as the next largest party in the House of Commons. Moreover, the party’s fortunes were flagging at the onset of the election and there was even speculation Singh’s leadership would not survive the campaign (Hui & Dickson, 2019). Thus, notwithstanding the surge of attention he received following the blackface controversy, Singh’s level of “political centrality” was undoubtedly lower than Obama’s.

Finally, racial identity and racial attitudes are simply not as salient in Canadian politics as they are in American politics (Banting & Thompson, 2016). The most prominent social cleavages in Canadian politics have historically been defined by regional economic differences and the divide between English- and French-speaking Canadians (Johnston, 2019). Of course, Canada is not free of racism and racial discrimination. While attitudes toward immigrants in Canada are clearly more positive than in many other countries (Reitz, 1988), at least one-third of Canadians have straightforwardly negative views of diversity and dislike immigration, multiculturalism, and racial minorities (Besco & Tolley, 2019). Recent polls confirm this: 40% think immigration is too high (Angus Reid, 2019), 33% think multiculturalism is bad for Canada (Research Co., 2019), and—especially relevant for the present study—37% said they could not vote for a candidate who, like Singh, was Sikh (Angus Reid, 2017; see also Bouchard, 2021). There are, furthermore, wide racial disparities in terms of income (Reitz & Banerjee, 2003) and audit studies in employment find roughly similar levels of discrimination in Canada as in the United States (Quillian et al., 2019). Nonetheless, in contrast to the U.S., issues of racial discrimination are rarely a focus of attention during Canadian elections, much less a point of sharp partisan disagreement, as noted above.Footnote 3 The implication is that Canadians may be less likely than Americans to spontaneously identify political leaders with their perceived racial groups.

Racial Spillover and Policy Attitudes: Experimental Evidence

Research Design

We begin our investigation of racial spillover with a study of policy attitudes. The essence of our approach is to exploit the natural variation that Singh provides on certain contextual moderators of racial spillover, while manipulating experimentally a key attitudinal moderator. Singh, as we argue above, is somewhat less likely than Obama to enjoy a relationship of perceived solidarity with racial minorities and their interests, given the relatively lower polarization of Canadian parties on questions of racial equality. Further, the lower salience of racial identities and attitudes in Canadian politics suggests that Singh, relative to Obama, may be less likely to be spontaneously associated with his perceived racial group. The logic of our experimental design, then, is to check the importance of these contextual differences by varying the salience of the connection between Singh and a set of his policies. Our reasoning is that, to the extent the salience manipulation strengthens the impact of racial attitudes on relevant policy judgments (in other words, promotes racial spillover), the particular form of solidarity with racial minorities that Obama enjoyed, and the special salience of racial identities and attitudes in the United States, cannot be necessary to the existence of racial spillover.Footnote 4

We note that this research design does not provide leverage on the importance, in racial-spillover dynamics involving Obama, of the president’s uniquely central position in the American political system—at least to the extent that this mechanism operated through the exposure it conferred on Obama’s policies and affiliations.Footnote 5 By design, the salience treatment in our experiment, described below, should effectively erase any variation between the two leaders in the level of exposure enjoyed by their policy commitments. Conversely, our observational study of party evaluations, reported in the next section, does provide some leverage on the role of the “political centrality” factor, albeit as part of a conjunction of contextual differences between the two cases.

Methods

Our experimental study was a module included in the online component of the 2019 Canadian Election Study (CES; for more information, see below) that was fielded to 800 respondentsFootnote 6 (although, in the analyses, the sample size is reduced by missing data). Interviews were completed between October 4th and 20th, 2019. Notably, the fieldwork began more than 2 weeks after the publication of the first blackface photos and, thus, it is reasonable to assume that a majority of our respondents will have been aware of the resulting controversy, including Trudeau’s subsequent dialogue with Singh.Footnote 7

We asked all respondents to evaluate five policy proposals prominently featured on the NDP’s website. In particular we asked about policies designed to: (1) “Make it easier to get Employment Insurance (EI) when people need it”; (2) “Stop racism and discrimination online by requiring offensive posts to be removed from social media”; (3) “Create a national pharma-care plan that pays for prescription drugs”; (4) “Promote the use of Indigenous languages”; and (5) “Ban single use plastics, such as plastic bags and straws”. The order of presentation of the policies was randomized and attitudes were elicited using a pair of 10-point scalesFootnote 8; in the analysis, we average responses to the measures.

The five policies were selected to create variation with respect to both attitude strength and baseline racialization. In this way, we are able to add to existing evidence regarding the importance of these attitudinal moderators of racial spillover (see, especially, Tesler, 2015). In view of Tesler’s expectation that racial spillover is decreasing in the strength and baseline racialization of the attitude in question, we expect the new and not-obviously racialized issue of banning single-use plastics to manifest the largest spillover effects. Two issues in relation to which we expect much less racial spillover, for reasons of attitude strength, are Employment Insurance and prescription drug coverage. Employment Insurance is a longstanding, familiar political issue in Canada and, as such, likely to be the object of strong attitudes (Carmines & Stimson, 1980). While prescription drug coverage is, strictly speaking, a new issue, its obvious connections to healthcare—a perennially important issue in Canadian elections—suggests attitudes toward it may have the properties of more crystallized (that is, stronger) issue attitudes. Lastly, while removing racist posts on social media and Indigenous language promotion are relatively new issues, they are manifestly connected to racial groupsFootnote 9; thus, we also expect less racial spillover on these issues than in regards to the ban on single-use plastics.

Our experimental manipulation is embedded in the wording of the policy questions and aims to subtly moderate the salience of Singh’s connection to the policy position in question (see Online Resource [OR] Appendix C). In the “party condition”, respondents are asked about “some policies the NDP is campaigning on” and a party logo is shown on the same screen above the policies. In the “Singh condition”, the questions are instead about “some policies Jagmeet Singh and the NDP are campaigning on”, and a photo of Singh appears above the policies. Importantly, the reference to the NDP in both conditions is designed to hold constant the salience of the party and, thus, better isolate the effect of making Singh more salient.

Our principal measure of racial attitudes is from the CES’s core questionnaire. The measure is a “thermometer”-type question, which asks, “how do you feel about racial minorities?” Responses are elicited using a 0–100 scale. One virtue of this question is that it has been asked on the CES since 1988, which allows us, in a later section of the paper, to track its influence in political attitudes over time.

We also discuss, below, various alternative specifications including additional variables. Question wordings and detailed results for these analyses are presented in Appendices C and A, respectively.

Results

We first examine the difference in the relationship between support for the policies and racial attitudes across the party and Singh conditions. For this first analysis we estimate an OLS regression model of policy support, averaging responses across the five policies, with each policy attitude scaled to the 0–1 interval. The dependent variable is, thus, average policy support, and the independent variables are the treatment condition, the racial thermometer (scaled from 0 to 1) and its interaction with the treatment, and controls for age, education, gender, unemployment, union membership, party identification, and province (all pre-treatment; question wording and coding details in OR Appendix C). The results are substantively the same with and without controls. We plot predicted values in Fig. 1 and report full results in OR Appendix B (see Online Table 1B).

Fig. 1
figure 1

Support for policies by feelings toward racial minorities. Left panel: predicted values with covariates set to means; right panel: difference between the party and Jagmeet Singh conditions. n = 747

Full size image

As the left panel of Fig. 1 makes clear, respondents who dislike racial minorities are less supportive of the NDP’s policies when Jagmeet Singh’s connection to them is made salient. At the bottom end of the racial thermometer, the difference is nearly one-tenth of the range of the policy support measure and statistically significant at the 95% level (see Fig. 1, right panel). At the top end of the scale there is a reverse effect—those most warmly disposed to racial minorities support the policies more when Singh is salient: the treatment effect is positive and statistically significant at the 95% level for the 13% of respondents with thermometer ratings of 100; the effect is significant at the 90% level for the 18% of the sample with thermometer ratings at 90 or above. This pattern of symmetrical results across the range of racial attitudes is consistent with Tesler’s (2012) finding of simultaneous activation of racial liberals and conservatives (pp. 698–700).

Notably, there is a strong correlation between the racial thermometer and the policy support measure even in the party condition. This is consistent with the observational evidence presented in the next section, which shows that evaluations of the NDP have been increasingly racialized since Singh’s ascent to the party leadership. Given the apparently high level of ex ante racialization of the NDP, therefore, our experimental estimate of the magnitude of the racial-spillover effect is conservative.

We next consider variation in the spillover effect by partisanship, given that the political significance of the phenomenon depends on its distribution across voters. If the effects of racial spillover on support for NDP policies were concentrated among (center-right) Conservative partisans, for example, then their political consequences would be of little concern to the (center-left) NDP, since the support of Conservative partisans is unlikely in any case. On the other hand, if racial-spillover effects cause a party’s own partisans and potential supporters to, for example, turn against their leader, then the effects may have serious political consequences. To examine this question, we repeat the analysis above within partisan groups. Figure 2 shows the predicted treatment effects by party identification.

Fig. 2
figure 2

Racial spillover by partisanship. Predicted values with covariates set to means; difference between the party and Jagmeet Singh conditions shown. n = 747

Full size image

There are clearly racial spillover effects for both NDP and Liberal partisans: those with negative feelings toward racial minorities are less likely to support policies when they are clearly connected to Jagmeet Singh. For both partisan groups, the treatment effects are quite sizable at the bottom end of the racial attitudes measure: the difference in policy support between the party and Singh conditions is nearly one-fifth of the range of the support measure (significant at the 95% level). Conversely, there is no evidence of similar effects for either Conservatives or non-partisans. Having said that, we have limited power to estimate effects within partisan groups, particularly among non-partisans, and none of the between-party differences in the treatment effect is statistically significant.Footnote 10

Finally, we examine racial spillover by policy attitude. Figure 3 is similar to Fig. 1, but shows a separate estimate for each policy (see OR Appendix B, Table 2B, for full regression estimates). Recall that, in view of Tesler’s expectation that racial spillover is inversely related to levels of attitude strength and prior attitude racialization, we anticipate the largest spillover effects on the relatively unfamiliar issue of banning single-use plastics. And indeed, on their face, the results indicate that attitudes toward the single-use plastics ban show the largest racial-spillover effect. Surprisingly, the next largest effect concerns the Employment Insurance (EI) issue—an issue in relation to which we had anticipated less spillover, owing to the issue’s relative familiarity. This result implies that, while EI is not commonly racialized in Canadian political discourse, there is room for such links to be created. Conversely, as expected, we have no clear evidence of racial spillover in relation to prescription drug coverage, which we reasoned may attract strong attitudes like other issues in the health care domain. Likewise, we see no racial spillover regarding removing racist social media posts or Indigenous language promotion, two issues with transparent connections to the interests of racialized peoples. Regarding between-policy comparisons in the treatment effect, the only significant difference is between Indigenous language promotion and the plastics ban (significant at the 95% level).Footnote 11

Fig. 3
figure 3

Racial spillover by policy. Predicted values with covariates set to means; difference between the party and Jagmeet Singh conditions shown. n = 747

Full size image

Alternative Specifications

We probe the specificity and robustness of the results in a number of alternative specifications. First, we check whether the effects we uncover apply only to racial attitudes. To the extent this is the case, it further confirms the theorized mechanism of racial spillover, that is, the activation of racial attitudes by racialized political figures. Accordingly, we estimate the model underlying Fig. 1, but, in place of the racial thermometer, substitute one of six alternative attitude measures (results reported in OR Appendix A, Table 3A; question wordings in OR Appendix C). Five of the measures fall within the broad domain of ethnic diversity and multiculturalism, but lack explicit racial content and focus on other concerns: perceptions of refugees; the rights of religious minorities; immigrants’ use of government benefits; immigrants’ adoption of Canadian values; and feelings toward immigrants. The only item explicitly focused on race queries agreement with the proposition that “[i]t is more difficult for non-whites than whites to be successful in Canada.” If the mechanism of racial spillover is activation of racial attitudes, then only this last “racial-inequality” item should interact with the experimental treatment. The results show that, indeed, the racial inequality item does interact with the treatment (p < 0.10). As for the other alternative measures, the interaction with the treatment is statistically insignificant and substantively small in each case, save for the interaction with feelings toward immigrants, measured with a standard thermometer (p < 0.05). Note, however, that the immigrant thermometer is very highly correlated with the racial thermometer in our data (r = 0.79), which may reflect the racialization of immigration in contemporary Canada.Footnote 12 Overall, then, we conclude that results using alternative attitudes provide further support for racial-spillover theory by showing that the phenomenon operates specifically through the activation of racial attitudes.

Second, we estimate a model that controls for attitudes toward Jagmeet Singh. After all, citizens commonly rely on leader evaluations to inform policy judgments (Lenz, 2012) and these evaluations, in turn, may be influenced by racial attitudes. As such, our experimental treatment may simply activate this familiar process, rather than racial spillover. As we report in OR Appendix A, Table 1A, while evaluations of Singh have a sizable effect on policy attitudes, this effect does not interact with the experimental treatment. Furthermore, even controlling for Singh evaluations and their interaction with the treatment, the interaction between the treatment and the racial thermometer is substantively unchanged and continues to be statistically significant.Footnote 13

Summary of Experimental Findings

Overall, we find clear evidence of racial spillover in Canadian policy attitudes. The effects are politically consequential (that is, they occur among Singh’s own co-partisans and ideologically adjacent Liberal partisans), involve racial attitudes specifically, and are robust to alternative specifications. These results suggest that racial spillover does not depend on President Obama’s particular relationship with racial minorities or the unique politics of race in the U.S. We also find suggestive evidence of moderation of racial spillover across policies consistent with Tesler’s arguments, though, contrary to expectations, we do find evidence of spillover in relation to one familiar policy area; a larger sample is required to more precisely identify racial spillover effects (and lack thereof) on different types of policies.

Racial Spillover and Party Evaluations: Over-Time Evidence

Next, we examine “real-world,” or observational, evidence of racial spillover in Canadian attitudes. To do so, we draw on data from the Canadian Election Study over 10 elections, from 1988 to 2019. The CES consists of surveys of nationally representative samples of Canadian citizens (for details, see OR Appendix C). Since 1988, interviews have been conducted primarily by phone. The 2015 and 2019 studies also included interviews with separately recruited online samples (a feature that we exploit in a robustness check, below).

Our analytical approach in this section exploits over-time change in the racial characteristics of the NDP leader—that is, the transition to Singh from his (White) predecessor—to identify the impact of racialized party leaders on evaluations of their political parties. We assume, consistent with our discussion of the case above, that Singh’s arrival will have transformed the perceived relationship between the NDP leader and racial minorities. The other proposed moderators of spillover are assumed to be constant within the design, including both contextual and attitudinal variables. Importantly, unlike the experiment, the present design does not interfere with attentional differences arising from the natural variation between Obama and Singh in political centrality—i.e., the contrast between the offices of U.S. President and leader of the third party in the Canadian House of Commons. Thus, to the extent the transition to Singh’s leadership strengthens the relationship between racial attitudes and evaluations of the NDP, we reason that neither peculiar features of U.S. race politics nor of the U.S. presidency are necessary to racial spillover.

The key variables in the analysis are a feeling thermometer for the NDP—asked in the same fashion as the racial thermometer employed above—and two racial attitudes indicators. If Jagmeet Singh’s leadership led to racial spillover, strengthening the link between racial attitudes and evaluations of his party, then we would expect the correlation between the NDP thermometer and racial attitudes to be higher in 2019 than it had been previously. The first racial attitudes measure is the racial minorities thermometer, which is asked in all years of the CES included in our analysis. The second racial attitudes measure is based on an item that asks, “How much should be done for racial minorities?,” and allows responses ranging from “Much more” to “Much less”. While the second item has been asked in three fewer elections than the first, using both indicators allows us to check the robustness of our results to variation in question wording.Footnote 14 In addition, the “do for minorities” item, which was asked on the phone and online in both 2015 and 2019, allows us to compare racial-spillover effects across survey modes.

The modeling approach involves estimating, for each combination of racial attitude and survey sample,Footnote 15 two OLS regression models: first a bivariate model, including only the racial attitude in question as a predictor, and then a multivariate estimate that adds a set of controls. The controls include partisanship, a set of standard demographic variables (age, education, gender, income, province, and ethnicity) and, to control for “progressive” political values generally, a feeling thermometer for feministsFootnote 16 (see question wordings in OR Appendix C). To improve the representativeness of the estimates, we utilize sampling weights provided by the CES investigators (see OR Appendix C for details).

We plot coefficient estimates for the racial thermometer in Fig. 4 (see OR Appendix B, Tables 4B and 5B, for full results). The results show clear evidence of an increased relationship between racial attitudes and NDP evaluations in 2019: while racial and party attitudes are correlated in most years, the correlation in 2019 is much higher than in any other year. In the bivariate models (upper panel), the difference is very large: the coefficient in 2019 is 0.36, roughly 50% larger than in any previous election. Even with a wide range of controls (lower panel), the results are similar: while the differences are somewhat smaller, the relationship between racial attitudes and feeling toward the NDP is stronger under Jagmeet Singh’s leadership than in any of the past 10 elections. Using seemingly unrelated regressions to test differences in the coefficients between elections, we find that the difference between 2019 and any previous election is statistically significant (the largest p-value is 0.02, obtained in the comparison with the 2006 model [multivariate results]).

Fig. 4
figure 4

Racial minorities and NDP thermometers, 1988–2019. OLS coefficients and 95% CIs shown

Full size image

Figure 5 plots the results of similar models using the “do for minorities” question and reveals a very similar pattern, with the 2019 coefficients approximately 50% larger than those in any previous year. Notably, the difference in the magnitude of the 2015 and 2019 coefficients appears roughly the same whether comparing estimates from the phone or online samples across these years. This result helps allay the worry that respondents’ reports of their racial attitudes may be shaded by social desirability concerns, which, if they were operating, ought to produce relatively weaker links between racial attitudes and NDP evaluations in the phone sample. Using seemingly unrelated regressions to test differences in the coefficients between elections, we again find that the differences between the 2019 samples (i.e., both phone and online) and any previous election are statistically significant (the largest p-value is 0.04, obtained in the comparison between the 2019 online sample and the 2008 sample [multivariate results]).Footnote 17

Fig. 5
figure 5

Do for racial minorities and NDP thermometer, 1993–2019. OLS coefficients and 95% CIs shown

Full size image

Two additional analyses provide further support for the above interpretations; we report these in OR Appendix A, but summarize the results here. First, we confirm that (observable) changes in sample composition cannot explain the 2015–2019 increase in the association between the racial thermometer and NDP evaluations. Specifically, pooling the 2015 samples (phone and online) and the 2019 online sample, we estimate between-sample differences in the association between the racial minorities and NDP thermometers, controlling for interactions between the racial thermometer and a set of demographic variables (age, education, gender, income, ethnicity, and province). These models account for the possibility that certain demographic groups may be both more likely to associate the NDP with racial minorities and overrepresented in the 2019 online sample (relative to either of the 2015 samples). The results (OR Appendix A, Table 11A) indicate, however, that the 2015–2019 increase in the association between the racial minorities and NDP thermometers is entirely robust to this additional specification. Second, we estimate models similar to those underlying the lower panel of Fig. 4, but substitute the thermometer measures for the other major parties, the Liberals and Conservatives. In this way, we evaluate one plausible alternative to the racial-spillover hypothesis: that the strengthening of the link between racial attitudes and views of the NDP reflects a general racialization of the party system in 2019. The results do not, however, support this alternative account: there is only mixed evidence of an increase in the correlation between the racial thermometer and party evaluations for the ConservativesFootnote 18 and no indication whatsoever of an increasing correlation involving the Liberal Party (see OR Appendix A, Figs. 3A and 4A).

Conclusion

There is considerable evidence of associations between racial attitudes and opinion toward a range of political objects. While much of the research finds such associations where one might expect them, given the objective features or symbolic content of the targets of evaluation, recent work on Tesler’s racial-spillover theory suggests racial attitudes can shape public opinion in far more subtle ways. Merely proposing a policy, it seems, can suffice to imbue responses to the policy with associations reflecting the perceived racial identity of the proposer.

Our paper adds to this emergent literature by helping to extend it beyond the case of Barack Obama’s impact on American public opinion, which has, understandably, been central to most previous research. In our experimental study we find, consistent with past research, considerable evidence of racial spillover in the influence of NDP leader Jagmeet Singh on Canadian policy attitudes. The effects are politically consequential in magnitude and, importantly, observed in relation to policies with no manifest racial content. In our analysis of over three decades of Canadian Election Study data, we find that links between racial attitudes and evaluations of the New Democratic Party were stronger in 2019 than they have been for at least ten elections. This not only provides evidence of a historical shift, but demonstrates that racial-spillover dynamics occur both in survey experiments and on the broader political landscape.

These findings, we argue, suggest that certain exceptional features of the Obama case may not be necessary conditions for racial spillover to operate. Obama’s high level of perceived solidarity with racial minorities and their interests, the uniquely central place of the presidency in the U.S. political system, the strong focus on racial identity and racial attitudes in American political life—none of these would seem to be essential foundations of the racial-spillover process. That said, we cannot say, with only two cases, whether such factors facilitate, in a probabilistic sense, racial spillover. In fact, we rather suspect that they do promote racial spillover, given the likelihood that they (jointly) reinforce linkages between presidents and their perceived racial groups, and between presidents and salient objects of political evaluation.

While the overall pattern of results fits neatly with research on racial spillover focused on the Obama case, we note one exception to this conclusion. While the survey experiment shows, consistent with earlier work, that racial spillover occurs in a new policy area (banning single-use plastics) where people are unlikely to have formed options, we also find spillover in regards to a more familiar policy (Employment Insurance)—a finding that is somewhat at odds with previous research. This suggests that the effects of racial spillover on well-known policies may deserve further research. We have relied in our analysis on the standard assumption that familiar policies will elicit strong attitudes. But, of course, many people may not have meaningful opinions even on prominent and longstanding policy debates. Alternatively, it may be that the power of racial attitudes is such that racial spillover has the potential to move even firmly held policy views.

At the most general level, the results in the paper help to sketch—and enlarge—the boundary conditions of the racial spillover theory. Barack Obama’s presidency was historic in many ways, which naturally suggests its political effects, including on attitudes and behaviour, may be exceptional. Yet the potential for racial spillover seemingly extends to a quite different political leader and context. To be sure, Jagmeet Singh’s ascent to the leadership of the Canadian NDP was historic in ways that resonate with Obama’s election as president. In a variety of important ways described in this paper, however, Singh’s political profile and significance is very different from Obama’s. The robustness of racial spillover to this variation suggests the phenomenon may travel to diverse contexts.