Political Behavior

, Volume 33, Issue 3, pp 433–454

Framing Labels and Immigration Policy Attitudes in the Iowa Caucuses: “Trying to Out-Tancredo Tancredo”


    • Department of GovernmentCentre College
  • David P. Redlawsk
    • Eagleton Institute of PoliticsRutgers University
  • Howard Sanborn
    • Department of International StudiesVirginia Military Institute
Original Paper

DOI: 10.1007/s11109-010-9141-x

Cite this article as:
Knoll, B.R., Redlawsk, D.P. & Sanborn, H. Polit Behav (2011) 33: 433. doi:10.1007/s11109-010-9141-x


We use an experiment built into a series of surveys of Iowa voters during the 2008 Iowa Caucus campaign to test the effect of differing group framing labels on immigration policy preferences. We find that certain framing labels matter, but only among Republican partisans for whom the immigration issue is important. We also find that issue importance produces more conservative policy preferences for Democrats as well as Republicans. We examine and discuss these results as well as their implications for the immigration debate, the interaction between issue salience and policy preferences, and the theory of political framing in general.


Immigration attitudesFraming labelsEthnic cuesIowa Caucuses


In 2008, Iowa took its prominent place once again as the first battleground of that year’s presidential election. For two years prior, Republican and Democratic candidates had tromped through the state, campaigning in pursuit of their party’s nomination. Democrats challenged the Bush policy on the war in Iraq, while Republicans re-iterated their support for anti-terrorism measures and lower taxes.

Throughout the campaign, candidates offered their positions on illegal immigration and how the government should treat undocumented aliens already living in the United States. Republican candidates varied dramatically in their stances. Congressman Tom Tancredo, for example, focused his entire (short-lived) presidential campaign on the dangers of illegal immigration, arguing that “the great tradition of the melting pot in America is not working. The melting pot is cracked.”1 Other Republicans, like John McCain, advocated for a “path towards citizenship” for those already in the country, although he eventually altered his position later in the election season. In general, however, Republican candidates like Tancredo and Fred Thompson tried to out-do each other with their focus on border security and taking a hard line with the illegal immigrants already in the country. In contrast, Democratic candidates avoided taking a firm stand on policy toward illegal immigrants when they could and, for the most part, were not pressed to do so by Iowa Democratic caucus-goers. Generally, Democratic candidates favored immigration reforms that would provide options for temporary guest worker visas and even a path through which an illegal immigrant could eventually become a U.S. citizen.

At first glance, Iowa might not seem a fertile ground for assessing views on illegal immigration, as the Latino population is relatively small, just under four percent. However, Iowa has one of the fastest Latino growth rates in the country; from 2000 to 2007, the Latino population grew by 45.2 percent. Subsequently, Latinos have become the single largest ethnic minority group in the state (Iowa Division of Latino Affairs, 2008). In some small cities, Latinos suddenly constitute a substantial portion of the population. For example, the small south-eastern Iowa towns of Conesville, West Liberty, and Columbus Junction are 59 percent, 41 percent, and 39 percent Latino, respectively. Overall, the Latino population is projected to reach nine percent of the state population by 2030.

We use this unique opportunity, then, to study how attitudes toward undocumented immigrants evolve in a state where citizens are just beginning to recognize this widespread change in demographics. Since many of the candidates use Iowa as springboard into the primary season, caucus-goers should be well-exposed to campaign rhetoric generally; those concerned about illegal immigration and the influx of non-whites to the state should be attuned to the candidates, in particular. To assess the effects of this environment, we utilize data collected during the 2007–2008 campaign season not only to analyze the policy preferences of Iowa caucus-goers but also to test how these voters responded to different issue frames presented throughout the campaign season. While the majority of respondents share a similar preference, we find that those most likely to be receptive to negative rhetorical framing labels are also more likely to choose immigration policy options that are more punitive than their counterparts. This effect is exacerbated dependent upon how the immigration issue is framed.

Elite Discourse and Public Opinion

How voters perceive an issue such as immigration policy is conditioned by the ways in which the issue is framed by elites. Such “frames in communication” (Chong and Druckman 2007b) are invoked by leaders to encourage citizens to think about an issue in particular ways; to develop certain “frames in thought” (Chong and Druckman 2007b) which lead to linkages between sets of values and policy issues, resulting in a framing effect (Druckman 2001). The particular words that elites use to describe a policy may lead to citizens “calling attention to some matters while ignoring others” (Iyengar and Kinder 1987). Nelson and Oxley (1999) argue that framing affects the importance people place on specific beliefs about an issue. The use of frames does not necessarily change some underlying existing belief, but instead changes the mix of considerations likely to condition citizens’ responses (Zaller 1992). Some frames are stronger than others, and elites seek out the frames to give them a competitive advantage in the marketplace of ideas. Druckman (2004) and Chong and Druckman (2007a, b) term this process “competitive framing.”

Framing can occur by using images, metaphors, analogies, and other methods for casting a policy in a particular light or associating it with other issues or concepts. When an issue is framed in more acceptable or positive terms, the public may consequently take a more favorable view of the issue itself, policy options surrounding it, or even a person or group referenced by the frame. For example, Nelson et al. (1997) conducted an experiment measuring how evaluations of a Ku Klux Klan rally varied depending on how the rally was framed. Those exposed to news stories presenting the Klan rally as a free speech issue were more likely to view the Klan itself favorably than those exposed to stories framing the rally as a public order and safety issue. Barreto et al. (2009) show similar results in framing candidate Barack Obama. Voters responding to a frame that focused on Obama’s Christian roots were more positive towards him than those who were not exposed to the frame, while a Muslim frame depressed evaluations. These effects, however, appeared only for those who accepted each frame’s premise that Obama could sympathize with the Christian or Muslim community.

As Barreto et al. (2009) find, not all frames are readily accepted by everyone who encounters them. Brewer (2001), in particular, argues that at least some citizens are active processors of frames and do not blindly accept them. Some may well resist particular frames while accepting others. Peffley and Hurwitz (2007) reinforce this point that citizens’ predispositions on particular issues can interact with framing statements to condition responses. For example, given the partisan nature of most framing (Lakoff 2004), we would not be surprised to find party preferences playing an important role in citizens’ responses to frames. Decades of research has shown that partisan beliefs act as a perceptual screen (Campbell et al. 1960), and more recent research on motivated reasoning (Taber and Lodge 2006; Redlawsk 2002; Redlawsk et al. 2010) leads us to expect that as frames in communication are received and processed, partisans will respond differentially to them.

Likewise, whether an issue is salient to an individual when the frame is received may condition how the frame is processed. We would expect those who care more about an issue to be more affected by frames about it, as their existing mix of considerations on the issue is likely to be deeper (Zaller 1992). Also, previous research on “core values” has shown that value hierarchies can serve to moderate opinions on issues under different framing conditions (Brewer 2001; Craig et al. 2005). Because issue salience can derive from one’s value hierarchies, we might expect that it can also serve to condition the effect of frames on people’s attitudes.

The use of framing has been evident in recent elite discourse on immigration. Over the course of the 2006 midterm and 2008 primary campaigns, commentators noticed a distinct pattern in how immigration was discussed. Republican politicians, favoring a more punitive approach to immigration reform, tended to refer to unauthorized residents of the United States as “illegal immigrants” or “illegal aliens.” Democrats, on the other hand, were more likely to use the phrase “undocumented immigrants” or “undocumented workers.” By employing different labels, these political elites were attempting to frame the issue to elicit support for their position. As a National Public Radio report explained: “If you call them ‘undocumented workers’ it’s not fearful – these are just people who lost their papers. If you call them ‘illegal aliens’ it’s something that’s frightening [and] concerning” (Martin 2006).

Labels, Groups, and Policy Preferences

It is evident that the immigration issue has been framed both positively and negatively by labeling the immigrants themselves. Why might such framing affect related policy attitudes? One possible explanation derives from social identity theory (Tajfel and Turner 1986) which posits that individuals sort the world into “in-groups” and “out-groups.” Threats from out-groups serve to strengthen in-group favorability. When certain groups are framed using language that reminds in-group members of the other’s out-group status, the out-group status is accentuated, increasing the salience of in-group commonality. Using labels such as “Mexican” or “illegal” to refer to unauthorized Latino residents of the United States could serve to increase the salience of their out-group status. This, in turn, could increase the importance of defending the dominant status of the majority in-group (Blumer 1958), leading to a greater likelihood of opposing reform measures that make it easier for members of the out-group to incorporate themselves into the in-group (like a path to citizenship for undocumented immigrants).

Alternatively, when the out-group status of the immigrants in question is emphasized by the use of a negative framing label, it may simply serve to increase the already present out-group disfavor that occurs automatically as part of the social identity hypothesis. When out-group disfavor is strengthened via the negative framing label, individuals may be less likely to support policies benefiting the disfavored out-group. The key difference between this line of theorizing and the first is that there is no dependence on a perceived threat to in-group incorporation. Although these theoretical mechanisms differ, the expected result would be the same.2

Existing research supports this hypothesized effect of framing labels on immigration policy attitudes. Augoustinos and Quinn (2003), for instance, examined attitudes toward immigrants in Australia. They found that Australian attitudes on the rights and privileges of immigrant groups were affected by whether or not the target group in question was referred to as “asylum seekers,” “refugees,” or “illegal immigrants.” From an American perspective, McAdams et al. (2008) report a study similar to ours using a survey experiment in the state of Ohio during the 2006 midterm campaign season. They show that framing the unauthorized Latino out-group in terms of “undocumented aliens” or “foreign workers” increased respondents’ likelihood of expressing support for a policy of mass deportation. However, this framing study suffered from a limited range of policy options—respondents could only agree or disagree with the policy of deportation.

Additionally, Brader et al. (2008) examine how preferences on immigration levels and Official English policies vary within a variety of news media frames. Using a series of controlled experiments, they introduce subjects to three distinct framing manipulations: (1) emphasizing either the economic cost or benefit of immigration to the United States, (2) featuring either Latino or European immigrants as the subjects of the news story, and (3) portraying the immigrants as either high-skilled or low-skilled workers. They find that the ethnic cues manipulation (Latino vs. European) produced the greatest change in reported policy preferences and that this effect is due to an emotional response. Cues based on ethnicity are more likely to produce anxiety (thus affecting immigration policy preferences in a more conservative direction) than cues based on economic costs or skill levels.

We seek to build upon and improve these findings in several ways. First, we provide a broader range of immigration reform policy options than were provided by McAdams et al. (2008) from the most lenient, allowing illegal immigrants to stay in the country with no consequences, to the most punitive, that of mass deportation. Second, while Brader et al. (2008) tested the effect of news media framing in a controlled experimental environment, we conduct a survey experiment during a heavily contested political campaign in Iowa. This allows us to directly test the effect of frames in a political campaign environment where the nature of the campaign rhetoric suggests that immigration was a salient issue.

Iowa and Immigration

Iowa makes for a unique testing ground of voter attitudes toward illegal immigrants, given the rapidly-growing Latino population. As recent scholarship demonstrates, non-Hispanic white attitudes toward immigrants are not driven by the size of the immigrant group in question, but rather by the rate at which that immigrant population is growing (Hopkins 2009). With the high growth rates in places like South Dakota, New Hampshire, Utah, and Maine, we may be able to apply our findings from this study of Iowa to other cases of states in the midst of a rapid, demographic shift.

Moreover, immigration was one of the most frequently-discussed issues during the 2008 Iowa election season. During the campaign, then-President George W. Bush had proposed a comprehensive immigration reform plan that allowed for a path to citizenship for illegal immigrants living in the United States. This legislation received a great deal of attention in the national media and was widely discussed by pundits, candidates, and elected officials before it was soundly defeated in Congress. Even local concerns about immigration surfaced; federal immigration enforcement agencies conducted a series of immigration raids at agricultural plants in the Iowa towns Marshalltown (2006) and Postville (2008).

Consequently, Iowans who supported the Republican Party consistently rated immigration as one of the most important issues to their vote in the presidential caucuses. Roughly 62 percent of likely caucus-goers rated immigration as either the most important issue or a very important issue during the campaign in our March 2007 survey of caucus-goers. As the campaign season progressed the issue became more important to Republicans; by October, 69 percent considered it either very important or the most important issue to their vote. Of those Republicans actually attending the caucus in January, just less than 54 percent felt the same. On the Democratic side, the issue was not quite as important, though many Democrats still cited it among their top voting issues (40 percent in March, 35 percent in October, and 23 percent at the caucus). In any case, this interest prompted Republican candidates, in particular, to work immigration reform into their stump speeches and to field many questions about it during town hall meetings (Kahn 2007).

For these reasons, the issue of illegal immigration was highly salient in the minds of Iowans during the campaign season, making Iowa an excellent environment in which to test the effect of framing on immigration attitudes. The confluence of a number of factors allowed for a unique and timely study of attitudes: the increasing salience of illegal immigration to most Americans, the prominence of Iowa in a campaign season that featured highly competitive contests for both parties’ nominations, and the prevalence of more than one elite frame for the issue of unauthorized immigration. Thus, we proceed to state our expectations for our study before offering the results of our analysis.


We contend that political candidates employ negatively-charged out-group labels like “illegal immigrant” to frame the immigration issue both to appeal to fervent supporters of deportation and to lead people to support more punitive immigration policy preferences. Brader et al. (2008) find experimental support for such candidate beliefs in showing that framing immigration using explicit ethnic cues causes people to be more conservative on immigration. We thus conducted two separate survey experiments to test these effects. One experiment (conducted in January 2008 on actual Iowa caucus-goers) sought to emulate the differences in elite discourse by presidential candidates. We modified the object of an immigration policy question on respondents’ preferred policy options in dealing with those already in the country illegally to refer either to “undocumented immigrants” or “illegal immigrants.” (Complete question wording for these experiments is available in Appendix A).

The other experiment (conducted in one of our pre-caucus surveys in October 2007) sought to test the effect of explicit ethnic cues. We asked likely caucus-goers the same policy question, but this time framed the object as either “undocumented immigrants” or “undocumented Mexicans.” It is important to note that these frames vary descriptions only of the targets of immigration policy (the immigrants), not the immigration policy options themselves.

For reasons discussed previously, we expect that when framing immigration policy alternatives in terms of either “illegal immigrants” or “undocumented Mexicans” instead of “undocumented immigrants,” individuals will be more likely to support more punitive immigration reform policies. This leads us to the following simple hypothesis:


Those who are exposed to the term “illegal immigrants” or “undocumented Mexicans” when asked about immigration reform preferences will be more likely to indicate support for more hard-line preferences, such as mass deportation over earned citizenship, than those who are exposed to the term “undocumented immigrants” when asked about their immigration reform preferences.

The implication of this hypothesis, then, is that the effects of the framing experiment are consistent across all demographic characteristics and partisan preferences. However, as was noted earlier, Republican candidates were far more likely to support punitive measures on the campaign trail than their Democratic counterparts. This is puzzling because, as reported by a number of polls in 2007, the most popular policy preference for both Republicans and Democrats was conditional citizenship (see, for example, Kiely 2007). If some Republican candidates voiced a strong preference for deportation while the majority of potential caucus-goers supported a much less punitive option, then we must consider why these candidates acted as they did. Looking at the Republican campaigns in Iowa leads us to suspect candidates did so to appeal to specific blocs of likely caucus-goers who would support them in January 2008. These voters might not comprise the majority of the party, but would be highly motivated to attend the caucus and vociferously support the candidate who best defended their views on a particularly resonant issue. We suspect that voters who do not give much thought to immigration policy (or at least, see it as unimportant) will be relatively unaffected by the frame as it does not cause them to process it with any depth, while those for whom the issue is important will respond even more strongly to cues about the ethnicity and legal status of these immigrants. This leads to:


Respondents who view immigration as important will be more likely to choose more punitive policy preferences when exposed to the term “illegal immigrants” or “undocumented Mexicans” than respondents who view immigration as unimportant.

In addition, given the partisan nature of both the immigration issue itself as well as the structure of the rhetoric on it in Iowa, Republican voters might be more likely to respond to the frames than Democratic voters, and Republicans for whom the issue is important might respond most strongly, which would help explain the observed rhetoric of Republican candidates.


Framing effects will be stronger for Republicans who consider the issue important. These Republicans will be more likely to indicate support for punitive policy preferences such as deportation when exposed to the term “illegal immigrants” or “undocumented Mexicans” than Republicans for whom the issue is unimportant as well as all Democrats.

Data and Methods

Our data come from a series of public opinion surveys of likely Iowa caucus-goers fielded by telephone in the months of March, August, and October 2007 in the middle of the intense presidential campaign season that characterized the 2008 Iowa Caucus contests. Additional data were collected from caucus attendees following the January 3, 2008 caucuses. All four surveys followed the same basic approach and were approximately 20 min in length. While we randomly selected callers of all races and ethnicities, we limit our analysis to non-Hispanic white respondents since the vast majority of the respondents in our surveys were white and because we are interested primarily in white attitudes on a racial policy preference.

Data for the analysis of framing effects comes from both our October 2007 survey of likely caucus-goers, with 793 respondents participating in the “Mexican” frame experiment, and our January 2008 survey of actual caucus-goers, with 496 respondents participating in the “illegal” frame experiment. Our dependent variable is a four-point ordinal variable constructed from the question asking respondents to select from among four policy options regarding illegal immigrants already living in the country. This variable was re-coded so that the more punitive policy preferences (i.e. mass deportation) correspond with higher values. We can safely operationalize this variable in an ordinal manner because immigration policy preferences can be conceptualized on an ideological spectrum from more liberal (“amnesty”, “path to earned citizenship”) to more conservative (mass deportation) as previous scholarship on racial policy preferences demonstrates (Kinder and Sanders 1996; Henry and Sears 2002).

We control for the effects of a number of other variables that have been shown to exert influence on individual immigration policy preferences. Family income and labor union membership are included to account for the economic competition hypothesis (Citrin et al. 1997; Alvarez and Butterfield 2000) which posits that anti-immigrant hostility is driven by competition over scarce economic resources like jobs. We also include the percent Latino residing in the respondent’s zip code to control for the effect of social context on a respondent’s out-group attitudes and related policy preferences. Theoretically, a higher percentage of Latinos could either lead to more conservative (Giles and Buckner 1993; Tolbert and Grummel 2003) or more liberal (Welch et al. 2001; Oliver and Wong 2003) immigration policy preferences. Also included are controls for common socio-demographic characteristics that have previously been shown to exert effects on immigration policy preferences. These include education (Citrin et al. 1997), gender (Hughes and Tuch 2003), age (Wilson 1996), and frequency of church attendance (Knoll 2009).


Hypothesis 1

To directly assess the effects of our immigration frames, we use a two-sample proportion test to see if there is a significant change in the rate of response for a particular policy, given the frame. In our October 2007 study of likely caucus-goers, where we randomly inserted the ethnicity (“Mexican”) frame into roughly half of the policy questions, we find little, if any, effect (see Table 1). The difference in support for deportation does appear to trend toward Hypothesis 1, as those that received the ethnicity frame responded at a higher rate in favor of deportation than those who did not receive the frame (23.1 percent to 20.3 percent). However, this difference is not statistically significant. When we investigate support for conditional citizenship, we find more support for our hypothesis. There is a nearly 5 point decrease in support for conditional citizenship when the word “Mexican” is inserted; this difference is statistically significant, albeit only at the p < 0.10 level.
Table 1

Percentage support for government policy concerning undocumented immigrants among likely Iowa caucus-goers


Mar 2007

Aug 2007

October 2007

January 2008

"Undocumented immigrants"

"Undocumented Mexicans"

"Undocumented immigrants"

"Illegal immigrants"

Deport all







Temporary guest worker







Earned citizenship







Permanent resident/no requirements














Note: March, August, and October are “likely” caucus-goers; January is actual caucus-goers

Source: University of Iowa Hawkeye Poll Data, 2007/2008

Two-sample proportion test conducted between frames. At p < 0.10, there is a significant difference in support for earned citizenship with the “Mexican” frame in October 2007. There is a significant difference (p < 0.10) for the temporary guest worker and permanent residence policies in January 2008

Likewise, in our January 2008 study of actual caucus-goers, we find little support for the effects of the “illegal” frame on respondents. The difference in support for deportation among those who received the “illegal” frame (20.5 percent) and those who did not (21 percent) is minimal and certainly not significant. There is more of a difference in support for conditional citizenship (55 percent without the frame, 59 percent with the frame), but this is only a difference of about 4 percent and is also not significant. The shift in support for a temporary work policy is, however, significant (22 percent, 15 percent) but it is difficult to interpret this result given the lack of significance in the other categories.

The simple results displayed in Table 1 suggest that our framing of the issue of immigration was not very powerful. We do not see strong effects for either framing immigrants by their country of origin (“Mexican” vs. unspecified) or by their legal status (“illegal” vs. “undocumented”). We offer a number of explanations for this non-finding. First, it may simply be that our experimental manipulations were not strong enough to draw out our hypothesized responses, although there is evidence that frames in survey experiments are actually usually stronger than real world framing effects (Barabas and Jerit 2010). A second, and more likely, explanation is that the frames did not operate with equal impact across all respondents. It is quite possible that the demographic and attitudinal characteristics of participants in our survey interacted with the effects of the framing experiment to condition responses; this condition has been discussed in prior scholarship (Brewer 2001), but has rarely been tested by framing researchers (but see Barreto et al. 2009). Hypotheses 2 and 3 specify expectations that require us to go beyond the simplistic assumption of equal frame effects.

Hypothesis 2

We test our second hypothesis by building a model of attitudes towards immigration policy pooling the data from all four of our surveys.3 As noted above, we use ordered logistic regression because the policy options provided to respondents that constitute our dependent variable can be clearly ranked from least punitive (earned citizenship) to most punitive (simple deportation). Thus, not only can we assess the likelihood of a respondent preferring a particular policy, but also observe shifts in support along this scale.4 We include variables to control for a number of demographic and contextual factors, as well as indicators for the ethnicity frame and for the illegal frame which serve as a further test of whether either frame has a significant effect on policy preference. We also include a dummy variable to represent our January sample, which was taken the week after the Iowa Caucus and is the only post-caucus survey. The results are presented in Table 2.
Table 2

Ordered logistic regression of undocumented immigrant policy preference for likely Iowa caucus-goers, 2007–2008


Model 1

Model 2

Model 3

Model 4



Issue importance

0.988** (0.114)

1.011** (0.132)

1.521** (0.206)

0.591** (0.187)

Post caucus survey

0.386* (0.170)

0.389* (0.171)

0.423† (0.248)

0.278 (0.238)


−0.041 (0.054)

−0.041 (0.055)

−0.007 (0.077)

−0.074 (0.078)


0.066 (0.074)

0.065 (0.074)

0.032 (0.098)

0.092 (0.117)


−0.099† (0.055)

−0.097† (0.055)

−0.016 (0.079)

−0.184* (0.079)

Born-again Christian

0.037 (0.127)

0.036 (0.127)

−0.017 (0.159)

0.108 (0.223)

Church attendance

0.043 (0.105)

0.052 (0.105)

−0.001 (0.163)

0.055 (0.140)

Labor union membership

0.092 (0.123)

0.091 (0.124)

0.067 (0.249)

0.113 (0.138)


0.330** (0.109)

0.347** (0.109)

0.210 (0.152)

0.439* (0.159)

Partisanship (caucus)

−0.595** (0.122)

−0.593** (0.122)

Urban population

−0.203 (0.139)

−0.218 (0.139)

−0.316† (0.190)

−0.135 (0.211)

Latino population

0.378 (1.482)

0.315 (1.486)

2.476 (2.111)

−1.393 (2.163)

Ethnicity framing label (Mexican)

0.136 (0.153)

0.383† (0.218)

0.959** (0.330)

0.258 (0.294)

Ethnicity frame × importance

−0.474 (0.298)

−1.194** (0.406)

0.189 (0.509)

Illegal framing label

−0.099 (0.215)

−0.277 (0.260)

0.045 (0.415)

−0.378 (0.334)

Illegal frame × importance

0.428 (0.331)

−0.011 (0.474)

0.760 (0.509)






LR χ2





Log likelihood










† p < 0.10, * p < 0.05, ** p < 0.01, two-tailed

Dependent variable: Policy preference toward undocumented immigrants (see Appendix A for complete question wording and coding). Higher values correspond to more conservative policy preferences

Source: University of Iowa Hawkeye Poll Data, 2007/2008

Note: Given the size of the standard error for the Latino population coefficient, we also ran our models without including this measure in the analysis. Our findings remain the same, both in coefficient size and significance

In Model 1, we confirm our earlier finding of no direct framing effects. We do find support for several individual differences as explanatory factors in attitudes toward immigration policy. We also find that those who actually attended their caucus were somewhat more punitive than those in the pre-caucus sample. Overall, policy preferences seem to be conditioned primarily on issue importance and partisanship. So as not to overstate the results presented in Table 2, however, we acknowledge that the r-squared values of Model 1, as well as for all models discussed below, are relatively low.

Of course, in order to test whether the frames have differential effects across respondents we must specify a model that interacts the frames and our hypothesized factors. As per our previous review of the literature, we look first to issue importance as a likely factor that might condition framing effects. Issue importance is a highly significant predictor of policy preference in Model 1: individuals who view immigration as important are more likely to prefer punitive measures (i.e. mass deportation) compared to those who do not think immigration is important.

In Model 2, we test the interaction between our frames and issue importance. We continue to see strong main effects for issue importance on policy preferences, as well as individual difference factors, including gender and partisanship. And, as with the initial model, we see no effects for our “illegal” frame. It appears that the legality of the immigrants simply does not condition policy preferences. But the story is very different for the ethnicity frame. When we classify immigrants as “Mexican,” voters are more punitive in their policy preferences than when ethnicity is unspecified.5 Interestingly, while the main effect appears only when the interaction with issue importance is included, the coefficient on the interaction term itself just misses statistical significance. Thus, we find no direct support for Hypothesis 2.

Interpreting ordinal logit coefficients is intuitively difficult, so we present predicted probabilities for the main effects for the ethnicity frame, issue importance, and partisanship in Model 2 in Table 3. As the importance of the issue of immigration increases, the likelihood of selecting a more punitive policy preference also increases substantially, as we proposed in Hypothesis 1. That is, those for whom the issue is most important are more likely to be subject to framing effects. An individual who says that immigration is a highly important issue is more than twice as likely to select deportation as his or her policy preference compared to a respondent reporting that immigration is not important (28 percent vs. 12 percent). We also see that the ethnicity frame has a main effect as well. Respondents given the ethnic frame are more likely to be more punitive, all else being equal. The likelihood of selecting the most punitive response goes from 18 percent for those unframed to 24 percent for those told that the immigrants are “Mexican.”
Table 3

Predicted probabilities for undocumented immigrant policy preference


Blanket amnesty

Earned citizenship














Ethnicity frame






















Note: Table entries are percentages. Predicted probabilities generated from analysis of Model 2 in Table 2

Source: University of Iowa Hawkeye Poll Data, 2007/2008

Hypothesis 3

Our final hypothesis predicts that the effect of framing may depend on partisanship as well as issue importance. After all, our results clearly indicate that Republicans and Democrats differ on how punitive their policy preferences are. The coefficient for party in Model 2 is strongly significant and negative, suggesting that Democrats are much less punitive. Examining the predicted probabilities for party in Table 3 confirms this. Likely Republican caucus-goers were 9 points more likely to select deportation and 11 points less likely to select conditional citizenship than their Democratic counterparts.

As illustrated by our third hypothesis, we consider whether the importance of the immigration issue may be playing a role in conditioning partisan responses to immigration and to our framing experiments. Given that candidates have incentives to focus on voters they expect to turn out, and given that it is regularly assumed that those who attend the Iowa Caucus are more conservative (or liberal for Democrats) than those who do not, candidates may have focused their rhetoric accordingly, expecting it to resonate with those who care most about the issue. It could be that while Republicans and Democrats across the board might both be most likely to support conditional citizenship, Republicans who view immigration as highly important might be more likely to support punitive measures like deportation. This would then explain the behavior of Republican presidential candidates, who might see this passionate group as a key voting bloc during the subsequent election.

To consider this possibility we examine a three-way interaction between the frames, issue importance, and party. Because making sense of such three-way interactions is notoriously difficult, we re-estimate Model 2 (Table 2) separately for Republicans (Model 3) and Democrats (Model 4). We see immediately that issue importance is significant and positive for both groups of partisans, i.e. those who care more about the issue are more punitive in their policy preferences. We also confirm that receiving the illegal frame has no effect for members of either party, while the ethnicity frame has no effect on Democrats. For Republicans, however, the effect of the ethnicity frame and its interaction with issue importance are strongly significant, which we explore below.

Turning first to the main effects of importance, as displayed in Fig. 1, we see that Republicans who view immigration as highly important are much more likely to be punitive, while those Republicans for whom immigration is of low importance are much more likely to choose a relatively lenient policy option—conditional citizenship. Low-salience Republicans are actually just as likely as low-salience Democrats to choose conditional citizenship. Somewhat unexpectedly, we find a similar, though much less dramatic, pattern for Democratic respondents (Fig. 2). It appears that, for both parties, issue importance has a similar effect: if a respondent considers immigration of high importance, regardless of partisanship, he or she is more likely to hold a punitive policy preference.
Fig. 1

Predicted probabilities of undocumented immigrant policy preference by issue importance, for Republicans. Note: Predicted probabilities generated from analysis of Model 3 in Table 2. Pr(Amnesty) = Predicted probability of the amnesty policy preference. Pr(Guest) = Predicted probability of the guest worker policy preference. Pr(Earned) = Predicted probability of the earned citizenship policy preference. Pr(Deport) = Predicted probability of the deportation policy preference. Source: University of Iowa Hawkeye Poll Data, 2007/2008

Fig. 2

Predicted probabilities of undocumented immigrant policy preference by issue importance, for Democrats. Note: Predicted Probabilities generated from analysis of Model 4 in Table 2. Pr(Amnesty) = Predicted probability of the amnesty policy preference. Pr(Guest) = Predicted probability of the guest worker policy preference. Pr(Earned) = Predicted probability of the earned citizenship policy preference. Pr(Deport) = Predicted probability of the deportation policy preference. Source: University of Iowa Hawkeye Poll Data, 2007/2008

Both the importance of immigration and the described ethnicity of the immigrants play a role in conditioning Republican policy options, while the ethnicity frame has little impact on Democrats. In Table 4, we present predicted probabilities of issue importance for Republicans only, sorted by whether the respondent received the “undocumented Mexican” frame or the “undocumented immigrant” frame. Among low-salience Republicans, those that received the frame were 59 percent likely to support conditional citizenship. Those that did not receive the frame were 18 points more likely (77 percent) to choose conditional citizenship, all else being equal. Thus, framing immigrants as “Mexican” dampened support for conditional citizenship even among those for whom the issue was of low importance.
Table 4

Predicted probabilities for undocumented immigrant policy preference among Republicans


Blanket amnesty

Earned citizenship



Ethnicity frame













No frame











Note: Table entries are percentages. Predicted probabilities generated from analysis of Model 3 in Table 2

Source: University of Iowa Hawkeye Poll Data, 2007/2008

This effect is much more dramatic among those Republicans who view immigration as an issue of high importance. When the ethnicity frame was not presented to these respondents, they were 45 percent likely to choose conditional citizenship. However, when the frame was presented, likely Republican caucus-goers who were most concerned about immigration were only 25 percent likely to select conditional citizenship. This group was also more likely to choose punitive measures: the likelihood of selecting deportation rose from 34 percent for those who had not received the frame to 57 percent to those who had received the frame. To make this clear, we graph the predicted probabilities of deportation and earned citizenship policy preferences among Republicans in Fig. 3. While low-salience Republicans are still 10 percent likely to be punitive (and 18 points less likely to support earned citizenship when framed with the immigrant’s ethnicity), high-salience Republicans are more than 23 percentage points more likely to choose deportation. They are also 20 percent less likely to choose earned citizenship. We observe no similar effects for Democrats.
Fig. 3

Predicted probabilities of undocumented immigration policy choice for Republicans by frame and issue importance. Note: Predicted probabilities generated from analysis of Model 3 in Table 2; x-axis figures are percentages. Source: University of Iowa Hawkeye Poll Data, 2007/2008


These results have a number of important implications. First, immigration was a highly salient issue for many Iowans in the 2007–2008 campaigns. This is not overly surprising given both the environment of Iowa with its growing Latino population and the prominence of the immigration issue in the national environment during the 2005 to 2007 time period.

Second, a majority of Iowa caucus-goers, both Republican and Democratic, supported an earned path to legalization and citizenship for undocumented immigrants currently in the United States. These policy preferences, however, were conditioned upon how important the immigration issue was to the individual caucus-goer. In general, if an individual believed immigration to be an important issue, he or she was more likely to support more punitive policy options. This effect holds, generally, for both Republicans and Democrats; at least in Iowa, immigration policy attitudes are not primarily determined by partisanship. This echoes the findings of Nieman et al. (2006), who find the same effect in California: those for whom the immigration issue is important have more conservative policy preferences, regardless of partisanship.6 The fact that these results have been found in a diverse state (California) as well as a relatively homogenous state (Iowa) speaks to their generalizability.

Third, and perhaps most importantly, we find that not all frames are created equal; some have no effect while others may have their effects conditioned, even amplified, by respondent characteristics and attitudes. Contrary to our expectations, framing the immigration policy issue in terms of “undocumented” vs. “illegal” did not influence policy preferences among Iowa caucus-goers in the 2008 primary season. However, framing the issue with an explicit ethnic cue, i.e. “Mexican” vs. “immigrant,” did influence policy preferences in the expected direction, but only among Republicans, especially for those who viewed immigration as particularly important.7

On one hand, these results suggest that for all the effort that presidential candidates made to tailor their language on the immigration issue to their respective audiences, it produced little effect on the policy preferences of Iowa caucus-goers (with some important exceptions to be discussed shortly). Our initial thinking was that a sample of Iowa caucus-goers would be a good place to test such frames, since the Iowa Caucus is a high-information environment, with much discussion about immigration as the campaign progressed. It may instead be the case, though, that since Iowa caucus-goers are among the most politically aware and attentive voters in the country (Tolbert et al. 2009), there was less chance of “moving” respondents by framing experiments. This may especially be the case considering that the elite rhetoric of presidential candidates may likely convey implicit messages about the positive/negative effect of immigrants on American society, regardless of whether or not specific labels are employed.

On the other hand, these results also demonstrate that explicit usage of ethnic cues can elicit more conservative policy preferences among certain groups, specifically Republicans who care more about immigration. It is difficult in contemporary American politics to discuss the immigration issue using race-neutral language. Whereas immigrants are rarely referred to explicitly in racial or ethnic terms, it is often the case, whether intentionally or not, that the ethnicity of the immigrant group in question (predominantly Latinos) is implicitly understood by both the candidates and the mass public. This effect may very well transfer to other situations where the ethnic cues are implicitly understood, even if they are not explicitly given.

Subsequently, we recognize two potential implications of our work. First, framing labels exerted an effect on policy preferences, but only among a particular sub-group of the population. In this case, negative framing labels did not appreciably alter immigration policy preferences among any Democrats or even Republicans for whom the issue was not as important. This finding adds to previous research on the effects of framing experiments and the common assumption that frames have similar effects on all respondents who receive them (Brewer 2001; Peffley and Hurwitz 2007; Barreto et al. 2009). While random assignment assures that those receiving a frame and those not receiving the frame will be similar in individual characteristics, the frames may in fact exert a different effect on subsets of those receiving them, requiring additional analysis to tease apart framing effects when comparing to a control group.

Second, despite anecdotal claims that we are now in a “post-racial” America, ethnic cues persist in their effectiveness in shifting how people view certain contentious public policies like immigration. This finding adds support for those who argue that racial prejudice continues to play a significant role in the formation of racial public policy preferences (Kinder and Sanders 1996, e.g.). Indeed, our findings are contrary to the arguments of those who contend that it is not race, but rather principled ideological objections about the proper role of government that are at the core of racial policy attitudes (Sniderman and Piazza 1993, e.g.).

We, therefore, extend the work of Brader et al. (2008) who demonstrate that specific ethnic cues from the news media affect immigration policy preferences to a much greater degree than cues related to economics or immigrant skill level. We find that ethnic cues, conditioned by a number of factors, do indeed affect policy preferences. Issue salience, particularly among Republicans, appears to have been one of these key characteristics; apropos, given the highly politicized campaign environment, the sharp growth in the Iowa Latino population, and the intense elite focus on the immigration issue. However, the conditioning effects hold only for a small, though vociferous, group of caucus-goers. Thus, despite the efforts of presidential candidates to strategically reference undocumented immigrants in an effort to curry political favor, the results of this study suggest that it may have made little difference to the policy preferences of most Iowa caucus-goers.


The subtitle of this paper is a reference to a claim that Tancredo made on November 28, 2007 that his fellow Republican contenders were trying to “out-Tancredo” him in their efforts to be perceived as the most hard-line on immigration policy (Boston Globe, November 29, 2007).


Beyond social identity theory, framing the policy target as “Mexican” compared to say, “Asian,” might lead to more support for deportation simply because Mexico is geographically closer to the United States, making it a more direct and less-expensive process to deport Mexican immigrants. Another possibility might be that there is less public sympathy for Mexican immigrants since they are rarely perceived as victims of human rights abuses like immigrants from African or southeast Asian countries. We thank an anonymous reviewer for bringing this to our attention. Since our data do not allow us to conclusively prove which theoretical mechanism is correct, we rely on prior research to argue that the social identity mechanism is more likely because racial policy attitudes (including immigration policy attitudes) are largely driven by a combination of racial threat, symbolic racism, and, to a smaller extent, principled political conservatism (see Kinder and Sanders 1996; Sniderman and Piazza 1993; Tolbert and Grummel 2003). This suggests that a “Mexican” frame is more likely to cue racial/ethnic concerns than other motivations related to geography or human rights considerations (see also Brader et al. 2008).


We pool our data to take advantage of the larger number of cases and the alternative framing experiments in a single analysis as well as to provide clarity in presentation of the results. The two separate framing experiments have the same “baseline” frame: “undocumented immigrants.” Each then has a different alternative frame (“undocumented Mexicans” or “illegal immigrants”). This allows us to consider the experiments as one in which there are two treatments and a control. Alternative specifications where we analyze the studies separately find substantially similar results. When we run the ordered logistic regression analyses independently for October 2007, when we conducted our ethnicity experiment, and January 2008, when we conducted our “illegal” experiment, we find similar and significant support for the effects of issue importance and party preference as we report here.


Alternate specification using a multinomial logit estimation confirms the findings from our ordered logit analysis. In comparison to the baseline category of “amnesty,” support for “conditional citizenship” is significantly conditioned by party preference and education. Support for “deportation” is conditioned by issue importance, party preference, and education. As we later argue, importance should matter much more to those choosing deportation because of the punitive nature of that option.


This assumes, of course, that the ethnicity frame does not implicitly convey an indicator of legal status. It is possible that some respondents assume that all immigrants from Mexico are undocumented, thus this finding may be at least partially attributable to perceived legal status as well as ethnicity.


This also complements previous research investigating the “pliability” of politically tolerant political attitudes. It has been shown that it is easier, using competing arguments and frames, to persuade those who are tolerant on certain issues to adopt more intolerant attitudes than it is to persuade those who are intolerant to adopt more tolerant attitudes. See Gibson (1998) and Peffley et al. (2001).


From these results, we are not able to conclusively state whether immigration rhetoric in the campaigns influenced the policy preferences of Iowa voters or if the frames merely helped uncover preferences that were already present. Rather, we simply identify whether shifting frames causes, ceteris paribus, changes in public opinion on the issue of immigration.



Authors are presented alphabetically and are equally responsible for the content of this report. Replication data are available from the authors upon request. A previous version of this paper was presented at the 2009 Conference for the Southern Political Science Association in New Orleans, LA. We are indebted to the editors, anonymous reviewers, and conference discussants for their extremely constructive feedback and suggestions. We also acknowledge and appreciate assistance from James Krueger of the University of Wisconsin-Oshkosh.

Copyright information

© Springer Science+Business Media, LLC 2010