Immigration Attitudes and Positive Messaging: Evidence From the United States

Abstract

This paper considers a messaging strategy to shift immigration preferences, arguing that if citizen attitudes in this issue-area build from several dimensions, then a positive message related to each dimension should move attitudes in a more favorable direction. It tests the first part using original survey data with directly comparable questions about whether immigration hurts/helps American culture/the economy/national security, providing evidence that all three dimensions currently support the preferences of voting-age citizens. It tests the second part by randomly presenting another sample with different messages about how labor immigration strengthens national security, creates new jobs, or enhances culture, finding that all three reduce anti-immigration attitudes with significant effects even within groups that are more opposed to immigration (namely, white Americans, those with less education, and partisan Republicans).

Keywords

immigration attitudes labor migration messaging national security

Among the primary features of economic globalization (e.g., international trade, cross-border flows of capital and labor), labor immigration stands as the most restricted. And immigration restrictions, especially in the democratic destination countries, are not hard to explain: a majority of citizens consistently express a preference against immigration openness in cross-national surveys (e.g., Facchini & Mayda, 2009). As Rosenblum and Cornelius (2012) argue on this point, “if there is a universal truth about immigration policy, it is that residents of industrialized states would prefer to see lower levels of immigration.”

While the average citizen arguably knows little about actual policy in this issue-area (Brader et al., 2008; Citrin & Sides, 2008; Wong, 2007), these survey results suggest that citizens appear to be at least somewhat familiar with the potential costs associated with external labor openness (e.g., greater competition for lower-skilled jobs, helping to explain why immigration opposition tends to be especially strong among less educated individuals). And while citizens may be less acquainted with the potential benefits, there are many positive effects associated with a more open labor immigration policy. The tension between achieving these benefits while also adhering to the democratic principle of public policy being set to majoritarian preferences underlies the “liberal paradox” in this issue-area (Hollifield, 1992).

Yet this liberal paradox, or the democratic immigration dilemma, could be reduced if more citizens came to hold favorable attitudes about external labor openness. This paper thus considers a varied messaging strategy to shift immigration attitudes, asking if positive messages and of what type can move attitudes in a more favorable direction. While there has been a lot of research focused on explaining individual-level attitudes in this issue-area, less has been directed at identifying “what might foster more positive immigration attitudes” (Bonilla & Mo, 2018). Correspondingly, there is “very little evidence indicating whether and how one can systematically decrease anti-immigration sentiments on a broad scale” (Facchini et al., 2016).

Given the importance of this research agenda, a growing community of scholars are working on related questions. For example, Grigorieff et al. (2016) provide experimental evidence showing that information about the proportion of immigrants improves attitudes about immigrants themselves, but not about immigration policy. Conversely, Facchini et al. (2016) demonstrate, using a Japanese sample, that informational treatments about how foreign workers could help the country’s economy create support for a more open external labor policy. However, Adida et al. (2018) find no change in attitudes towards Syrian refugees from either an information treatment or a perspective-taking exercise, although Bonilla and Mo (2018) report more positive immigration policy attitudes for American respondents exposed to a bridging treatment linking immigration to the bipartisan issue of human trafficking. However, Hopkins et al. (2019) find no significant effect in experiments providing respondents with accurate information about the extent of immigration, which non-Hispanic Americans tend to overestimate. Most recently, Williamson et al. (2021) show that a perspective-taking message about family history increases support for more open immigration.

With this mixed evidence about effectiveness, it remains important to further establish if positive messages can influence attitudes about labor immigration policy. And more importantly, if messaging can be effective, then we also need to better establish what kind of information can produce positive effects. As Facchini et al. (2016) discuss: “To ensure [the] effectiveness of information campaigns, figuring out the specific immigration related benefits to which native citizens will most relate, and targeting different audiences with group-specific information, is a task that requires further experimentation.” Our paper takes on this task.

First, we present the argument that if citizen preferences concerning immigration build from multiple dimensions (e.g., effects related to culture, the economy, and national security), then positive messages related to each of these dimensions should shift their attitudes in a more favorable direction. Our argument does not claim that these are the only possible considerations that could influence policy preferences in this issue-area. Indeed, Newman et al. (2015) show the importance of humanitarian concerns, which appear especially important for immigration policy related to refugees and family reunification. However, in this paper more focused on immigration policy related to foreign labor, we limit our analysis to economic, cultural, and national security dimensions.

Second, we test the “if” portion of our argument using original survey data and directly comparable questions about whether immigration hurts or helps American culture/the economy/national security. After demonstrating how these questions broadly capture separate dimensions related to policy attitudes in this issue-area, we present survey results showing that all three dimensions currently support the policy preferences of voting-age American citizens. On this basis, American immigration attitudes appear not to be driven only or primarily by cultural considerations as argued by some scholars (e.g., Hainmueller & Hopkins, 2014; Newman & Malhotra, 2019; Sniderman et al., 2004); they appear similarly linked to economic and national security considerations.

Third, we test the “then” portion of our argument experimentally. This experiment randomly presents a second sample of voting-age American citizens with three different positive “messages” (defined as information packaged within a persuasive frame) about labor immigration. One message concerns the national security benefits, the second presents some economic advantages related to job creation, and the third considers how foreign workers enhance American culture. We find that all three messages reduce anti-immigration attitudes and that they even have an effect within groups that are more opposed to external labor openness (e.g., white Americans, citizens with less education, and partisan Republicans). Finally, we discuss some policy implications associated with these results, namely how messaging might operate in the real-world and its limitations in the face of counter-messages.

The Argument

As outlined above, the argument advanced in this section has two parts: if (1) citizens base their immigration policy preferences on multiple considerations (e.g., those related to culture, economics, and national security), then (2) a variety of different positive messages about labor immigration should make their attitudes more favorable. We state this argument as an “if/then” proposition because this structure helps to establish on what basis we expect our central proposition (i.e., the second part) to be true. And the first part is stated conditionally since not all scholars may accept it as valid; accordingly, it will be tested in the next section. Before explaining why we believe this two-part argument to be true, it is useful to discuss why it might be false.

Beginning with the second part, while individuals have little factual knowledge about the extent of immigration and its effects (especially the benefits), this subject is nonetheless very salient (Mellon, 2014), even serving as a primary issue for many voters in the 2016 Brexit referendum (Meleady et al., 2017) and in 2016 U.S Presidential election (Gimpel, 2017). Given its salience, opinions may be strongly anchored (Kustov et al., 2021), making it hard for any positive message about labor immigration to move policy preferences. As discussed by Lawrence and Sides (2014), “people tend to resist changing their attitudes.” And when new and contrary “information is impossible to avoid, people may ignore it, discount it, or rationalize it away.” Hopkins et al. (2019) similarly argue that “attitudes towards immigration are grounded partly in stable predispositions, often established early in life and reinforced by later socialization” thus rendering “these attitudes resistant to information that challenges existing beliefs.”

Shifting to the first part of our argument, one might counter that immigration preferences are not strongly based on multiple considerations (as we argue here). Instead, they are largely grounded upon a single primary concern: immigration represents a broad cultural threat related to the skin color, religion, language, and social practices of those from foreign countries. Reviewing the literature, Hainmueller and Hopkins (2014) note the “dominant economics-versus-culture framing” in the research program devoted to explaining immigration attitudes. However, they strongly favor the latter, arguing that “immigration attitudes are shaped by sociotropic concerns about its cultural impacts – and to a lesser extent its economic impacts” (ibid, emphasis added), even concluding that “[a]s an explanation of mass attitudes toward immigration, the labor market competition hypothesis has repeatedly failed to find empirical support, making it something of a zombie theory” (ibid, emphasis added). This conclusion parallels the argument advanced earlier by Sniderman et al. (2004) that “considerations of national identity dominate those of economic advantage in evoking exclusionary reactions to immigrant minorities” and the results more recently presented by Newman and Malhotra (2019) showing that the skill premium preference in terms of immigration policy, often interpreted as economic sociotropism (i.e., individuals believe that more skilled immigrants are better for the economy), may simply be racism in disguise.

On this basis, one would not expect any message about the economic or national security benefits associated with labor immigration to significantly influence citizen attitudes since they are not strongly based on such concerns. And even if attitudes were potentially moveable (contrary to the argument above), then one might expect that only messages about the cultural benefits should be able to shift them; messages about the economic and national security benefits associated with labor immigration would not have the same positive effect.

To advance our two-part argument, we respond to the concerns just raised. First, we accept that immigration is a highly salient issue for many voters. However, the salience of an issue does not mean that attitudes cannot be moved. In fact, salience combined with the lack of information in this issue-area may lead some individuals to consider new information and adjust their preferences. Stated differently, because individuals care about immigration, they have an incentive to pay attention to new information in this issue-area (Brader et al., 2008). And since they are unfamiliar with this information, it has the potential to influence their preferences, especially when it is packaged within a persuasive frame of another important issue (Bonilla & Mo, 2018). Indeed, while Hopkins et al. (2019) find no effect for those who received a purely informational treatment to correct overestimates about the percentage of the population that is foreign-born, they also conclude that information packaged within persuasive “frames [what we define as ’messages’] might speak more directly to the cultural and economic threats that many American believe stem from immigration, and they may resonate with people’s stable predispositions in a way that facts lacking such frames do not.”

Second, we do not dispute the claim that immigration preferences are based on broad cultural considerations. However, there is no shortage of literature showing how individual-level attitudes in this issue-area also appear to have an economic foundation (e.g., Bearce & Roosevelt, 2019; Gerber et al., 2017; Hanson et al., 2007; Hix et al., 2021; Malhotra et al., 2013). Indeed, Facchini et al. (2016) already demonstrate that different messages about how foreign workers could help the Japanese economy foster more favorable immigration attitudes, and it seems hard to explain these results if Japanese citizens only based their preferences on the cultural costs/benefits.

Going further, the “culture versus economics” debate does not consider another frame, about how immigration represents a potential threat to national security (e.g., Lahav & Courtemanche, 2012; Rudolph, 2003; Teitelbaum, 1984; Weiner, 1993). Indeed, the Trump administration regularly employed a national security logic to justify its efforts to further restrict immigration. For example, the opening line in a 2018 White House statement on migration stated: “Our current immigration system jeopardizes our national security and puts American communities at risk” (White House, 2018). As then Homeland Security Secretary Kirstjen Nielsen offered to explain the Trump Administration’s “zero tolerance” immigration policy: “This is a national security issue” (Karson, 2018). Our claim is not that national security considerations necessarily underlie individual-level immigration policy preferences in all national contexts and at all points in time. Instead, we argue that at least in the present American context, which is where we test our argument in the next two sections, preferences in this issue-area are based on multiple considerations, including culture, the economy, and national security.

If American preferences are based currently on multiple considerations, then we also expect that a variety of different positive messages about labor immigration (e.g., about how it strengthens national security, helps the national economy, and enhances American culture) should make attitudes more favorable in this issue-area. Especially given the real-world evidence offered by Flores (2018) that the effect of negative immigration messages is ephemeral, our argument does not advance any claim that the effect of these positive messages should be long-lasting; accordingly, we only test their short-term impact on individual attitudes. But Flores also found no effect (either short or long term) for positive messages about immigration around the 2016 U.S. Presidential election, thus, marking the United States as a potentially “hard” case for finding even a short-term effect for positive messages.

Given that our two-part argument will be tested on samples of voting-age American citizens, we state our hypotheses with reference to this sample. Our first hypothesis (H1) posits that in addition to culture, Americans base their immigration policy preferences on economic and national security considerations. Building on this logic, our second hypothesis (H2) proposes that different messages about how labor immigration strengthens national security, helps the economy, and enhances American culture should all reduce opposition to immigration openness. With our empirical focus on the United States, it is important to state that our hypothesis testing should be treated as a quantitative case study. While our argument potentially applies elsewhere, we cannot claim that the specific results presented in the next two sections necessarily fit other countries or other points in time (for the American case).

Survey Evidence

In this section, we test our first hypothesis that Americans currently base their immigration preferences on multiple considerations, including those related to culture, economics, and national security. Stated differently, H1 implies that immigration policy preferences should be significantly correlated with cultural considerations related to immigrants, with economic considerations related to the same, and with national security considerations. While we might have simply treated the first part of our argument as an untested assumption underlying the second, it merits empirical consideration for two related reasons.

First, as a final statement in their experimental study of information and immigration attitudes, Grigorieff et al. (2016) argue that we still “need to get a better understanding of how people form their political attitudes” in order to predict what information might influence their immigration policy preferences. Our survey takes a step in this direction. Second, while there is already a large literature on the determinants of immigration attitudes, as reviewed by Hainmueller and Hopkins (2014: 24 emphasis added), “relatively few studies have considered the impact of sociotropic economic considerations alongside sociotropic cultural considerations, making it unclear how much weight to accord each explanation.”

The Survey

The last point above is crucial to the design of our survey instrument. When scholars have sought to test culture versus economics, this exercise has often compared “apples to oranges.” Cultural considerations are usually captured using other individual-level attitudes (e.g., racial attitudes), while economic considerations tend to be measured in terms of a material factor (e.g., educational attainment). To the extent that attitudes naturally correlate more readily with other attitudes (Fordham & Kleinberg, 2012), it is not surprising to observe that cultural attitudinal variables often appear as stronger predictors of immigration attitudes than material variables. One thus needs to construct directly comparable measures of how respondents think about this issue-area in terms of culture, economics, and national security.

We do this in our survey using three simple and parallel questions (presented in a random order): “In your opinion, do immigrants help or hurt American culture?” “In your opinion, do immigrants help or hurt the American economy?” and “In your opinion, do immigrants help or hurt national security?” For each question, respondents have the same set of possible answers (presented in a random direction). We use their responses to create three attitudinal independent variables, labeled Hurts Culture, Hurts Economy, and Hurts Security, with hurts a lot = 2, hurts a little = 1, no effect = 0, helps a little = −1, and helps a lot = −2. The wording in these three queries is deliberately concise, varying only in terms of a single basic concept, in order to make them comparable with each other. Our wording is also intentionally neutral in an effort to reduce social desirability bias, which might be more present had we asked respondents to agree/disagree with statements like “immigrants reduce American racial purity” or “immigrants take jobs away from Americans.”

Following these Hurts… queries and the demographic/partisanship questions (described below), we ask the question used for our dependent variable, labeled Decrease Immigration: “What is your opinion about the number of immigrants that should be allowed to enter the United States?” Presented in a random direction, respondents have five possible answers, which we code in the following manner: decrease a lot = 2, decrease a little = 1, remain the same = 0, increase a little = −1, and increase a lot = −2. Descriptively, the immigration policy preferences in our sample are consistent with the stylized fact in the first paragraph of this paper: the mean value for Decrease Immigration is positive, indicating a general preference for greater immigration restrictions. Indeed, only 29% of our respondents prefer to increase immigration, either “a lot” or “a little.”

Our survey instrument also includes a set of demographic/partisanship queries used to create our control variables. Our control set includes dichotomous indicators for gender (Male), race (White, Black, and Hispanic with other races as the omitted category), working status (Employed and Retired) and partisanship (Democrat and Republican with independents as the omitted category). It also includes a cardinal measure for Age and ordinal measures for both family Income Category (less than $25,000 = 0, $25,000–$49,999 = 1, $50,000–$74,999 = 2, $75,000–$99,999 = 3, and $100,000 and greater = 4) and Education (less than high school degree = 0, a high school degree or GED = 1, some college but no degree = 2, a 2-year tertiary degree = 3, a 4-year Bachelors degree = 4, and a post-Bachelors degree = 5). Descriptive statistics for all of our survey variables are presented in Table 1.

Table 1.

Survey Descriptive Statistics.

	Mean	S.D.	Min.	Max.
Decrease immigration	.16	1.26	−2	2
Hurts culture	−.42	1.38	−2	2
Hurts economy	−.38	1.41	−2	2
Hurts security	.06	1.25	−2	2
Male	.49	.50	0	1
Age	44	17	18	99
White	.68	.47	0	1
Black	.12	.33	0	1
Hispanic	.13	.34	0	1
Education	2.9	1.5	0	5
Income category	2.1	1.5	0	4
Employed	.63	.48	0	1
Retired	.18	.38	0	1
Democrat	.35	.48	0	1
Republican	.32	.47	0	1

N=1050.

Our survey was conducted in May 2019 using Qualtrics facilities, and the sample consists of 1050 voting-age American citizens, who are nationally representative across five dimensions: (1) age group, (2) gender, (3) race, (4) income category, and (5) region. Qualtrics recruited this sample using an opt-in methodology, making it nationally representative in these dimensions through quotas. Information about the size of the quota buckets in each of these dimensions is provided in Online Appendix 1, Table 5.

Before regressing Decrease Immigration on the three Hurts… variables to test H1, it is important to address the possibility that these independent variables are not measuring anything distinct from each other. It has often been argued that expressed immigration attitudes are merely visceral reactions about foreigners. If this was true, then one would expect that individuals would respond in more-or-less the same way to all three “immigrants help or hurt…” queries, meaning that they capture nothing much that is specific to the queried dimension (i.e., American culture, American economy, or national security). The different mean values in Table 1 for the three Hurts… variables suggest that this was not the case for our survey sample, but we need to go further in demonstrating how these variables each capture something different.

In Table 2, we present the bivariate correlations among these three independent variables, showing that it does not exceed 0.55. Furthermore, when conducting a factor analysis using these three variables, they do indeed load to a single factor with an eigenvalue greater than 1.0. But this single factor loading should be expected since all three variables are intended to capture greater “opposition towards immigrants” (which would, in fact, be an appropriate label for this factor). However, the “uniqueness” value for each variable, capturing the percentage of its variance that is not explained by the created factor is relatively high (>50%), indicating that Hurts Culture (uniqueness = .53), Hurts Economy (uniqueness = .51) and Hurts Security (uniqueness = .59) each measure something that is different from the other two variables about why citizens are more opposed to immigrants.¹

Table 2.

Bivariate Correlation Matrix.

	Hurts Culture	Hurts Economy	Hurts Security
Hurts culture	1.00
Hurts economy	.55*	1.00
Hurts security	.48*	.50*	1.00

N=1050. Statistical significance: * p<.05 (one-tailed).

Testing H1

Our estimates in Table 3 begin with a model that includes only the demographic/partisan controls since we

Table 3.

Survey Models of Decrease Immigration.

	No Attitudinal IVs	Add Attitudinal IVs	Only Attitudinal IVs
Hurts culture		.21* (.04)	.22* (.04)
Hurts economy		.15* (.03)	.15* (.04)
Hurts security		.22* (.03)	.25* (.04)
Male	−.01 (0.08)	.04 (.06)
Age	.016* (.003)	.014* (.003)
White	.24* (.12)	.13 (.09)
Black	.04 (.14)	.006 (.137)
Hispanic	.28 (.17)	.23 (.12)
Education	−.11* (.03)	−.05* (.02)
Income category	−.05 (.35)	−.03 (.03)
Employed	−.001 (.083)	−.13 (.08)
Retired	−.18 (0.17)	−.21 (.14)
Democrat	−.45* (0.07)	−.13 (.08)
Republican	.30* (0.09)	.06 (.09)
Constant	−.23 (0.20)	−.13 (.17)	.29* (.04)
R²	.12	.33	.29
N	1050	1050	1050

OLS estimates with robust standard errors clustered by state in parentheses.

Statistical significance: * p < .05 (two-tailed).

expect the addition of the attitudinal independent variables (Hurts…) to weaken their effects.² The estimates in the first model accord with results already observed in the immigration policy literature: older Americans are significantly more in favor of decreasing immigration, as are white Americans. Those with more education are significantly less in favor of decreasing immigration, as are partisan Democrats. Consistent with this latter result, partisan Republicans significantly favor decreased immigration (compared to independents as the omitted partisan category).

As expected, these demographic/partisan relationships tend to attenuate when the attitudinal independent variables are added in the second model of Table 3 to test H1. And consistent with our first hypothesis, all three of the Hurts… variables enter with statistically significant positive coefficients. Although we do not think that there should be strong social desirability bias in these Hurts… independent variables both because of the generic wording in the underlying queries and because the responses were obtained from an anonymous internet survey without face-to-face interviews, it is nonetheless important to consider the possibility of social desirability bias and how this might affect our results. If citizens oppose immigration but do not want to express their belief that immigrants hurt the culture/economy/national security, then this should weaken the relationship between the former variable and the latter set. Thus, when regressing Decrease Immigration on the Hurts… variables as done in Table 3, social desirability bias would create a Type II error (i.e., a false negative), which does not create support for H1. Likewise, if social desirability bias affects all three Hurts… variables to more-or-less the same extent, then it should not be problematic when comparing their coefficients.

Given that these variables capture issue-specific attitudes about immigrants, it is perhaps not surprising to observe their strong association with a preference to Decrease Immigration, so it is also important to compare their coefficients, which is possible given their parallel structure/coding. In the second model, the point estimate for Hurts Security is the largest (0.22) with Hurts Economy as the smallest (.15). Regarding this latter result, it should be expected that the Hurts Economy coefficient would be somewhat attenuated since our survey was conducted at a time when the official unemployment rate in the United States was near its low for the 21st century (3.6%), potentially reducing economic concerns related to immigrant competition for jobs. However, even in this context, economic considerations nonetheless remain important for understanding American policy preferences in this issue-area. Furthermore, when we test to see if the three Hurts… coefficients are significantly different from each other, they are not so at conventional levels of statistical significance.³

In the third model of Table 3 when we drop the demographic/partisan control variables, the three Hurts… coefficients all remain statistically significant, becoming somewhat larger compared to the parallel point estimates in the second model. However, there are no statistically significant differences between any Hurts.. coefficient in the third model compared to the same in the second, showing the stability of our results testing H1. Indeed, as further evidence on this point, while all three Hurts… coefficients are statistically different from zero with at least 95% confidence in the third model, they remain as not statistically different from each other at the same confidence level.⁴

Group-Specific Relationships

Of course, the results in Table 3 describe the “average” voting-age American citizen, and there may be important variation in how individuals within different groups presently form/support their expressed immigration policy attitudes. In particular, we focus on white Americans, those with less education, and partisan Republicans as the groups who strongly prefer to Decrease Immigration, following the results in the first column of Table 3. Working from the simple specification in the third column of Table 3, we estimate three additional Decrease Immigration models, interacting the Hurts… variables with group-specific measures. The first group measure is the White dummy, thus, separating Americans who report in this majority racial category from those who report as racial minorities (non-Whites). The second is a dummy variable for respondents with at least a 4-year Bachelors degree to separate those with less skill/education (<Bachelors) from those with more (≥Bachelors). And the third model contains two sets of interactions with the Hurts… variables, using both the Democrat and Republican dummies (with Independent as the omitted partisan category).

In Figure 1, we report the group-specific marginal effect from these three interaction models. Figure 1 is organized in three panels, one for each Hurts… variable. Thus, one can compare the three marginal effects for each group vertically (e.g., the first column is for non-White respondents and the second for White both coming from the same interaction model). While all three dimensions significantly support white American attitudes, the same is not quite true for non-whites, whose immigration policy preferences are not (unsurprisingly) significantly related to broad cultural concerns about immigrants. The immigration policy preferences of less educated Americans (<Bachelors) are supported by all three dimensions, as are those with more skill and education (≥Bachelors). Finally, Republican immigration attitudes are not significantly supported by broad economic considerations, perhaps because this consideration becomes redundant next to cultural and national security considerations. However, the attitudes of those who report as a Democrat or as an Independent appear based on all three dimensions.⁵

Figure 1.

Marginal effect of Hurts… for different groups.

Thus, with certain group exceptions that we have just discussed, the immigration attitudes of American citizens currently appear to be generally supported by broad cultural, economic, and national security considerations. This is an important demonstration because it provides further evidence that while citizen preferences in this issue-area are certainly based on cultural concerns, they are also based on those related to the economy and national security. We thus turn to our second hypothesis that different messages about how labor immigration strengthens national security, helps the economy, and enhances American culture should all make citizen preferences more favorable in this issue-area.

Experimental Evidence

The Experiment

We test H2 experimentally, randomly presenting a second sample of voting-age American citizens with one of three different positive messages about the effect of immigration next to a control group receiving no message.⁶ Given our focus on labor immigration (rather than on immigration related to refugees and/or family reunification), our messages all deal with the benefits of having more foreign workers in the United States. However, these messages differ based on the role of immigrant workers (either as soldiers, business-owners, or cultural innovators) and their associated benefits (either related to national security, the economy, or American culture).⁷ The informational treatments provide facts that are easy to find online, so they approximate what the mass public might encounter if they were searching for such information. They are also deliberately short so that they provide content that one could deliver in a brief political message (either in print, on television, or online).

As shown in Online Appendix 2, each treatment includes a textual vignette with approximately 70 words written using a parallel structure. These messages also include a photographic image to help anchor the textual content and a follow-up question to help ensure that the respondent actually reads the text. These follow-up questions are also used as our treatment check. While in Online Appendix 2, we provide citations for the facts within the vignettes,⁸ these citations were not shown to respondents because we do not want the treatment effects (if any) to be based on perceptions about source credibility. We intend this exercise to be an information/framing experiment and not an endorsement experiment. Before proceeding, it is important to state that other messages may be more-or-less effective in making immigration attitudes more favorable in the short-term. Thus, to the extent that we find positive effects for these three messages, one can only say that it is possible to move attitudes with some message in that dimension. One certainly cannot conclude that any message about the benefits of labor immigration should have the same effect.

Following the random assignment of these treatments, respondents received the same query about immigration policy as used in our survey: “What is your opinion about the number of immigrants that should be allowed to enter the United States?” Our attitudinal dependent variable (also coded in the same manner) thus retains the label of Decrease Immigration. H2 expects Security Treatment, Economy Treatment, and Culture Treatment to take on negative signs, indicating a lesser preference to Decrease Immigration compared to the untreated control group.

This messaging experiment was conducted in July 2019 using Qualtrics facilities. Nationally representative in terms of gender and age group, our experimental sample includes 1001 voting-age American citizens; 250 received each of the three treatments with another 251 in the control group. As evidence of successful randomization, we present our descriptive statistics in Online Appendix 1, Table 6 to demonstrate the demographic/partisan balance across the sub-samples. In Online Appendix 1, Table 7, we also present the results of a multinomial logit with the treatment group as the dependent variable to show that our demographic/partisanship variables do little to explain assignment to these treatments.⁹

Experimental Results

Table 4 presents our experimental results using two different samples (the full sample and then dropping those who failed the treatment check) and two different specifications (excluding and then including the control variables) to assess their robustness. Assuming successful randomization so that no controls are needed, the first model considers the full sample, following the logic that survey experiments tend to overstate real-world treatment effects because many citizens fail to receive the treatment equivalent. The results show that all three treatments have a significant effect in reducing opposition to immigration, consistent with the expectations of H2. The negative point estimate is the largest for Culture Treatment (−0.47), followed by Security Treatment (−0.45), and the smallest for Economy Treatment (−0.35). While all three …Treatment coefficients are statistically different from zero, none are statistically different from the other treatments.¹⁰ Thus, one might conclude that all three positive messages are more-or-less equally effective in reducing anti-immigration attitudes. In substantive terms, these coefficients represent a roughly 10% reduction in the preference to Decrease Immigration.

Table 4.

Experiment Models of Decrease Immigration.

	Full	Full	Passed	Passed
	Sample	Sample	Check	Check
Security treatment	−.45* (.11)	−.43* (.11)	−0.42* (0.13)	−.45* (.12)
Economy treatment	−.35* (.13)	−.41* (.13)	−0.54* (0.13)	−.62* (.12)
Culture treatment	−.47* (.12)	−.41* (.12)	−0.54* (0.12)	−.45* (.10)
Male		.07 (.08)		.12 (.10)
Age		.015* (.003)		.017* (.003)
White		.30* (.12)		.44* (.18)
Black		−.02 (.15)		.13 (.16)
Hispanic		.24 (.14)		−.17 (.18)
Education		−.11* (.03)		−.10* (.03)
Income category		.004 (.028)		.008 (.036)
Employed		−.01 (.10)		−.05 (.09)
Retired		−.17 (.12)		−.35 (.14)
Democrat		−.33* (.11)		−.33* (.13)
Republican		.45* (.10)		.48* (.10)
Constant	.56* (.09)	−.17 (.26)	.56* (.09)	−.46 (.28)
R²	.02	.18	.03	.21
N	1001	1001	742	742

OLS estimates with robust standard errors clustered by state in parentheses.

Statistical significance: * p < .05 (two-tailed).

In the second model of Table 4, we add the control variables to the full sample regression to address any concern that our randomization was not completely successful. Even with a different right-hand side specification, we find very similar results: all three …Treatment variables have a significant effect in reducing preferences for a restrictive immigration policy, but none of these informational treatments are significantly different from the others. Correspondingly, there are no statistically significant differences between any the …Treatment coefficients in the second model when compared to the same in the first, consistent with successful randomization.

Of course, not all respondents were successfully treated by our positive messages based on their ability to correctly answer the follow-up question. Indeed, about 26% of our respondents failed this treatment check. In preliminary experiments using Mechanical Turk (MT) convenience samples, we found a lower percentage failing the treatment check. But MT respondents can be rated, thus, giving them an incentive to pay closer attention to the content of our vignettes. However, the respondents in our population-based Qualtrics sample had no such incentive, so it is perhaps not surprising that a higher percentage failed the treatment check.

It is thus important to consider if the experimental results would change when we drop the respondents who were not successfully treated. In Table 4, we do this in the third and fourth models both excluding and then including the control set. Not surprisingly, the negative coefficients associated with our positive messages tend to get larger when compared to their corresponding coefficients in the full sample models. However, the basic results remain unchanged: all three of the …Treatment coefficients are statistically different from zero and not statistically different from each other. Thus, our experimental results appear robust both across samples and with different right-hand side specifications.

However, do these messages successfully reach American citizens who show the most opposition in this issue-area? We showed in the first model of Table 3 that white Americans, citizens with less education, and partisan Republicans were especially opposed to a more open immigration policy. Do their attitudes move in the opposite direction when treated with a positive message? And if so, then what messages are more effective in reducing the anti-immigration preferences of American citizens within these groups? If our messages only influence those who are already comparatively supportive of immigration (e.g., non-white Americans, citizens with more education, and partisan Democrats), then while the “average” American might be more supportive, polarization would also be expected to increase in this issue-area (e.g., a greater distance between Republicans and Democrats in terms of their immigration policy preferences).

We explore this possibility in a series of models where the …Treatment variables are interacted with group-specific measures (White in one model, having at least a Bachelors degree in second, and the Democrat and Republican dummies in a third).¹¹ These interaction models are thus directly parallel to those used for the group-specific marginal effects from our survey data in Figure 1. In Figure 2, we plot the same set of group-specific marginal effects for each …Treatment variable from our experimental data. Figure 2 is organized in the same manner as Figure 1 with three stacked panels (one for each …Treatment variable), thus, allowing for the vertical comparison of the three treatment effects for each group. Focusing on white Americans, all three treatments appear effective in reducing their opposition to immigration. Only the Economy Treatment shows a significant effect for non-white Americans, but it should be difficult to further reduce the anti-immigration preferences within our racial minority group since their attitudes are already less negative in this issue-area compared to majority white Americans.

Figure 2.

Marginal effects of …Treatment for different groups.

Focusing next on those with less education (<Bachelors), one can also observe that all three treatments successfully reduce their preference to Decrease Immigration. But for those with more education (≥Bachelors) and already more favorable immigration attitudes, the Security Treatment has no significant impact in further reducing their preference to Decrease Immigration, while the Culture Treatment and the Economy Treatment are both effective in this direction. Finally, while the Security Treatment and the Economy Treatment are effective in reducing anti-immigration preferences among Republicans, the Culture Treatment is statistically insignificant. For Democrats with already more favorable immigration attitudes, the Security Treatment had no significant effect, which is consistent with the fact that the immigration attitudes of Democrats in our survey sample were only weakly supported by security considerations (as shown in Figure 1).

Discussion

This paper has considered a varied messaging strategy to shift immigration attitudes in a more favorable direction. We argued that if policy preferences in this issue-area build from several dimensions (e.g., immigration effects related to culture, the economy, and national security), then positive messages about labor immigration related to each of these dimensions could reduce anti-immigration attitudes. We tested the first part of this argument using original population-based survey data with directly comparable questions about whether respondents believe that immigration hurts or helps American culture/the American economy/national security. These survey results showed that all three dimensions currently support the policy preferences of voting-age American citizens in this issue-area. We then tested the second part of the argument, randomly presenting a second sample of voting-age American citizens with three different messages related to either national security, job creation, or culture. These experimental results demonstrated that not only do all three messages reduce anti-immigration attitudes, they also have an effect within groups that are most opposed to a more open immigration policy in the United States, namely, white Americans, those with less education, and partisan Republicans.

Our results arguably complement those reported by Williamson et al. (2021). Their perspective-taking exercise about family histories might be read as a positive cultural message and to the extent that cultural considerations support immigration policy preferences, it would be expected to improve such attitudes. Likewise, they fit with the messaging results reported by Bonilla and Mo (2018). To the extent that Newman et al. (2015) shows that citizens also base their immigration policy preferences on humanitarian considerations, Bonilla and Mo’s bridging treatment linked to human trafficking might be read as a humanitarian message about the benefits of immigration.

Having summarized our argument/results and how they fit with the existing literature, we now conclude with a brief discussion of their policy implications. Recognizing the need to increase citizen support for immigration if one desires policy in this issue-area to become less restrictive in democratic destination countries like the United States, it has been difficult to find effective strategies for reducing popular opposition to immigration, especially ones that can be scaled up from small experimental samples (Facchini et al., 2016). However, the positive messages considered in this paper arguably fit these criteria (both effective and scalable). Indeed, our messages about immigration were relatively short, and similar messages could be used as political campaign ads, either in print, on television, or over social media.

However, it is also important to recognize that during a national debate over immigration reform, for example, positive messages would surely be met by negative counter-messages (Kuklinski et al., 2000), which would likely reduce the effectiveness of the former. But perhaps citizen attitudes about immigration are currently so unfavorable because the message environment in the United States is so strongly dominated by negative messages (e.g., immigrants threaten national security particularly post-2001 and immigrants take jobs from natives especially after the 2008 recession). This understanding would suggest that the addition of positive messages might still help to attenuate anti-immigration attitudes since such messages are not so present in the current political/economic climate. Thus, new experiments presenting respondents with both positive and negative immigration messages stand as one important next step in this growing research program.

Another next step would be to refine the positive messages in this issue-area. American attitudes about immigration policy tend to be quite negative, while attitudes about immigrants are often more positive, consistent with Schildkraut’s (2020) concept of “ambivalent attitudes” in this issue-area. What type of citizens are more likely to hold these ambivalent attitudes (i.e., more favorable attitudes about immigrants compared to immigration openness) and what kind of messages have the greatest impact on their attitudes concerning the latter? Can attitudes about external labor openness be shifted more with positive messages about the benefits of immigration or with positive messages about immigrants once they have already entered the national economy? The messages considered in this paper did not distinguish between these two dimensions of policy in this issue-area: external openness versus the internal treatment of immigrants (Bearce & Hart, 2019).

Supplemental Material

sj-pdf-1-apr-10.1177_1532673X221078276 – Supplemental Material for “Immigration Attitudes and Positive Messaging: Evidence From the United States”

Supplemental Material, sj-pdf-1-apr-10.1177_1532673X221078276 for “Immigration Attitudes and Positive Messaging: Evidence From the United States” by David H. Bearce and Ken Stallman in American Politics Research

Footnotes

Acknowledgements

An earlier draft of this paper was presented at APSA 2019 in Washington, DC and in the Department of Political Science at the University of Colorado Boulder. We thank Christina Boyes, John Griffin, Diana Kim, Josh Strayhorn, and Jenny Wolak for helpful comments and suggestions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

David H. Bearce

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

David H. Bearce is a professor of Political Science and International Affairs at the University of Colorado Boulder. He studies international political economy with a focus on immigration, trade, and currency policy.

Ken Stallman is a Ph.D. candidate in Political Science at the University of Colorado Boulder. He studies labor migration policy with a focus on immigration openness and immigrant rights.

References

Adida

C. L.

Platas

M. R.

, (2018). Perspective taking can promote short-term inclusionary behavior toward Syrian refugees. Proceedings of the National Academy of Sciences, 115(38), 9521-9526. https://doi.org/10.1016/S0165-1765(03)00032-6.

Norton

E. C.

(2003). Interaction terms in logit and probit models. Economics Letters, 80(1), 123–129. https://doi.org/10.1016/s0165-1765(03)00032-6.

Bearce

D. H.

Hart

A. F.

(2019). Labor migration numbers and rights: do they trade off or advance together? International Interactions, 45(1), 28–53. https://doi.org/10.1080/03050629.2019.1525137.

Bearce

D. H.

Roosevelt

(2019). A sometimes hidden economic dimension to individual immigration preferences: Cross-national evidence in support of the labor competition hypothesis. Political Research Quarterly, 72(4), 894–909. https://doi.org/10.1177/1065912918819863.

Bonilla

C. H.

(2018). Bridging the partisan divide on immigration policy attitudes through a bipartisan issue area: The case of human trafficking. Journal of Experimental Political Science, 5(2), 107–120. https://doi.org/10.1017/xps.2018.3.

Brader

Valentino

N. A.

Suhay

(2008). What triggers public opposition to immigration? Anxiety, group cues, and immigration threat. American Journal of Political Science, 52(4), 959–978. https://doi.org/10.1111/j.1540-5907.2008.00353.x.

Citrin

Sides

(2008). Immigration and the imagined community in Europe and the United States. Political Studies, 56(1), 33–56. https://doi.org/10.1111/j.1467-9248.2007.00716.x.

Facchini

Margalit

Nakata

(2016). Countering public opposition to immigration: The impact of information campaigns. IZA Discussion Papers. No. 10420.

Facchini

Mayda

(2009). The political economy of immigration policy. United Nations Development Programme, Human Development Reports. Research Paper 2009/3.

10.

Flores

R. D.

(2018). Can elites shape public attitudes toward immigrants? Evidence from the 2016 US presidential election. Social Forces, 96(4), 1649–1690. https://doi.org/10.1093/sf/soy001.

11.

Fordham

B. O.

Kleinberg

K. B.

(2012). How can economic interests influence support for free trade? International Organization, 66(2), 311–328. https://doi.org/10.1017/s0020818312000057.

12.

Gerber

A. S.

Huber

G. A.

Biggers

D. R.

Hendry

D. J.

(2017). Self-interest, beliefs, and policy opinions: Understanding how economic beliefs affect immigration policy preferences. Political Research Quarterly, 70(1), 155–171. https://doi.org/10.1177/1065912916684032.

13.

Gimpel

(2017). Immigration policy opinion and the 2016 presidential vote: Issue relevance in the Trump-Clinton election. https://cis.org/sites/default/files/2017-12/gimpel-2016-vote.pdf.

14.

Grigorieff

Roth

Ubfal

(2016). Does information change attitudes towards immigrants? Representative evidence from survey experiments. IZA Discussion Papers. No. 10419.

15.

Hainmueller

Hopkins

D. J.

(2014). Public attitudes toward immigration. Annual Review of Political Science, 17(1), 225–249. https://doi.org/10.1146/annurev-polisci-102512-194818.

16.

Hanson

G. H.

Scheve

Slaughter

M. J.

(2007). Public finance and individual preferences over globalization strategies. Economics and Politics, 19(1), 1–33. https://doi.org/10.1111/j.1468-0343.2007.00300.x.

17.

Hix

Kaufmann

Leeper

T. J.

(2021). Pricing immigration. Journal of Experimental Political Science, 8(1), 63–74. https://doi.org/10.1017/xps.2020.6.

18.

Hollifield

(1992). Immigrants, markets, and states: The political economy of postwar Europe. Harvard University Press.

19.

Hopkins

D. J.

Sides

Citrin

(2019). The muted consequences of correct information about immigration. Journal of Politics, 81(1), 315–320. https://doi.org/10.1086/699914.

20.

Karson

(2018, June 21). DHS Secretary Nielsen casts immigration crisis as 'a national security issue. ABC News. https://abcnews.go.com/Politics/dhs-secretary-nielsen-casts-immigration-crisis-national-security/story?id=56057714.

21.

Kuklinski

J. H.

Quirk

P. J.

Jerit

Schwieder

Rich

R. F.

(2000). Misinformation and the currency of democratic citizenship. Journal of Politics, 62(3), 790–816. https://doi.org/10.1111/0022-3816.00033.

22.

Kustov

Laaker

Reller

(2021). The stability of immigration attitudes: Evidence and implications. Journal of Politics. Forthcoming. https://doi.org/10.1086/715061.

23.

Lahav

Courtemanche

(2012). The ideological effects of framing threat on immigration and civil liberties. Political Behavior, 34(3), 477–505. https://doi.org/10.1007/s11109-011-9171-z.

24.

Lawrence

Sides

(2014). The consequences of political innumeracy. Research and Politics, 1(2), 1–8. https://doi.org/10.1177/2053168014545414.

25.

Malhotra

Margalit

C. H.

(2013). Economic explanations for opposition to immigration: Distinguishing between prevalence and conditional impact. American Journal of Political Science, 57(2), 391–410. https://doi.org/10.1111/ajps.12012.

26.

Meleady

Seger

C. R.

Vermue

(2017). Examining the role of positive and negative intergroup contact and anti‐immigrant prejudice in Brexit. British Journal of Social Psychology, 56(4), 799–808. https://doi.org/10.1111/bjso.12203.

27.

Mellon

(2014). Internet search data and issue salience: The properties of Google trends as a measure of issue salience. Journal of Elections, Public Opinion and Parties, 24(1), 45–72. https://doi.org/10.1080/17457289.2013.846346.

28.

Newman

B. J.

Hartman

T. K.

Lown

P. L.

Feldman

(2015). Easing the heavy hand: Humanitarian concern, empathy, and opinion on immigration. British Journal of Political Science, 45(3), 583–607. https://doi.org/10.1017/s0007123413000410.

29.

Newman

B. J.

Malhotra

(2019). Economic reasoning with a racial hue: Is the immigration consensus purely race neutral? Journal of Politics, 81(1), 153–166. https://doi.org/10.1086/700109.

30.

Rosenblum

Cornelius

(2012). Dimensions of immigration policy. In Rosenblum

Tichenor

(Eds.), The Oxford handbook of the politics of international migration (pp. 245–273). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780195337228.013.0011.

31.

Rudolph

(2003). Security and the political economy of international migration. American Political Science Review, 97(4), 603–620. https://doi.org/10.1017/s000305540300090x.

32.

Schildkraut

(2020). Ambivalence in American public opinion about immigration. In Berinsky

(Ed.), New Directions in Public Opinion (3rd ed., pp. 315–337). Routledge.

33.

Sniderman

P. M.

Hagendoorn

Prior

(2004). Predisposing factors and situational triggers: Exclusionary reactions to immigrant minorities. American Political Science Review, 98(1), 35–49. https://doi.org/10.1017/s000305540400098x.

34.

Teitelbaum

M. S.

(1984). Immigration, refugees, and foreign policy. International Organization, 38(3), 429–450. https://doi.org/10.1017/s0020818300026801.

35.

Weiner

(1993). International migration and security. Westview Press.

36.

White House (2018). National security threats—Chain migration and the visa lottery system. https://trumpwhitehouse.archives.gov/articles/national-security-threats-chain-migration-visa-lottery-system/.

37.

Williamson

Adida

C. L.

Platas

M. R.

Prather

Werfel

S. H.

(2021). Family matters: How immigrant histories can promote inclusion. American Political Science Review, 115(2), 686–693. https://doi.org/10.1017/s0003055420001057.

38.

Wong

C. J.

(2007). “Little” and “big” pictures in our heads: Race, local context, and innumeracy about racial groups in the United States. Public Opinion Quarterly, 71(3), 392-412. https://doi.org/10.1093/poq/nfm023.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.32 MB