Skip to main content
Free

Social Exclusion and Political Identity: The Case of Asian American Partisanship

Abstract

How does social exclusion on the basis of racial/ethnic identity affect partisanship and political attitudes? Drawing on sociological research on the group basis of partisanship and psychological theories of social identity, we contend that exclusion at the individual level stemming from racial/ethnic group membership can affect political identity. People who feel that a political party excludes them from the American social fabric based on their race/ethnicity should be less likely to perceive that party as serving their group’s interests and therefore less likely to support that party. We apply our theory to Asian Americans, an understudied minority population that is becoming increasingly politically relevant. Through both a large-scale, representative survey and a novel laboratory experiment, we find empirical support for our principal hypothesis. Our findings partly explain why Asian Americans are overwhelmingly likely to identify as Democrats and advance an identity-oriented explanation of partisanship in American electoral politics.

How does social exclusion on the basis of race or ethnic identity affect partisanship and political attitudes? Despite the many theoretical and empirical advances made in the fields of immigrant and minority political behavior—in addition to the long-standing study of group membership and partisan affiliation—there remain few studies that examine the central role of social exclusion in the formation of political identity. This is surprising given that partisan politics in the United States, historically and currently, has involved both overt and subtle forms of exclusionary political communication (see, e.g., Mendelberg 2001). Further, as many scholars have noted, immigrant political identity does not fit neatly within existing paradigms of American political behavior, such as theories that conceive of party identification as largely a product of early socialization or parental attitudes (e.g., Campbell et al. 1960; Green, Palmquist, and Shickler 2002; Jennings and Niemi 1968). Nor is it well informed by the school of research that views party identification as the result of individuals assessing competing party platforms and electoral performance over long periods of time (e.g., Bartels 2000; Fiorina 1981). Consequently, we provide an alternative explanation of the sources of party attachments.

Our theoretical argument builds on insights from the political science literature on the group bases of partisanship, as well as psychological approaches to understanding social identity. We contend that exclusion at the individual level stemming from racial or ethnic group membership can affect political identity. People who feel that a political party excludes them from the American social fabric should be less likely to perceive that party as serving their group’s interests and therefore should be less likely to support or affiliate with that party. The empirical test of the argument is applied to Asian Americans, a group we argue is both highly relevant to the study of social exclusion and political identity, as well as a minority population whose partisan affiliation has received limited attention in research on political behavior, despite their increasing relevance to US electoral politics. We test our claim using both high-quality, observational data and an innovative laboratory experiment that implements a racial microaggression, a “real world” exclusionary intervention that captures an experience that an individual might have in his or her daily life. We leverage a mixed-method approach, using different sources of evidence, as each approach has unique strengths and limitations. While large-scale surveys are highly representative, laboratory studies allow for precise and controlled randomized interventions. Moreover, a laboratory context offers the ability to have personal interactions that can more strongly cue social exclusion.

Reassuringly, our principal hypothesis is supported by different methodological approaches. Observational data from a nationally representative survey of Asian Americans (Ramakrishnan et al. 2008) demonstrates robust positive correlations between Democratic partisanship and reported racial victimization, which proxies for feelings of social exclusion. We combine this analysis with an original experiment that exogenously increases feelings of social exclusion and examines downstream political consequences to isolate the causal effects of social exclusion per se. If social exclusion causes Asian Americans to support the Democratic Party (and if they believe that the Republican Party is less likely to make them feel “American”), then priming exclusion based on their race/ethnicity in an experimental setting should cause Asian Americans to adopt more Pro-Democratic Party positions. We indeed find that Asians who are subjected to a seemingly innocuous exclusionary cue are more likely to believe that Republicans do not represent their interests, be generally favorable toward Democrats, and identify with the Democratic Party.

Our account of group-based exclusion influencing party identification, as well as the empirical evidence from studying the Asian American population, have implications for understanding the contemporary nature of American electoral politics more broadly. A prominent theory by Bawn et al. (2012) conceptualizes political parties as coalitions of policy-demanding groups (e.g., business, labor, religious denominations). In the American context, other relevant groups could be based on ascriptive characteristics such as sex, race, and ethnicity (see also Achen and Bartels [2016] and Zaller [2012]). Our results are consistent with this theoretical account and suggest that Democrats may have an advantage retaining not only Asian Americans but also other ethnic minority groups that feel socially excluded. The Democratic Party has attracted Asian Americans, a fast-growing, increasingly politically engaged constituency, and we argue that this is based partly on not being tied to exclusionary policies and rhetoric.

The article is organized as follows. We first motivate the empirical case of Asian American partisanship. We then review the related theoretical literature that lays the groundwork for our core hypothesis and explains the rationale of leveraging Asian Americans as a highly relevant test case. We then describe our observational analysis of a nationally representative survey. Next we discuss the methodology and findings of the experimental study that illustrates the importance of social exclusion for minority party identification. We conclude with a discussion of the implications of these findings and pathways for future research.

The Case of Asian American Partisanship

We test our argument using Asian Americans for several related reasons. First, Asian Americans are the nation’s fastest-growing racial/ethnic group (US Census Bureau 2013) and therefore represent an important case for understanding the political implications of social exclusion.1 Among places of origin, Asia recently surpassed Latin America as the main source of documented immigration to the United States; according to the latest census data, about 36% of all new immigrants in 2010 were Asian compared to 31% who were Hispanic (Barrera 2013). People of Asian descent constitute 5% of the US population and are projected to comprise approximately 9% by 2050 (File 2013). Of 16 million Asian Americans, 3.9 million voted in the 2012 elections, accounting for nearly 4% of all voters (a 500,000-person increase from 2008). In some states, they make up a considerably higher proportion of the electorate; for instance, 12% of California voters are Asian American (Baldassare et al. 2015). Since 1996, the number of Asian American voters has increased by 105%, in contrast to a 13% increase in the number of white voters (File 2013).

Second, this rapid increase in Asian immigration has been accompanied by an unmistakable and puzzling pattern of strong Democratic affiliation. In the 2008 and 2012 presidential elections, Barack Obama received 64% and 73% of the Asian American vote share, respectively (Wilkinson 2012). This general democratic orientation is confirmed in other large surveys (e.g., Hajnal and Lee 2011; Wong et al. 2011). The 2012 Pew Research Center Study of Asian Americans found that 51% identify with the Democratic Party versus 27% with the Republican Party. According to the 2008 and 2012 Cooperative Congressional Election Studies (CCES), which have national samples of US adults with large numbers of Asian Americans interviewed, Asians are significantly more likely than whites to exhibit liberal responses to survey items on party identification, ideology, and vote choice. Their responses are much more similar to blacks and Hispanics than whites with respect to these variables (see fig. A1 in the appendix). Such behavior is a major challenge to income-based explanations of vote choice and party identification (Gelman et al. 2009), as Asians are on average much wealthier than white Americans (DeNavas-Walt, Richardson, and Stringfellow 2010) but are more likely to vote for Democrats (see fig. A2).

These demographic and political patterns underscore the need to better understand the bases of Asian American political behavior. Relative to the well-developed literatures on the partisan orientation and voting patterns of other minorities such as African Americans and Hispanics (e.g., De la Garza 2004; Fraga et al. 2006; Nicholson and Segura 2005; Uhlaner and Garcia 2005), the theoretical and empirical literature on Asians remains relatively limited. A few studies jointly analyze the political participation and attitudes of both Hispanic/Latino and Asian immigrants (e.g., Wong 2008), but this research does not theorize about the political orientation of Asians as a distinct immigrant/ethnic group in the United States. Moreover, the existing scholarship that focuses solely on the political behavior of Asian Americans mainly examines questions related to political participation, as opposed to political identity (e.g., Lien 2001; Lien et al. 2001; Wong 2008; Wong et al. 2011). The paucity of research on Asian American partisanship and voting behavior is unfortunate as this minority group is rapidly increasing in political, cultural, and economic influence.

A smaller literature has begun to explore Asian American partisan affiliation; however, this body of work mainly addresses the question of what explains whether Asians have any type of party identification at all, rather than focusing on Asian Americans’ selection of a particular political party. This agenda is a variant of the previous research on political engagement. Hajnal and Lee (2011) provide a general account of why ethnic minorities and white independents select a political party. Wong (2000), studying Asian and Latino immigrants, finds that factors that allow for greater assimilation (e.g., length of time in the United States, citizenship status, and English proficiency) are correlated with developing a party attachment. Moreover, this work builds on much older surveys that may not reflect contemporary political patterns and use more localized samples that may not reflect the national political landscape. For instance, Cain, Kiewiet, and Uhlaner (1991), relying on a 1984 survey in California, found that Asian Americans were more likely to be Republicans than Latinos, a pattern that is no longer true today.2

When examining the literature on Asian American political behavior, two trends emerge. First, Asians are less likely to participate in politics than other ethnic groups. Second, even if they participate, Asians are less likely to align with a political party. Consequently, it is unsurprising that most research on this group has focused on these two empirical patterns. However, given that Asian American political participation has been increasing and an overwhelming majority of today’s Asian American voters are rejecting candidates from the Republican Party, explaining the partisanship of Asian Americans is of scholarly importance. By focusing on the less-explored question of why Asian Americans align more with the Democratic Party, we begin to address a research gap and build on extant literature on Asian American political behavior.

There have been recent efforts to employ larger, national samples of Asian Americans and examine partisan orientation more directly. These studies confirm that Asian Americans are currently more likely to identify with the Democratic Party (e.g., Hajnal and Lee 2011; Wong et al. 2011). However, these researchers draw on previous assimilation models to argue that Asians who have lived in the United States longer are wealthier and more educated should be more likely to be Republican, without focusing on the current pattern of Asian American party identification, leading to our focus on social exclusion as an important causal factor in the formation of political identity.

The Role of Social Exclusion: Theoretical Overview

Group identity and party affiliation

Our core argument and research design are motivated by long-standing literature on social groups and party coalitions, as well as research from psychology on the political importance of social identity. Regarding the former, recent theoretical advances refashion parties as building coalitions of groups (including groups based on shared ethnicity or nationality) and argue that electoral shifts occur when cleavages form around some groups (Bawn et al. 2012). This innovation challenges theories that focus only on individual voter preferences in a retrospective voting framework. It partly draws on an intellectual agenda in American electoral politics that emphasizes the roles of different group affiliations in explaining party support, and the importance of such group coalitions as a basis for explaining macrolevel patterns of partisan identification and vote choice.3

An early example of this research is Kleppner’s classic The Cross of Culture (1970), which brought to the fore the relevance of cultural and religious concerns (the “cross of culture”) over pure economic-based accounts (the “cross of gold”) in explaining voting behavior. Analyzing voting patterns in the 1890s from Ohio, Michigan, and Wisconsin, Kleppner (1970) posited that the Republican Party benefited electorally in the late 1890s by making their religious appeals more inclusive to nonpietistic Christians, while Democratic religious rhetoric became more exclusionary. In his view, the rapidly changing electoral coalitions in the American Midwest at the turn of the century could be traced to exclusionary rhetoric on the basis of religious identity. We apply a similar argument in explaining Asian American partisanship today, with an analogous focus on social exclusion as a key causal factor.

Later studies established substantively large correlations between individual membership with a particular group and partisan identification, often using serial American National Election Studies data or national exit poll data. Such studies also measured the group-level components of each party’s electoral coalition over time to better understand the sources of party support (Axelrod 1972, 1986; Stanley, Bianco, and Niemi 1986; Stanley and Niemi 1991, 2006).4 The theoretical explanations linking specific group affiliations (and in particular, immigrant-group affiliations) to partisan choice are somewhat inconsistent in the literature, but one convincing summary account is provided by Miller and Wlezien (1993), who argue that a party-group connection occurs because group members associate parties with specific policies that are beneficial or detrimental to the group. Drawing on social identity theory, which is discussed further below, they claim that initial group affiliation serves psychological needs related to self-conception (Miller et al. 1981). The group in turn provides an information cue to members regarding its view about which party benefits the individual.5 In their account, party leaders can influence the connection between parties and groups, for example, by explicitly catering to a group. A complementary logic can be found in Blumer (1958), who conceives of groups as politically competing for scarce resources; by extension, groups might ally with parties that promise said resources.

The salience of social identity

The literatures on (1) the incorporation of immigrant groups and (2) group-level affiliation as a correlate of party choice, however, tend to eschew psychological approaches that stress the importance of social identity at the individual level. Research on the emergence of identity choice demonstrates that group-based social identities can become activated and politicized once their salience is heightened. Myriad studies in the fields of political and social psychology document the ease and importance of priming different group identities and attendant support for such in-groups (and/or hostility toward out-groups) (Tajfel 1974; Tajfel and Turner 1986).6 Huddy (2013) provides a recent, comprehensive overview of studies that explore the conditions under which a group identity translates into political cohesion and attitude change. She also documents potential categories of group identities and the conditions under which identity activation can have political consequences. We draw from her review the finding that individuals can have multiple group identities, and that making salient one identity activates preferences related to that identity (and importantly, for our account, political preferences and attitudes toward parties).

Two particular insights related to identity salience are relevant for our main hypothesis. First, situational factors can make some identities salient and increase attachment to a group (Simon 2004). As Huddy notes in her discussion of this relevant strand of the self-categorization literature, social identities are “driven almost completely by one’s immediate perceptual context …. From their perspective, identities vary, in part, because social categories such as age or gender vary in salience across situations” (2013, 517).7 Second, a situational factor that increases “separateness” can activate the salience of one’s minority status. Huddy contends that “salience is heightened by any factor that increases the ‘separateness’ and ‘clarity’ of a category, and one of the factors most likely to increase a category’s clarity is minority status” (2013, 758).8 These insights suggest that priming a group identity based on its distinctiveness (on any dimension) can provoke attachments to that group identity. Brader and Marcus (2013) and Mackie, Devos, and Smith (2000) show that emotions activate certain group identities, but we are unaware of any priming studies on social exclusion in a political context. Related studies also show that “defensive” identification with one’s race is possible, when primed about discourse about other groups (Andreychik and Gill 2009).9

Missing from the psychological account of social identity is an argument that links exclusion at the individual level on the basis of a group membership to political attitudes and party affiliation. We bridge the group-based (i.e., sociological) and the individual-based (i.e., psychological) approaches to advance an argument linking social exclusion to political orientation. We argue that when an individual is made to feel excluded on the basis of her group membership, such exclusion should affect partisanship. This is because exclusion is a particularly relevant and important way of priming a salient identity and stimulating emotional reactions. Individuals who perceive exclusion based on their ethnic background may link such behavior to exclusion of their ethnic group as a whole. Such feelings of social exclusion can be unrelated to public policies and may be individualistic. Nonetheless, individuals who feel that one party excludes them from the social fabric are less likely to perceive the party as serving their group’s interests, and therefore should be less likely to support or affiliate with the party. Currently, because the Democratic Party is largely viewed to be the party with a policy agenda more beneficial for ethnic minority groups (Carmines and Stimson 1989; Lee 2002), individual-level feelings of social exclusion should be linked to greater support for Democrats. If social exclusion is an important causal factor in partisan affiliation, then exogenously activating feelings of exclusion should lead people to link these feelings with their existing views of the political parties.

Linking social identity to Asian American partisanship

The above argument about exclusion and political identity travels broadly to any group that could be made to feel excluded from US politics or society, but it may be particularly applicable to Asian Americans. The premise that social exclusion matters strongly for Asians is consistent with qualitative work on political views toward this group. Chang (2004) shows how the reaction of the US public to the 1996 “Asian Donorgate” campaign finance controversy—where Asian Americans were depicted as outsiders attempting to buy influence with US politicians on behalf of foreign governments—conveys the perception of Asian Americans as perpetually foreign. Kim finds that Asians are often viewed more positively by white Americans than other minority groups and yet viewed as “permanently foreign and unassimilable” (Kim 2000, 16). This attitude toward Asians acts as a barrier to political participation (Kim 2007) and leads to the exclusion of Asians from civic membership (Kim 2000).

The potential significance of social exclusion as a determinant of Asian American political behavior is also motivated by findings in psychological research that document white citizens’ conflation of “American” with “whiteness,” and their perceptions that Asian Americans are less American. This research finds that whites are more likely to be viewed as Americans than Asians (Devos and Banaji 2005; Sidanius et al. 1997; Smith and Zarate 1992; Stroessner 1996).10 Other studies using an array of methodological approaches similarly observe that Asians face among the strongest social barriers to social assimilation (Devos and Heng 2009; Devos and Ma 2008; Liang, Li, and Kim 2004; Yogeeswaran and Dasgupta 2010). Although Asian Americans are perceived as less American, Asian Americans themselves are just as likely as white Americans to identify themselves as American and have explicit and implicit patriotic attitudes (Cheryan and Monin 2005; LaFrombroise, Coleman, and Gerton 1993; Sidanius et al. 1997). The impact of such exclusionary feelings can be magnified in other political contexts, such as advertisements, political rhetoric, and policy positions on issues related to Asians. To the extent that the Democratic Party is seen as less exclusionary, we theorize that greater feelings of exclusion from the United States make Asian identification with the Democratic Party more likely. This can be due to perceptions of exclusionary behaviors by Republican political elites as well as Republican identifiers in the mass public.

Further, as many scholars of minority political behavior emphasize, social context shapes the impact of individual-level variables on political identification, and this context may be unique for Asians. A key aspect of this context is that Asians as a group are perceived differently from other minority groups, and this perception molds group consciousness and interactions (Kim 2000; Lien 2000; Wong 2000). For instance, while viewed as a “model minority” with desirable traits (Chou and Feagin 2008), as noted above, Asians are simultaneously perceived as less “American” and therefore perhaps unequal citizens, whether due to stereotypes based on their physical appearance (as they have great difficulty “passing” as stereotypical white Americans) or due to other visible or cultural characteristics such as accent, dress, food, language, religion, and the like (e.g., Devos and Banaji 2005; Sidanius et al. 1997). This is perhaps in contrast to African Americans and Hispanics/Latinos, who have longer histories of residence in the United States, and with respect to the latter group, more geographic proximity to the United States.

Our hypothesis that social exclusion plays a role in Asian American party identification also builds on the social psychological literature that notes the importance of “everyday” discrimination for feelings of social exclusion among minority groups. This literature documents how commonplace the sense of social exclusion may be for racial minorities. Asian Americans are more likely than whites to be victims of racial microaggressions, which are defined as brief and commonplace verbal, behavioral, or environmental indignities. Such aggression can communicate negative racial slights, leading to perceptions of exclusion (e.g., Sue et al. 2007). We operationalize these microaggressions in both the observational and empirical data analyses, which we turn to in the next section.

By no means are we suggesting that feelings of social exclusion are unique to Asian Americans. Indeed, many previous studies largely focusing on Latino politics and immigration have discussed how policies at the aggregate level can directly exclude minority groups as well. These studies detail how anti-immigrant rhetoric and related policies (such as English-only ballot initiatives or explicit entry restrictions) target Latinos and treat them as “alien” to US culture and therefore less deserving of the benefits of citizenship (Schmidt Sr. 2000). As Segura and Rodrigues (2006) write regarding Latino and other immigrant populations, “immigrant populations are uniquely vulnerable to suspicions that they constitute an unassimilable ‘other’” (380). Kinder and Kam (2009) discuss the enduring relevance of ethnocentrism as a basis for some white Americans’ policy and partisan preferences.11 We build on this research to focus on the effect of individual-level exclusion on partisan affiliation.

Evidence from an Observational Study

Before presenting the causal test of our hypothesis using experimental data, we first assess whether there is correlational evidence in a high-quality, probability sample of Asian Americans. We analyze the 2008 National Asian American Survey (NAAS), which interviewed a nationally representative sample of Asians (Ramakrishnan et al. 2008).12 This is among the most detailed data on Asian American political preferences to date, consisting of a telephone sample of 5,159 Asians. The majority of the sample were US citizens (76%), and 11% were born in the United States. The average household in the study reported an income in the $50,000–$75,000 range. The two largest groups were of Chinese or Taiwanese descent (26%) and of South Asian descent (22%).

The dependent variable of interest for these analyses is party identification.13 Approximately 39% of the sample identified with the Democratic Party; 19% identified with the Republican Party (these figures include self-reported independents who lean toward a party). A third of the sample reported “not thinking in these [partisan] terms,” and 9% reported “don’t know” or refused to answer the party identification question. A striking feature of this distribution is the plurality of respondents who do not select one of the two major political parties. Other research has explored this empirical pattern; we extend this literature by examining determinants of partisan choice among identifiers. Party identification is measured on a 6-point scale (strong Democrat, not strong Democrat, lean Democrat, lean Republican, not strong Republican, strong Republican). We rescaled the party identification measure to lie between 0 and 1, with higher values representing Democratic identifiers. We describe below how we handle missing values on the dependent variable.

To proxy for feelings of social exclusion, we construct a binary measure indicating whether a respondent reported that he or she has been a victim of racial discrimination.14 Respondents reported whether they had ever been racially discriminated against in each of the following situations: (1) unfairly denied a job or fired, (2) unfairly denied a promotion at work, (3) unfairly treated by the police, (4) unfairly prevented from renting or buying a home, (5) unfairly treated at a restaurant or other place of service, or (6) been a victim of a hate crime. Nearly 40% of the sample reported being a victim in at least one of these situations.15 The racial incidents in these categories can be plausibly argued to temporally precede the dependent variable; therefore, the social exclusion variable can be considered exogenous. In other words, the variable does not represent self-reported attitudes on feelings of discrimination but rather recollections of specific incidents. Of course, people may misremember or misreport experiences based on their political attitudes, but the survey item likely obviates the inferential problems involved with correlating two attitudes measured in the same survey with each another. Nevertheless, we address potential issues of causal inference below via an experiment in which we exogenously increase feelings of social exclusion.

We control for standard demographic and political variables that are associated with partisan identification, as well as additional variables possibly relevant for Asians: gender, income, education, age, percentage of time spent living in the United States, religiosity, citizenship status, and liberal-conservative ideology.16 Gender, religiosity, and citizenship status are all binary indicators; all other demographic variables are linear transformations coded to lie between 0 and 1.17 By recoding both the dependent and independent variables in this manner, we can interpret a regression coefficient as representing a 100β percentage-point increase in the dependent variable associated with moving from the lowest to highest possible value of the independent variable. Summary statistics for the variables of interest can be found in the appendix (see table B1).

Table 1 reports results of OLS regressions where the dependent variable is the 6-point measure of party identification. Many respondents did not respond to the party identification question that constitutes the dependent variable. We conducted the analysis in various ways to address missing data on party identification. Models 1–2 display the results from listwise deleting respondents who did not answer the question; this approach eliminates approximately 42% of the sample. Models 3–4 display results of the same OLS estimations by coding respondents who did not identify with a party as the midpoint of the scale (0.5). Both of these approaches make different assumptions about nonrespondents, and the fact that the results are similar across model specifications increases confidence in our results. Further, we estimated multinomial logit models where refusal to answer the question is treated as a unique response category (see table B2). We also imputed missing values for party identification (see table B3). Both of these alternative analytical approaches yielded similar results.

Table 1.

OLS Regressions Predicting Asian American Partisan Identification

 Model 1Model 2Model 3Model 4
Social exclusion.04***.04***.03***.03***
 (.01)(.01)(.01)(.01)
Female.05***.04***.02***.02***
 (.01)(.01)(.01)(.01)
Income−.01−.02.00.00
 (.03)(.03)(.02)(.02)
Education.11***.12***.07***.07***
 (.02)(.02)(.01)(.01)
Age−.31*−.12−.23**−.12
 (.14)(.13)(.08)(.08)
Age2.19.04.18*.10
 (.15)(.15)(.09)(.09)
Percentage of life spent in US.12***.13***.10***.11***
 (.03)(.03)(.02)(.02)
Religious−.08***−.05***−.03***−.02**
 (.02)(.01)(.01)(.01)
US citizen−.06***−.06***−.02*−.02*
 (.02)(.02)(.01)(.01)
Ideology.36***.28***
  (.02) (.01)
Constant.63***.34***.54***.35***
 (.04)(.04)(.02)(.03)
Observations2,9932,9935,1595,159
Adjusted R2.05.13.03.09

Notes. Dependent variable is partisan identification [0 (Strong Republican)→1 (Strong Democrat)]. Regression standard errors are in parentheses. Columns 1 and 2 report models excluding respondents who did not report a party identification. Columns 3 and 4 report models recoding missing values on party identification as the midpoint.

*. p < .05.

**. p < .01.

***. p < .001 (one-tailed).

View Table Image

Reported racial victimization (a measure of social exclusion) is consistently positively and statistically significantly correlated with Democratic Party identification.18 Across all models, our proxy for social exclusion increases identification with the Democratic Party by 3 to 4 percentage points depending on the specification (see table 1). Note that income remains uncorrelated with partisan identification across all specifications, while higher education is positively correlated with Democratic identification, as is length of time in the United States. Female Asians are more likely to be Democrats, while identifying with a religion and US citizenship are positively associated with Republican identification. The significance of social exclusion in predicting party identification is robust to the inclusion of a control for political ideology (see columns 2 and 4).19

We also estimated an alternate specification where the dependent variable is identification with the Democratic Party in binary terms (coded as 1 if the respondent leans toward the Democratic Party, is a not-strong Democrat, or a strong Democrat and 0 otherwise). The results of the logistic regressions can be found in appendix table B4. Whereas the previous analyses took into account the extremity of association with a party, this analysis solely predicts an individual crossing the cut point from being a Republican to being a Democrat. This represents a strong test as we are predicting switching from one party to another, not simply being more attached to a particular party. The coefficient on social exclusion remains positive and significant. The estimated marginal effect of being a victim on the probability of identifying as a Democrat is about 7 to 8 percentage points.

We find that the effect of social exclusion does not vary across income brackets within the Asian population; we find no interaction effect between income and our victimization measure (see table B5). Also, we do not observe heterogeneity with respect to country of origin, consistent with our theoretical expectation that the effects of social exclusion affects Asian Americans broadly speaking. This suggests it is appropriate to pool across different ethnic groups in line with our theoretical focus.20

Existing observational approaches (including the analysis just described, as well as examples from the extant literature) that use diverse samples of Asians show a correlation between perceptions of social exclusion and Democratic partisan identification, providing some prima facie evidence in support of our hypothesis. The relationship between social exclusion and partisan identification in this analysis of observational data is modest (yet is comparable in size to religiosity and one-third as large as education). Although the survey item used to measure social exclusion is an imperfect proxy, it is the best available measure in a large survey to capture recall of instances of perceived social exclusion. We conducted our own original experiment, in part, to more directly operationalize social exclusion. Observational analyses also remain vulnerable to concerns about omitted variable bias and reverse causation, raising the possibility that such correlations might be spurious. For instance, political orientation itself may predict whether people are sensitive to social exclusion. Additionally, variables such as education or cosmopolitanism could be related to both party identification and perceptions of exclusion. Our experimental research design described below builds on observational findings by isolating exclusion and showing the impact of an exogenous change in social exclusion on party identification.

Experimental Evidence

In this study we exogenously manipulated feelings of social exclusion by randomly exposing individuals to a “real world” cue designed to make them feel as if they are not welcome in the United States. We tested whether making Asians feel this way made them less likely to support Republicans as they might associate the party with these negative feelings. If social exclusion is a causal factor that can explain party affiliation, then manipulating it should increase positive views of the Democratic Party and identification with it.

Below we demonstrate through various outcome measures that the experiment captures this theoretical pathway. Asian respondents who were exogenously made to feel socially excluded based on their Asian background were more likely to view the Republican Party as exclusionary (as their feelings of exclusion were linked to their views of the party) and consequently less likely to exhibit positive feelings for the Republican Party. Our study builds on a voluminous literature in political and social psychology on priming (Bargh 1982). In a plethora of laboratory and survey contexts, a range of subtle, external interventions have been shown to activate hypothesized behavioral outcomes and attitude changes. We draw on research that shows how primes can unconsciously activate specific personality traits, change task performance (physical and cognitive), affect stereotypes and attitudes toward out-groups, and change political attitudes (e.g., Berger, Meredith, and Wheeler 2008; DeMarree, Wheeler, and Petty 2005; Lodge and Taber 2005). The mechanisms linking primes to specific outcomes depend on the outcome of interest, but we focus on testing the specific prime of a racial microaggression on social exclusion, which has been shown to occur in other laboratory contexts.

Procedures and design

We conducted a laboratory experiment where the measured political outcomes were both survey-based and behavioral. The study was conducted in the behavioral lab of a major research university; subjects were paid $10 for participating. This study had to be conducted in person in the laboratory (as opposed to with a more general sample over the Internet) because the experimental manipulation requires interpersonal contact. We opened the study to Asians and whites based on demographic background data of the respondents collected well before the experiment took place. Subjects were told that they were participating in a study about current events. The study was conducted between November 26 and December 7, 2012; 114 subjects participated; 61 were of self-reported Asian descent and 53 described themselves as white.21 Upon entering the facility, a white female research assistant welcomed the subject and was instructed to privately assess and document whether the subject was of Asian descent.22 The subject was then guided to a computer by the assistant. The experimental protocol proceeded as follows: (1) the assistant was provided with a list of which subjects were randomly assigned to be in the treatment condition and receive a racial microaggression; (2) the assistant was instructed to say the following to each subject assigned to the treatment condition before he or she began the survey, “I’m sorry; I forgot that this study is only for US citizens. Are you a US citizen? I cannot tell.” If the subject was a US citizen, the assistant was instructed to say “OK, go ahead” and have the respondent start the survey; if the subject was not a US citizen, the assistant was instructed to pause and then say “it’s OK, go ahead.” This procedure was also applied to white respondents so that we could assess any baseline effect of the treatment unrelated to racial microaggression (e.g., simple rudeness). Randomization was successful from the perspective of achieving balance on observable demographic characteristics (see table C1).

The racial microaggression employed in the study was a microinvalidation—“verbal comments or behaviors that exclude, negate, or nullify the psychological thoughts, feelings, or experiential reality of a person of color” (Sue et al. 2007, 278). The invalidating question asked by the assistant to treated subjects is similar to commonplace questions such as “Where are you really from?”; “Where were you born?”; and “You speak good English.” In other words, Asian participants assigned to the treatment condition were made to feel like foreigners in their own country. White participants should not have had a similar reaction, making them an important benchmark for comparison.23 This intervention builds on and is a modification of social exclusion or racial microaggression interventions in other laboratory contexts (e.g., Cheryan and Monin 2005). It also accurately captures a common interaction in daily life among some Asians when they interact with other individuals who do not assume Asians are either citizens or “American” (Sue et al. 2007).24

Outcome measures

Subjects then completed an online survey that measured political attitudes. All measures were coded such that higher values reflect negative views of the Republican Party relative to the Democratic Party. Descriptive statistics of our outcome measures can be found in table C2.25

Closed-minded

Study participants were asked: “How well does the term ‘closed-minded’ describe [Republicans/Democrats]?” (response options: “very well,” “somewhat well,” “slightly well,” and “not well at all”). To develop a measure of how much more closed-minded the respondents felt Republicans were relative to Democrats, the response to the Democratic item was subtracted from the response to the Republican item. This resulting difference measure is a 7-point scale, recoded to range from 0 (meaning that Democrats are viewed to be maximally—with respect to the measure’s range—more closed-minded than Republicans) to 1 (meaning that Republicans are viewed to be maximally more closed-minded than Democrats).

Ignorant

Respondents were asked: “How well does the term ‘ignorant’ describe [Republicans/Democrats]?” (response options: “very well,” “somewhat well,” “slightly well,” and “not well at all”). As with the closed-mindedness question, we create a measure of how much more ignorant a given participant felt Republicans were relative to Democrats by differencing out a participant’s response to the question on Democrats from the response to the question on Republicans. This measure is again a 7-point scale, recoded to range from 0 (meaning that Democrats are viewed to be maximally—with respect to the measure’s range—more ignorant than Republicans) to 1 (meaning that Republicans are viewed to be maximally—with respect to the measure’s range—more ignorant than Democrats).

Represent interests

Study participants were asked: “How well do you think the [Democratic/Republican] Party is likely to represent the interests of people like yourself?” (response options: “very well,” “somewhat well,” “slightly well,” and “not well at all”). Again, we subtract a participant’s response to the question when asked about the Republican Party from the response to the same question asked about the Democratic Party. This difference measure is a 7-point scale, ranging from 0 (meaning that the Republican Party is viewed to be maximally—with respect to the measure’s range—more representative of people like them than the Democratic Party) to 1 (meaning that the Democratic Party is viewed to be maximally—with respect to the measure’s range—more representative of people like them than the Republican Party).

Net likes

The net likes measure is a combination of the following task questions: (1) “Is there anything in particular you like about the [Democratic/Republican] Party?”; and (2)“Is there anything in particular you dislike about the [Democratic/Republican] Party?” Each of the four questions was followed by the following directive: “Please list as many responses as you like, listing each response in a separate field. If there is nothing that you like about the [Democratic/Republican] Party, just skip ahead.” Respondents were given up to ten fields to enter likes/dislikes. The net likes measure is a difference-in-difference measure. For each party, we compute a net likes measure by assessing how many more positive traits are listed as opposed to negative traits. We then take the difference between the net likes listed by a participant for the Democratic Party and that of the Republican Party. This measure is rescaled to be between 0 and 1, such that a higher number means that (compared to when thinking about the Democratic Party) the respondent listed fewer things they “liked” about the Republicans relative to “disliked” about the Republicans. This task required a great deal of effort on the part of respondents and therefore can be interpreted as a behavioral manifestation of liking or aversion toward a political party.

Feeling thermometer

We asked respondents to report their warmth to the parties on feeling thermometers: “On a scale of 0 to 100, where 0 represents a completely negative opinion and 100 represents a completely positive opinion, how would you rate the [Republican/Democratic] Party?” We assess how much more negative a participant views the Republican Party relative to the Democratic Party by differencing out a participant’s response to the feeling thermometer question about Republicans from the response to the identical question about Democrats. This new measure is rescaled to lie between 0 and 1, ranging from 0 (meaning that the respondent has a maximally—with respect to the measure’s range—positive opinion of Republicans relative to Democrats) to 1 (meaning that the respondent has a maximally—with respect to the measure’s range—positive opinion of Democrats relative to Republicans).

Party identification

Participants’ party identification was measured through a sequence of questions, where participants were first asked: “Generally speaking, do you think of yourself as a Republican, a Democrat, an independent, or what?” (response options: “Republican,” “Democrat,” “independent,” and “other”). Those who answered “Republican” or “Democrat” were then asked the follow-up question: “Would you call yourself a strong [Democrat/Republican] or a not very strong [Democrat/Republican]?” (response options: “strong” and “not very strong”). Participants who answered the first question with either “independent” or “other” were subsequently asked: “Do you think of yourself as CLOSER to the Republican Party or to the Democratic Party?” (response options: “Republican Party” and “Democratic Party”). Responses were combined to create a 6-point scale, which is recoded to lie between 0 and 1, where higher values reflect stronger identification with the Democratic Party.

Pro–Democratic Party (PDP) index

We averaged the six outcome variables described above into a single additive index reflecting a latent variable of pro-Democratic attitudes. The advantage of this averaged measure is that it nets out measurement error associated with any one of the index components.

Manipulation check

As a manipulation check, subjects also completed a behavioral task where they listed as many US politicians they could think of on the spot. The logic of the measure is that if the racial microaggression offends Asians, they may desire to compensate by showing how much they know about American politics in an attempt to feel less excluded and prove themselves as more “American.” Consequently, they should also spend more time answering this question. This procedure follows from what was found in similar microaggression interventions in other laboratory contexts (e.g., Cheryan and Monin 2005). The goal of the manipulation check is to ensure that the treatment is properly manipulating the theoretical construct of interest (i.e., feelings of exclusion).26

Results

We present regression results predicting the outcome variables with a dummy for the racial microaggression treatment, a dummy for whether a respondent is Asian, and the interaction between these two dummy variables (see table 2). We use the white respondents as a baseline group to observe the effect of the microaggression per se. The quantity of interest, which is captured in the interaction term and plotted in figure 1, is the difference between Asians and whites in their response to the microaggression treatment on the various outcome variables. The treatment effect among whites allows us to determine the causal effect of the experimental manipulation, netting out any direct effect of the treatment unrelated to social exclusion based on ethnicity (e.g., perceived rudeness or awkwardness of the research assistant’s statement). More formally, the estimand of interest is (Y¯t,aY¯c,a)(Y¯t,wY¯c,w), where Y¯ represents the average of the outcome variable, and the subscript denotes whether respondents were assigned to the treatment (t) or control (c) groups, and whether the respondent was Asian (a) or white (w).

Table 2.

Experimental Study Results

Dependent VariablePDP IndexClosed-MindedIgnorantRepresent InterestsNet LikesFeeling Therm.Party IDTime TakenNames Listed
β1: Microaggression−.03−.02−.03−.07−.06−.04.03−.25−2.80
treatment(.05)(.05)(.05)(.06)(.04)(.05)(.07)(.23)(2.00)
β2: Asian−.08−.04−.03−.11*−.14***−.09*−.05−.65**−3.96*
respondent(.04)(.05)(.05)(.05)(.04)(.05)(.07)(.22)(1.88)
β3: Treatment x.13*.13*.14*.18*.10*.11.10.69*5.78*
Asian(.06)(.07)(.07)(.08)(.06)(.07)(.09)(.32)(2.73)
Constant.67***.66***.62***.73***.62***.70***.69***4.84***8.10***
 (.03)(.04)(.04)(.04)(.03)(.03)(.05)(.15)(1.32)
Observations114114114114114114114114114
Adjusted R2.02.02.03.02.08.01.01.05.02

Notes. Regression standard errors are in parentheses. Baseline categories are control group and white respondents.

*. p < .05.

**. p < .01.

***. p < .001 (one-tailed).

View Table Image
Figure 1.
Figure 1.

Experimental study results. The figure displays effect sizes with 95% confidence intervals.

Before delving into the main results, we note that the manipulation check was successful. Compared to white respondents, Asians responded to the racial microaggression treatment by listing 5.78 more US politicians vis-à-vis the control group. This difference is statistically significant (p=.02). Further, the effect of the treatment on the time Asians took to complete the survey was 88 seconds more than the treatment effect among whites (p=.02).27

The treatment affected Asian respondents’ views of the Republican Party compared to whites, generally increasing their negative views of the party (see the third row of table 2). The simple intervention of making an Asian subject feel excluded with respect to “Americanness” and being suspected of not being a US citizen increased negative dispositions toward the Republican Party and increased positive views of the Democratic Party compared to the baseline treatment effects among white respondents. The microaggression treatment reduced affinity toward the Republican Party as measured by the Pro-Democratic Party (PDP) Index (13 percentage points; p=.02). The results indicate a clear association between a sense of exclusion and positive feelings toward the Democratic Party at the expense of the Republican Party.

When examining the components of the index separately, we again see consistent, positive effects on pro-Democratic attitudes as a result of the treatment, although some are larger than others. Compared to their view of Democrats, Asians in the treatment group viewed Republicans to be more closed-minded (13 percentage points) and more ignorant (14 percentage points). This suggests that priming social exclusion activates latent associations respondents had between the Republican Party and negative traits associated with racial discrimination. Accordingly, Asians in the treatment condition are 18 percentage points less likely to believe that the Republican Party represents their interests. Also in response to the treatment, Asians listed fewer things they “liked” about Republicans (10 percentage points) and rated Republicans 11 percentage points less favorably than Democrats on the feeling thermometer. The negative associations therefore led Asian respondents to have more negative feelings toward the Republican Party. All of these effects produced lower Republican identification among Asian respondents in the treatment condition. Asian respondents in the treatment group are 10 percentage points more Democratic on the 6-point partisan identification scale, more than half a response category. The treatment effects on the feeling thermometer and party identification outcomes do not quite achieve standard levels of statistical significance given that these are fairly rigid and enduring political dispositions. Nonetheless, the other, stronger results demonstrate that the perceptions of the parties (i.e., the ingredients of party identification) were affected by making respondents feel socially excluded.28 Overall, this study provides evidence of a causal link between social exclusion and orientation toward the Democratic party.

Discussion

As Wilkinson (2012) writes, “If you were a black-haired Buddhist from Taipei or a brown-skinned Hindu from Bangalore, which party would instinctively seem more comfortable?” This glib statement in our view does capture some aspects of how excluded groups might adopt political identities. The role of social exclusion, despite being tested in other social psychological contexts, has not been systematically examined as a possible determinant of partisan orientation at the individual level. Nor has it been tested in a way that allows for identifying causal effects. In our empirical approach, we find strong evidence that in the case of Asian Americans, exclusion based on group membership affects party affiliation.

This study makes three main contributions. First, we introduce a novel theoretical account of political identity. To develop our argument about the importance of social exclusion, we bridge sociological work on the group basis of partisanship with psychological theories of social identity. In doing so, we conceive of group-based exclusion as an individual-level phenomenon. Second, we apply our theory to an understudied minority population, but one that is becoming increasingly politically relevant to contemporary American politics. Third, we introduce a novel experimental approach that allows us to exogenously manipulate social exclusion and assess downstream political consequences, allowing for the causal identification of the effect of exclusion per se.

There are several opportunities for building on this study to contribute to a broader research agenda. Most obviously, the main theoretical insight and empirical approach employed in this article can be applied to many other groups for which social exclusion may be a relevant variable. The experimental design could be replicated on African American, Hispanic, Middle Eastern, or even European American subjects. Further, we focused on social exclusion in a laboratory context to capture a racial microaggression or exclusionary interaction that Asian Americans might receive in everyday life. Other studies might simulate other types of microaggressions, or more overt aggression in an explicit political context (such as campaign advertising), and do so in a broader survey context. Survey-based studies could either mimic such racial microaggressions or have subjects read explicitly partisan (or nonpartisan) content that has rhetoric that makes them feel excluded.

Because our theoretical mechanism of interest was predicted to apply to Asian Americans broadly, we have collapsed all “Asians” into a single ethnic category. In spite of the immense diversity within the Asian American population, there is value in considering all Asians Americans as a group as a first step. There is evidence that the majority of Asian Americans adopt, in part, a “pan-Asian” identity (Lien 2001; Lien, Conway, and Wong 2003). For example, one study found that 60% of Asian Americans accepted the panethnic term as part of their identity (Lien et al. 2003). Literature on panethnic identity has also shown that “exclusionary action, threats, or discrimination directed toward ethnic, linguistic, or cultural groups activate panethnic identities and group formation” (Okamoto and Mora 2014, 230), and Massey and Sanchez (2010) found that hostility toward immigrants and discrimination within the United States cultivated a Latino panethnic identity among Latin American immigrants. Research on panethnicity has demonstrated that external threats intensify group cohesions as group members band together in defensive solidarities (Espiritu 1992). A racially defined group can suffer punishments as a result of an externally imposed membership in a larger group. In other words, group members can suffer sanctions due to the activities of others who resemble them and not for their own behavior (Light and Bonacich 1988), which can promote panethnic support. Like Latino panethnic identity, scholars have noted that a pan-Asian identity was fueled through defensive mobilization. According to Espiritu (1992), anti-Asian activities “necessarily lead to protective pan-Asian ethnicity” (134). Our laboratory social exclusion treatment makes the common fate of Asian Americans salient by examining the impact of social exclusion based upon their background as Asians, and as such, we expect the dynamics to be similar across Asian American ethnicities. In fact, our manipulation check in the experiment indicates strong evidence that regardless of ethnicity, our social exclusion treatment provokes a general desire for those excluded to show they should be included. This evidence suggests that regardless of ethnicity, making Asians feel excluded based on their appearance can make them resentful, as they desire to be acknowledged and included in the American social fabric.

Nonetheless, moving beyond the approach here and treating each Asian American ethnicity separately would be a valuable advancement for testing the social exclusion hypothesis. Some research on Asian Americans emphasizes that Asian Americans, regarding some preferences, perhaps should not be conceived of as a unified political community (Nakanishi 1991; Tam 1995). Future research that treats them as a disparate population rather than a monolithic group is therefore encouraged. However, given the difficulty of obtaining such targeted samples and the value in considering Asian Americans as a singular electoral coalition, this research offers a starting point. More generally, we hope that this study inspires novel conceptions of party identification that might be unique to immigrant and marginalized groups, of which social exclusion is only one.

We thank Matt Barreto, Larry Bartels, María Jiménez Buedo, Tony Carey, Sapna Cheryan, Wendy Tam Cho, Josh Clinton, John Geer, Zoltan Hajnal, Marc Hetherington, Jane Junn, Cindy Kam, Taeku Lee, Pei-te Lien, Benoit Monin, Efrén Pérez, Karthick Ramakrishnan, Candis Watts Smith, Janelle Wong, and John Zaller for helpful discussions and comments. We also thank seminar participants at Loyola Marymount University, New York University, The Ohio State University, Oxford University, Princeton University, University of North Carolina, and Vanderbilt University. This article was presented at the annual meetings of the American Political Science Association, the Midwest Political Science Association, and the Western Political Science Association. Finally, we thank the editors and reviewers for their constructive feedback. Sonya Bendriem and Guilherme Russo provided excellent research assistance.

Notes

Alexander Kuo () is an associate professor in the department of politics and international relations and tutor at Christ Church, the University of Oxford, Manor Rd, Oxford OX1 3UQ. Neil Malhotra () is professor of political economy, Stanford Graduate School of Business, 655 Knight Way, Stanford, California 94305-7298. Cecilia Hyunjung Mo () is an assistant professor of political science at Vanderbilt University, Nashville, Tennessee 37203-5721, and the W. Glenn Campbell and Rita Ricardo-Campbell National Fellow and Robert Eckles Swain National Fellow (2015–2016) at Hoover Institution, Stanford University, Stanford, California 94305-6003.

Original data collection was approved by the Stanford University IRB (25840). Data and supporting materials necessary to reproduce the numerical results in the article are available in the JOP Dataverse (https://dataverse.harvard.edu/dataverse/jop). An appendix with supplementary material is available at http://dx.doi.org/10.1086/687570.

1. Asian Americans are defined here as those persons living (permanently) in the United States and whose ancestry can be traced to East Asia (e.g., China, Japan, and Korea), Southeast Asia (e.g., Cambodia, Indonesia, Laos, Malaysia, the Philippines, Singapore, Thailand, and Vietnam), and South Asia (e.g., Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, and Sri Lanka).

2. In the 2012 US presidential election, Barack Obama garnered 73% of the Asian American vote, exceeding his support among Hispanics voters (71%) (Wilkinson 2012).

3. As Axelrod (1972) noted in his empirical analysis of the group-level question, Key (1955) argued that a party usually dominates by cobbling together a stable winning coalition. Axelrod’s analysis focused on the following groups’ importance for maintaining the electoral strength of the Democratic Party in the post–New Deal era: blacks, Catholics, the poor, Southerners, and unions. The theoretical emphasis on group cleavages and party coalitions is also present in Schattschneider (1975).

4. This research agenda is related to the immense cross-national literature on the relevance of different cleavages for the formation of politically relevant groups, and the conditions under which such groups form and politically compete (e.g., Lipset and Rokkan 1967).

5. The party-group connection can change over time if the number of people who identify with a group changes or perceptions of the group’s connection with the party change.

6. For one review on priming and identity salience, see Devos and Banaji (2003). For a general theoretical summary of the social psychological literature on how individual affiliation with groups can be politicized, see Simon and Klandermans (2001).

7. Further, group salience can enhance the political effects of a strong identity. For example, one study finds that white Americans were more supportive of spending on minority education when their national identity was made salient but were less supportive of the same program when their racial identity was made salient (Transue 2007).

8. For other research on how category salience plays a clear role in shaping identity, see studies by McGuire and colleagues, who report evidence that children with an ethnic minority in their classroom (and whose own ethnicity is therefore salient) are more likely to describe themselves in terms of their ethnicity; and children in families where there are more members of the opposite gender are more likely to mention their gender when describing themselves (McGuire and Padawer-Singer 1976; McGuire et al. 1978).

9. Some observational work on minorities finds that members of ethnic and racial groups identify primarily as American and only secondarily as members of their ethnic or racial group, despite the greater salience of minority group status in the United States (Citrin, Wong, and Du 2001; Sears et al. 1999).

10. See also Masuoka and Junn (2013) for a discussion of the importance of racial hierarchy in the United States in explaining attitudes toward immigrant groups.

11. In related research on African Americans, some argue that defense in the white US population of “American” values (supposedly individualism or self-reliance) can in fact be a guise for anti-black preferences, with the implication that such groups should be excluded from social policies because “they” do not share such values (Kinder and Sanders 1996).

12. This observational analysis partially replicates similar analyses reported in Chapter 6 of Hajnal and Lee (2011), although we focus explicitly on the choice of Asians to identify with the Democratic Party. See Lee (2008) for a discussion of data collection and for further details on the demographic composition of the NAAS sample.

13. Respondents were asked: “Generally speaking, do you think of yourself as a Republican, Democrat, independent, some other party, or do you not think in these terms?’’ where the order of Republican and Democrat in this question was randomized (response options: “Republican,” “Democrat,” “independent,” “other party (specify),” “Do not think in these terms’’). Those who answered “Republican” or “Democrat” were then asked “Would you call yourself a strong Republican/Democrat or not a strong Republican/Democrat?” Those who answered “independent” were asked the follow-up question: “Do you think of yourself as closer to the Republicans or the Democrats?”

14. The text of the preamble to the questions reads: “We are interested in the way you have been treated in the United States, and whether you have ever been treated unfairly because of your race, ancestry, being an immigrant, or having an accent.” Respondents were then asked: “Have you ever been unfairly denied a job or fired?” (response options: “Yes” or “No”). A similar wording followed for the remaining five discrimination questions.

15. About 18% report being a victim in one of these categories; 10% report having been a victim in two of these categories; and 7% report having been a victims in three or more categories. We also included these levels as dummy variables and did not observe any significant differences between those coefficients, leading us to collapse responses in a binary fashion.

16. Due to missing data in the independent variables, we present models that include binary indicators of missing data and recode individuals missing on each of the control variables as 0 on those variables. This allows us to not list-wise delete any data, while allowing for an intercept shift for respondents who did not answer various questions.

17. Gender is captured through a binary variable, which is 1 if a respondent is female and 0 otherwise. Education is coded on a 5-point scale: did not graduate from high school, high school graduate, some college, college degree, postgraduate degree. Income is coded on an 8-point scale representing increasing income categories (see table B1 for the income categories). Religious is coded as 1 if a respondent chooses a religion and 0 otherwise. Political ideology is coded on a 6-point scale with higher values corresponding to being liberal. Percentage of time spent living in the United States is coded as the fraction of a person’s age spent in the United States.

18. Given that our hypotheses are directional, we employ one-tailed hypothesis tests throughout.

19. We estimate models both including and excluding political ideology given the concern that ideology may actually stem from initial attachment to a political party.

20. We also investigated whether the relationship between social exclusion and partisanship was more pronounced among Asians who were more politically engaged and therefore may have a clearer conception of differences between the parties. We find that voter registration status, past turnout, intended turnout, and political interest do not moderate the relationship. However, political news consumption significantly moderated the relationship between exclusion and Democratic party identification; moving across the news consumption scale increased the relationship by 5 percentage points (p = .05). A similarly large moderating relationship was observed for the extent of political activism (6–7 percentage points), but this effect did not reach conventional levels of statistical significance. Detailed results are presented in table B6.

21. Due to constraints on sample size in the experiments, we did not have the statistical power to run analyses separately for different subgroups of Asian Americans. However, this is less important for our specific research question, given that we theoretically expect social exclusion to affect Asian Americans as a whole. This is not to say that the effects are equivalent across all ethnic subgroups. Future research should delve into the obvious diversity of the Asian American community. We return to this issue in the Discussion section.

22. The research assistant’s assessment of respondents’ ethnicity matched respondents’ self-reports in all cases.

23. We also expect noncitizens to feel excluded by the treatment. Even though they are not citizens, they may feel singled out given the presumption of their citizenship status based on their physical appearance. We found no significant difference in treatment effects between citizens and noncitizens; the difference in the treatment effect on the Pro-Democratic Party (PDP) index, which is described in the next section, between citizens and noncitizens was substantively small and did not achieve standard levels of statistical significance (p = .89 [two-tailed test]).

24. Subjects were debriefed afterward that the interaction with the research assistant was part of the study design.

25. In table C2, we also report demographic characteristics of respondents. As shown in the appendix, while most respondents were college students, the sample was diverse with respect to gender and religiosity. In sum, 88% of respondents were US citizens.

26. We did not ask an explicit post-treatment question about feelings of victimization or racial identity because this would prime respondents and inflate treatment effects.

27. To reduce skewness in the completion time variable, we took its natural log. The p-value from the test examining the difference-in-difference in the nonlogged measure was p = .03.

28. The variables closed-minded, ignorant, represents interests, and net likes are all strongly and significantly (p < .001) positively correlated with the feeling thermometer and party identification measures.

References