Ethnonationalism and White immigration attitudes

In this article, I explore how much of White Americans' opposition to immigration — opposition that is often grounded in fears of the threat that immigration poses to the robustness of America's national identity — is shaped by ethnonationalism, a set of beliefs concerning which traits are important for being a “ true ” American. Drawing on data from the 2016 American National Election Studies (ANES), I examine how ethnonationalism shapes White Americans' attitudes towards immigrants. I find that ethnonationalism is positively associated with anti-immigrant attitudes among Whites, with the effect size of my ethnonationalism measure being larger than that of any other variable in my ordinary least squares (OLS) regression model. Critically, this includes a number of variables that are already known to be strongly predictive of White attitudes towards immigration, such as Republican partisanship. Beyond anti-immigrant attitudes, I also find that ethnonationalism also predicts support for policies that would restriction immigration.

of the US (Hanson et al., 2007), generalized ethnocentrism (Kinder & Kam, 2010) and White consciousness (Jardina, 2019). The debate concerning White immigration attitudes has become even more important since Donald Trump's victory in the 2016 election, with a number of studies noting the salient relationships between antiimmigrant sentiment and higher levels of support for Trump among White voters (Hooghe & Dassonneville, 2018;Reny et al., 2019;Sides et al., 2019). The crux of the debate is whether the immigration policies championed by Trump, and that some Whites voters with latent xenophobic attitudes ostensibly endorsed, is motived by economic anxiety or whether Trump and the current immigration attitudes of White base of support are products of a "cultural backlash" (Mutz, 2018;Norris & Inglehart, 2019).
In this article, I build on the existing corpus of White immigration opinion by offering an additional explanation for White opposition to immigration. A hypothesis I seek to explore is whether anti-immigrant attitudes among Whites in the Trump era are shaped by a set of pervasive beliefs concerning the robustness of America's national identity. The pervasive set of beliefs concerning which traits are important for being "truly" American is defined as ethnonationalism . I theorize that ethnonationalism is a predictor of White opposition to immigration that is related to-although one that is not necessarily reflective of-other predictors of White opposition to immigration, including ethnocentrism, White consciousness, and negative economic evaluations.
In this respect, I seek to make two contributions to the existing scholarship. First, despite the fact that scholars have previously correlated beliefs concerning what it means to be "truly" American with higher level of opposition to immigration, 1 there are no analyses that synthesize these beliefs into a broader framework and test their robustness as a predictor of immigration opinion. However, this is a critically important lacuna in the existing scholarship, given that traits such as ancestry, nativity, language, and participation in the customs and traditions of the nation state are all theoretically important "ethnosymbolic" markers of national identity (A.D. Smith, 2009).
Because nationalism scholars hypothesize that these ethnic markers are tapping into a latent belief system about what makes one a "true" member of the nation state, the ethnonationalism construct is, hypothetically speaking, a more consequential predictor of White immigration opinion than any one of these individual ethnic markers are alone. I therefore address this important lacuna by combining four items concerning the importance of various cultural markers for being a "true American" into a single measure of ethnonationalism. Taken together, the ethnonationalist belief system may equal or even rival the effects of measures of White group identities and psychological predispositions on White immigration attitudes. Thus, if ethnonationalism exhibits a comparable or larger effect size relative to these other indicators, we can take this as strong evidence that beliefs about national identity are more consequential for understanding White opposition to immigration than theories of narrow racial group interest or generalized, ethnocentric, attitudes.
To test these expectations, I use data from the 2016 ANES to explore the relative power and influence of ethnonationalism as a predictor of White immigration opinion. My findings are consistent across a variety of model specifications. Overall, I find that ethnonationalism is a robust predictor of negative attitudes towards immigrants themselves, with the size of the standardized coefficient for ethnonationalism being larger than a number of variables that are known to be substantively associated with such attitudes. I also find that ethnonationalism is strongly associated with support for policies advocated for by Trump that would restrict immigration, including support for building a wall along the US-Mexico border. Finally, I find evidence that high levels of ethnonationalism make the effects of anti-immigrant sentiment on favourable estimations of Trump himself more salient. These findings suggest that ethnonationalism is a potentially important moderator when it comes to understanding how xenophobic attitudes lead White majorities to exhibit positive feelings towards radical political actors such as Trump.

| EXPLAINING WHITE OPPOSITION TO IMMIGRATION
White Americans exhibit higher levels of opposition relative to other racial and ethnic groups (Abrajano & Hajnal, 2017). A large amount of the scholarship on White opposition to immigration has focused on the effects of "economic anxiety" as predictors of anti-immigrant attitudes. The "economic anxiety" thesis offers two rationales for why material self-interest predicts opposition to immigration (Hainmueller & Hopkins, 2014). The first of these rationales focuses on the hypothesized effects of immigration on the real wages of White individuals. This labourmarket competition hypothesis is rooted in the assumption that the income received from work is a substantive factor in individual wellbeing. In the US context, individuals are likely to believe that immigrants increase the relative supply of low skilled labour, which resultingly leads to lower wages for native workers in manual occupations (Scheve & Slaughter, 2001). Given that native workers in manual occupations are especially likely to be vulnerable to these effects, they are thus likely to prefer restricting immigration for fear of how immigrant workers increase competition for their employment and wages (Scheve & Slaughter, 2001). Some research has found that individuals employed in occupational sectors are the most vulnerable to greater economic competition from immigrants and are thus the most likely to oppose immigration. Nonetheless, evidence in support of the labour-market competition hypothesis has been somewhat mixed. A multitude of studies find little evidence that individuals who are more likely to experience labour market competition exhibit higher levels of opposition to immigration (Hainmueller et al., 2015;Malhotra et al., 2013;Valentino et al., 2019). Though these studies cast doubt on the robustness of the labour market competition hypothesis, recent literature indicates that this hypothesis should not be dismissed outright. For instance individuals with low transferable skills in the labour market may articulate a subjective sense of job insecurity, and consequently higher levels of hostility towards immigrants (Pardos-Prado & Xena, 2019).
The second of these rationales is known as the fiscal burden hypothesis. The fiscal burden hypothesis focuses on the fiscal impacts of immigration. In the fiscal-burden model, low-skilled immigrants are assumed by natives to be a net burden on the public finances. Consequently, increases in low-skilled immigration also increases fiscal pressure to raise taxes or reduce per-capita transfers for public spending. These effects are likely to be especially apparent in a federal system such as the US, where some states have generous public services and a large number of immigrants.
Consistent with this hypothesis, evidence indicates that native workers with higher incomes are more opposed to immigration (Hanson et al., 2007). Despite these findings, it is important to qualify that there is disagreement within the immigration literature about the fiscal contributions of immigrants themselves. Indeed, US states with the fastest growing immigrant populations experience lower increases in state income taxes and smaller increases in per-capita welfare expenditures (Hainmueller & Hiscox, 2010).
Another paradigm for explaining anti-immigrant attitudes among Whites places a greater emphasis on the additional importance of cultural factors as opposed to purely material self-interest. This model posits that negative evaluations of groups that motivate attitudes in separate domains of White public opinion apply to attitudes towards immigration, too (Citrin et al., 1997). Anti-immigrant hostility can also be relatively undifferentiated across groups (Sniderman et al., 2000). Kinder and Kam (2010) make such an argument using their generalized measure of ethnocentrism in the Us Against Them. They posit that ethnocentrism is a psychological predisposition in which Whites divide the world into ingroups versus outgroups. In this model, the bifurcation of ingroups and outgroups is not neutral. Rather, Whites assign virtuous assessments to ingroup members (i.e., other Whites), while evaluating outgroup members as dangerous and untrustworthy. This generalized predisposition towards outgroups (i.e., ethnocentrism) has been found to be a robust predictor of White opposition to immigration in recent studies (Miller, 2020).
Finally, Whites' anxiety and concern for their ingroup plays an important role in shaping White opposition to immigration. In this model, Whites do not simply express a greater preference for restricting immigration because of group-specific outgroup attitudes (for instance, White animus towards Latinos). While these attitudes may certainly be a factor in opposing immigration, the primary argument is that Whites are concerned that the large influxes of non-White immigration threaten their dominance over America's culture and it's political and economic institutions.
Consistent with this theorizing, measures of White identity and White consciousness have been found to predict White opposition to immigration across a variety of measures, including negative attitudes towards immigrants themselves, as well a greater preference for immigration reduction and support for specific policies and would restrict immigration (Jardina, 2019).

| ETHNONATIONALISM AND THE THREAT TO " AMERICANNESS"
A substantial body of literature indicates that many individuals also view immigration as a threat to the robustness of America's national identity (Citrin et al., 2007;Wong, 2017). Some conceptualizations of the national ethnos and its boundaries can easily accommodate immigrants, while others draw a tight boundary around what it means to be "true" members of the national ethnos, meaning that immigrants are not so easily subsumed into the national fold. The concept of American ethnos is thus the set of ideas and sentiments that form the conceptual framework of US nationhood.
Most accounts of nationalism in the US literature aim to gauge the specific set of criteria as to what it means to be a "true" American. The crux of the debate in the literature has been between creedal (or civic) forms of nationalism on the one hand, and those rooted in ethnocultural understandings of American national identity, on the other (Bonikowski & DiMaggio, 2016;R.M. Smith, 1997;Walzer, 1990). Civic nationalism embraces the liberal tradition of tolerance and universalism. By contrast, ethnic nationalism (or ethnonationalism) draws strong boundaries around what it means to be "truly" American by emphasizing the importance of characteristics such as ancestry, birthplace, language, and shared customs and traditions. These civic and ethnic conceptualizations of nationhood and American national identity have been in a state of flux throughout the history of the Republic, with the US often oscillating between the openness to newcomers inscribed on the Statue of Liberty, to patterns of nativist exclusion (Higham, 2002). Americans have also expressed support for both varieties of American nationalism. For instance, in a study of the subjective characteristics that defined national identity,  found that a sample of Californians expressed strong support for civic conceptualizations of American national identity but also exhibited high levels of support for linguistic criteria.
Nationhood cannot be understood without taking into account the ethnic compositions of the nation state.
A.D. Smith (2009), for instance, proposes an ideal-typical theory of nationhood, where members of the national ethnos share a number of important attributes, including a "myth" of common ancestry, and an association with a specific "homeland." In addition to these ethnic traits, there are a number of mechanisms that engender ethnic selfrenewal. One of these mechanisms is cultural borrowing, which may entail co-ethnics learning the language of the host country. Another of these is popular participation, which is exemplified in participation in the customs and traditions of the nation state. A.D. Smith's (2009) ethnosymbolism theory thus provides us with a specific set of criteria for understanding what it means to be "truly" American. Consistent with these criteria, those with salient levels of ethnonationalism believe that having "American" ancestry 2 and being born in the US are important markers of Americanness, while speaking English and participation in America's customs and traditions are important mechanisms for the renewal of American ethnos.
Scholars have found that individuals who subscribe to elements of the ethnonationalist belief system are more likely to oppose immigration. A substantial majority of these studies use data from the 1996 and 2004 waves of the University of Chicago/NORC General Social Study (GSS). These waves of the GSS contained a number of items that asked respondents whether a wide-ranging criterion were important for being "truly" American. For instance, Schildkraut's (2005) analysis of the 1996 GSS data found that individuals who thought that being born in the US was an important marker for being "truly" American were more likely to support restricting immigration. Moreover, this relationship held despite controlling for a number of socio-political and demographic variables that are known to be predictive of opposition to immigration, including Republican partisanship (Hajnal & Rivera, 2014). Similarly, using both the 1996 and 2004 waves of the GSS, Wong (2010) found that those who thought nativity was important for being "truly" American were also more likely to oppose birthright citizenship and oppose giving citizenship to legal permanent residents. 3 It is not surprising to find that those who place an emphasis on the importance of English are also likely to oppose immigration. As Schildkraut (2005) observes, if assimilation into the host society is a central norm that immigrants are often perceived as violating, language is perhaps the most visible signal of that norm.
Theiss-Morse (2009) finds that more than 90% of the American public agree that one must speak English if they are to be "truly" American. Consistent with these trends, some experimental studies have tested the effects of incidental exposure to Spanish among English-speaking White Americans. For instance, Newman (2013) found that exposure to Spanish increased perceptions of cultural threat among White participants, leading them to express negative attitudes towards immigrants. The influence of these beliefs is also reflected in a large literature that speaks to the success of the "English only" movement in shaping policy towards immigrants over the last few decades, with a number of states passing statutes or amending their state constitutions to mandate monolingualism (Citrin, Reingold, Walters, et al., 1990;Tatalovich, 2014). As noted by Piatt (1990), supporting the "English only" movement became a socially acceptable way of tapping into Whites' fears about being outnumbered by immigrants.
As these studies demonstrate, the prevailing trend in the existing scholarship has been to examine one component of the broader ethnonationalist belief system (for instance, nativity or language) and test its robustness as a predictor of opposition to immigration. On the one hand, this approach has been useful in helping us better understand which of these respective components may have the greatest amount of explanatory power when we aim to understand why so many Whites express a robust opposition to immigration. However, and consistent with ethnosymbolism theory (A.D. Smith, 2009), it is critically important to note that ancestry, nativity, language and participation in customs and traditions of the nation state are markers that are thought to function as part of a broader belief system pertaining to what it means to be a "true" member of the nation state. Therefore, while it is useful to be aware of the relative power and influence of each of these traits on Whites' attitudes towards immigrants, we nonetheless need to account for the possibility that, taken together, they are tapping into a set of broader popular attitudes towards US nationhood and America national identity.
There are a number of important steps to understanding the potential influence of ethnonationalism on antiimmigrant attitudes. First, those with salient levels of ethnonationalism use a distinct set of criteria (American ancestry, being born in the US, speaking English, and participation in American customs/traditions) to make sense of what it means to be "truly" American. Second, individuals with these beliefs, but who would otherwise not hold a distinct set of views towards immigrants, need a salient set of issues to emerge to fill that gap. Third, political elites frame issues such as immigration in a manner that appeals to these ethnonationalist beliefs (Zaller, 1992)-in the case of radical political actors such as Trump, this includes the perceived threat that immigrants or other ethnic outsiders pose to the robustness of American ethnos . In this way, ethnonationalist beliefs are activated by a specific frame regarding the "threat" of immigration promoted by elites, with the activation of these frames being dependent on the degree of resonance between the latent beliefs themselves and support for a popular message (Bonikowski, 2017).
In sum, my claim that ethnonationalism is a substantive predictor of White opposition to immigration requires three types of evidence. The first is that ethnonationalism should be a predictor of negative attitudes towards immigrants themselves (H1). Importantly, these hypothetical effects through ethnonationalism should be independent of the effects of other broad measures of White group attitudes that are likewise known to shape attitudes towards immigrants, especially ethnocentrism (Kinder & Kam, 2010), and White consciousness (Jardina, 2019). Perhaps unsurprisingly given its hypothetical associations with xenophobic attitudes, ethnonationalism was a robust predictor of White vote choice for Trump in the 2016 US Presidential election . This also suggests a potential connection between the tendency of the ethnonationalist belief system to draw rigid boundaries around what it means to be "true" members of the American ethnos and support for policies advocated by Trump that would exclude those (i.e., immigrants) who do not meet the criteria for being "truly" American. In substantive terms, this should equate to higher levels of support for policies such as deporting migrant children and building a wall along the US-Mexico border (H2).

| DATA AND METHODS
To test these theoretical expectations, data are taken from the 2016 ANES Time Series Study. The 2016 ANES is nationally representative, dual-mode survey, conducted between November 7, 2016, and January 8, 2017, that was designed to collect data on US voter behaviour and public opinion in time for the 2016 US Presidential election. The face-to-face component of the study used a stratified, clustered address-based sampling design to recruit N = 1,181 participants, while the online component used a similar address-based sampling design to recruit a total of N = 3,090 participants. Altogether, a total of N = 4,271 participants were recruited. Because the objective of the paper is to gauge whether ethnonationalism is a substantive correlate of the immigration attitudes of White Americans, I selected a subsample of respondents who identified as White and non-Hispanic only (N = 3,030).

| Dependent measures
To measure negative attitudes towards immigrants themselves, I construct an index of anti-immigrant attitudes using survey items from the 2016 ANES (Sides et al., 2019). The first item asked respondents whether they thought that immigrants were good for America's economy. The second item asked respondents whether they thought that immigrants increased crime rates in the US. And the third item asked whether they thought that immigrants generally harm America's culture. The second and third items were reverse coded such that higher values were indicative of negative attitudes towards the effects of immigration on US crime rates and culture. After recoding, the items were computed into a single index ranging between 0 and 1 (Cronbach's α = .79).
To measure support for restrictionist immigration policies, I use survey items about immigrant-targeted policies.
The 2016 ANES contains a number of items that ask respondents whether they favour or oppose certain policies for immigrants in the US. Specifically, I include items for whether a White respondent favours (1) building a wall along the US-Mexico border, (2) sending back children which were brought to the US illegally and (3) ending birthright citizenship. Items are coded such that higher values correspond to higher levels of opposition to birthright citizenship, higher levels of support for sending back immigrant children and building a wall with Mexico, and higher levels of support for ending birthright citizenship. When computed into an additive index, the three items exhibit a Cronbach's α = .66, suggesting that they are tapping into the same underlying attitude.
To measure affect for Trump, I use the standard feeling 101-point feeling thermometer. The thermometer ranges between 0 and 100, with a minimum score of 0 indicating "very cool or unfavourable" estimations of Trump, and a maximum score of 100 indicating "very warm or favourable" estimations.

| Ethnonationalism
The key independent measure is ethnonationalism. Ethnonationalism is an additive index made of four items that ask how important a number of traits are for being "truly American." The four traits were "to be have been born in the US," "to have American ancestry," "to speak English" and "to follow America's customs and traditions." Possible responses for each item ranged from 1 = very important to 4 = not important at all. 4 The four items were reverse coded so that higher values were indicative of a respondent placing greater emphasis on the importance on each trait for being "truly American." The items exhibited correlations with one another. 5 The rescaled index ranges between 0 and 1 and has a Cronbach's α = .77.

| Ethnocentrism
To measure ethnocentrism, I rely on a series of items concerning group stereotypes from the 2016 ANES (Kinder & Kam, 2010). 6 A single measure of ethnocentrism was computed out of the group stereotype items. The variable was constructed as Ethnocentrism = in-group favouritism 7 + out-group negativity. 8

| White consciousness
To measure White consciousness, I rely on the same three items used by Jardina (2019). The first item asked how important being White was to a respondent's identity. The second item asked how important it is for Whites to work together to change laws that are unfair to Whites. And the third item asked how likely it is that many Whites are unable to find jobs due to employers hiring minorities instead. Possible responses for the first two items ranged between 1 = extremely important and 5 = not at all important, while possible responses for the third item ranged from 1 = extremely likely to 5 = not at all likely. All three items were reverse coded and then computed into a single index (Cronbach's α = .58). 9 In order assess the degree of similarity between ethnonationalism, ethnocentrism and White consciousness, I performed a correlation analysis. The correlation analysis indicated moderate correlations between the three measures of White ingroup attitudes. However, the magnitude of these correlations was not sufficiently high enough to raise concerns related to multicollinearity. 10 Additional explanatory variables include a number of measures that are known to be associated with White attitudes towards immigration, negative economic evaluations (Hickel & Bredbenner, 2020;Miller, 2020), Christian nationalism (Baker et al., 2020) and authoritarianism (Brandt & Reyna, 2014;Hetherington & Suhay, 2011). 11 Negative personal economic evaluations are measured using a 5-point ordinal item that asks how worried a respondent is about their own financial situation, with possible responses ranging between 1 = extremely worried and 5 = not at all worried. 12 Negative national economic evaluations are measured using a 5-point ordinal item that asks whether the current national economy is good or bad, with possible responses ranging between 1 = very good and 5 = very bad.
The 2016 ANES does not contain the standard six-item battery for Christian nationalism (Whitehead et al., 2018). Therefore, I use a 7-point ordinal item that asks respondents how patriotic they think Christians are (possible responses ranging between 1 = patriotic and 7 = unpatriotic) 13 as a proximate measure of Christian nationalism.
Lastly, the measure of authoritarianism asks respondents to choose between four pairs of "desirable qualities" and to indicate which trait is more important for a child to have. The pairs were "independence or respect for elders," "curiosity or good manners," "obedience or self-reliance" and "being considerate or being well-behaved." In the computed measure of authoritarian attitudes, respondents received a score ranging between 0 and 4, such that respondent scored 4 if they chose all of authoritarian traits (respect for elders, good manners, self-reliance and being well-behaved) and a score of 0 if they chose all of the less authoritarian traits.

| Controls
Models are also estimated with a number of socio-political, demographic and structural covariates. Party ID is a 7-point ordinal item ranging between 1 = strong Democrat and 7 = strong Republican. Ideology is a 7-point ordinal item ranging between 1 = extremely liberal and 7 = extremely conservative. Age is measured continuously in years.
Females are coded as 1 and males as 0. Education is a 16-point ordinal item ranging between 1 = less than first grade and 16 = doctorate degree. Family income is a 28-point ordinal item ranging between 1 = under $5,000 and 28 = $250,000 or more. 14 Respondents who are unemployed are coded as 1, and those otherwise as 0. Religion is coded using five denomination categories, born again Protestant, Catholic, other Christian, non-Christian and the religiously unaffiliated, with non-born-again Protestants serving as the reference category. Religious service attendance is a 5-point ordinal item ranging between 1 = more than once a week and 5 = never. 15 Lastly, respondents residing in the South are coded as 1, and those residing elsewhere as 0.

| Ethnonationalism and anti-immigrant attitudes
The first step is to assess whether ethnonationalism is a substantive predictor of anti-immigrant attitudes. Figure 1 depicts the results of a linear model estimated using ordinary least squares (OLS) regression in which the anti-immigrant index is regressed against the ethnonationalism, the set of explanatory variables and the sociopolitical and demographic covariates. Points to the right of the reference line along the x axis in Figure 1 indicate a positive relationship between a variable of interest and the dependent measure, or more salient anti-immigrant attitudes, while points to the right of the x axis indicate a negative relationship, or less salient anti-immigrant attitudes.
As indicated by Figure 1, ethnonationalism is positively associated with anti-immigrant attitudes among Whites (p < .001). The magnitude of this relationship is also substantial, with the size of the standardized coefficient for ethnonationalism being larger than that of ethnocentrism and White consciousness. It is also noteworthy how ethnonationalism exhibits a larger effect size than other variables that are similarly known to be associated with antiimmigrant attitudes among Whites, including negative economic evaluations (Hickel & Bredbenner, 2020;Miller, 2020), authoritarian attitudes (Craig & Richeson, 2014) and Republican partisanship (Hajnal & Rivera, 2014).
The negative and significant coefficient for age (p < .01) is particularly surprising; however, scholarship consistent with the "competitive threat" model suggests that younger individuals face greater competition with immigrants in the labour market (Gorodzeisky & Semyonov, 2018), potentially accounting for the direction of the age coefficient.
To get a more substantive approximation of the magnitude of these effects, I use postestimation. Figure 2  Finally, the bottom panel in Figure 2 indicates that an individual with the lowest levels of White consciousness scores .35 on the normalized anti-immigrant scale. By contrast, that same individual scores .47 on the anti-immigrant F I G U R E 1 Ethnonationalism and anti-immigrant attitudes among whites. Points represent the size of each standardized OLS coefficient. The capped lines are 95% confidence intervals. Sample limited to Whites only. Data are weighted. **p < .01, *** p < .001. Full model estimates presented in Section A5 of the supporting information. Source: 2016 ANES scale at the highest levels of White consciousness. Therefore, moving from least to most conscious on the White consciousness scale is associated with just a .12 increase in anti-immigrant attitudes. In sum, the results of the postestimation point to ethnonationalism being a more substantive predictor of anti-immigrant attitudes than White consciousness among Whites, and its effect size being smaller than that of ethnocentrism. 16 These effects through ethnonationalism are also robust to a host of alternative model specifications. A contemporary debate in the literature concerns the effects of perceived job insecurity and Whites' negative attitudes towards immigration (Melcher, 2020). To account for these effects, I limit my sample to Whites in employment only, and include additional controls for perceived job insecurity and for whether a respondent reported having a cut to their working hours or pay. 17 In this model, the size of the standardized coefficient for ethnonationalism is actually larger than that in the OLS model in Figure 1 and remains significant at the p < .001 level. Another debate in the literature concerns the effects of class identity on negative attitudes towards immigrants, with working-class Whites feeling as though they have the most to lose from greater competition (Gest, 2016). To account for the effects of class ID, I control for a respondents' subjective class identification. 18 In this model, the size of the standardized coefficient for ethnonationalism is once again larger than that in the OLS model in Figure 1 and exhibits the same level of statistical significance (p < .001). Full estimates of the alternative models are presented in Section A6 of the supporting information file.

| Ethnonationalism and support for restrictionist immigration policies
In the next analysis, I estimate the relationship between ethnonationalism and support for restrictionist immigration policies among Whites. Figure 3 depicts the results of three ordered probit models in which the dependent measures F I G U R E 2 Anti-immigrant attitudes as a function of Ethnonationalism, ethnocentrism, and white consciousness. Points represent the predicted opinion on immigration at each level of the ethnonationalism, ethnocentrism and White consciousness scales. The capped lines are 95% confidence intervals. Predicted probabilities calculated by holding all other variables in OLS model constant or at their respective mean values. Source: 2016 ANES are again regressed against the ethnonationalism measure, each of the explanatory variables and the socio-political and demographic covariates. Points to the right of the reference line on the x axis indicate a positive relationship between a variable of interest and a greater preference for a given policy of immigration restriction, while points to the left of the reference line indicate a negative relationship, or a lower preference. Figure 3, the ethnonationalism measure is positively associated with White support for building a wall along the US-Mexico border (p < .001), for sending back individuals brought to the US illegally as children (p < .001) and for ending birthright citizenship (p < .001). Figure 3 also suggests that Kinder and Kam's (2010) construct is a stronger predictor of White support for restrictionist immigration policies than ethnonationalism.

As indicated by
However, ethnonationalism appears to be a stronger predictor of support for restrictionist immigration policies than Jardina's (2019) measure of White consciousness. As indicated here, the coefficients sit close to the reference line across models suggesting that the effects of White consciousness are not distinguishable from zero.
To ensure that ethnonationalism is primarily tapping into attitudes towards immigration and not some other latent trait such as a general, negative, disposition towards marginalized groups, it is also worth considering support for policies where ethnonationalism should not exhibit a substantive effect among Whites. One area of public opinion where we should not expect ethnonationalism to exhibit substantive effects is when it comes to understanding support for policies that would discriminate against LGBTQI+ individuals. Rather than being shaped by beliefs about American national identity, the literature instead indicates that support for anti-LGBTQI+ policies is likely to be driven by factors such as moral traditionalism, contact with LGBTQI+ individuals and religiosity (Brewer, 2003;Merino, 2013;van der Toorn et al., 2017).
To test this possibility, I estimate two models gauging Whites' support for allowing business owners to refuse service to same sex couples and support for mandating that transgender individuals use the bathroom of the gender that they were assigned at birth. Consistent with the literature on LGBTQI policy attitudes, these models also control for moral traditionalism and intergroup contact. 19 Across both models, the coefficient for ethnonationalism does not exhibit the conventional p < .05 level of statistical significance. The results of the alternative models thus lend weight to the assumption that ethnonationalism is tapping into salient attitudes towards immigrants, but not other marginalized groups such as LGBTQI+ individuals. Full estimates of these models are presented in Section A8 of the supporting information.
F I G U R E 3 Ethnonationalism and White support for Restrictionist immigration policies. Points represent the size of each the ordered probit coefficient. The capped lines are 95% confidence intervals. All variables z-transformed to obtain standardized coefficient estimates. Sample limited to Whites only. Data are weighted. Full estimates presented in Section A7 of the supporting information. Source: 2016 ANES 6 | DISCUSSION Is ethnonationalism a robust predictor of immigration attitudes that is independent of the effects of other measures of White group attitudes such as ethnocentrism (Kinder & Kam, 2010) and White consciousness (Jardina, 2019)?
And does ethnonationalism have significant implications for Whites' immigration policy preferences? The answer to both of these questions is yes. The findings demonstrate that ethnonationalism is a distinct and substantive correlate of White opposition to immigration. Ethnonationalism is a statistically measurable construct, and its influence is not diminished by a number of measures that are already known to be substantively associated with anti-immigrant attitudes.
These findings contribute to our existing understanding of White opposition to immigration in a number of important ways. First, much of the previous work on White opposition to immigration focuses on the effects of factors such as ethnocentrism (Kinder & Kam, 2010), White consciousness (Jardina, 2019), negative economic evaluations (Hickel & Bredbenner, 2020;Miller, 2020) and authoritarian attitudes (Craig & Richeson, 2014) but ignores the influence of individuals' concerns with the robustness of America's national identity altogether.
Moreover, the few studies that have explored whether ethnonationalist beliefs are predictive of anti-immigrant attitudes have only explored the effects of one component at a time (Schildkraut, 2005;Wong, 2010). However, and consistent with A.D. Smith's (2009) theory of ethnosymbolism, I have argued that it is critically important that we see factors such as ancestry, nativity, language and participation in the customs and traditions of the nation state as a set of markers that together form part of a wider belief system that pertain to what it means to a be "true" American. In line with A.D. Smith's (2009) theorizing, I have shown that these ethnonationalist beliefs all exhibit correlations with one another and load onto the same factor, indicating that they are tapping into the same latent set of beliefs. Furthermore, as evidenced by the size of the standardized coefficient for ethnonationalism in Figure 1, this set of beliefs appears to be the strongest predictor of anti-immigrant attitudes for Whites. Importantly, ethnonationalism was a stronger predictor of anti-immigrant attitudes than a number of variables that have long been known to shape Whites' opposition to immigration, including Republican partisanship (Hajnal & Rivera, 2014).
Second, my finding that ethnonationalism is associated with higher levels of support for Trump-era policies that would restrict immigration builds upon our existing understanding of why so many Whites increasingly favour a change in direction to post-1965 national policymaking on immigration. To understand this contemporary opposition, it is useful to point out that current debates regarding immigration policy are reflective of the nativist sentiment that has usually followed after large waves of immigration to the US (Higham, 2002). In the early 20th Century, for instance, those who viewed immigrants as a threat to the robustness of America's national identity cloaked their support for policies that would restrict immigration in the rhetoric of "preserving" national homogeneity (Grant, 1916;Ross, 1914). Similarly, those who have been the most fervent supporters of policies such as deporting migrant children and building a wall along the US-Mexico border today fear a further decline in national homogeneity which may only be halted if current demographic trends are reversed by enacting policies that would restrict immigration.
To those with salient ethnonationalist beliefs, then, support for these policies functions as a means of reversing America's post-1965 tradition of being an ethnoculturally plural nation.
In sum, the findings presented here offer an additional explanation for why so many Whites continue to exhibit robust levels of opposition to immigration. While concerns related to the robustness of America's national identity have not played a focal role in the contemporary literature on White immigration opinion, a centrepiece of Trump's campaign and subsequent Presidency has been to denigrate those that he does not see as "true" Americans, whether it be by calling Ted Cruz an "anchor baby" during the 2016 Republican primaries (Diamond, 2016) or Tweeting at four Democratic Congresswomen to "go back to their own countries" (Yglesias, 2019). Beliefs regarding whether criteria such as being born in the US make one "truly American" appear to be similarly distributed among the broader White adult populace and have been shown in this paper to be strongly predictive of White opposition to immigration across a variety of measures.

| Limitations and future directions
It is also important to highlight the limitations of the current study, as well as to point to directions for future research into the effects of ethnonationalist beliefs on anti-immigrant attitudes. One limitation of the current study is that I have only examined how ethnonationalism affects the immigration attitudes of non-Hispanic Whites. While there is evidence to suggest that ethnonationalist beliefs are not uniformly distributed among ethnoracial groups (Schildkraut, 2007), there is nonetheless reason to suspect that some groups may exhibit ethnonationalist beliefs of a similar salience to non-Hispanic Whites. A potentially interesting group to examine is Hispanics who identify as White. An emerging literature highlights that Hispanics who express an attachment to White identity are more likely to exhibit higher rates of political conservatism relative to Hispanics who do not identify as White (Filindra & Kolbe, 2020). Given the robust links between political conservatism and immigrant-based xenophobia, these ideological differences may also lead White Hispanics to exhibit more conservative policy opinions across a range of issues, including greater support for immigration restriction. Future research should test this possibility further by comparing the relative differences between non-Hispanic White and Hispanic White immigration preferences, as well as any intra-Hispanic differences. 5 Bivariate correlations for the ethnonationalism items are presented in Section A2 of the supporting information. 6 In these items, White respondents were presented with a seven-point scale on which they had to rate the characteristics of a given ethnoracial group. The scales were based on a series of paired antonyms. The first antonym was hardworking versus lazy. A score of 1 indicated that respondents thought all the people in a given ethnoracial group were hardworking. A score of 4 indicated that most people in the group were not particularly close to one end or the other. And a score of 7 indicated that most people in the group were lazy. White respondents were first asked to rate themselves on this scale. Afterward, Whites were presented with the same scale again, but were instead asked to rate African Americans, Hispanics and Asian Americans. This process was then repeated for the second antonym-peaceful versus violent. 7 In-group favouritism was calculated by summing the scores for the two ANES items in which White respondents had to rate their own group on the hardworking/lazy scale and peaceful/violent scale. The two items have a Cronbach's α = .76. This summed score was then divided by two to create an average score. Responses for the items in which Whites had to rate their own group were reverse coded so that a higher score responded to a higher rate of perceived in-group virtuousness. The formula for calculating the in-group favouritism score is In-group favouritism = (trait 1 in-group score + trait 2 in-group score)/2. scale. Whites' ratings of non-White ethnoracial groups were averaged to create a single score for all out-groups across both scales. The two average scores were then summed and divided by 2 to create an average score that represented out-group negativity. Thus, the formula for calculating the out-group negativity score is Out-group negativity = (trait 1 average outgroup score + trait 2 average out-group score)/2. 9 This Cronbach's α score was just below the accepted .70 threshold for internal consistency. 10 Bivariate correlations for the three measures of White ingroup attitudes are presented in section A3 of the Supplemental Information file. 11 While Hispanic and undocumented immigrant affect are measures that are both known to be associated with higher levels of opposition to immigration among Whites (Ramirez & Peterson, 2020), I do not account for these factors in my estimation strategy because both measures are likely to function as posttreatment variables (Montgomery et al., 2018). That is, how people feel towards Hispanics and undocumented migrants is likely to exist as a function of ethnonationalist attitudes. As such, to avoid introducing significant bias into my coefficient for the ethnonationalism term, models are estimated without controlling for Whites' feelings towards Hispanics and undocumented immigrants. To account for the possibility of these items introduce posttreatment bias, I performed ancillary analysis where models were estimated with additional controls for Hispanic and undocumented immigrant affect. Ancillary models are presented in Section A4 of the supporting information. 12 The item is reverse coded such that higher values are indicative of negative personal economic evaluations. 13 The item is reverse coded such that higher values are indicative of Whites thinking that Christians are patriotic.
14 Respondents who refused to provide their incomes were omitted from the analysis. 15 The item is reversed coded such that higher values are indicative of frequent religious service attendance. 16 The discrepancy between the standardized beta weights for ethnonationalism and ethnocentrism in Figure 1 versus the postestimation in Figure 2 is partly explained by the fact that the ethnocentrism measure has a larger range than the ethnonationalism measure. The range for the standardized ethnonationalism measure ranges between À2.469 and 1.578, while the range for the standardized ethnocentrism measure ranges between À6.064 and 4.688. The ethnocentrism scale is composed of nine, 7-point ordinal items, leading to 63 possible values. By contrast, the ethnonationalism measure is composed of just four, four-point ordinal items, leading to just 16 possible values. The greater number of values for the ethnocentrism measure leads to greater measure variation and, by implication, the possibility for greater movement as one moves from least to most ethnocentric along the ethnocentrism scale. 17 Perceived job insecurity is a 5-point ordinal item that asks employed respondents how worried they are about losing their job in the future, with possible responses ranging between 1 = not at all and 5 = extremely. The latter variable is a dichotomous variable where a value of 1 indicates that an employed respondent has had their working hours or pay cut by their employer, and 0 for otherwise. 18 Class ID is a constructed dichotomous variable where a value of 1 indicates that a respondent identifies as working-class, and 0 for otherwise. 19 Moral traditionalism is an additive index of four items that asked whether respondents agreed or disagreed with a number of statements concerning societal adherence to variety of "traditional" norms. These statements were (1) the world is changing and we should adjust, (2) newer lifestyles are breaking down society, (3) we should be more tolerant of other moral standards, and (4) there should be more emphasis on traditional family values. The second and fourth items were reverse coded such that higher values were indicative of respondents agreeing that newer lifestyles break down society and that there should be more emphasis on traditional family values. After recoding, the items were computed into a single index ranging between 0 and 1 (Cronbach's a = .73). The latter variable is a dichotomous variable where a value of 1 indicates that a respondent knows a friend or family member who is LGBTQI+, and 0 for otherwise.