American Political Science Review (2019) 113, 2, 442–455

doi:10.1017/S0003055418000813 © American Political Science Association 2018. This is an Open Access article, distributedunder the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives licence (http://creativecommons.org/licenses/by-nc-nd/4.0/), whichpermits non-commercial re-use, distribution, and reproduction in anymedium,provided theoriginalwork is unalteredand is properlycited. The written permission of Cambridge University Press must be obtained for commercial re-use or in order to create a derivative work.

Does Exposure to the Refugee Crisis Make Natives More Hostile?DOMINIK HANGARTNER ETH Zurich and London School of Economics

ELIAS DINAS European University Institute and University of Oxford

MORITZ MARBACH ETH Zurich

KONSTANTINOS MATAKOS King’s College London

DIMITRIOS XEFTERIS University of Cyprus

Although Europe has experienced unprecedented numbers of refugee arrivals in recent years, thereexists almost no causal evidence regarding the impact of the refugee crisis on natives’ attitudes, policypreferences, and political engagement. We exploit a natural experiment in the Aegean Sea, where

Greek islands close to the Turkish coast experienced a sudden andmassive increase in refugee arrivals, whilesimilar islands slightly farther away did not. Leveraging a targeted survey of 2,070 island residents anddistance toTurkeyasan instrument,wefind thatdirect exposure to refugeearrivals induces sizableand lastingincreases in natives’ hostility toward refugees, immigrants, and Muslim minorities; support for restrictiveasylumand immigrationpolicies; andpolitical engagement to effect suchexclusionarypolicies. Since refugeesonly passed through these islands, our findings challenge both standard economic and cultural explanationsof anti-immigrant sentiment and show that mere exposure suffices in generating lasting increases in hostility.

INTRODUCTION

Recent years have seen thehighest levels of forceddisplacement, up to 66 million, since the after-math of World War II. Since 2015, more than

three million individuals, predominantly from Muslim-majority countries such as Syria, Afghanistan, and Iraq,

have applied for asylum in Europe alone. Althoughapplications declined in 2016 and 2017 after theimplementation of the EU–TurkeyAgreement, overallnumbers are expected to remain high as asylum seekerscontinue to flee protracted conflicts and persecution.

The dramatic increase in asylum seekers and thechaotic management of refugee flows have dominatednews cycles across Europe, led to social tensions andpolitical conflict, and triggered, at times, even violentpolitical backlash (Bansak, Hainmueller, and Hang-artner 2016; Marbach and Ropers 2018). In manycountries, extreme-right parties have sought to leveragenatives’ anxieties to mobilize voters and enact morerestrictive asylum policies. Several recent studies thatfocusedonAustria (Steinmayr2018), France (Vertier andViskanic 2018), and Greece (Dinas et al. forthcoming)have all corroborated that refugee migration is animportant factor fueling the rise of extreme-right parties.

Beyond its impact on aggregate vote shares forextreme-right parties, little is known, however, abouthow refugee migration shapes the attitudes, policypreferences, and political behavior of European citizens.This lacuna is problematic for our theoretical under-standing of the short- and long-term consequences of therefugee crisis on Europe’s political landscape and themechanism under which contact with refugees alleviatesor catalyzes concerns. Similarly, policymakers who aretaskedwith designing amore cooperative asylum systemneed a deeper understanding of how and when refugeearrivals trigger local opposition and political backlash(Bansak, Hainmueller, and Hangartner 2017).

Why do we know so little about the impact of refugeemigration on natives’ attitudes, preferences, andbehavior? First, in order to go beyond readily availableaggregate data on voting behavior and drill down toindividual-level outcome measures, one has to field asurvey in areas with varying levels of refugee exposure.

Dominik Hangartner, Associate Professor, Center for Comparativeand International Studies, ETH Zurich; Immigration Policy Lab,Stanford University and ETH Zurich; and Department of Govern-ment, London School of Economics, d.hangartner@lse.ac.uk.

Elias Dinas, Swiss Chair in Federalism, Democracy and Interna-tional Governance, Robert Schuman Centre for Advanced StudiesandDepartmentof Social andPolitical Sciences,EuropeanUniversityInstitute; and Associate Professor, Department of Politics andInternational Relations, University of Oxford, elias.dinas@eui.eu.

Moritz Marbach , PostDoc, Center for Comparative and Inter-national Studies, ETHZurich; and Immigration Policy Lab, StanfordUniversity and ETH Zurich, moritz.marbach@gess.ethz.ch.

Konstantinos Matakos, Assistant Professor, Department of PoliticalEconomy, King’s College London, konstantinos.matakos@kcl.ac.uk.

Dimitrios Xefteris, Assistant Professor, Department of Economics,University of Cyprus, xefteris.dimitrios@ucy.ac.cy.

The survey was fielded by the Survey Unit of the University ofMacedonia according to our sampling frame and in line with the ethicspolicy of the London School of Economics and Political Science forhuman subjects research. DH, ED, MM, KM, and DX conceived anddesigned the research;ED,DH, andKMdesigned the survey;ED,KM,MM,DX, andDHhelpedwith data collection;MMandDHperformedstatistical analyses; and DH, MM, KM, and ED wrote the manuscript.We thank Alexandra Dufresne, Marc Helbling, Dan Hopkins, andDavid Laitin for detailed comments. This paper benefited from thequestions and suggestions from numerous seminar and conferenceparticipants. The usual disclaimer applies. ED and KM acknowledgefinancial support from the British Academy for a Small LeverhulmeGrant (project number CTR00340) and DH support from the Lev-erhulme Trust. Replication files are available at theAmerican PoliticalScience Review Dataverse: https://doi.org/10.7910/DVN/XGVQDT.

Received: August 15, 2017; revised: April 5, 2018; accepted:November 6, 2018. First published online: December 27, 2018.

442

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

Second, akeyproblem facedbyall studiesof the impact ofrefugee migration is that refugee flows are far from ran-domly assigned. StartingwithRavenstein’s (1885) “law ofmigration,” scholarshaveexplored themigrant’sdecision-making process that determines whether—and whereto—migrate, and there is ample evidence that a similarcalculus also applies to forcefully displaced people (e.g.,Neumayer 2005). Within the financial and legal con-straints thatasylumseekers face, they tend toflee toplacesthat are feasible to travel to, but alsowhere they are, interalia, welcome. This latent hospitality toward refugees ishard to measure empirically but will typically confoundthe relationship between refugee arrivals and outcomemeasures of hostility (measured after the arrivals tookplace). Ifone ignores this importantconfounderandrunsacross-sectional regression of, say, exclusionary attitudeson refugee arrivals, the estimated correlation might wellbe negative (indicating that more arrivals alleviate hos-tility), even if the true causal effect is positive.

In this paper, we address both issues by fielding atargeted survey of European citizens to measure theimpact of the refugee crisis on their attitudes towardrefugees, immigrants, and Muslim minorities; their pref-erences regarding asylum, immigration, and integrationpolicies; and their political engagement to enact policiesaffecting refugees. By focusing the comparison on similarrespondents in comparable areas, some of whom weredirectly exposed to the refugee crisis and someofwhomatmost indirectly so, we can contribute to the ongoingdebate about the consequences of refugee migration andempirically answer the question whether exposure to therefugee crisis triggers or alleviates natives’ hostility.

We focus on Greece, one of the European countriesmost affectedby the refugee crisis.With its long coastlinealong the EU external border, its proximity to the

Turkish coast, and many difficult-to-patrol islands,Greece quickly became the main entry point to Europefor most Syrian and Afghan refugees. Figure 1 (leftpanel) shows that of the 1.4 million new asylum seekerswho have reachedEuropean territory between 2015 and2016 by passing the Mediterranean Sea, more than onemillion arrived in Greece (UNHCR 2016). Figure 1(right panel) indicates that in Greece, these arrivalshappened within a very short time window: from abaseline of almost zero at the beginning of 2015, thenumber of monthly arrivals peaked at 211,000 at theheight of the crisis in lateOctober 2015. After theMarch2016 Agreement between the European Union andTurkey called for the return of refugees arriving inGreece from Turkey, effectively closing off this migra-tion route, arrivals of asylum seekers in Greece imme-diately decreased to 3,000 per month by the end of 2016.

These boat arrivals not only took place in a short timewindow, but were also highly concentrated amongabout a dozen Greek islands in the Aegean Sea. Whileislands closest to the Turkish coast received up to fiverefugee arrivals per resident between the onset ofthe refugee crisis and the implementation of theEU–Turkey agreement, on islands only slightly fartheraway the number was essentially zero. As we discusslater in detail, we leverage this distance to the Turkishcoast as an instrument for refugee arrivals. This solvesthe aforementioned selection issue by focusing theanalysison theexogenouspartof thevariance in refugeearrivals that is solely driven by proximity to Turkey andignoring the endogenous part that is driven by preex-isting differences in attitudes toward refugees.

Several factors facilitate our empirical strategy. First,distance to theTurkish coast causes dramatic variation inthenumberof refugeearrivals.This allowsus tocompare

FIGURE 1. Mediterranean Refugee Arrivals

Note: Right panel shows the number of sea arrivals across countries at the external border of Europe for 2015 (dark blue) and 2016(light blue). Left panel shows number of monthly refugee arrivals in all of Greece (gray) and on the Aegean islands (dark blue).

Does Exposure to the Refugee Crisis Make Natives More Hostile?

443

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

similar residents on neighboring islands, some of whichreceived hundreds of thousands of passing refugees,while others slightly farther away receivednone. Second,geographical proximity and the fact that they oftenbelong to the same administrative unit ensures thatislands with and without arrivals are identical acrossmany observable and unobservable characteristics, suchas the nature and quality of political competition andregional government, provision of public services(education, healthcare, and the judiciary), and access toEU funds. Third, among those islands with refugeearrivals, refugees were concentrated in particularUNHCR hotspots, allowing us to also leverage within-island variation as a complementary mechanism test.

We find that direct exposure to the refugee crisis hasstatistically and politically meaningful effects on natives’exclusionary attitudes, preferences over asylum andimmigrationpolicies, andpolitical engagement.Exploitingthe exogenous variation in refugee arrivals caused bydistance to the Turkish coast—our instrument—we findthat respondents directly exposed to the refugee crisisexperiencea1/4 standarddeviation(SD) increase intheiranti-asylum seeker and anti-immigrant attitudes as wellas a 1/6 SD increase in their anti-Muslim attitudes.Comparedtorespondentsonunexposed islands, theyaremore likely to oppose hosting additional asylum seekersand to support the ban from school for asylum seekers’children and are less likely to donate to UNHCR and tosign a petition that lobbies the government to providebetter housing for refugees. Across all these outcomes,we find that direct exposure to the refugee crisis has along-term impact on natives’ hostility. In a companionstudy,wealsoexamine theshort-termimpactofexposureto the refugee crisis on voting behavior in this setting(Dinas et al. forthcoming).

Our study makes four contributions. First, it providessome of the first causal evidence on the impact of therefugeecrisisonexclusionaryattitudes, policypreferences,and political engagement of European citizens. Second,rather than focusing on contemporaneous, and potentiallyexacerbated, effects of refugeearrivals, our studyallowsusto look at the long-term consequences, as we conductedour survey in February and March 2017. Ever since theimplementation of the March 2016 Agreement betweenthe European Union and Turkey, refugee flows wereminimal, so that our survey measures the persistence ofpast shocks about a year after the crisis. Third, in Greece,the refugee crisis coincided with economic turmoil andfinancial austerity imposed by the so-called troika con-sisting of the European Commission, the EuropeanCentral Bank, and the IMF, as well as the chaotic man-agement of refugee flows at the local, national, andEuropean level. All these factors are believed to catalyzeanti-immigrant sentiment (Dinas et al. forthcoming).

Fourth, to understand how island residents experi-enced the refugeecrisis, it is important tonote thatalmostall refugees left the islands of first arrival within a veryshort period, typically within twenty-four hours, tocontinuetheir journeyviaAthens totheGreekmainland,and other European countries. The passing presence ofrefugees on these islands allows us to make importantinferences about the drivers of anti-refugee sentiment.

On the one hand, the transient nature of arrivals all buteliminated the chance for sustained interactions betweennatives and refugees. For contact to reduce tensions,scholarshaveproposeda fairlystrict setofconditions thathave to be met, such as friendly acquaintance within acooperative context (rather thanaone-off exchangewitha stranger) (Allport 1954; Pettigrew1998;VanLaar et al.2005). These conditions were clearly not met in thepresent context, and our study therefore identifies the“pure” effect of exposure to the refugee crisis.Hence,wehave little reason to expect that exposure reduced hos-tility among the local population.

On the other hand, because refugees quickly left theislands for other European countries, the usual materi-alist concerns that immigrants competewith natives overscarce resources such as jobs or welfare benefits (BoboandHutchings 1996; Scheve and Slaughter 2001) do alsonot apply in this context. The same is true for ideationalconcerns that immigrants change the culture and identityof the host country (Golder 2003). Based on thesestandard theories of the economic and cultural drivers ofanti-immigrant sentiment, we would not expect thatrefugee arrivals trigger exclusionary reactions.

Our finding that hostility prevails in the Greek Aegeanislands suggests that mere exposure to the chaos of therefugee crisis generates a feeling of threat that can activatelatent predispositions against immigrants and mobilizesupport for exclusionary policies (Sniderman, Hagen-doorn, and Prior 2004). The next section summarizes thepredictions of the most prominent theories of intergroupcontact and conflict for the present context. Afterdescribing the survey sample and measures, the mainresults, and the mechanism tests, we return to furtherexploringthe implicationsofourfindingsforourtheoreticalunderstanding of the drivers of exclusionary reactions.

CONTACT, CONFLICT, AND THREAT IN THECONTEXT OF THE GREEK REFUGEE CRISIS

A vast literature has investigated how proximity to andcontact with immigrants and refugees shape the atti-tudes of natives.While a full review is beyond the scopeof this paper, we focus on the theories and mechanismsmost relevant for the context of the refugee crisis andrefer themorebroadly interested reader to theexcellentreviews of Paluck and Green (2009) and Hainmuellerand Hopkins (2014).

One stream of research originating from Allport’sseminal work (1954) assesses how, and under whichconditions, contact between themajority ingroup and anethnic outgroup can yield increased levels of empathyand understanding (Barlow, Louis, and Hewstone 2009;Berg 2009; Finseraas and Kotsadam 2017; McLaren2003; Pettigrew and Tropp 2006) and decrease dis-criminatory behavior (Scacco and Warren 2018). Thesestudies show that if ingroups and outgroups (i) shareequal status and common goals, (ii) find themselves in acooperative rather than competitive environment, and(iii) operate under a well-defined set of norms or regu-lations, contact can reduce prejudices. And althoughAllport (1954) was the first to point out that these

Dominik Hangartner et al.

444

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

conditions are often hard to meet in reality, contacttheory seems to be robust to violations of some of thesecriteria (Paluck, Green, and Green 2017; Pettigrew andTropp 2006). Recent works on the mechanism under-lying theseeffectspoint totheroleofcontact in increasingknowledgeabout the outgroup, encouraging perspectivetaking, and reducing intergroup anxiety (Barlow et al.2012; Pettigrew and Tropp 2008).

Proximity, however, does not necessarily lead tocontact. One example that illustrates this distinctioncomes from Enos (2014), who randomly assignedphysical proximity to outgroup members by placingSpanish-speaking confederates on the Boston subway,without facilitating face-to-face interaction. The resultwas an increase in exclusionary attitudes among thewhite-majority group. Observational studies thatexamine the correlation between the share of immi-grants and anti-immigrant sentiment have yielded anambiguous picture. Whereas some studies suggest thatethnic diversity is accompanied by lower levels ofintergroup hostility (Sturgis et al. 2014; Zorlu 2017),others find exactly the opposite (Kaufmann and Harris2015; Putnam2007; Schlueter andScheepers 2010). Thelatter results are often interpreted as evidence in favorof the realistic group conflict hypothesis: thepresence ofan outgroup in sufficient numbers generates, accordingto this line of argument, competition for scarceresources (Blalock 1967; Blumer 1958; Bobo 1983;Bobo and Hutchings 1996; Forbes 1997; Hardin 1997).

Closely related to theoriesof realisticgroupconflict isarapidly increasing stream of political science researchthat studies the drivers of anti-immigrant sentiment.Numerous explanations have been suggested, yet theytend to cluster into two broad categories: economicconcerns and cultural concerns (Golder 2003; Hainmu-eller and Hopkins 2014). Economic concerns can resultfromat least three, complementary, sources. Natives canperceive that asylum seekers increase competition in thelabormarketand thusenhancetheriskofunemployment(Mayda 2006; Scheve and Slaughter 2001); alter themeans-testedwelfare distribution, converting them fromnetbeneficiaries tonetcontributors(FacchiniandMayda2009); and lead to congestion effects, defined as theconcern that in the short run, supply of public goods andservices, such as public housing, health, and education,cannot follow the rapid increase in demand caused byimmigration (Cavaille and Ferwerda 2017).

Cultural concerns refer to a more general tendency toperceiveoutgroupnormsandvaluesas incompatible andpotentially in conflict with those of the majority ingroup(Brown 2000; Connor 1994). Religious, ethnic, or lin-guistic differences are often treated as signals of culturaldistance, triggering exclusionary preferences as a way ofprotecting the ingroup’s customs and traditions (Bansak,Hainmueller, and Hangartner 2016; Brader, Valentino,and Suhay 2008; Ivarsflaten 2005). Previous research hasfound that these cultural or identity concerns are oftenmore important forexplainingopposition to immigrationin Western countries than economic concerns (see, e.g.,Hainmueller and Hopkins 2014).

Whatpredictionsdothese theoriesmake inthecontextof theGreek refugee crisis? The fact that asylum seekers

and refugees quickly passed through the Aegean islandsposits a challenge for the contact hypothesis since theconditions under which contact can reduce prejudice areclearly not met in this context. Time for interaction withthe same person wasminimal, and casual contact did notoccur in a cooperative context and happened in theabsence of common norms. Furthermore, the potentialfor friendship was limited since the local population waswell aware that refugees are leaving the island with thenext available ferry to Athens.1

But for the very samereason,wewouldalsonot expectthat economic and cultural concerns play a major role inthis context. Refugees did not enter the labor market onthese islandsanddidnotrequestwelfarebenefitsorapplyfor public housing. The accommodation costs of therefugee camps, where a few of the refugees stayed for awhile before continuing their journey, were mostlyshoulderedby the central government and theEuropeanUnion. For the same reason, perceptions of threat tonational identity should also be minimal, as it soonbecame obvious that virtually all asylum seekers wouldquickly leave the islands.Accordingly, standard theoriesof cultural and economic concerns would predict thatpassingmigrationflowsdonot triggera lasting increase inanti-immigration sentiment.

The massive numbers of refugee arrivals, albeittransient in nature did, however, clearly disrupt thedaily life and routines of the native island population.Numerous reports in the local press highlight theinability of the local and national authorities to effec-tively provide medical support,2 sanitary services,3 andwaste collection4 for the passing refugees.Mountains ofrubbish and open defecation surrounding arrival anddeparture hotspots sparked concerns about the spreadof diseases. Chaotic scenes and the presence of largenumber of staff and volunteers from the UNHCR, theEU, the National Center of Disease Control and Pre-vention, and various NGOs further compounded theseeffects to create the impression of a state of emergency.In fact, the mayor of Lesvos, Spyros Galinos, asked thecentral government to declare a state of emergency akinto that followingmajor natural disasters.Hewrote in his(open) letter to the central government: “the whole

1 A fraction, about 11,000of themore thanonemillionasylumseekersthat arrivedon theAegean islands, is still hosted in refugee campson afew islands. These refugee camps, the largest of which is Lesvos’Moria, are, however, located in the rural countryside, which severelylimits the potential for interaction. Furthermore, as the robustnesstests in Appendix E9 show, our results are robust to excludingrespondents in the proximity of these camps.2 See, e.g., the editorial “The State is late but… volunteers and themunicipality are rushing [to respond to the crisis]” (author trans-lation) in the newspaperEmpros on September 5, 2015. http://www.emprosnet.gr/koinonia/75136-kratos-argei-alla-ethelontes-kai-dimos-trehoyn.3 See, e.g., the editorial “Social Explosion is in sight” (author trans-lation) in the newspaper Dimokratis on September 3, 2015. www.dimokratis.gr/index.php?article52015-9-3_orati_i_koinoniki_ekrixi.4 See, e.g., Marinos Orfanos in the newspaper Empros: “The port [ofMytilene] has become a vast dumpster bin and a space of open-airtoilets” (author translation) on September 2, 2015. http://www.emprosnet.gr/koinonia/75750-zamtrakis-aperanti-homateri-kai-horos-ypaithrion-afodeytirion-limani.

Does Exposure to the Refugee Crisis Make Natives More Hostile?

445

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

situationand the inability [of the central administration]to handle it leaves local communities at the verge ofsocial explosion […] the non-controlled nature of therefugee crisis has already turned this problem into athreat that is proportionate to a large scale naturaldisaster […] the rule of law needs to be reinstated […],public health needs to be safeguarded and the socialorder of the island needs to be maintained, and themigrants need to be treated humanely according to theprovisions of the UN Charter for human rights.”5

The mayor’s letter and the newspaper reports makeclear that this disruption of everyday life was not a merenuisance for the local population,butwasperceivedasanupsetof the socialorderandgenerateda feelingof threat.Recentworkonsocial threat suggests that theperceptionof a threatening environment can activate otherwiselatent predispositions against immigrants and otherminorities (Feldman and Stenner 1997; Hetheringtonand Weiler 2009; Stenner 2005). This process can havetwo complementary consequences: first, the chaos onarrival islands might act as a situational trigger that“galvanizes” (Sniderman, Hagendoorn, and Prior 2004)the attitudes of those already predisposed againstimmigration.Second, perceived threatandanxietymightalsomobilize support for exclusionarypoliciesaboveandbeyond those that already hold hostile attitudes (Brader,Valentino, and Suhay 2008; Sniderman, Hagendoorn,and Prior 2004). Since perceptions of threat and anxietyareemotional triggers, theymakeit likely that individualswill not only change their attitudes or beliefs, but alsotake political action (Brader 2006; Brader, Valentino,and Suhay 2008).

Furthermore, the referenced newspaper articles sug-gest that perceived levels of social threat were highest inthe vicinity of the arrival and departure hotspots. In thelocations where refugees gathered, the failure of theGreek government and the European Union in man-aging the refugee flows was most visible, and their dis-ruptionofnatives’everyday lifemost severe.Hence, akinto Hopkins’ (2010) politicized places hypothesis, weexpect tofindthestrongest increases inoutgrouphostilityfor natives living in close proximity of the hotspots thatexperienced the refugee crisis most intensely.

Taken together, these factors suggest that socialthreat can explain how the chaotic management ofrefugee flows activates exclusionary reactions toimmigrants and ethnic minorities, even in a contextwhere these flows are temporary and transient. In theabsence of contact, as well as economic or culturalconcerns, the mere exposure to the refugee crisis suf-fices to crystallize latent anti-immigrant attitudes andmobilize citizens to support exclusionary policies.

EMPIRICAL STRATEGY

To estimate the causal effect of exposure to the refugeecrisis on natives’ exclusionary attitudes, policy prefer-ences, and political engagement, we leverage the

distance of an island to the Turkish coast as an instru-ment for refugee arrivals. This instrumental variableapproach allows us to resolve the selection bias asso-ciated with refugees choosing their arrival island basedon preexisting levels of hostility toward outgroups.

As shown inFigure 2, islands very close to theTurkishcoast, such as Samos,Chios, andLesvos, receivedmorethan five arrivals per island resident within the14 months between the onset of the crisis in early 2015and implementation of the EU–Turkey agreement inMarch2016.Thenumberof arrivals rapidlydecreases asthe distance to Turkey increases, with islands fartherthan 50 kilometer away not registering any refugees.Asindicated by the blue line, the relationship between percapita refugee arrivals and logged distance can be wellapproximated using linear regression. Table D10 inAppendix provides the estimates from these first-stageregressions of the per capita number of arrivals ondistance and a binary indicator of arrivals on distance,respectively.Thefirst stage for thebinary treatment, ourmain specification, is very strong (F-statistic5 133.75).The reason for this becomes clear in Figure 2: while allbut two islands that are less than 50 kilometers awayfrom Turkey receive refugees, no island more than 50kilometers away register any arrivals. As robustnesstests, we also estimate the first stage using a continuousmeasure of the number of arrivals per capita and the lognumber of arrivals per capita.6 The corresponding F-statistics are 9.22 and 14.20, respectively, around thecritical value proposed by Stock and Yogo (2002).

One frequent concern with instrumental variableanalysis is the credibility of the exclusion restriction. Inour context, a violation of the exclusion restrictionwould mean that islands slightly farther away from theTurkish coast not only register fewer refugee arrivals,but are also different in other, preexisting, character-istics such as the sociodemographic composition orpolitical preferences of the local population. For-tunately, we can conduct a variety of placebo tests thatshow that this is not a concern. Table 1 provides theestimates from two-stage least squares regressions, ourmain specification further discussed below, of the pla-cebo outcomes on a binary indicator for refugeearrivals, instrumented with distance to the Turkishcoast. For all placebo outcomes—vote choice for themain parties Golden Dawn, Nea Dimokratia, andPASOK in the January 2015 election (prior to the onsetof the refugee crisis) and thedemographic indicators forfemale, old versus young (split at the median age), andtertiary education—we find no significant effect. TableD11 inAppendix extends the analysis to smaller partiesthat also ran in the January 2015 elections and a linearspecification for education and age. None of the esti-mated coefficients is significantly different from zero,thereby supporting our confidence in the validity of theinstrument.

5 A copy of the letter is available here: https://tinyurl.com/yc4ruefh.

6 We are grateful to an anonymous reviewer for pointing us to the log-log specification.Hence, this robustness testwasnotprespecified in thepre-analysis plan.

Dominik Hangartner et al.

446

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

SURVEY

In order to assess the impact of refugee arrivals in theAegean Sea, our sampling frame consists of allinhabited Greek islands. Between February 22 andMarch 7, 2017, we successfully interviewed 2,070respondents, split equally between treated and controlislands and proportional to the population of eachisland. All interviews were conducted using computer-assisted telephone interview (CATI) by calling landline

numbers, which were randomly selected from the tel-ephoneregistry.7Withineachhousehold, the respondentwas selectedusing the next-birthdaymethod, an efficientprocedure for selecting a sample that is representative of

FIGURE 2. Visualization of the First-Stage Estimates

Note: The blue line indicates the OLS regression of the cumulative number of refugee arrivals per island resident on the logged distance tothe Turkish coast. The red line shows the analogous scatterplot smoother. The size of the dots is proportional to the number of surveyrespondents.

TABLE 1. Placebo Outcomes

GD S/A ND PASOK Female Age Education

Treatment 0.00 20.06 0.04 0.02 20.03 20.01 20.03(0.008) (0.029) (0.024) (0.013) (0.030) (0.032) (0.024)

Constant 0.02*** 0.39*** 0.21*** 0.07*** 0.57*** 0.50*** 0.47***(0.006) (0.021) (0.020) (0.010) (0.021) (0.024) (0.020)

N 2,046 2,046 2,046 2,046 2,046 2,046 2,046

Note: 2SLS regressionestimates (with cluster-robust standard errors in parentheses) of placebooutcomesof refugee arrivals, instrumentedwith the island’s distance to the Turkish coast. Outcomes are vote choice for: Golden Dawn (GD), SYRIZA/ANEL (S/A), Nea Dimokratia(ND), andPASOK in the January 2015election; demographic indicators for female, older thanmedianageof 50, and tertiary education.* p , 0.05; ** p , 0.01; *** p , 0.001.

7 The survey was fielded by the Survey Unit of the University ofMacedonia according to our sampling frame and in linewith the ethicspolicy of the London School of Economics and Political Science forhuman subjects research.

Does Exposure to the Refugee Crisis Make Natives More Hostile?

447

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

all household members (Salmon and Nichols 1983). Thecumulative response rate (RR3) as defined by theAmerican Association for Public Opinion Research is8%, which is about the same as the response rate forCATI interviews in the USA (Keeter et al. 2017). Onepotential concern is that thewillingness tobe interviewedis correlated with distance to the Turkish coast, whichmight confound the analysis. However, this is not thecase: the probability of being interviewed is very similarfor residents living close and farther away from theTurkish coast, and Welch’s t-test cannot reject the nullhypothesis of no difference at conventional levels (p .0.094, two-tailed).

Themainanalysisusingall adult respondentspresentedbelow uses survey weights that are proportional to thepopulation size of the island under the constraint that therespondents of treated and control islands receive half ofthe aggregate weight each. Item nonresponse was gen-erally low (on average 4.9% across all measures) but inorder to remove any imbalances and increase the effi-ciency of our estimates, we use multiple imputation toaddress missing answers. The robustness section belowshows that all results also hold with alternative weightingstrategies, without weighting, and list-wise deletion.Furtherdetailson thesampling frame, surveyweights,andmultiple imputation are provided inAppendix SectionA.

Outcome Measures

For the outcomes, we employ standard indicators forattitudes toward outgroup minorities, policy prefer-ences concerning refugee and voluntary migration, andquasi-behavioralmeasures of political engagement.Weuse a series of feeling thermometer questions aboutvarious outgroups (relative to the majority group) asexplicit measures of racial attitudes. Past research hasshown that they are highly correlated with behavioralmeasures of discrimination (McConnell and Leibold2001). The additional attitudinal questions are largelybased on the Global Attitudes Survey (Wike, Stokes,and Simmons 2016), but answers are recorded on a five-point scale to decrease acquiescence response bias(Revilla, Saris, and Krosnick 2014). The questionseliciting policy preferences are adapted from Bansak,Hainmueller, and Hangartner (2017) and tailoredtoward the Greek case. Lastly, the quasi-behavioraloutcomes attach real-life consequences to therespondents’ answers to questions about politicalpreferences and engagement in the refugee context(Brader, Valentino, and Suhay 2008).

Thefirst setofoutcomesmeasuresanti-asylumseekerpolicy preferences and attitudes.Weasked respondentswhether they agreed that refugee children should bebanned from schools (from 1 to 5), an issue that washotly debated at the time of the survey; that Greeceshould decrease the number of people towhom it grantsasylum (1–5); and that asylum seekers aremore likely tocommit terrorist attacks (1–5) and crimes (1–5) and area burden on the country (1–5).

The second set of outcomes measures more generalanti-immigrant policy preferences and attitudes. Weasked respondents if they support a decrease in the

numberof economicmigrants (from1 to 5) and strongerborder protection (1–5). We also measured the differ-ence in their feeling thermometer scores (0–100)between Greek Muslims and Muslim immigrants, andGreek Christians and Christian immigrants.

The third set of outcomes provides quasi-behavioralmeasures of respondents’ willingness to politicallyengage to change asylum policies and donate to supportrefugees. We asked respondents whether we shouldnotify their Member of Parliament (MP) on their behalfthat they support a decrease (or increase) of the numberof people to whom Greece grants asylum (from 1 to 5);whether (0/1) and how much (0–100) they wanted todonate to UNHCR if they won the EUR 100 raffle thatwas part of the survey; and whether they would like tosign an existing petition that lobbies the government toprovide better housing for refugees (0/1).

The fourth set of outcomes measures general anti-Muslim policy preferences and attitudes. We askedrespondents whether they believed that the politicalrepresentation of the Greek Muslim minority in theGreek parliament should be reduced, remain the same,or be increased (1–5). We also measured the differencein their feeling thermometer scores (0–100) betweenChristian immigrants andMuslim immigrants andaskedwhether they believe that most Muslims are not willingto integrate (1–5) and are likely to support extremistgroups (1–5).8 The question wording and outcomecoding are reported in Appendix Section A.3 and A.4.

Lastly, for eachof the four sets of outcomes,webuilt asummary scale that combines the different measures byextracting the first component of a polychoric principalcomponent analysis (PCA) (Olsson 1979). AppendixSectionA.3 provides details on the PCAand shows thatfor each scale, thefirst component explains a large shareof the total variance. Averaging responses acrossmultiple items measuring the same latent construct notonly reduces bias and variance from random meas-urement error that is common in survey research(Hainmueller,Hangartner, andPietrantuono2017),butalso alleviates worries about false-positive significancetests by focusing the analysis on a few main outcomes.

Treatment and Instrument

To measure refugee arrival, our treatment variable, weuse island-level registry data collected by UNHCR(2016). Figure 3 shows the strong clustering of arrivalson islands close to the Turkish coast. We code a binaryindicator equal to one if UNHCR recorded a positivenumber of arrivals and 0 otherwise. As robustness tests,we also code two continuous measures based on thecumulative number of arrivals between January 2015and March 2016 as a fraction of the number of islandresidents and the log (11) of this ratio, respectively.

8 We also included a battery of questions that measure general atti-tudes toward outgroups with questions such as: “For being trulyGreek, it is important to (i) havebeenborn inGreece; (ii) beingable tospeak Greek; (iii) being Christian Orthodox.” We do not find anyeffect of refugee arrivals on these outcomes. The results are shown inAppendix Section D.

Dominik Hangartner et al.

448

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

Our instrument is the logarithm of the shortest path(henceforth distance) between Turkish coast and thepopulation center of amunicipality (typically an island),which we calculate using Google Maps. This distancemeasurecorrelateshighly (r50.997)withanalternativedistance measure based on the polygon’s centroid of amunicipality in Greece and the closest point to theTurkish polygon in GIS software.

Statistical Analysis

To estimate the effect of exposure to the refugee crisis,wefit two-stage least squares (2SLS)models inwhichweregress the outcome on our binary exposure indicatorcontrolling for respondent demographic characteristics.We instrument the exposure dummy with a continuousinstrument as described above. For all analyses, wecluster standard errors by municipality, of which thereare 92 in ourmain specification. The standard errors arevirtually identical if we cluster at the island level (seeAppendix Figure E7).

The sampling frame, questionnaire, coding rules forcovariates and outcomes (including the summary scalesextracted from the PCA), and statistical analyses were

all prespecified in a pre-analysis plan posted to thePolitical Science Registered Studies Dataverse beforewe conducted the survey and can be retrieved at doi:10.7910/DVN/FFEUEH.

RESULTS

In Figure 4, we present the effect estimates for theinstrumented treatment. Tables D4–D9 in Appendixreport the full regression tables. The black estimatesshow the impact of exposure to the refugee crisis onthe first principal components for each summarymeasure. Exploiting exogenous variation in refugeearrivals, we find that respondents on directly exposedislands experience an increase of about 0.24 SD intheir hostility toward asylum seekers (p, 0.001, two-tailed), 0.23 SD in their hostility toward immigrants (p5 0.001), and 0.13 SD in their hostility toward Mus-lims (p 5 0.041). Finally, we find that immediateexposure has a strong impact on our summarymeasure of the quasi-behavioral outcomes andincreases anti-refugee political engagement by 0.33SD (p , 0.001).

FIGURE 3. Map of the Aegean Sea

Does Exposure to the Refugee Crisis Make Natives More Hostile?

449

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

Looking at the outcomes separately, we see consistentchanges in policy preferences (blue estimates) and atti-tudes (greenestimates) and that the formerare generallylarger. With regard to anti-asylum seeker preferences,we find that respondents on exposed islands are morelikely to support banning refugee children from schools(point estimate: 0.41units onafivepoint scale,p, 0.001)and decreasing the number of people to whom Greecegrants asylum(0.26,p, 0.001). Thesepolicy preferencesare particularly remarkable because refugees onlypassed through the islands and were not sending theirchildren toAegeanschools.Wealsofindthatexposure tothe refugee crisis leads to an increase in anti-asylumseeker attitudes: exposed respondents believe that asy-lum seekers are more likely to commit terrorist attacks(0.25, p5 0.002) and crimes (0.14, p5 0.102) and are aburden on Greece (0.12, p 5 0.097).

With regard to general anti-immigrant policy prefer-ences,wefindthat respondentsonexposed islandspreferfewer economic migrants (0.24, p, 0.001) and strongerborder protection (0.08, p 5 0.0126). Turning to thefeeling thermometer scores, we see that exposedrespondents prefer Greek Muslims over Muslim immi-grants (0.09, p 5 0.015) but find no effect on the dif-ference between Christian Greek natives and Christianimmigrants (20.002, p 5 0.956).

Looking at anti-Muslim policy preferences, we findthat exposure to the refugee crisis significantly weakenssupport for the political representation of the Muslimminority in theGreek parliament (0.10, p5 0.043). Thiseffect on domestic policy is remarkable as Greece hashad a sizable Muslim minority since the OttomanEmpire, and the issue of adequate parliamentary rep-resentation of Greek Muslims is unrelated to the

FIGURE 4. 2SLS Regression Estimates of the Impact of Refugee Arrivals

Note: 2SLS regression estimates (with 95% confidence intervals based on cluster-robust standard errors) of the impact of refugee arrivals,instrumented with the island’ s distance to the Turkish coast, on respondents’ attitudes (green), policy preferences (blue), and PCA-basedsummary measure (black).

Dominik Hangartner et al.

450

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

refugee issue. While respondents on arrival islandsmore strongly prefer Christian overMuslim immigrants(0.12, p 5 0.014), we do not find any differences inrespondents’ beliefs about the willingness of Muslimsliving inGreece toadoptGreek culture (0.12, p50.081)or support extremist groups (0.09, p 5 0.271).

Turning to the quasi-behavioral outcomes, we findthat exposure to the refugee crisis increases politicalengagement in support of more restrictive asylumpolicies. Confirming the results based on stated pref-erences discussed above, residents on exposed islandsaremore likely to lobby theirMP to grant fewer asylumrequests (0.24 unit increase on five point scale, p ,0.001). They are also 14 percentage points less likely todonate to the UNHCR (p , 0.001) and donate onaverageEUR11 less (p, 0.001). Lastly, these residentsare 7 percentage points less likely to sign a petition thatlobbies the national Greek government to providebetter accommodation for asylum seekers (p 5 0.005).

Most of these effects are not only statistically significantbut also politically important. To benchmark the results,we can compare the effect size for the outcome thatmeasures support for decreasing the number of grantedasylum requests to a recently conducted survey coveringfifteen European countries (Bansak, Hainmueller, andHangartner 2016) that asked the same question.Appendix Figure D1 shows that the 0.26 unit increaseresulting fromdirectexposure to the refugeecrisis is aboutequivalent to moving from the second most liberal(Norway) to the second most restrictive country (Hun-gary) of the fifteen surveyed countries.

Robustness

One concern is that there might be unobserved charac-teristics thatdifferbetween islandscloserandfartherawayfrom the Turkish coast that are also correlated withpreferences over asylumpolicies, thereby invalidating theignorability assumption for our instrument.We conduct aseries of tests to show that this concern is unwarranted.

First, we control for respondents’ vote choice in thelast election prior to the refugee crisis in January 2015.With thepresenceof thefiercely anti-immigrantGoldenDawn, we can expect that support for different parties iscorrelated with anti-immigrant and refugee attitudesand therefore a good proxy for preexisting differencesamong islands that are closer and farther away fromTurkey. Figure E2 inAppendix shows that this concernis unfounded. Adding binary indicators for voting forone of the parties running for office (plus abstention) ascovariates has little effect on our estimates. These nulleffects further corroborate the validity of the exclusionrestriction discussed in the previous section. Alter-natively, we can also use municipal-level vote sharesfrom the January 2015 election to adjust for preexistingdifferences in political preferences. Appendix FigureE3 shows that all results remain unchanged.

As a second check, we constrain the sample to islandsthat are relatively close to the Turkish border. If there isanunobserved confounder correlatedwith distance, theeffect estimates for the subsample of islands within a255-kilometer (the midpoint of our distance variable)

perimeter of Turkey should be different from theestimates from the full sample. Figure E4 in Appendixshows that this is not the case and that the subsampleestimates are very similar.

Third, we re-estimate all IV regressions but use thecontinuous measures of refugee arrivals per islandresidents instead of the binary indicator. Figure E5 inAppendix shows the results, which follow the samepattern that we have observed for the binary arrivalindicator. This finding not only confirms the robustnessof our main results, but also reveals that the intensity ofarrivals, i.e., the intensive margin, matters. While thebinary estimates show that the arrival of refugeestriggers hostile reactions among natives, the continuoustreatment estimates indicate that their relative numberalso matters: for 10 additional refugee arrivals perresident, the anti-refugee, anti-immigrant, and anti-Muslim preferences increase by 0.31 SD (p , 0.001),0.29 SD (p , 0.001), and 0.17 SD (p 5 0.008), respec-tively. The results for log arrivals are very similar, interms of both effect size and significance.

Lastly, in order to gauge the impact of the weightingand multiple imputation procedure on the results, weestimate ourmain regression with weights proportionalto the number of voters of the island (independent oftreatment status); with weights provided by the surveycompany balancing the sample by gender, age, andpopulation density; with equal weights per respondent;andwith andwithoutmultiple imputation.AsFigureE6inAppendix shows, all estimates remain robust to thosealternative specifications.

Mechanisms

This section explores different mechanisms that predictwhen, and for whom, exposure to the refugee crisistriggers exclusionary reactions. First, we considerwhether respondentswho receive their incomeprimarilyfrom tourism reactmore strongly to refugee arrivals, dueto material concerns about the impact on their revenue.To test for thismechanism,wereplicate themainanalysisbut subset the sample to respondents whose incomestems exclusively, partially, or not at all from tourism.FigureE11 inAppendix shows the results.While there issome heterogeneity in treatment effects for respondentsdepending on their main source of income, there is noconsistent pattern and most of these differences are notstatistically significant. This finding further supports theargument that not egocentric economic concerns, butrather the chaotic management of the refugee crisis,triggered hostility among the local population.

Second, we test whether the impact of the refugeecrisis is strongest in the vicinity of the hotspots, whereperceived threat arguably reached its highest levels. Toanswer this question, we leverage a complementaryempirical strategy that focuses on islands that registeredarrivals and exploit the within-island variation of thedistance between respondents’ township and the closestrefugee hotspot. As hotspots, we code (UNHCR) ref-ugee camps and the ports from which refugees con-tinued their journeys via ferries to Athens and beyond.Figure 5 shows a map of these hotspots.

Does Exposure to the Refugee Crisis Make Natives More Hostile?

451

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

Figure 6 presents the results from the within-islandanalysis. We run a series of OLS regressions with fixedeffects for each municipality, where we use the sameoutcomes but respondents’ distance to the hotspot asthe treatment variable. We specify three differentfunctions for the treatment: a binary indicator for livingwithin a 10-kilometer radius of a hotspot, a continuousmeasure of distance in kilometers, and logged distance.

Despite the difference in the empirical strategy and thefocus on highly local variation in exposure to the refugeecrisis, where spillover effects on island residents slightlyfartherawayare likely, thefindingsareremarkablysimilarto our main estimates identified from between-islandvariation. Proximity to refugee hotspots significantlyand substantially increases respondents’ hostility towardasylum seekers and economic immigrants and alsoincreases their support for exclusionary policies. Thispattern is consistentwith the idea that the highly localizedandspatiallyconcentratedchaos surroundingthehotspotscreated a feeling of threat and triggered exclusionaryreactions among natives.

A third mechanism relevant for our theoretical under-standingof thedriversofanti-immigrantattitudesconcerns

“galvanization effects” (Sniderman, Hagendoorn, andPrior 2004), i.e., the idea that situational triggers dis-proportionally affect the reactions of those who arealready predisposed to feeling threatened. If so, wewould expect that those with the highest level of pre-existing anti-immigrant preferences are most stronglyaffected by the perceived threat induced by the refugeecrisis (Hetherington andWeiler 2009).Alternatively, wemight believe that instead of further hardening theattitudes of already hostile citizens, threat might inducecitizens with relatively positive attitudes toward out-groups and, absent the crisis, low levels of threat,becoming more hostile (Stenner 2005).

In order to discriminate between these two com-peting hypotheses, we run a series of quantile instru-mental variable regressions (Chernozhukov andHansen 2005), which enable us to estimate the treat-ment effect at different deciles of the outcome dis-tribution. Figure E10 in Appendix presents the effectestimates for the four principal components. The gen-eral pattern is that the treatment effects are fairlyconstant across the different deciles,meaning that thosewith less exclusionary attitudes react similarly negative

FIGURE 5. Location of Hotspots Based on Data From the UNHCR

Dominik Hangartner et al.

452

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

to the refugee crisis as those with already strongexclusionary attitudes. While some of the estimates areslightly larger for respondents with more hostile views,these differences are not always statistically significant.

We also perform an additional test that splits thesample into voters who supported either right-wing(New Democracy) or extreme-right (Golden Dawn)parties versus not voting for one of these two parties inJanuary 2015, the last pretreatment election. If theimpact of the crisis amplifies partisan differences, wewould expect that the effect ismore pronounced amongright-wing or extreme-right voters (Gravelle 2016,2018). If, however, exposuremutes partisandifferences,we would expect larger effects among voters for leftistand centrist parties (Branton et al. 2007).9 Figure E8 inAppendix shows again that the effects are fairly similarbetween right-wing/extreme-right and centrist/leftistvoters. A Welch’s t-test indicates that for all fourcomponents, the differences between the subgroups are

not statistically significant (p 5 0.270, p 5 0.261, p 50.639, p 5 0.981). Together, the two analyses seem toprovide little evidence for either hypothesis, suggestinginstead that the exposure effects spread quite uniformlyacross the ideological spectrum.

CONCLUSION

This study exploits a natural experiment in the AegeanSea to examine the impact of the refugee crisis on theattitudes,policypreferences,andpolitical engagementofnatives. Using distance to the Turkish coast as aninstrument for exposure to the refugee crisis,wefind thatresidents of islands that experience large and suddeninfluxes of refugees becomemore hostile toward asylumseekers, immigrants, andMuslims, and aremore likely tosupport and lobby for more restrictive asylum policiesthan natives in similar islands that receive fewer or noasylumseekers.Substituting thebinarywithacontinuousarrival measure, we find that exposure to the refugeecrisis impacts attitudes at both extensive and intensivemargins, i.e., that natives’ hostility to refugee andimmigrant outgroups grows proportionally to the num-ber of arrivals. These findings are further supported by

FIGURE 6. Within-Island OLS Estimates (with 95% Confidence Intervals Based on Cluster-RobustStandard Errors) of the Effect of Distance to the Hotspot on Natives’ Attitudes

Note: Thefirstmodelusesabinary indicatorequal toone if the respondent lives less than10kilometer fromahotspot, thesecondmodelusesacontinuousmeasure of distance in kilometers and the third model the log of continuousmeasure. Estimates are based on imputed data andweighted by the number of voters and the municipality’s treatment status. Standard errors are clustered by township (Ncluster 5 125).

9 This test was suggested by an anonymous reviewer and was notspecified in the pre-analysis plan. The complement to voting for theNew Democracy or Golden Dawn is voting for SYRIZA, ANEL,PASOK, KKE, ANTARSYA, The River, Democrats’ and Socialists’Movement,EK, any other party not listed in the survey questionnaire,and people who did not vote.

Does Exposure to the Refugee Crisis Make Natives More Hostile?

453

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

complementarywithin-island analyses, which reveal thatmost of the political backlash against refugee arrivals isspatially concentrated. These effects are large enough tobe politically significant. For example,with respect to thequestionofwhether the government shoulddecrease thenumber of people to whom it grants asylum, the differ-ences in opinion amongGreek residents on exposed andnonexposed islands are as big as the differences inopinionamongcitizensof very restrictiveandvery liberalcountries in Europe.

While themassive refugee arrival onGreek islands isan extreme case, people in many places have beenexposed to past and ongoing refugee crises in similarways. Since refugees, unlike othermigrants, rarely havethe ability to travel by plane, they typically pass throughseveral countries until they reach their destination. Forexample, in the context of the current refugee crisis,millions of citizens from transit countries observedAfghan and Syrian refugees traveling to WesternEuropevia Jordan,LebanonandTurkey toGreece, andfrom there along theBalkan route via FYRMacedonia,Bosnia andCroatia or Serbia andHungary toAustria.10

Natives living along this route have experienced thetransit of refugees and how it disturbed everyday life invery similar ways as the residents of theAegean islands.We therefore expect that our findings from Greecewould also apply to theseEuropean countries, aswell asseveral Northern African transit countries.

We believe that our findings have important impli-cations forour theoretical understandingof thedriversofanti-refugee and anti-immigrant preferences. First, ourstudy shows that in cases in which refugee arrivals aremassive but also transient, they spark natives’ hostilityacrossawiderangeofoutcomes.Refugees left thearrivalislands typicallywithin24hours,which severely limits thepotential formeaningful interactionbetweennatives andrefugees. In Allport’s (1954) words, the best we couldhopefor in thepresentcontext is“casualcontact,”butnot“acquaintance.” Hence, we would not expect that con-ditions for the contacthypothesis to alleviatehostility aremet here (Pettigrew 1998). For the same reason, how-ever, we also do not expect that standard theories ofeconomic or cultural concerns would apply (e.g., Golder2003; Scheve, andSlaughter 2001), since refugeesdid notstay long enough on arrival islands to threaten theeconomic, political, or cultural prerogatives of the nativepopulation (Blalock 1967).

Even if the transient refugee arrivals did not threatenprerogatives of the local population, they did disrupttheir lives. The inability of the local and Europeanauthorities to effectively manage the refugee flows andprovide medical support and sanitary services causedchaotic scenes at the hotspots and sparked concernsabout the spread of diseases. Our findings of a uniformeffect of exposure to the refugee crisis across the samplesuggest that this threat triggered exclusionary reactionsnot only among those already predisposed againstimmigration, but also among respondents who

otherwisewould exhibit inclusionary attitudes andhavenot voted for (extreme) right-wing parties in the past.

Second, the fact that almost twelve months elapsedbetween the passing of the last refugee in March 2016and the fielding of our survey in 2017 suggests that theexclusionary reactions documented here are not short-lived or an exacerbated reaction to a contemporaneoustrigger, but rather entrenched and persistent. Hence,this finding is difficult to reconcile with the idea thatrepeated exposure to an outgroup alone, i.e., withoutmeaningful contact, is sufficient to mitigate initialnegative reactions (Enos 2014).

Third, we find that exposure to large numbers ofasylum seekers causes natives to become more hostilenot only toward refugees, but also toward economicmigrants and Muslims, including native Muslims whohave been residing in Greece for centuries. In times ofenduringly high levels of refugee migration, this findingof ahostile spillovereffect fromoneoutgroup toanother,unrelated minority is concerning from a normativeperspectiveandan important avenue for future research.

SUPPLEMENTARY MATERIAL

To view supplementary material for this article, pleasevisit https://doi.org/10.1017/S0003055418000813.

Replication materials can be found on Dataverse at:https://doi.org/10.7910/DVN/XGVQDT.

REFERENCES

Allport, Gordon W. 1954. The Nature of Prejudice. Cambridge, MA:Addison-Wesley.

Bansak, Kirk, Jens Hainmueller, and Dominik Hangartner. 2016.“How Economic, Humanitarian, and Religious Concerns ShapeEuropean Attitudes toward Asylum Seekers.” Science 354 (6309):217–22.

Bansak, Kirk, Jens Hainmueller, and Dominik Hangartner. 2017.“Europeans Support a Proportional Allocation of Asylum Seek-ers.” Nature Human Behaviour 1 (7): 0133.

Barlow, Fiona Kate, Stefania Paolini, Anne Pedersen, Matthew J.Hornsey, Helena R. M. Radke, Jake Harwood, Mark Rubin, andChris G. Sibley. 2012. “The Contact Caveat: Negative ContactPredicts Increased Prejudice More than Positive Contact PredictsReduced Prejudice.”Personality and Social Psychology Bulletin 38(12): 1629–43.

Barlow, Fiona Kate, Winnifred R. Louis, and Miles Hewstone. 2009.“Rejected! Cognitions of Rejection and Intergroup Anxiety asMediatorsof the Impact ofCross-GroupFriendshipsonPrejudice.”British Journal of Social Psychology 48 (3): 389–405.

Berg, Justin Allen. 2009. “Core Networks and Whites? Attitudestoward Immigrants and Immigration Policy.” Public OpinionQuarterly 73 (1): 7–31.

Blalock, Hubert M. 1967. Toward a Theory of Minority-GroupRelations. New York: Wiley.

Blumer, Herbert. 1958. “Race Prejudice as a Sense of Group Posi-tion.” Pacific Sociological Review 1 (1): 3–7.

Bobo, Lawrence. 1983. “Whites’ Opposition to Busing: SymbolicRacism or Realistic Group Conflict?” Journal of Personality andSocial Psychology 45 (6): 1196–210.

Bobo, Lawrence, and Vincent L. Hutchings. 1996. “Perceptions ofRacial Group Competition: Extending Blumer’s Theory of GroupPosition to a Multiracial Social Context.” American SociologicalReview 61 (6): 951–72.

10 See Steinmayr (2018) for a study on how passing (and settled)refugees affects voting behavior in Austrian border municipalities.

Dominik Hangartner et al.

454

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13

Brader, Ted. 2006. Campaigning for Hearts and Minds: How Emo-tional Appeals in Political Ads Work. Chicago, IL: University ofChicago Press.

Brader, Ted, Nicholas A. Valentino, and Elizabeth Suhay. 2008.“WhatTriggersPublicOpposition to Immigration?Anxiety,GroupCues, and Immigration Threat.” American Journal of PoliticalScience 52 (4): 959–78.

Branton, Regina, Gavin Dillingham, Johanna Dunaway, and BethMiller. 2007. “Anglo Voting on Nativist Ballot Initiatives: ThePartisan Impact of Spatial Proximity to the US-Mexico Border.”Social Science Quarterly 88 (3): 882–97.

Brown, Rupert. 2000. “Social Identity Theory: Past Achievements,Current Problems and Future Challenges.” European Journal ofSocial Psychology 30 (6): 745–78.

Cavaille, Charlotte and Jeremy Ferwerda. 2018. “HowDistributionalConflict over In-Kind Benefits Generates Support for Anti-Immigrant Parties.” Working Paper.

Chernozhukov,Victor, andChristianHansen. 2005. “An IVModel ofQuantile Treatment Effects.” Econometrica 73 (1): 245–61.

Connor, Walker. 1994. Ethnonationalism: The Quest for Under-standing. Princeton: Princeton University Press.

Dinas, Elias,KonstantinosMatakos,DimitriosXefteris, andDominikHangartner. Forthcoming. “Waking up the Golden Dawn: DoesExposure to the Refugee Crisis Increase Support for Extreme-Right Parties?” Political Analysis.

Enos, Ryan D. 2014. “Causal Effect of Intergroup Contact onExclusionary Attitudes.” Proceedings of the National Academy ofSciences 111 (10): 3699–704.

Facchini, Giovanni, and Anna Maria Mayda. 2009. “Does the WelfareStateAffect IndividualAttitudes toward Immigrants?EvidenceacrossCountries.” The Review of Economics and Statistics 91 (2): 295–314.

Feldman, Stanley, and Karen Stenner. 1997. “Perceived Threat andAuthoritarianism.” Political Psychology 18 (4): 741–70.

Finseraas, Henning, and Andreas Kotsadam. 2017. “Does PersonalContactwithEthnicMinoritiesAffectAnti-ImmigrantSentiments?Evidence from a Field Experiment.” European Journal of PoliticalResearch 56 (3): 703–22.

Forbes, HughDonald. 1997.Ethnic Conflict: Commerce, Culture, andthe Contact Hypothesis. New Haven Yale University Press.

Golder,Matt. 2003. “ExplainingVariation in theSuccess ofExtremeRightParties inWesternEurope.”ComparativePoliticalStudies36(4):432–66.

Gravelle, Timothy B. 2016. “Party Identification, Contact, Contexts,and Public Attitudes toward Illegal Immigration.” Public OpinionQuarterly 80 (1): 1–25.

Gravelle, Timothy B. 2018. “Partisanship, Local Context, GroupThreat, andCanadianAttitudes towards Immigration andRefugeePolicy.” Migration Studies 6 (3): 448–67.

Hainmueller, Jens, and Daniel J. Hopkins. 2014. “Public Attitudestoward Immigration.”Annual Review of Political Science 17: 225–49.

Hainmueller, Jens, Dominik Hangartner, and Giuseppe Pie-trantuono. 2017. “Catalyst or Crown:DoesNaturalization Promotethe Long-Term Social Integration of Immigrants?” AmericanPolitical Science Review 111 (2): 256–76.

Hardin, Russell. 1997. One for All: The Logic of Group Conflict.Princeton, NJ: Princeton University Press.

Hetherington,Marc J., & JonathanD.Weiler. 2009.Authoritarianismand Polarization in American Politics. Cambridge: CambridgeUniversity Press.

Hopkins, Daniel J. 2010. “Politicized Places: Explaining Where andWhen Immigrants Provoke Local Opposition.” American PoliticalScience Review 104 (1): 40–60.

Ivarsflaten,Elisabeth. 2005. “ThreatenedbyDiversity:WhyRestrictiveAsylum and Immigration Policies Appeal to Western Europeans.”Journal of Elections, Public Opinion & Parties 15 (1): 21–45.

Kaufmann, Eric, and Gareth Harris. 2015. “‘White Flight’ or PositiveContact? Local Diversity andAttitudes to Immigration in Britain.”Comparative Political Studies 48 (12): 1563–90.

Keeter,Scott,NickHatley,CourtneyKennedy,andArnoldLau. 2017.“What Low Response Rates Mean for Telephone Surveys.”Working Paper, Pew Research Center.

Marbach, Moritz, and Guido Ropers. 2018. “Not in My Backyard: DoIncreases in ImmigrationCausePoliticalViolence?”WorkingPaper.

Mayda, Anna Maria. 2006. “Who Is against Immigration? A Cross-Country Investigation of IndividualAttitudes toward Immigrants.”The Review of Economics and Statistics 88 (3): 510–30.

McConnell,AllenR., andJillM.Leibold. 2001.“RelationsAmong theImplicit Association Test, Discriminatory Behavior, and ExplicitMeasures of Racial Attitudes.” Journal of Experimental SocialPsychology 37 (5): 435–42.

McLaren, Lauren M. 2003. “Anti-Immigrant Prejudice in Europe:Contact, Threat Perception, and Preferences for the Exclusion ofMigrants.” Social Forces 81 (3): 909–36.

Neumayer, Eric. 2005. “Bogus Refugees? The Determinants ofAsylum Migration to Western Europe.” International StudiesQuarterly 49 (3): 389–409.

Olsson, Ulf. 1979. “Maximum Likelihood Estimation of the Poly-choric Correlation Coefficient.” Psychometrika 44 (4): 443–60.

Paluck, Elizabeth Levy, and Donald P. Green. 2009. “PrejudiceReduction: What Works? A Review and Assessment of Researchand Practice.” Annual Review of Psychology 60: 339–67.

Paluck,ElizabethL., SethA.Green, andDonGreen. 2018.“TheContactHypothesis Re-evaluated.” Behavioral Public Policy, Forthcoming.

Pettigrew, Thomas F. 1998. “Intergroup Contact Theory.” AnnualReview of Psychology 49 (1): 65–85.

Pettigrew, Thomas F., and Linda R. Tropp. 2006. “A Meta-AnalyticTest of Intergroup Contact Theory.” Journal of Personality andSocial Psychology 90 (5): 751.

Pettigrew, Thomas F., and Linda R. Tropp. 2008. “How Does Inter-group Contact Reduce Prejudice? Meta-Analytic Tests of ThreeMediators.” European Journal of Social Psychology 38 (6): 922–34.

Putnam, Robert D. 2007. “E Pluribus Unum: Diversity and Com-munity in the Twenty-First Century the 2006 Johan Skytte PrizeLecture.” Scandinavian Political Studies 30 (2): 137–74.

Ravenstein, Ernst Georg. 1885. “The Laws of Migration.” Journal ofthe Statistical Society of London 48 (2): 167–235.

Revilla, Melanie A, Willem E. Saris, and Jon A. Krosnick. 2014.“Choosing the Number of Categories in Agree–Disagree Scales.”Sociological Methods & Research 43 (1): 73–97.

Salmon, Charles T., and John Spicer Nichols. 1983. “The Next-Birthday Method of Respondent Selection.” Public OpinionQuarterly 47 (2): 270–6.

Scacco, Alexandra, and Shana S. Warren. 2018. “Can Social ContactReduce Prejudice and Discrimination? Evidence from a FieldExperiment in Nigeria.”American Political Science Review 112 (3):654–77.

Scheve, Kenneth F., and Matthew J. Slaughter. 2001. “Labor MarketCompetition and Individual Preferences over Immigration Policy.”Review of Economics and Statistics 83 (1): 133–45.

Schlueter, Elmar, and Peer Scheepers. 2010. “The Relationshipbetween Outgroup Size and Anti-Outgroup Attitudes: A Theo-retical Synthesis and Empirical Test of Group Threat -and Inter-group Contact Theory.” Social Science Research 39 (2): 285–95.

Sniderman, Paul M., Louk Hagendoorn, and Markus Prior. 2004.“Predisposing Factors and Situational Triggers: ExclusionaryReactions to Immigrant Minorities.” American Political ScienceReview 98 (1): 35–49.

Steinmayr, Andreas. 2018. “Contact Matters: Exposure to Refugeesand Voting for the Far-Right.” Working Paper.

Stenner, Karen. 2005. The Authoritarian Dynamic. Cambridge:Cambridge University Press.

Stock, James H., and Motohiro Yogo. 2002. “Testing for WeakInstruments in Linear IV Regression.” Working Paper 284,National Bureau of Economic Research.

Sturgis, Patrick, Ian Brunton-Smith, Jouni Kuha, and Jonathan Jackson.2014. “Ethnic Diversity, Segregation and the Social Cohesion ofNeighbourhoods inLondon.”EthnicandRacialStudies37(8):1286–309.

UNHCR. 2016. “Emergency Information Sharing Web Portal.”Available at http://data.unhcr.org/mediterranean/documents.php.

Van Laar, Colette, Shana Levin, Stacey Sinclair, and Jim Sidanius.2005. “The Effect of University Roommate Contact on EthnicAttitudes and Behavior.” Journal of Experimental Social Psy-chology 41 (4): 329–45.

Vertier, Paul, and Max Viskanic. 2018. “Dismantling the ‘Jungle’:Migrant Relocation and Extreme Voting in France.” CESifoWorking Paper Series 6927, CESifo Group Munich.

Wike, Richard, Bruce Stokes, and Katie Simmons. 2016. “EuropeansFearWaveofRefugeesWillMeanMoreTerrorism,FewerJobs.” InPew Research Center, Available at: http://pewrsr.ch/29ziWV3.

Zorlu,Aslan. 2017. “Attitudes towardAsylumSeekers in Small LocalCommunities.” International Migration 55 (6): 14–36.

Does Exposure to the Refugee Crisis Make Natives More Hostile?

455

Dow

nloa

ded

from

htt

ps://

ww

w.c

ambr

idge

.org

/cor

e. IP

add

ress

: 54.

39.1

06.1

73, o

n 24

Jul 2

020

at 0

4:38

:20,

sub

ject

to th

e Ca

mbr

idge

Cor

e te

rms

of u

se, a

vaila

ble

at h

ttps

://w

ww

.cam

brid

ge.o

rg/c

ore/

term

s. h

ttps

://do

i.org

/10.

1017

/S00

0305

5418

0008

13