Within these theoretical constraints, this review aims to provide a comprehensive list of experimental studies on the effect of communication interventions on attitudes to immigration. The methodology by which the articles were selected, acting as a further set of constraints, was that the article must have been found within the first 150 article results from one of three Google Scholar searches for, respectively, “attitudes to immigration experiment” or “public opinion immigration experiment” “policy preferences immigration experiment”. Despite some earlier criticism that Google Scholar was not as universal in its coverage as other scholarly search engines, such as Web of Science, more recent studies have shown that this is no longer true, making it a relatively reliable single source (e.g. Halevi et al., 2017). Naturally, there was a very large amount of overlap between these three searches and all three included a large number of results that did not meet the criteria of being either (1) An experiment; (2) Aiming to change attitudes to immigration (rather than, for example, broader attitudes such as to outgroups generally or tests of social desirability bias); or (3) Endogenous to the capabilities of communicators (i.e. not dealing with macroeconomics, terrorist attacks, or migration flows, etc., nor testing the effects of deliberation or citizens’ forums, etc.) All those studies listed in the references section below constitute the 68 experimental studies, except those indicated with an “†” symbol.
This results in a set of articles with the descriptive elements—in terms of year of publication, method, country, journal, and type of immigration attitude—described in Table 2 and Appendix 1, below. Methodologically, the studies are overwhelmingly survey experiments (broadly defined), but also include some relevant conjoint, lab, natural, and quasi-natural experiments. Nearly half of the studies are based in the USA, with most of the rest in (often multiple) European countries and a few in Australia, Canada, Japan, Israel, South Korea, and Turkey. In terms of journals, we see greater variation, with most relating to three respective disciplines: political science, migration studies, and communication science. In terms of type of attitudes, a slight majority tested what affect policy preferences (either amount of immigration or to a lesser extent who can enter) and a large minority studied perceptions (overwhelmingly the effect of immigration, though occasionally the scale), while two studied personal prejudice against immigrants. The theories being tested varied substantially and were often multiple at the same time or were too idiosyncratic to be easily categorised. However, they roughly fit into the nine sections below for our purposes and, even within each of those, most often concentrate on “economic competition” and/or “cultural threat” (see Dennison & Geddes, 2021, for review).
Evidence on effectiveness is summarised qualitatively and also by the proportion of studies in which the authors find relevant statistically significant effects. Despite its shortcomings, statistical significance is used rather than effect sizes because: (1) This review limits its ambition to identifying what is effective and what is not, rather than to what extent; (2) Whether an article finds evidence that passes a certain threshold of likelihood of being replicated in the population makes an easily comparable metric; (3) Given the various measures, concepts, and contexts used in the various article, comparability of effect sizes would be highly problematic.
Providing information, correcting misperceptions and “myth-busting”
The most common forms of strategic communication used both by migration policymakers and tested by academics are those that seek to change various forms of attitudes to immigration by providing new information, typically either explicitly or implicitly attempting to correct “misperceptions”, either about migration numbers or its effects.
Correcting/providing information on migrant stocks and flows
Citizens have been repeatedly shown to overestimate the proportion of immigrants in their countries, cultural and religious differences, and migrants’ economic weakness (Alesina et al., 2021). Although such misperceptions are by no means unique to the field of migration—with citizens also likely to misperceive rates of inequality (e.g. Hauser & Norton, 2017)—they have been shown to correlate with anti-immigration views (Nadeau et al, 1993; Sides & Citrin, 2007). Studies from the USA document how corrections about the size of minority populations or by priming the annual level of immigration fail to change citizens’ immigration-related opinions (Hopkins et al., 2019; Lawrence & Sides, 2014; Sides & Citrin, 2007). In Denmark, Jørgensen and Osmundsen (2020) show that giving correct information about welfare usage, crime rates, and the proportion of the population that are migrants has no effect on policy preferences, even though individuals update their factual beliefs after considering the correct information.
Similarly, one experiment using information in narrative form—a short video about a hardworking immigrant woman described in prosaic text—also failed to change immigration attitudes (Alesina et al., 2021). These findings suggest that individuals discard counter-attitudinal information and reinterpret new information in selective ways to uphold their pre-existing views of the world and applicable narratives, for example, by normatively concluding that the actual immigration rate is “still too high”. This is ‘especially troublesome for democracy: if people can interpret information as they wish, they can always distort the causal chain from factual reality to political judgments’ (Jørgensen & Osmundsen, 2020: 2). Furthermore, this suggests that, as Hopkins et al. (2019) explain: ‘‘Misperceptions of the size of minority groups may be a consequence, rather than cause, of attitudes toward those groups.’’
However, Grigorieff et al. (2016) use an experiment in the US to show that providing individuals with information about the number of immigrants in their country makes them less likely to argue that there are too many of them. They also show that providing individuals with comprehensive information about immigration improves attitudes to existing immigrants and convinces conservatives to favour increasing legal immigration but does not change immigration policy preferences regarding legalisation and deportation. The effects were shown to still be present four weeks later. Furthermore, Bareinz and Uebelmesser (2020) show that a bundle of information on both the share and the unemployment rate of foreigners robustly decreases welfare state concerns about immigration in Germany, and that the quantity of information has a positive impact on its effect whereas the provision of information only on the share of foreigners has no effect. Conversely, Wiig (2017) also showed that information about the employment rate of immigrants in Norway (60 percent) causes individuals to rate their preferences for immigration policy more strictly.
Florio (2020) carried out an experiment in schools in Rome (aged 13–17) in which half of the classes were exposed to information—an expert informing the students about statistics on immigration numbers in Italy and in the world, as well as key origin and destination countries and expenditures and revenues generated in Italy over the course of two-hours—and the other half were exposed to contact—meeting a Mauritanian refugee in their class over two hours and reading a book about his journey three weeks beforehand. Overall, the information treatment was shown to increase positivity to a greater extent than the contact treatment, in terms of policy preferences and perceived numbers, though neither affected attitudes to immigrants.
Conversely, Blinder and Schaffner (2020) show that providing individuals with information about immigration flows—”Approximately 1.2 million legal immigrants came to the US in 2016”—make preferences for legal immigration more negative, particularly for Democrat voters, while information about Donald Trump policy proposals—”President Trump has endorsed a plan that would set levels of legal immigration to 540,000 per year “—make preferences for legal immigration more negative, particularly for Republican voters. Notably, Margalit and Solodoch (2022) show that presenting immigration information in terms of stocks rather than flows results in more positive immigration policy preferences, which they argue is the result of the sense of moral obligation elicited towards those already in one’s country.
Fact checking on the effects of migration
In terms of vote choice, Barrera et al. (2020) show that exposure to misinformation by populist radical right leader Marine Le Pen had the effect of changing vote intentions in France, upon which fact-checking corrections had no countereffect. Swire et al. (2017) and Nyhan et al. (2020) reached similar conclusions regarding misinformation from Donald Trump, concluding that voters take fact-checking “literally but not seriously”.
However, in terms of attitudes, Facchini et al. (2016) carried out a large-scale experiment in Japan, showing that exposing individuals to new information on potential social and economic benefits reduced opposition to immigration, increased support for temporary visas and even increased willingness to petition politicians. It was shown to make little difference whether the information was presented in statistical form or the form of a personal story. Effects were shown to persist 10–12 days later, albeit between one and two thirds smaller. Nakata (2017) further showed that the effects did not vary by age, gender, or education. In the US, Haaland and Roth (2020) show that presenting research about the labour market impact of the Mariel boatlift affects immigration policy preferences and willingness to sign petitions, as well as perceived wages and employment (but not fiscal or cultural) effects in that specific case. Effects on policy preferences for both low- and high-skilled immigration were still visible one week later. They (2020: 2) conclude that ‘an information treatment based on research evidence can be effective in changing beliefs and policy views for Republicans and Democrats alike, even on a highly contested issue such as immigration.’ Similarly, Igarashi and Ono (2020) show that feelings of hostility toward immigrants decrease when individuals receive positive information about immigration, while exposure to negative information does not necessarily change their attitude.
Furthermore, Hameleers et al. (2020) show that exposure to fact checking reduces attitudinal polarisation and belief in misinformation about immigration. Carnahan et al. (2021, see also Grigorieff et al, 2016) show that repeated exposure to 500– to 600-word fact checking website articles on immigration had stronger and longer effects, observable four weeks after the initial test. Keita et al., (2021) use a natural experiment in Germany, in which some newspapers disclose the national origin of criminals and some do not, to show that consistently doing so reduces self-reported concern about immigration by providing a realistic overview of how many crimes are perpetrated by Germans and non-Germans respectively. Theorin et al., (2021) expose participants to a variety of fictional tweets—some with a negative message on immigration, some with a positive one, and some in ‘episodic’ (or narrative) format and some in thematic (or informative) format—showing that none of the four types have a statistically significant effect on attitudes to free movement.
Appeals to emotions
One of the more common recommendations by migration communication practitioners has been to focus on “emotions” instead of “facts” in order to persuade. Lecheler et al. (2015: 819) test emotional reaction in participants to four types of immigration frames both posed negatively and positively, each of which was expressed in a news article about a career event for immigrant women in Amsterdam. All the frames caused emotional reactions among participants—contentment, compassion, enthusiasm, hope, anger, fear, and sadness—but only some emotions went on to affect attitudes to immigration: most notably enthusiasm and anger. Theorin (2021; compare to Theorin et al., 2021, above, on null effects of information and emotion) shows that reading longer news articles has an effect, with emotions having a greater impact than perceptions.
Other studies have shown that emotions act as a mediating variable on the effect of providing information. Morisi and Wagner (2020) show that positive information about politics and politicians reduces populist attitudes (rather than attitudes to immigration) but that the effect disappears when voters are in an angry emotional state and is lower when they are in a fearful state. Brader et al. (2008) show that those citizens moved by negative and ethnic out-group based news stories about immigration are those with high self-reported emotional anxiety. Chkhaidze et al. (2021) exposed participants to one of four versions of a passage about an increase in immigrants in one town. Each version included all identical facts and figures and differed in only a single word at the beginning of the passage, describing the increase in immigrant labour as either an “increase,” a “boost,” an “invasion,” or a “flood”, with large effect on participants’ attitudes to the increase in immigration and the predictions about its effects on the economy.
Emphasising self-interest and common interest
Most studies show that appeals to self-interest, either economically or otherwise, are ineffective at changing attitudes to immigration. Hainmueller and Hiscox (2010) use a survey experiment to show that both low-skilled and highly skilled natives strongly prefer highly skilled immigrants over low-skilled immigrants and that rich and poor natives are equally opposed to low-skilled immigration in general, undermining labour market competition theories of opposition to immigration and suggesting instead sociotropic considerations. Similarly, Schaub et al. (2021) show that anti-immigration sentiment is unaffected by the presence of refugees in respondents’ hometowns overall: on average, they record null effects for all outcomes, which they interpret as supporting a sociotropic perspective on immigration attitudes. However, part of this overall lack of effect is because right- and left-leaning individuals are both drawn to the centre following increased local presence of immigrants. Harell et al. (2012) demonstrate that, in Canada and the US, income level has no effect on the extent to which citizens prefer immigrants with high skills. These results are in line with larger macro-economic studies suggesting that what worries citizens are the economic costs of unemployment and dependence on the welfare state more than direct competition from immigrants for jobs. In both countries, the skill level of the immigrant was far more important in determining whether individuals believed that they should be allowed in than ethnic background.
Offering a more nuanced interpretation, Dancyfier and Donnelly (2013) show that individuals employed in growing industrial sectors are more likely to support immigration than are those employed in shrinking sectors, giving evidence in support of the labour market competition thesis. Naumann et al. (2018: 1009) use European data to show that ‘rich natives prefer highly skilled over low-skilled migration more than low-income respondents do’, which they argue suggests an economic concern over the fiscal burden of immigrants because ‘these tax concerns among the wealthy are stronger if fiscal exposure to migration is high.’ Moreover, Hix et al (2021) show that individuals are more willing to admit immigrants when restriction is shown to carry costs, with egocentric considerations more powerful than sociotropic ones. Jeannet (2018) takes an instrumental variable approach to show that retirement has no effect on attitudes to immigration and that retired individuals are more likely to have restrictive policy preferences when immigration is framed as unskilled, just as workers do. This undermines the labour market hypothesis that immigration policy preferences are driven by job competition fears and instead supports the notion of sociotropic determinants.
Indeed, the evidence in favour of the effectiveness of appeals to common interests—or “sociotropic” concerns—is strong. Solodoch (2020) uses a survey experiment that asks both natives and immigrants of various origins to evaluate different profiles of visa applicants to the Netherlands, showing that opposition to immigration is primarily driven by sociotropic concerns and to a far lesser extent by the ethnic basis of the would-be immigrant, with those of immigrant origin no more favourable to profiles of visa applicants of their own ethnic background. Valentino et al. (2019) offer similar results supporting a sociotropic economic thesis against a labour market competition thesis. Kustov (2021) theorises that “parochial altruists”—citizens who are both nationalistic and altruistic, which he shows to be a high proportion of the population of the UK—fit their immigration policy preferences according to the effect that they perceive it to have on their fellow citizens. This study undermines the argument that anti-immigration sentiment is rooted in ethnic animus or selfishness but instead suggests that immigration will be positively received if citizens can be convinced that it is good for their fellow citizens, particularly those less well off.
Emphasising conformity or diversity
Whereas many strategic communication campaigns on migration have emphasised the positive aspects of diversity, others have sought to emphasise the similarity or conformity of migrants and migration within the host society. Hopkins et al. (2014) use survey experiments to show that, among native-born Americans who regularly hear the Spanish-language in day-to-day life, exposure to the language can induce anti-immigration attitudes, supporting the cultural threat theory of attitudes. Newman et al. (2012) find a similar link between incidental exposure to Spanish and anti-immigrant sentiment and policy preferences. Hopkins (2015) present participants in an experiment with six manipulated news clips that include an immigrant stating: “I’ve worked hard, always paid my taxes. I’d really like the chance to be an American citizen”, however, the immigrant is randomised according to language—fluent Spanish, fluent English, or accented English—and dark or light skin tone. Skin tone is shown to have no effect on support for a new legalisation policy for unauthorised immigrants, while accented English is shown to have the most positive effects, theoretically explained as demonstrating a willingness of migrants to learn English. Ostfeld (2017) took a similar approach, exposing participants to a story about a family of undocumented immigrants living in the United States who were at risk of deportation. Both cultural assimilation (language, food, and social integration) and skin tone and physical features were shown to make a difference (see also Alarian & Neureiter, 2021).
Kaufmann (2019) similarly shows an ‘assimilation prime’, which stresses continuity over change and reassures white respondents that immigration will leave the boundaries and size of the majority group unaffected leads to support for greater immigration while a ‘diversity prime’, which stresses change and urges ethnic majority respondents to embrace it, accept their group’s ethnic decline, and focus instead on the ethnically neutral civic nation-state as the embodiment of their collective identity leads to more negative policy preferences. These effects were particularly strong amongst working class participants and populist radical right voters. Sobolewska et al. (2017) also show in the Netherlands and the UK that several social integration measures as well as the economic integration measures—having local friends, women working, voting, not being religious (in the UK), going to the pub or community centre, and cooking national food—increases the extent to which citizens view those immigrants as integrated.
The last few years have seen many conjoint experiments, in which preferences of which type of immigrants according to various variables, are preferred by citizens are revealed by the choices they make when selecting between two hypothetical migrant profiles. Typically, these show that natives are more sympathetic to descriptions of migrants fleeing persecution or having a job rather than simply coming for a better life, having legal migration status, high education levels and language skills, and shareing the country of destination’s religion.
Indeed, Bansak et al. (2016) show across 15 European countries that asylum seekers who have higher employability, more consistent asylum testimonies and severe vulnerabilities, and are Christian rather than Muslim received the greatest public support. These results are similar to Iyengar et al. (2013) who show that skill level is important when evaluating would-be immigrants, whereas cultural attributes—as measured by Middle Eastern nationality and Afrocentric appearance—have little effect (see Turper et al., 2015, and Valentino et al., 2019, for similar results). Strabac et al. (2014) show that in Norway, Sweden, the UK and the US, Muslim immigrants are not viewed more negatively than immigrants in general. However, Ha et al. (2016) use a survey experiment to show that South Koreans are more favourable to North Korean defectors than ethnic Korean Chinese or guest-workers from Indonesia. España-Nájera and Vera (2020) also use a survey experiment in California to demonstrate that favoritism for high-skilled immigrants drops when they add the Hispanic descriptor, but that legal status outweighs any possible anti-Hispanic sentiment.
More specifically, Czymara and Schmidt-Catran (2017) show that women in Germany were considerably more likely to prefer male immigrants to female immigrants prior to the 2015/16 New Year’s Eve attacks in Cologne, though this preference disappeared afterwards, whereas men were more likely to discriminate by country of origin and not by gender. Hellwig and Sinno (2017) show that, in the UK, the type of migrant affects the perceived threat felt by immigration: Eastern Europeans provoke economic and criminality threats whereas Muslims do not; Muslims provoke security and cultural threats whereas Eastern Europeans do not. Knoll et al. (2011) show that labelling irregular immigrants as “illegal” versus “undocumented” has no effect on immigration policy preferences, but that, among Republicans and especially those for whom immigration is a “most important issue”, the term “immigrants” elicits less opposition than Mexicans.
However, Wright et al. (2016) show that attribute-based judgements—related to the characteristics of the immigrant(s)—pale in comparison to categorical judgements related to issues of justice and fairness that explain public opposition to irregular migration, given moral convictions about adherence to rules. Hedegaard (2021) uses a conjoint experiment to show that climate migrants are perceived to be less deserving of permanent residency than migrants who typically could qualify for asylum, but more deserving than economic migrants. Overall, these studies suggest that communication describing migrants in terms of fairness and regularity are likely to be the most effective.
Emphasising common ground
Other communication strategies have sought to emphasise areas of common ground—issues on which most people agree—between opponents and supporters of migration. Bonilla and Mo (2018) use a ‘bridging frame’ to create a connection between a previously defined issue—human trafficking, concern about which is high and bipartisan in the US—and immigration policy, showing that treating participants with a ‘bridging frame’ reduces opposition to immigration among Republicans in relation to a control frame (about human trafficking but without any ‘bridging’), and more consistently than an information-based ‘learning frame’ or an American Dream-based ‘values frame’, which has no effect. These findings are consistent with the power of narratives built on areas of consensus (Dennison, 2021). Similarly, Van Klingeren et al. (2017) use survey experiments in the Netherlands to show that presenting politicians as divided and conflictual on an issue such as immigration exacerbates attitudinal polarisation.
Appeals to empathy
Appeals to empathy remain relatively understudied in the academic literature. Here, two studies looking at the use of a humanitarian message are overviewed. Though not the same as appealing to empathy, the two are likely to have some similar components, with the main difference that humanitarian values are abstract whereas eliciting genuine empathy means eliciting the feelings that another, for example, a migrant may have and imagining oneself in their position (see Dennison, 2021, for overview of empathy in the use of narratives). Newman et al. (2015) show that appealing to humanitarian values in White Americans elicits lower anti-immigration sentiment among participants who score high on empathy and that the power of a humanitarian frame outweighs that of a simultaneous threat frame, when presented together. Getmansky et al. (2018) show that a humanitarian message focussing on saving innocent women and children affects Turkish perceptions of Syrian refugees to become more positive, whereas focus on their militant ties cause greater negativity; their messages on economic costs and the ethnic balance in Turkey have no effect. Relatedly, Madrigal and Soroka (2021) show that presenting identical images of individual rather than a group of migrants reduces anti-immigration sentiment amongst those with high threat sensitivity. Heizler and Israeli (2021) use the natural experiment of the death of Alan Kurdi to show that the drowning of an identified individual lead to people becoming more positive about immigration, whereas the drowning of over a thousand unidentified immigrants in two consecutive events in April 2015 produced no observable change in public sentiment. This suggests that individual stories may be more powerful than statistical data.
Although the effects of who is delivering the message have been regularly postulated by policymakers and NGOs (see Dennison, 2020), particularly in terms of using migrants as messengers, this study found no recent experimental tests of this. Instead, messenger effects have been tested in terms of authority figures. Donnelly et al. (2020) show that there are no systematic differences in the effects of pro-immigration messaging when delivered by politicians, unions or businesses in a survey experiment in Canada, Germany and the UK. Margolis (2018) showed that a religiously-loaded radio message from an pro-immigration American evangelical organisation demobilised evangelical opponents of immigration, whereas an identical secular version, with no religious references, did not. The religious version included two pastors asking listeners to join a movement of Christians that supports immigration solutions rooted in biblical values. Relatedly, Wright and Citrin (2011) show in the US that hostility to immigration protesters decreases when they are shown waving U.S. flags as opposed to Mexican ones, but this effect does not translate to more moderate policy attitudes on immigration.
Appealing to identity
One’s self-identity can be defined as the extent to which an individual sees membership of a group, either nationally, ethnically, or religious (such as being Dutch, European, Arab, or Jewish), as integral to themselves or is attached to it. This concept has been applied to voting and political attitudes (Carl et al., 2019; Dennison et al., 2020; Hooghe & Marks, 2005). Sniderman and Gagendoorn (2007) show that when Dutch respondents are primed with a reminder of their national identity and group belonging, they give more negative attitudes to immigration. However, when Breton (2015) made the same prime—“people belong to different types of groups. One of the most important and essential of these groups is the nation to which you belong to. In your case, you belong to the Canadian nationality. Each nation is different’—it had no effect on immigration attitudes, nor did asking participants about the importance of their Canadian identity to them, which Breton theorises as the result of a different conception of national identity in Canada than in Europe.
Bloom et al. (2015) prime survey participants—American Catholics, Turkish Muslims, and Israeli Jews—with their religious identity, which they show increases positivity to immigrants with a similar religious or ethnic background—particularly among conservatives—but increases opposition to other, distinct groups. Lazarev and Sharma (2017) make a similar finding regarding religious identity (both Muslim and Sunni) primes and Turkish attitudes and behaviour towards Syrian refugees. Wojcieszak and Garrett (2018) show that priming American participants with their national identity leads to greater opposition to immigration among those already opposed, but has no effect on immigration supporters, and that this effect operates both directly and via the news media that they choose to consume.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.