National Trauma and the Fear of Foreigners: How Past Geopolitical Threat Heightens Anti-Immigration Sentiment Today

Abstract

This paper introduces a historical, macro-political argument into the literature on anti-immigration sentiment, which has mainly considered individual-level predictors such as education or social capital as well as country-level factors such as fluctuations in labor market conditions, changing composition of immigration streams, or the rise of populist parties. We argue that past geopolitical competition and war have shaped how national identities formed and thus also contemporary attitudes toward newcomers: countries that have experienced more violent conflict or lost territory and sovereignty developed ethnic (rather than civic) forms of nationalism and thus show higher levels of anti-immigration sentiment today. We introduce a geopolitical threat scale and score 33 European countries based on their historical experiences. Two anti-immigration measures come from the European Social Survey. Mixed-effects, ordinal logistic regression models reveal strong statistical and substantive significance for the geopolitical threat scale. Furthermore, ethnic forms of national identification do seem to mediate this relationship between geopolitical threat and restrictionist attitudes. The main analysis is robust to a wide variety of model specifications, to the inclusion of all control variables known to affect anti-immigration attitudes, and to a series of alternative codings of the geopolitical threat scale.

Over the past two decades, a new literature on anti-immigration sentiment in Europe has emerged that uses large-scale, multi-country surveys to understand where and why such sentiment is more pronounced. This research has explored how individual characteristics such as education and religiosity, competition between natives and immigrants for jobs and housing, perceptions of collective threat depending on the size and composition of immigrant streams, or the rise and fall of populist, right-wing parties determine how open or hostile to future immigration Europeans are. This paper addresses three limitations of this literature. First, this wide range of individual-level and country-level factors do not explain well which European populations are more hostile to immigrants than others. Second, existing studies emphasize temporally proximate factors to the exclusion of long-term processes. Third, the existing literature focuses on causal forces that operate within countries independent of one another, while relations between countries might matter as much, especially in Europe, where intense geopolitical competition and frequent wars have shaped processes of state formation and nation-building in ways that may well influence contemporary attitudes toward immigration.

Combining comparative-historical research and quantitative analysis of survey data, this paper introduces a macro-political, historical legacy argument into the debate. We argue that the varied geopolitical histories of countries, particularly whether national sovereignty and/or territory have been lost or threatened, help explain which countries are more hostile toward immigration—above and beyond the causal forces that have already been identified in the literature. Past geopolitical threat centers national identities on shared ancestry and ethnic commonality, rather than on the civic bonds to a state that seemed under siege and whose future appeared uncertain. This legacy of ethnic nationalism, in turn, makes individuals less welcoming to ethnically different immigrants.

In order to assess this argument empirically, we combine two dimensions of geopolitical threat into a scale: loss of territory or independence in the past and recurrent internal or external conflict. We then test whether the experience of geopolitical threat leaves a legacy of contemporary anti-immigration sentiment using data on 33 countries from the European Social Survey. We focus on two dependent variables: resistance toward potential immigrants of a different race/ethnicity than the host-country majority, and resistance toward potential immigrants from poor countries outside Europe.

Mixed-effects, ordinal logistic regression models reveal a substantively important and statistically significant association between the geopolitical threat scale and anti-immigration attitudes. This result is robust across a range of alternative model specifications and different codings of the geopolitical threat scale. This suggests that Europe's tumultuous geopolitical history has played a crucial role in shaping patterns of openness and closure toward immigrants. In a concluding, more tentative analysis, we examine the mediating mechanism foreseen by our theory and show that geopolitical threat is associated with ethnic forms of national identification, which in turn influences anti-immigration attitudes. We also test a second possible mediating mechanism based on the idea that geopolitical threat increases the political salience of nationalism overall, independent of its form, and thereby acts a breeding ground for nationalist, populist parties that in turn galvanizes anti-immigrant sentiments (as argued by Bohman [2011]). We show that the citizens of countries where nationalism is more prevalent in the discourse of political parties are not more restrictionist once we control for the level of geopolitical threat.

A Historical, Macro-political Theory of Anti-Immigration Attitudes

Whether emphasizing the individual or country level, existing efforts to account for attitudes toward immigration focus on contemporary causal forces that occur within countries conceived as independent from each other—for example, the amount of education a given individual has attained, or the size of a country's immigrant population. We believe a more historically informed and relational perspective can augment our understanding of anti-immigration sentiment. The geopolitical history of Europe is a history of winners and losers. Some nation-states, such as Austria, emerged from an empire lost, and others, such as the Czech Republic, owe their existence to such imperial collapse; in Europe's wars, some countries have gained or maintained territory, while others have had national territory taken away—in the case of Hungary, as much as three-quarters of its former domains. Some of today's countries, such as Latvia, lost independence for generations. Other countries, such as Spain, were or still are haunted by domestic violent conflict and the threat of secession, while others have a long and continuing history of conflict with nearby states, such as between Turkey and Greece.

What are the mechanisms that link threats to sovereignty and territory with anti-immigration attitudes? We argue that specific forms of nationalism and national identification provide the connection. In line with Triandafyllidou (1998), we focus on the negative, exclusionary sides of national identification. “The history of each nation,” she argues, “is marked by the presence of significant others that have influenced the development of its identity by means of their ‘threatening’ presence” (Triandafyllidou 1998, p. 593). Such threats are more or less important in the various histories of nation-state formation in Europe: Switzerland was never invaded by another country, never lost sovereignty, and never experienced a secessionist civil war, while Hungary lost substantial territory and was subjected to foreign rule for much of its modern history.

Such experiences will influence how the nation will be imagined and how its relationship to the state will be conceived; that is, they will affect forms of national identification. International war and domestic conflict as well as attendant losses of sovereignty or territory will make the nation, not the state, the main locus of loyalty and identification. In the modern world, the state is supposed to rule in the name of a nation, promote the welfare of its members, guarantee their equal treatment before the law, and protect them from the dangers of alien rule (Wimmer 2002). In extreme cases, such as Hungary, the state has been unable to live up to this ideal; the focus of identification and loyalty is therefore shifted to, and rests henceforth with, the national community. As a consequence, the “ethnic” aspects of national identity will be more fully developed than the “civic” aspects: the boundaries of the national community will be defined by shared ethnic descent as well as common culture and language, rather than by the institution of citizenship that binds the population to “its” state (Brubaker 1992).

How does ethnic nationalism translate into anti-immigration sentiment? First, as Brubaker has shown in his well-known historical comparison of France and Germany, ethnic forms of defining national boundaries make it more difficult to embrace immigrants of different origins as members of the national community, leading to more exclusionary boundary definitions in citizenship laws and immigration policies. Second, ethnic nationalism also leads to more restrictive attitudes toward the prospects of immigration by ethnic others. Pehrson, Vignoles, and Brown (2009) find that there is more anti-immigration prejudice in countries where national membership is imagined as a matter of ethnic ancestry or common language (see also Hjerm [1998]). Sides and Citrin (2007, pp. 479–80) link this finding to social identity theory's expectation that “the innate tendency towards ‘in-group favoritism’ is more intense when the group in question has great emotional significance”: while immigrants are by definition outsiders in the context of the nation-state, for those who imagine the nation in ethnic, rather than civic terms, their national identity has more emotional resonance and therefore sentiments toward immigrants and immigration should be more negative—an expectation that Sides and Citrin (2007) empirically confirm.

We thus build on these various findings and put them into a larger, geopolitical, and historical perspective by showing empirically where emphasis on the ethnic and cultural aspects of national belonging comes from: from traumatic or conflictual experiences of nation-state formation. We also extend the social-psychological mechanisms discussed above in two ways. First, a particular form of national identity, once emerged, will subsequently be “locked” into path dependency. It will become routinized in taken-for-granted templates of how the boundaries of national membership are defined, encoded in school curricula that transmit such templates from generation to generation (Darden 2013), hard-wired into life worlds through everyday symbolic practices (see the “banal nationalism” of Billig [1995]), and sedimented in collective memories that are transmitted through oral histories and cultural imaginaries (Assmann 2011). Conformingly, individuals who were young when these traumatic events occurred should not be more hostile to immigrants than those who have not experienced these events but have been socialized into an ethnic nationalism that emerged from them.

Second, over time the attitudes toward non-national others no longer concern only those nations or ethnic groups seen as responsible for traumatic geopolitical experiences. Rather, they become detached from the original configuration of actors and generalized to all foreigners. For example, immigrants from Turkey are unwelcome not only in countries who have been in direct conflict with Turkey in the past (such as Greece), but also in countries whose historical traumas had no connection whatsoever with Turkey or the Ottoman Empire. This generalization process is consistent with a common finding in social identity theory research, namely “that outgroup members are seen as more similar to each other than are ingroup members” (Brown 2000, p. 750), creating a subjective homogenization of the former, independent of objective indicators of difference. In our argument, the homogenized “outgroup members” are all those falling outside the boundaries of the ethnic nation. To be sure, not all variation in attitudes toward different ethnic or national outgroups is eroded in this process: Greeks still have more positive attitudes, on average, toward Serbian immigrants than toward those from Turkey. But at the end of this process, we expect to find more inter-country variation in citizen attitudes toward non-national others than intra-country variation in citizen attitudes toward different types of non-citizens. Below, we will offer some preliminary tests of this outgroup homogenization hypothesis, showing that the size of the immigrant population from countries that have in the past represented a threat to the focal country does not affect anti-immigration sentiments in a significant way.

This approach parallels other efforts to develop a geopolitical approach to anti-immigration sentiment that would complement the usual focus on individual-level and country-level variables related to labor market competition, party politics, demographic and cultural threats posed by different levels and types of immigration, and so forth (for a review, see Ceobanu and Escandell [2010]). Legewie (2013), exploiting the fact that a large-scale European survey was fielded before and after the bombing of a dance club for tourists in Bali, shows that such international events significantly affect attitudes toward immigration. His research also indicates, however, that this effect dissipates after a few weeks. We maintain this focus on how geopolitical events affect attitudes toward immigration, but we move the argument in a more historical and structural direction by emphasizing past events that, by virtue of their political importance and psychological impact, have left durable legacies for how citizens of different countries view the prospect of immigration by non-national others. Empirically, this means that our analysis will focus on the relatively constant differences between countries, not the ups and downs in anti-immigration sentiment caused by political events, economic cycles, and the like. As discussed below, this focus on relatively stable country differences captures a very important—but of course not the sole—component of the overall variance in anti-immigration sentiment in contemporary Europe.

In sum, we argue that individual citizens confront the prospect of immigration not only as less or more educated, as less or more religious, as less or more right-wing in their politics, as employed or unemployed, as situated in an economy that is less or more robust, or as living in a country with fewer or more foreign residents—to name some of the most common independent variables in research on anti-immigration attitudes. Individual citizens also confront the prospect of immigration as the inculcated members of “nations” that have had more or less traumatic experiences with foreign countries and peoples. These experiences structure nationalist narratives of us and them, leading to more or less emphasis on the nation or the state, which in turn affects the average disposition toward immigration of ethnic others. Past histories of lost wars, threat of secession, compromised sovereignty, or territorial loss thus translate into contemporary resistance toward immigration from outside the imagined national community (to cite the proverbial formula of Anderson [1991]).

Geopolitical Threat: Dimensions and Cases

Which geopolitical experiences are important? Territory and sovereignty are the main pivots of the modern world order of nation-states (Wimmer 2002). We thus distinguish two dimensions of geopolitical threat: losses of territory and/or sovereignty (the loss dimension) and threats of such losses by virtue of ongoing or recurring conflict, whether external or internal (the conflict dimension). The temporal horizon relevant for the argument is the period during and after the transition to the nation-state, that is, after a country has been founded or reorganized as a modern nation-state (data on nation-state creation are from Wimmer and Min [2006]).¹ This is because contemporary nationalist discourses, collective identities, and symbolic representations refer mostly to the national past, that is, to the period after “our country” first appeared on the geopolitical map, while earlier periods, although often included in nationalist narratives, for example in the form of “golden ages” of previous statehood, are arguably less salient in collective memories.

Each dimension entails two subdimensions: two types of actual loss (territory; independence) and two types of potential loss due to conflict (external; internal). Figure 1 presents the two major dimensions and their subdimensions. The shading indicates expectations for anti-immigration attitudes; darker shades entail greater hostility. If during or after the transition to the nation-state a country has experienced losses of both territory and independence as well as recurrent or ongoing external and internal conflicts (the upper left of figure 1), we expect its citizenry to be much more resistant to immigration than that of a country that has experienced none of these (the lower right portion of figure 1). Robustness tests in the online appendix explore to what extent these four subdimensions independently influence anti-immigration sentiment.

Figure 1.

Geopolitics and citizen attitudes toward immigrants and immigration

The loss and conflict dimensions are distinctive but interrelated. Many countries lost territory or sovereignty due to internal or external conflict. Hungary and other countries behind the Iron Curtain, for example, lost sovereignty to the Soviet Union after the Second World War. Because they have produced losses of territory or sovereignty/independence in the past, ongoing conflicts are perceived as threats to both in the present. Indeed, most contemporary conflicts involve territorial issues (see Huth [1996]), including but not limited to irredentist claims between states (such as in contemporary Ukraine), secessionist domestic conflict (as in Spain or Turkey), or contested sovereignty (as in Kosovo).

Our conception of geopolitical threat differs from geopolitical weakness in the classical international relations sense of the term (as measured, e.g., in the National Material Capabilities dataset). By geopolitical threat, we do not mean vulnerability to territorial invasions or other incursions on sovereignty in some hypothetical sense. We do take into account such vulnerability as additional factors to be considered (by including GDP per capita as a control variable and considering newly independent countries as particularly threatened in the robustness analysis). But it is not part of our conception of geopolitical threat, which is based on actual losses for the nation-state and actual conflicts that threaten the same, not on contemporary military-political power in the global arena.

The next several paragraphs conceptualize each of the geopolitical threat subdimensions and score countries on them. Recognizing that geopolitical threat may be more or less severe, we assign three scores: 2 indicates a severe threat, 1 a less severe threat, and 0 no threat. Following the textual discussion, we present a tabular summary of the coding (Table 1). With the rationale for each country's score on the geopolitical threat scale thus established, the next major section quantitatively tests our geopolitical theory of anti-immigration attitudes.² A Robustness section details the numerous recodings of the threat scale we undertook to ensure that it does not depend on the details of the coding decisions.

Loss of territory: We assign a value of 2 in the following historical circumstances. First, nation-state formation and major territorial losses were concurrent events for a handful of countries that are former land empires in Europe. These include Austria, Russia, and Turkey.³ Second, two other countries lost a very substantial amount of their territory (i.e., between ca. 30 and 70 percent) after nation-state formation, namely Hungary and Germany (Rothschild 1974; Turnock 2007). Note that we don't consider the loss of overseas colonial territories as having similarly traumatic consequences, but we do modify this coding decision in robustness tests reported in the online appendix.

We assigned the value of 1 to Bulgaria, Finland, Italy, and Romania (Allison 1985; Knippenberg and Markusse 1999; Paasi 1995), which lost some territory after lost wars but nowhere near the scale of Hungary and Germany. We also assigned a score of 1 when larger countries split into two, which often entailed a significant loss of territory. However, unlike the cases assigned a value of 2, the loss of territory was due to a failed political union, not a defeat in international war. No foreign power forced the split of the Czech Republic and Slovakia, or of Norway and Sweden (Lindgren 1959; Aarebrot 1982), or of Great Britain and Ireland. We thus assume this to have less traumatic consequences.

Some cases are more complicated. Germany did forcibly take Schleswig and Holstein from Denmark, but after World War I the latter was able to reclaim the portion most heavily populated by Danes (Rerup 1980; Rostgaard 2008). Denmark lost Iceland through a failure of union, however, and we thus assign it a score of 1. For the Czech Republic, we count the loss of Slovakia as a failure of union, but not vice versa for Slovakia. This is because Czechs represented the core group of Czechoslovakia, and Slovaks the peripheral one—in the same way that the English were the core group and the Irish a peripheral one in the pre-1922 United Kingdom (see, e.g., Leff 1997; Stein 1997) and the Croats and Slovenes represented peripheral groups in Yugoslavia. In the aftermath of World War II, Poland lost a significant portion of its eastern territories but also gained substantial territory to the west that was populated by Polish speakers (Carter 2007; Knippenberg and Markusse 1999). We therefore do not count Poland as a case of territorial loss.

Loss of independence: A large group of countries have a score of 2 on this dimension. They formed as nation-states between the 1860s and 1910s and then lost independence to the Soviet Union after World War II, whether de jure (Estonia, Latvia, Lithuania) or de facto (Bulgaria, Czech Republic, Hungary, Poland, and Romania) (Beissinger 2002).

Countries whose independence was heavily compromised receive a score of 1. After nation-state formation following World War I, Austria was prohibited from forming a union with Germany. It was then occupied by the Allies after World War II (Bluhm 1973; Hoffman 1951; Johnson 2008; Thaler 2001). Germany also was occupied after World War II, and part of the current nation-state, East Germany, was under Soviet domination for several decades after that war. The Great Powers imposed a foreign king on Greece for much of this country's post-Ottoman history, and this king had substantial power for much of this period (Clogg 2002). Foreign meddling by Great Britain and the United States in the Greek Civil War was also substantial (Kofas 2003; Miller 2009; Stefanidis 2007; Wittner 1982). Cyprus independence has been compromised in various ways ever since it became a nation-state in 1960, particularly due to intermittent interventions and meddling by Greece as well as due to Turkish military occupation (Aktar, Kizilyürek, and Özkirimli 2010; Anastasiou 2008; Attalides 1979; Ker-Lindsay 2011). Once again, the Czech Republic and Slovakia represent complex cases. When Czechoslovakia lost independence to the Soviets, neither of these countries officially existed. Following the logic outlined in the territorial loss subsection, we assign a score of 2 on the loss of independence dimension to the Czech Republic, and a 1 to Slovakia.

External conflict: External conflict in Europe is mostly a thing of the past, settled largely by two world wars. But there are exceptions. As the above discussion implies, Greece, Turkey, and Cyprus are involved in an ongoing conflict of varying intensity over the status of Cyprus. The Cyprus issue in fact is but one source of conflict between Turkey and Greece, which together represent “a classic ‘adversarial dyad’” (Heraclides 2010, p. 3). Cyprus, Greece, and Turkey thus merit a score of 2 on the external conflict dimension. Joining them is Russia, due to its recurrent border incursions in former imperial territories such as Georgia (Forsberg 1995; Ambrosio 2009). Two other countries were assigned a score of 1: Ireland, due to the conflict in Northern Ireland, and Croatia because of recent war and territorial disputes that, however, seem largely to be settled (Ramet 2006).

Internal conflict: While the geopolitical threat posed by external conflict is overt, internal conflict indirectly threatens territorial integrity. A score of 2 is assigned when there is at least one ongoing or recurrent conflict involving significant secessionist claims by minority groups. Note again that anticolonial independence movements in overseas territories do not count as secession, a coding decision we check in robustness tests. Belgium receives this score due to the recurrent Flemish-Walloon conflict, as do Cyprus (Turkish vs. Greek Cypriots), Spain (Basques and Catalans), Great Britain (Northern Ireland; Scotland), Russia (e.g., Chechnya), and Turkey (Kurds).⁴ A score of 1 is assigned to cases where minorities have not politically mobilized for independence, but the state perceives them as representing a potential threat to territorial integrity (see, e.g., legal treatment of Russians in Estonia and Latvia). We also assign a score of 1 to Ukraine on the basis of Crimeans’ push for independence or union with Russia in the first years of the 1990s (Furtado 1993); we assign a 1, rather than a 2, because the conflict was short-lived and, until several years after the time frame of our analysis, which ends in 2010, appeared to have been resolved (Sasse 2007).

How can we aggregate these four scores into a single measure? We opted for a simple additive scale, summing the scores for each of the four dimensions. The additive scale ranges from zero for countries with the least geopolitical threat to eight for those who score the maximum on all dimensions. We opted for this simple mode of aggregation in the absence of strong empirical or theoretical reasons for doing otherwise. Treating each dimension as equivalent is therefore the most sensible approach. In the Robustness section, we show that each dimension has similar effects on the outcome as the aggregated score, thus increasing our confidence in the latter.

Table 1 (see below) presents the coding for each country. We see that about half of the countries have very low geopolitical threat scores (0–1), another 13 countries have moderate scores (2–3), and four countries are in the higher range (4–6).

Table 1.

Country Scores for Geopolitical Threat

Country	Loss		Conflict
	Territory	Independence	Internal	External	Total Score
CH Switzerland	0	0	0	0	0
FR France	0	0	0	0	0
IS Iceland	0	0	0	0	0
LU Luxembourg	0	0	0	0	0
NL Netherlands	0	0	0	0	0
NO Norway	0	0	0	0	0
PT Portugal	0	0	0	0	0
SI Slovenia	0	0	0	0	0
DK Denmark	1	0	0	0	1
FI Finland	1	0	0	0	1
HR Croatia	0	0	0	1	1
IE Ireland	0	0	0	1	1
IT Italy	1	0	0	0	1
SE Sweden	1	0	0	0	1
SK Slovakia	0	1	0	0	1
UA Ukraine	0	0	1	0	1
BE Belgium	0	0	2	0	2
ES Spain	0	0	2	0	2
LT Lithuania	0	2	0	0	2
PL Poland	0	2	0	0	2
AT Austria	2	1	0	0	3
BG Bulgaria	1	2	0	0	3
CZ Czech Republic	1	2	0	0	3
DE Germany	2	1	0	0	3
EE Estonia	0	2	1	0	3
GB United Kingdom	1	0	2	0	3
GR Greece	0	1	0	2	3
LV Latvia	0	2	1	0	3
RO Romania	1	2	0	0	3
HU Hungary	2	2	0	0	4
CY Cyprus	0	1	2	2	5
RU Russian Federation	2	0	2	2	6
TR Turkey	2	0	2	2	6

Data and Methods

Dependent Variables

To examine the relationship between the geopolitical threat scale and citizens’ attitudes toward immigration, we use data from the European Social Survey (ESS), which has been conducted every two years since 2002. We use data from all 33 countries that participated at least once between 2002 and 2010. For each country we use the most recent survey available, which for most countries is the fifth round (2010). For Iceland, Italy, and Luxemburg we use the second round (2004), and for Latvia, Austria, Romania, and Turkey we use the fourth round (2008). So that we can clearly identify the influence of past geopolitical threats on anti-immigration attitudes of native majority members, we drop from the sample all immigrants and all those with immigrant parents.

All ESS rounds included three items that gauge the respondent's desired level of immigration for different migrant groups using a four-point scale ranging from “allow many” to “allow none.” One question refers to potential immigrants who are not of the same ethnicity: “How about people of a different race or ethnic group from most [country] people?” The second question refers to potential immigrants who come from poor countries outside Europe: “How about people from the poorer countries outside Europe?” A third question, which is first in the survey sequence, refers to potential immigrants who might be considered of the same race/ethnicity as the host-country majority: “To what extent do you think [country] should allow people of the same race or ethnic group as most [country] people to come and live here?”

This third question offers an opportunity to further test our geopolitical threat theory: Such threat should affect attitudes toward only migrants not considered part of the nation. Co-ethnics, by contrast, will appear as closer to the national community, and may even be considered, through nationalist lenses, as “coming home” into the national family and thus perceived as non-threatening. To test these propositions, we run models with all three questions as separate dependent variables.

The data don't allow us to conclusively say what kind of potential migrant respondents have in mind. For example the question about immigrants “of same race or ethnicity” could be interpreted in different ways, depending on whether the respondent makes a subjectively meaningful distinction between race and ethnicity. To test whether or not respondents understand these questions in terms of race (distinguishing Europeans from non-Europeans), we conducted additional analysis of two ESS rounds that asked a set of relevant questions (rounds 1 and 7). In line with our conceptualization of the dependent variable, this additional analysis shows that respondents seem to interpret the question about immigrants of different race or ethnicity as referring to non-national others, whether these hail from outside Europe (supposedly of different race) or from within Europe (of the same race). We also note here that existing research evidences the cross-cultural validity, from Vladivostok to Gibraltar, of both the same race/ethnicity and different race/ethnicity items (Meuleman and Billiet 2012).

The first section of table 2 provides summary statistics for the three dependent variables. While respondents were more hostile toward immigrants from poor countries outside Europe as compared to those of a different race/ethnicity, the distribution is very similar across variables: fewer than 15 percent would like to allow many to come, while about one-third responded “allow some” or “allow a few.” In contrast, attitudes toward potential immigrants of the same race/ethnicity are much more inclusive, in line with our expectations. Almost a quarter (23 percent) would allow many to come, and another two-thirds would allow some or a few.

Table 2.

Descriptive Statistics for Dependent and Independent Variables

Dependent variable Allow many/few people:	Of different race/ethnic group	From poor countries outside Europe	Of same race/ethnic group
many (1)	13	12	23
some (2)	35	32	40
a few (3)	33	34	25
none (4)	19	23	12
Pct. missing	5	5	4
Aggregated to the country level
Mean	2.55	2.65
Std. deviation	0.32	0.34
Independent variables:
Individual level	Mean	Sd	Pct Missing
Age	47.4	18.5	>0.01
Education	12.0	4.1	1.1
Religiosity (0 to 10)	4.8	3.0	1.1
Trust in Parliament (0 to 10)	4.0	2.7	2.6
Trust in people (0 to 10)	4.8	2.5	0.4
Attendance of services	2.7	1.5	0.7
Social meeting frequency	4.8	1.6	0.7
Discrete variables	Pct.		Pct Missing
Married (ever)	73
Male	46		0.04
Member of union	41
Supervisor	23
Father's ed (prim or less)	34
secondary	44
tertiary	15
missing	7
Unemployed in prev. 12 m	12
Money is tight	10
Have s.o. to discuss intimate matters	87		1.7
Location: (large city)	23		0.02
Suburb	10
Small town	30
Village	32
Farm/rural	6
Country-level variables	Mean	Sd	Pct Missing
Economic indicators
GDP per capita 1k  USD	26.7	13.4
Pct. change in GDP per  capita	1.8	2.5
Percent foreign born	10.8	7.1
GINI (post-tax)	30.2	5.0
Migration & diversity
Pct. foreign born	10.8	7.1
Asylum applications  per 1 k pop	0.8	1.6
Migration flow per 1 k  pop	8.2	7.1	9.1
Inequality and redistribution
Post-tax GINI	30.2	5.0
Diff. pre to post tax  GINI	16.6	4.1	24.2
Safety net exp. 1 k EUR	5.7	4.2	9.1
Globalization (KOF Index)	81.3	7.3
Geopolitical threat scale	1.9	1.7

Our analysis focuses on relatively stable differences between countries in these measurements of anti-immigration sentiment, while also acknowledging that other factors (many of which we control for in the empirical analysis) explain other, temporally less stable portions of the overall variance. Is this emphasis on relative stability empirically meaningful? There was on average a 0.17 difference between the minimum (1) and the maximum (4) restrictiveness toward immigrants recorded in any available survey year for the same country. These changes over time within countries pale in comparison with differences between countries, on which our analysis focuses: the latter are 7.5 times more pronounced than the former. Country differences remain relatively constant over time: across all the survey years, 21 of the 32 countries remain in their quartile of a rank order ranging from least to most anti-immigrant countries. Among the 11 that move to the adjacent quartile, only three had more than a quarter-point change on the four-point scale.

But what about general trends affecting all countries in similar ways—which our theory and analysis would not be able to make sense of? While there was a movement in the 1990s toward more anti-immigration attitudes in all 12 European countries analyzed by Semyonov, Raijman, and Gorodzeisky (2006), in some countries this trend has gone in the other direction in the 2000s. Based on our data, this was the case in Bulgaria, Estonia, France, Latvia, Lithuania, Norway, Russia, Sweden, and Turkey, while Czechs, Italians, Romanians, Slovaks, and Ukrainians became more anti-immigration on average. In the other 19 countries, there is no clear trend. Focusing on the differences between countries that remain relatively stable over time, as our theory and measurement do, therefore is a reasonable approach. This is supported by additional analysis: our results remain substantially identical for every single wave of the ESS survey (see online appendix).

Control Variables

Individual level

An extensive review of the quantitative literature on immigration attitudes identified all significant individual-level variables in previous research. In addition to age and gender, these are:

• – Place of residence: A long line of research shows a positive relationship between urban residence and social tolerance, including toward immigration (e.g., Green 2009). The ESS asked respondents whether they resided in a big city, a suburb, a small town, a country village, or a farm in the countryside. We enter this information as a series of dummy variables with big cities as the omitted reference category.

• – Educational attainment: As a measure of respondents’ socioeconomic status, we include years of education. More educated individuals tend to express more welcoming attitudes toward immigrants (e.g., Hainmueller and Hiscox 2007).

• – Father's education: The ESS gives information on the educational attainment of the respondents’ fathers, which we coded into three categories: primary education or less, secondary education, and tertiary education or more. We group missing cases (7 percent) into a separate dummy variable.

• – Economic insecurity: Research suggests that economic insecurity makes people less welcoming toward immigrants (e.g., Scheepers, Gijsberts, and Coenders 2002). We include two relevant indicators: whether respondents report a spell of unemployment that lasted 12 months or longer; and whether respondents find it very difficult to live off their current household income.

• – Religiosity (e.g., Davidov et al. 2008): We use two ESS variables to measure respondents’ religiosity, which is supposed to lessen anti-immigration sentiment. The first is respondents’ ranking of themselves on a scale from 0 (least religious) to 10 (most religious). The second measures how often respondents attend religious services (every day, more than once a week, at least once a month, etc.). We enter both variables as linear predictors with higher values indicating higher religiosity.

• – Social capital (e.g., Rydgren 2009): We measure social capital, which should make individuals more open to immigration, using three variables: whether a person has someone with whom to discuss intimate matters; a linear measure (0 to 10) of respondents’ assessment whether most people can be trusted or, in contrast, “you can't be too careful”; and how often one meets socially with friends, colleagues, or relatives, which is scored on a seven-point scale ranging from “never” to “every day” and entered as a linear predictor.

• – Institutional trust (e.g., Ceobanu and Escandell 2008): Trust in the political system is purported to decrease anti-immigration attitudes. We measure this using an item that queries respondents’ degree of trust in their country's parliament; 0 indicates no trust and 10 signals complete trust.

Country level

Previous research has examined the influence of various country-level variables on immigration attitudes. We group these into four categories—economics, immigration history and diversity, inequality and redistribution, and globalization.

• – Economic indicators (e.g., Meuleman, Davidov, and Billiet 2009): To index economic conditions and trends, we use GDP per capita and the country's unemployment rate, along with measures of change for both. For GDP per capita, we draw on the Penn World Tables (Heston, Summers, and Aten 2012) and use a measure that is adjusted for purchasing power parity at 2005 constant prices. As measures of trend, we calculate the relative change from the year prior to the survey as well as change relative to three years before the survey. For the unemployment rate, we use data from the CIA World Factbook. We also calculate the relative change from the previous survey year.

• – Migration and diversity (e.g., Coenders, Lubbers, and Scheepers 2009): Three indicators measure the existing immigrant stock as well as migration flows. Drawing mainly on Eurostat data, we calculate the share of the population that is foreign born in the survey year or the closest year for which such data are available. We also use data on asylum applications and migration flow (again standardized to the country population) assembled by Eurostat and the UNHCR.

• – Inequality and Redistribution (e.g., Jesuit, Paradowski, and Mahler 2009): As a measure of economic inequality, we use the post-tax Gini coefficient, drawing on data from the OECD and Eurostat. We also consider two redistribution measures: the change in Gini coefficient due to taxation and redistribution, and social protection expenditures. For the former, we use pre-tax Gini coefficients calculated by the OECD; the social protection expenditures (in Euros) are collected by Eurostat.

• – Globalization: Researchers have explored, with varying results, the relationship between globalization and both national identity (Kunovich 2009; Ariely 2012) and attitudes toward immigrants (Kaya and Karakoç 2012). We include the KOF index for globalization (Dreher 2006), which ranges from 1 to 100.

Analysis

Since we will analyze ordered categorical dependent variables in a multilevel data structure, we use a mixed-effects ordinal logistic regression model. To check the robustness of results, we also estimate an ordered logistic regression and use clustering at the country level to adjust the standard errors. The results are virtually identical.

Because we have a relatively small number of country-level observations, adding all country-level variables at once creates a degrees-of-freedom problem. We thus went through a multi-step process, using the geopolitical threat scale together with one group of variables at a time (i.e., all economic, all indicators of immigration and diversity, and so on). We then retained the significant variables from each group for the final model specification.

One could argue that many of our control variables both at the country level and the individual level are endogenous or “downstream” from the processes we are examining and thus introduce post-treatment bias. To test for this, we estimate a simple bivariate relationship between geopolitical threat and anti-immigration attitudes and also a set of multilevel models that exclude all country-level control variables and include only basic demographic controls on the individual level. In both cases, results (not shown here) confirm the main analysis we present below.

Results

Main Analysis

Does the level of geopolitical threat (GPT) affect attitudes toward immigration? For the sake of brevity, we discuss only the same race/ethnicity and different race/ethnicity items. The results for potential immigrants from poor countries outside Europe are substantively the same as for the different race/ethnicity item—suggesting, in line with our argument, that all non-national others are seen through the same lens.

As a first approximation, figure 2 presents simple bivariate relationships. We plot the average response on the immigration question (scored from 1 to 4, with 4 representing the greatest restrictiveness toward potential immigrants) against the GPT rating of each country. The left panel concerns immigrants of a different race/ethnicity. Although there is a fair amount of dispersion at each level of GPT, a clear positive relationship is visible, as indicated by the gray regression line. According to this simple bivariate relationship, a one-unit increase in the GPT scale is associated with about a 0.1 point increase in the average anti-immigration attitudes. To put this in perspective: a two-point move on the GPT rating (which is a bit more than one standard deviation), or a move from Italy (GPT = 1) to Great Britain (GPT = 3), for example, would, given a standard deviation of 0.32 on the dependent variable, translate into more than half a standard deviation (0.61) increase in the level of anti-immigration attitudes.

Figure 2.

Average level of attitudes toward immigrants from different ethnic/racial background (left panel) and same ethnicity race (right panel) on a scale from 1 to 4, with indicating more restrictive preferences plotted against the geopolitical threat scale

In contrast, the relationship between the GPT scale and attitudes toward immigrants of the same race/ethnicity, shown in the right panel of figure 2, is much weaker; even as a bivariate relationship, it does not reach levels of conventional statistical significance—in line with our expectation that past GPT affects only attitudes toward non-national others.

The multilevel regression model (table 3) shows that these relationships hold once we account for variation in other variables. The statistical significance of the GPT scale is strong (p < 0.001), and it holds up under a wide range of alternative specifications (on this, more below). One can also compare the effect size of a one-unit increase in the GPT scale to the effects of other variables. For example, the effect of one more point on the geopolitical scale (corresponding to just over half a standard deviation) is comparable to that of four years less education, or moving from a large city to a farm or rural area. The coefficients of the control variables conform to findings in the literature.

Table 3.

Mixed-Effects Ordinal Logistic Models of Citizen Restrictiveness toward Immigrants

	Of different ethnicity		From poor countries		Of same ethnicity
	Estimate	z value	Estimate	z value	Estimate	z value
Male	0.04	2.38*	0.07	3.76***	−0.02	−0.87
Age (in decades)	0.08	12.14***	0.09	14.15***	0.04	6.62***
Married (ever)	0.05	2.09*	0.07	2.93**	0.04	1.48
Location: Suburb (a)	0.06	1.79.	0.04	1.07	0.06	1.66.
Small town	0.07	2.85**	0.06	2.48*	0.08	3.14**
Village	0.20	8.12***	0.15	5.92***	0.18	7.28***
Farm/rural	0.24	5.63***	0.16	3.79***	0.18	4.30***
Education (years)	−0.06	−23.93***	−0.05	−20.42***	−0.06	−21.59***
Member of union	−0.08	−4.02***	−0.07	−3.59***	−0.10	−5.02***
Supervisor	−0.10	−4.72***	−0.08	−3.91***	−0.12	−5.75***
Father's ed: secondary (b)	−0.06	−2.51*	−0.05	−1.89.	−0.07	−2.95**
tertiary	−0.28	−8.88***	−0.24	−7.54***	−0.25	−7.90***
missing	−0.03	−0.83	−0.05	−1.24	0.00	0.12
Unemployment spell 12m+	0.01	0.35	0.04	1.35	−0.03	−1.27
Money is very tight	0.20	6.50***	0.18	5.67***	0.22	6.93***
Religiosity	0.01	2.54*	0.00	0.12	−0.01	−2.82**
Attendance of services	0.02	2.10	0.01	1.72.	0.03	3.46**
Trust in people	−0.09	−21.99***	−0.09	−22.05***	−0.08	−21.33***
Social meeting freq.	−0.05	−7.92***	−0.04	−6.50***	−0.06	−9.67***
Confidant	−0.26	−9.04***	−0.24	−8.55***	−0.24	−8.52***
Trust in Parliament	−0.07	−17.72***	−0.06	−15.18***	−0.05	−13.34***
Geopolitical threat	0.25	4.45***	0.23	3.31***	0.06	1.18
Pct. GDP change	−0.09	−2.74**	−0.09	−1.99*	−0.06	−1.93.
Pct. foreign born	0.02	1.47	0.01	0.97	0.00	−0.20
GINI (post tax)	−0.03	−1.26	−0.02	−0.84	0.02	1.06
Globalization index	0.02	1.55	0.01	0.76	0.05	3.52***
Threshold coefficients
1|2	−2.98		−2.98		−2.77
2|3	−0.90		−0.90		−0.78
3|4	0.91		0.91		0.87
Observations	47,111		46,875		47,160
Countries	33		33		33
Variance of random Effects	0.20		0.31		0.19

Note: *** p < 0.001 ** p < 0.01 * p < 0.05 . p < 0.1

Omitted reference categories: (a) Large city, (b) Primary education or less.

Table 4, which summarizes predicted probabilities at all levels of the independent variable, provides the most direct information on the effect size of the GPT scale. Looking at the first two columns in table 4, we see that, holding other variables constant at their means or median values, in a country with a GPT score of zero we expect nearly two-thirds of citizens (62 percent) to answer in the category “allow many” or “allow some” immigrants. At a value of two (which is about average), in contrast, we expect just below half (48 percent) in those two categories. And at the other end of the spectrum, in a country with the highest threat score of 6, we predict only 26 percent to answer in either of these two inclusive response categories.

Table 4.

Predicted Marginal Answer Distributions of Attitudes toward Immigrants of a Different Race/Ethnicity at Different Values of the Geopolitical Threat Scale

Geopolitical Threat scale	Admit many	Admit some	Admit a few	Admit none
0	0.18	0.44	0.29	0.10
1	0.14	0.41	0.32	0.12
2	0.11	0.37	0.36	0.15
3	0.09	0.34	0.38	0.19
4	0.07	0.29	0.40	0.23
5	0.06	0.25	0.41	0.28
6	0.05	0.21	0.41	0.33

Note: Other variables fixed at their mean (continuous) or modal value (discrete).

As mentioned above, the analysis of attitudes toward immigrants from poor countries outside of Europe produces virtually identical results. These findings also speak to the hypothesis, elaborated in the Theory section, that GPT not only affects attitudes toward immigrants from the specific country that posed a threat in the past, but becomes generalized to immigrants from all countries. Since past threat was almost exclusively inflicted by other European countries, immigrants from poor non-European countries would not fit the specific threat profiles for any of the countries we examined. To further test this hypothesis, we introduce the contemporary number of immigrants from countries that posed a threat in the past as a control variable. As reported in the online appendix, this variable neither reaches standard levels of significance nor affects our main results.

The results for the same race/ethnicity variable are also in line with theoretical expectations. The point estimate is small (just 0.06), and the t-value of 1.18 does not reach statistical significance. It is thus meaningful to differentiate attitudes toward co-national immigrants from attitudes toward non-national immigrants, rather than combining all answers to the corresponding questions into a single factor, as is often done in quantitative research (e.g., Meuleman, Davidov, and Billiet 2009; Davidov and Meuleman 2012).

Robustness checks

The online appendix details a series of robustness checks, which we briefly summarize here. The first set concern the coding of the key independent variable, the GPT scale. In the development of the scale, we went through several iterations, a meaningful exercise given the uncertainty of some coding decisions; all of the alternative scales produced substantively and statistically significant results. After settling on the version that entailed the most parsimonious dimension specification and, in our view, the most accurate scoring (which was not, we should add, the version that produced the highest coefficients and strongest statistical significance), we tested four alternative codings: (1) to account for the possibility that in recently independent countries (such as Latvia) citizens might feel especially insecure in the international environment, we created a version of the scale that adds two extra threat points to these countries; (2) we added a threat point (loss of sovereignty) to all countries occupied by Nazi Germany during World War II; (3) we created a version that counts loss of overseas colonies as an additional point on the GPT scale and counts it as two points if this loss was accompanied by anti-colonial wars of liberation; and (4) we counted only those territories as losses that were populated by members of the core nation, rather than ethnic minorities. The results for all these different specifications of the main independent variable conform to the analysis presented above.

Second, we show that results are not driven by a handful of cases that assume extreme values on the GPT scale by re-estimating the model with a subset of countries.

Third, to test whether summing up all four dimensions is justifiable, we use each of the four components (loss of sovereignty, loss of territory, external conflict, internal conflict) of the GPT scale as an independent variable in a separate regression.

Fourth, because our argument emphasizes the importance of long-term social processes and relative stability over time, GPT should predict levels of anti-immigration sentiment not just in a single year but over a range of years. To test this, we aggregated all available ESS rounds by country and year; we find, again, identical results. Finally, and along similar lines, we reproduced the main analysis for every single wave of the ESS and plotted the bivariate relationship between geopolitical threat and the mean anti-immigration attitude per country for each survey year. As appendix figure 1 shows, the results are substantially identical as those for figure 2 above.

Mediating factors: Ethnic nationalism

We now turn to an empirical evaluation of whether ethnic nationalism mediates between a past of GPT and contemporary anti-immigration attitudes. Recall our argument that GPT produces forms of national identification focused on shared ethnic ancestry rather than on the political bonds of citizenship. This form of nationalism, we suggested, represents one of the main causes of anti-immigration sentiment. The ESS does not provide measures of national identification at the individual level, so we conduct this analysis at the country level using data from the ISSP module on National Identity from 2003. It contains a question indexing the degree to which membership in the nation is considered a matter of ethnic ancestry: “Some people say that the following things are important for being truly [NATIONALITY]. Others say they are not important. How important do you think it is … to have [NATIONALITY] ancestry?” Unfortunately, only 20 countries participated in both the ESS and ISSP 2003, and two of these don't provide responses on this item, which leaves a sample of 18. The analysis reported below is therefore rather tentative.

We also explore an alternative mediator suggested by Bohman (2011): the presence of parties making nationalist claims that may in turn foster anti-immigration sentiment. To test this, we use data from the Comparative Manifesto Project (CMP), which offers a rich content analysis of party platforms. Following Bohman (2011, p. 463), we use the variable “nationalistic articulation with the aim of defining and consolidating the national identity.” We include all manifestos from 1990 to the year of the ESS data (2010 for most countries). These data cover all 33 countries in our analysis.

As shown in the top panel of figure 3, the simple bivariate relationship—not taking the two mediators into account—between GPT and the average level of anti-immigration sentiment amounts to a standardized coefficient of 0.52. In the bottom panel of figure 3, we introduce the two mediating variables.⁵ In line with our theoretical expectations, GPT strongly predicts levels of ethnic nationalism, which in turn predict anti-immigration attitudes. Once this variable is included, the direct path leading from GPT to restrictionist attitudes is attenuated by about one-third.

Figure 3.

Path model relating geopolitical threat to country average attitudes toward immigration. Graphic does not represent a correlation between the two mediating variables as well as residual errors. Coefficients displayed are standardized

Significance levels: * p < 0.05 *** p < 0.001

GPT also robustly predicts nationalist party platforms; however, with GPT included in the regression model, party-articulated nationalism does not predict anti-immigration attitudes. In contrast to Bohman (2011), therefore, we find no evidence that nationalist party platforms are an independent, alternative causal mechanism that mediates the relationship between GPT and anti-immigration attitudes.

Conclusion

To advance our understanding of why the citizens of some countries are more hostile to immigration than those of others, we expanded the analytical horizon beyond the boundaries of individual nation-states and the temporal confines of the recent past. The statistical and substantive significance of the geopolitical threat scale as well as the robustness of the results across a range of alternative model specifications and coding rules suggest that Europe's history of nation-state formation has played an important role in shaping patterns of resistance and openness to immigration. Geopolitical competition and war, and the losses of territory and sovereignty—that these may entail, shape the nature and salience of national boundaries, which in turn influence the citizenry's openness to immigrant aliens as potential members of the nation. Consistent with this argument, a conflict-ridden or painful history of nation-state formation leads to more ethnic forms of national identification, which in turn increase hostility against immigrants who are perceived as non-national others, but not against immigrants who share the same national origin.

We note here that there is a residual direct effect of geopolitical threat when taking forms of nationalism (ethnic versus civic) into account, as shown in the mediation analysis above. This suggests that other processes, not covered by our theoretical argument, are important in translating geopolitical histories into national narratives that stimulate greater degrees of antagonism toward foreigners. While we tentatively ruled out that nationalist party platforms could play such a mediating role, more research is needed to explore additional mechanisms, such as the role of school curricula (see Darden [2013]) or mass media discourses (see Millas [1991]).

This research contributes to growing efforts to take seriously the relationship between geopolitics and immigration (Nagel 2002; Hyndman 2012; Legewie 2013). Our findings are in line with research that questions the mainstream focus on either competition for jobs or other individual resources (Hainmueller and Hiscox 2010) and that instead places long-term processes of nation-state formation at the center of analysis (Wimmer 1997, 2002). Our argument and findings are also in line with Onraet, van Hiel, and Cornelis's (2013) recent shift of the analytical focus from individual-level to country-level sources of threat—though they study right-wing attitudes, not anti-immigrant sentiment. Rather than collective threats of a short-term nature, such as immigration flows or changing economic conditions (see evidence provided by Davidov and Meuleman [2012]), we find that what matters are long-accumulating threats to the territorial integrity and political sovereignty of a nation.

It will be important to assess how well our argument travels to other contexts. We would predict, for example, that citizens of Chile should be more open to potential immigrants of different origin than either Bolivia or Peru, given Chile's geopolitical victories over these countries in the early nineteenth century. Insofar as our European findings do travel to other contexts, a more general theoretical point could be made: that in analyzing the social world, we would do well to keep in mind that “the majority of actors are the dead” (Archer 1996, p. 696, paraphrasing Comte). In less aphoristic terms, legacies of geopolitical competition, and the different forms of state formation and nation-building that they produced, play a crucial role in shaping contemporary political attitudes across countries and continents.

Notes

1

For countries that became independent, distinct polities only very briefly in the aftermath of the First World War (e.g., Ukraine, Slovenia), we use the second, more recent date of nation-state formation. This coding decision does not affect our results.

2

A full bibliography is available upon request.

3

Alternatively, one could argue that only those lost imperial territories that were populated by the empires’ core nation were perceived as a loss, which was not the case for Austria, Russia, and Turkey. The coding changes entailed by this interpretation do not affect our results, as discussed in more detail in the online appendix. We maintain that the formation of nation-states in crucibles of geopolitical loss of a falling empire is equivalent to losing territory some years after nation-state formation. We also maintain that losing contiguous territory in the context of these European land-based empires is qualitatively different from losing overseas colonies. Changing the latter coding decision also does not affect our results.

4

Consistent with the main lines of our argument, peripheral movements that involve a tiny share of the population and territory and do not threaten the integrity of the nation-state in its current form—for example, the Corsican issue in France or the Tyrolean one in Italy—are not scored as experiencing ongoing and recurrent internal conflict.

5

We estimate this model using a full information maximum likelihood (FIML) estimator, which allows us to address the problem of missing data on the “ethnic nationalism” variable and retain the full set of 33 observations.

About the Authors

Wesley Hiers is Visiting Lecturer of Sociology at the University of Pittsburgh. His research examines the historical and political processes that shape ethnos-based social closure and inequality. He is the author of “Party Matters: Racial Closure in the Nineteenth-Century United States,” Social Science History 37(2) [2013], and coauthor of Politicized Ethnicity: A Comparative Perspective (Palgrave Macmillan, 2016). He teaches theory, methods, intro, political sociology, and a variety of race-ethnicity courses.

Thomas Soehl is Assistant Professor of Sociology at McGill University. His research examines the socio-political attachments of migrants, the ways in which host societies transform migrants, and how migration challenges modern nation-states. His recent work on acculturation and the inter-generational transmission of culture and socio-economic characteristics has been published in the American Journal of Sociology, the International Migration Review, and the Journal of Ethnic and Migration Studies.

Andreas Wimmer is the Lieber Professor of Sociology and Political Philosophy at Columbia University. His research brings a long-term and globally comparative perspective to the questions of how states are built and nations formed, how individuals draw ethnic and racial boundaries between themselves and others, and which kinds of political conflicts and war results from these processes. Using new methods and data, he continues the old search for historical patterns that repeat across contexts and times. His most recent book publication is Nation Building: Why Some Countries Come Together While Others Fall Apart (forthcoming, Princeton University Press).

Supplementary Material

Supplementary material is available at Social Forces online, http://sf.oxfordjournals.org/.

Author notes

This project has benefited from the generous financial and intellectual support of the Foundation for Population, Migration, and Environment in Zurich (Project 11-42). A version of this paper was presented at the 2014 meeting of the Association for the Study of Nationalities at Columbia University in New York City. We thank participants for their insightful feedback. All authors contributed equally. Names are listed in alphabetical order. Direct all correspondence to Wesley Hiers, Department of Sociology, 2421 WWPH, University of Pittsburgh, 230 S. Bouquet St., Pittsburgh, PA 15260. Telephone: [phone]; e-mail: .