What Triggers Public Opposition to Immigration? Anxiety, Group Cues, and Immigration Threat


  • Data collection for this article was made possible by Time-sharing Experiments for the Social Sciences (TESS) and grants from the Howard R. Marsh Center for the Study of Journalistic Performance and the Horace H. Rackham School of Graduate Studies. Antoine Banks, Andrea Benjamin, and Eric Groenendyk provided excellent research assistance. For helpful comments on this research at various stages, we thank Jake Bowers, Nancy Burns, Jamie Druckman, Zachary Elkins, Stanley Feldman, Rob Franzese, Marty Gilens, Don Green, Greg Huber, Leonie Huddy, Don Kinder, John Lapinski, Skip Lupia, George Marcus, David Mayhew, Tali Mendelberg, Joanne Miller, Ken Scheve, John Sweeney, Cara Wong, and anonymous reviewers, as well as audiences and discussants at MPSA and APSA annual meetings, Dartmouth College, the University of Michigan, New York University, Princeton University, and Yale University.

Ted Brader is associate professor of political science, University of Michigan, Center for Political Studies, 4242 ISR, 426 Thompson Street, Ann Arbor, MI 48106-1248 (tbrader@umich.edu). Nicholas A. Valentino is Mike Hogg professor of community affairs, University of Texas at Austin, Department of Government, 1 University Station A1800, Austin, TX 78712-0119 (nvalenti@austin.utexas.edu). Elizabeth Suhay is a doctoral candidate in political science, University of Michigan, Department of Political Science, 5700 Haven Hall, Ann Arbor, MI 48109-1045 (suhay@umich.edu).


We examine whether and how elite discourse shapes mass opinion and action on immigration policy. One popular but untested suspicion is that reactions to news about the costs of immigration depend upon who the immigrants are. We confirm this suspicion in a nationally representative experiment: news about the costs of immigration boosts white opposition far more when Latino immigrants, rather than European immigrants, are featured. We find these group cues influence opinion and political action by triggering emotions—in particular, anxiety—not simply by changing beliefs about the severity of the immigration problem. A second experiment replicates these findings but also confirms their sensitivity to the stereotypic consistency of group cues and their context. While these results echo recent insights about the power of anxiety, they also suggest the public is susceptible to error and manipulation when group cues trigger anxiety independently of the actual threat posed by the group.

Immigration surged onto the national agenda following the 2004 election, as politicians wrangled over reforms on what is perceived to be a growing problem for the United States (U.S.). Public concern followed, with 10% of Americans by 2006 naming it the most important problem facing the country, the highest level in 20 years of polling by Pew Research Center. It became a major issue in the 2008 presidential election. Republican anger at John McCain for his support of the Comprehensive Immigration Reform Act contributed to the near collapse of the early front-runner's campaign in 2007. A few months later, Republicans in the lead-off Iowa caucuses picked immigration as the country's most important problem more than any other issue (including war, terrorism, and the economy).1 Debates about immigration have flared at intervals throughout U.S. history (Tichenor 2002). Current episodes reflect mounting pressures from heavy immigration and an expanding Latino electorate. West Europe also has experienced a rising tide of migrants, spurring bitter debates over how to deal with the newcomers and a growth in electoral position taking (Fetzer 2000; Sniderman, Hagendoorn, and Prior 2004; Sniderman et al. 2000). Even though the issue has cut across partisan divisions in the past, activists and politicians increasingly are taking the debate to the American public.

Public debates about immigration, like those in other domains, often suggest the interests, values, or lifestyles of citizens are in harm's way. Elite discourse tends to emphasize adverse consequences for jobs, taxes, crime, schools, cultural norms, or social harmony (Simon and Alexander 1993). It is not surprising, therefore, that opponents of immigration always outnumber supporters. Nonetheless, public opinion on the issue can be volatile, with opposition spiking to nearly two-thirds in some years and dropping to as little as one-third in others (Lapinski et al. 1997; Simon and Lynch 1999). Why do Americans find the “threat” of immigration more or less compelling? As with many issues, Americans tend to be poorly informed and uncertain about immigration (Scheve and Slaughter 2001) and much of what they learn comes through the mass media. We suspect that the way journalists and politicians portray immigration plays a significant role in activating (or assuaging) opposition.

Our aim is to ascertain which features of public debate affect popular support for immigration. We begin with an assumption about the essential nature of immigration as a political issue: it generates conflict between groups—citizens versus noncitizens, English speakers versus foreign language speakers, whites versus nonwhites (King 2000). Discourse and opinion on immigration is “group centric” (Nelson and Kinder 1996). Immigration attitudes seem rooted substantially in group identity and prejudice (Citrin et al. 1997; Kinder 2003). In contrast to this line of work, some studies stress the role of economic factors in explaining significant individual and cross-national variation in immigration opinion (Quillian 1995; Scheve and Slaughter 2001). Whether prior studies emphasize “symbolic” group factors or “real” costs, they say little about how or why opposition to immigration changes, especially in the short term. Sniderman and colleagues (2004), however, show that political attitudes toward immigration are shaped by “situational triggers” as well as predisposing factors.

Can ethnic or racial group cues act as situational triggers that change how the American public responds to immigration? Many observers regard as commonplace the notion that support for immigration depends on who the immigrants are. Many opponents of immigration claim, however, that opposition is based on the consequences of immigration, not the identity of immigrants. Moreover, prominent recent scholarship has been either silent on the question or found that opinion in some European countries does not depend on the race or ethnicity of immigrants (Scheve and Slaughter 2001; Sniderman, Hagendoorn, and Prior 2004; Sniderman et al. 2000). We believe immigrant identity matters in the United States. In particular, we suspect that news about the potential threat of immigration will have a larger impact on whites' attitudes when the immigrants being discussed are nonwhite. Testing this suspicion is our starting point, because researchers have yet to establish that group cues affect how individuals respond to information about immigration. Ascertaining the existence of such effects, however, does not explain why they occur. Our primary goal is to understand why the race or ethnicity of immigrants, above and beyond arguments about the consequences of immigration, drives opinion and behavior.

If the race or ethnicity of the salient immigrant group influences opinion, there are at least two potential explanations. First, group cues might alter public perceptions about the severity of the problem. Negative stereotypes about Latinos in particular might boost concerns about cultural assimilation, consumption of scarce public resources, crime, and so on. Thus when the news media highlight Latinos in discussions about immigration, white citizens may come to believe that immigrants pose an even greater problem than if white Europeans were featured. This strengthening of the perception that immigration has a negative impact might then spur greater opposition.

Although perceived harm is a plausible mediator of group cue effects, we suspect a different mechanism is at work. Racial or ethnic cues may trigger emotional reactions, such as anxiety, which may cause changes in opinion and behavior independently of changes in beliefs about the severity of the immigration problem. This argument differs dramatically from traditional research on political threats, which tends to highlight the impact of cognitively based perceptions (Miller and Krosnick 2004). Recent research suggests that emotions play a central role in both political judgment and behavior (Marcus, Neuman, and MacKuen 2000). When anxiety is triggered by threatening cues or conditions, it can facilitate opinion change and motivate information seeking and political action. Though prior scholarship posits the emotional power of groups (Kinder and Sanders 1996), ours is the first to test the emotional impact of group cues on political attitudes and behavior (cf. Hutchings et al. 2006).

Although our predictions are consistent with recent work on anxiety, their normative tone differs. Several studies stress the potential benefits to learning and rational decision making that follow from anxiety increasing openness to new information and decreasing reliance on predispositions (Brader 2006; Marcus, Neuman, and MacKuen 2000; Valentino et al. 2008). These effects, however, also make people more susceptible to manipulation if the information at hand is skewed or untrue. In addition, if anxiety is triggered independently of the information provided by stereotypic group cues, the public may react more strongly than the facts warrant in some situations and not strongly enough in others.

In this article, we examine how key aspects of elite discourse influence opposition to immigration. We expect that public opinion about immigration depends at least in part on what citizens read or hear in the news. We expect that news stimulates greater opposition when it highlights costs rather than benefits of immigration. But we also think that such reactions are tied to the immigrant groups made salient in the news. Group cues imbue the discussion of costs and benefits with emotional significance. In other words, group images cause changes in attitudes and behavior by triggering an emotional reaction, rather than by simply changing beliefs about the severity of the problem.

Public Discourse and Opinion on Immigration

Our interest in immigration opinion is driven in part by the perennial importance of the issue. Debate over immigration has heated up in recent years, with concerns about its economic and cultural impact contributing to a backlash against globalization (Cohen 2001; Huntington 2004; Scheve and Slaughter 2001). After the 9/11 attacks, the issue prompted new concerns about national security. As politicians and activists bring these concerns further into the public spotlight, their appeals are countered by an increasingly influential and vocal Hispanic minority. In the 1990s, Governor Pete Wilson rode to reelection in California championing Proposition 187, which denied government services to immigrants. This “victory,” however, may have damaged Republican Party prospects in the nation's most populous state by alienating Hispanic citizens. George W. Bush took the opposite tack by proposing to liberalize immigration policy and received record levels of support from Hispanics in the 2004 election. Beginning in 2006, local and national efforts to expand criminal sanctions against illegal immigrants and their employers led to massive protests. At the same time, conservative Republicans became increasingly outraged by congressional attempts to pass reform legislation that would balance border control with pathways to legal status for illegal immigrants.

Immigration's status on the national agenda has waxed and waned throughout history (Tichenor 2002). Although immigrant nationalities have changed, there have been striking similarities in the central issues and character of the debate across time. Though rising levels of immigration carry both costs and benefits, the emphasis of public discourse tends to be on the former (Simon and Alexander 1993). Politicians and activists have always voiced concerns about the economic and cultural threat immigration poses to Americans (Cohen 2001). News coverage from 1995 to 2005 was twice as likely to stress the costs of immigration as the benefits (Pande 2006).

Anti-immigration rhetoric also often has group overtones and, explicitly or implicitly, makes a distinction between stigmatized ethnic or racial groups, such as Hispanics, and “good” immigrants, such as those from Canada, Ireland, or Poland (Huntington 2004; King 2000). Again, from 1995 to 2005, 84% of immigration stories in major papers mentioned specific groups (Pande 2006). Twice as many stories referred to Hispanics as the next most mentioned group (East Asians), and more stories spotlighted Hispanic migrants than migrants from all other regions combined.

Popular attitudes often seem to mirror these patterns in elite discourse. Public opinion about immigration runs heavily toward opposition (Simon and Lynch 1999), with a plurality of Americans preferring to reduce the number of immigrants or hold it constant. The share of the public favoring increases in immigration is consistently tiny, usually hovering around 10%. Surveys suggest opposition is tied to both “realistic” economic concerns and “symbolic” cultural concerns, but especially the latter (Citrin et al. 1997; Kinder 2003; Scheve and Slaughter 2001; Sniderman et al. 2000). Although research has turned up numerous correlates of opinion, we know little about the effects of political discourse. There is evidence, however, that opposition to immigration (in the Netherlands) is responsive to situational triggers such as the relative salience of national identity or of economic versus cultural threats (Sniderman, Hagendoorn, and Prior 2004). Finally, we also know significantly less about what motivates citizens to act on their attitudes.

In order to understand how public discourse affects popular opposition to immigration, we examine three characteristics of that discourse: emphasis on costs and benefits, ethnic identity cues, and portrayal of immigrants as low- or high-skilled workers. Prior research indicates that negative consequences and group cues are salient in U.S. news coverage of immigration. References to the occupation or skill level of immigrants are not rare, but typically less pronounced than the other two factors. Nonetheless, many economic models view skill level as critical to workers' perceptions or experience of the harm they face from migrants (Scheve and Slaughter 2001). In addition, ethnic identity cues, such as pictures of Hispanic immigrants, may signal more than ethnicity. Today the popular image of immigrants in the American mind is based on the stereotype of low-skilled Hispanic laborers. Perhaps reactions to Hispanic immigrants are primarily based on the economic skills and needs they are perceived as bringing, rather than on their ethnicity per se. In order to isolate better the effect of ethnic cues, we try to disentangle them from skill level.

The Mediating Role of Beliefs and Emotions

Our second major goal in this article is to examine the psychological mechanism underlying the effects of group cues in immigration discourse. Why does the salience of particular ethnic or racial groups in elite debate affect public opinion? We already have considerable evidence that such cues influence opinion in other policy domains by priming attitudes toward the group in question. Brief references to groups that benefit from a policy as well as subtle visual cues in news or political ads can cause group attitudes to play a larger role in opinion formation (Gilens 1999; Mendelberg 2001; Nelson and Kinder 1996; Valentino, Hutchings, and White 2002). Group priming may work differently, however, in the explicitly group-based context of immigration, a possibility we explore elsewhere (Brader and Valentino 2005). If group priming is most evident when cues are subtle (Mendelberg 2001; Valentino, Hutchings, and White 2002), then any effect of the relatively explicit cues in immigration discourse may function by directly triggering emotions or changing perceptions of threat.

We consider the extent to which immigrant group cues change political attitudes and behavior by altering beliefs about consequences, triggering emotions, or both. One classic route to persuasion is by changing the beliefs or perceptions about the world that form a basis for attitudes (Eagly and Chaiken 1993). Belief change does not require exposure to direct arguments. Ethnic or racial group cues may carry implications for how citizens perceive the nature or severity of any harm (or benefit) from immigration. For example, migrants from non-English-speaking countries might suggest cultural tensions, whereas migrants from poor countries might suggest economic competition and a strain on public coffers. A citizen's policy preference may depend on such factual perceptions, on what she believes to be true about the policy and the problem it is meant to address. Thus, if group cues change beliefs about the severity of the immigration “problem,” opinion change may follow.

A long line of research also suggests that eliciting emotions can motivate changes in attention, opinion, and behavior (Eagly and Chaiken 1993). For example, Hot Cognition theory contends that almost any object—social groups, politicians, policy issues—triggers automatic positive or negative affect based on a person's past evaluations of the object. This initial emotional reaction biases subsequent information processing in a direction consistent with the initial reaction (Lodge and Taber 2005). Another approach, Affective Intelligence theory, distinguishes between two basic emotional systems (Marcus, Neuman, and MacKuen 2000). The “disposition system,” which generates enthusiasm based on feedback about a person's progress toward her goals, regulates interest and reliance on habit. The “surveillance system” generates anxiety or fear in response to potential threats. Anxiety in turn prompts information seeking and a greater focus on contemporary information, thereby increasing the likelihood of learning and persuasion (Valentino et al. 2008). Because anxiety focuses attention on available information, the direction of persuasion depends on the implications of that information (e.g., whether it suggests immigration is harmful or beneficial). Finally, since emotions are motivational impulses (Lodge et al. 2006), they also make it more likely that individuals will take political action (Brader 2005, 2006; Valentino et al. 2006). Might outgroup cues in news about immigration trigger anxiety? We think so. Although political scientists have only just begun to research “emotional triggers” in politics, there is a good deal of work in psychology demonstrating the capacity of groups to evoke emotions (Mackie and Smith 2003).

Of course, belief change and emotional responses are not mutually exclusive. Both may mediate the effects of group cues on immigration opinion. After all, beliefs about the severity of immigration problems and emotions about that same topic are often strongly correlated, and each may affect opinion formation (Isbell and Ottati 2002). We expect that, overall, elite debate about immigration influences perceptions about the severity and nature of the immigrant “threat”and elicits anxiety or other emotional reactions. Nonetheless, prior scholarship on threat has often looked either at the role of perceptions of harm or anxiety, but rarely both. Where both have been considered, scholars have offered differing assessments, such as that perceived threat and anxiety are closely related elements in the same causal chain (Marcus et al. 1995, 2005), that perceived threat matters while anxiety does not (Miller and Krosnick 2004), or that perceived threat and anxiety are strongly correlated but have distinct effects (Huddy et al. 2005).

We expect that cues to the identity of immigrants will principally influence public opinion and political action by triggering anxiety or other emotions. Ethnic and racial groups are potent, affect-laden symbols in American society (Mackie and Smith 2003; Sears 2001). Moreover, in the ordinary course of politics, citizens often arrive at judgments without devoting much attention or effort to learning, remembering, and sifting through details. They are guided instead by “gut” reactions and habits of mind (Lodge and Taber 2005; Marcus, Neuman, and MacKuen 2000). Under typical circumstances, therefore, we expect the emotional impact of group cues to shape opinion and behavior, even though it is plausible for such cues to affect opinion by changing perceptions about the nature or severity of a threat. To the extent group cues trigger anxiety about immigration, we expect changes in opinion and action consistent with the information at hand. Likewise, if anxiety makes opinion change and action more likely, then we expect anxiety to mediate these observed effects of outgroup cues.

Our expectations depart in two ways from the positive normative implications of anxiety stressed in Affective Intelligence theory. First, while the theory predicts openness to new information when anxiety is aroused, people should be open to available arguments, not necessarily valid ones. When people focus on the information at hand, they are also more vulnerable to the ways in which that information can be skewed or manipulated. This may lead to undesirable outcomes. Second, we argue that triggers for anxiety may, but need not, correspond to realistic or valid threats. Just as campaign ads evoke emotions with staged imagery often unrelated to the political message (Brader 2006), news reports may trigger emotions not only with the “hard” information they provide, but also with the way in which the story is packaged. Thus, group cues in immigration news may trigger anxiety independently of actual information about the severity of the threat. Although fear of outgroups is not necessarily irrational, it may cause individuals to react more strongly than “facts” warrant as well as to treat equivalent situations unequally (i.e., based solely on the salient group). Such distortions in public opinion can in turn provide incentives, or justifications, for officials to enact biased or overreaching public policies.


Our main theoretical goals then are (1) to examine the impact of group cues in immigration discourse on perceptions of threat and emotions and (2) to test whether changes in those perceptions, emotions, or both mediate the impact of discourse on opinion and behavior. The theoretical discussion yields three main hypotheses. First, the impact of information about the consequences of immigration on attitudes and behavior will depend on which immigrant groups are salient. We expect greater opposition when news emphasizing the harm from immigration contains images of Latino migrants than when it portrays white European migrants.

Next we will explore why, if at all, such an ethno-racial effect occurs. To do so, we will examine the impact of news on beliefs and emotions about immigration. We know from decades of research that individuals who are exposed to a one-sided message from a trusted source are more likely to accept the message content as true (Eagly and Chaiken 1993). Therefore, few would be surprised if a credible news report that stressed the costs of immigration increased the belief that immigration is harmful. Given their highly symbolic nature (Sears 2001), group cues should be potent triggers of emotion. Specifically, outgroup cues should be a potent source of anxiety in news about immigration. In one sense, all immigrants are members of an outgroup (i.e., “foreigners”). But the history of immigration makes clear that some immigrants, those ethnically or racially distinct from the native majority, are regarded as more alien than others (King 2000). We expect white Americans to respond with stronger, more negative emotions when elite debate focuses on Latino immigrants. Our second hypothesis, therefore, is that while any news emphasizing the costs of immigration should increase the perception that immigration is harmful to Americans, it will be more likely to trigger anxiety when the story calls attention to Latino, as opposed to white European, immigrants.

Opinions and behavior in turn may change in response to shifting perceptions of harm, emotional reactions, both, or neither. We expect most people to update their opinions and take new action only when their routine is interrupted by the experience of anxiety (Brader 2006; Marcus, Neuman, and MacKuen 2000). Thus, our third hypothesis is that opinion change, information seeking, and political action should be more likely to the extent news about immigration arouses anxiety; in other words, anxiety should mediate these effects. According to Affective Intelligence theory (Marcus, Neuman, and MacKuen 2000), anxiety should directly stimulate a search for information relevant to the threat and motivate political action aimed at removing the threat. Because anxiety facilitates persuasion indirectly, however, by shifting attention to available information, the direction of opinion change should be consistent with that information. For example, if Latino cues increase anxiety, as our second hypothesis suggests, then we expect those anxious citizens to become more opposed to immigration after reading threatening news about the issue. If those anxious citizens, however, encountered reassuring news, then they should become more supportive (or less opposed) to immigration.

Finally, on one point we are less certain. Will we in fact observe increased anxiety if stigmatized outgroup cues accompany otherwise reassuring information about immigration (e.g., that it has beneficial consequences for Americans or that immigrants bring more skills/assets than expected)? Given the automatic nature of emotional triggers (Damasio 2000), we expect that such group cues, once encoded with such negative affect, will initiate an anxious response in all situations. However, this automatic reaction is also revised and updated continuously on the basis of new information (Lazarus 1991; LeDoux 1996), so that anxiety may dissipate quickly, perhaps even before an individual is consciously aware of the feeling, if the trigger is accompanied by reassuring signals. This perspective does not leave us with a clear prediction about whether anxiety will change in response to Latino cues embedded in more reassuring news about immigration. However, research on political tolerance suggests that anxiety is more likely when disliked political groups are portrayed in a threatening, rather than reassuring, context (Marcus et al. 1995). A second possibility is that the emotional “trigger” is not the ethno-racial cue itself, but rather a stereotypic image of that group. Recent research finds that cognitive processes such as priming (Mendelberg 2001; Valentino, Hutchings, and White 2002) and emotional processes (Mackie and Smith 2003) can depend on stereotype consistency. Thus, news inconsistent with stereotypes about Latinos may elicit less anxiety than stereotype-consistent stories. If this is the case, we should expect greater anxiety when Latino immigrants are portrayed as low-skilled workers with potentially harmful consequences for American society than when they are portrayed as high-skilled or potentially beneficial. The competing hypothesis, in this case, is that Latino immigrants will trigger greater anxiety regardless of how they are portrayed.

To test the hypotheses, we conducted two experiments in the summer and fall of 2003. Experiments allow us to distinguish the causal influence of information about costs and benefits of immigration, information about immigrant skill level, and cues of immigrant ethnicity. The data also shed light on the psychological mechanism. Perceptions of harm and anxiety are often highly correlated, and self-reports of each are affected by predispositions. In an experiment, however, we can focus on changes in perceptions or emotions in response to manipulations of news content and, in so doing, discern better which are mediating any observed effects.

Experiment 1

We embedded an experiment in a nationally representative survey of 354 white, non-Latino adults. The survey was conducted via WebTV by Knowledge Networks, which maintains a large, randomly contacted respondent pool by offering free Internet access in exchange for occasional participation in surveys. The company randomly selects participants from its pool for each survey.2 Subjects from 46 states took part in ours. The median interview was 16 minutes. The median subject was 46 years old with some college and a $45,000 household income. Fifty-two percent of the subjects were women. Thirty-one percent identified as Republicans, and 28% as Democrats.

Prior to the experiment, subjects were told they would take part in “a survey regarding a number of current social issues.” The survey first measures political predispositions and economic outlook. Immediately prior to the stimulus, subjects were told that we wanted their reactions to a story that “had been in the news lately.” After reading, subjects answered questions about immigrants and immigration. An appendix provides wording for all questions used here.

The web-based platform allows us to deliver stimuli matching those in actual news coverage. The study employed a 2 × 2 design with a control group. We manipulated the ethnic cue by altering the picture and name of an immigrant (white European versus Latino) featured in a mock New York Times report about a governors' conference on immigration.3 We also manipulated the tone of the story, focusing either on the positive consequences of immigration for the nation (e.g., strengthening the economy, increasing tax revenues, enriching American culture) or the negative consequences (e.g., driving down wages, consuming public resources, undermining American values). Tone was also manipulated by portrayal of governors as either glad or concerned about immigration and citizens who had had either positive or negative interactions with immigrants.4 Every story stated that immigration to the United States is increasing and will continue to do so.5

Our goal in manipulating group salience was to test the impact of the ethnic identity cue itself. We strove to vary the cue in a way that held constant any new information about the nature and severity of immigration's impact. Each article contained a picture of a “recent immigrant” who appears to be either Latino or European. A caption read: “[Jose Sanchez/Nikolai Vandinsky] is one of thousands of new immigrants who arrived in the U.S. during the first half of this year.” To maximize control, a graphic artist altered the images so that only the heads differed (i.e., bodies, dress, and background were identical). We selected two faces from 40 photos based on ratings by an independent panel of eight naïve judges. Judges rated how much each male appeared ethnically European or Hispanic, as well as wealthy, attractive, friendly, and law abiding. The faces chosen for our manipulation were maximally distinct on the dimension of ethnicity, but statistically indistinct on other traits.6

The article included a quote (identical across conditions) from the men on their feelings about coming to the United States and identified their national origin as Mexico or Russia. We named specific countries to conform with standard news practice and, more importantly, to maintain control over what subjects called to mind. Prior research has shown that when information is omitted, subjects tend to “fill in” the details (Gilliam and Iyengar 2000). For the Latino immigrant, we expected whites to assume Mexican nationality even if it were not mentioned. We were less certain about the assumptions respondents would make about a fair-skinned immigrant who appears to be of European descent. We chose Russia for several reasons. First, most Americans have heard of it; picking a country many do not know (e.g., Slovenia) would be equivalent to naming no country at all. Second, we wanted to hold the fact of language differences constant. Third, we ruled out countries linked to earlier waves of immigrants with which many white Americans identify, such as Ireland or Italy. Finally, working-class Russians have been emigrating to the United States in fairly large numbers but are not branded with a racialized stigma (Rumbaut and Portes 2001). To the extent Russian nationality has negative associations, ours is a conservative choice that may dampen effects.

The Impact of News on Opinion, Information Seeking, and Action

Our first hypothesis is that news emphasizing costs should boost opposition to immigration more powerfully when it is accompanied by images of stigmatized groups than when it is accompanied by racial ingroup members. To test this hypothesis, we examine effects on both opinion and political behavior, starting with the former. We tapped immigration preferences via a standard question about whether immigration to the United States should be increased, decreased, or kept the same. We also asked a more specific policy question: whether English should be the official language of the country. Figure 1 displays the mean opinion across conditions.

Figure 1.

Impact of News Frame and Ethnic Cue on Immigration Policy Preferences

Note: Each graph displays the mean policy opinion of the four treatment conditions. The “baseline” control group mean values for each panel are 2.65 (top) and 2.44 (bottom).

There is a main effect on overall preferences for the news story's emphasis on consequences (Mdiff= .42, F = 11.92, p < .001) and an interaction effect between emphasis and ethnic cues (F = 3.24, p < .073).7 Respondents in the negative (i.e., immigration is costly) Latino condition expressed the most opposition to immigration, while those in the positive Latino condition expressed the most support. Although emphasis on costs versus benefits works similarly for the European and Latino versions, the negative effect is much larger for stories with Latino cues, consistent with our first hypothesis. The impact of the threatening story is over two times stronger when Latino cues are present.8 This effect is substantively large, shifting opinion 15% of the scale. The interactive pattern of support for English-only laws is similar. The negative Latino story boosts opposition by 7% of the scale, and the interaction is marginally significant (F = 3.80, p < .052).

Next we examine the impact of our news stories on political behavior. Our goal was to improve upon typical survey self-reports of willingness to engage in activity. Our items were designed so that subjects would believe they were requesting information from real organizations or taking action on the issue. First, we asked if they would like more information about immigration from a variety of sources, including nonpartisan research centers, the U.S. government, pro-immigrant groups, and anti-immigrant groups. This question taps information-seeking behavior by leading subjects to believe they would be contacted by the group(s). Second, subjects could request that an email message be sent on their behalf to members of Congress indicating their support for a law reducing the number of new immigrants. This item measures political mobilization by invoking real contacting of elected officials. Subjects were told the email would contain their name, city, and opinion on the issue.

In all, 28% of respondents requested information and half of these requested it from more than one source. Despite the diversity of sources, the decision to request materials from any source was positively correlated with the decision to do so from others. We therefore count the number of sources requested as a measure of general information seeking. Figure 2 (top panel) displays the mean number of requests by respondents in each condition. The pattern only partially matches expectations. The slope for Latino cues is similar to what we have seen previously: respondents request nearly three times as much information on average when the news is negative. The slope runs the other direction, though insignificantly so, for European cues. Thus the interaction effect is significant (F = 4.05, p < .045). Surprisingly, however, the European cue produces higher demand for information compared to the Latino cue, regardless of the frame (Mdiff= .31, F = 8.44, p < .004).

Figure 2.

Impact of News Frame and Ethnic Cue on Information Seeking and Political Action

Note: The “baseline” control group mean values for each panel are 0.51 (top), 0.14 (middle), and 0.31 (bottom).

In an effort to clarify this unanticipated main effect, we compare the treatments to the global control group, in which subjects did not read a story about immigration. The mean number of requests in the control group is 0.51, suggesting the negative Latino cell is closest to “baseline” level of requests. The greatest departure from this baseline is the positive Latino cell (t = 2.63, p < .009). It appears, therefore, that pairing Latino immigrant cues with a positive message about immigration actually suppresses the demand for information. While the overall pattern of results for Latino cues is consistent with our expectations, their relation to the results for European cues is surprising. The latter appear to push information seeking above the baseline. For now, we are left to speculate that the European immigrant cue is sufficiently novel that it piques the respondent's curiosity, particularly when benign consequences are emphasized.9

We see the same pattern for information requests from anti-immigration groups. The negative Latino story boosts the proportion who request information from .01 to .10, while the negative European story again suppresses requests from .20 to .13. The interaction effect is significant (F = 4.97, p < .027). Once again, information seeking is dramatically suppressed from the control group baseline of 0.14 when Latino cues are paired with positive news (t = 2.81, p < .006). The main effect for ethnic cue is also significant (Mdiff= .11, F = 10.08, p < .002), with European cues producing higher levels of interest compared to Latino cues regardless of the frame. In sum, the results for information seeking do not conform as neatly to expectations as other results. The European cue produces unexpectedly high information demand. However, the pattern for information seeking when Latino cues are present matches that for opinion change: negative Latino cues result in higher demand than positive ones.

Finally, we examine whether the news story affects a realistic measure of political action. The third panel in Figure 2 displays the proportion of respondents in each condition choosing to send a message to Congress supporting a reduction in immigration to the United States. The now-familiar interaction reemerges (F = 5.38, p < .021), with the negative Latino story mobilizing opposition to immigration. The effect is substantial as well: over 45% of respondents in the negative Latino condition send an anti-immigration message, 19 percentage points higher than the positive Latino condition and 14 points higher than the control group. In contrast, the tone of the story has no effect on contacting when the European immigrant is pictured.

Beliefs about Potential Harm versus Emotional Reactions

So far our results confirm the common speculation that responses to immigration news depend on the immigrant group made salient. The next important theoretical question is why this happens. Do people believe Latino immigration is worse than immigration from other groups? Or is the effect grounded in more visceral (automatic), emotional reactions? Our second hypothesis states that, while news about the costs and benefits of immigration may affect its perceived impact, Latino cues should trigger anxiety. Shortly after reading the story, subjects reported their views on the likelihood that immigration will have a “negative financial impact on many Americans” and/or a “negative impact on the way of life in many American communities.” Summed responses form a perceived harm scale (α= .82). Figure 3 (top) shows the effect of the news manipulations. The main effect of news emphasis on perceived harm from immigration is large (13% of the scale) and significant (F = 12.83, p < .001). Consistent with prior research (Holbrook et al. 2001), the framing effect is asymmetric: negative information produces a sizeable increase over baseline levels (Mcontrol= 3.59), while perceived harm in the benefits frame is nearly identical to that in the control group. We observe similar results for perceptions of economic and cultural harm when analyzed separately.

Figure 3.

Impact of News Frame and Ethnic Cue on Perceived Harm from Immigration and Emotional Reactions to Immigration

Note: The “baseline” control group mean values for each panel are 3.59 (top), 3.63 (middle), and 1.64 (bottom).

What about emotions? Subjects completed a post-test battery of six questions asking how they “feel” about increased immigration—anxious, proud, angry, hopeful, worried, excited. Individuals often report a mix of fear and anger that seems to constitute the sort of anxiety response discussed by Marcus, Neuman, and MacKuen (2000). But recent work also draws attention to the fact that anger can emerge as a distinct response from anxiety, with distinct effects (Huddy, Feldman, and Cassese 2007). We checked for differentiation in the self-reports and effects of anxiety and anger, but found they emerge as part of a common factor and perform similarly when analyzed separately. Therefore, we add responses for anxious, worried, and angry into a single scale of anxiety (α= .85).10 We examine enthusiasm responses as well, by combining the proud and hopeful items into a scale (α= .78).11

The middle and lower panels in Figure 3 show the results for emotions. A main effect emerges for the cost-benefit emphasis of the news on both anxiety (Mdiff= 1.16, F = 10.53, p < .001) and enthusiasm (Mdiff= .45, F = 4.07, p < .045). Consistent with the second hypothesis, however, respondents report much higher levels of anxiety about immigration when an emphasis on costs is paired with the image of a Latino immigrant than when it is paired with a European immigrant (F = 6.68, p < .009).12 When the immigrant is Latino, the negative story boosts anxiety by over 1.8 points on a 0–9 scale (i.e., 20%) compared to the positive story (t = 4.16, p<.001). When the immigrant is European, the negative news story increases anxiety by only one-sixth that amount, a statistically insignificant difference (t = 1.04, p < .298). In contrast, in the third panel of the figure, there is no interaction effect on enthusiasm (F = 0.90, p < .344). These results suggest that white Americans don't necessarily become anxious when reminded about immigration and its harmful potential, nor when reminded that some immigrants are nonwhite. They experience substantially greater anxiety when negative consequences of immigration are paired with a stigmatized outgroup.13 Furthermore, enthusiastic reactions do not depend at all on the ethnic cues paired with news about the issue.

Do Emotions or Beliefs Mediate the Effects of Outgroup Cues on Opinion and Behavior?

We have found so far that bad news containing Latino cues strongly drives opposition. We also found that ethnic cues do not change beliefs about the severity of the immigration problem. Anxiety, however, is powerfully triggered by the same kind of news that moves opinion the most. Is anxiety mediating the effect of racialized news coverage on opinion about immigration? Are the more anxious citizens the same ones whose opinions change and who are motivated to take action? We expect the effects of threatening news are mediated by negative emotions.

For mediation to occur, the explanatory variable must affect both the dependent variable and the proposed mediating variable, and the effect of the explanatory variable on the dependent variable must diminish when the mediating variable is controlled. Table 1 displays results for four dependent variables. Similar evidence of emotional mediation occurs for all the dependent variables examined in this study. The first column of estimates shows the impact of the experimental manipulations on anxiety. In the rest of the table, the first of two columns for each outcome variable shows the unmediated effects of the experimental manipulations. Our primary interest is in the second of the two columns for each outcome variable, when anxiety is added to the equation. In every case, anxiety about immigration influences opinions and action. More importantly, there is a sizeable drop in the interaction term when anxiety is in the model. In order to determine our confidence in these results, we employ the standard Sobel test for mediation (Baron and Kenny 1986; MacKinnon et al. 2002). As Table 1 indicates, the interaction effect is mediated significantly by anxiety for all four variables. The conjunction of Latino cues and negative news about immigration influenced levels of anxiety. Anxiety then caused shifts in policy attitudes, information seeking, and political action.

Table 1.  Anxiety Mediates the Interaction Effect of Ethnic Cues and Cost-Benefit Emphasis on Anti-Immigration Attitudes and Behavior (Experiment 1)
 Effect on MediatorEffects on Political Attitudes and Behavior
Anxiety about ImmigrationDecrease ImmigrationSupport English OnlyRequest Anti-Immigration InformationSend Anti-Immigration Message to Congress
  1. Note: Coefficients (and standard errors) are estimates from ordinary least-squares regression (for Anxiety), maximum likelihood estimation of ordered probit models (for Decrease Immigration and Support English Only), and maximum likelihood estimation of probit models (for Request Anti-Immigration Information and Send Anti-Immigration Message to Congress). All models include controls for education, income, age, and gender.

  2. *p < .10, **p < .05, ***p < .01 (two-tailed).

Emphasis (costs)– 0.14– 0.19– 0.34– 0.46– 0.16– 0.21
Cues (Latino)– 0.06– 0.19– 0.12– 0.34*– 0.29– 1.46***–1.39***– 0.37– 0.28
Cues × Emphasis0.18***0.49*0.200.52*0.341.25**0.98*0.76**0.50
Anxiety about Immigration  2.25*** 1.38*** 1.75*** 1.76***
(0.27) (0.28) (0.44) (0.31)
Sobel Test (z)  2.45*** 2.28** 2.16** 2.34**
R2 (adjusted)0.18        
% Predict Correctly
% Improved 7.718.70.0–

Belief about the severity of the immigration problem does not mediate the interactive effect of ethnic cues and news emphasis. This is apparent from the fact that the conditions for mediation are not met (i.e., the explanatory variable does not cause changes in the mediator). Nonetheless, we confirm this by repeating the mediation analyses, substituting perceived harm for anxiety. Table 2 reminds us there is a main effect of news emphasis on perceived harm, but no interaction effect. Thus, while perceived harm strongly correlates with opinions and actions, controlling for it does not significantly diminish the interaction effect. In each case, the interaction effect remains significant and the Sobel test for mediation does not approach significance. Although respondents perceived a threat from immigration, these perceptions did not mediate effects found in this study.14 A change in beliefs about the harm from immigration was not sufficient to change opinions or behavior.

Table 2.  Perceived Harm Does Not Mediate the Interaction Effect of Ethnic Cues and Cost-Benefit Emphasis on Anti-Immigration Attitudes and Behavior (Experiment 1)
 Effect on MediatorEffects on Political Attitudes and Behavior
Perceived Harm from ImmigrationDecrease ImmigrationSupport English OnlyRequest Anti-Immigration InformationSend Anti-Immigration Message to Congress
  1. Note: Coefficients (and standard errors) are estimates from ordinary least-squares regression (for Perceived Harm), maximum likelihood estimation of ordered probit models (for Decrease Immigration and Support English Only), and maximum likelihood estimation of probit models (for Request Anti-Immigration Information and Send Anti-Immigration Message to Congress). All models include controls for education, income, age, and gender.

  2. *p < .10, **p < .05, ***p < .01 (two-tailed).

Emphasis (costs)0.09*0.230.02– 0.14– 0.30– 0.34– 0.46– 0.16– 0.39
Cues (Latino)– 0.05– 0.19– 0.15– 0.35*– 0.30– 1.46***–1.42***– 0.37– 0.28
Cues × Emphasis0.060.49*0.48*0.53*0.49*1.25**1.15**0.76**0.68**
Perceived Harm from Immigration  2.44*** 1.35*** 1.61*** 2.43***
(0.27) (0.26) (0.48) (0.36)
Sobel Test (z)  0.85 0.85 0.83 0.85
R2 (adjusted)0.11        
% Predict Correctly 46.156.757.356.389.289.669.674.2
% Improved 7.718.30.5–

Alternative Explanations for the Emotional Significance of Group Cues

Consistent with our expectations, we find that stigmatized group cues—in this case, Latinos—imbue news about immigration with greater emotional significance. This in turn increases the impact of the news story on attitudes and behavior. Up until now, we have assumed on the basis of prior research that Latino cues generate stronger emotions because of negative affect stemming from ethnocentrism (Kinder 2003) or racial prejudice (Sears 2001). But Latino cues may hold emotional significance for some reason other than the racial or ethnic stigma attached to that group by white Americans. Perhaps the emotional significance stems instead from the size or economic position of the group, since Latinos constitute the largest current pool of immigrants to the United States and are associated with a poorer, less skilled workforce. The question of why white Americans react so negatively to a potential threat from immigration when Latino (rather than European) immigrants are salient holds clear substantive implications.

An alternative explanation for our results begins with the claim that European and Latino cues trigger distinct beliefs about the impact of immigration. Specifically, responses may be tied to beliefs about the sheer size of each group in the current pool of immigrants and/or about the relative skill level of each group. In the first case, the belief that Latinos make up a larger share of recent immigrants to the United States might lead citizens to take the potential threat from immigration more seriously when it is linked to Latinos (Quillian 1995). In the second case, the stereotype that Latino immigrants are poor and unskilled might lead citizens to worry about the greater costs to society associated with an influx of less skilled labor (Scheve and Slaughter 2001).

We can use data from our initial experiment to address the possibility that reactions are proportional to the size of the immigrant group. First, perceptions of threat should not have increased equally in reaction to both European and Latino cues, as they did in the first experiment, if respondents linked the impact of immigrants to the size of the group. Nonetheless, the survey asked respondents for their “best guess” as to the percentage of recent immigrants who have come from Latin America and from Europe. The mean guesses were 55% from Latin America and 20% from Europe, though individual guesses were highly variable.15 If the ability of Latino cues to modulate levels of anxiety stems from the perception that Latino immigrants are far more numerous than European immigrants, then this perception should either increase along with anxiety or moderate the interaction effect that gives rise to anxiety. To test these possibilities, we examine the impact of the experimental manipulations on the perceived gap between Latino and European immigration levels (i.e., Percent Latino − Percent European). We find no evidence that the interaction which generates anxiety also affects perceived levels of immigration (F = 0.48, p < .489). Instead, Latino cues widen the perceived gap by nearly seven percentage points (F = 3.88, p < .050), and the gap is largest, 42%, for the positive Latino condition. Similarly, the interaction effect on anxiety is not moderated by the perceived gap in immigration levels. In sum, beliefs about the size of immigrant groups are influenced by group cues, but this does not explain changes in anxiety, opinion, or behavior.

Another explanation suggests the interaction effect is driven by assumptions about the skill level of Latino versus (white) European immigrants. If people assume that Latino immigrants are less skilled than Europeans, then this may explain their anxiety. In this case, low-skilled immigrants should trigger anxiety regardless of ethnicity. This contrasts with the notion that the effect is driven by race or ethnicity alone. Of course, these factors are not mutually exclusive. Skill levels may be an important aspect of the negative stereotype. If so, we would expect an interaction between ethnicity and skill level, such that only the depiction of low-skilled Latino immigrants would boost anxiety. We performed a second experiment to explore this possibility.

Experiment 2

We recruited 267 subjects from the local area, including during a festival that attracts hundreds of thousands of visitors from across the country. They were offered $20 for taking a “public opinion survey” about current events and completed a self-administered questionnaire at one of a dozen computer stations. For comparison with the first experiment, we limit our analysis to the 220 non-Latino white subjects. Fifty-eight percent of these subjects were women. The median subject was 45 years old with a bachelor's degree and a household income of roughly $52,000. Twenty-four percent of subjects identified as Republicans and 45% identified as Democrats.

Subjects first answered questions regarding predispositions and political knowledge. They then read the following message: “For this study, we have collected a large number of articles that have appeared recently in newspapers around the country. The computer is going to select one of these newspaper articles at random for you to read. When you are finished with the article, we will ask you some questions about it. Please read carefully.” After reading the article, they answered more questions and then were fully debriefed before leaving.

The experiment employed a 2 × 2 design with a control group. Again a mock New York Times article was used.16 All of the stories included the same basic information as the negative frame from the first study. As before, we manipulated the salient ethnicity using a photo. The caption identified the immigrant as Jose Sanchez or Nicholas Van Dyke, the latter reflecting a slight alteration from the first experiment—a Dutch instead of Russian surname.17 The photos featured the identical faces from the first experiment. The second factor involved the implication that immigrant workers are either high- or low-skilled. The first condition noted concerns about high-skilled job competitors and quoted high-skilled native workers who had witnessed colleagues losing jobs to immigrants. The second condition included similar content, except that the relevant natives and immigrants were low-skilled. The subtitle also made a reference to “high-skilled” or “low-skilled” immigrants. All subjects were told that immigration is increasing and that it will have negative consequences. We focus, in other words, on the threatening half of the design from the first experiment.18

The effects of the second experiment were more subtle than the first and harder to detect due to the smaller sample size. Manipulations of the news story did not significantly change policy preferences or requests for information, but they did affect the probability of sending an anti-immigration message to Congress. We focus on this instance of political action, since our goal is to see which threatening story subjects found most compelling and whether the same psychological mechanism is at work as in the first study. Figure 4 shows the results for perceived harm, anxiety, and sending a message to Congress. In the third panel, we see an interaction between ethnicity and skill level emerges for sending a message (F = 3.77, p < .054). When the immigrant story featured a Latino, the proportion of subjects who requested an anti-immigration message be sent to their congressperson increased from .11 in the high skill condition to .24 in the low skill condition. When the European cue was present, opposition was actually higher (.21) after reading about high-skilled rather than low-skilled immigrants (.16), though this difference is not significant.

Figure 4.

Impact of Skill Level Emphasis and Ethnic Cue on Perceived Harm from Immigration, Anxiety about Immigration, and Political Action

Note: The “baseline” control group mean values for each panel are 2.26 (top), 1.31 (middle), and 0.19 (bottom).

Is the interaction effect on behavior driven by a change in beliefs or emotions?Figure 4 tells a clear story. The middle panel shows a large interaction between ethnic cue and skill level on anxiety (F = 7.94, p < .005). The low-skilled, Latino immigrant cue boosts anxiety by .77 points on a 0–6 scale, compared to the high-skilled Latino cue. There is a tiny drop in anxiety from the high-skilled to the low-skilled condition when the immigrant is European. No interaction occurs for perceived harm, displayed in the top panel of Figure 4 (F = 0.23, p < .633). There is a modest counterintuitive main effect of ethnicity, however: immigration was more likely seen as having a negative impact when European, not Latino, cues were present (Mdiff= .38, F = 3.68, p < .057). Despite this, only the low-skilled Latino cue triggered anxiety.19 Consistent with expectations, anxiety and not perceptions again mediate the impact of the news on political action (see Table 3: Mediated Model 1).

Table 3.  Anxiety but Not Perceived Harm Mediates the Interaction Effect of Ethnic Cues and Skill-Level Emphasis on Anti-Immigration Behavior (Experiment 2)
 Effects on Potential MediatorsEffects on Sending an Anti-Immigration Message to Congress
Anxiety about ImmigrationPerceived National Harm of ImmigrationPerceived Personal Harm of ImmigrationUnmediated ModelMediated Model 1Mediated Model 2Mediated Model 3
  1. Note: Coefficients (and standard errors) are estimates from ordinary least-squares regression (for Anxiety and Perceived Harm), and maximum likelihood estimation of probit models (for Sending an Anti-Immigration Message). All models include controls for education, income, age, gender, and ideology.

  2. *p < .10, **p < .05, ***p < .01 (two-tailed).

Skills Emphasis (low)–– 0.58– 0.54– 0.61– 0.67*
Ethnic Cues (Latino)– 0.05– 0.110.01– 0.56–0.48– 0.47– 0.67*
Cues × Skills Emphasis0.20***0.04– 0.041.01**0.711.08**1.17**
Anxiety about Immigration    1.86***  
Perceived National Harm of Immigration     1.28*** 
Perceived Personal Harm of Immigration      1.69***
Sobel Test (z)    2.16**0.43– 0.49
R2 (adjusted)    
% Predict Correctly   81.381.381.378.8
% Improved– 0.8

One possible objection to our findings about the mediating role of anxiety is that the measures of perceived harm ask about the likely impact on Americans in general. Perhaps anxiety captures the subject's belief that immigration is likely to affect her personally. Fortunately, we can address this concern with data from the second experiment. Subjects were also asked how likely it was that immigration would have a “negative financial impact on you or your family” and/or have a “negative impact on the way of life in your community.” We combined responses to form a measure of perceptions of personal harm (α= .72). Manipulations of the news article produced neither main nor interactive effects on perceived personal harm, and this holds for both the combined scale and the individual items (p > .25 for all factors). As a result, the mediating role attributed to anxiety cannot be explained by changes in beliefs about personal harm any more than by changes in beliefs about national harm (see Table 3: Mediated Models 2 and 3).

The second experiment also supports our principal hypotheses: immigration opinion depends on the immigrant group made salient, and the effects of these group cues on political behavior are mediated by anxiety. However, it also supports the corollary prediction that outgroup cues generate greater anxiety when they are consistent with stereotypes. The power of Latino group cues to trigger anxiety does not stem simply from ethnicity, or from the implied skill level of immigrants, but rather from a combination of these factors. News about immigration increases anxiety and provokes greater opposition among white Americans when it highlights low-skilled Latino migrants and emphasizes the negative consequences of immigration. In the presence of reassuring reports inconsistent with the stereotype, beliefs may change, but anxiety, attitudes, and behavior do not.


We find substantial support for two key propositions: (1) group cues in immigration discourse can elicit anxiety and (2) changes in anxiety, not perceived threat, mediate the impact of these cues on opinion and political behavior. Conventional wisdom suggests that opposition to immigration grows with the severity of the harm immigrants pose. Another popular but untested suspicion is that such opposition depends on the identity of salient immigrant groups. Our results show that, while news emphasizing the costs of immigration boosts the perception that immigration is harmful, ethnic cues strongly condition emotional reactions to this news. Stigmatized outgroups, in this case Latino immigrants, trigger negative emotions when costs are emphasized. This emotional trigger matters. Even when citizens in our study perceived the harmful consequences of immigration identically, hostile attitudes and actions flared only when group cues elicited anxiety.

Previous scholarship has shown how group attitudes are correlated with immigration opinion, but no study to date has shown so clearly that racial cues powerfully alter immigration opinion. More importantly, perhaps, we show how group cues seem to have this effect. Two explanations are possible: (1) group cues change beliefs about the severity of the immigration problem; or (2) group cues trigger emotions, and these emotions (not beliefs about severity) drive opposition. In our study, perceptions of the likely harm from immigration were quite responsive to news emphasis on costs or benefits, but perceptions alone were not sufficient to move opinions or provoke political action. This finding held regardless of whether the threat was perceived as economic or cultural in nature, or as personal or collective in scope. In contrast, group cues and the emotions they triggered proved critical to changing attitudes and behavior. Citizens felt more threatened by Latino immigration, not European immigration, and this feeling triggered opposition to immigration and multilingual laws, prompted requests for information, and led people to send anti-immigration messages to Congress. Although group cues also changed beliefs on occasion (e.g., about the relative size of immigrant groups), those changes did not account for effects on policy opinions or political behavior.20

Our results also have broader implications for research on the roles of both groups and emotions in politics. For example, the findings bring new questions into view for the study of group politics. First, how does the emotional impact of group cues fit with their role in activating prejudice? We know subtle group cues can affect opinion by increasing the salience of group-related dispositions (Mendelberg 2001; Nelson and Kinder 1996). However, scholars have found that anxiety-eliciting stimuli tend to decrease the salience of predispositions in judgment (Brader 2005; Marcus, Neuman, and MacKuen 2000), while group priming seems to occur via heightened accessibility of group schemas in memory (Valentino, Hutchings, and White 2002). Second, how does individual variation in predispositions such as symbolic racism (Sears 1988) moderate emotional reactions to cues such as the ones in our immigration stories (cf. Marcus et al. 1995)? While at first blush one might guess that those high in racial animus might react more negatively to the cue and therefore be even more powerfully affected by the story, it is not necessarily so. If racial animus is a product of previous, recurring negative emotions toward a group, additional exposure may neither boost the emotion above that already elevated baseline, nor have any added impact on opinions or behavior. More research on the emotional impact of groups promises to further our understanding of the socialization and activation of racial predispositions.

In particular, future work needs to clarify how group cues trigger emotions. Psychologists have begun to study emotional reactions to groups (Mackie and Smith 2003) but rarely examine the ways groups get invoked or implicated subtly in public discourse. In the present studies, anxiety from Latino cues embedded in the news seemed to evaporate when those same cues were accompanied by unexpectedly reassuring information about the benefits of immigration or the higher skill status of migrants. Why? We speculate that anxiety arousal may hinge on stereotypic depiction of the group. When stereotypes are undermined, the emotional impact wanes. When details are simply absent, people “fill in” the missing pieces with negative stereotypes (Gilliam and Iyengar 2000), and the emotional impact remains. An alternative explanation is that Latino cues did unconditionally elicit anxiety, but we did not detect it because the available information was immediately reassuring. In this more complicated account, the lack of a corresponding positive shift in opinions might be explained by the asymmetric impact of negative and positive information (Holbrook et al. 2001).

What do our findings have to say about theories of emotion and politics? Although our study was not designed to adjudicate among competing theories of emotion, our results are broadly consistent with tenets of at least two leading approaches. In keeping with the expectations of Hot Cognition theory (Lodge and Taber 2005), we find that affective processes, rather than conscious beliefs, guide political reasoning. All three types of outcomes we examined—policy opinions, information seeking, and political action—moved in a direction consistent with the negative affect that was triggered (i.e., in an anti-immigration direction). In line with Affective Intelligence theory (Marcus, Neuman, and MacKuen 2000), we find asymmetric effects for positive and negative emotions and, specifically, that anxiety facilitates opinion change in the direction of currently available information. Similarly, we find that anxiety is positively associated with threat-relevant information seeking. Both theories predict that emotions provide an “essential motivational thrust” to carry out one's goals (i.e., act on one's attitudes; Lodge et al. 2006, 30; Marcus, Neuman, and MacKuen 2000), and we indeed find that anxiety-eliciting news not only changed opinions, but also mobilized opposition into political action.21

We also contribute to research on emotion and politics in significant ways. While recent work helps to establish the causal role of anxiety (Brader 2006), our study verifies that anxiety actually mediates the effects of threatening political stimuli. Moreover, we demonstrate the impact of anxiety on actual behavior, rather than self-reports. Our findings underscore the importance of emotion and yet depart from recent research emphasizing its beneficial effects. One implication of the present study is that citizens' responses to new information may be held hostage to the past, in the form of anxiety induced by outgroup images that may or may not hold contemporary relevance. Emotions serve an adaptive function in alerting us to important changes in our environment (Lazarus 1991), but this hard-wiring can leave us overly responsive to irrelevant symbols and too weakly responsive to what has changed. Our results also illustrate the way in which the outcome of the anxiety-stimulated judgment process is conditional on the information available. Structural biases in public discourse that lead some consequences, characteristics, or groups to be emphasized over others may be picked up and amplified by anxious citizens. Candidates or interest groups can exploit vulnerabilities for political gain, but journalists also may be party to influencing the public in this manner, driven solely by news imperatives or their own stereotypes (Gilens 1999).

In order to shed light on the mechanisms underlying dynamics in public opinion, the present study follows earlier work by using experiments to “mimic elite discourse” (Kinder and Sanders 1996). Our findings have serious implications for public debates over policy issues. Political rhetoric about public policies is often laden with references to religious, national, and racial groups. When cues about stigmatized outgroups are available, as they often are, political elites can imbue information about the issue with emotional significance for the audience. To understand the impact of elites on mass opinion, one must attend to the power of feeling, not just seeing, threats.


Question Wording

EMOTIONS. “Now, moving on, we would like to know how you feel about increased immigration. The following questions will ask you how you feel when you think about the high levels of immigration to this country. How [anxious (that is, uneasy)/proud/angry/hopeful/worried/excited] does it make you feel?” (Very, somewhat, a little, or not at all?)

BELIEFS. Perception of National Harm: “In your opinion, how likely is it that immigration will have a negative financial impact on many Americans?”“Now setting aside any economic concerns, in your opinion, how likely is it that immigration will have a negative impact on the way of life in many American communities?”Perception of Personal Harm: “How likely is it that immigration will have a negative financial impact on you or your family?”“Again setting aside any economic concerns, how likely is it that immigration will have a negative impact on the way of life in your community?”

OPINIONS. Immigration: “Do you think the number of immigrants from foreign countries who are permitted to come to the United States to live should be increased a lot, increased a little, left the same as it is now, decreased a little, or decreased a lot?”English Only: “Do you favor a law making English the official language of the United States, meaning government business would be conducted in English only, or do you oppose such a law?”

BEHAVIOR. Information Seeking: “There are many organizations that put out information on immigration. Would you be interested in receiving more information about this issue from any of the following types of groups?”Contacting: “There are currently several proposed laws before Congress that would substantially decrease the number of legal immigrants that come to this country. Would you like us to send an email message to members of Congress indicating either your support or opposition to these reforms?”


  • 1

    Unlike periodic open-ended polling by Pew on the most important problem, entrance and exit polls offer voters a choice among a small number of issues. Pew data are publicly available on their website (http://people-press.org), while marginal results for exit polls can be found on the sites of most national news outlets (e.g., see http://politics.nytimes.com/election-guide/2008/index.html).

  • 2

    According to information provided by Knowledge Networks, participation rate in their panel at the time was 56%. The completion rate among panelists contacted for our survey was 77%.

  • 3

    The complete text of the stimulus materials is available upon request.

  • 4

    Subjects in the control group read an article on cell phone use and car crashes. In our analyses, we determine effects by running analyses of variance (ANOVA) on the 2 × 2 factorial design of the treatments; the control group is not included in those analyses. We provide control group means in the figures as a reference for “baseline” levels of opinion or behavior.

  • 5

    Cell sizes: 72 (control), 73 (neg. Lat.), 72 (neg. Eur.), 69 (pos. Lat.), 68 (pos. Eur.).

  • 6

    The mean rating for “how ethnically Hispanic” (1–5 scale) the man appears is 3.9 for the Latino photo and 1.0 for the European photo. This difference is highly significant (t = 8.40, p < .001). The mean rating for “ethnically European” is 1.4 for the Latino photo and 4.4 for the European photo. This difference is also highly significant (t = 9.72, p < .001). On the four other trait dimensions, the difference in mean ratings did not exceed 0.4 points nor approach statistical significance.

  • 7

    All reports of statistical significance are based on two-tailed tests and controls for education, age, income, and sex. Despite randomization, the distribution across cells is a bit uneven (e.g., subjects in the positive European cell are slightly better educated than those in other cells (t = 2.13, p < .03)).

  • 8

    If the effect of the news frame is analyzed separately for both versions, the boost in opposition falls short of significance when paired with the European immigrant cue (t = 1.49, p < .139) and is strongly significant when paired with the Latino cue (t = 3.49, p < .001).

  • 9

    This possibility merits future investigation. People may seek information because they need it, like it, or find it novel or surprising. We also checked whether information seeking, especially in the positive frame or European cue stories, might be tied to enthusiasm, but it is not.

  • 10

    This is consistent with most work on Affective Intelligence theory (Marcus et al. 2000). However, our battery is not well suited to detecting anger as distinct from anxiety because it only contains one anger term (Marcus et al. 2006). Thus, one could reinterpret the scales as negative and positive affect, rather than as anxiety and enthusiasm specifically. We did not design the study to test competing accounts of emotional structure, so factor analysis cannot definitively adjudicate between one-, two-, or multidimensional models.

  • 11

    In principal factors analysis, “excited” did not load as strongly on the enthusiasm factor and seemed to tap general arousal. We dropped it, though its inclusion does not change our findings.

  • 12

    The effect is identical for anger (F = 5.59, p < .019) and anxiety (F = 5.52, p < .020).

  • 13

    The difference in results for beliefs and emotions cannot be explained by a ceiling effect for perceived harm (the scale is 0 to 6) nor by the distribution of responses (the standard deviation accounts for a nearly identical proportion of each scale, differing by less than one percent).

  • 14

    Similarly, enthusiasm neither meets the conditions for mediation nor passes a Sobel test.

  • 15

    There is no right answer. But, if one looks at foreign-born U.S. residents in 2000, 51% were from Latin America and 14.3% were from Europe (U.S. Census Bureau 2001).

  • 16

    Subjects in the control group read an article about fast-food chains.

  • 17

    The switch in nationality offers an opportunity to see how sensitive effects are to the European nationality with which Latinos are contrasted. The Dutch were part of more “traditional” waves of immigrants to the United Sates. The cue is also more subtle in this study, as neither the immigrant nor his nationality are mentioned in the body of the story. The photo, however, remains prominent.

  • 18

    Cell sizes: 42 (control), 47 (low sk. Lat.), 44 (low sk. Eur.), 44 (high sk. Lat.), 43 (high sk. Eur).

  • 19

    The fact that European cues increase perceived harm seems tied to education and income. As education decreases, for example, the difference drops until perceived harm is higher for Latino cues. However, the interaction effect for anxiety persists regardless of education or income.

  • 20

    At some level, “beliefs” about harm and “feeling threatened” must be related. But how? Offering a complete answer to that question would require access to more intrusive types of data. At the nonconscious level, we know situations can trigger anxiety only when the brain perceives potential harm; yet emotions often occur without conscious awareness and thus without triggering changes in beliefs that would be detected in a self-report (Damasio 2000). Emotions and beliefs likely exert reciprocal influence, especially by the time they manifest themselves consciously (Lazarus 1991; LeDoux 1996).

  • 21

    Although there is considerable overlap in how these two theories conceptualize emotion, they do posit some distinct mechanisms and effects. First, while each emphasizes a single basic negative and positive emotion, Affective Intelligence theory predicts a third emotion, anger, will sometimes emerge with consequences distinct from anxiety (Marcus et al. 2006). Second, while Hot Cognition theory expects both positive and negative affect to prime emotionally congruent considerations (Lodge and Taber 2005), Affective Intelligence theory predicts that enthusiasm and anxiety trigger distinct modes of judgment in which either reliance on habit or effortful processing of current considerations holds sway (Marcus et al. 2000). Third, whereas Hot Cognition theory predicts that negative affect is a source of bias, Affective Intelligence theory predicts that anxiety promotes less biased information seeking.