Fairweather Cosmopolitans: Immigration Attitudes in Latin America During the Migrant Crisis

Abstract

What explains voter attitudes toward immigration in Latin America? This article argues that increased refugee arrivals moderate the impact of social identities on immigration attitudes. We propose that informational cues associated with increased immigration make cosmopolitan identities less important—and exclusionary national identities more important—determinants of immigration preferences. Analyzing 12 Latin American countries from the 2017–2022 wave of the World Values Survey, we demonstrate that cosmopolitanism is positively associated with pro-immigration attitudes, but only in countries experiencing low-to-moderate refugee inflows. Conversely, nationalism is negatively associated with pro-immigrant attitudes, and increasingly so as refugee inflows increase. The uneven distribution of refugee migration has therefore reshaped public opinion in Latin America by moderating the effects of competing social identities (i.e., cosmopolitanism and nationalism). These findings contribute to broader debates on the behavioral impacts of immigration by highlighting an indirect mechanism by which increased immigration may generate anti-immigrant hostility.

Scholars have long been interested in studying the impact of refugee crises on immigration attitudes. This literature is fixed on two opposing arguments. First, advocates of the contact hypothesis posit that, under select conditions, increased ingroup-outgroup interactions and growing ethnic diversity can reduce prejudices toward foreigners (e.g., Allport 1954). Refugee arrivals may consequently push immigration attitudes in a more open direction (Liebe et al. 2018; Newman 2013) or at least keep native attitudes stable (Kustov et al. 2021; Lancaster 2022; Stockemer et al. 2020). Alternatively, proponents of realistic group conflict theory (RGCT) argue that sudden influxes of migrants generate competition over resources (e.g., Blalock 1967; Forbes 1997) and stoke natives’ concerns about immigrants as a threat to their culture or safety (Brader et al. 2008). In these ways, even transient refugee flows can cultivate popular support for anti-immigrant policies and far-right xenophobic parties (Hangartner et al. 2019; Vasilakis 2018).

Not only are the effects of refugee flows on public opinion not obvious, but until recently, quantitative research on immigration attitudes has overwhelmingly focused on the Global North while neglecting public opinion in developing and middle-income countries. ¹ This empirical gap is noteworthy given that developing countries currently host the most refugees due to their geographic proximity to major migrant-sending states (United Nations High Commissioner for Refugees 2023). Furthermore, in contrast to many Western democracies, ² Latin America stands out for its more welcoming set of refugee policies (e.g., Acosta and Freier 2018; Betts et al. 2023; Cantor 2018; Hammoud-Gallego 2022; Hammoud-Gallego and Freier 2023).

This more liberal approach toward accepting refugees has nonetheless faced an unprecedented test. Since 2015, over 7 million people have left Venezuela (Gandini and Selee 2023), with hundreds of thousands also continuing to flee from gang violence and abject poverty in the Northern Triangle region of Central America. ³ Most of these migrants have moved to fellow Latin American countries, notably Colombia, Peru, and Ecuador (United Nations High Commissioner for Refugees 2023). International organizations such as the World Bank have lauded Latin America for exemplifying “regional solidarity in the face of an unprecedented migration” (World Bank 2019, 5). Yet, during this same period, Latin America has moved in a “securitist” direction (Brumat and Vera Espinoza 2024), reflecting more closed policies that have coincided with the rise of anti-immigrant parties of the “new right” (Domenech 2017). ⁴ Due to fears of incentivizing further migration, most of the region has proved unwilling to extend Venezuelans refugee status under the Cartagena Declaration (Freier and Parent 2019; Freier et al. 2022). Even in countries like Colombia that have pioneered accommodative refugee policies, xenophobic rhetoric and popular anti-immigrant sentiment are a major concern.

Why do some Latin Americans support openness to migrants while others do not? And how, if at all, have unprecedented refugee flows reshaped public opinion about immigration within Latin America? In contrast to the contact hypothesis and RGCT, we propose an indirect pathway by which refugee crises generate shifts in immigration attitudes. Drawing on identity-based motivation (IBM) theory, we argue that refugee inflows make cosmopolitan identities less important and nationalist identities more important for natives’ attitudes toward immigrants. This occurs due to contextual cues stemming from a refugee crisis: actual or perceived competition over scarce resources, increased social comparison with immigrant groups, and political rhetoric that primes exclusionary national identities. Accordingly, refugee arrivals generate more opposition to immigration by simultaneously attenuating the effect of cosmopolitanism (a negative predictor of anti-immigration sentiment) and accentuating the effect of nationalism (a positive predictor of anti-immigration sentiment).

We test our argument across 12 different Latin American countries using cross-national survey data from the World Values Survey. Our results are broadly consistent with “fairweather” cosmopolitanism: respondents that ascribe to a cosmopolitan identity (i.e., self-described world citizens) possess more open immigration attitudes, but only in countries experiencing low-to-moderate refugee inflows. At the same time, nationalism is associated with greater anti-immigration sentiment, and increasingly so as refugee inflows increase.

In addition to elucidating public opinion during Latin America’s migrant crisis, these findings contribute to the broader literature on immigration preferences by highlighting an unexplored pathway by which refugee inflows incite opposition to immigration: refugee inflows moderate the impact that voters’ social identities have on their immigration preferences, causing even self- described cosmopolitans to adopt more nationalist-looking attitudes. We also demonstrate the generalizability of our theory by expanding our sample to include migrant-receiving countries outside of Latin America.

Cosmopolitanism, Nationalism, and Immigration Attitudes

Whereas political economists have traditionally viewed immigration preferences as driven by concerns about labor market competition (Scheve and Slaughter 2001) and public finance (Hanson et al. 2007), research from behavioral and political psychology consistently finds non-material considerations to be more powerful predictors of immigration attitudes (Hainmueller and Hopkins 2014). According to this research program, anti-immigration sentiment is primarily rooted in low-tolerance attitudes—including nationalism, ethnocentrism, and racism—as well as sociotropic beliefs about how immigration influences the culture, economy, and security of the nation as a whole (e.g., Hainmueller and Hiscox 2007). These attitudes are tied to perceptions of migrants as a distinct outgroup that threatens the nation’s social fabric and natives’ dominant socioeconomic position. Tolerance-based explanations also clarify trends in the empirical evidence left unexplained by egocentric preference models, such as discrimination against specific immigrant cohorts on the basis of national origin, race, language, and religion (e.g., Creighton and Jamal 2015; Hainmueller and Hangartner 2013; Hopkins 2015).

The idea of cosmopolitanism rests at the core of these theories. Cosmopolitanism can be conceptualized as a social identity in which individuals perceive themselves as a citizen of the world (e.g., Bayram 2019; Ossewaarde 2007; Vertovec and Cohen 2002), viewing “humanity as their moral community and subscrib[-ing] to the idea of moral obligations owed to all human beings” (Bayram 2019, 760). In this way, cosmopolitans value improving the welfare of all people—even those culturally and geographically distant from themselves—and possess a natural affinity for global cooperation (Buchan et al. 2011), respect for international law (Bayram 2017), and tolerance for opposing ideas (Stoeckel and Ceka 2023). These same predispositions encourage cosmopolitans to abandon negative stereotypes toward outgroups and instead acknowledge the benefits of cultural diversity and global integration. For this reason, too, cosmopolitan identities underlie sociotropic considerations driving preferences toward globalization (Mansfield and Mutz 2009).

Cosmopolitanism is therefore a natural companion to pro-immigration attitudes. Not only are cosmopolitans inherently appreciative of the social and economic benefits brought by migrants, but their attachment to world citizenship as an identity reinforces more liberal immigration preferences. As individuals begin to self-identify as world citizens, they instinctively adopt beliefs and values congruent with typical cosmopolitan views (Tajfel 1981). Consistent with this, survey research demonstrates that cosmopolitanism is associated with more open immigration attitudes (e.g., Betts 1988; Hainmueller and Hiscox 2007; Pryce 2018; Taniguchi 2021).

National identities, by contrast, are more multidimensional. Although strong national identities frequently imply positive affect toward one’s nation, these identities encompass elements of both nationalism and patriotism, which make divergent predictions about immigration attitudes. Nationalism refers to feelings of national superiority and an emphasis on outcompeting other nations or cultures (Feshbach 1994), while patriotism is a self-referential attitude (de Figueiredo Jr and Elkins 2003, 178), comprising love for one’s country and strong feelings of attachment to national ideas and culture. Whereas nationalism tends to correlate with anti-immigrant sentiment, the relationship between patriotism and immigration preferences is less clear (de Figueiredo Jr and Elkins 2003). Moreover, the precise meaning that individuals ascribe to national identity can strongly influence policy attitudes (e.g., Acevedo and Meseguer 2022; Schildkraut 2013; Taniguchi 2021; Wright 2011). For instance, civic conceptions of citizenship—or the normative belief that good citizens should themselves be closely connected to the nation’s culture and politics—may not produce opposition to immigration in the same way as ethnic or racial conceptions of citizenship (Schildkraut 2013).

Nevertheless, the connection between social identities and immigration attitudes remains incomplete and oversimplified. Scholars commonly present cosmopolitan and national identities as fixed dispositional traits that unconditionally push voters’ policy attitudes in a more open or closed direction. Traditional conceptions of cosmopolitanism, for instance, paint society in terms of a stark cosmopolitan-parochial divide with the former class of citizens possessing more globalist attitudes by virtue of their higher educational attainment (Bean 1995; Betts 1988). Similarly, typologies of migrant-receiving countries have separated “settler” nations from “non-immigrant” nations, with the latter category consisting of citizens that place stronger emphasis on ethnic and national identity (Freeman 1995; Meissner 1992). Although these static conceptions of identity are parsimonious, we contend that dispositional factors such as cosmopolitanism must be considered in tandem with situational factors. Certain contexts encourage cosmopolitans to think “like cosmopolitans.” However, other contexts can render cosmopolitan identities less relevant, encouraging otherwise tolerant individuals to form attitudes based on other values and beliefs.

Compounding this issue is the fact that individuals juggle multiple social identities, sometimes concurrently ascribing to identities that appear contradictory. Most notably, Bayram 2019) demonstrates that most cosmopolitans have an integrated dual identity that compromises both world citizenship (i.e., a negative predictor of anti-immigration sentiment) and nationalism (i.e., a positive predictor of anti-immigration sentiment). This raises the question of how individuals reconcile multiple competing identities to form a single policy attitude.

Fairweather Cosmopolitanism

The Latin American refugee crisis—and increasing international migration, more generally—raises a critical question about the role of cosmopolitanism in different migratory contexts: Do cosmopolitans remain pro-immigration even when facing substantially increased immigration levels?

Our argument begins with the fact that people ascribe to different social identities and that these identities motivate individuals’ preferences and choices (Tajfel et al. 1979). Social identities involve individuals’ self-perception—a perception “informed by one’s membership in groups defined by some shared attribute, such as language, religion, or race” (Kalin and Sambanis 2018, 240). Social identities place individuals within an ingroup (“us”) juxtaposed with a certain outgroup (“them”), in turn incentivizing individual actions by prescribing desirable behavior. Because people acquire esteem from their group membership (Leary and Baumeister 1995), individuals pursue choices that align with the values of their ingroup. This identity-congruent behavior may occur even when such choices go against one’s material self-interest.

The idea that social identity motivates individual behavior underpins the connection between cosmopolitanism and immigration preferences. Although cosmopolitanism fails to clearly delineate an outgroup (i.e., “us” comprises all humans), a cosmopolitan identity imparts globalist and pro-tolerance values, setting clear boundaries on identity-congruent behavior. By virtue of being world citizens, cosmopolitans are expected to adopt policy preferences that are consistent with furthering the social welfare of people outside their local or national community. Cosmopolitans are therefore more committed to globalization, including more open immigration policies.

However, people frequently adopt multiple social identities that complicate the pursuit of identity-congruent behavior. Vote choice provides one example of this. During election time, voters may simultaneously self-identify on the basis of race, partisanship, religion, and any other number of ascriptive traits. This is likely to make some form of identity-incongruent behavior unavoidable. For instance, while voting for candidate A may align with the perceived preferences of one’s religious group, voting for candidate B might better align with the prescriptions of one’s party membership. Thus, even if preferences are identity-based, attachment to different group identities often makes “correct” behavior difficult to discern.

Cosmopolitans are not exempt from this conundrum. As Bayram (2019) demonstrates, even self-categorized cosmopolitans tend to cling to a national identity. Because national identities often develop early in one’s life and are continuously reinforced through public discourse, national identities comprise a core part of one’s self that is unlikely to be jettisoned, even when embracing world citizenship. This results in individuals merging world and national citizenship into an “integrated dual identity” (Bayram 2019, 759), in which citizens contemporaneously maintain close attachment to their nation and the global community. We confirm this empirical regularity in our data, finding that feelings of closeness to the world are positively associated with both national pride and closeness to one’s nation. ⁵ In the context of immigration policy, this creates a dilemma for how cosmopolitans form their immigration policy preferences: Despite the fact that cosmopolitans express a normative commitment to open borders, their enduring attachment to their nation demands nativist policy preferences that privilege citizens at home.

We posit that context—specifically, the migration pressures faced by receiving countries—is essential for understanding how natives’ national and cosmopolitan identities affect their immigration policy preferences. This argument is rooted in identity-based motivation (IBM) theory and its assertion of dynamically constructed social identities (Fisher et al. 2017; Oyserman 2009, 2015). While people favor identity-consistent behavior, IBM theory posits that different contextual factors can trigger which identities are most salient when informing preferences. Considering the multitude of identities that people ascribe to, “the ones that will matter are those that come to mind in context and are experienced as relevant to the situation at hand” (Oyserman 2015, 5). The oft-cited rally-around-the-flag effect demonstrates one instance of this (Mueller 1970). Times of national crisis tend to accentuate national identity while at the same time attenuating the effect that other dispositional characteristics have on presidential approval. Political economy research similarly shows how characteristics of the welfare state partially dictate the prevalence of national identities (Shayo 2009), in turn affecting popular support for immigration policy (Bearce and Connell 2023). Social identities are therefore sensitive to one’s environment, making them potentially unstable predictors of behavior.

In this sense, cosmopolitan identities are indeed real and tend to pull policy preferences in the direction of openness and inclusiveness. However, exposure to sudden increases in immigration increases the salience of competing social identities, prompting cosmopolitan individuals to de-prioritize their world citizenship and instead rely more strongly on exclusionary national identities when forming immigration policy attitudes. Below, we outline several mechanisms by which refugee flows diminish the role of cosmopolitanism: 1) perceptions of material threat; 2) social comparison; and 3) political communications emphasizing exclusionary national identities. These mechanisms are complementary, non-exclusive, and may not necessitate direct contact between natives and refugees. Rather, media coverage of immigration (and political debates around it) can increase perceptions of material and cultural threat, consequently making cosmopolitan identities less decisive predictors of immigration attitudes.

First, individuals may experience or perceive heightened native-immigrant competition over resources during times of high immigration, in turn causing natives to think more based on an exclusionary national identity. Often, this competition is grossly exaggerated. In fact, the majority of empirical work finds that immigration has a negligible effect on native wages while boosting overall productivity and innovation in receiving countries (Peri 2014). ⁶ Threats to natives’ general material and physical security are often similarly more imagined than real. For example, despite perceptions that immigrants increase crime, increasing crime rates in Colombia have (ironically) been driven by crime against immigrants (Knight and Tribin 2023).

Yet, public misconceptions about the economic effects of immigration still prove consequential and are hard to correct (Hopkins et al. 2019). Research in major immigrant-receiving countries indicates that perceptions of material threat are common among natives. Majorities of Colombians, Peruvians, and Ecuadorians believe that immigration “lowers salaries and worsens working conditions” (Rivero 2019, 8). Such concerns also accompany worries about competition for social services and fears about crime. In an environment that is perceived to be zero-sum, national citizenship becomes an increasingly salient group identity used by natives to declare themselves as the “rightful” recipients of jobs and state resources. In this way, perceived material threats make national identities more influential—and cosmopolitan identities less influential—for immigration attitudes. ⁷

Second, immigration pressures encourage natives to cling more closely to their national identities as a means to positively distinguish themselves from migrant groups. As proponents of social identity theory argue, group identities are an important basis of self-esteem (Tajfel et al. 1979; Turner et al. 1987). Individuals are motivated to maintain social identities that favorably distinguish them from other groups and, notably, ingroup favoritism intensifies when group identities are threatened (Voci 2006). Visible immigration can buttress the significance of natives’ identities, leading natives to emphasize positive features of their national identities that distinguish themselves from non-citizens. This is perhaps especially the case with visibly poorer refugees: As economic disparities between migrants and natives become more visible, natives attribute greater significance to their national identity to separate themselves from foreigners and uphold a positive group-image. Consistent with this mechanism, Fouka and Tabellini (2022) find that Mexican immigration in the United States makes native status a more salient ingroup-outgroup distinction for white Americans, lowering prejudice toward Black Americans but increasing prejudice toward foreigners.

Lastly, migration influxes stoke nationalist rhetoric that draws a clear distinction based on citizenship and further stimulates national identities over cosmopolitan identities. As Sniderman et al. (2004) find, priming individuals to think in terms of their national identities increases opposition to immigration. Much of the discourse associated with refugees in migrant-receiving countries likely has this effect. For instance, Ecuadorian presidential candidate, Andrés Arauz, voiced objections over accepting more Venezuelan refugees, stating “our priority is to serve the Ecuadorian people” [emphasis added] (Americas Quarterly 2021). Similarly, Bogotá mayor Claudia López has delivered stigmatizing remarks toward migrants, once labeling Venezuelan immigrants as “violent” and stating, “they [the government] offer everything to Venezuelans, what guarantees are left for Colombians?” [emphasis added] (Soto 2021).

Crucially, elites cues accompanying migration pressures not only encourage natives to lean on their national identity but can also redefine its meaning. This point relates to another key proposition of IBM theory: Situational cues determine not just which identities are salient but also what it means to ascribe to an ingroup (Oyserman 2009). Unfortunately, sudden influxes of immigration tend to prompt elites to cater to ethnic or racial conceptions of national identity which are more closely tied to anti-immigrant sentiment than civic conceptions of national identity (Schildkraut 2013). As Valentino et al. (2013) show, the large-scale arrival of specific migrant cohorts also effectuates media coverage that focuses on negative cues about ethnic outgroups. Thus, the political rhetoric surrounding immigration tends to evoke an exclusionary national identity that works against the idea of globalization fostering greater supranational attachment.

Readers may question whether the arrival of culturally and ethnically similar refugees poses a substantial threat to (and thereby accentuates) natives’ national identities in the Latin American context. However, evidence from the region indicates that anti-immigrant prejudices do not strictly follow ethnic or linguistic lines. For example, in a conjoint experiment, Colombians preferred Peruvian migrants to more-ethnically-similar Venezuelans (Argote and Daly 2024). Further, in Costa Rica—a primary destination for Nicaraguan migrants—about 42 percent of individuals agree or strongly agree that “Costa Rican culture is threatened” by immigrants [emphasis added] (Malone 2019, 11). Studies such as these suggest that cultural threats are socially constructed, making national identities important even in the context of intraregional migration.

In sum, abnormally high immigration is accompanied by contextual cues that increase the importance of exclusionary national identities and decrease the importance of cosmopolitan identities. Pairing this with the stylized fact that cosmopolitan (national) identities are associated with more open (restrictive) immigration attitudes, we arrive at a theory of fairweather cosmopolitanism: Individuals that self-identify as world citizens possess more open immigration attitudes—but only when migration pressures are low. As migration pressures increase, these same individuals rely more on their national identity to form their policy preferences, washing out the effect of cosmopolitanism. This does not necessarily mean that cosmopolitans are insincere or that they drop their cosmopolitan identities in the face of high levels of immigration. Rather, we argue that context shifts which identities individuals subconsciously rely on as they form their preferences. Put simply, cosmopolitans do not always think like cosmopolitans:

H_1A: When the rate of refugee inflows faced by the receiving country is relatively small, cosmopolitanism is positively associated with pro-immigration attitudes.

H_1B: As the rate of refugee inflows faced by the receiving country increases, there is a weaker positive relationship between cosmopolitanism and pro-immigration attitudes.

Whereas existing research highlights the connection between nationalism and anti-immigrant sentiment (e.g., Schildkraut 2013; Taniguchi 2021; Wright 2011), a further observable implication of our theory is that the impact of nationalism on immigration attitudes should intensify during times of increased immigration, as national identity gains salience:

H_2A: When the rate of refugee inflows faced by the receiving country is relatively small, nationalism is negatively associated with pro-immigration attitudes.

H_2B: As the rate of refugee inflows faced by the receiving country increases, there is a stronger negative relationship between nationalism and pro-immigration attitudes.

Research Design

We examine the effects of cosmopolitanism and nationalism on immigration attitudes using individual-level data from Wave 7 of the World Values Survey (WVS). This data is particularly useful for our purposes due to its cross-national coverage and its available measures of cosmopolitanism, nationalism, and immigration attitudes. Importantly, too, the WVS introduced a new battery of immigration-related questions in Wave 7. ⁸

Respondent interviews were conducted between 2017 and 2022, depending on the country. We subset the World Values Survey to include only Latin American countries, leaving us with 12 countries and more than 1,000 respondents in each country. Table S1, in the Supplemental Information, shows the 12 country cases, the number of respondents in each case (before listwise deletion), and the year of data collection. We measure the dependent variable, pro-immigration attitudes, with an index of six items (α= 0.68). The first of these items asks respondents “to evaluate the impact of [immigrants]…on the development” of their country. The remaining five items asks about the perceived effects of immigration: respondents were asked whether they agreed or disagreed that immigration increases crime, strengthens cultural diversity, increases the risks of terrorism, increases unemployment, and leads to social conflict. We recoded all six items so that higher values indicate more pro-immigration attitudes and combined them into a mean index. We then rescaled the immigration index between 0 (strongly anti-immigration) and 100 (strongly pro-immigration) to ease interpretation. The main advantage of using an index over a single-item measure is that it mitigates measurement error and concerns about question wording. In the appendix, we provide additional models that estimate the effect of cosmopolitanism and nationalism on each of these six items individually. ⁹

Figure 1 shows the distribution of the immigration index by country, with dashed lines indicating country means. As shown, there is considerable variation across countries: attitudes about immigration are most negative in Colombia—the migrant-receiving country most affected by the Latin American refugee crisis. In comparison, attitudes about immigration are relatively positive in wealthier and less affected Southern Cone countries, including Argentina, Chile, and Uruguay.

Figure 1. Distribution of the immigration index by country. Larger values indicate more pro-immigration attitudes.

Our main right-hand side variables are cosmopolitanism, nationalism, and immigration flows. The first two of these are individual-level variables drawn from the WVS. We measure cosmopolitanism using a single question drawn from a battery of five items asking respondents “how close they feel” to their town, country, continent, and so on. The cosmopolitanism item asks respondents how close they feel to “the world,” with responses ranging from “not close at all” (0) to “extremely close” (3). Higher values therefore indicate a more cosmopolitan identity. Consistent with other measures of cosmopolitanism (e.g., Bayram 2017; Buchan et al. 2011), this item taps into individuals’ closeness to the global community in contrast to more parochial attachments to town or country. ¹⁰ For our purposes, this measure of cosmopolitan identity is preferable to index-based measures of cosmopolitan orientation (see Pichler 2012), which themselves include attitudes about immigrants (our dependent variable). Research on the measurement of cosmopolitanism also confirms that “closeness to the world” is correlated with other cosmopolitan attitudes (Pichler 2009).

Recall that national identities include elements of both patriotism and nationalism, with the former being a less exclusionary orientation than the latter (Feshbach 1994; de Figueiredo Jr and Elkins 2003). For the purposes of testing H_2A−2B, we intend to capture the more nationalistic component of national identity, which should strengthen as a predictor of anti-immigrant sentiment as refugee flows increase. Our nationalism measure is drawn from a question asking respondents “how proud” they are to be their nationality. While this is not the cleanest proxy of nationalism, we can be reasonably confident that the measure approximates an exclusionary national identity due to its relationship with other variables in the WVS data. Figure S2 in the appendix demonstrates that national pride is negatively associated with trust for other nationalities, affinity toward neighbors of a different race or language, and favorable beliefs about the contribution of migrants. Conversely, de Figueiredo Jr and Elkins (2003) find that feelings of closeness to one’s nation—the most obvious alternative measure in the WVS—loads more heavily onto the patriotic dimension of national identity.

We expect that the effect of national and cosmopolitan identities depends on exposure to immigration and immigration-related messaging. To test this, we rely on country-level indicators of exposure to the Latin American immigration crisis. Our primary indicator is the percent change in the refugee population between 2015 and 2020 based on data from the United Nations Population Division. ¹¹ We use percent change following evidence that sudden increases in immigration are more perceptible—and therefore more consequential for immigration attitudes—than the absolute number of immigrants (Hopkins 2010). We also use change in the refugee stock (rather than total immigrant stock) given that the Latin American immigration crisis involved significant flows of refugees fleeing from Venezuela and Haiti. In any case, change in the refugee population is highly correlated with change in the immigrant population, and we find similar results using the latter indicator. ¹² Across our 12-country sample, the median value is a 61-percent increase in the refugee population. The maximum increase occurred in Colombia, where the arrival of Venezuelan refugees drove a 171-percent increase in the refugee population. The minimum value occurred in Venezuela, where the refugee population decreased by 18.8 percent amidst economic and political turmoil.

The 12-country sample used here includes all of Latin America’s largest migrant-receiving countries (as well as some key migrant-sending countries, such as Venezuela) and is therefore well suited for exploring how, if at all, changes in refugee inflows moderated the effects of natives’ identities in Latin America. Nevertheless, our sample of 12 countries is admittedly quite small for conventional hypothesis testing using multilevel models. Accordingly, we assess support for our hypotheses in multiple steps. First, we examine the bivariate relationship between cosmopolitan or national identities and immigration attitudes by destination country, allowing us to assess where these social identities are most impactful. This also permits us to pinpoint outlier countries where national and cosmopolitan identities are more important for immigration attitudes than expected. We then proceed to a more conventional multilevel modeling framework that introduces a series of control variables and a cross-level interaction between the percent change in refugees and our individual-level indicators of social identity. Lastly, we present a series of robustness checks that corroborate the main results.

Results

Correlations by Country

Figures 2 and 3 present correlations between cosmopolitan (and national) identities and the pro-immigration index by country. These correlations are from ordinary least squares (OLS) models regressing the pro-immigration index on our indicators of cosmopolitan and national identities. We plot these correlations against the percent change in the refugee population between 2015 and 2020. ¹³

Figure 2. Estimated coefficients on cosmopolitanism from OLS models of pro-immigration attitudes.

Consistent with H_1A, Figure 2 suggests cosmopolitanism and pro-immigration attitudes tend to be positively correlated in countries with modest refugee inflows. We find positive and statistically significant correlations in five of the six countries below the median value of percent change in the refugee population (i.e., Nicaragua, Guatemala, Ecuador, Bolivia, and Brazil). For example, a one-point increase in the cosmopolitanism measure is associated with a 1.2-point increase in the pro-immigration index in Bolivia and a 1.8-point increase in Brazil. ¹⁴ In fact, Venezuela—the country of origin for most intra-Latin American refugees—is the only country below the median where we find an insignificant correlation between cosmopolitan identity and pro-immigration attitudes.

In line with H_1B, the positive correlation between cosmopolitanism and pro-immigration attitudes is attenuated as refugee flows increase. It is insignificant and near zero in four of the six countries with higher-than-average increases in their refugee populations (i.e., Argentina, Colombia, Mexico, and Peru). We suspect that perceived native-immigrant competition and anti-immigrant discourse make cosmopolitan identities less salient, thereby weakening the typical positive relationship between cosmopolitanism and pro-immigration attitudes in these destination countries. Uruguay and Chile are notable outliers in that cosmopolitanism and pro-immigration attitudes remain positively correlated despite relatively high levels of refugee inflows.

Figure 2 illustrates that the correlation between cosmopolitanism and pro-immigration attitudes tends to be weaker in countries with higher-than-average refugee inflows. Figure 3 mirrors this, showing stronger negative correlations between nationalism and pro-immigration attitudes in countries experiencing more refugee migration. Contrary to H_2A and previous public opinion research, we do not find any statistically significant correlations between national identity and immigration attitudes in the six countries with below-average refugee inflows. Furthermore, the correlation is only negative in two of these countries: Guatemala and Venezuela. In comparison, correlations between national identity and immigration attitudes are negative and statistically significant in four of the six countries with higher-than-average levels of immigration—a finding that is consistent with H_2B. Increased native-immigrant competition and opportunistic anti-immigrant discourse likely strengthen the link between national identity and anti-immigration attitudes in these countries. Yet, as with cosmopolitanism, there are still outliers: We find near-zero correlations in Chile and Mexico despite relatively high refugee inflows.

Figure 3. Estimated coefficients on national pride from OLS models of pro-immigration attitudes.

It is worth noting that the major outliers in Figures 2 and 3—Chile and Uruguay—are more highly developed and more strongly pro-immigrant than other countries in the sample. While we should not make too much of these simple correlations, cosmopolitan identities appear to be more firmly entrenched in these countries, leaving the link between cosmopolitanism (or nationalism) and pro-immigration attitudes less affected by immigration flows. In addition to this, higher levels of development may make these countries better able to accommodate inflows of relatively high-skilled migrants, resulting in less migrant-native competition and insulating the connection between cosmopolitanism and pro-immigration attitudes.

Multilevel Models

The correlations presented above offer tentative support for the argument that the effects of cosmopolitan and national identities depend on immigration rates. Of course, these bivariate correlations may be spurious—driven by other individual- or country-level variables that strengthen (or weaken) the effects of competing identities on immigration attitudes. We therefore turn to multilevel models that allow us to estimate average effects while controlling for potential confounds. ¹⁵ Specifically, we test our conditional hypotheses through cross-level interactions between the percent change in the refugee population and our two indicators of cosmopolitan and national identity. According to our hypotheses, both interaction terms should be negative. As the rate of refugee inflows increases, we expect a smaller positive effect of cosmopolitan identity and a larger negative effect of national identity.

We opt for a fairly parsimonious specification given the limited number of upper-level units (i.e., 12 countries). At the country-level, we control for level of economic development using ln(GDP per capita). Previous work indicates a link between GDP and pro-immigration attitudes (Boateng et al. 2021; Leong 2008). As a proxy for development, GDP per capita captures countries’ capacity to accommodate immigration flows and, to some extent, cultural changes thought to follow from economic modernization (Inglehart 1997). At the individual level, we control for demographics that are likely related to cosmopolitan and national identities as well as immigration attitudes. These include respondents’ self-reported income group, educational attainment, gender, employment status, and an urban-rural indicator variable. We also include a control for interpersonal trust, which should be positively correlated with both pro-immigrant attitudes and cosmopolitanism. Variable descriptions, including the original survey text, are provided in the Supplemental Information.

Table 1 presents estimates from multilevel models of pro-immigration attitudes. Models 1 and 3 test for the unconditional effects of cosmopolitanism and nationalism. In line with the conventional wisdom, we find a positive effect of cosmopolitan identity and a negative effect of national identity on pro-immigration attitudes. The estimate on the cosmopolitan identity measure indicates that, all else equal, moving from the minimum to the maximum value is associated with a 2.8-point increase in the immigration attitude index. Meanwhile, Model 3 indicates that moving from the minimum to the maximum value of the national identity measure is associated with a 3.9-unit decrease in the immigration attitude index. Thus, on average, more cosmopolitan-identifying respondents tend to be more pro-immigration while more nationalist-identifying respondents tend to be more anti-immigration. However, the interaction terms in Models 2 and 4 indicate that the effects of cosmopolitan and national identities depend on rates of refugee inflows. The estimates on the cross-level interaction terms are negative and statistically significant at the 90-percent level in Models 2 and 4. In other words, increasing refugee inflows appear to weaken the positive effect of cosmopolitanism while amplifying the negative effect of nationalism.

Table 1. Multilevel Models of Pro-immigration Attitudes

Note: ***, **, and * indicate statistical significance levels of .01, .05 and .10 percent, respectively. Random Intercepts at the Country Level. Higher Values of the DV Indicate More Pro-immigration Attitudes

To interpret these interactions, we plot the marginal effects of cosmopolitan and national identities across different rates of refugee inflows in Figures 4 and 5. Figure 4 shows that the estimated effect of cosmopolitanism is positive and statistically significant at low-to-moderate values of refugee inflows. For instance, where there is a 50-percent increase in the refugee population, a one-unit increase in the cosmopolitan identity measure is associated with an approximately 1-point increase in the pro-immigration index. The positive effect of cosmopolitanism then diminishes as refugee inflows increase, becoming near zero and insignificant once the rate of refugee inflows exceeds 120 percent.

Conversely, Figure 5 shows that the estimated effect of national identity is near zero in countries with minimal refugee inflows but becomes increasingly negative and significant as refugee inflows increase. Specifically, when the percent change in the refugee population is below about 30 percent, the estimated effect of national identity is near zero and insignificant. Above that point, the negative effect of nationalism increases with the rate of refugee immigration. When the refugee immigration rate is about 100 percent, for example, a 1-unit increase in the national identity measure is associated with an approximately 1.3-point decrease in the pro-immigration index.

Figure 4. Marginal effect of cosmopolitan identity conditional on % change in refugee population. Note: Estimates are from Model 2. 95% confidence intervals are shown.

Figure 5. Marginal effect of national identity conditional on % change in refugee population. Note: Estimates are from Model 4. 95% confidence intervals are shown.

The marginal effects presented in Figures 4 and 5 are consistent with fairweather cosmopolitanism. We find that cosmopolitanism is significantly and positively associated with pro-immigration attitudes—but only in countries with low-to-moderate levels of refugee inflows. By contrast, nationalism is negatively associated with pro-immigration attitudes—but only in countries with at least moderate levels of immigration.

In addition, we find a negative and fairly substantial direct effect of refugee immigration on pro-immigration attitudes. Based on Model 1, a 10-percentage point increase in the refugee population is, on average, associated with a 1.3-point decrease in the pro-immigration index. This correlation suggests that individuals are sensitive to refugee immigration (or its accompanying political discourse) and departs from research showing the relative stability of immigration attitudes in the United States and Europe (Kustov et al. 2021; Mader and Schoen 2019).

Robustness and Generalizability

A potential concern with the results presented above pertains to the single-item measures of cosmopolitanism and nationalism. To be sure, random measurement error on either variable would attenuate its effect on immigration attitudes, but we have little reason to think that this measurement error should be correlated with immigration flows so as to drive the interaction effects. Nevertheless, as an alternative measure of cosmopolitanism, we take the z-score of respondents’ “closeness to the world” relative to their average reported closeness to their town, region, country, and continent.

Using z-scores is not our preferred method for measuring cosmopolitanism since, by construction, the measure cannot adequately reveal the existence of “nationalist cosmpolitans” (Bayram 2017)—or respondents that possess equally strong national and cosmopolitan identities. As we emphasize, social identities are not zero-sum, and cosmopolitans are more likely to possess national identities and national pride than non-cosmopolitans. ¹⁶ Moreover, z-scores of zero could denote respondents that either maintain weak social identities or strong social identities across the board. For this reason, Table S2 in the appendix should be interpreted as capturing the relative importance of a global identity as opposed to its absolute importance (compare to Wright et al. 2012).

Still, the results for cosmopolitanism remain unchanged when using this alternative measure. As shown in Table S2, we again find a positive constituent term for Cosmopolitanism (p < 0.01) and a negative Cosmopolitanism × % Change in Refugee Population interaction term (p < 0.10), indicating that the positive relationship between cosmopolitanism and pro-immigration attitudes diminishes with increasing refugee inflows. When calculating the z-score of respondents’ closeness to country, we find a negative constituent term (p < 0.01) but a positive interaction term (p < 0.10). In other words, refugee inflows tend to attenuate the effects of “closeness to country.” Yet, these results should be interpreted with a grain of salt since the “closeness to country” variable behaves much differently than the national pride variable, in turn raising concerns about its validity as a measure of nationalism. In stark contrast to National Pride, feelings of closeness to country are associated with greater trust toward other nationalities and greater affinity toward neighbors of a different race or language. ¹⁷ This suggests the results in Table S2 may be picking up the patriotic component of national identity.

One might also worry that these interactions are driven by omitted country-level variables. It is certainly possible that the effects of cosmopolitanism and nationalism depend on other country-level variables that are correlated with refugee inflows. Indeed, we would expect such correlations based on evidence that destination-country characteristics influence migration flows (Lewer and Van den Berg 2008; Fitzgerald et al. 2014). To examine this possibility, we sequentially control for interactions between cosmopolitanism (and nationalism) and other country-level variables that are typically associated with increased migration flows. Most notably, migrants will tend to seek out more affluent countries with a higher standard of living and ample job opportunities. We therefore interact our measures of social identity with ln(GDP per capita) and national unemployment rates. In addition, liberal democracies are expected to receive more immigration due to the attractiveness of expansive political rights for migrants (Nikolova 2017) as well as the legal-constitutional pressures on democratic governments to accept unwanted immigration (Joppke 1998). We account for this by interacting our cosmopolitanism and nationalism measures with the quality of democracy in the destination country using V-Dem’s Polyarchy Index (Coppedge et al. 2020).

As shown in Table S3, the Cosmopolitanism × % Change in Refugee Population interaction term is negative and statistically significant (p < 0.05) when controlling for additional interactions between cosmopolitanism and these three country-level variables. We find similar results for the effects of nationalism as well. As shown in Table S4, the National Pride × % Change in Refugee Population interaction term is negative and significant or near-significant (with two-tailed p-values of 0.04, 0.09, and 0.28) when controlling for interactions with GDP per capita, unemployment, and democratic quality. Note also that the interactions between nationalism and these three country-level variables are themselves insignificant. These results are broadly consistent with the hypothesis that cosmopolitanism and nationalism are more closely moderated by levels of refugee immigration than by other country-level variables. ¹⁸

A final question concerns the generalizability of the results beyond the countries in our sample. While our primary focus is on the Latin American refugee crisis, we leverage the WVS data to estimate the model on a larger sample of 54 countries, including some major European migrant destinations. As shown in Table S5 of the appendix, our results are similar when using a larger cross-national sample. Like with Latin American results, cosmopolitanism tends to associate with more pro-immigration attitudes while nationalism is associated with more anti-immigration attitudes on average. Further, the interactions between these variables—cosmopolitanism and nationalism—and the percent change in the refugee population are both negative and statistically significant (p < 0.05). This corroborates the finding that increases in the refugee population attenuate the positive effect of cosmopolitanism while amplifying the negative effect of nationalism, and suggests that this pattern extends beyond Latin America.

Discussion

Why do some citizens support open immigration policies while others do not? And how, if at all, has the Latin American refugee crisis reshaped public attitudes toward immigration? The migration literature demonstrates that immigration attitudes are strongly contingent on non-material factors, such as cosmopolitanism and nationalism. Building on this, we proposed a theory of fairweather cosmopolitanism in which the effects of these social identities are themselves dependent on migration pressures. We then leveraged the varied impact of intraregional migration in Latin America to provide evidence that refugee inflows attenuate the positive effect of cosmopolitanism while accentuating the negative effect of nationalism. We find evidence that cosmopolitanism is more fickle than normally assumed: it is correlated with pro-immigration attitudes where immigration pressures are modest, but tends to not be significantly associated with immigration openness in countries experiencing large refugee shocks.

One limitation of this study is that we rely on cross-sectional (i.e., static) data to model a temporal process. This raises endogeneity concerns—for instance, that immigrants might prefer countries where individuals are, on average, more cosmopolitan. To some extent, we can mitigate these concerns: The main findings are robust when including interactions with other country-level variables thought to make countries more attractive to migrants, ¹⁹ as well as when controlling for country-specific means of nationalism and cosmopolitanism. ²⁰ These specifications increase our confidence that refugee flows do, in fact, condition the effects of cosmopolitanism and nationalism on immigration attitudes. Nevertheless, future work using experimental designs or time series data will be essential for better isolating these effects.

A further question pertains to the mechanisms by which increased immigration flows shape the relative impacts of cosmopolitanism and nationalism. We proposed three (non-mutually exclusive) mechanisms: real or perceived competition between immigrants and natives; increased social comparison; and elite discourse that primes exclusionary national identities. We provide some evidence for the elite discourse mechanism in the supplemental materials, showing that the effect of cosmopolitanism is more strongly moderated by immigration flows among individuals who consume more news media. ²¹ This suggests that mediated political discourse accompanying mass immigration is partly responsible for dampening the effect of cosmopolitanism. We also find that immigration flows more strongly moderate cosmopolitanism among urban respondents. ²² This is perhaps suggestive evidence that contact with immigrants—who tend to relocate to cities—depresses the effect of cosmopolitanism through social comparison or perceived resource competition. ²³ Nevertheless, we regard these tests as very tentative evidence indicating only the plausibility of these theoretical mechanisms. Further research with more fine-grained data on respondents’ information sources and contact with immigrants is required to fully disentangle the causal story.

Taking a step back, these findings contribute to a growing research program studying immigration attitudes in the Global South—the region that hosts the most refugee migration (United Nations High Commissioner for Refugees 2023) and continues to receive large flows of transit migration. Perhaps due to more muted linguistic and religious differences, some scholarship has suggested a greater role for economic self-interest (Harris et al. 2018; Lawrence 2011) and fiscal burden concerns (Meseguer and Kemmerling 2018) in driving immigration preferences within developing countries. Yet, cultural concerns, nationalism, and other non-material explanations still carry weight outside the Global North (e.g., Acevedo and Meseguer 2022; Alrababa’h et al. 2021; Cogley et al. 2019). ²⁴ Our findings suggest that this is partly due to natives juggling competing social identities that delineate different in- and outgroups, depending on the context. Furthermore, national identities themselves are not solely based on ethnic or religious dimensions, but also historical legacies and national origin (Acevedo and Meseguer 2022). Scholars should continue giving greater attention to contextual and historical factors that underlie immigrant hostility in the Global South.

This article is also one of the first to focus on the public response to the recent wave of forced migration in Latin America. Latin America has traditionally stood out from other regions due to its more generous asylum policies (Hammoud-Gallego 2022; Hammoud-Gallego and Freier 2023). Despite this, substantial gaps persist between the rhetoric of governments and policy implementation (Acosta and Freier 2015). This has been especially true with the onset of mass Venezuelan emigration and growing South-South migration (Aron Said and Castillo Jara 2022; Brumat and Vera Espinoza 2024; Zapata and Tapia Wenderoth 2022). Our analysis offers one explanation for what Acosta and Freier (2015) label as “schizophrenic” public opinion driving policy responses in the region—inclusive in theory but rejecting unwanted immigration in practice (Acosta and Freier 2015, 689). A silver lining here is that the positive effect of cosmopolitan identity persists up until very high levels of immigration. But even so, an unfortunate implication of our results is that cosmopolitanism is least influential for public opinion precisely when accommodative immigration policies are most important. Distinguishing between ordinary and extraordinary times is therefore essential for a complete understanding of immigration attitudes.