Introduction
According to the bulk of social scientific literature, negative attitudes towards immigrants and immigration are strongly associated with having low levels of income and education and a right-wing political affiliation, as well as with being White, old, and male. There is also a consensus that those who live in low-density places are negatively oriented towards immigrants and immigration, but only because such places attract or retain people with the aforementioned characteristics with negative attitudes, or because such places, being ethnically and racially homogeneous, afford little or no interaction with diverse populations. This assumption is marred by two problems. First, when scholars and commentators contrast higher- and lower-density areas and the attitudes of their inhabitants, they ignore the autonomous effects of population density—a powerful morphological social fact that was central for Durkheim and Simmel and which has been largely ignored since. Second, both in the scholarship and the public discourse on attitudes towards immigrants and immigration, we observe a conflation of documented and undocumented immigration. Yet attitudes to these types of immigration might well be divergent. This paper tries to rectify these two problems. We will focus on contemporary American attitudes towards undocumented immigration and immigrants—with some comparison, at the end, of these attitudes with those towards documented immigration and immigrants. While we will also consider the autonomous effects of those variables that are considered to have strong effects (such as income, education, race, political affiliation, and interaction with undocumented immigrants), our main focus will be the autonomous and hitherto ignored causal role of population density. This effort will also allow us to empirically capture cosmopolitanism, often seen idealistically as the celebration of diversity.
Scholarship on Attitudes Towards Immigrants and Immigration
According to a recent poll, Americans consider immigration to be the number-one issue facing their country [Pew Research Center 2018]. Immigration has also clearly been at the forefront of US politics at least since Donald Trump’s presidential victory in 2016 in the wake of a largely nationalist campaign marked by the slogan “Make America Great Again.” So, not surprisingly, since then, scores of surveys have been conducted that have consistently shown party affiliation to be a profound and enduring predictor of how Americans feel about immigrants and immigration policies [Schain Reference Schain2018]. Less obviously, sociologists and political scientists—but also media pundits—point to individual attributes such as income, education, age, and gender in accounting for attitudes towards foreign-born populations in the United States as well as in Europe. The general consensus is that immigrants are seen as an economic threat by lower-income groups [Donato and Massey Reference Donato and Massey2016; Ebert and Okamato Reference Ebert and Okamato2015; Chiricos et al. Reference Chiricos, Stupi, Stults and Gertz2014; Filindra and Merkowitz Reference Filindra and Merkowitz2013; Manewska and Achterberg Reference Manevska and Achterberg2013; Hopkins Reference Hopkins Daniel2010; Mughan and Paxton Reference Mughan and Paxton2006; Quillian Reference Quillian1995]. Further, it is argued that the less educated are more strongly opposed to immigration than the highly educated [see, for example, Cavaille and Marshall Reference Cavaille and Marshall2019; Hainmueller and Hopkins Reference Hainmuller and Hopkins2014; Lancee and Sarrasin Reference Lancee and Sarrasin2015; Van Dalen and Henkens Reference Dalen Henrik and Henkens2005]. For many scholars, lower levels of educational attainment lead to less occupational success, which makes those affected vulnerable to economic competition from immigrants and thereby fosters negative attitudes towards them among those groups [Mayda Reference Mayda2006; Scheve and Slaughter Reference Scheve and Slaughter2001]. Others have pointed out that education can in itself, with or without economic advancement, inculcate cosmopolitan attitudes, and lack of education can do the opposite [Ford and Goodwin Reference Ford and Goodwin2014; Pecoraro and Ruedin Reference Pecoraro and Ruedin2016].
Race is another often-cited factor. Some researchers have claimed that Whites, as the dominant majority racial group in the United States, seek to retain their hegemony and are simply more anti-immigrant than others [Massey Reference Massey1995]. Others blame the recent status decline of the native white majority for their negative attitudes towards immigrants’ [Alba and Foner Reference Alba and Foner2015]. Hochschild [Reference Hochschild2017] argues that for many white Americans, immigrants, among other groups, are seen as ‘cutting in line’ to achieve the American dream. Many scholars have similarly claimed that the white working classes feel economically excluded and that anti-immigrant xenophobia is at the crux of their populist anger towards the “corrupt elite” [Algan et al. Reference Algan, Guriev, Papaioannou and Passari2017; Oliver and Rahn Reference Oliver and Rahn2016; Eatwell and Goodwin Reference Eatwell and Goodwin2016; Golder Reference Golder2016; Norris and Inglehart Reference Norris and Inglehart2019; Mudde Reference Mudde2007; Noury and Roland Reference Noury and Roland2020]. There are also those who point to hostilities towards immigrants within the African American community as stemming from competition between disadvantaged groups [Johnson, Farrell, and Guinn Reference Johnson, Farrell and Guinn1997; Rodriguez Reference Rodriguez1999; Sanchez Reference Sanchez1999]. The general thinking here is that economic self-interest, perception of cultural threat, and political ideology make lower-status whites, as well as others with less education and income, hostile to immigration. The larger the size of immigrant populations, and the more those populations grow, the more hostile these attitudes become [Kaufmann and Goodwin Reference Kaufmann and Goodwin2018]. Some studies, however, have not found race to be a significant determinant in this regard [e.g. Chandler and Tsai Reference Chandler and Tsai2001].
Place of residence has also recently received attention in these debates. Many have signalled a strong correlation between population density and positive attitudes [Lichter and Ziliak Reference Lichter and Zilliak2017]. Much ink has been spilt regarding Trump’s success in rural America and his lack of support in cities.Footnote 1 According to a recent Pew Research Center analysis of the American Community Survey (2012–2016), 57% of rural residents believe that growth in the number of outsiders “threatens traditional American customs and values,” while only 35% of urban residents feel this way. Likewise, in European countries such as the UK and France, there seems to be a divide between the parochial anti-immigration provinces and cosmopolitan pro-immigration cities [Geddes and Scholten Reference Geddes and Scholten2016; Alba and Foner Reference Alba and Foner2015; Guilluy Reference Guilluy2019]. Van Dalen and Henkens, for instance, find that in the Netherlands, anti-immigration sentiments are stronger among those respondents living in localities with low population densities [2005].
Population Density and Cosmopolitanism
There are two limitations to the existing research: (i) its approach to population density; (ii) its conflation of documented and undocumented immigrants and immigration.
First, despite their stress on a “geography of polarization” [Alba and Foner Reference Alba and Foner2015], for immigration scholars, place in itself—and in particular the population density of where one lives—does not usually play an autonomous causal role in shaping attitudes towards immigration. Population density is treated in three ways. The first way is to treat it as a background control in descriptive analysis.Footnote 2 The second way is to think of it as an epiphenomenon, without causal power. For many, high-density areas simply retain or attract people with the individual attributes associated with pro-immigrant attitudes (such as high income and education, liberal political positions, etc.), while low-density areas do the opposite. So density per se does not matter, and the relationship between density and pro-immigrant attitudes is a simple correlation. Maxwell, among others, has argued that large European cities have more positive attitudes to immigration than rural areas because they have larger percentages of residents who are highly educated or professionals. Here, the key mechanism is one of similar people, all with pro-immigrant attitudes, selecting themselves into denser areas [Maxwell Reference Maxwell2019]. Similarly, it is often argued that low-density places are economically stagnant locations populated by resentful parochial types, and high-density places are vibrant ones, to which successful cosmopolitans flock.
The third way is to think of population density as something that facilitates interaction with immigrants, which then leads to pro-immigrant attitudes [e.g. Pettigrew and Tropp Reference Pettigrew and Tropp2006]. The idea is that dense settings are both more ethnically diverse and more generative of interaction, and that interaction with immigrants results in more cosmopolitan attitudes. According to Wessendorf [Reference Wessendorf2014], urbanites, unable to avoid contact with those who are different, end up becoming more civil-minded. In this line of thinking, density is seen as a proxy for intergroup interaction, which seems to confirm studies stating that contact diminishes intergroup prejudice [Pettigrew and Tropp Reference Pettigrew and Tropp2008; Allport Reference Allport1954; van Heerden and Ruedin Reference Heerden and Ruedin2019; Kaufmann and Harris Reference Kaufmann and Gareth2015]. Others argue that greater ethnic diversity reduces social solidarity [Putnam Reference Putnam2007; for a nuanced perspective, see Laurence, Schmid, and Hewstone Reference Laurence, Schmid and Hewstone2018]. And some evidence indicates that exposure to different groups while competing with them over scarce resources can generate negative attitudes [Dancygier Reference Dancygier2010; Quillian Reference Quillian1995].
Changing population demographics [Hopkins Reference Hopkins Daniel2010; Kaufmann Reference Kaufmann2017] in dense and segregated environments [Enos Reference Enos2017] can exacerbate aversion to immigrants. At the same time, several scholars have proposed that more positivity emerges when natives and immigrants interact as equals in intimate relationships, workplace environments, or on the street [McLaren Reference McLaren Lauren2003; Pettigrew and Tropp Reference Pettigrew and Tropp2006; Stolle et al. Reference Stoll, Petermann, Schmid, Schönwälder, Hewstone, Vertovec, Schmitt and Heywood2013]. Yet there is no consensus on whether diversity in itself yields more favourable attitudes towards immigrants. Van Dalen and Henkens [Reference Dalen Henrik and Henkens2005] have found that in the Netherlands, the ethnic concentration of the neighbourhood in which people reside does not affect their views on foreigners or immigration policy [2005]. They also argue that natives perceive foreigners more favourably when they meet them at work and school, whereas contact while going out worsens attitudes [Laurence, Schmid, and Hewstone Reference Laurence, Schmid and Hewstone2018].
Useful as this debate is, however, it understands population density as having a causal force only insofar as it facilitates interaction with immigrants. But this is an impoverished understanding. Population density has significant effects on lived experience regardless of the characteristics of those who live in densely populated places, such as their income, education, political and cultural beliefs, race, ethnicity, age, and gender. Population density is also distinct from size and social heterogeneity, and it does not necessarily lead to increased interaction with heterogeneous populations. In fact, living one’s daily life in close quarters to others might increase feelings of aversion. In any case, such interaction does not necessarily lead to the development of more favourable attitudes to these populations. A central characteristic of public life in denser settings—one that is experienced profoundly in everyday life while ignored by much of social science—is that people share spaces and are visually available to one another at a higher rate. This effect of density is different from and independent of social heterogeneity and of interaction with others since it is unusual for us to interact with the people that we see on the street and with whom we share the pavement. It is also independent of the demographic, socio-economic, and political characteristics of the people who live their lives in proximity to each other. There is reason to think that this characteristic of public life in more densely populated places might have significant effects on our attitudes towards others—especially those markedly different from us, including immigrants.
By focusing on population density, we build on a morphological perspective that was at the heart of classical social theory, a perspective that has been largely abandoned in contemporary sociology. In The Division of Labor in Society, Durkheim poignantly called attention to the important role played by population density (which he called “material density,” differentiating it from “moral or dynamic density,” which has to do with frequency of social interaction) in social life. He argued that material density is the main determinant of the degree of division of labour [Durkheim (1893) Reference Durkheim1984: 201–208]. The gist of Simmel’s famous article on the mental life of the metropolis was similar in its focus on the sociological and psychological effects of high density that we find in urban settings, and especially the blasé outlook of the pedestrian [Reference Simmel1950: 413–416]. While Goffman is well known for his analyses of the interactional order, his most important concept may well be civil inattention, the mutual ignorance that we exercise vis-à-vis one another in public spaces [Reference Goffman1972]. In the same vein, Louis Wirth argued that personality characteristics associated with urban life, such as cosmopolitanism—that is, an open attitude towards non-natives—were a consequence of the three characteristics that defined a city: (a) increased population size, (b) density of settlement, and (c) heterogeneity of inhabitants and group life [Reference Wirth1938].
Wirth famously conceptualized the city as “a relatively large, dense, and permanent settlement of socially heterogeneous individuals” [Ibid.: 8]. Yet, subsequent reflection and research neglected density and focused on population size and heterogeneity. Many scholars have erroneously conflated density with population size, even though the two are different things that are not even necessarily related.Footnote 3 This conflation is probably in part due to the fact that American cities tend to be sprawling, which has led to an underemphasis on density and an overemphasis on population size [Wilkinson Reference Wilkinson2019]. Insofar as attitudes towards immigrants are concerned, it has been largely assumed by urban sociologists that population size and density—again, the latter is usually reduced to the former—are only generative of favourable attitudes towards others to the extent that they create diversity and interaction between different groups.Footnote 4 A similar position can also be found among social theorists such as Richard Sennett, who argue that the rich public life of cities is due to their social heterogeneity and to the possibilities for interaction that they engender [Reference Sennett1977]. This perspective not only reduces density to social interaction but also ignores a significant possibility: that interaction between different groups, especially between natives and immigrants, can intensify their negative attitudes towards each other.
An equally important problem with the existing scholarship on attitudes towards immigrants and immigration is that there has been very little effort within it to differentiate between attitudes towards documented and undocumented types. Instead, it largely focuses on public opinion towards immigration/immigrants as a whole and rarely explores potential distinctions between attitudes towards legal and unauthorized forms of immigration [Buckler 2008]. This is reflected in the choice of instruments used in large surveys. The most frequently used question in US studies asks respondents’ attitudes towards increasing/decreasing ‘the number of immigrants to America nowadays’ or ‘the number of immigrants from foreign countries permitted to come to the United States’ [e.g., Newman Reference Newman2013]. European studies frequently use even-less-defined measures, asking whether ‘immigration’ should be increased/reduced with little indication of the legal/illegal status of the immigrants [e.g., Kaufmann Reference Kaufmann2017], or posing questions about ‘people who come to live here from other countries’ [e.g., Laurence and Bentley Reference Laurence and Bentley2018]. Others combine measures of attitudes towards legal and illegal immigration into a single score of conservative immigration attitudes [Knoll Reference Knoll Benjamin2013; Berg and Morley Reference Berg Justin and Morley2014].
This conflation is reflected in the most frequently used survey item, which asks respondents what the government should do with respect to the current influx of immigration: increase, decrease, or maintain it.Footnote 5 Yet the two types of immigration need to be differentiated. First, attitudes to documented and undocumented immigrants and immigration might well diverge. The United States is a country of immigrants. According to a report by the Pew Research Center (2018), an overwhelming 76% of Americans considered immigration “a good thing for the US.”Footnote 6 So it is possible for someone to support documented immigration and even clamour for more of it while being critical of undocumented immigration for one of a number of reasons. These might include: an unyielding legalistic mentality; the lower socio-economic status of undocumented immigrants; the stigmas and negative stereotypes attached to this population; or the differential threat that the respondent attributes to undocumented immigrants as opposed to documented ones. Second, a significant portion of immigrants in the United States are undocumented (23% according to Pew, but twice as much according to some economists), and disputes surrounding immigration have mostly involved the status of undocumented immigrants from south of the border.Footnote 7 Political polarization between the parties mostly plays out around attitudes on the issue of unauthorized entry.Footnote 8
We believe that the causes of attitudes towards undocumented immigrants and immigration in the United States are worth studying in themselves, especially given the political zeitgeist [Mudde Reference Mudde2007]. But we also think that favourable attitudes towards undocumented immigrants and immigration are very strong empirical indicators of cosmopolitanism. We think that cosmopolitanism is always a question of degree, and not simply a matter of general openness to all “others.” More specifically, we think that a positive attitude towards undocumented immigration and immigrants bespeaks higher cosmopolitanism than similar attitudes towards documented immigrants and immigration, as the former type risks contradicting the normative claims of the nation-state. This is in effect in the very etymology of the term “cosmopolitanism”; the word cosmopolis, in ancient Greek, refers to a “world state.”Footnote 9 The cosmopolitan has traditionally been defined as a “citizen of the world.” The term has been applied—sometimes positively, other times negatively— to individuals from particular ethnic or religious groups—such as Jews—who are spread across the world and thereby considered by some not to be full members of the polity that controls the territory they live in. But to the extent that cosmopolitanism is touted as a worldview, it tends to refer to the idea that all human beings should belong to the same political community, so that legal barriers between nation-states are illegitimate or at least problematic. Most people are not that cosmopolitan: pace John Lennon,Footnote 10 such barriers are legitimate for most. But it is possible to differentiate people by how close they are to this ideal by considering their attitudes to those who cross national barriers in violation of the laws of the nation-state in question. Using this empirical indicator, we will also be able to study cosmopolitanism positively, as a moral stance that can potentially conflict with the normative framework of the nation-state. This will also allow us to see whether this moral position has any material causes or conditions.
Based on an analysis of nationally representative US samples from two distinct surveys, this paper considers the causes of American attitudes towards undocumented immigrants and immigration. At the end of the paper, we compare these attitudes with those towards documented immigrants and immigration. We hypothesize that higher population density will, net of demographic, socio-economic, and political characteristics, generate more favourable attitudes towards undocumented immigrants and immigration, and, thereby, higher cosmopolitanism. We claim in general (i) that density is a central morphological variable in the making of social values and norms, and (ii) that it generates its effects at least in part through visual exposure to and space-sharing with generalized others, as opposed to through social interaction. In particular, we argue that higher population density makes city dwellers’ attitudes more favourable towards undocumented immigrants and immigration than the attitudes of those in more rural areas,Footnote 11 and that this is true independent of the respondents’ status in terms of the factors that most people identify cities with, such as more frequent interaction with different groups and a higher number of people with higher education, income, and liberal viewpoints.
Data and Methods
For this study, we draw on data from two sources provided by the Pew Research Center. The first is the 2018 Global Attitudes & Trends Survey (GATS), a cross-national project covering 27 countries including the United States.Footnote 12 We use the American subsample (N = 1,500) containing a wealth of information on natives’ attitudes to, among other topics, immigrants and immigration. The other source is the June 2018 Political Survey (N = 2,002), which was designed to gauge American public opinion on a host of domestic and global issues, including pro- and anti-immigrant sentiments and preferences.Footnote 13 Both surveys contain probability samples of Americans (ages 18 and over) in 50 states plus the District of Columbia. Data were collected via phone interviews in English and Spanish for the Political Survey. The combined sample (consisting of 500 landline and 1,500 mobile-phone interviews) is weighted using an iterative technique that matches gender, age, education, race, and region to parameters from the 2016 Census Bureau’s American Community Survey one-year estimates, and population density to parameters from the Decennial Census. For the American subsample of GATS, telephone interviews in English and Spanish were similarly conducted. The sample design consisted of the list-assisted Random Digit Dial (RDD) probability sample of landline households (25% of sample) and list-assisted RDD frame of mobile numbers (75% of sample). Individual weights are provided based on gender, age, education, race, Hispanic origin and nativity, region, population density, phone use, and probability of selection. More methodological and other technical details are available from the Center’s website.Footnote 14
In testing our central hypothesis, we use three outcome variables: Deportation, Negative Attitude, and No Policy Support. The first is taken from the GATS; the other two come from the Political Survey. Deportation, a dichotomous variable, is based on the interviewees’ survey responses concerning whether undocumented immigrants in the United States should be deported. Negative Attitude is operationalized using an item measuring Americans’ general feelings towards undocumented immigrants; and No Policy Support is measured using answers (yes/no) to a question asking respondents whether they support the government’s policy of granting permanent status to the children of undocumented immigrant parents in the United States. Our main predictor is population density, measured at the county level. Since we want to assess the independent effect of this variable, we control for and also consider the effects of the individual attributes—age, gender, race, education, household income, religion, political orientation, and party affiliation—that are given prominence by scholars and commentators in their explanations of attitudes to immigrants and immigration. This control is also essential because, in the United States, people in more densely populated areas tend to be better educated, more affluent, and more politically liberal. To measure population density, we take advantage of a crucial piece of information available in both surveys: density quintiles based on the three-digit 2010 FIPS (Federal Information Processing Standards) codes, which uniquely identify counties and county equivalents in the United States, from the respondents’ self-reported zip codes.Footnote 15 We created five dummy categories to indicate density at the county level (e.g., lowest density = “1st quintile”, highest density = “5th quintile”). Prior research heavily relies on the relatively crude dichotomy between urban and rural respondents or unreliable subjective indicators to measure geographic density [Frasure-Yokley and Wilcox-Archuleta Reference Frasure-Yokley and Wilcox-Archuleta2019]. Ours is an improved alternative in terms of its accuracy and validity. We measure density at the county level. This is because county—as opposed to neighbourhood, district, or state—corresponds best to the full lived experience of individuals as they reside, work, and move around in their local area as part of their everyday routines.
For the two data sets, we adjust for a common set of confounders or compositional effects [see Maxwell Reference Maxwell2019]: age, gender, race, education, household income, religion, political orientation, and party affiliation. We also ran sensitivity analyses by incorporating additional factors that correlate with the outcome measures. For the GATS subsample, we include Community Type, which taps respondents’ subjective assessments of their residential communities in terms of geographic location or size (“a big city,” “suburbs,” “country village,” etc.). Inclusion of this variable provides a more stringent test of our hypothesis concerning the causal role of objective population density, controlling for a subjective understanding of one’s own place of residence. Using the Political Survey data, we provide further robustness checks by taking into account a critical factor shown to influence anti-immigrant views, a variable we call Outgroup Contact. Since we want to consider whether population density has an effect on attitudes towards undocumented immigrants and immigration, we need to control not only for natives’ individual attributes, but also for their interaction with undocumented immigrants. Typically, in the literature, intergroup interaction is not directly gauged but inferred from the size of migrant stock, leading to conflicting findings [Laurence et al. Reference Laurence, Schmid, R. Rae and Hewstone2019]. We overcome this limitation by utilizing answers to two questions. The first is a question that is rarely asked in surveys: “How often do you personally come into contact with immigrants who speak little or no English?” Including this variable allows us to control for the interaction between natives and undocumented immigrants, since the latter are much less likely than documented immigrants to be proficient in English (34% vs. 57% in 2016).Footnote 16
Lastly, to further adjust for the possibility of self-selection, i.e., that more cosmopolitan-oriented people tend to inhabit denser places, we add a variable (Moral Objection) from the Political Survey sample that controls for personal convictions about whether it is morally wrong to assist “illegal immigrants.” To the extent that more liberal and open-minded individuals gravitate towards living in more urban places (or tend to flee from more rural alternatives), those places characterized by higher population density can be expected to contain a greater number of residents who would not morally object to accepting such outsiders. Controlling for this variable, a proxy for value orientation that accounts for the growing urban–rural cultural chasm in the United States [Cramer Reference Cramer2016]Footnote 17 further allows for a more conservative testing of our main argument.
Data from the 2018 versions of GATS and the Political Survey are hierarchically structured, i.e., individuals are nested in higher-level units (50 states plus Washington, D.C.), creating the well-known issue of data clustering, which can lead to biased parameter estimates [Raudenbush and Bryk Reference Raudenbush and Bryk2002]. To minimize this methodological concern, we ran a series of multilevel logit (for Deportation and No Policy Support) and ordinal logistic (for Negative Attitude) models with random intercepts. In addition to the individual-level covariates described above, at the contextual level of analysis, we included State Density, an aggregate measure taken from the US Census Bureau website. As a sensitivity analysis, in models not shown, we also used Urban Population in lieu of the state-density variable. Main results were consistent (available on request). In addition, we estimated multilevel models by including additional contextual measures such as state economic index, ethnic heterogeneity, and immigrant size. For the sake of simplicity, we show and discuss our findings without including state-level covariates since these do not alter the main results in any significant way.
However, to demonstrate the robustness of our main findings regarding the impact of population density on anti-immigrant attitudes, we estimated two additional sets of models (as shown in Tables 7 and 8) and discuss the findings below. The statistical results shown in all the tables are based on unit-specific models adjusting for state-level random effects. Exact wordings for all the survey items and coding procedures for variable construction are shown in Table 1. Table 2 contains descriptive statistics for the two nationally representative US samples from the June Political Survey 2018 and the GATS 2018. We group-mean centred all non-dichotomous covariates and applied the individual weights provided in the data sets to adjust for the probability of selection and non-response bias. Data analyses were performed using the latest version of the statistical package HLM 8 [Raudenbush et al. Reference Raudenbush, Bryk, Cheong, Congdon and Toit2011].
Summary of variable definitions and coding criteria
Data source: 2018 Pew Research Center’s June Political Survey and Global Attitudes & Trends Survey (US subsample)
Descriptive statistics for variables in the two datasets
Results
In the GATS 2018, 51% of the sample said that immigrants who were in the country illegally should be deported. For the residents of those countries with the least dense populations (1st quintile county density), the figure is 7 percentage points higher (58%), whereas for the residents of the most densely populated counties (5th quintile), it is 15 percentage points lower (36%). The percentage difference across density quintiles is thus considerable: 22%. In the June Political Survey, 22% did not support a policy of granting legal status to the children of undocumented immigrant parents. Answers are also strongly bifurcated: among those who live in the least densely populated counties, the average is almost doubled (42%), while it is only around 17% for those living in the most densely populated counties. Here also, the percentage difference attributable to population density is considerable: 25%. Lastly, the mean value for the last outcome measure was 2.8 (approximating the answer choice “somewhat unsympathetic”), with a half-point difference between the most and the least densely populated counties.
Our hypothesis states that population density, independent of both individual attributes and of intergroup interaction, is significantly related to attitudes towards undocumented immigrants and immigration. We argue that population density is only a powerful predictor of positive attitudes towards undocumented immigrants and immigration—hence of cosmopolitanism—insofar as it affects everyday experience, causing us to live our public lives in close physical proximity to others. This is why we measure it at the county level—the level at which we spend most of our time. According to our hypothesis, state-level density should not matter as it does not have an effect on how we live our everyday lives. If we live in a dense city, the fact that the state that the city happens to be in is full of mountains, and so the state itself is not dense, is irrelevant to our everyday experience. To verify that this is so – that density matters at the county level and not at the state level – as a preliminary test before investigating our main hypothesis, we estimated a set of null or unconditional models (not shown) to see whether living in a more densely populated state (not county) is associated with greater receptivity towards undocumented immigrants.
To that end, we regressed the three dependent variables separately on State Density (converted to z-scores), the results of which appear in Table 3. As is customary, we first check the degree of data dependence with respect to Deportation (τ = 0.04, χ2 =74.81, p < 0.05), No Policy Support (τ = 0.05, χ2 = 87.42, p = 0.001), and Negative Attitude (τ = 0.06, χ2 = 106.49, p < 0.001).Footnote 18 Having confirmed significant clustering for all three, we proceed with multilevel analysis. According to Model 1 in Table 3, State Density is negatively, albeit marginally (p < 0.1), related to Deportation. This relationship, however, falls below the conventional significance level when we enter the set of dummies for population density measured at the county level (ref.: 5th quintile, or “the densest”), which is shown in Model 2. That the association between state density and anti-immigrant sentiment becomes fully mediated implies the causal influence of population density at work, parameter estimates for which are all highly significant (varying from p < 0.01 to p < 0.001). Our argument finds additional support when considering the other two outcomes: No Policy Support and Negative Attitude. In both cases, as shown in Models 4 and 6 respectively, we see that State Density again loses its statistical significance when the dummy indicators are incorporated (ref.: 1st quintile, or “the least dense”). Replacing the state-density measure with the percentage of urban population replicates the results. From this, we can infer that, on average, more densely populated states tend to be slightly more “anti-immigrant” (although this is not statistically significant at the conventional level). However, a closer inspection reveals within-state variations concerning how Americans feel about unauthorized immigrants, variations that stem from population density operationalized at the county level. The contrast between state and county levels makes our findings on the relationship between population density at the county level and positive attitudes towards undocumented workers all the more striking.
Results from estimating the relationships between state and county density measures and anti-immigrant attitudes
Note: p < 0.001 *** p < 0.01 ** p < 0.05 * p <0.1 #
Do the above effects concerning the impact of population density (at the county level) hold after controlling for background variables? As has been observed, the composition of individual residents varies according to density of place: people in denser areas tend to be better educated, of higher income, and more politically liberal, etc. To account for potential confounding, therefore, control variables are introduced. Tables 4 through 6 contain the findings corresponding to the three outcome measures (Deportation, No Policy Support, Negative Attitude), respectively. According to Model 1 in Table 4, among the eight socio-economic and demographic controls, only four are significantly positively related to Deportation: gender, race, ideology, and party identification. Being a Protestant also raises the odds, but only marginally. Specifically, on average, white Republican males who describe themselves as politically conservative are more likely to say that America’s unauthorized immigrants should be deported. Odds are about 46% greater (p < 0.05) for men than for women. They are 69% greater for whites than for non-whites. What is more, the odds of supporting deportation are about 50% higher for those who are more politically conservative (p < 0.001), and they are almost three times higher for Republicans vis-à-vis Democrats and Independents (OR = 2.9, p < 0.001). Contrary to some previous findings, and contrary to common wisdom, low income and education do not predict negative attitudes towards undocumented immigrants and immigration.
Multilevel estimates from regressing Deportation on population density
Note: p < 0.001 *** p < 0.01 ** p < 0.05 * p < 0.1 #
To examine factors that correlate with No Policy Support, we move onto Model 1 in Table 5. Here, we find that only gender, being male (OR = 1.6, p < 0.01); ideology, being more conservative (OR = 1.5, p < 0.001); and party identification, being Republican (OR = 3.1, p < 0.001) matter in the direction and magnitude comparable to results from the previous table. With respect to Negative Attitude, as illustrated in Model 1 in Table 6, being male (OR = 1.6, p < 0.001), white (OR = 1.7, p < 0.001), more conservative (OR = 1.4, p < 0.001), and Republican (OR = 2.9, p < 0.001) once again emerge as significant predictors. In addition, we find that older Americans are more likely to hold a negative attitude towards immigrants (p < 0.001). Despite slight variations, we thus witness two patterns across the alternative outcome measures: first, and surprisingly, education and income are not related to Americans’ views on undocumented immigrants. This is quite a remarkable finding, given the consensus both in scholarly and journalistic circles which takes for granted the importance of these variables in this regard. Second, political affiliation and self-evaluation of political ideology are robust indicators of negative attitudes. But since undocumented immigration is so central to the general political debate in the United States, this finding is hardly very surprising—and is even somewhat tautological, as there are many who are Republican or Democrat, conservative or liberal primarily because of their attitudes towards undocumented immigrants and immigration, and because having a certain attitude towards these is becoming more and more central to being a Democrat or Republican, liberal or conservative.
Multilevel estimates from regressing No Policy Support on population density
Note: p < 0.001 *** p < 0.01 ** p < 0.05 * p < 0.1 #
Multilevel estimates from regressing Negative Attitude on population density measures and controls
Note: p < 0.001 *** p < 0.01 ** p < 0.05 *
We now examine the main evidence with respect to our hypothesis testing pertaining to the (county) population density measures. For Deportation, as shown in Model 2 in Table 4, all four dummy variables are significant. The positive sign for them indicates that, ceteris paribus, living in less dense counties vis-à-vis the reference (most dense) category is associated with significantly higher odds of wanting to remove unwanted foreigners. That is, population density implies cosmopolitanism or openness towards outsiders. And the effect size is substantial. The odds of living in the least dense county, for example, raise the odds of support for deportation by more than two and a half times (OR = 2.6, p < 0.001). This, most strikingly, almost rivals the effect of being Republican vis-à-vis Democrat or Independent (OR = 2.9, p < 0.001).
Next, concerning No Policy Support, a similar trend emerges in Model 2 in Table 5. Here, the signs for density dummies are negative since the reference category consists of the least, not most, dense counties (1st quintile). We do not find a significant difference between 2nd quintile and the reference category (countryside): towards the lower end of population density, people’s immigration-related attitudes are not sharply differentiated along the geography of their residence. When we compare those in the least dense areas with the rest (3rd through 5th quintiles), however, the difference is stark. Controlling for a host of variables (compositional factors), being embedded in a denser community is thus generally associated with a more cosmopolitan outlook on unauthorized immigrants. And the same can be said of our third and last variable, Negative Attitude, as shown in Table 6. Empirical support is relatively limited here, however: only residents in the most densely populated counties hold more positive attitudes (indicated by the negative sign for the 5th quintile) in comparison with their counterparts living in the least populated counties (p < 0.001). No other regional difference is found. What this suggests is that density has a stronger effect when it comes to policy on immigrants (the question of whether to deport them or deny them legal residency) than it does on general attitudes towards them. This makes intuitive sense. There is a strong and significant difference between urban and non-urban dwellers concerning the forced deportation of unauthorized immigrants, which is a politically charged issue in America, especially under the Trump administration administration. However, this striking difference subsides when we consider a relatively less sensitive issue, namely natives’ general predisposition towards undocumented immigrants.
For a stricter test of our thesis, we run sensitivity analyses. Model 3 in Table 4 incorporates Community Type, a variable based on the respondents’ subjective understanding of where they live (e.g., a big city or the countryside). This is important, since a person who thinks and believes that he or she is embedded in a “small town” may adopt correspondingly “rural” values and attitudes in general. Conversely, a self-recognized member of an “urban metropolis” may adopt those that are commensurate with cosmopolitan living. Now, it is possible that the effect of population density, as reported above, is actually mediated by the perception of one’s broader residential community and the associated values and attitudes. If so, including this variable would reduce the magnitude and the significance level of our categorical density indicators. We find that it does not. In fact, Community Type is not a significant predictor, while the parameter estimates for the density measures remain robust. In other words, population density lowers anti-immigrant sentiments, above and beyond subjective density.
Can the effect of density simply be a result of a self-selection? It is plausible, even likely, that people who are more cosmopolitan move into large urban areas. We address this concern by adding the variable Moral Objection into the analysis. As Model 3 in Tables 5 and 6 indicates, considering it wrong to assist unauthorized immigrants is highly significantly related to No Policy Support (OR = 5.3, p < 0.001) and Negative Attitude (OR = 4.2, p < 0.001), as indicated by the magnitudes of estimated odds ratios. Importantly, this variable controls for a person’s overall orientation on the issue of unauthorized immigration; it is also a proxy for their broader cultural environment, closely correlated with population density. Whether we approve of a policy intended to help the children of immigrants without legal status, or are favourably predisposed to them in general, is largely a function of how we morally evaluate the issue at hand. Controlling for this critical factor, we still find that living in a denser community significantly reduces the likelihood of Americans having anti-immigrant views and preferences as measured by these two outcome variables.
Finally, we add Outgroup Contact to the analysis to take into account the effect of the frequency of interaction with immigrants with no or limited English ability. The latter implies that the immigrants have either recently arrived or have not been fully incorporated, and, as we have seen, such immigrants are more likely to be undocumented. Though we cannot be conclusive, given its wording, this survey item likely intends to capture negative experience. As such, we anticipate a positive sign for the estimated coefficient. In Model 4 of Table 5, we find exactly that: interaction with those who speak little or no English (outgroup contact) raises the odds of not supporting the policy of granting permanent status to undocumented immigrant children (p < 0.05). This is an important finding insofar as it contradicts the contact argument regarding interaction and tolerance. And since we find that people who live in denser areas are more likely to interact with such immigrants with limited English-speaking ability, and since this interaction has a negative effect on attitudes, it is all the more remarkable that higher-density places are still significantly more favourable to undocumented immigrants than lower-density places where such contact is less frequent. By comparison, as we see in the last model in Table 6, the negative relationship that outgroup contact has with policy support does not exist with negative attitudes towards unauthorized immigrants. In both cases, however, the main takeaway is clear: adjusting for moral objection and outgroup contact, along with various background controls, the negative relationship between population density and anti-immigrant sentiments remains statistically significant.
Indeed, as we have demonstrated across varying model specifications, the density–cosmopolitanism link remains robust. For two of our outcome variables (Negative Attitudes and No Policy Support), we estimate additional models by including the percentage of unauthorized immigrants at the state level (Unauthorized%). As a robustness check, we also measure country-level density as an interval variable (Residential Density). According to Model 1 in Table 7 (for Negative Attitudes) and in Table 8 (for No Policy Support), the coefficient for Unauthorized% is significant and negative, meaning states with a higher proportion of undocumented immigrants exhibit a more favourable attitude towards them. We can interpret this as meaning that Americans living in close proximity to a larger number of foreign-born members of the community are more open to them. Throughout Models 1–3 in both tables, parameter estimates for Residential Density are all negatively significant. That is, living in a denser county indicates more favourable, i.e., cosmopolitan, attitudes towards undocumented immigrants. These findings thus lead us to conclude that—based on a multilevel analysis of two different data sets and using three different outcome variables while controlling for a host of background variables at individual and contextual levels—the hypothesized link between population density and outgroup attitudes is not a statistical artefact but is supported by robust empirical evidence. In the absence of counterfactual examples, making a strong causal claim is problematic [Gangl Reference Gangl2010].Footnote 19 Testing the true causal nature of our hypothesis would require the acquisition of experimental data collected from subgroups exposed to various “treatments” of population density. Because of the impracticality of such a study design, we included appropriate confounders. Nevertheless, our results should be interpreted with caution, since we cannot fully adjust for selection bias (pro-immigrant individuals gravitating towards cosmopolitan, i.e., high-density, urban centres).
Robustness check for Negative Attitude by including percentage of unauthorized immigrants
Robustness check for No Policy Support by including percentage of unauthorized immigrants
Discussion and Conclusions
This paper has several surprising findings. We found that income and education do not affect attitudes towards undocumented immigrants and immigration. We also found that population density—in terms of the location in which a person spends most of their everyday public life—has a significantly positive effect over and above individual-level factors: education, income, race, political attitude, age, gender, and contact with undocumented immigrants. Population density, simply the number of people residing in each unit of area, generates cosmopolitanism, demonstrated by a favourable attitude towards those who are not only non-members of the political community but who are on national territory in violation of the laws set by the nation-state. What is more, population density does this per se and not because of characteristics that are associated with it, such as high education, high income, interaction with immigrants, liberalism, being non-White, being a woman, and being young.
Population density is obviously not the only factor generating tolerance towards non-native populations. In effect, the United States’s official ideology is that it is a country of immigrants. Furthermore, some of those who have negative views of undocumented immigrants and immigration are well disposed towards documented immigrants and immigration. To better capture the role of population density in cosmopolitanism, we also looked at the relationship between population density and attitudes towards documented immigration by considering responses to the following question (models available on request): “Should LEGAL immigration into the US be kept at its present level, increased, or decreased?” We found that density does not have an impact on the responses, with the inclusion of control variables, including interaction with immigrants who speak no or poor English. That is, in the fully specified model, density effects become mediated by individual-level covariates, highlighting the causal primacy of the composition of individual residents. The relationship between density and pro-immigrant sentiments thus holds only in the case of undocumented immigrants and immigration. While Americans show a similar level of favourability towards documented immigration, those who live in higher-density areas are more likely to ignore the legal status of immigrants, which shows higher cosmopolitanism.
How can we account for the relationship between density and cosmopolitanism? For most scholars and commentators, high density is conflated with large population size, and it is almost synonymous with high levels of interaction—usually with heterogeneous populations. But high density is not large size, and higher density does not necessarily lead to higher levels of interaction in the case of either homogeneous or heterogeneous populations. The most important characteristic of density—that it creates higher general (yet superficial) visibility and space-sharing, as opposed to higher levels of social interaction—is ignored by most. We argue that it is higher visibility and space-sharing that account for density’s cosmopolitan effects—and this is why we control for interaction as well as for the individual characteristics. At least in the United States, living in a denser place means that one is more likely to spend more time walking on the street or other public spaces, or in mass transit than driving. Public spaces in denser areas are more numerous and more crowded. The result is that one is visible to a larger number of others (in very dense places, to countless others), but only fleetingly so to any specific other. Further, higher density means that one is more likely to share the same spaces with a larger number of others, who are also more likely to move away from that space and be replaced by others. All of this will produce a general habituation to the presence of larger numbers of others around us. What is more, any person that we see will be more likely to be seen as a generalized other, as an object without qualities, rather than as an individualized or typified other. Our high-level visual availability to one another and our constant space-sharing in dense places—and the denser the place, the more this is so—are an unintentional by-product of the myriad activities that we undertake in public. As a result, the denser the location in which we live our everyday lives, the more heterogeneous and unchosen those whom we see and with whom we share the same spaces will be. Hence, habituation and generalization will be applied to more diverse populations. This is especially so at the county level in the United States. Even if an individual’s block or neighbourhood is racially or ethnically segregated—as is often the case, for instance, in New York City—a heavier reliance on mass transit will expose them more to those who are different from them but share the same spaces with them.
There is more human presence, and thus a more vibrant public sphere, in higher-density places, which leads their residents to take everyday physical proximity to others more for granted. But proximity does not mean more interaction. Public spaces are livelier in higher-density places, and in those spaces a great deal of our daily life is spent in the presence of strangers whom we can always technically interact with in some way. But public interactions that take place within public settings—when we shop, for example—tend to be superficial, ephemeral. Less interactions than transactions, such encounters largely take place between fungible beings. Further, most of the time we are in public in denser places, we do not interact with anyone; we go about our own business, carrying on a protective blasé attitude, civilly inattentive to others. Most proximity and space-sharing in dense areas does not involve interaction at all. In fact, the more one sees others and the more one shares space with them, the more one wants and needs to avoid interacting with them. Georg Simmel [Reference Simmel1950], Jane Jacobs [Reference Jacobs1961], and Erving Goffman [Reference Goffman1972] saw this both factual and normative principle at work a long time ago: in cities, due to their density, we learn to ignore all others and, as far as possible, reduce them to anonymous, generalized beings, as long as they do not seem to present a physical threat. This undiscriminating indifference of the city dweller to all other city dwellers produces a general, egalitarian tolerance of all, including undocumented immigrants, who are simply fleeting beings to be ignored just like anyone else. The denser the place, the more this is the case.
In lower-density areas, by contrast, one is more likely to spend more of one’s time not seeing others and simply driving. This is especially so in America. Public spaces are emptier; fewer people are visible on the street as well as other public spaces. Further, there is less space-sharing; there are often no pavements. At the same time, the less dense the place, the more noticeable those we see in public spaces become. When we are less likely to see people, we are more likely—and even expected—to say “hi” when we see someone. It is harder, physically and normatively, to ignore neighbours on the street in rural settings. As a result, anonymity is harder to achieve, and it is less acceptable. Seeing someone in less dense contexts is more of an event. What is more, the people that we see are not only more noticed; they are also subject to more intense scrutiny and judgement. They are not generalized others; they are more likely to be individualized, and this often involves reducing them to types. We also demand more of them; we don’t just want them to be beings that we can ignore. And the more different they are from us, the more exigent we become. We discriminate between them in all kinds of ways. The less dense the environment, the more likely that we will interact with those we meet on the street, and the less superficial that interaction. Moreover, those we see and interact with—usually in the places that we drive to—are more likely to be those we have chosen to see and interact with. Visual and physical exposure to others is less of an unintentional by-product of some other activity. As a result, in contradistinction to life in denser settings, we spend our time in homophilous circles, and we are less likely to be habituated to people different from ourselves and to think of them as generalized others. To repeat: public life in dense settings involves a phenomenology based on visibility and space-sharing, and not interaction. The denser the setting, the more this is so. In their research on shopping streets, Kasinitz and Zukin found very little interaction with strangers—but a lot of awareness of being in the presence of people who were different and whose presence the participants had not chosen to be in [Reference Kasinitz and Zukin2016]. Interaction can in fact undermine tolerance, as we saw with low-status immigrants in the United States. By contrast, the lived experience of higher density—general but superficial visibility and constant space-sharing—is an important cause of cosmopolitanism.
Cosmopolitanism is often defined in a normative and idealist way by scholars. For Nussbaum, it is an ideal of allegiance to humanity as a whole [Reference Nussbaum1994]; for Appiah, a support for tolerance that involves multiple alliances to various forms of community [Reference Appiah2007]. According to Beck, cosmopolitanism is an “outlook” defined by “[…] the internalized otherness of others, ability to see oneself from the viewpoint of those who are culturally other” [Reference Beck2006: 153]. Hannerz says that it is “an orientation, a willingness to engage with the Other” [1990: 239]. These definitions are too abstract. To make the concept more open to empirical inquiry requires us to see that cosmopolitanism, as a morality, stands in opposition to the idea of the nation-state with its borders, its strict privileging of its own citizens over others, and the absolute right that it gives its citizens to control who can come in and stay and who cannot. Few are pure cosmopolitans. Still, the denser the location in which a person lives, the more universalistically indifferent they are to the legal citizenship status of others, and the more cosmopolitan.Footnote 20
Another problem with the existing thinking on cosmopolitanism is that it is seen as an advanced moral position achieved through the active celebration of humanity and diversity and through interaction. By contrast with this rosy view, which lacks empirical support, we advance a less grandiose understanding of cosmopolitanism, an understanding that stresses the material conditions of the phenomenon. Cosmopolitanism is a general and unintended habituation caused in part by constant visual exposure and physical proximity to people in general—both of these things being the outcome of population density. And it is this lived experience that makes cities particularly cosmopolitan, and not the individual attributes of their inhabitants, nor the fact that cities allow for a higher rate of interaction with immigrant populations. Constant visual exposure and physical proximity create a general, and thus universal, indifference to all others, and it is this superficial yet egalitarian indifference which is the basis of cosmopolitanism.
Our research has limitations. First, we were unable to control for ethnic diversity at the county level. It is possible that density generates more diversity and hence more proximity to undocumented immigrant populations. Large European cities, for instance, tend to have larger immigrant populations than other areas [Garbaye Reference Garbaye2005]. Wirth argued that demographic diversity was a function of population size and density [1938], and Hall and Lee found the same with suburban diversity [Reference Hall and Lee2010], but without differentiating between size and density. Further research should try to find ways to control for ethnic diversity. Yet this limitation should not substantially affect our findings: since we do control for interaction with low-status immigrants (those who speak little English), if ethnic diversity is part of the reason why dense counties are more favourable towards undocumented immigrants and immigration (assuming that dense counties are indeed more diverse), this would not be because of more interaction between natives and undocumented immigrants but because of the space-sharing and visual exposure that density imposes on both of these groups.
Relatedly, if there are indeed more immigrants in denser counties and if being an immigrant is associated with more pro-immigration attitudes (which is quite possible), this could account for some of the difference between higher-density and lower-density counties. But we deal with this problem by controlling for race (being non-Hispanic White), albeit with lingering limitations concerning White immigrants of European descent in the sample. Also, the outgroup contact variable that we analysed exclusively involves interaction with immigrants who speak little or no English. There are exceptions, of course. Some undocumented immigrants speak English very well, and there are documented immigrants who speak little or no English. But, as we saw above, there is a strong relationship between English linguistic ability and legal status. Moreover, it is impossible to measure interaction directly: people cannot know whether the immigrant—or the assumed immigrant—that they are interacting with is documented or not. Finally, this interaction variable, whatever its limitations, is nevertheless an improvement over existing research, which usually does not control for intergroup contact. It offers us a more stringent test of our population-density argument. Without this control, one could object that density is simply about more frequent interaction, an alternative explanation that we empirically reject. In any case, a more comprehensive set of questions tapping different qualities of intergroup contact involving documented (those fluent in English) and undocumented (those who are not) immigrants would clearly help advance our understanding of the issues at hand.
Sociologists and political scientists as well as pundits have pointed to income and education as significant factors in varying American attitudes towards foreign-born populations residing in the country. By looking at the case of undocumented immigrants, something that is rarely done, we have found this generalization to be incorrect. We also found, at least with respect to one outcome variable (No Policy Support), that interaction with low-status or “newly arrived” immigrants negatively affects attitudes towards undocumented immigration. Further, we identified a critical yet neglected factor underlying attitudes towards undocumented immigrants and immigration: population density. This factor is a contextual one: it has to do with our lived experience of the proximity of others as we lead our daily lives. But, as we have seen, it has little to do with interaction. Many sociologists erroneously reduce everyday life in the public sphere to social interaction. Yet most of the time, when we are in the presence of others, when we are visible to them and sharing the same space with them, we are not interacting with them. Nonetheless, this is, as we have seen, far from inconsequential; it is in effect the basis—or at least one very important basis—of cosmopolitan attitudes, whereas interaction itself might undermine them. Superficial visibility of all to all as anonymous, fleeting beings that should and can be safely ignored generates a general tolerance. However, interaction with the undocumented immigrant, as is the case with all members of vulnerable groups, can undermine this tolerance by further increasing his or her particularized visibility, by making him or her stand out from others. The more stigmatized the group from which the undocumented immigrant comes, and the more he or she can be readily identified as a member of that group, the more the group is seen as some sort of a threat, and the less the cosmopolitan tendency described in this paper will operate. This logic applies to all minorities. High and particularized visibility—as opposed to superficial and generalized visibility—poses a threat for subordinate groups. For instance, the more Jews became visible as Jews in urban life with the emancipation from the ghetto, which increased interaction with Gentiles, the more antisemitism rose in 19th-century Europe. However, we are not arguing that interaction with non-natives can never increase cosmopolitan attitudes—only that interaction, especially with low-status groups, is always risky.
This paper stresses the role of morphology in the making of attitudes. It is harder to think of a deeper morphological variable than population density, something that has been overtaken by size and ethnic/racial composition in the studies of cities. Yet population density has immediate and very powerful effects.Footnote 21 In our account, morphology affects attitudes through lived experience. This is markedly different from most social scientific explanations, which rely on interests or values, or a combination of the two. One effect of higher density that warrants particular attention is that it generates exposure to people we have not chosen to be exposed to; that is, to those who are unlike ourselves. Exposure here often involves little or no interaction, and it is usually superficial; still, the habituation that results might be more civic—albeit unintentionally so—than the effects of the voluntary associations praised by commentators from Tocqueville [Reference Tocqueville2001] to Putnam [Reference Putnam2001], in which a person is more likely to associate with people like themselves. The quintessential public sphere in denser places is the street, the public transport, or the neighbourhood café. The quintessential public sphere in a less dense place is more likely to be the church or the bowling club. Streets, underground trains, and cafés contain more heterogeneous populations who were already there before we entered these spaces; here, visual exposure and space-sharing, usually with little or no interaction, are more likely to lead to a superficial, yet real and egalitarian tolerance of all others. Churches or bowling clubs, as they are voluntarily created, attract and consist of more homogeneous populations; here, a high level of interaction among similar others generates not only more community, but also lower tolerance of outsiders and of non-conformity in the people that one associates with.
In conclusion, this paper has proposed a novel understanding of cosmopolitanism: one that is not based on a celebration of difference, but rather on indifference. There is an irony here. Cosmopolitanism grows in cities because they are denser; but the more strongly it grows, the more it becomes indifferent to citizenship in the legal, contemporary sense of the term: that is, membership of a national polity. Understanding cosmopolitanism requires us to reconsider the original sense of citizenship: residence in a city, in a place with high density. Understanding it in this simple sense—as opposed to doing so in terms of the legal and moral load that the term has acquired over time—has profound sociological effects.
