Anti-Muslim policy preferences and boundaries of American identity across partisanship

Abstract

Does American identity predict preferences for anti-democratic policies that aim to marginalize Muslim Americans? Absent significant priming of inclusive elements of American identity, we argue that individuals with stronger attachments to American identity will be less likely than their counterparts to reject a range of anti-Muslim policies that are antithetical to principles of religious liberty and equality. Across three surveys and multiple measures, American identity powerfully predicts preferences for curbing the civil liberties of Muslim citizens. Particularly striking is the finding that the effect of American identity spans the partisan divide; it consistently explains the endorsement of exclusionary policies among self-identified Democrats, who typically hold more progressive policy positions toward minority groups than Republicans. Overall, our study highlights the contradictory and exclusionary nature of American identity, which has important implications for minority groups constructed as outside the boundaries of Americanness.

Scholars increasingly recognize the independent and important role that attitudes toward Muslims play in American politics. This emerging body of research finds that negative affect toward Muslims is held by a substantial portion of the electorate (Gerteis et al., 2020; Lajevardi, 2020; Mogahed et al., 2018; Panagopoulos, 2006), influences candidate evaluations (Calfano et al., 2020; Kalkan et al., 2018) and the prospect of substantive representation (Lajevardi, 2018), predicts vote outcomes and policy evaluations (Jardina and Stephens-Dougan, 2021; Lajevardi and Abrajano, 2019; Tesler, 2021), and shapes partisan preferences (Tesler, 2021).

While much of this work has been informative, it has largely overlooked the unique influence that American identity may exert in explaining citizens’ endorsement of anti-democratic policies targeted at Muslims. This is an important omission given that electoral politics across Western democracies have seen national identity reconfigure the lines of what and for whom parties and politicians stand (Hobolt and De Vries, 2016; Moffitt, 2020; Schildkraut, 2011; Uslaner, 2017; Wong, 2017). The primacy of national identity is now arguably the dividing line in national politics, particularly as it pertains to deciding whose rights should be protected or curtailed (Uslaner, 2017). For example, national identity was on display during Trump’s 2016 election campaign and presidency (Bedolla, 2021; Sides et al., 2017), in shaping support for Brexit (Henderson et al., 2017), and in the rise of the Sweden Democrats (Elgenius and Rydgren, 2019) and the AfD in Germany (Rosellini, 2020).

In contrast to nativist elements, it has long been argued that national identity in the U.S. is fundamentally different from its expression in other countries in that American identity rests on a set of political principles or shared democratic beliefs (Citrin et al., 1994). While American ethos is often described in such terms, research has found strong evidence that American identity is thoroughly infused with “whiteness,” with racial and ethnic minority groups often conceived of as second-rank citizens (Devos and Banaji, 2005). The evidence that national identity is not readily ascribed to low-status group is supported by social dominance theory (Sidanius et al., 1997), leading to the conclusion that “… behind the deliberate endorsement of inclusive principles lurks a more exclusionary national identity” (Devos and Mohamed (2014, p. 2), see also Katzenstein et al. (2010)).

Nonetheless, research is lacking on the relationship between American identity and policies that implicate Muslims in specific. This oversight is noticeable given that American identity has been uniquely leveraged against US Muslims—due to various historical, geopolitical, and socio-cultural factors—in ways that are arguably distinct from other minoritized groups. Filling this scholarly gap and building on the extant literature, we therefore consider the case study of Muslim Americans. As a group whose racial, ethnic, religious, and immigrant backgrounds are viewed with suspicion and often perceived to pose both a realistic and symbolic threat to the “American” way of life (Argyle et al., 2018; Kalkan et al., 2009; Lajevardi et al., 2023; Oskooii et al., 2019), we expect American identity to predict preferences for anti-Muslim policies even across partisan lines. We argue that beyond the psychological difficulty of viewing Muslim citizens as Americans, strong American identifiers will go a step further and show a proclivity to endorse or not reject policies that clearly undermine beliefs about religious freedom and egalitarianism.

To test this claim, we draw on three original public opinion surveys organized into two studies. Our analyses show that individuals with a stronger attachment to American identity are not only more likely to hold negative attitudes toward Muslim Americans but also significantly more likely than their counterparts to endorse or not reject policies that aim to curb or even eliminate their civil liberties, namely restricting their First and Second Amendment rights. These findings not only highlight the contradictory nature of American identity (Schildkraut, 2014) but also illustrate that some identifiers are willing to sacrifice their espoused principles to keep certain groups clearly outside the boundaries of “American-ness.” Perhaps more importantly, the analyses reveal that the exclusionary nature of American identity functions similarly among partisans of all stripes, indicating that even high American-identifying Democrats cannot be counted on to express opposition toward anti-Muslim policies that are antithetical to values of egalitarianism and religious liberty.

This article makes at least two key contributions. First, it shows that American identity is an important construct that should be accounted for when examining opposition to Muslim Americans. Second, it illustrates that the exclusionary and contradictory nature of American identity is much more pronounced and overt than previously assumed; even in the case of Muslim citizens of the United States, American identifiers may forgo their cherished principles and end up endorsing a range of anti-democratic policies.

1. Theory and expectations

National identity is often invoked in American politics, typically to stir the tide against permissive immigration policies (Huddy and Khatib, 2007). American identity has long been associated with the country’s Anglo-Saxon, Protestant, and Northern European populations (Smith, 1988), contrary to the notion that America is a “nation of immigrants” compromised of people from all backgrounds (Greene et al., 2020). Even citizenship held by racialized others, as Masuoka and Junn (2013) have observed, is an insufficient condition for full membership in the American polity and claims to the rights and privileges that come along with it.

As such, while some believe that “American-ness” comes from performative acts, like pledging allegiance to the flag (Theiss-Morse, 2009), it has been so thoroughly infused with “whiteness,” that, for many, being American and white are nearly indistinguishable from one another (Devos and Banaji, 2005). Thus, there has long been a concern about whether people of color can become American or whether they represent a threat to the so-called American way of life (Smith, 1988). Concerns over perceived outsiders are particularly visible among those who most strongly report a subjective or internalized sense of belonging or attachment to the nation. Studies have shown that these individuals tend to oppose bilingual education policies (Huddy and Sears, 1995) and favor proposals, such as making English the official language and printing election ballots only in English (Schildkraut, 2003).

Building on this work, we conceive of American identity as a social identity, which demarcates in-group and out-group members of the nation based on racial, cultural, and religious cleavages. We contend that American identity is an important construct that researchers should consider when examining public opinion toward Muslim Americans. Arguably no group has been more vilified as anti-American as have Muslims. In recent times, several unprecedented policies that clearly challenge American ethos have been espoused by politicians, some of which have been introduced into state legislatures. Examples range from the infamous Muslim travel ban and “extreme vetting” practices of certain nationals such as immigration form DS-5535 to mosque restrictions and the indiscriminate surveillance of Muslim places of worship and neighborhoods.

1.1. American identity and attitudes towards Muslims

There are several reasons why Muslim Americans are considered outsiders in their own home. To begin, most Muslims are not perceived to be racially “white” (d’Urso, 2022; Maghbouleh, 2020), ¹ the majority are foreign-born, and they do not share the same religious traditions as the majority population. It also goes without saying that American national identity was reinvigorated following the tragic 9/11 attacks (Hutcheson et al., 2004; Li and Brewer, 2004), which not only unified the public, but also launched the country into a global War on Terror, where Islam and its adherents were arguably and fairly indiscriminately constructed to be the enemy of America.

Following the attacks on 9/11, Islam was associated with global terrorism (Powell, 2018), and American identity was regularly invoked as a shield against perceived external threat characterized as Islamic terrorism (Cainkar, 2009; Ewing, 2008). As such, a number of policies have been proposed, justified, and enacted through the lens of serving as protective measures safeguarding American values and lives against Islamic terrorism (Patel and Koushik, 2017). Therefore, Islam is perceived by some to not only pose a symbolic and cultural threat but also an existential one to the American way of life (Lajevardi et al., 2023; Oskooii et al., 2019; Saleem et al., 2017). What’s more, the involvement of the U.S. in conflicts over the past two decades in predominantly Muslim-majority countries has not only made Islam inherently more salient in American political discourse, it has also linked Islam and Muslims with adversaries in these conflicts (Beydoun, 2023). Finally, Islam is often portrayed by elites and the media as being incompatible and at odds with Western or American values (Lajevardi, 2020; Nacos and Torres-Reyna, 2007; Sides and Gross, 2013; Tesler, 2018), resulting in hostility, hate, and resentment towards Muslims in the US (Hobbs et al., 2023; Lajevardi, 2021). By being positioned as the bastion of democratic freedoms and being perceived as at odds with Islam, American identity is often seen as contrary to and arguably superior to Muslim practices and beliefs.

Given that Muslim Americans are viewed and constructed as outsiders on multiple fronts, we expect those with stronger attachments to American identity to not only evaluate Muslim Americans as a group unfavorably, such as endorsing orientalist notions of this population, but also show greater support for anti-Muslim policies that are antithetical to values of egalitarianism and religious liberty. Stated differently, Muslim Americans present a unique case study to evaluate the extent to which American identifiers are committed to upholding the principles they claim to endorse.

1.2. American identity and partisanship

Finally, we also consider whether American identity operates similarly across partisan lines. It remains an open question as to whether American identity exerts a uniform influence across different political partisans as it pertains to their preferences for anti-Muslim policies.

On the one hand, research has shown that American identity can lessen the divide among cross-partisans (Levendusky, 2018), so it is possible that American identity can shape policy attitudes similarly for Democrats and Republicans. This may be due to American identity not being imbued with political ideology (Huddy and Khatib, 2007), unlike patriotism. As such, American identity may predict similar anti-Muslim policy positions across the partisan aisle, a hypothesis that we believe is also important to test.

It is also possible that American identity may be more influential in weighing on political preferences depending on how each party has used American identity in its rhetoric. For instance, those parties that intertwine national identity with anti-Muslim rhetoric may activate this link between American national identity and anti-Muslim policies for their partisans. At the elite level, Democrats and Republicans have somewhat diverged in their stance toward Muslims. For example, research suggests that Democratic legislators are more responsive to Muslim constituents than are Republicans (Lajevardi, 2018; Martin, 2009), left-leaning cable news media offers more positive portrayals of Muslims relative to right-leaning news (Bleich and van der Veen, 2021; Lajevardi, 2021), ² and Democratic legislators generally express more positive views toward Muslims than their Republican counterparts (Lajevardi and Spangler, 2022).

Nonetheless, there are several reasons why American identity might be activated – or not – for citizens of different partisan stripes. Turning first to Democrats, those with stronger American identity might be susceptible to endorsing anti-Muslim policies, because they, like Republicans, are subject to the same national security narratives that often cast suspicion on Muslim communities. Such Democrats may then prioritize perceived security over civil liberties, despite their party affiliation. Moreover, given that national identity is often conflated with a certain racial and cultural background, Democrats who more strongly identify with American identity than their counterparts may support policies that align with a homogenized vision of national identity. At the same time, there are also good reasons to expect differences in the relationship between American identity and anti-Muslim policy endorsement between Democrats and Republicans. Among Democrats, many prioritize egalitarian and liberal values (e.g., civil rights), which could counteract inclinations toward supporting exclusionary policies. Moreover, Democrats are often a diverse group of people along the lines of race, ethnicity, and religion, and so they may be more likely to engage in interfaith solidarity, embracing Muslim Americans as part of a broader religious or cultural coalition against discrimination. Finally, if their American identity is rooted in civic nationalism, and especially in the country’s democratic ideals and institutions, Democrats may be predisposed to oppose policies that they view as undemocratic.

Republican American identifiers, meanwhile, may especially exhibit both susceptibility to and resistance against anti-Muslim policies due to various factors within their party’s stances and the broader political landscape. Republican politicians often emphasize strong national security policies. Given the association of Muslims with terrorism in certain media and political discourses, Republican citizens may be influenced to support restrictive policies as a means of safeguarding the nation, particularly those with stronger American identity attachments. Cultural conservatism may also play a role in shaping support for such policies. For example, those Republicans who hold traditional and more conservative views on the boundaries of American culture and identity may perceive Muslims as a threat to these values, leading to support for policies that limit the influence of what they see as non-American cultures. However, there is reason to believe, given the libertarian streak in the party, which prioritizes individual freedoms, that Republicans with stronger than weaker attachments to American identity would oppose policies that infringe on civil liberties.

Given these factors, we are agnostic as to the direction and magnitude of the influence American identity may have on support or rejection of anti-Muslim policies across the partisan divide but deem such inquiry necessary to better understand the relationship between American identity and anti-Muslim policy preferences across the partisan aisle.

2. Data and measures

To examine the relationship between American identity and anti-Muslim policy attitudes, we conducted two studies. Study 1 relies on a nationally representative survey fielded in 2018 by the Cooperative Congressional Election Study (CCES) through YouGov. Our module was included in the pre-election wave (September 27–November 5) of the questionnaire and includes 721 observations. ³

Study 2 utilizes two original, cross-sectional surveys that were hosted on Qualtrics and conducted online in English through opt-in panels of adult U.S. respondents via Lucid survey sampling firm. ⁴ Relative to the CCES data set, these surveys provide us with additional outcome measures and a much larger sample. The first was fielded between June 17–July 7, 2019 (N = 3,733) and employed quotas for gender, party identification, age, region, and race. The second was fielded between August 17–August 20, 2020 (N = 1,876) and includes over-samples of Black, Latinx, and Asian American respondents. ⁵ While the Lucid datasets are not nationally representative of the overall U.S. population, research finds that its demographic makeup corresponds reasonably well to high-quality datasets when weighted to population benchmarks (Tausanovitch et al., 2019). To this end, we created survey weights for both datasets using the 2019 American Community Survey (ACS) 5-year estimates for sex, age, race, and education.

2.1. Outcome variables

Both studies use three identical questions related to the most salient, group-centric anxieties over American Muslims – commonly referred to as Muslim American Resentment (MAR) – that have been extensively validated and utilized (Lajevardi, 2020; Lajevardi and Oskooii, 2018). The first question taps into classic assimilation concerns directed at American Muslims and immigrant-based populations more generally, by asking respondents to indicate how strongly they agree or disagree with the contention that “Most Muslim Americans lack basic English language skills.” The second and third questions are proxies for long-held orientalist tropes of Muslims, by asking respondents to indicate agreement or disagreement with the following statements: “Muslim Americans sometimes do not have the best interests of Americans at heart” and “Muslim Americans, in general, tend to be more violent than other people.” Consistent with previous work, we combined these items and created an additive scale and then normalized (min-max) the variable to range from value 0 to 1. Cronbach’s scale reliability coefficients for each survey are as follows: 2018 CCES $\alpha $ = 0.829; 2019 Lucid $\alpha $ = 0.790; 2020 Lucid $\alpha $ = 0.819.

We rely on the MAR scale to determine whether American identity predicts negative affect toward Muslim Americans. This assessment is important because if attitudes toward Muslim Americans are unrelated to individuals’ attachments to American identity it would undermine the argument we have outlined with respect to the relationship between American identity and anti-Muslim policy preferences.

Next, each study includes questions about a variety of policy proposals that aimed to, or in some cases, have already curbed the civil liberties of Muslim citizens. All measures were normalized to range from 0 to 1, with the highest value indicating support for the restrictive policy in question. We provide summary statistics in the supplementary appendix in Tables A1–A3.

In total, we inquired about six different policies, two of which were included across both studies. The first of the two questions probed attitudes about the infamous “Muslim Travel Ban,” which garnered a lot of attention during the 2016 presidential election season and was immediately enacted through an Executive Order in January 2017 by President Trump. To gauge levels of endorsement for this travel ban, respondents were asked whether they agree or disagree with a ban that would “limit Muslim Americans from re-entering the US if they have left for any reason (i.e., vacation, work, longer visits).” The second question asks respondents to indicate their support for the increasingly popular anti-Sharia bills, which have been considered and enacted by many state legislatures.

We also asked several questions that were not included in both studies. In Study 1, we asked respondents about then-candidate Trump’s proposal to establish a Muslim registry. The 2019 Lucid survey in Study 2 also included a question about Republican presidential contender Ted Cruz’s proposal to “empower law enforcement to patrol and secure Muslim neighborhoods before they become radicalized.” In the 2020 Lucid survey (Study 2) we added a weapon ban question that gained traction after Omar Mateen, responsible for the Orlando nightclub shooting, lawfully purchased the guns used in the attack. Specifically, new discussions emerged on whether Muslim citizens should be subject to more checks and scrutiny or even prevented from buying weapons. Finally, we asked respondents in both Study 2 surveys to indicate the extent to which they would be willing to pass laws to restrict the number of mosques or Islamic centers from being built in the U.S. The desire to impose such religious restrictions dates back to the 2010 Park51 Islamic Center controversy in lower Manhattan, New York (Oskooii et al., 2019), but has since expanded to many other communities across the country.

2.2. Explanatory and control variables

Our key explanatory variable is American identity, which is a subjective or internalized sense of belonging or attachment to the nation – the sense of being or feeling American (Citrin et al., 2001). Unlike patriotism, American national identity is not ideologically divisive; it encompasses all those who feel close to their country, which includes both Democrats and Republicans (Citrin et al., 2001; Huddy and Khatib, 2007). Across both studies, American identity is an additive index (normalized to range from 0-1) composed of four commonly used items that do not conflate American identity with ideologically divisive concepts such as symbolic or uncritical patriotism (Huddy and Khatib, 2007; Oskooii et al., 2019).

More specifically, respondents were asked to indicate how strongly they agreed or disagreed with the following statements: (1) My American identity is an important part of my “self,” (2) Being an American is an important part of how I see myself, (3) I am proud to be an American, and (4) Sometimes, I dislike being an American. This scale ranges from value 0 to 1, with the highest value indicating stronger American identity. ⁶ We note here that respondents in the 2020 Lucid sample were not presented with the fourth statement, and, therefore, the American identity scale was constructed with only the first three items. However, comparing weighted mean values ( $\mu $ : 0.738, 0.730, and 0.759) and Cronbach’s scale reliability coefficients ( $\alpha $ : 0.865, 0.811, and 0.886) across each survey suggests a fair amount of consistency and internal validity. In the supplementary appendix Figures A1–A3 we also provide the variable’s distribution by partisanship, which shows that partisans of all stripes can exhibit a strong sense of being or feeling American, although a somewhat higher proportion of Republicans report stronger attachment to the nation than Democrats or Independents. ⁷

Finally, our analyses account for standard demographic controls (education, income, sex, age, and race), political interest, political ideology (liberal-conservative), and dummy variables for party identification. We also control for Trump's approval, given that a greater proportion of his supporters may express a strong attachment to American identity and also harbor negative feelings towards out-groups, particularly Muslim Americans.

3. Results: Study 1

We begin by examining the relationship between American identity and negative evaluations of American Muslims before examining policy preferences. To this end, we estimated ordinary least squares (OLS) models where the MAR scale was regressed on American identity. ⁸ As Table A4 shows, American identity is statistically associated with MAR in both bivariate and multivariate regression models. This suggests that higher American identifiers are more likely than their lower-identifying counterparts to hold more negative attitudes toward Muslim Americans, even after accounting for confounding factors like political ideology and partisanship.

Having established a connection between American identity and anti-Muslim affect, Table A5 reports six OLS models, where responses to three policy questions were regressed on American identity. In both bivariate and multivariate models, American identity is strongly associated with anti-Muslim policy proposals (at $p \lt $ 0.001). For ease of interpretation and to determine the practical importance of the findings, Figure 1 presents model coefficients with 90% confidence bands. The coefficient values signify how much the mean of each policy measure changes given a single unit shift in American identity while holding all other covariates constant. As shown, American identity exerts a substantively strong influence on policy preferences. For example, a single unit increase in American identity corresponds to an average increase of 0.18 points in support of the Muslim ban, which ranges from value 0 to 1. We found similar effect sizes for a Muslim registry proposal (0.16) and support for anti-Sharia legislation (0.15). Outside of American identity, Trump's approval is the only other consistent and strong predictor of endorsement of exclusionary policies. For instance, a single unit increase in Trump approval corresponds to an increase of 0.35 support for the Muslim ban, which the President instituted. In fact, Trump's approval plays such a significant role in that it partially or fully soaks up the effect of political ideology and partisanship; in separate models without Trump's approval, both ideology and partisanship predict policy attitudes.

Figure 1. The relationship between American identity and anti-Muslim policy attitudes (Study 1: 2018 CCES).

Note: OLS Coefficients with 90% CIs are derived from regression results reported in Table A5.

We also examined heterogeneous effects of partisanship by estimating each of the aforementioned policy models by subsets of self-identified Democrats and Republicans (see Table A6). This approach is particularly useful in determining if American identity bifurcates the opinions of Democrats, who, on average, are less likely than Republicans to endorse anti-Muslim policies. Figure 2 suggests this to be the case. Across all outcome measures, higher American-identifying Democrats are statistically more likely than their lower-identifying counterparts to express relatively more support for anti-Muslim policies. A single unit increase in American identity among Democrats increases support for the Muslim ban and anti-Sharia laws by an average of 0.17 to 0.19 points, while it increases support for a Muslim registry by 0.37 points. Perhaps due to ceiling effects, we did not find that American identity is associated with policy attitudes among Republicans. However, the findings of the two surveys in the second study lead to another conclusion.

Figure 2. The relationship between American identity and anti-Muslim policy attitudes by partisans (Study 1: 2018 CCES).

Note: OLS Coefficients with 90% CIs are derived from regression results reported in Table A6.

4. Results: Study 2

Study 2 allows us to replicate the analyses across two original surveys conducted in 2019 and 2020. ⁹ We begin by analyzing the relationship between American identity and MAR. The results in Tables A7 and A8 illustrate that American identity is positively associated with MAR in both bivariate and multivariate models (at $p \lt $ 0.001). Consistent with Study 1, higher American identifiers are more likely than their lower identifying counterparts to believe that most Muslim Americans are not assimilated, do not hold the best interest of Americans at heart, and that they tend to be more violent than other people. More specifically, a single unit increase in American identity increases MAR by an average of .19 points in 2019 and .26 points in 2020 while keeping other covariates constant.

Clearly, higher American identifiers are more likely than lower American identifiers to believe in some of the most common orientalist notions of Muslims as previously documented (Lajevardi, 2020; Oskooii et al., 2019). By not viewing Muslim Americans as “prototypical Americans” and regarding them as an existential threat to the American way of life, we expect individuals with stronger American identity attachments to show more willing less to endorse anti-Muslim policies as well, even if such endorsements are antithetical to what it means to be an American. In Tables A9 and A10 we report OLS regression results where responses to five policy questions are regressed on American identity. Across the board, American identity is positively associated with every single policy outcome measure (at $p \lt $ 0.001). Figures 3 and 4 show that a single unit increase in American identity corresponds to an increase of between .13 to .24 points in support of each outcome measure ranging from 0 to 1.

Figure 3. The relationship between American identity and anti-Muslim policy attitudes (Study 2: 2019 Lucid).

Note: OLS Coefficients with 90% CIs are derived from regression results reported in Table A9.

Figure 4. The relationship between American identity and anti-Muslim policy attitudes (Study 2: 2020 Lucid).

Note: OLS Coefficients with 90% CIs are derived from regression results reported in Table A10.

The 2020 Lucid analyses reveal similar results. All else equal, a single unit increase in American identity increases support for the Muslim ban by 0.23 points, anti-sharia law and weapon ban by about 0.25 points, and opposition to the construction of mosques by 0.27 points. Once again, these relationships are independent of the influence that partisanship and political ideology exert on the endorsement of anti-Muslim policies. Across both surveys, Trump approval is associated with support for each policy position but does not drown out the effect of American identity, as it does for partisanship.

The last set of models reported in Tables A11 and A12 address the question of whether the effect of American identity on policy attitudes persists among Democrats and Republicans, or are confined to one party. As Figures 5 and 6 show, both higher American-identifying Democrats and Republicans are much less likely than their lower-identifying counterparts to reject policies that are at odds with cherished American values of religious liberty and equality. In sum, the results suggest that the effect of American identity on anti-Muslim policy attitudes function similarly across party lines.

Figure 5. The relationship between American identity and anti-Muslim policy attitudes by partisans (Study 2: 2019 Lucid).

Note: OLS Coefficients with 90% CIs are derived from regression results reported in Table A11.

Figure 6. The relationship between American identity and anti-Muslim policy attitudes by partisans (Study 2: 2020 Lucid).

Note: OLS Coefficients with 90% CIs are derived from regression results reported in Table A12.

5. Additional analyses

In this section, we report a set of additional analyses across the two studies (and three data sets) to further investigate the relationship between American Identity and policy preferences.

We begin by considering how strong a particular confounder (or group of confounders) would have to be to change our conclusions. In considering how vulnerable our results may be to omitted variable bias, we conducted a series of sensitivity analyses using the approach and R package (“sensemakr”) developed by Cinelli and Hazlett (2020). As our benchmark covariate, we used Trump’s approval to bind the possible strength of unobserved confounders given that attitudes toward Trump were among the strongest predictors of anti-Muslim policy preferences. ¹⁰ The results of the aforementioned sensitivity analysis for every single outcome variable across each of the three data sets is reported in Appendix Tables B1–B11. Each table displays the robustness value (RV), which summarizes the types of confounders that would problematically change the research conclusions. For instance, Table B1 shows that the RV of American Identity coefficient is 14.4%. This means that unobserved confounders explaining at least 14.4% of the residual variance of both the treatment and the outcome would explain away the estimated treatment effect. Stated differently, any confounder explaining less than 14.4% of the residual variance of both the treatment and the outcome would not be strong enough to bring down the estimated effect to zero. The robustness value accounting for statistical significance ( $\alpha $ = 0.05) is also shown in the table, which suggests that confounders would need to be about half as strong (7.8%) to make the estimate not “statistically significant.” Finally, the lower right corner of the Table shows the strength of association that a confounder as strong as Trump Approval would have to be (i.e., 3.3%–10.9%). Since the robustness value of 14.4% is higher than either quantity, the results suggest that such a confounder could not fully eliminate the point estimate. We find this trend in every single table except for Table B6, in which the Trump approval upper bound of 13.3% is higher than the RV of 9.2%.

Overall, we believe that confounding as strongly associated with the treatment as Trump approval is not likely plausible particularly since we are considering policy preferences that directly implicated Trump or were in line with his stance toward Muslims, such as the infamous “Muslim Travel Ban.” As such, we do not believe that omitted variable bias will likely explain away the observed findings, but allow the reader to make such judgments for themselves. ¹¹

Next, we consider whether the relationship between American identity similarly holds among Whites and People of Color (POC). Previous work on this topic presents conflicting findings. Greene et al. (2020) find differences between White Americans and Black, Latino, and Asian Americans in levels of attachment to American identity as well as belonging to the nation. In a similar vein, Masuoka and Junn (2013) suggest that American identity is less of a motivator for the politics of POC relative to White Americans. Other research suggests that Black Americans are less attached to American identity than Whites on certain items, but are more nativist overall than Whites (Citrin et al., 2001). Nonetheless, a growing body of scholarship suggests that American identity might be as high (at least) among Black American adolescents as it is for White Americans (Martinez-Fuentes et al., 2020) ¹² and that American identity is a powerful predictor of support for restrictionist immigration policies and anti-immigrant candidates among Latinx Americans (Hickel Jr et al., 2020).

Given these mixed findings, we remain open to how attachment to national identity may shape anti-Muslim policy attitudes by race and decided to examine if American identity differentially shapes anti-Muslim policy preferences among White and POC respondents. In Appendix Tables C1–C3 we present models in which we regressed each outcome variable on American identity among White and POC respondents. The results suggest that American identify functions fairly similarly across White and POC respondents. While we caution cross-model coefficient size comparisons, we note that the magnitude of the relationship between American identity and endorsement of anti-Muslim policies is particularly similar in models using data from the 2020 Lucid survey, which entails a more even distribution of White and POC respondents.

Finally, we consider whether respondents’ tendency to consistently agree or disagree with each of the American identity items presented to them could undermine the validity of the survey data and consequentially impact our findings. To consider the extent to which the presence of response bias, in particular acquiescence bias, may be inflating the relationship between American identity and policy outcomes or even be responsible for it, we took two straightforward and easy-to-follow approaches.

First, we examined the share of the respondents in the CCES and 2019 Lucid who strongly disagreed or strongly agreed with all the four American identity items regardless of whether the items were positively or negatively worded. We found that only less than 0.5% of CCES and less than 3% of Lucid respondents displayed such a behavior. This suggests that only a very small percentage of the study participants exhibited a tendency to strongly agree or disagree with each item regardless of their content. ¹³ While this may be indicative of response bias, the very small portion of participants that exhibited this type of selection behavior diminishes the concern that our data suffers from large-scale or systematic bias that would undermine or be responsible for explaining the relationship between American identity and the various outcome measures.

Second, we calculated the share of respondents who scored 0 or 1 on the American identity scale, which means respondents who strongly disagreed and strongly agreed on the three positively worded items across all the three data sets and on the reverse coded negatively worded item in the CCES and 2019 Lucid. We found that only 1.9% of CCES, 0.9% of 2019 Lucid, and 1.4% of 2020 Lucid respondents scored a 0. Conversely, 25.9% of CCES, 22.6% of 2019 Lucid, and 31.4% of 2020 Lucid respondents scored a 1. Given that many U.S. respondents are likely to view Americanness in a positive light, these percentages are not surprising and do not appear to be abnormally inflated. Nevertheless, we reran all the models excluding such respondents and found that in 10 out of 11 total models (see Appendix Tables D1–D3) the American identity coefficient is statistically significant and very similar in size to the coefficients in models in which such respondents were included. Therefore, we do not believe that response bias is likely responsible for the consistent relationships between American identity and the policy outcomes we documented across three surveys and various model specifications.

6. Conclusion and discussion

Our research contributes to existing scholarship on national attachments and attitudes toward minority groups and applies its insights to appraisals of American Muslims and policies that implicate the population. The results highlight the contradictions between the values that American identifiers claim to cherish and the choices they make when presented with restrictive policies aimed to exclusively marginalize Muslim Americans. Across two studies and multiple policy measures, we consistently find that the strength of individuals’ American identity predicts endorsement of policies that have already curbed or aim to curb the civil liberties and rights of Muslim Americans. Furthermore, we find that this relationship is not moderated by partisanship. That is, partisans across the aisle with stronger American identity are less likely than their counterparts to reject policies at odds with the values of equality and religious liberty.

We believe that our work presents new avenues for future research. While previous studies have clearly shown that not all ethnic, racial, or religious minority groups are equally viewed as American (Devos and Mohamed, 2014), it would be interesting to evaluate whether the effects found in this study replicate to policies that implicate other minority groups. For example, does American identity predict support for policies that aim to take gun rights away from Asian Americans or Black Americans? One could make the argument that Muslim Americans present such a unique case of ‘outsiders’ that American identifiers are willing to forgo their professed values when Muslims are the subject of discriminatory policies, and not necessarily when other groups’ rights are similarly being threatened. A cross-group comparison study will help scholars identify for which populations and policies American identifiers may be more or less likely to undermine their principles for. While we cannot be certain, we anticipate that American identity in the current climate might be more predictive of the willingness to strip away the civil liberties of American Muslims compared to other minority populations who are not perceived to be outsiders on so multiple fronts (e.g., race, religion, and nativity) and perceived to pose both a realistic and symbolic threat. It would also be interesting to explore whether the findings here extend to subsets of partisans, such as libertarian Republicans, who may be opposed to government intervention as it particularly pertains to civil liberties.

Finally, and on a more optimistic note, there is reason to believe that American identifiers can, under certain circumstances, shift their attitudes and show less willingness to endorse anti-Muslim policies that stand at odds with their principles. Recent panel studies on the Muslim Travel Ban show that while high American identifiers were more likely than their counterparts to endorse President Trump’s executive action, a wave of swift and one-sided political communication highlighting the incompatibility between the ban and American values nudged some high identifiers to oppose the ban (Collingwood et al., 2018; Oskooii et al., 2019). Therefore, if inclusive elements of American identity can gain salience under the right conditions, some American identifiers may, at least temporarily, change their preferences.