In recent years, debates over immigration have come to the forefront in countries around the world, contributing both to the election of Donald Trump (e.g., Hooghe & Dassonneville, Reference Hooghe and Dassonneville2018; Sides et al., Reference Sides, Tesler and Vavreck2017) and to the United Kingdom’s withdrawal from the European Union (e.g., Goodwin & Milazzo, Reference Goodwin and Milazzo2017). As immigration plays an increasingly central role in politics, understanding the sources and nature of immigration attitudes is more important than ever.
Scholars have debated a variety of explanations for anti-immigration attitudes (for a review, see Hainmueller and Hopkins Reference Hainmueller and Hopkins2014), including economic concerns (e.g., Hainmueller et al., Reference Hainmueller, Hiscox and Margalit2014; Scheve & Slaughter, Reference Scheve and Slaughter2001), status threat (Craig & Richeson, Reference Craig and Richeson2014; Major et al., Reference Major, Blodorn and Major Blascovich2018), perceived threat (De Vreese & Boomgaarden, Reference De Vreese and Boomgaarden2016; Kentmen-Cin & Erisen, Reference Kentmen-Cin and Erisen2017), emotions (Brader et al., Reference Brader, Valentino and Suhay2008; Erisen et al., Reference Erisen, Vasilopoulou and Kentmen-Cin2020), and racial prejudice (Hartman et al. Reference Hartman, Newman and Bell2014). Recent work points to more fundamental roots of immigration attitudes, suggesting that anti-immigration attitudes are partially a by-product of an evolved psychological system—the behavioral immune system—that facilitates disease avoidance. According to this theory, out-group members are tagged as potential disease threats, motivating avoidance. This argument is supported by a number of psychology studies (e.g., Navarrete & Fessler, Reference Navarrete and Fessler2006), but it has been tested most comprehensively by Aarøe, Petersen, and Arceneaux (Reference Aarøe, Petersen and Arceneaux2017). If correct, this theory has novel implications for our understanding of immigration attitudes, including the promise of intergroup contact and the nature of political ideology.
The evolutionary logic behind the behavioral immune system (BIS) hypothesis makes it both provocative and challenging to test. However, one key implication of the BIS hypothesis is universality (for related discussion, see Petersen Reference Petersen2015). That is, the BIS hypothesis ought to hold across a variety of political and cultural contexts. Existing research, however, has provided only a narrow set of tests. As discussed by Aarøe, Petersen, and Arceneaux (Reference Aarøe, Petersen and Arceneaux2017), the psychological literature has relied almost entirely on student samples from Canada. These authors improve upon this body of research by testing the BIS hypothesis in nationally representative samples in both the United States and Denmark. These findings greatly strengthened the body of evidence for the BIS hypothesis, but the evidence remains limited to a few select countries. Thus, one of the key implications of the theory has been subjected to an extremely limited test.
In this article, we provide a more thorough test of the BIS hypothesis, drawing on nationally representative samples from Norway, Sweden, Turkey, and Mexico, as well as two diverse samples from the United States. These data provide the most comprehensive test of the universality of the BIS hypothesis to date. Overall, we find strong support for the BIS hypothesis while raising new questions about how it operates across different political contexts.
Disgust and the behavioral immune system
The BIS hypothesis stems from a large psychological and physiological literature on the nature and origins of disgust. The BIS is a set of psychological mechanisms designed to detect the potential presence of a pathogen and motivate a series of behavioral responses that minimize the chances of infection (for a review, see Schaller & Park Reference Schaller and Park2011). Disgust plays a key role in this system as the emotional response to perceived disease threats that motivates avoidant behavior (Oaten et al., Reference Oaten, Stevenson and Case2009). A wide variety of evidence supports this hypothesis. For example, common pathogen vectors, such as bodily fluids, pests, and rotten foods, reliably elicit disgust across cultures (Curtis & Biran, Reference Curtis and Biran2001). Images of disease threats also reliability elicit disgust and do so at a greater rate than similar images that are not tied to disease threat (Curtis et al., Reference Curtis, Aunger and Rabie2004). Thus, the emotion of disgust seems to have evolved specifically to help avoid infection.
While everyone is expected to show a disgust response to potential disease threats, there are individual differences in the strength of this tendency, known as pathogen disgust sensitivity. Women, in particular, tend to score higher in disgust sensitivity (Tybur et al., Reference Tybur, Bryan, Lieberman, Caldwell Hooper and Merriman2011). This trait is both heritable (Sherlock et al., Reference Sherlock, Zietsch, Tybur and Jern2016) and moderately stable over time (Olatunji et al., Reference Olatunji, Adams, Ciesielski, David, Sarawgi and Broman-Fulks2012), but it is theorized to respond to environmental variation in individual health and the benefits of contact with potential pathogens (Tybur et al., Reference Tybur, Lieberman, Kurzban and DeScioli2013). However, there is an emerging consensus that pathogen disgust sensitivity is the best available indicator of the BIS (Tybur et al., Reference Tybur, Frankenhuis and Pollet2014).
While there is variation in pathogen disgust sensitivity, people tend to be overly sensitive to potential pathogen cues. This is because the potential cost of failing to avoid a disease is typically much higher than the potential cost of missing out on a meal or an opportunity for social interaction (Haselton & Nettle, Reference Haselton and Nettle2006). As a result, many phenotypically abnormal individuals are tagged as potential disease threats. For example, people who are more sensitive to disease threats tend to have more negative attitudes toward obese people (Lieberman et al., Reference Lieberman, Tybur and Latner2012; Park et al., Reference Park, Schaller and Crandall2007), make harsher judgments of unattractive individuals (Park et al., Reference Park, van Leeuwen and Stephen2012), and avoid physical contact with disabled people (Park et al., Reference Park, Faulkner and Schaller2003). These findings suggest that many forms of social stigmatization and exclusion may be driven by implicit disease concerns and feelings of disgust (Kurzban & Leary, Reference Kurzban and Leary2001).
Given that the function of disgust is to avoid pathogens, the primary behavioral output is avoidance. Disgust thus works as an implicit germ theory of disease. People avoid close physical contact with any person or object that is perceived as a potential contamination threat (Park et al., Reference Park, van Leeuwen and Chochorelou2013). This insight has proven valuable in explaining a variety of political attitudes. For example, people who are higher in disgust sensitivity are more likely to oppose interracial dating and marriage (Kam & Estes, Reference Kam and Estes2016), hold more negative attitudes toward transgender people (Miller et al., Reference Miller, Flores, Haider-Markel, Lewis, Tadlock and Taylor2017; Vanaman & Chapman, Reference Vanaman and Chapman2020), and worry about genetically modified foods and vaccines (Clay, Reference Clay2017; Clifford & Wendell, Reference Clifford and Wendell2016). Some of the clearest findings come from studies on homelessness. People who are higher in disgust sensitivity are more likely to support exclusionary policies that promote distance from homeless people, such as banning panhandling and banning sleeping in public (Clifford & Piston, Reference Clifford and Piston2017). However, disgust sensitivity did not emerge as a predictor of policies that are less directly relevant to physical distance, such as government aid to homeless people.Footnote 1 Moreover, the effects of disgust sensitivity on support for exclusionary policies could not be explained by negative affect. These findings suggest that negative affect is not a prerequisite for the desire to avoid social contact.
Beyond directly motivating the avoidance of potentially infected others, the BIS also drives a variety of psychological dispositions that facilitate avoidance of potentially infected others and help enforce rules designed to inhibit the spread of infection (e.g., hygienic and dietary norms). At the country level, higher levels of parasite stress are associated with a suite of psychological dispositions related to adherence to in-group norms, such as authoritarianism (Murray et al., Reference Murray, Schaller and Suedfeld2013), conformity and obedience (Murray et al., Reference Murray, Trudeau and Schaller2011), collectivism (Fincher et al., Reference Fincher, Thornhill, Murray, Schaller, Anderson, May, Bond, Smith, Bouchard, McGue, Cohen, Cukur, Gusman, Carlo, Dobson, Carper, Ewald, Faulkner, Schaller and Stoehr2008), and group-oriented morality (van Leeuwen et al., Reference van Leeuwen, Park, Koenig and Graham2012). At the individual level, various indicators of pathogen threat and disgust sensitivity are linked with increased conformity (Wu & Chang, Reference Wu and Chang2012), right-wing authoritarianism (Liuzza et al., Reference Liuzza, Lindholm, Hawley, Gustafsson Sendén, Ekström, Olsson and Olofsson2018), and lower social trust (Aarøe et al., Reference Aarøe, Osmundsen and Petersen2016). A 30-nation study found that pathogen disgust sensitivity is consistently tied to traditionalism, and more strongly so, than to intergroup dominance (Tybur et al., Reference Tybur, Inbar, Aarøe, Barclay, Barlow, de Barra, Becker, Borovoi, Choi, Choi, Consedine, Conway, Conway, Conway, Adoric, Demirci, Fernández, Ferreira, Ishii and Žeželj2016). These dispositions all serve to enforce tradition and in-group norms while decreasing contact with strangers. In short, they all fulfill a basic disease avoidance strategy.
Disgust sensitivity and immigration attitudes
Drawing on the logic of the BIS, scholars have argued that disgust sensitivity may help explain opposition to immigration. As discussed earlier, the BIS is overly cautious, tagging many individuals and objects as potential disease threats. Just as this process results in obese, homeless, or disabled people being tagged as disease threats, it may also represent many members of racial, ethnic, and cultural out-groups as potential threats. This could be attributable to phenotypical group differences, such as skin tone, as well as cultural differences in diet, hygiene, or sexual practices. Indeed, disgust sensitivity consistently predicts opposition to immigration, primarily for ethnically and culturally distinct out-groups (Aarøe et al., Reference Aarøe, Petersen and Arceneaux2017; Faulkner et al., Reference Faulkner, Schaller, Park and Duncan2004). A recent series of experiments provide evidence that this pattern is a by-product of pathogen avoidance mechanisms, rather than a direct response to out-group membership (Petersen, Reference Petersen2017; van Leeuwen & Petersen, Reference van Leeuwen and Petersen2018). In other words, disgust sensitivity motivates anti-immigration views because an overly sensitive system tags phenotypical and cultural differences as potential indicators of illness.
The predictions of the BIS model fit well within the existing research on immigration. For example, numerous studies have found that White Americans tend to react more negatively to ethnically and culturally distinct immigrants. For example, several studies have shown that country of origin weighs heavily in immigration attitudes (Brader et al. Reference Brader, Valentino and Suhay2008; Erisen and Kentmen-Cin 2017; Hainmueller & Hangartner Reference Hainmueller and Hangartner2013; Hartman et al., Reference Hartman, Newman and Bell2014). Similarly, higher levels of acculturation, signaled by language proficiency, consistently influence immigration attitudes (Hainmueller & Hopkins, Reference Hainmueller and Hopkins2015; Newman et al., Reference Newman, Hartman and Taber2012; Sniderman et al., Reference Sniderman, Hagendoorn and Prior2004). Thus, many of the characteristics of immigrants that seem to drive opposition to immigration fit within the BIS theory.
One of the most novel implications of the BIS theory, however, has not been extensively tested. Because of the evolutionary origins of the BIS, pathogen disgust sensitivity ought to be linked to anti-immigration attitudes in a wide variety of cultural contexts. However, a systematic literature review revealed that nearly all of the research testing the BIS hypothesis had been conducted in Canada or the United States, and most of this work relied on student samples (Aarøe et al., Reference Aarøe, Petersen and Arceneaux2017). There are, of course, some exceptions in the literature. For example, researchers found a significant relationship between disgust sensitivity and opposition to immigration in a student sample in Switzerland (Green et al., Reference Green, Krings, Staerkle, Bangerter, Clemence, Wagner-Egger and Bornand2010). Other studies have used convenience samples in other countries to assess the relationship between disgust sensitivity and concepts related to immigration, such as social sominance orientation (data from 30 nations; Tybur et al., Reference Tybur, Inbar, Aarøe, Barclay, Barlow, de Barra, Becker, Borovoi, Choi, Choi, Consedine, Conway, Conway, Conway, Adoric, Demirci, Fernández, Ferreira, Ishii and Žeželj2016) and travel bans in the context of a pandemic (data from Singapore; Moran et al., Reference Moran, Goh, Kerry and Murray2021), but they did not directly measure immigration attitudes. More recent work has used large, nationally representative samples in both the United States (Kam & Estes, Reference Kam and Estes2016) and Denmark (Aarøe et al., Reference Aarøe, Petersen and Arceneaux2017). Overall, however, the existing body of research relies heavily on convenience samples and evidence from the United States and Canada.
The heavy reliance on data from the United States and Canada is problematic for at least two reasons. First, it is possible that the link between disgust sensitivity and immigration attitudes is culturally bound. For example, patterns of political rhetoric or media coverage within a particular country may create a connection between disgust and immigration where one might not otherwise exist. Second, these findings have also focused almost entirely on the attitudes of Whites toward Latino and African immigrants. Thus, examining the BIS hypothesis across a wide variety of countries and cultural contexts is critical for testing the theory.
The common use of convenience samples, and college samples in particular, also raises questions about the generalizability of past work. While some research suggests that convenience samples and representative samples tend to produce similar relationships between personality traits and political variables (Vitriol et al., Reference Vitriol, Larsen and Ludeke2019), there is also evidence that these relationships tend to be overestimated in convenience samples (Clifford et al., Reference Clifford, Jewell and Waggoner2015). This may be due to higher levels of political knowledge, and thus more constrained belief systems in convenience samples (Kalmoe, Reference Kalmoe2020).
Overall, the BIS hypothesis has received support across a number of studies. However, in spite of the universal applicability of the hypothesis, it has been tested in only a narrow set of countries and often relies on convenience samples. In the next section, we provide novel tests of the BIS hypothesis in five nations: Norway, Sweden, Turkey, Mexico, and the United States.
A comparative approach to immigration attitudes
The context for immigration varies across our five countries in a number of ways that influence immigration attitudes. To begin with, economic conditions differ substantially. Mexico and Turkey are distinct from the other three countries in having a lower gross domestic product (GDP) per capita as well as higher poverty and inequality rates than the United States, Sweden, and Norway. Turkey, the United States, and Mexico have lower levels of educational attainment among migrants than Norway and Sweden (Organisation for Economic Co-operation and Development [OECD], 2022). The United States and Norway have more similar rates of economic participation between foreign-born and native populations than the other three countries (OECD, 2022). These countries also vary in the generosity of their welfare policies. Norway and Sweden are distinct from the other three countries in having higher social expenditures, as a percentage of their GDP, than the average reported for the OECD (OECD, 2022). Overall, the economic conditions in Mexico and Turkey make them more likely to perceive immigrants as an economic threat than the United States, Norway, and Sweden. Notably, according to our theory, economic factors should be less relevant to disgust sensitivity, but this variation across countries helps establish the generalizability of claims about the role of disgust sensitivity.
Our five focal countries also differ substantially in the social and cultural context for immigration. In Norway, Sweden, and the United States, international migrants make up between 15% and 20% of the population. However, this number is much lower in Turkey (7%) and Mexico (less than 1%; United Nations, 2019). The composition of these immigrant groups differs as well. Refugees make up a small share of international migrants (less than 15%) in all of our cases except Turkey, where nearly two-thirds of all migrants are refugees. Citizens tend to be more accepting of immigrants who are fleeing war or persecution, rather than seeking economic opportunities (Bansak et al., Reference Bansak, Hainmueller and Hangartner2016; Hainmueller & Hopkins, Reference Hainmueller and Hopkins2015).
Perhaps most importantly for our theory, there is variation in the ethnic and cultural similarity between immigrant groups and their host population, which plays a large role in immigration attitudes. For example, shared language and religious identity all influence acceptance of immigrants (Bansak et al., Reference Bansak, Hainmueller and Hangartner2016; Hainmueller & Hopkins, Reference Hainmueller and Hopkins2015). Norway and Sweden are both relatively ethnically homogenous in comparison to the United States, Turkey, and Mexico. Ethnic differences between the host population and immigrant groups are perhaps most stark in Norway and Sweden, where the political focus has been on Middle Eastern immigrants, particularly Muslims. In Turkey, however, Syrian immigrants share a religious identity with a majority of the host population. Although there are more political and social dissimilarities between Turkey and Syria, on religious grounds, a significant majority of the Syrian refugees residing in Turkey are Sunni Muslims, the main religious sect in the country. In Mexico, there are relatively few cultural and ethnic differences between the host population and immigrant groups. Qualitative evidence illustrates how both groups can fake their accents and learn a few key words to either blend in with the crowd or tap into people’s sympathy (Acosta-García & Martínez-Ortiz, Reference Acosta-García and Martínez-Ortiz2015; Castañeda et al., Reference Castañeda, Manz and Davenport2002).
Overall, while disgust sensitivity should play a broad role in immigration attitudes, the expectations are clearest for Norway and Sweden, and the weakest for Mexico.
Overview of the studies
To provide a more comprehensive test of the BIS theory of immigration attitudes, we rely on six surveys conducted in five countries. Each study differed in methodology and measures, which we detail here. Crucially, however, the Norway, Sweden, Turkey, and Mexico surveys are all probability samples of the population, while the US surveys rely on internet panels.
US sample 1 . Respondents were recruited through Qualtrics Panels in May 2016. After excluding inattentive respondents, 786 respondents completed the survey. The sample is not nationally representative, but invitations to participate in the survey were balanced on census demographics (age, gender, ethnicity, and census region) and partisanship. As a result, the sample is highly diverse and similar to the population in many dimensions.
US sample 2 . A total of 2,462 respondents were recruited through the Lucid platform in April 2019. Lucid uses a quota sample to match to US Census Bureau demographic margins on gender, ethnicity, education, region, age, and income. Lucid samples tend to closely resemble the demographic composition of nationally representative samples (Coppock & McClellan, Reference Coppock and McClellan2019).
Norwegian sample . Data from Norway come from the Norwegian Citizen Panel (Norsk Medborgerpanel) hosted by the University of Bergen. Participants were drawn based on a probability sample of the general Norwegian population, with an average active participant pool of about 10,000. Each wave of the survey constitutes a representative cross-section of the Norwegian population. Our data are drawn from wave 3 (October–November 2014), and a subsample of 619 respondents completed our questions about disgust sensitivity. The overall panel recruitment rate for wave 3 was 23.1 percent.
Swedish sample . Data from Sweden come from the Citizen Panel (Medborgarpanelen or MP), which is an online panel survey from the Laboratory of Opinion Research (LORE) hosted at the University of Gothenburg. Each wave of the Citizen Panel is a probability sample recruited through multiple modes of contact. Our questions were included in LORE Citizen Panel 21, which was fielded between May 31 and June 23, 2016. A total of 1,396 respondents completed our disgust sensitivity battery.
Turkish sample . Data come from a stratified random probability sample of 1,224 Turkish voters, which was fielded by Infakto RW. The distribution of the sample across geographical areas and provinces is based on the Nomenclature of Territorial Units for Statistics classification in order to cover the whole country including urban and rural settlements. These interviews include an oversample from four municipalities (Adana, Mersin, Şanlıurfa, and Mardin) in the south and southeastern parts of the country, where Syrian refugees have settled in larger numbers.Footnote 2 Interviews were conducted face-to-face during May 5–18, 2017. The average length of the interview was approximately 24 minutes. According to American Association of Public Opinion Research (AAPOR) standards, the response rate in our study was 19%, the cooperation rate was 36%, and the refusal rate was 34%.
Mexican sample . Data come from a quarterly omnibus survey fielded by Buendía and Laredo in May 2019. The study consists of a stratified probability sample of 1,000 Mexican adults enrolled to vote, 18 years and older, and residing in housing units within the national territory. Interviews were conducted face-to-face. According to AAPOR standards, the response rate was 59%, the cooperation rate was 84%, and the refusal rate was 11%.
Measures and descriptive statistics
Disgust sensitivity
With a few minor variations, all five studies included the same seven-item measure of pathogen disgust sensitivity, a subscale of the Three Domains of Disgust Scale (TDDS; Tybur et al., Reference Tybur, Lieberman and Griskevicius2009).Footnote 3 Although there are alternative measures of disgust sensitivity, the pathogen disgust subscale of the TDDS is argued to be the best available measure of the BIS (Tybur et al., Reference Tybur, Frankenhuis and Pollet2014). The scale asks respondents to imagine a series of scenarios, such as seeing mold on leftovers in your refrigerator, and then rate how disgusting that scenario is on a scale ranging from “not disgusting at all” (1) to “extremely disgusting” (5). The scale has been extensively validated using a variety of methods (Olatunji et al., Reference Olatunji, Adams, Ciesielski, David, Sarawgi and Broman-Fulks2012; Tybur et al., Reference Tybur, Lieberman and Griskevicius2009). Olatunji and colleagues (Reference Olatunji, Adams, Ciesielski, David, Sarawgi and Broman-Fulks2012) use a behavioral avoidance task (e.g., would the participant be willing to touch an object) as well as an image-viewing galvanic skin conductance measure to demonstrate that the TDDS is correlated with behavior and physiology.Footnote 4 Additionally, pathogen disgust sensitivity, as measured by the TDDS, is substantially heritable (Sherlock et al., Reference Sherlock, Zietsch, Tybur and Jern2016) and reliably perceived by others (Karinen et al., Reference Karinen, Tybur and de Vries2019). Evidence also suggests that sex differences in pathogen disgust sensitivity using the TDDS are small, and the scale measures the same constructs across the sexes (Tybur et al. Reference Tybur, Bryan, Lieberman, Caldwell Hooper and Merriman2011; but see Balzer & Jacobs, Reference Balzer and Jacobs2011).
In addition to being translated into each country’s native language, there is one difference in how the scales were administered. In discussion with colleagues in Sweden, it was determined that a scenario involving a cockroach would be more familiar to respondents if it were replaced with a scenario regarding a mouse. This item was also used in the Turkish sample. Additionally, we were unable to include all seven items in the Swedish survey, so one item was excluded from the scale. The full text of the items is shown in Table 1.
Measuring pathogen disgust sensitivity.
Notes: * p < .05; ** p < .01; *** p < .001, two-tailed. All variables scaled to range 0–1.
Immigration attitudes
For some of the surveys, we had considerable control over the content of the immigration questions asked, but not for all surveys. As a result, most of our questions were not asked in all surveys, and not all were asked in identical form. As detailed here, we focus our attention on five questions that were asked in similar form in at least three countries. These questions involve preferred immigration levels, whether immigrants bring disease, whether immigration threatens national identity, whether immigrants should have access to social welfare programs, and whether begging or panhandling should be banned. Regarding the last item, panhandling is strongly associated with immigrants in many countries, though perhaps less so within the United States. In addition to analyzing these items, we also use all of the available items in each study for factor analysis, and so we discuss each additional item in the text below.
US study 1 (Qualtrics) . In addition to four common items, the US 1 study included two questions about accepting 75,000 Syrian refugees into the United States and providing financial aid to countries hosting Syrian refugees.
US study 2 (Lucid) . In addition to two common items, the US 2 study included a question about accepting Syrian refugees into the United States.
Norwegian study . In addition to two common items, the Norwegian study asked whether “Norwegian Muslims have greater loyalty to other Muslims in the world than to people in this country.”
Swedish study . In addition to four common items, the Swedish study asked two questions about encouraging immigrants to leave Sweden and whether police should be able to interrogate anyone who they believe is in the country illegally.
Turkish study . In addition to three common items, the Turkish study asked about encouraging immigrants to leave, giving citizenship to refugees who make large financial investments in Turkey, and whether refugees should be given the right to work.
Mexican study . Four common items were included.
Results
We begin our analysis with a discussion of descriptive statistics across each country. All p-values reported in the text are two-tailed. The pathogen disgust scale has not been widely used in comparative research, so it is worth some attention to measurement. Table 1 displays the item means, scale means, and standard deviations for each study. To maintain consistency with the following analyses, we rescale each item to range from 0 to 1. The seven items formed a reliable scale in each country, with alphas ranging from 0.68 (Sweden) to 0.85 (Turkey). Thus, we take the arithmetic mean of the seven items as our measure of disgust sensitivity.
Past work has found consistent gender differences, with women more likely to report higher levels of pathogen disgust sensitivity (e.g., Tybur et al., Reference Tybur, Bryan, Lieberman, Caldwell Hooper and Merriman2011). Given the importance of gender differences in past research, we examine whether these differences replicate in each of our samples. The results are shown at the bottom of Table 1. In each case, we find that women score significantly higher than men (ps < .05), with effect sizes ranging from 0.03 (Mexico) to 0.08 (Turkey), or about 0.16 to 0.50 standard deviations. These findings support the generalizability of gender differences in pathogen disgust and the utility of the scale.
Turning to immigration attitudes, Table 2 displays the question wording and descriptive statistics for the five common items. Starting at the top, four studies asked about immigration levels. In the US samples, 39% and 44% of respondents wanted to decrease immigration levels, while 34% in Sweden wanted to halt all immigration. In Mexico, 69% of respondents reported wanting to decrease low-skilled immigration, and 83% of our Turkey sample reported wanting to decrease the number of refugees allowed into the country. While these questions are not directly comparable, there is clearly considerable anti-immigration sentiment in each country.
Immigration attitudes and descriptive statistics.
Note: All variables scaled to range 0–1.
While the BIS is expected to operate at an implicit level, we were able to include a question in four of our studies assessing the extent to which people explicitly associate immigrants with disease. In three countries—the United States, Turkey, and Mexico—a majority of respondents (54% to 58%) agree that immigrants increase the risk of disease outbreaks. Only in Sweden does this fall below a majority, but still 35% agree with the statement. Thus, many people make an explicit association between immigrants and disease, and this association is not isolated to any individual country. This suggests that this association is not simply a function of the local political context.
Our studies also shared three questions assessing the perceived social and economic costs of immigration. Starting with social costs, a substantial proportion of respondents indicated that immigration posed a threat to their country’s national identity in Sweden (36%), Norway (45%), and Mexico (39%). Turning to economic benefits, US respondents were the most opposed to undocumented immigrants receiving welfare benefits (71%), while close to half of the respondents opposed this policy in Sweden (49%), Turkey (53%), and Mexico (42%). A similar number in Norway opposed refugees receiving the same social assistance as citizens (45%). Finally, three countries also included a question about banning panhandling or begging, which is commonly associated with immigrants. Support was lowest in the United States (47%, 40%), where the association between panhandling and immigrants is weaker, while support was high in Norway (65%) and Turkey (85%).Footnote 5 Overall, there is considerable anti-immigration sentiment in all five countries, including Turkey and Mexico, whose native populations have greater cultural and ethnic similarity to the dominant immigrant population. Sweden, on the other hand, tended to display the lowest levels of anti-immigrant sentiment.
The effects of disgust sensitivity on immigration attitudes
In this section, we turn to our core tests of the relationship between pathogen disgust sensitivity and immigration attitudes. We again focus our attention on the common items across our studies and model each outcome using ordinary least squares (OLS). Following previous work (Aarøe et al., Reference Aarøe, Petersen and Arceneaux2017), we control for basic sociodemographics (age, gender, and education) and political identity (ideological identification and/or partisan identity, as available).Footnote 6 Additionally, we control for race and ethnicity, as appropriate. Specifically, we include a dummy variable for White respondents in the United States and a dummy for Kurdish respondents in Turkey.
For the broadest test of the BIS hypothesis, we conducted a factor analysis of all of the available immigration questions in each country, including the noncommon items discussed earlier (see Appendix for details). Each analysis is restricted to a single factor and estimated using maximum likelihood. All items are scored such that higher values correspond with greater opposition to immigration. To aid interpretation, disgust sensitivity is standardized. Figure 1 plots the OLS coefficients for each country (see Appendix for model details).Footnote 7
Disgust sensitivity and opposition to immigration (factor score). Figure displays estimated regression coefficients and 95% confidence intervals.
As is clear, the BIS hypothesis is supported across all countries. All coefficients are statistically significant, even when applying the Holm correction for multiple comparisons. Of course, because different sets of items were used to construct the factor scores, the coefficients may not be directly comparable across countries. However, it is worth examining the effects within each country. Across the various models, a one standard deviation increase in disgust sensitivity is associated with an increase in opposition to immigration that ranges from 0.09 standard deviations to 0.21 standard deviations. For comparison, we examine education, which has long been considered an important force in immigration attitudes (e.g., Hainmueller & Hiscox, Reference Hainmueller and Hiscox2007). Across these same models, the standardized magnitude of holding at least a college degree ranges from .09 to .25. Thus, the effects of disgust sensitivity are quite similar in magnitude to the effects of a variable that has long been considered a crucial factor in immigration attitudes.Footnote 8 As shown in Table 3, whether or not we control for ideological and partisan identification makes little substantive difference, consistent with only weak relationships between disgust sensitivity and these variables (Kam & Estes, Reference Kam and Estes2016; Terrizzi et al., Reference Terrizzi, Shook and McDaniel2013).
Robustness checks on the effects of pathogen disgust sensitivity.
Note: Standardized coefficients presented for pathogen disgust sensitivity in the corresponding model. * p < .05; ** p < .01; *** p < .001, two-tailed. Blank cells represent models that cannot be estimated due to unavailable measures.
We can make better comparisons across countries by focusing on single questions that were similar or identical across multiple countries. The coefficients for disgust are displayed for each outcome and each country in Figure 2 (see Appendix for model details). All dependent variables are scored such that higher values indicate greater opposition to immigration. The top panel displays the results for increasing or decreasing immigration levels. Disgust sensitivity has a positive and significant effect in all three countries available (Turkey, Norway, United States), and of roughly similar magnitude. Standardized effects range from .07 to .13. To examine whether there is meaningful variation across countries, we estimated two alternative models: one with country dummy variables, and one in which country dummies are interacted with disgust sensitivity, allowing the relationship to vary across the country. Model fit is slightly better for the simpler model according to the BIC and Bayes factor, but the evidence is equivocal (for further details, see Appendix).
Effects of disgust sensitivity on immigration attitudes. Figure displays estimated regression coefficients and 95% confidence intervals.
Turning next to whether immigrants spread disease, the effect is positive and significant in all four countries we tested (United States, Mexico, Turkey, Sweden). The next panel down displays results for the question of whether immigration threatens the identity of the host nation. The effect of disgust is positive and significant for Sweden and Norway, but it is small and not statistically significant in Mexico. However, we cannot rule out a modest association in this case. The last two panels focus on policies that are more oriented toward economics. In all five cases, including Turkey and Mexico, disgust is associated with opposition to providing welfare to immigrants and there is little apparent variation in effect size. Finally, the effect of disgust on banning begging is positive and significant for all three countries. Overall, the effects of disgust sensitivity are largely consistent across the five questions that were asked in at least three countries, supporting the generalizability of the relationship between disgust sensitivity and immigration attitudes.
Robustness to alternative explanations
Additional variables measured in each dataset allow us to test alternative explanations of the relationship between disgust sensitivity and immigration attitudes and explore the robustness of this relationship. To simplify our analyses, we focus on the immigration factor score as the outcome in each case.Footnote 9 We estimated a series of models predicting immigration attitudes as a function of disgust sensitivity and alternative sets of covariates. To briefly summarize these results, Table 3 displays standardized coefficients for disgust sensitivity in each model with asterisks denoting statistical significance. Blank cells represent models that cannot be estimated due to unavailable covariates.
To summarize the findings, disgust sensitivity is significantly related to anti-immigration attitudes in every single model we estimated. This includes models with no covariates, models with only demographics, and models that include ideological identity and/or partisanship (in addition to demographics). Thus, our findings are robust to a variety of specifications and not driven by suppression effects (Lenz & Sahn, Reference Lenz and Sahn2021). More importantly, the effect size is largely unaffected by the inclusion of covariates.
The lower half of the table presents a series of models that allow us to test a number of alternative hypotheses. Each of these models includes all of the controls described earlier. The primary alternative explanation is that disgust sensitivity is related to social conservatism, but that disgust has no unique effects on immigration attitudes. For example, it may be that disgust sensitivity motivates social conservatism because of attitudes toward sexuality, such as gay rights and abortion, and the relationship between disgust sensitivity and immigration attitudes is only due to their shared relationship with social conservatism (Billingsley et al., Reference Billingsley, Lieberman and Tybur2018; Tybur et al., Reference Tybur, Inbar, Güler and Molho2015). On this view, the link between disgust sensitivity and immigration attitudes is spurious. We can test this in multiple ways in the United States and Norway. First, one of our samples (US 1) includes the full TDDS, including sexual disgust sensitivity. If the relationship between pathogen disgust sensitivity and immigration attitudes is entirely driven by attitudes connected to sexuality, then controlling for sexual disgust sensitivity should eliminate the effect of pathogen disgust sensitivity (Billingsley et al., Reference Billingsley, Lieberman and Tybur2018; Shook et al., Reference Shook, Terrizzi, Clay and Oosterhoff2015). However, consistent with our theory, pathogen disgust sensitivity remains significantly related to immigration attitudes, while sexual disgust sensitivity is not. Additionally, both US studies and Norway include issue attitudes related to sex, such as abortion, same-sex marriage, polygamy, and transgender rights. In each study, we created an index of these social policy attitudes and added them as a control.Footnote 10 If the relationship between pathogen disgust sensitivity and immigration attitudes is merely a product of more general social conservatism, then controlling for social policy attitudes should eliminate any effect of pathogen disgust. However, in each case, the effect of pathogen disgust remains positive and significant. Thus, social conservatism and sexual attitudes do not explain the consistent relationship between disgust sensitivity and immigration attitudes.
In addition, we have a host of other variables that are related to immigration attitudes. Similar to Aarøe, Petersen, and Arceneaux (Reference Aarøe, Petersen and Arceneaux2017), we find that controlling for the Big Five personality traits does not affect our inferences. Neither do controls for moral disgust sensitivity, authoritarianism, or humanitarianism substantively affect our results. Overall, the results are quite robust to alternative explanations and model specifications.
Conclusion
As the topic of immigration increasingly takes center stage in politics, understanding the nature of immigration attitudes among the mass public is as important as ever. Psychological research suggests that the intensity of immigration attitudes may derive in part from an evolved disease avoidance mechanism. In this article, we expanded on this research by providing the broadest test of this explanation to date. Our results show consistent support for the BIS hypothesis across five countries that vary considerably in economic and social conditions, as well as cultural distance between the host and immigrant populations. These findings support the universality of the BIS hypothesis.
While our findings rely on observational methods, they are consistently robust to a variety of alternative potential explanations, including controls for partisanship and ideological identification, issue-based measures of social conservatism, sexual disgust sensitivity, as well as other personality traits and psychological dispositions. Additionally, the magnitude of the effects of disgust sensitivity is similar to the effects of education, a variable that has long been concerned a central factor in explaining immigration attitudes. Thus, we confirm a substantively important relationship between pathogen disgust sensitivity and immigration attitudes that is not readily explained by alternative factors. Nonetheless, there is recent evidence that many stable personality traits may not be exogenous to politics (Bakker et al., Reference Bakker, Lelkes and Malka2021; Boston et al., Reference Boston, Homola, Sinclair, Torres and Tucker2018; Luttig, Reference Luttig2021), and thus further research is needed to identify a causal effect of disgust sensitivity.
While we found consistent support across all five countries we studied, the findings were similar across countries. This runs contrary to the expectation that greater ethnic and cultural similarity between immigrant and host populations should decrease the magnitude of the relationship between disgust sensitivity and opposition to immigration. However, we were unable to directly measure perceived ethnic or cultural similarity and may have had insufficient statistical power to observe variation across countries. Thus, an important avenue for future work is to explore the specific mechanisms that drive the relationship between disgust sensitivity and immigration attitudes, and how the levels of the mechanisms vary across countries and contexts.
Understanding the source and nature of anti-immigration attitudes may help provide new insight into dealing with this form of intergroup conflict. Social contact has long been considered one of the most promising solutions to conflict. However, those who are high in disgust sensitivity are the most likely to avoid such contact (Aarøe et al., Reference Aarøe, Petersen and Arceneaux2017), creating a potential roadblock to this solution. Additionally, people high in disgust sensitivity may be more likely to have negative experiences with intergroup contact, in turn affecting intergroup attitudes (Sirin et al., Reference Sirin, Valentino and Villalobos2017). Thus, gaining an understanding of the emotional roots of immigration attitudes may help to understand the challenges to attitude change.
Supplementary Materials
To view supplementary material for this article, please visit http://doi.org/10.1017/pls.2022.6.
