Political Behavior

, Volume 36, Issue 1, pp 143–163

Decoding Prejudice Toward Hispanics: Group Cues and Public Reactions to Threatening Immigrant Behavior

Authors

    • Department of Government and Justice StudiesAppalachian State University
  • Benjamin J. Newman
    • Department of Political ScienceUniversity of Connecticut
  • C. Scott Bell
    • Department of Political ScienceFlorida State University
Original Paper

DOI: 10.1007/s11109-013-9231-7

Cite this article as:
Hartman, T.K., Newman, B.J. & Scott Bell, C. Polit Behav (2014) 36: 143. doi:10.1007/s11109-013-9231-7

Abstract

Consistent with theories of modern racism, we argue that white, non-Hispanic Americans have adopted a “coded,” race-neutral means of expressing prejudice toward Hispanic immigrants by citing specific behaviors that are deemed inappropriate—either because they are illegal or threatening in an economic or cultural manner. We present data from a series of nationally representative, survey-embedded experiments to tease out the distinct role that anti-Hispanic prejudice plays in shaping public opinion on immigration. Our results show that white Americans take significantly greater offense to transgressions such as being in the country illegally, “working under the table,” and rejecting symbols of American identity, when the perpetrating immigrant is Hispanic rather than White (or unspecified). In addition, we demonstrate that these ethnicity-based group differences in public reactions shape support for restrictive immigration policies. The findings from this article belie the claim of non-prejudice and race-neutrality avowed by many opponents of immigration.

Keywords

ImmigrationPrejudiceHispanicExperiment

One key question that arises from debates about immigration in the U.S. is the degree to which restrictive immigration policies, and public support for them, are motivated by prejudice toward Hispanics. American lawmakers and citizens who support restrictive policies claim that ethnic prejudice, which is defined as antipathy toward a particular group or its members (Allport 1954; Brown 2010; Stangor 2009),1 plays no role in their anti-immigration positions. Instead, these individuals argue that they have legitimate concerns about the economic and cultural consequences of unfettered immigration, as well as a desire to impose sanctions on immigrants for law-violating behaviors. In other words, the key issue is the immigrants’ transgressive behaviors irrespective of their racial or ethnic identity. For example, Roy Beck, the Executive Director of Numbers USA, an immigration reduction organization, argues that “the chief difficulties that America faces because of current immigration are not triggered by who the immigrants are” or “some vision of a homogeneous white America.”2 Echoing these sentiments, Alabama state lawmakers recently defended one of the country’s most strident immigration laws, HB 56, by publicly proclaiming that they “don’t have a problem with Hispanics as people”3 but with “illegal immigrants entering the state and taking jobs away from the people of Alabama.”4

Despite these reassurances, immigration advocates argue that these anti-immigration efforts are motivated in part by bigotry and represent a veiled attack on the larger Hispanic population. For instance, Isabel Rubio, executive director of the Hispanic Interest Coalition of Alabama, contended that “at its core HB 56 is aimed at the Latino community, not the entire immigrant community.”5 Indeed, the claim that opposition to immigration is fueled by prejudice toward Hispanics is supported by opinion research demonstrating that negative stereotypes and affect toward Hispanics serve as significant predictors of support for restrictive immigration policies among the American public (Burns and Gimpel 2000; Citrin et al. 1997; Hood and Morris 1997; Valentino et al. 2013). Further, recent research demonstrates that Americans associate negativity with Hispanic relative to Anglo immigrants, and this implicit bias predicts anti-immigration policy preferences (Perez 2010). In sum, while many opponents of immigration claim that their aversion centers upon the behavior of immigrants and not their ethnic identity, critics on the left of the debate, along with findings from the opinion research, suggest that it is by and large the Hispanic identity of immigrants—not their behavior—that matters most in driving public opposition to immigration.

In this article, we test whether Hispanic ethnic identity plays a role in shaping public reactions to threatening and law-violating behaviors, as well as influencing support for restrictive immigration policies. In addition to disclosing a potential ethnicity-based bias in mass opinion on immigration, we engage our topic with the larger question in mind of whether the expression of prejudice toward Hispanics has gone “underground” and become “coded” in racially or ethnically neutral terms in the contemporary U.S. In line with theories of modern racism, which have long been used to study white-black relations and anti-black prejudice, we argue that individuals understand that openly derogating minorities is socially unacceptable (Crandall et al. 2002), as it contradicts core American norms of equality and tolerance (McClosky and Zaller 1984; Sullivan et al. 1982), as well as the narrative that “America is a nation of immigrants.” Instead, the expression of anti-Hispanic sentiment among white Americans6 has been funneled into the citation of specific behaviors that are deemed inappropriate, either because they are formally illegal or economically and culturally threatening.

We present data from a series of nationally representative, survey-embedded experiments to demonstrate that white respondents do not treat threatening immigrant behavior equally with respect to different immigrant groups. Instead, transgressions such as remaining in the country without legal documentation, working without paying taxes, and failing to support traditional symbols of American culture and identity, are considered more offensive if committed by Hispanic than non-Hispanic immigrants. In addition to addressing a timely issue in American politics, these findings make an important contribution to the opinion research on immigration. Prior demonstrations of the importance of anti-Hispanic sentiment in shaping opinion on immigration are largely correlational in nature (e.g., Burns and Gimpel 2000; Citrin et al. 1997; Hood and Morris 1997) and reliant upon explicit measures of prejudice that likely underestimate true levels due to social desirability concerns (e.g., Devine et al. 2002; Piston 2010). Our experimental design and analyses extend this literature by demonstrating the causal role of an immigrant’s ethnic identity in shaping public opposition to immigration in a manner not reliant upon self-reported prejudice toward Hispanics. Taken together, our findings demonstrate that the ethnic identity of an immigrant influences how Americans react to transgressive behaviors by immigrants, and that these ethnicity-based group differences in public reactions shape support for restrictive immigration policies. Ultimately, our findings suggest that the focus within popular political discourse on the “illegality” and “threats” of immigrants may indeed serve as a coded means of expressing antipathy toward specific immigrant minorities.

Hispanic Immigrants and Modern Racism Revisited

At more than 50 million people, Hispanics have supplanted African Americans as the largest, and arguably the most salient, minority group in the United States. While white-black relations in the U.S. provided the predominant intergroup context in which to study prejudice for the better part of the 20th century, substantial and persistent influxes of Latin American immigrants have brought Hispanics to the forefront of the study of intergroup relations in American politics in the 21st century. While in no way diminishing the importance of the study of prejudice toward African Americans or denying its continued existence, white Americans’ attitudes toward Hispanics undeniably constitute an emerging and increasingly important arena in which to study the dynamics of prejudice in contemporary American society. And, just as the study of anti-black prejudice is intricately entangled with the analysis of public opinion within the social welfare and affirmative action policy domains (Gilens 1999; Kinder and Sanders 1996), the study of anti-Hispanic prejudice is, and will likely remain, closely tied to the analysis of public opinion on immigration policy. We aim to apply the lessons learned from the study of prejudice toward Blacks in guiding our thinking about the operation of white Americans’ prejudice toward Hispanics, specifically as it pertains to preference formation and public discourse on immigration. In so doing, we turn to theories of modern racism and their tenets as the basis for developing our hypotheses regarding the nature and operation of anti-Hispanic prejudice within the U.S.

Modern racism is an umbrella term for a set of theories addressing a “new” form of racism in the U.S. that emerged following the Civil Rights Movement. Known as “modern racism” (McConahay 1986), “symbolic racism” (Kinder and Sears 1981; Sears 1988), “subtle racism” (Pettigrew and Meertens 1995), “racial resentment” (Kinder and Sanders 1996), and “aversive racism” (Gaertner and Dovidio 1986), these theories, while distinct, all contain the same basic tenets that: (1) white Americans recognize changes in societal norms since the 1960s, which make it unacceptable to freely express “old fashioned” forms of racial prejudice based upon open bigotry toward Blacks; and (2) as a result, a subtle, covert form of racism has emerged to supplant earlier, overt expressions of prejudice against African Americans. In essence, scholars of modern racism argue that mass racial antipathy among Whites persists, but that it has found more inconspicuous and covert avenues of expression in the form of opposition to social welfare and liberal racial policies, as well as the belief that Blacks violate traditional American values such as individualism, self-reliance, and the Protestant work ethic.

The key insight of the modern racism perspective is that prejudice became “coded” (Hurwitz and Peffley 2005; Mendelberg 2001; Valentino et al. 2002), whereby antipathy toward Blacks among elites and the masses was reconstituted as a defensibly race-neutral and “principled” political opposition to policies intended to advance the interests of black Americans. Thus, rather than appealing to explicitly racist considerations, this new coded language of opposition to pro-black policies was founded upon seemingly non-racial concerns over the rate of societal change, the fairness of special treatment by government to specific groups, and the status of traditional values such as hard-work and merit-based rewards. While the modern racism perspective is not without its critics or shortcomings (e.g., see Sniderman and Carmines 1997), there is a substantial body of evidence in support of its claims (e.g., see Sears 1988). Moreover, it provides a useful framework for theorizing the operation of prejudice toward Hispanics in the contemporary debates over immigration policy in the U.S. In adapting modern racism theory from a white-black to a white-Hispanic intergroup context, the central prediction would thus be that Whites’ antipathy toward Hispanic immigrants would be expressed in a covert manner—coded by the usage of language and reference to concerns that are not explicitly racial.

To determine the content of this hypothesized coded language, we need only turn to the opinion research and popular political discourse on immigration in the U.S., both of which strongly suggest that the most common types of concerns voiced by anti-immigrant politicians, media pundits, interest groups, and citizens alike pertain to (1) the law-violating behaviors of immigrants who enter or remain in the country illegally, (2) the threats posed by immigrants to material resources, and (3) the threats posed by immigrants to American culture (Beck 1996; Chavez 2001, 2008; Chomsky 2007; Citrin et al. 1997; Cohen 2001; Cornelius and Rosenblum 2005; Huntington 2004; Paxton and Mughan 2006; Simon and Alexander 1993). According to a modern racism perspective, these three broad types of concerns may serve as the coded vehicles through which mass anti-Hispanic sentiment is expressed, given that these concerns are explicitly nonracial and thus defensibly devoid of overt prejudicial content. Thus, our primary hypothesis is that a portion of Whites’ opposition to immigration in the contemporary U.S. is at least partially rooted in anti-Hispanic prejudice, but that this prejudice is concealed as overt concern over law-violating behavior, as well as economic and cultural threats posed by immigrants. We label this the coded prejudice hypothesis.

Survey-Embedded Experiments

One way in which public opinion scholars have assessed prejudice toward minority group members is with survey-embedded experiments. Largely established by the 1991 National Race and Politics Survey,7 this approach involves asking respondents to report their reactions to norm- or value-violating behaviors committed by an individual whose racial identity is manipulated across randomly assigned experimental groups. For instance, in one experiment, subjects were asked to report their level of anger in response to either “a man” or “a black man” who “collects welfare because he is too lazy to get a job.” This type of experimental group cue manipulation was developed to reveal anti-black sentiment by enabling researchers to observe whether respondents report more negativity when the perpetrator of a norm-violating behavior is black rather than of an undefined racial identity. Within this general experimental paradigm, prejudice is revealed through differential reactions to hypothetical behaviors because these behaviors are held constant and only the racial identity of the actor is varied.

To test our coded prejudice hypothesis, we drew upon this basic experimental paradigm as the foundation for designing four survey-embedded experiments. Each of these experiments was crafted to present respondents with an immigrant who is engaging in a law-violating or threatening behavior, and the primary manipulation concerns whether the perpetrating immigrant is ethnically Hispanic or some other ethnicity. For ease of discussion, and to be conceptually consistent with extant theoretical frameworks in the intergroup conflict and immigration opinion literatures, we classify these violations, or “transgressions,” in terms of realistic and symbolic threats posed by immigrants to American society (Newman et al. 2012; Sides and Citrin 2007; Sniderman et al. 2004; Stephan et al. 2009). Realistic threats focus on concern over material resources such as jobs, wages, or taxes (Citrin et al. 1997), while symbolic threats concern violations committed by an outgroup to an ingroup’s core set of values, symbols, or cultural norms (Kinder and Sears 1981; Sears and Kinder 1985). Both types of transgressions can be actual or imagined (Cornelius and Rosenblum 2005), and it is perceived violations that often lead to antipathy toward immigrants (Sniderman et al. 2004) and support for restrictive immigration policies (Citrin et al. 1990, 1997; Hood and Morris 1997). Prior experimental research has demonstrated that information about the costs and benefits of immigration influences policy preferences and political action differentially depending upon the national origins of the immigrants in question (Brader et al. 2008). However, our work is the first to date that explores possible variation in how citizens evaluate a range of transgressive immigrant behaviors as a function of the ethnic identity of the perpetrating immigrant. In so doing, our work holds the promise of disclosing how public condemnations of the putative transgressions committed by immigrants may contain significant ethnicity-based biases and thus serve as a coded vehicle for the expression of antipathy toward Hispanics.

Sample

A total of 275 white, non-Hispanic adults participated in our nationally representative, random-digit-dial (RDD) telephone survey conducted by the Appalachian Survey Research Laboratory from March 22nd to June 15th, 2010.8 While modest in size, our sample is large enough to provide sufficient power to test for mean differences in responses across experimental groups. We also note that our sample of white Americans compares quite well on many key demographic variables to other nationally-recognized RDD samples. For comparison, we have included demographic information for white respondents from the 2006 Pew Research Center Poll on Immigration and the 2008–2009 American National Election Panel Study (see Table 1).9 Our sample contains slightly more females (57 %) than males with ages ranging from 18 to 96 years old (median age = 55 years old). The median household income range of the sample is $60,000–$80,000, and 43 % of respondents have earned a 4-year college degree. We find that 38 % of respondents identified themselves as Republicans, 26 % as Independents, 29 % as Democrats, and 7 % as other.10 Ideologically, the sample consisted of 37 % conservatives, 42 % moderates, and 21 % liberals.
Table 1

Sample comparisons by key demographic variables

 

Decoding prejudice, 2009

Pew research poll: immigration, 2006

ANES panel study, 2008–2009

Female

57.0 %

52.4 %

56.1 %

Age (median years)

55.0

51.0

52.0

College graduate

42.9 %

39.1 %

43.1 %

Income (median category)

$60,000 to $79,999

$50,000 to $74,999

$60,00 to $74,999

Employed

57.5 %

64.1 %

63.3 %

Democrat

29.1 %

27.3 %

31.0 %

Independent

25.8 %

32.0 %

31.7 %

Republican

37.8 %

35.4 %

37.3 %

Other/no party affiliation

7.3 %

5.2 %

N

275

1409

3292

The “Pew Research Center Poll: Immigration,” sponsored by the Pew Research Center for the People & the Press and the Pew Hispanic Center, was conducted between February 8th and March 7th, 2006. The 2008–2009 American National Election Panel Study recruited respondents via telephone to participate in online surveys from January 2008 and September 2009. Summary statistics are from white respondents only. Percentages do not add to 100 due to rounding and missing values. Whenever possible, derived variables were used in the ANES sample

The “Overstay Visa” Experiment

The objective of the first experiment in our survey was to test the coded prejudice hypothesis in reference to arguably the most pervasive issue within popular political debate over immigration, namely undocumented, or “illegal,” immigration. The central argument that some political elites and Americans have made to explain their anti-immigrant sentiment and policy positions is that many immigrants violate the law by illegally entering and residing within the U.S. The underlying race-neutral logic of this argument is that opposition to immigration exists in response to the transgressive behavior of illegal entry and residence, irrespective of the ethnic identity of the immigrants committing such violations. If this contention were true, then it should make no difference whether the immigrant is Hispanic, European, or of some other origin; instead, the illicit activity should be paramount in driving opposition.

To test whether this argument truly holds, or if there is indeed an ethnicity-based group bias underlying it, we designed a simple experiment, which we call the “Overstay Visa” Experiment. In this experiment, we randomly assigned survey respondents to one of three hypothetical scenarios in which we varied the ethnic origins of an undocumented immigrant. The exact wording of this manipulation read as follows: “Suppose someone from (Mexico/Britain) enters the U.S. with a short-term visa but then stays in this country longer than legally authorized. In your view, how serious is this offense on a scale from 0 to 10, where ‘0’ means ‘not at all’ and ‘10’ means ‘very serious’?”11 We chose this form of illegality because it subtly identifies the immigrant’s legal status without referring to them explicitly as an illegal immigrant. Further, we chose Mexico as the specific Hispanic origin for our experiment because individuals from Mexico comprise the majority of the legal and undocumented Hispanic immigrant populations within the U.S. (Dockterman 2011; Passel 2006; Passel and Cohn 2011). In addition to the Mexican versus British group cues, we also included a control condition, which simply referred to “someone” without identifying any particular country of origin. We decided to include an ethnically undefined immigrant to serve as a true control condition to compare responses to the immigrants from Mexico and Britain (e.g., see Gaines et al. 2007). If prejudice is not a factor in determining Whites’ attitudes, then respondents should not judge unauthorized immigrants from Mexico any differently than they do those from the United Kingdom (or those without a defined country of origin). If, on the other hand, respondents do differentiate by nationality, such that they view transgressions by Hispanics as being more offensive than those of an undocumented immigrant from Britain or an unnamed country, then we can assume that negative group affect toward Hispanics is at work.

The “Under the Table” Experiment

The goal of the second experiment was to move beyond the former by confronting participants with a particularly egregious realistic violation and test for the presence of group-based differentiation in respondents’ reactions. In our “Under the Table” Experiment, we present respondents with a scenario in which the same undocumented immigrant from the prior experiment is now failing to pay taxes on earned income: “Now suppose this person with an expired visa (from Mexico/Britain) is working ‘under the table’ and does not pay taxes on this income.” Failing to pay taxes not only violates the law, but also norms of fairness. As such, this particular transgression should serve as an important argument to justify opposition to immigration, since people commonly perceive that undocumented immigrants use more services (e.g., healthcare, education, etc.) than they pay for in their share of taxes. As in the Overstay Visa Experiment, this experiment also included a neutral control condition where respondents are asked to react to an ethnically undefined immigrant committing the transgression of working illegally and not paying taxes. For consistency, assignments to experimental group cue conditions for this experiment were the same as the Overstay Visa Experiment.12

The “Foreign Flag” Experiment

The previous two experiments were designed to test whether white Americans would deem realistic transgressions as more flagrant offenses when committed by Hispanic rather than non-Hispanic immigrants. Following these initial realistic experiments, our survey shifted in focus to symbolic, culturally-oriented transgressive behaviors. In contrast to arousing legal or material concerns, these behaviors center upon the perceived failure of an immigrant to assimilate to American society, which involves threats to important symbols of American identity and culture. Once again, paying attention to symbolic violations is important, given that a good deal of opinion research on immigration has found that cultural concerns often matter more than economic concerns in shaping public opposition to immigration (Citrin et al. 1990; Hood and Morris 1997; Sides and Citrin 2007; Sniderman et al. 2004). Thus, pointing to cultural grievances related to immigration may serve as a particularly available and coded way to express antipathy toward some ethnic groups over others.

In the “Foreign Flag” Experiment, we focus on the symbolic transgression of displaying a national flag from one’s home country rather than the American flag. National flags are potent symbols of cultural identity (e.g., see Schatz and Lavine 2007), and symbolic behaviors involving the use of flags by immigrants have been documented to resonate with American elites and the masses. For example, in defending his argument that Hispanic immigration poses an unprecedented threat to American culture, Huntington (2004) recounts how Mexican immigrants protesting California’s Proposition 187 marched through the streets of Los Angeles waving scores of Mexican flags while holding American flags upside down. Our assumption is that displaying the flag of another country should be interpreted as a significant symbolic violation in that these individuals will be seen as not taking pride in the U.S. For the purposes of this experiment, respondents were presented with the following scenario: “Suppose someone in your neighborhood chose not to display the American flag but instead displayed the national flag of (Mexico/Canada) in their front yard.” Respondents were randomly assigned to either a Mexico or Canada group cue condition,13 and were then asked to report how offended they would be in response to this behavior on a scale of 0 (“not at all offended”) to 10 (“very offended”).

The “Foreign Team” Experiment

In addition to displaying a foreign flag, rooting for a foreign team in a sporting event may serve as a salient symbolic transgression in the eyes of many American citizens. Once again, Huntington (2004) provides anecdotal evidence by discussing how Mexican Americans booed the U.S. national anthem and assaulted American players during a U.S.-Mexico soccer match in Los Angles in 1998. According to Huntington, this transgressive behavior indicates a dramatic rejection of American identity (and assertion of Mexican identity), which is an integral feature of the supposed cultural threat posed by Hispanic immigrants to American society. To simulate this cultural violation, we created the “Foreign Team” Experiment: “Suppose that you are watching the Olympic Games. How offended would you be if someone were cheering for Team (Mexico/Canada) to win a gold medal instead Team USA?” We opted for an Olympic match in part because our survey was conducted only a few weeks after the 2010 Winter Olympic Games at a time when national sports pride should have been relatively salient. Respondents were randomly assigned to either the Mexico or Canada condition for this fourth and final experiment.

Results

The effects of our group-cue experimental manipulations across our realistic and symbolic experiments are depicted in Fig. 1 as mean differences across conditions in seriousness ratings for each offense. Above the bars, we provide regression coefficients and associated significance levels based upon regressions of the perceived seriousness of each transgressive behavior on the relevant experimental treatment dummy variables (the Hispanic group cue always served as the excluded category). We begin with the results from the Overstay Visa Experiment, which are presented in the leftmost portion of Fig. 1. Clearly, respondents viewed residing in the U.S. without legal documentation as a major transgression, as the mean response for all three conditions was well above the scale midpoint. However, consistent with our coded prejudice hypothesis, respondents differentiated by the immigrant’s ethnicity, such that they rated the Hispanic immigrant as committing a more serious offense than the other immigrants: The mean seriousness ratings for Mexican immigrants was .76, while for British and unspecified immigrants it was .67 and .68, respectively. Regression results confirm that the perceived offense was significantly less in the British (B = −.09, SE = .04, p < .05) and the undefined conditions (B = −.09, SE = .04, p < .05) relative to those who received the Mexican group cue. It is also worth noting that seriousness ratings did not significantly differ when the hypothetical immigrant hailed from Britain rather than from an unspecified country (B = −.01, SE = .04, p = .88).
https://static-content.springer.com/image/art%3A10.1007%2Fs11109-013-9231-7/MediaObjects/11109_2013_9231_Fig1_HTML.gif
Fig. 1

Perceived seriousness of each violation by experimental condition. Bars represent the mean levels of perceived seriousness of each offense, and entries above the bars reflect the mean difference between the Mexican group cue treatment and the non-Hispanic or control conditions (i.e., entries are unstandardized regression coefficients). *p < .05, **p < .01, ***p < .001 level. Significance levels are based upon two-tailed hypothesis tests

The results from this first experiment demonstrate that white Americans, while holding true to a general opposition to law-violating immigrant behavior, nevertheless do react differently to the transgression of being in the country illegally depending on the ethnicity of the perpetrating immigrant. To build on this initial result and test whether respondents punished Hispanics who work illegally and fail to pay income taxes more than other immigrants committing the same offense, we turn attention to the mean offense ratings by experimental condition for the Under the Table Experiment. Figure 1 reveals that respondents deemed this type of offense as the worst transgression (of the four experiments); yet, despite having such a high level for the perceived offense across the three conditions, we nonetheless observe respondents differentiating among immigrants by group membership. The mean seriousness ratings for each of the experimental conditions are .88 for Hispanic immigrants, .84 for British immigrants, and .80 for unspecified immigrants. Respondents deemed the offense of working illegally and not paying taxes significantly less serious when perpetrated by an undefined immigrants (i.e., “someone”) than a Mexican immigrant (B = −.08, SE = .03, p < .05). Additionally, we find directional support for our coded prejudice hypothesis as the immigrant working under the table from Britain was rated less harshly than the Mexican immigrant doing the same, however, this effect failed to attain conventional levels of statistical significance (B = −.04, SE = .03, p = .20). Once again, white respondents did not make any distinction between British versus generic undocumented immigrants (B = .03, SE = .03, p = .24).

Next, we turn to the results from our symbolic threat experiments. The results from our Foreign Flag Experiment, which are presented in the third portion of Fig. 1, reveal that respondents viewed this type of transgression as a less serious offense in general than the earlier realistic transgressions; however, participants again indicated that they were significantly more offended when a Hispanic immigrant committed this type of cultural offense: The mean seriousness rating for Hispanic immigrants is .54 compared to only .32 for Canadian immigrants (B = −.22, SE = .05, p < .001), which corresponds to a 68 % increase in seriousness ratings for Hispanics. We find a similar pattern of results for the Foreign Team Experiment, which are presented in the rightmost portion of Fig. 1. As hypothesized, respondents were significantly more offended when someone cheers for the Mexican national team compared to the Canadian team: The mean response to this type of symbolic transgression is .20 for Hispanic immigrants but only .11 for Canadian immigrants (B = −.09, SE = .04, p < .05). This finding is in line with the results from the prior symbolic offense experiment, as well as those from the two preceding realistic experiments, in demonstrating that white Americans are significantly more affronted by transgressive behavior when committed by a Hispanic rather than stereotypically white immigrant.

What is important to reiterate is that if Whites were concerned about transgressive or threatening immigrant behavior in a manner devoid of ethnic prejudice, then we should not have found significant differences in perceived offensiveness across experimental conditions. Our results show otherwise. We find significant and persistent differences in offense perceptions revealing a consistent bias against Hispanic immigrants.

Alternative Hypotheses

The experimental paradigm we drew upon in designing our survey experiments suggests the differences in offense perceptions observed across experimental conditions, because the behavior of the immigrants in each experiment is held constant, are solely attributable to a change in the ethnicity of the immigrant committing the transgression. Although we believe that the results from our experiments demonstrate prejudice against Hispanics by clearly depicting differential reactions toward individuals due to their group membership, we acknowledge that there may be alternative explanations for the observed results that deserve some discussion. For instance, it is possible that our survey respondents reacted more negatively to the Mexican immigrant than to other immigrants because of actual or perceived differences in their populations, rates of illegal entry, skill levels, or any other characteristics associated with one immigrant group relative to the others. In other words, what we have labeled as anti-Hispanic prejudice could simply reflect genuine concerns about Hispanics’ higher rates of illegal entry into the U.S. and subsequently larger undocumented population, as well as their lower socioeconomic status relative to British or Canadian immigrants. Thus, one alternative hypothesis is that respondents could have been more concerned about the prevalence of the transgressive behaviors rather than the specific ethnic identity of the offending immigrant. A skeptic could argue that participants may have been willing to overlook what they perceived to be an isolated incident, but they would have had more difficulty ignoring widespread disregard for the law.

A second and related alternative hypothesis is that some respondents in our realistic experiments may have felt economically threatened by a large influx of low-skilled workers with whom they might be in direct competition (Scheve and Slaughter 2001). This economic threat hypothesis has some face validity given the negativity of recent media coverage, in which Hispanic immigrants have been described as posing threats to labor market competition, increasing the consumption of public services, and heightening the tax burden on Americans (Brader et al. 2008; Chavez 2001; Valentino et al. 2013). Indeed, this economic competition explanation is consistent with supporters of Alabama’s HB 56, who have claimed that they are simply interested in “protecting American jobs.”

To address these two alternative hypotheses, we first remind readers that both of our realistic experiments included an ethnically undefined immigrant to serve as a true control condition. Note that this ethnically undefined cue simply referred to “someone” committing a set of transgressive behaviors, rather than the more loaded term “immigrant,” which is likely to be associated with Hispanics (Valentino et al. 2013). Moreover, the generic individual in our hypothetical scenarios should not have conveyed any systematic information about the offending immigrant’s group size, rate of illegal entry, skill level, or any other potentially important attributes. Recall that we discovered that our control condition operated exactly as we had expected: Respondents reacted more negatively to the Hispanic immigrant compared to the non-Hispanic immigrant, whether the immigrant was specifically identified as British or simply as “someone.”

To further dispel these alternative hypotheses, we conducted additional analyses in which we tested whether respondents’ perceptions about the size of the illegal immigrant population that is of Hispanic origin, as well as respondents’ education and income levels, moderated the effects of our group cue manipulations. If concerns about the size of the undocumented Hispanic immigrant population or economic threat were driving our results in the realistic, and to a lesser extent, symbolic experiments, then we would expect individuals who deem illegal immigration as a particularly prevalent issue or are most vulnerable to the economic threat of labor market competition with low-skilled workers to react most negatively to the transgressive behaviors of the Mexican immigrant. The results of our moderated regressions, in which we interacted a treatment dummy variable14 with the perceived proportion of the Hispanic population that is undocumented, education, and income,15 are presented in Table 2. Across our four experiments, we find no evidence that any of these alternative factors moderate the effect of being told a Mexican immigrant is perpetrating each transgression on reported seriousness ratings. In other words, our main treatment effects are not conditional upon individual differences in beliefs about the pervasiveness of Hispanic illegal immigration, education (and likely skill levels), or income. These null findings are consistent with recent research by Valentino et al. (2013), who find no evidence that economic vulnerability affects negative beliefs about immigration, as well as others who show that economic self-interest factors little into opinion on immigration (Citrin et al. 1997; Hainmueller and Hiscox 2010).16
Table 2

Testing alternative explanations: moderated regression models of group cues on seriousness ratings

 

Realistic transgressions

Symbolic transgressions

Overstay visa

Work under table

Display foreign flag

Cheer for foreign team

Treatment

 Hispanic group cue

.06

(.11)

−.00

(.09)

−.02

(.16)

−.06

(.12)

Moderators

 % Hispanic illegal

.25**

(.08)

.10

(.06)

.37**

(.12)

.21*

(.09)

 Education

−.13†

(.07)

−.17**

(.06)

−.41***

(.11)

−.29***

(.08)

 Income

−.06

(.09)

−.09

(.08)

.05

(.13)

−.09

(.10)

Interactions

 Hispanic cue × % Hispanic illegal

.09

(.13)

.08

(.10)

−.02

(.19)

−.04

(.15)

 Hispanic cue × education

.04

(.13)

.00

(.10)

.09

(.18)

.20

(.14)

 Hispanic cue × income

−.06

(.14)

.07

(.11)

.01

(.21)

−.08

(.16)

Constant

.67***

(.06)

.92***

(.05)

.49***

 

.29***

(.07)

N

237

239

239

240

.13

.13

.13

.11

Entries are unstandardized regression coefficients from ordinary least squares regression. Significance levels are based upon two-tailed hypothesis tests

 p < .10; * p < .05; ** p < 01; *** p < .001 level

Indirect Effects of Group Cue Treatments on Policy Preferences

In addition to identifying a systematic bias against Hispanics in public reactions to transgressive immigrant behavior, we are also interested in the larger political ramifications of how this “hidden” bias toward Hispanics influences support for restrictive immigration policies. To this end, we estimate a series of path models to test the indirect effect of the Mexico treatment on two key immigration policy items through its impact on perceived violation seriousness. The first immigration policy, labeled Deport Illegals, concerns attitudes toward undocumented immigrants already residing within the U.S.: “What do you think should happen to undocumented immigrants who have lived and worked in the United States for at least two years: Should they be given a chance to keep their jobs and eventually apply for legal status, or should they be deported back to their native country?” Responses to this question have been coded so that a value of 1 means “deported,” while 0 means “apply for legal status.” The other policy item, labeled Border Fence, concerns preventing undocumented immigration from crossing the border: “Do you support or oppose building a fence along the U.S.-Mexico border to prevent people from illegally entering the country?” Responses to this 4-point item ranged from “strongly support” to “strongly oppose,” where higher scores indicate a preference for restricting the border. For our mediational analyses, we used a dichotomous group cue treatment variable, which is coded “1” for those receiving the Hispanic group cue and “0” for those receiving the non-Hispanic group cue.17

We present the results of our mediational analyses18 in Fig. 2 (for full results see Table 3), where the upper panel displays the path models for the deportation dependent variable, and the lower panel shows the path models affecting preferences for a U.S.-Mexico border fence. The leftmost portion of each panel redisplays the significant direct effects of the group cue treatments on the violation seriousness ratings, which now serve as mediators in the path analyses. Next, the rightmost portion of each panel demonstrates that each perceived offense mediator significantly and substantially increases support for restrictive immigration policies. Last, the indirect effects, or “causal mediation effects” (Imai et al. 2011; see also Múthen 2011), are italicized and displayed above or below each violation mediator. The indirect effects can be interpreted as the change in the probability of supporting a restrictive immigration policy corresponding to a change in the value of the perceived offence mediator produced by moving from the control to treatment conditions. The results reveal that the causal mediation effects are significant in seven of the eight path models (i.e., p < .10), and on average, the indirect effects amount to a .07 increase in the probability of support for restrictive immigration policies across our path models.
https://static-content.springer.com/image/art%3A10.1007%2Fs11109-013-9231-7/MediaObjects/11109_2013_9231_Fig2_HTML.gif
Fig. 2

Indirect effects of group cue treatments on restrictive immigration policy preferences. NotesN ~ 275. Entries are the direct and indirect effects from eight path models estimated using Imai et al.’s (2011) mediation package in R. The direct effects for (1) the binary treatments on the continuous mediators are OLS coefficients, and (2) the mediators on the categorical outcome variables are probit or ordered probit coefficients. The indirect effects (listed above the mediators) are the change in the probability of supporting a restrictive immigration policy corresponding to a change in the value of the perceived offence mediator produced by moving from the control to treatment conditions. † p < .10, *p < .05, **p < .01, ***p < .001 level. Significance levels are based upon two-tailed hypothesis tests

Table 3

Mediated effects of experimental treatments on policy attitudes

 

Effect on mediator

Effect on policy attitudes

Violation seriousness

Deport illegals

Border fence

I. Overstay Visa Experiment

Group cue treatment

.09**

(.03)

.03

(.18)

−.18

(.14)

Violation seriousness

  

2.36***

(.35)

2.32***

(.28)

 Mediated effects of group cue treatment

  Total effect

  

.09

 

.02

 

  Indirect effect

  

.07**

 

.07*

 

  Prop. total effect mediated

  

.87

 

4.68

 

II. Under the Table Experiment

Group cue treatment

.06*

(.03)

.07

(.17)

−.12

(.14)

Violation seriousness

  

2.40***

(.44)

2.25***

(.33)

 Mediated effects of group cue treatment

  Total effect

  

.09

.01

  

  Indirect effect

  

.06**

.05*

  

  Prop. total effect mediated

  

.69

.82

  

III. Foreign Flag Experiment

Group cue treatment

.22***

(.05)

−.22

(.17)

.05

(.14)

Violation seriousness

  

1.32***

(.22)

1.05***

(.18)

 Mediated effects of group cue treatment

  Total effect

  

.02

.10

  

  Indirect effect

  

.10***

.08*

  

  Prop. total effect mediated

  

4.21

.78

  

IV. Foreign Team Experiment

Group cue treatment

.09*

(.04)

.01

(.16)

−.00

(.13)

Violation seriousness

  

.84**

(.28)

1.09***

(.24)

 Mediated effects of group cue treatment

  Total effect

  

.03

.04

  

  Indirect effect

  

.03

.04†

  

  Prop. total effect mediated

  

.93

1.00

  

N ~ 275. Entries are the direct and indirect effects from 8 path models estimated using Imai et al.’s (2011) mediation package in R. The direct effects for (1) the binary treatments on the continuous mediators are OLS coefficients, and (2) the mediators on the categorical outcome variables are probit or ordered probit coefficients. The indirect effects are the change in the probability of supporting a restrictive immigration policy corresponding to a change in the value of the perceived offence mediator produced by moving from the control to treatment conditions. Significance levels are based upon two-tailed hypothesis tests

p < .10; * p < .05; ** p < 01; *** p < .001 level

These significant causal indirect effects reveal that receiving the Hispanic group cue increased support for restrictive immigration policies by significantly heightening the perceived seriousness of each realistic and symbolic transgressive behavior. That is, respondents in the Hispanic group cue condition deemed overstaying one’s visa, working under the table, displaying a foreign flag, and cheering for a foreign team as a more egregious violation than those in the non-Hispanic conditions, and these heightened violation perceptions in turn directly increased support for two distinct restrictive immigration policies. In essence, the results from these mediational analyses demonstrate that the disclosed ethnicity-based group bias in public reactions to transgressive immigrant behaviors has meaningful policy effects. Specifically, the decoded amount of antipathy toward immigrants of Hispanic origin compared to other immigrants significantly feeds into support for restrictive immigration policies. If support for restrictive immigration policies were truly based upon concern over law-violating behaviors, and if this concern lacked any concealed ethnic prejudice, then the only significant path observed in our figure would be the one linking perceived seriousness of offense to policy preferences. Our data reveal that this is simply not the case.

General Discussion

As we delve further into the 21st century, ethnic change and increasing diversity will undoubtedly serve as potent forces shaping the social and political scene in the U.S. Yearly influxes of immigrants from Latin America, along with high birth rates among Hispanic households, will continue to place Hispanics at the center of political debates and conflict over issues of immigration, multiculturalism, and other policy areas associated with racial and ethnic minorities, such as affirmative action and social welfare. To date, debates about immigration policy have assumed an important position alongside other social and cultural issues in defining the major division between the political left and right in the American mass public. The critical question lurking underneath these debates about immigration in contemporary American politics is the role of prejudice as a contributing factor to this political polarization.

In turning to public political discourse for an answer, one is left with a stalemate. Opponents of immigration defend themselves against accusations of racism with non-racial rationalizations for their preferred policies. These rationalizations center upon concerns over law enforcement, economic opportunity and well-being, and the protection of cherished cultural norms and institutions. Pro-immigration groups and citizens on the left of the issue retort that these justifications merely serve as a veil for ethnically-motivated attacks on Hispanic populations throughout the country. Scholarly research on public opinion toward immigration generally supports the latter’s claims, as prejudice and negative stereotypes toward Hispanics bolsters support for restrictive immigration policies. That being said, intergroup threat research also finds that concern over the economic and cultural impacts of immigration serve as powerful sources of support for restrictive policies, even when controlling for prejudice toward Hispanics (e.g., Citrin et al. 1997). Yet, one limitation of much of this research is that it is largely correlational in nature. Ultimately, popular and academic treatments of the sources of public support for anti-immigrant policies leaves an uneasy degree of uncertainty concerning the role of prejudice in shaping opposition to immigration.

In this article, we conducted a series of survey-embedded experiments to tease out the distinct role of prejudice in shaping public opinion on immigration. Across four separate experiments, our analyses demonstrate that while the American public does care about immigrants’ transgressive behaviors, they also significantly distinguish between the immigrants who are engaging in these behaviors. Consistent across our studies, we find that white Americans take significantly greater offense to transgressions like being in the country without legal documentation, working illegally and not paying taxes, and rejecting symbols of American culture and identity, when the perpetrating immigrant is Hispanic rather non-Hispanic. One major implication of the findings from our studies is that it provides an evidentiary basis for viewing the claim of race-neutrality and non-prejudice among opponents of immigration as suspect. Indeed, our studies uncover an important “hidden” ethnicity-based group bias in public reactions to immigrants. Beyond this, however, the most novel feature of our findings is the demonstrated effects of this hidden bias on policy attitudes, where the portion of reported offense in response to transgressive immigrant behaviors explained by manipulated variation in the perpetrating immigrant’s ethnic identity significantly influenced immigration policy preferences.

In sum, from the opinion research on immigration we know that there is a significant degree of anti-immigrant sentiment among the public, and that this corresponds to support for a variety of controversial restrictive immigration policies at the federal and state level. What the findings from our survey experiments suggest is that a unique portion of this sentiment is grounded in group-based prejudice. Interestingly, however, this prejudice is hidden under the surface because it is coded into the race-neutral language of concern over the threatening behavior of immigrants. Our findings suggest that the importance of threat in predicting attitudes may in part be due to its current employment as a “coded” vehicle through which prejudice toward Hispanics is expressed and translated into political opposition to immigration.

One issue left open for future research is whether the “coding” of prejudice toward Hispanics is a conscious versus unconscious process. For example, do citizens, who are simultaneously aware of their own prejudice and social norms discouraging its expression, consciously alter their language of opposition to immigration? Or, do the biases we observed operate by a process of unconscious mobilization of unacknowledged racial antipathies? Additionally, to what extent are coding processes—be they conscious or unconscious—achieved by citizens alone or with the aid of elite frames that implicitly harness prejudice as their underlying source of popular resonance? We believe that addressing these types of questions and issues would indeed be a fruitful direction for future research.

The larger significance of decoding mass antipathy toward Hispanics pertains to determining pathways toward the achievement of meaningful and over-due immigration reforms, as well as the promotion of social and civil harmony among Whites and growing immigrant minorities in the U.S. If concerns over illegality, economic competition, and cultural assimilation are the culprits standing in the way of these goals, then the solutions, beyond the provision of available factual information to quell these concerns, would be law enforcement, employment eligibility verification systems in conjunction with policies to create more and new jobs for Americans, and policies aimed at facilitating the assimilation of immigrants. If, on the other hand, the culprit is a simple, yet masked, ethnic prejudice toward salient immigrant groups, then the solutions would shift toward official efforts aimed at reducing and undermining mass prejudice through programs promoting positive intergroup contact between white Americans and residentially proximate Hispanic populations.

Footnotes
1

Allport’s (1954) original definition of prejudice also included the condition that the antipathy be “based upon a faulty and inflexible generalization.” However, as Brown (2010) and Stangor (2009) note, most scholars have subsequently dropped that requirement from the operational definition of prejudice.

 
2

Excerpts accessed from Numbers USA’s organizational website: https://www.numbersusa.com/content/. Emphasis in italics were added by authors and do not appear in the original quote.

 
3

Quote from Alabama State Representative Kerry Rich. Retrieved from: http://www.sandmountainreporter.com/news/local/article_b58d77c2-9920-11e0-8a09-001cc4c03286.html.

 
5

Excerpt taken from an interview given by Rubio on Democracy Now! on October 20, 2011. Retrieved from: http://www.democracynow.org/shows/2011/10/20.

 
6

In our usage, the terms “Whites” and “white Americans” exclude individuals who would identify themselves as ethnically Hispanic.

 
7

Data collected by the Survey Research Center, University of California, Berkeley, from February 1 to November 21, 1991.

 
8

A list-assisted method of random digit-dialing (RDD) was used to obtain phone numbers in the sample from all 48 contiguous states, including the District of Columbia. Within selected households, individuals 18 years and over were chosen at random for participation. Multiple attempts were made at each contact number (as many as seven attempts) in order increase response rates and give potentially eligible respondents a reasonable opportunity to participate in the survey. Moreover, households and individuals who were initially unwilling to participate in the survey were contacted multiple times in an attempt to persuade them to participate. Calls were staggered over times of day and days of the week to maximize the chance of making contact with potential respondents. In total, 6,032 telephone numbers were dialed, each of which was given a final disposition: 3,763 numbers were deemed ineligible (e.g., nonworking, businesses, etc.), 1,109 numbers were of unknown eligibility (always busy, never answered, etc.), and the remaining 1,160 numbers were coded as eligible households (275 completes, 304 refusals, 36 non-whites, 70 language unable, and 475 callbacks). We used two methods of determining levels of participation in this survey: (1) The Cooperation Rate (AAPOR Formula #4) was 51.4 %; and (2) the Response Rate (AAPOR Formula #4) was 22.6 %. The response rate is a very conservative estimate of participation, while the cooperation rate adjusts for the fact that many phone numbers in the list are non-eligible.

 
9

The “Pew Research Center Poll: Immigration” was sponsored by the Pew Research Center for the People & the Press and the Pew Hispanic Center. A total of 6,003 surveys were completed between February 8th and March 7th, 2006. The 2008–2009 American National Election Panel Study consists of an Internet panel of 4,240 Americans recruited via RDD sampling methods.

 
10

Percentages do not total 100 % because of rounding.

 
11

For ease of interpretation, we recoded this and all subsequent variables from 0 to 1.

 
12

For instance, a respondent who was presented with the scenario of a Mexican immigrant overstaying his or her visa was later asked to evaluate the seriousness of this same Hispanic immigrant working without paying income taxes.

 
13

Unlike the previous realistic experiments, our symbolic experiments involved only two conditions, and respondents were randomly assigned to either condition for each individual symbolic experiment. Our decision to exclude the race-neutral control condition for the symbolic experiments stemmed from our concern that the prior exposure to group cues in the realistic experiments could prime respondents in the neutral condition. Thus, participants in an unidentified symbolic condition, had we included one, could conceivably be primed to think of that “someone” as Mexican, British, or truly undefined, which would undermine the integrity of contrasts in responses to respondents assigned to the explicit Mexico or Canada conditions. To eliminate this possibility, our symbolic experiments explicitly specify the ethnic origins of the immigrant in question. In addition, we changed the ethnicity of the non-Hispanic immigrant in our symbolic experiments from British to Canadian to demonstrate that our results hold across different white immigrants.

 
14

As there were no significant differences in main effects between the non-Hispanic treatment conditions in our realistic experiments (i.e., British and undefined cues), we opted to combine them into a single category. Thus, we used a dichotomous group cue variable for our analyses (1 = Mexico; 0 = non-Hispanic). The coding for the symbolic experiments was similar (1 = Mexico; 0 = Canadian).

 
15

Perceptions about the illegal immigrant population of Hispanic origin came from responses to the following question: “If you had to guess, what percentage of the Hispanic immigrant population is living in the U.S. without legal documentation?” Responses ranged from “0” to “100 %,” with a mean of 42.3 % and a standard deviation of 24.6 %. Education is a 6-point scale, where a graduate degree serves as the highest category. Household income is a 8-point scale based upon $20,000 increments, and missing values were imputed in Stata based upon gender, education, age (and its squared term), and employment status. For ease of interpretation, all variables were recoded from 0 to 1.

 
16

Kinder and Kam (2010) demonstrate that ethnocentrism, or “a predisposition to divide the human world into in-groups and out-groups” (p. 8), strongly predicts anti-immigrant sentiment. They argue that individuals are predisposed to favor their ingroup at the expense of outgroups, and that antipathy toward outgroups should increase as a function of the cultural, linguistic, and ethnic distance of an outgroup to one's ingroup. According to this approach, the operative mechanism underlying our experimental findings could be general aversion to outgroups and “prejudice broadly defined” (Kinder and Kam 2010, p.52), rather than group-specific prejudice toward Hispanics divorced from an encapsulating ethnocentrism. While this alternative and more general framework could account for findings such as ours, this hypothesis is directly challenged by evidence that specific attitudes toward Hispanics, not ethnocentrism, influence immigration policy preferences (Valentino et al. 2013). In light of these countervailing findings, we should note that the primary goal of this article is to test for the existence of bias toward Hispanics by determining whether individuals evaluate the transgressive behaviors of Hispanic immigrants more negatively than those of non-Hispanic immigrants. Adjudicating whether the demonstrated bias in our experiments stems from prejudice toward Hispanics embedded within general ethnocentrism is beyond the scope of this article.

 
17

Once again, we opted to collapse the British and undefined treatments from our realistic experiments because we found no significant differences in main effects between these conditions.

 
18

To estimate the mediated effects of our group cue treatments on policy preferences, we used the mediation package in R (Tingley et al. n.d.) to regress (1) a continuous measure of perceived offensiveness of a given violation on a dichotomous group cue treatment variable using OLS, and (2) a categorical immigration policy item on the perceived offensiveness of a given violation, as well as a dichotomous treatment variable, using probit or ordered probit link functions.

 

Acknowledgments

We would like to thank Dan Hopkins, Shanna Pearson-Merkowitz, Nick Valentino, and Cara Wong for their helpful suggestions on earlier versions of this article. We also thank Dustin Landers for his help during the data collection phase of this project.

Copyright information

© Springer Science+Business Media New York 2013