《社会保障概论》课程教学资源（课后阅读资料）APSR_Volume112. Issue 4.Nov 2018_Exclusion_and_Cooperation_in_Diverse_Societies_Experimental_Evidence_from_Israel

American Political Science Review (2018)112.4.742-757 doi:10.1017/S0003055418000266 American Political Science Association 2018 Exclusion and Cooperation in Diverse Societies:Experimental Evidence from Israel RYAN D.ENOS Harvard University NOAM GIDRON Hebrew University of Jerusalem and Princeton University t is well-established that in diverse societies,certain groups prefer to exclude other groups from power and often from society entirely.Yet as many societies are diversifying at an increasingly rapid pace,the need for cross-group cooperation to solve collective action problems has intensified.Do preferences for exclusion inhibit the ability of individuals to cooperate and,therefore,diminish the ability for societies to collectively provide public goods?Turning to Israel,a society with multiple overlapping and politically salient cleavages,we use a large-scale lab-in-the-field design to investigate how preferences for exclusion among the Jewish majority predict discriminatory behavior toward Palestinian Citizens of Israel.We establish that preferences for exclusion are likely symbolic attitudes,and therefore stable and dominating of other attitudes;are held especially strongly by low-status majority group members;and powerfully predict costly non-cooperation.This preferences/behavior relationship appears unaffected by mitigating factors proposed in the intergroup relations literature.The demonstrated influence of symbolic attitudes on behavior calls for further examination of the social roots of exclusionary preferences. t is well-understood that across groups and soci- regation in the United States,white citizens had to de- eties,people hold discriminatory attitudes toward cide whether to cooperate with African Americans cit- social outgroups.One of the most politically conse- izens or to retreat into enclaves of non-cooperation, quential ways these attitudes can manifest is in prefer- thus hindering the provision of public goods such as ences for exclusion,including exclusion from political schools.In South Africa,a similar challenge was faced institutions and power(Sidanius and Pratto 2001),the with the end of apartheid and the increase in inter- "imagined community"of a nation (Anderson 1983), actions across racial groups.Currently,anti-immigrant or the country itself via restrictive immigration poli- political parties and candidates in the United States cies(Citrin and Sides 2008).Yet,despite such barriers and Western Europe have gained support,yet immigra- and opposition,the ethnic and religious composition tion to these places continues;to what degree are sup- of many Western democracies continue to diversify porters of anti-immigrant policies willing to cooperate and,in recent decades,at an accelerating pace (Putnam with immigrants in their communities?And in Israel. 2007).Within this context of growing diversity,cooper- are Jewish citizens willing to cooperate with Palestinian 是 ation across social groups is necessary for building and Citizens of Israel(PCI)-despite the prevalence of ex- maintaining successful and well-functioning societies clusionary preferences toward this large and growing (Habyarimana et al.2009;Singh 2011).A crucial ques- national minority？ tion then emerges:Can individuals with a strong pref- Put in more general terms,the issue at stake is erence for outgroup exclusion nevertheless put these whether and to what degree exclusionary attitudes hin- preferences aside to cooperate and solve challenges of der cooperation.This question has implications for collective action? both the success of outgroup members in making po- There are many examples of the relevance of this litical and economic gains and the ability of society as question.Following the decline of de jure racial seg- a whole to work together for the greater good.Are ex- clusionary attitudes separable from behaviors?Even if members of the hegemonic group express a preference Ryan D.Enos is a Professor of Government,Department of Govern- for the exclusion of outgroup members,will they still ment,Harvard University,1737 Cambridge Street,Cambridge,MA work with outgroup members for their mutual benefit 02138 (renos@gov.harvard.edu). Noam Gidron is a Research Fellow at the Niehaus Center for in the creation and allocation of public goods? Globalization and Governance at Princeton University and an As- On the one hand,there is reason to believe that at- sistant Professor of Political Science,Hebrew University,Mount Sco- titudes are separable from behaviors.Cooperation is pus,Jerusalem (Noam.Gidron@mail.huji.ac.il). often treated as a strategic choice that can be updated Both authors contributed equally.Support for this research was with repeated interactions with outgroup members and provided by the Harvard Center for Jewish Studies,the Harvard Academy for International and Area Studies,and the Multidisci. other learning processes (Axelrod 2006).There is also plinary Program in Inequality and Social Policy at Harvard Uni- evidence that private attitudes toward outgroups may versity.A previous version of this article was presented at the 2017 not predict certain behaviors because the behaviors are Midwest Political Science Association Annual Meeting and the 2017 guided by social norms(Paluck 2009)or because atti- Toronto Political Behaviour Workshop.We thank Stanley Feld- tudes,as reported in surveys,are unstable and do not man and Eric Arias for comments and Riley Carney and Alexan- der Sahn for additional assistance.Replication files are available at reflect the more careful deliberation that may proceed the American Political Science Review Dataverse:https://doi.org/10. behavior (Zaller 1992).Yet,on the other hand,atti- 7910/DVN/DAR560 tudes may strongly predict non-cooperative behaviors Received:June 14,2017;revised:February 26,2018;accepted:May in light of evidence that exclusionary preferences are a 14,2018.First published online:July 13,2018. matter of deeply held prejudicial distaste(Hainmueller 742

American Political Science Review (2018) 112, 4, 742–757 doi:10.1017/S0003055418000266 © American Political Science Association 2018 Exclusion and Cooperation in Diverse Societies: Experimental Evidence from Israel RYAN D. ENOS Harvard University NOAM GIDRON Hebrew University of Jerusalem and Princeton University I t is well-established that in diverse societies, certain groups prefer to exclude other groups from power and often from society entirely. Yet as many societies are diversifying at an increasingly rapid pace,the need for cross-group cooperation to solve collective action problems has intensified. Do preferences for exclusion inhibit the ability of individuals to cooperate and, therefore, diminish the ability for societies to collectively provide public goods? Turning to Israel, a society with multiple overlapping and politically salient cleavages, we use a large-scale lab-in-the-field design to investigate how preferences for exclusion among the Jewish majority predict discriminatory behavior toward Palestinian Citizens of Israel. We establish that preferences for exclusion are likely symbolic attitudes, and therefore stable and dominating of other attitudes; are held especially strongly by low-status majority group members; and powerfully predict costly non-cooperation. This preferences/behavior relationship appears unaffected by mitigating factors proposed in the intergroup relations literature. The demonstrated influence of symbolic attitudes on behavior calls for further examination of the social roots of exclusionary preferences. I t is well-understood that across groups and soci￾eties, people hold discriminatory attitudes toward social outgroups. One of the most politically conse￾quential ways these attitudes can manifest is in prefer￾ences for exclusion, including exclusion from political institutions and power (Sidanius and Pratto 2001), the “imagined community” of a nation (Anderson 1983), or the country itself via restrictive immigration poli￾cies (Citrin and Sides 2008). Yet, despite such barriers and opposition, the ethnic and religious composition of many Western democracies continue to diversify and, in recent decades, at an accelerating pace (Putnam 2007).Within this context of growing diversity, cooper￾ation across social groups is necessary for building and maintaining successful and well-functioning societies (Habyarimana et al. 2009; Singh 2011). A crucial ques￾tion then emerges: Can individuals with a strong pref￾erence for outgroup exclusion nevertheless put these preferences aside to cooperate and solve challenges of collective action? There are many examples of the relevance of this question. Following the decline of de jure racial seg￾Ryan D. Enos is a Professor of Government, Department of Govern￾ment, Harvard University, 1737 Cambridge Street, Cambridge, MA 02138 (renos@gov.harvard.edu). Noam Gidron is a Research Fellow at the Niehaus Center for Globalization and Governance at Princeton University and an As￾sistant Professor of Political Science, Hebrew University,Mount Sco￾pus, Jerusalem (Noam.Gidron@mail.huji.ac.il). Both authors contributed equally. Support for this research was provided by the Harvard Center for Jewish Studies, the Harvard Academy for International and Area Studies, and the Multidisci￾plinary Program in Inequality and Social Policy at Harvard Uni￾versity. A previous version of this article was presented at the 2017 Midwest Political Science Association Annual Meeting and the 2017 Toronto Political Behaviour Workshop. We thank Stanley Feld￾man and Eric Arias for comments and Riley Carney and Alexan￾der Sahn for additional assistance. Replication files are available at the American Political Science Review Dataverse: https://doi.org/10. 7910/DVN/DAR56O. Received: June 14, 2017; revised: February 26, 2018; accepted: May 14, 2018. First published online: July 13, 2018. regation in the United States, white citizens had to de￾cide whether to cooperate with African Americans cit￾izens or to retreat into enclaves of non-cooperation, thus hindering the provision of public goods such as schools. In South Africa, a similar challenge was faced with the end of apartheid and the increase in inter￾actions across racial groups. Currently, anti-immigrant political parties and candidates in the United States and Western Europe have gained support, yet immigra￾tion to these places continues; to what degree are sup￾porters of anti-immigrant policies willing to cooperate with immigrants in their communities? And in Israel, are Jewish citizens willing to cooperate with Palestinian Citizens of Israel (PCI)—despite the prevalence of ex￾clusionary preferences toward this large and growing national minority? Put in more general terms, the issue at stake is whether and to what degree exclusionary attitudes hin￾der cooperation. This question has implications for both the success of outgroup members in making po￾litical and economic gains and the ability of society as a whole to work together for the greater good. Are ex￾clusionary attitudes separable from behaviors? Even if members of the hegemonic group express a preference for the exclusion of outgroup members, will they still work with outgroup members for their mutual benefit in the creation and allocation of public goods? On the one hand, there is reason to believe that at￾titudes are separable from behaviors. Cooperation is often treated as a strategic choice that can be updated with repeated interactions with outgroup members and other learning processes (Axelrod 2006). There is also evidence that private attitudes toward outgroups may not predict certain behaviors because the behaviors are guided by social norms (Paluck 2009) or because atti￾tudes, as reported in surveys, are unstable and do not reflect the more careful deliberation that may proceed behavior (Zaller 1992). Yet, on the other hand, atti￾tudes may strongly predict non-cooperative behaviors in light of evidence that exclusionary preferences are a matter of deeply held prejudicial distaste (Hainmueller 742 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Exclusion and Cooperation in Diverse Societies and Hiscox 2010:Hainmueller and Hangartner 2013; predicted by some social-psychology theory,levels Hopkins,Sides,and Citrin 2016). of exclusionary preferences are highest among low- These two approaches have very different implica- status Jews (the relatively poor and uneducated ultra- tions for addressing the challenges faced by diverse Orthodox population)-we explore the nature of these societies.If cooperation is mostly a strategic choice preferences and find that(3)exclusionary attitudes ap- then economic incentives,interactions,or learning can pear symbolic in nature,indicating they are stable and result in a stable equilibrium of cooperation.But if powerfully affect other attitudes.We then turn to the non-cooperation reflects prejudicial exclusionary pref- behaviors associated with these attitudes,and (4)find erences,it might not be addressed simply by material that the cooperation of Jewish and PCI is strongly pre- incentives for cooperation or by providing new infor- dicted by preferences for exclusion:Jews who endorse mation about outgroups. greater exclusion of Arabs,that is,greater social dis- The connection between cooperation and exclusion tance,systematically cooperate less with PCI.Finally, remains underexplored in the theoretical and empiri- we ask if this strong connection between attitudes and cal literature because the two components of the re- behaviors can be moderated by factors thought to pro- 4 lationship,cooperative behaviors and intergroup atti- mote cooperation and we find that(5)the relationship tudes,tend to be studied separately in political science. holds even when accounting for perceptions of Arabs' Political scientists have developed a powerful standard trustworthiness,suggesting that outgroup exclusionary toolkit for studying cooperative behaviors but have fo- attitudes do not merely reflect statistical,stereotype- cused on institutional and contextual moderators of based discrimination.We also look at measures of re- cooperation rather than on intergroup attitudes.Thus peated interaction between groups and find that the whether exclusionary attitudes,in fact,predict discrim- connection between exclusionary attitudes and coop- inatory behavior remains understudied,as reflected in eration is unaffected.In short,we find that exclusionary long-standing debates about the topic in other disci- attitudes are a robust predictor of cooperative behav- plines,including sociology and psychology (LaPiere ior,one that appears to be deeply rooted in individual 4r元 1934;Pager and Quillian 2005;Paluck 2009). psychology and not easily modifiable. We combine these two elements,deploying a large- Substantively,the strong link between attitudes and scale multi-site lab-in-the-field study and in-depth sur- behaviors suggests that diverse societies must di- vey in Israel to examine actual costly cooperation and rectly address the sources of exclusionary preferences how it is related to exclusionary attitudes of the Jew- to overcome collective action problems rather then ish majority toward PCI.We measure cooperative merely focus on the material benefits of coopera- behaviors using an economic decision-making game, tion,provide information about the trustworthiness the public goods game,that captures the challenge of (or other stereotypes)of minorities,or assume that cooperation in diverse societies (Habyarimana et al. repeated interactions will by themselves induce co- 2009).In measuring exclusionary preferences,we rely operation.These findings resonate with important re- on an underutilized measure among political scientists: search on the limited ability of material benefits to ad- social distance (Bogardus 1926).The social distance dress core elements of national conflicts in general and scale captures individuals'preference for sharing social the Palestinian-Israeli conflict in particular (Manekin space with a member of another group,by asking them Grossman,and Mitts 2016).Additionally,our findings to choose the degree of proximity to which they would that low-status members within the hegemonic major- accept outgroup members,ranging from a family rela- ity are,in fact,more likely to hold exclusionary pref- tive (closest)to none at all (most distant),with other erences toward and to practice non-cooperation with relationships in between.This scale is widely used in the low-status minority (and see also Gidron and Hall psychology and sociology but has largely been over- (2017)),sheds light on the deep barriers for political looked by political scientists,especially in the context cooperation across low-status groups. of behavioral games.Despite the relative lack of atten- Theoretically,these findings contribute to the lit- tion to social distance,because it captures the inclusion erature on intergroup relations in diverse societies or exclusion of minority groups,it has implications for (Alesina,Baqir,and Easterly 1999;Lieberman and Mc- the study of immigration,pluralism,and a wide range Clendon 2013;Uslaner 2012),which has largely ne- of other scholarship.To our knowledge,this is the first glected psychological characteristics.Our analysis also direct exploration of the relationship between social contributes to research on ethnic and racial discrimi- distance and cooperation.In the aggregate,such a rela- nation,not only in political science,but also in sociol- 四 tionship is likely to carry major implications for diverse ogy and economics(Charles and Guryan 2011;Hain- societies. mueller and Hangartner 2013;Pager and Shepherd In this article,we make several contributions.First 2008:Zussman2013). we explore the levels of exclusionary preferences Methodologically,our combination of survey and among the Jewish majority toward PCI.We find that experimental evidence allows us to address a long- (1)levels of exclusionary attitudes are high:(2)as standing debate regarding the relationship between discriminatory attitudes and discriminatory behaviors Because collecting behavioral data is often costly and This minority group is also sometimes referred to as Israeli Arabs We follow previous literature on the topic and use the term Pales- cumbersome,much research assumes,without empiri- tinian Citizens of Israel (PCl)(Canetti-Nisim,Ariely.and Halperin cal validation,that survey measures of discrimination 2008). are proxies for discriminatory behaviors(Pager and 743

Exclusion and Cooperation in Diverse Societies and Hiscox 2010; Hainmueller and Hangartner 2013; Hopkins, Sides, and Citrin 2016). These two approaches have very different implica￾tions for addressing the challenges faced by diverse societies. If cooperation is mostly a strategic choice, then economic incentives, interactions, or learning can result in a stable equilibrium of cooperation. But if non-cooperation reflects prejudicial exclusionary pref￾erences, it might not be addressed simply by material incentives for cooperation or by providing new infor￾mation about outgroups. The connection between cooperation and exclusion remains underexplored in the theoretical and empiri￾cal literature because the two components of the re￾lationship, cooperative behaviors and intergroup atti￾tudes, tend to be studied separately in political science. Political scientists have developed a powerful standard toolkit for studying cooperative behaviors but have fo￾cused on institutional and contextual moderators of cooperation rather than on intergroup attitudes. Thus, whether exclusionary attitudes, in fact, predict discrim￾inatory behavior remains understudied, as reflected in long-standing debates about the topic in other disci￾plines, including sociology and psychology (LaPiere 1934; Pager and Quillian 2005; Paluck 2009). We combine these two elements, deploying a large￾scale multi-site lab-in-the-field study and in-depth sur￾vey in Israel to examine actual costly cooperation and how it is related to exclusionary attitudes of the Jew￾ish majority toward PCI.1 We measure cooperative behaviors using an economic decision-making game, the public goods game, that captures the challenge of cooperation in diverse societies (Habyarimana et al. 2009). In measuring exclusionary preferences, we rely on an underutilized measure among political scientists: social distance (Bogardus 1926). The social distance scale captures individuals’ preference for sharing social space with a member of another group, by asking them to choose the degree of proximity to which they would accept outgroup members, ranging from a family rela￾tive (closest) to none at all (most distant), with other relationships in between. This scale is widely used in psychology and sociology but has largely been over￾looked by political scientists, especially in the context of behavioral games. Despite the relative lack of atten￾tion to social distance, because it captures the inclusion or exclusion of minority groups, it has implications for the study of immigration, pluralism, and a wide range of other scholarship. To our knowledge, this is the first direct exploration of the relationship between social distance and cooperation. In the aggregate, such a rela￾tionship is likely to carry major implications for diverse societies. In this article, we make several contributions. First we explore the levels of exclusionary preferences among the Jewish majority toward PCI. We find that (1) levels of exclusionary attitudes are high; (2) as 1 This minority group is also sometimes referred to as Israeli Arabs. We follow previous literature on the topic and use the term Pales￾tinian Citizens of Israel (PCI) (Canetti-Nisim, Ariely, and Halperin 2008). predicted by some social-psychology theory, levels of exclusionary preferences are highest among low￾status Jews (the relatively poor and uneducated ultra￾Orthodox population)—we explore the nature of these preferences and find that (3) exclusionary attitudes ap￾pear symbolic in nature, indicating they are stable and powerfully affect other attitudes. We then turn to the behaviors associated with these attitudes, and (4) find that the cooperation of Jewish and PCI is strongly pre￾dicted by preferences for exclusion: Jews who endorse greater exclusion of Arabs, that is, greater social dis￾tance, systematically cooperate less with PCI. Finally, we ask if this strong connection between attitudes and behaviors can be moderated by factors thought to pro￾mote cooperation and we find that (5) the relationship holds even when accounting for perceptions of Arabs’ trustworthiness, suggesting that outgroup exclusionary attitudes do not merely reflect statistical, stereotype￾based discrimination. We also look at measures of re￾peated interaction between groups and find that the connection between exclusionary attitudes and coop￾eration is unaffected. In short, we find that exclusionary attitudes are a robust predictor of cooperative behav￾ior, one that appears to be deeply rooted in individual psychology and not easily modifiable. Substantively, the strong link between attitudes and behaviors suggests that diverse societies must di￾rectly address the sources of exclusionary preferences to overcome collective action problems rather then merely focus on the material benefits of coopera￾tion, provide information about the trustworthiness (or other stereotypes) of minorities, or assume that repeated interactions will by themselves induce co￾operation. These findings resonate with important re￾search on the limited ability of material benefits to ad￾dress core elements of national conflicts in general and the Palestinian-Israeli conflict in particular (Manekin, Grossman, and Mitts 2016). Additionally, our findings that low-status members within the hegemonic major￾ity are, in fact, more likely to hold exclusionary pref￾erences toward and to practice non-cooperation with the low-status minority (and see also Gidron and Hall (2017)), sheds light on the deep barriers for political cooperation across low-status groups. Theoretically, these findings contribute to the lit￾erature on intergroup relations in diverse societies (Alesina, Baqir, and Easterly 1999; Lieberman and Mc￾Clendon 2013; Uslaner 2012), which has largely ne￾glected psychological characteristics. Our analysis also contributes to research on ethnic and racial discrimi￾nation, not only in political science, but also in sociol￾ogy and economics (Charles and Guryan 2011; Hain￾mueller and Hangartner 2013; Pager and Shepherd 2008; Zussman 2013). Methodologically, our combination of survey and experimental evidence allows us to address a long￾standing debate regarding the relationship between discriminatory attitudes and discriminatory behaviors. Because collecting behavioral data is often costly and cumbersome, much research assumes, without empiri￾cal validation, that survey measures of discrimination are proxies for discriminatory behaviors (Pager and 743 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Ryan D.Enos and Noam Gidron Ouillian 2005).However.currently"most of the exist- sures a person's willingness to participate in relation- ing literature on discrimination finds that stated atti- ships of varying degrees of closeness with a member of tudes are practically useless in explaining behaviour' a group in order to capture "personal-group relations" (Zussman 2013,436).2 In fact,recent scholarship has (Bogardus 1933)-that is,the affective feelings of an also claimed that even implicit measures of prejudice individual toward a group.Psychological studies have (Greenwald,McGhee,and Schwartz 1998)are poor treated social distance as a "commonly accepted gen- predictors of discriminatory behavior(Mitchell and eral measure of ethnic prejudice"(Weaver 2008,779)3 Tetlock 2017).By showing the strong and robust con- but for the purposes of political science research,it is nection between exclusionary attitudes and coopera- especially useful because it captures behavioral inten- tion,we present a counterpoint to these claims.We tions(Binder et al.2009)about the political question of suggest that a possible path forward in research on prej- exclusion,so that a stated desire to exclude may trans- udice and discrimination,rather than relying on noisy late into political behavior,such as voting (see,for ex- self-reported measures of behavior or poorly under- ample,Hainmueller and Hangartner(2013)). stood laboratory constructs,is to make use of the social The scale,developed by Bogardus,measures the de- distance scale next to well-validated tools of behavioral gree to which respondents prefer to exclude outgroup economics. members by asking whether they would accept a mem- Relatedly,our evidence for a direct connection be- ber of the group at increasing levels of closeness.The tween attitudes and behaviors also represents an im- scale ranges from family relative(minimal distance)to portant improvement over much of the literature on friend,neighbor,coworker,citizen,visitor,and none intergroup relations-a literature plagued by unre (maximal distance).Agreement with any one of the solved controversies over the nature and meaning of items implies agreement with the previous items.For survey attitudes (e.g..Sniderman and Tetlock (1986): instance,it is assumed that if someone will accept a per- Hochschild (2000)).Because we focus on questions son as a family member,they will also accept them as that directly measure exclusion and connect these to visitor and everything in between revealed behaviors in an economic game,our outcomes This scale has attractive properties that may improve are less likely than many others to reflect artifacts over other common measures of intergroup attitudes. of measurement error(Achen 1975),capture nonat- Political scientists often measure intergroup attitudes titudes (Zaller 1992),or suffer from false positives through culturally specific stereotypes (e.g.,asking if a (Kramer 1986).It is,perhaps,not surprising if a sur- group is "intelligent"(Kinder and Kam 2009))through vey attitude is correlated with another survey attitude questions specific to one group,such as African Amer- measuring a similar concept,especially given the large icans (e.g.,Tarman and Sears(2005));or through ques- menu of survey items often available to researchers. tions about specific policy measures,such as immigra- However,as discussed above,it is not obvious that a tion (e.g.,Enos (2014)).While these measures can all survey attitude will correlate with a behavior,espe- certainly be useful,group-specific measures make it cially a potentially costly one like non-cooperation.By difficult to compare attitudes across different groups demonstrating this connection,we move beyond deter- (e.g.,African Americans and Muslims)and individu- mining the meaning of survey attitudes by examining als may also hold negative feelings about a group with- other survey attitudes and,instead,show that these at- out endorsing specific stereotypes (Cuddy,Fiske,and S5.501g titudes are meaningful because they are tied to costly Glick 2007).Furthermore,attitudes about specific poli- behaviors. cies are problematic for capturing the attitudes of the large majority of most mass publics that have low en- EXCLUSION gagement in politics and hold unstable attitudes(Zaller 1992). We define exclusion as closing all or part of a society The social distance scale.in contrast,was intended from certain groups of people.Conflicts over the legal as a general measure to be used across multiple groups exclusion of low-status populations have been central (Bogardus 1926).It captures basic affective attitudes: to the politics of countries across the globe (Sidanius a person does not have to endorse specific stereotypes and Pratto 2001)and are at the heart of many political to know that they do not want to have a person from conflicts in advanced democracies.We focus on pref- an outgroup as a spouse or coworker.This allows erences for exclusion,not just from the country in the comparisons across groups,so that,for example,in the form of restricting immigration,but from national and American context.attitudes about African Americans subnational communities,in the form of excluding in- can be compared to attitudes about Latino immigrants dividuals from local institutions,such as a workplace. Thus,in Israel,we can usefully compare exclusion to- To measure exclusion,we turn to the concept of so- ward PCI to exclusion of other social groups,for in- cial distance.Social distance is a commonly used con- stance,as we do in this study,different groups of Jews. cept in sociology and psychology (see,for example Other measures commonly found in political science, Liviatan,Trope,and Liberman (2008)),but has seen lit- such as feeling thermometers,have similar properties tle use in political science (but see,in the Israeli context. Halperin,Canetti-Nisim,and Pedahzur(2007)).It mea- See also Marger(2003)and Simpson and Yinger(2013). For example,Binder et al.(2009,848)argues that social distance "generalizes to the outgroup as a whole without any reference to a 2 For a review of related literature,see Pager and Shepherd(2008). specific context or interaction.' 744 3

Ryan D. Enos and Noam Gidron Quillian 2005). However, currently “most of the exist￾ing literature on discrimination finds that stated atti￾tudes are practically useless in explaining behaviour” (Zussman 2013, 436).2 In fact, recent scholarship has also claimed that even implicit measures of prejudice (Greenwald, McGhee, and Schwartz 1998) are poor predictors of discriminatory behavior (Mitchell and Tetlock 2017). By showing the strong and robust con￾nection between exclusionary attitudes and coopera￾tion, we present a counterpoint to these claims. We suggest that a possible path forward in research on prej￾udice and discrimination, rather than relying on noisy self-reported measures of behavior or poorly under￾stood laboratory constructs, is to make use of the social distance scale next to well-validated tools of behavioral economics. Relatedly, our evidence for a direct connection be￾tween attitudes and behaviors also represents an im￾portant improvement over much of the literature on intergroup relations—a literature plagued by unre￾solved controversies over the nature and meaning of survey attitudes (e.g., Sniderman and Tetlock (1986); Hochschild (2000)). Because we focus on questions that directly measure exclusion and connect these to revealed behaviors in an economic game, our outcomes are less likely than many others to reflect artifacts of measurement error (Achen 1975), capture nonat￾titudes (Zaller 1992), or suffer from false positives (Kramer 1986). It is, perhaps, not surprising if a sur￾vey attitude is correlated with another survey attitude measuring a similar concept, especially given the large menu of survey items often available to researchers. However, as discussed above, it is not obvious that a survey attitude will correlate with a behavior, espe￾cially a potentially costly one like non-cooperation. By demonstrating this connection, we move beyond deter￾mining the meaning of survey attitudes by examining other survey attitudes and, instead, show that these at￾titudes are meaningful because they are tied to costly behaviors. EXCLUSION We define exclusion as closing all or part of a society from certain groups of people. Conflicts over the legal exclusion of low-status populations have been central to the politics of countries across the globe (Sidanius and Pratto 2001) and are at the heart of many political conflicts in advanced democracies. We focus on pref￾erences for exclusion, not just from the country in the form of restricting immigration, but from national and subnational communities, in the form of excluding in￾dividuals from local institutions, such as a workplace. To measure exclusion, we turn to the concept of so￾cial distance. Social distance is a commonly used con￾cept in sociology and psychology (see, for example, Liviatan, Trope, and Liberman (2008)), but has seen lit￾tle use in political science (but see,in the Israeli context, Halperin,Canetti-Nisim, and Pedahzur (2007)). It mea- 2 For a review of related literature, see Pager and Shepherd (2008). sures a person’s willingness to participate in relation￾ships of varying degrees of closeness with a member of a group in order to capture “personal-group relations” (Bogardus 1933)—that is, the affective feelings of an individual toward a group. Psychological studies have treated social distance as a “commonly accepted gen￾eral measure of ethnic prejudice” (Weaver 2008, 779)3 but for the purposes of political science research, it is especially useful because it captures behavioral inten￾tions (Binder et al. 2009) about the political question of exclusion, so that a stated desire to exclude may trans￾late into political behavior, such as voting (see, for ex￾ample, Hainmueller and Hangartner (2013)). The scale, developed by Bogardus, measures the de￾gree to which respondents prefer to exclude outgroup members by asking whether they would accept a mem￾ber of the group at increasing levels of closeness. The scale ranges from family relative (minimal distance) to friend, neighbor, coworker, citizen, visitor, and none (maximal distance). Agreement with any one of the items implies agreement with the previous items. For instance,it is assumed that if someone will accept a per￾son as a family member, they will also accept them as visitor and everything in between. This scale has attractive properties that may improve over other common measures of intergroup attitudes. Political scientists often measure intergroup attitudes through culturally specific stereotypes (e.g., asking if a group is “intelligent” (Kinder and Kam 2009)) through questions specific to one group, such as African Amer￾icans (e.g., Tarman and Sears (2005)); or through ques￾tions about specific policy measures, such as immigra￾tion (e.g., Enos (2014)). While these measures can all certainly be useful, group-specific measures make it difficult to compare attitudes across different groups (e.g., African Americans and Muslims) and individu￾als may also hold negative feelings about a group with￾out endorsing specific stereotypes (Cuddy, Fiske, and Glick 2007). Furthermore, attitudes about specific poli￾cies are problematic for capturing the attitudes of the large majority of most mass publics that have low en￾gagement in politics and hold unstable attitudes (Zaller 1992). The social distance scale, in contrast, was intended as a general measure to be used across multiple groups (Bogardus 1926). It captures basic affective attitudes: a person does not have to endorse specific stereotypes to know that they do not want to have a person from an outgroup as a spouse or coworker.4 This allows comparisons across groups, so that, for example, in the American context, attitudes about African Americans can be compared to attitudes about Latino immigrants. Thus, in Israel, we can usefully compare exclusion to￾ward PCI to exclusion of other social groups, for in￾stance, as we do in this study, different groups of Jews. Other measures commonly found in political science, such as feeling thermometers, have similar properties 3 See also Marger (2003) and Simpson and Yinger (2013). 4 For example, Binder et al. (2009, 848) argues that social distance “generalizes to the outgroup as a whole without any reference to a specific context or interaction.” 744 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Exclusion and Cooperation in Diverse Societies but responses to these questions tend to have little vari- ditions under which groups can overcome individual ation,calling into question their ability to usefully dis incentives to defect"(Grossman and Baldassarri 2012 criminate between attitudes (Krosnick 1991).Finally, 965).As Baldassarri(2015.367)notes,"designed to in- the widespread use and cross-cultural applicability of duce a social dilemma,[public goods games]capture this scale allows it to be compared to measurements how players balance self-interest and the well-being taken in other contexts,both in existing and future of the group."Habyarimana et al.(2009)argue that studies.5 the public goods game"captures the challenge of pub- A crucial question for predicting the relationship lic goods provision directly"and use results from such between exclusionary attitudes and cooperation games to argue that the failure to cooperate across eth- whether these preferences are symbolic attitudes.In nic groups in social situations analogous to the public the intergroup context,symbolic attitudes are attitudes goods game is the primary reason that diverse societies developed around affective responses to a particular fail to allocate desirable public goods. group that are socialized early in life,are stable over a Habyarimana et al.(2009)attribute this lack of coop- lifespan,and tend to dominate other attitudes(Tarman eration across ethnic groups to a lack of norms of coop- and Sears 2005).The canonical example of symbolic at eration.In other social science work,variation in play titudes is attitudes toward blacks and other racial and in the public goods game is ascribed to differences inin- ethnic groups in the United States(Sears and Henry stitutions(Alexander and Christia 2011).culture (Hen- 2003).Recent empirical work in American politics sug- rich et al.2006),geographic context(Enos 2017),or gests that attitudes toward immigrants are also sym- statistically based stereotypes (Fershtman and Gneezy bolic attitudes(Hopkins,Sides,and Citrin 2016). 2001).The focus in political science and economics Whether social distance should be understood as a on norms and institutions as determinants of cooper- symbolic attitude is important because it speaks to its vc士 ative behavior is understandable given the intellectual likelihood of dominating other attitudes and also of foundation of both disciplines.Yet this focus may ne- changing in the face of shifting demographics.If ex- glect important sources of variation.Cooperation is 4r元 clusionary preferences are symbolic attitudes,they are also likely structured by individual-level differences,in- likely to dominate other attitudes,meaning other atti- cluding psychological traits such as the attitudes asso- tudes will be shaped by the symbolic attitudes-not the ciated with exclusionary preferences.Robust findings other way around.This means that symbolic attitudes from psychology point to this connection:The cogni- are predicted to strongly affect opinion when brought tive biases associated with ingroup favoritism cause to bear on political questions.For example,turning to individuals to seek maximum distinctiveness between the American context.classic literature demonstrates groups,even when it is costly to their own group(Tajfel how whites'attitudes toward blacks dominates other et al.1971;Tajfel and Turner 1979;Turner and Oakes considerations in policy questions from school busing 1986).In other words,when choosing how to allocate (Kinder and Sears 1981)to health care (Tesler 2012). money,anti-outgroup or pro-ingroup biases (Brewer and Miller 1984)cause people to forgo allocations that COOPERATION are beneficial to their own group or mutually benefi- cial to both groups,to select allocations that maximize We connect preference for exclusion with cooperative the difference in monetary payout between groups.A behaviors.Cooperation is necessary to maintain public bias for maximizing differences would predict defec- goods provision,from roads to schools(Habyarimana tion in a public goods game,rather than mutually bene- ficial cooperation.Furthermore,evolutionary psychol- 8 et al.2009),and for the operation of democratic insti- tutions,such as legislatures (Axelrod 2006)-but the ogists have hypothesized that the evolution of the cog- logic of collective action means that cooperation is of- nitive adaptations for social exclusion are a result of ten difficult to achieve (Olson 1971).Given the central selective pressures for efficient within-group cooper- importance of cooperation,social scientists have de- ation and between-group competition (Kurzban and veloped a toolkit for measuring it,including the pris- Leary 2001).5 Also drawing on evolutionary reasoning, oner's dilemma or public goods game.This game re- Sidanius and Pratto(2001)argue that the competition wards participants for mutual cooperation,but rewards between groups to maintain group-based status hier- them more for defecting and allowing the other player archies and to exclude low-status groups from power to carry the cost.This tends to lead to mutual defection causes a range of discriminatory and non-cooperative where neither player cooperates. behavior. 四 The public goods game has been argued to mimic Despite the reasons to believe that exclusion and the dynamics underlying the challenges to societies in cooperation are related,as noted above,the relation- allocating public goods and the tendency for diverse ship between discriminatory attitudes and discrimina- societies in particular to fail at doing so.It "is the con- tory behaviors is heavily contested in the literature, ventional behavioral experiment used to study the con- and other scholarship may point to reasons to believe these attitudes and behaviors should be unrelated.In However,it should be noted that the meaning of levels of closeness in the social distance scale may vary across groups,so that "family rel- Quoting Kurzban and Leary (2001,195):"Indeed,if adaptations ative"may imply a different level of closeness in different cultures.It for within-group cooperation are designed for between-group com does,nevertheless,seem reasonable to assume that the ordinal rank- petition,then the psychology of inclusion and cooperation requires ings of the levels will be consistent across most cultures. a concurrent psychology of social exclusion and discrimination." 745

Exclusion and Cooperation in Diverse Societies but responses to these questions tend to have little vari￾ation, calling into question their ability to usefully dis￾criminate between attitudes (Krosnick 1991). Finally, the widespread use and cross-cultural applicability of this scale allows it to be compared to measurements taken in other contexts, both in existing and future studies.5 A crucial question for predicting the relationship between exclusionary attitudes and cooperation is whether these preferences are symbolic attitudes. In the intergroup context, symbolic attitudes are attitudes developed around affective responses to a particular group that are socialized early in life, are stable over a lifespan, and tend to dominate other attitudes (Tarman and Sears 2005).The canonical example of symbolic at￾titudes is attitudes toward blacks and other racial and ethnic groups in the United States (Sears and Henry 2003). Recent empirical work in American politics sug￾gests that attitudes toward immigrants are also sym￾bolic attitudes (Hopkins, Sides, and Citrin 2016). Whether social distance should be understood as a symbolic attitude is important because it speaks to its likelihood of dominating other attitudes and also of changing in the face of shifting demographics. If ex￾clusionary preferences are symbolic attitudes, they are likely to dominate other attitudes, meaning other atti￾tudes will be shaped by the symbolic attitudes—not the other way around. This means that symbolic attitudes are predicted to strongly affect opinion when brought to bear on political questions. For example, turning to the American context, classic literature demonstrates how whites’ attitudes toward blacks dominates other considerations in policy questions from school busing (Kinder and Sears 1981) to health care (Tesler 2012). COOPERATION We connect preference for exclusion with cooperative behaviors. Cooperation is necessary to maintain public goods provision, from roads to schools (Habyarimana et al. 2009), and for the operation of democratic insti￾tutions, such as legislatures (Axelrod 2006)—but the logic of collective action means that cooperation is of￾ten difficult to achieve (Olson 1971). Given the central importance of cooperation, social scientists have de￾veloped a toolkit for measuring it, including the pris￾oner’s dilemma or public goods game. This game re￾wards participants for mutual cooperation, but rewards them more for defecting and allowing the other player to carry the cost.This tends to lead to mutual defection, where neither player cooperates. The public goods game has been argued to mimic the dynamics underlying the challenges to societies in allocating public goods and the tendency for diverse societies in particular to fail at doing so. It “is the con￾ventional behavioral experiment used to study the con- 5 However, it should be noted that the meaning of levels of closeness in the social distance scale may vary across groups, so that “family rel￾ative” may imply a different level of closeness in different cultures. It does, nevertheless, seem reasonable to assume that the ordinal rank￾ings of the levels will be consistent across most cultures. ditions under which groups can overcome individual incentives to defect” (Grossman and Baldassarri 2012, 965). As Baldassarri (2015, 367) notes, “designed to in￾duce a social dilemma, [public goods games] capture how players balance self-interest and the well-being of the group.” Habyarimana et al. (2009) argue that the public goods game “captures the challenge of pub￾lic goods provision directly” and use results from such games to argue that the failure to cooperate across eth￾nic groups in social situations analogous to the public goods game is the primary reason that diverse societies fail to allocate desirable public goods. Habyarimana et al. (2009) attribute this lack of coop￾eration across ethnic groups to a lack of norms of coop￾eration. In other social science work, variation in play in the public goods game is ascribed to differences in in￾stitutions (Alexander and Christia 2011), culture (Hen￾rich et al. 2006), geographic context (Enos 2017), or statistically based stereotypes (Fershtman and Gneezy 2001). The focus in political science and economics on norms and institutions as determinants of cooper￾ative behavior is understandable given the intellectual foundation of both disciplines. Yet this focus may ne￾glect important sources of variation. Cooperation is also likely structured by individual-level differences,in￾cluding psychological traits such as the attitudes asso￾ciated with exclusionary preferences. Robust findings from psychology point to this connection: The cogni￾tive biases associated with ingroup favoritism cause individuals to seek maximum distinctiveness between groups, even when it is costly to their own group (Tajfel et al. 1971; Tajfel and Turner 1979; Turner and Oakes 1986). In other words, when choosing how to allocate money, anti-outgroup or pro-ingroup biases (Brewer and Miller 1984) cause people to forgo allocations that are beneficial to their own group or mutually benefi￾cial to both groups, to select allocations that maximize the difference in monetary payout between groups. A bias for maximizing differences would predict defec￾tion in a public goods game, rather than mutually bene￾ficial cooperation. Furthermore, evolutionary psychol￾ogists have hypothesized that the evolution of the cog￾nitive adaptations for social exclusion are a result of selective pressures for efficient within-group cooper￾ation and between-group competition (Kurzban and Leary 2001).6 Also drawing on evolutionary reasoning, Sidanius and Pratto (2001) argue that the competition between groups to maintain group-based status hier￾archies and to exclude low-status groups from power causes a range of discriminatory and non-cooperative behavior. Despite the reasons to believe that exclusion and cooperation are related, as noted above, the relation￾ship between discriminatory attitudes and discrimina￾tory behaviors is heavily contested in the literature, and other scholarship may point to reasons to believe these attitudes and behaviors should be unrelated. In 6 Quoting Kurzban and Leary (2001, 195): “Indeed, if adaptations for within-group cooperation are designed for between-group com￾petition, then the psychology of inclusion and cooperation requires a concurrent psychology of social exclusion and discrimination.” 745 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Ryan D.Enos and Noam Gidron particular,strategic behavior in situations like the tion and are subject to discrimination by state institu- public goods game may not reflect prejudicial atti- tions (Okun and Friedlander 2005). tudes like preferences for exclusion-after all,a central There is long-standing research on Israeli public premise of both cognitive psychology and behavioral opinion about both domestic and international Jewish- economics is that the mind is characterized by two sys- Arab tensions(Smooha 1987,1992,2002,2004:Gubler tems (Kahneman 2003),one of which makes the fast, and Kalmoe 2015;Gubler,Halperin,and Hirschberger heuristic decisions associated with affective associa- 2015:Zeitzoff 2014.2016).with clear evidence for tions,like exclusionary preferences,and another which widespread prejudice toward PCI.Pedahzur and Yishai makes the slower,more deliberate decisions that char- (1999)document"deep resentment toward the Arabs," acterize strategic choice.The latter system is known to with 56.5%of Jewish respondents in their sample op- overrule the former when the stakes are high.Labora- posed to granting Arab and Jewish citizens equal so- tory cooperation games are intentionally made to be cial rights.Bar and Zussman (2017)show that around costly and invoke this sort of strategic behavior. 40%of Jewish Israelis would be willing to pay more to Indeed,there are real-world examples of the separa- receive services from Jewish workers rather than from tion of strongly held prejudices and cooperative behav- Arab workers and Zussman(2013)reports that more ior in costly situations.For example.Axelrod (2006). than half of the respondents in his sample would prefer drawing on the logic of the public goods game.de- not to have an Arab neighbor. scribes the system by which opposing forces in World The PCI are not only a national-religious minority War I,despite the presumably strong feelings involved, within a state defined by Jewish nationality;they are 元 developed cooperative systems of"live and let live"to also often perceived as a security threat or a "fifth col- avoid the devastating costs of trench warfare umn"in the context of Israel's armed conflicts with In the face of these contrasting theoretical intuitions its surrounding Arab neighbors(Canetti-Nisim,Ariely, and empirical findings,our analyses take the first step and Halperin 2008).Smooha(2004)shows that a sub- toward showing that exclusion and cooperation are stantial share of Jewish Israelis believe that PCI sup- strongly related and,as such,further shed light on the port terrorism and may rebel against Israel in the fu- challenges that diversifying societies face in overcom- ture.This makes the case of exclusion of PCI poten- ing barriers for cooperation.Because we cannot ran- tially informative for thinking about minority groups domly assign a preference for exclusion,we cannot,of elsewhere:perceptions of security threats affect atti- course,speak directly to the causal effect of exclusion- tudes toward Muslim immigrants to Western countries ary attitudes on cooperation.However,even though (Hellwig and Sinno 2016)and stereotypes of orga- demonstrating so is not our focus,we do show that ex- nized,even transnational,criminality are often asso- clusionary attitudes can be characterized by associa- ciated with Latin American immigrants to the United tions that suggest they are developed early in life and States.Furthermore,the situation of a minority group are,therefore,causally prior to cooperative behavior. having cultural and familial ties to neighboring states 是 can also be found elsewhere,again such as with Latino INTERGROUP RELATIONS IN ISRAEL immigrants to the United States. We focus on the general issue of exclusion by Jewish With its high levels of diversity along multiple dimen- Israelis toward PCI,but also on the particular ques- 5.501g sions,Israel provides a fertile case for the study of in- tion of the attitudes of ultra-Orthodox Jews.Ultra- tergroup relations.While there are,of course,unique Orthodox Jews are distinctive among the Jewish ma- features of Israeli society,it is a case that is useful in jority because of their religious and social traditions understanding social dynamics in other societies that including low formal education and widespread non- are becoming increasingly heterogeneous.As Canetti- participation in the workforce,resulting in a popula- Nisim,Ariely,and Halperin (2008,92)write,"Israel's tion that is substantially poorer than the general Jew- ethno-national character as a Jewish state,the ongoing ish population.Furthermore,there are also barriers Arab-Israeli conflict,and the complex relations be- for cooperation between the ultra-Orthodox and the tween Jews and Arabs in Israel [...have turned Israel rest of the Israeli Jewish population(Enos and Gidron into a laboratory conducive to the study of the devel- 2016).The ultra-Orthodox thus present an informative opment of negative political attitudes toward various point of reference in their relations with the PCI be- minority groups.' cause they are,arguably,close to PCI on a social hi- We focus on intergroup relations between Jewish erarchy.According to prominent social-psychological and PCI,a highly salient social-political cleavage de- theories,they may therefore hold more exclusionary fined on religious and nationalistic lines.PCI con- attitudes due to greater perceived threat to their rel- stitute around 20%of the Israeli population.As a ative status (Blumer 1958;Bobo and Hutchings 1996; marginalized minority.the PCI are characterized by a Sidanius and Pratto 2001).Competition over resources low socioeconomic status and labor market participa- is often most relevant to the low-income portion of a majority group that shares social welfare institutions L with low-status minority groups (Bobo and Hutch- 7 Note that we examine Israeli-Jews'attitudes and behaviors toward ings 1996).3 Of course,less well-off subgroups of the PCI,as opposed to non-citizen Palestinians or citizens of neighboring Arab countries,the West Bank and Gaza.For research on Israeli- Jews'attitudes toward non-citizen Arabs,see Inbar and Yuchtman- 8 Although,notably,in the Israeli context,ultra-Orthodox and PCI Yaar(1986).The 20%figure does not include the West Bank or Gaza are largely institutionally separated,so competition over institutions populations. such as schools is not present. 746

Ryan D. Enos and Noam Gidron particular, strategic behavior in situations like the public goods game may not reflect prejudicial atti￾tudes like preferences for exclusion—after all, a central premise of both cognitive psychology and behavioral economics is that the mind is characterized by two sys￾tems (Kahneman 2003), one of which makes the fast, heuristic decisions associated with affective associa￾tions, like exclusionary preferences, and another which makes the slower, more deliberate decisions that char￾acterize strategic choice. The latter system is known to overrule the former when the stakes are high. Labora￾tory cooperation games are intentionally made to be costly and invoke this sort of strategic behavior. Indeed, there are real-world examples of the separa￾tion of strongly held prejudices and cooperative behav￾ior in costly situations. For example, Axelrod (2006), drawing on the logic of the public goods game, de￾scribes the system by which opposing forces in World War I, despite the presumably strong feelings involved, developed cooperative systems of “live and let live” to avoid the devastating costs of trench warfare. In the face of these contrasting theoretical intuitions and empirical findings, our analyses take the first step toward showing that exclusion and cooperation are strongly related and, as such, further shed light on the challenges that diversifying societies face in overcom￾ing barriers for cooperation. Because we cannot ran￾domly assign a preference for exclusion, we cannot, of course, speak directly to the causal effect of exclusion￾ary attitudes on cooperation. However, even though demonstrating so is not our focus, we do show that ex￾clusionary attitudes can be characterized by associa￾tions that suggest they are developed early in life and are, therefore, causally prior to cooperative behavior. INTERGROUP RELATIONS IN ISRAEL With its high levels of diversity along multiple dimen￾sions, Israel provides a fertile case for the study of in￾tergroup relations. While there are, of course, unique features of Israeli society, it is a case that is useful in understanding social dynamics in other societies that are becoming increasingly heterogeneous. As Canetti￾Nisim, Ariely, and Halperin (2008, 92) write, “Israel’s ethno-national character as a Jewish state, the ongoing Arab-Israeli conflict, [and] the complex relations be￾tween Jews and Arabs in Israel [...] have turned Israel into a laboratory conducive to the study of the devel￾opment of negative political attitudes toward various minority groups.” We focus on intergroup relations between Jewish and PCI, a highly salient social-political cleavage de￾fined on religious and nationalistic lines. PCI con￾stitute around 20% of the Israeli population.7 As a marginalized minority, the PCI are characterized by a low socioeconomic status and labor market participa- 7 Note that we examine Israeli-Jews’ attitudes and behaviors toward PCI, as opposed to non-citizen Palestinians or citizens of neighboring Arab countries, the West Bank and Gaza. For research on Israeli￾Jews’ attitudes toward non-citizen Arabs, see Inbar and Yuchtman￾Yaar (1986). The 20% figure does not include the West Bank or Gaza populations. tion and are subject to discrimination by state institu￾tions (Okun and Friedlander 2005). There is long-standing research on Israeli public opinion about both domestic and international Jewish￾Arab tensions (Smooha 1987, 1992, 2002, 2004; Gubler and Kalmoe 2015; Gubler, Halperin, and Hirschberger 2015; Zeitzoff 2014, 2016), with clear evidence for widespread prejudice toward PCI. Pedahzur and Yishai (1999) document “deep resentment toward the Arabs,” with 56.5% of Jewish respondents in their sample op￾posed to granting Arab and Jewish citizens equal so￾cial rights. Bar and Zussman (2017) show that around 40% of Jewish Israelis would be willing to pay more to receive services from Jewish workers rather than from Arab workers and Zussman (2013) reports that more than half of the respondents in his sample would prefer not to have an Arab neighbor. The PCI are not only a national-religious minority within a state defined by Jewish nationality; they are also often perceived as a security threat or a “fifth col￾umn” in the context of Israel’s armed conflicts with its surrounding Arab neighbors (Canetti-Nisim, Ariely, and Halperin 2008). Smooha (2004) shows that a sub￾stantial share of Jewish Israelis believe that PCI sup￾port terrorism and may rebel against Israel in the fu￾ture. This makes the case of exclusion of PCI poten￾tially informative for thinking about minority groups elsewhere: perceptions of security threats affect atti￾tudes toward Muslim immigrants to Western countries (Hellwig and Sinno 2016) and stereotypes of orga￾nized, even transnational, criminality are often asso￾ciated with Latin American immigrants to the United States. Furthermore, the situation of a minority group having cultural and familial ties to neighboring states can also be found elsewhere, again such as with Latino immigrants to the United States. We focus on the general issue of exclusion by Jewish Israelis toward PCI, but also on the particular ques￾tion of the attitudes of ultra-Orthodox Jews. Ultra￾Orthodox Jews are distinctive among the Jewish ma￾jority because of their religious and social traditions, including low formal education and widespread non￾participation in the workforce, resulting in a popula￾tion that is substantially poorer than the general Jew￾ish population. Furthermore, there are also barriers for cooperation between the ultra-Orthodox and the rest of the Israeli Jewish population (Enos and Gidron 2016). The ultra-Orthodox thus present an informative point of reference in their relations with the PCI be￾cause they are, arguably, close to PCI on a social hi￾erarchy. According to prominent social-psychological theories, they may therefore hold more exclusionary attitudes due to greater perceived threat to their rel￾ative status (Blumer 1958; Bobo and Hutchings 1996; Sidanius and Pratto 2001). Competition over resources is often most relevant to the low-income portion of a majority group that shares social welfare institutions with low-status minority groups (Bobo and Hutch￾ings 1996).8 Of course, less well-off subgroups of the 8 Although, notably, in the Israeli context, ultra-Orthodox and PCI are largely institutionally separated, so competition over institutions such as schools is not present. 746 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Exclusion and Cooperation in Diverse Societies hegemonic population having politically relevant ex- and under our direct supervision.Our sample includes clusionary preferences toward low-status minorities 439 subjects,all of whom are Jewish Israelis.Since has obvious parallels in the other societies(Gidron and we are interested in the implications of exclusion,we Hall 2017).For example,the tendency for poor whites choose to focus on the hegemonic majority group be- in the United States to oppose the social integration of cause that is the group with the power to exclude.We African Americans has a long been noted (Key 1949). used quotas for gender and age to generate a balanced In some analyses below,we divide the sample into sample on these covariates.While the sample is broadly secular and ultra-Orthodox based on respondents'self- representative of the Jewish population of Israel,ultra- identification.By dividing the sample,we can see Orthodox Jews are intentionally overrepresented (see whether the exclusionary preferences of the low-status Table A.1 in the Online Appendix). ultra-Orthodox group are higher than those of other Respondents were selected to participate in our Jews.Additionally,we compare preferences for exclu- study using a random walk strategy,with a partici- sion and cooperative behaviors of Jews toward PCl pation rate of about 17%.Participation took around with preferences for exclusion and cooperative behav- 40 minutes and was conducted inside participants' iors toward ultra-Orthodox and secular Jews.This al- homes.Participants were told that this research deals lows us to compare the preferences and behaviors to- with "Israeli society"and worked independently on ward a Jewish outgroup to preferences and behaviors computers provided by our fieldworkers.Compensa- toward the PCI outgroup. tion for participation was determined by randomly se- lecting the outcome of one of the economic decision- DATA AND RESEARCH DESIGN making games. Participants were first asked to play a public goods Our data was collected through lab-in-the-field experi- game with three rounds in a random order:against ments across twenty locations in Israel,with wide varia a PCI,Jewish secular,or ultra-Orthodox opposing tion in the local proportion PCI.Not only does using a player.The opposing players were all real people, 4r元 laboratory allow for careful measurement of play in the whose decisions in the public goods game,as we ex- economic games necessary for this design,but bringing plained to subjects,were recorded in advance.10 Par- the lab to the respondents-rather than the other way ticipants were shown a picture of the opposing play- & around-increases the external validity of the results in ers next to their name,age,and city of residence.The two primary ways.First,we are able to have a sample names provided a strong indicator whether the op- that is more representative of the Israeli Jewish popu- posing player was Jewish or PCI,and the distinctive lation than could be obtained when relying on univer- clothes of the ultra-Orthodox players clearly distin- sity students-a limitation that may be especially prob- guished them from Jewish secular players.Almost 95% lematic when needing variation in exclusionary inter- of our players were able to identify the ethnicity of the group attitudes.College students tend to have a distinct opposing player based on these cues. psychological profile(Sears 1986;Jones 2010),includ- The structure of our public goods game is drawn ing strong norms of equality and low levels of preju- from previous work on intergroup relations in eth- dice (Sidanius et al.2008)and their play in economic nically diverse societies (Habyarimana et al.2009). games varies substantially from the play of other pop- In each round,participants were given twenty Israeli ulations (Henrich et al.2006).Second,and of particu Shekels(NIS)and had to decide whether to cooperate lar importance for the focus of this research,stationary by sharing the full sum or defect by keeping the full sum to themselves.After announcing their decision. 8 laboratory experiments,in contrast to lab-in-the-field experiments,are limited in their ability "to inform the they were informed of the opposing player's decision. study of cooperation in social dilemmas"(Grossman which we recorded in advance.In line with the standard 2011).And as Baldassarri(2015)notes,lab-in-the-field procedure of the public goods game,payoffs were mul- experiments,especially-as in our case-when compli- tiplied by 1.5 and divided equally between the two par- mented by additional survey data,can overcome the ticipants.This means that if both sides cooperated,each limited ability of laboratory experiments to represent participant ended the game with 30 NIS.If one cooper- the contexts in which group identities and norms of co- ated and the other did not,the person who cooperated operation operate. received 15 NIS and the person who did not cooperate Fieldwork and data collection were conducted dur received 35 NIS.If both sides did not cooperate,both ing the summer of 2013 by a professional survey team remained with their initial sum of 20 NIS.Thus,as is the eys standard in the public goods game,cooperation is the Data was collected in the following cities:Ashdod,Kiryat Malachi. mutually beneficial strategy. Elad,Arad,Bet Shemesh,Kiryat Gat,Haifa,Bnei Brak,Tve- We measure cooperation as a binary variable mea- ria,Safed,Rehovot,Zichron Yaakov,Ofakim,Netivot,Modi'in Makabim-Reut,Tel Aviv.We also sampled four neighborhoods in sured by play in the game against the PCI player,coded Jerusalem:Neve Yaakov,Ramat Shlomo,City Center,and Kiryat one if the subject cooperated,zero otherwise.Over- Yovel.The share of non-Jewish (mostly PCI)population in these all,only 33%(N =147)of subjects cooperated with locations vary from 0 to 37%,according to Israel's Central Bureau the PCI player,while 61%cooperated with the secular of Statistics census data from 2008.In analysis to follow,we use non- Jewish population rather than percent PCI because Israeli Census /:sony does not include ethnicity of non-Jewish residents and some of the 10 Note that playing against pre-recorded moves of opposing players non-Jewish population may not be PCI.However,it is reasonable to has been used successfully in previous lab-in-the-field studies (Enos assume that the overwhelming share of this population is PCI. and Gidron 2016;Whitt and Wilson 2007). 747

Exclusion and Cooperation in Diverse Societies hegemonic population having politically relevant ex￾clusionary preferences toward low-status minorities has obvious parallels in the other societies (Gidron and Hall 2017). For example, the tendency for poor whites in the United States to oppose the social integration of African Americans has a long been noted (Key 1949). In some analyses below, we divide the sample into secular and ultra-Orthodox based on respondents’ self￾identification. By dividing the sample, we can see whether the exclusionary preferences of the low-status ultra-Orthodox group are higher than those of other Jews. Additionally, we compare preferences for exclu￾sion and cooperative behaviors of Jews toward PCI with preferences for exclusion and cooperative behav￾iors toward ultra-Orthodox and secular Jews. This al￾lows us to compare the preferences and behaviors to￾ward a Jewish outgroup to preferences and behaviors toward the PCI outgroup. DATA AND RESEARCH DESIGN Our data was collected through lab-in-the-field experi￾ments across twenty locations in Israel, with wide varia￾tion in the local proportion PCI.9 Not only does using a laboratory allow for careful measurement of play in the economic games necessary for this design, but bringing the lab to the respondents—rather than the other way around—increases the external validity of the results in two primary ways. First, we are able to have a sample that is more representative of the Israeli Jewish popu￾lation than could be obtained when relying on univer￾sity students—a limitation that may be especially prob￾lematic when needing variation in exclusionary inter￾group attitudes. College students tend to have a distinct psychological profile (Sears 1986; Jones 2010), includ￾ing strong norms of equality and low levels of preju￾dice (Sidanius et al. 2008) and their play in economic games varies substantially from the play of other pop￾ulations (Henrich et al. 2006). Second, and of particu￾lar importance for the focus of this research, stationary laboratory experiments, in contrast to lab-in-the-field experiments, are limited in their ability “to inform the study of cooperation in social dilemmas” (Grossman 2011). And as Baldassarri (2015) notes, lab-in-the-field experiments, especially—as in our case—when compli￾mented by additional survey data, can overcome the limited ability of laboratory experiments to represent the contexts in which group identities and norms of co￾operation operate. Fieldwork and data collection were conducted dur￾ing the summer of 2013 by a professional survey team 9 Data was collected in the following cities: Ashdod, Kiryat Malachi, Elad, Arad, Bet Shemesh, Kiryat Gat, Haifa, Bnei Brak, Tve￾ria, Safed, Rehovot, Zichron Yaakov, Ofakim, Netivot, Modi’in￾Makabim-Reut, Tel Aviv. We also sampled four neighborhoods in Jerusalem: Neve Yaakov, Ramat Shlomo, City Center, and Kiryat Yovel. The share of non-Jewish (mostly PCI) population in these locations vary from 0 to 37%, according to Israel’s Central Bureau of Statistics census data from 2008. In analysis to follow, we use non￾Jewish population rather than percent PCI because Israeli Census does not include ethnicity of non-Jewish residents and some of the non-Jewish population may not be PCI. However, it is reasonable to assume that the overwhelming share of this population is PCI. and under our direct supervision. Our sample includes 439 subjects, all of whom are Jewish Israelis. Since we are interested in the implications of exclusion, we choose to focus on the hegemonic majority group be￾cause that is the group with the power to exclude. We used quotas for gender and age to generate a balanced sample on these covariates.While the sample is broadly representative of the Jewish population of Israel, ultra￾Orthodox Jews are intentionally overrepresented (see Table A.1 in the Online Appendix). Respondents were selected to participate in our study using a random walk strategy, with a partici￾pation rate of about 17%. Participation took around 40 minutes and was conducted inside participants’ homes. Participants were told that this research deals with “Israeli society” and worked independently on computers provided by our fieldworkers. Compensa￾tion for participation was determined by randomly se￾lecting the outcome of one of the economic decision￾making games. Participants were first asked to play a public goods game with three rounds in a random order: against a PCI, Jewish secular, or ultra-Orthodox opposing player. The opposing players were all real people, whose decisions in the public goods game, as we ex￾plained to subjects, were recorded in advance.10 Par￾ticipants were shown a picture of the opposing play￾ers next to their name, age, and city of residence. The names provided a strong indicator whether the op￾posing player was Jewish or PCI, and the distinctive clothes of the ultra-Orthodox players clearly distin￾guished them from Jewish secular players. Almost 95% of our players were able to identify the ethnicity of the opposing player based on these cues. The structure of our public goods game is drawn from previous work on intergroup relations in eth￾nically diverse societies (Habyarimana et al. 2009). In each round, participants were given twenty Israeli Shekels (NIS) and had to decide whether to cooperate by sharing the full sum or defect by keeping the full sum to themselves. After announcing their decision, they were informed of the opposing player’s decision, which we recorded in advance. In line with the standard procedure of the public goods game, payoffs were mul￾tiplied by 1.5 and divided equally between the two par￾ticipants.This means that if both sides cooperated, each participant ended the game with 30 NIS. If one cooper￾ated and the other did not, the person who cooperated received 15 NIS and the person who did not cooperate received 35 NIS. If both sides did not cooperate, both remained with their initial sum of 20 NIS. Thus, as is the standard in the public goods game, cooperation is the mutually beneficial strategy. We measure cooperation as a binary variable mea￾sured by play in the game against the PCI player, coded one if the subject cooperated, zero otherwise. Over￾all, only 33% (N = 147) of subjects cooperated with the PCI player, while 61% cooperated with the secular 10 Note that playing against pre-recorded moves of opposing players has been used successfully in previous lab-in-the-field studies (Enos and Gidron 2016; Whitt and Wilson 2007). 747 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Ryan D.Enos and Noam Gidron FIGURE 1.Social Distance by Group ☐UO toward Secular Secular toward UO 吕 UO toward Arabs g Secular toward Arabs .0 ds 8 relative neighbor friend coworker citizen' visitor one elati coworker neighbo citiz visitor ne 4号 (a)ultra-Orthodox (UO) (b)Secular player(N=268)and 63%cooperated with the ultra- Because social distance is rarely used in political sci- & Orthodox player(N=277).11 ence literature,this will help us to interpret the mean- After completing the experimental games,respon- ing of the variable and its relationship with coopera- dents were asked a series of demographic survey tion. questions.They were then asked for their opinions In Figure 1,we show the distribution in our sam- about intergroup relations in Israel,including how ple of social distance attitudes among ultra-Orthodox they would position different outgroup members- Jews toward secular Jews and PCI,and secular Jews including PCI-along the social distance scale,using toward ultra-Orthodox Jews and PCI.12 The distribu- the following wording:"Below are some groups of peo- tion of Jews'preferred social distance from PCI is strik- ple in Israel.Look at each of them and say which is ing:only a minority of Israeli Jews,either secular or the closest relationship you would find acceptable for ultra-Orthodox,expresses a willingness to have even each group.For example,if you would accept some- minimal interpersonal relationships with PCI.The high one from a group living on your street,but not as a share of ultra-Orthodox Jews who would prefer PCI to close friend,then you would choose neighbors."Re- not be citizens of Israel is especially noteworthy:over spondents were asked to choose from relative,friend. 60%of ultra-Orthodox respondents would prefer not neighbor,coworker,citizen,visitor,and none.We also to admit PCI to Israel at all-not even as visitors.Less asked participants about the trustworthiness of differ- than 10%would event accept PCI as coworkers,and ent groups in Israel,including PCl,using the following levels of acceptance for closer relationships are vanish- wording:"Below are some groups of people in Israel. ingly small. For each,please mark how much you trust people from Notably,for both ultra-Orthodox and secular Jews, that group.”Possible responses were“none,”“little,” the distribution of exclusion toward a Jewish outgroup “some,”and“alot." is starkly different.Even though many secular Jews have marked hostility toward ultra-Orthodox (Enos RESULTS and Gidron 2016).secular Jews are far more accept- ing of this group than of the PCI outgroup,with a ma- Before exploring the relationship between exclusion- jority willing to accept ultra-Orthodox as neighbor or ary preferences,as measured by social distance and co closer but a majority not willing to accept PCI in any operation,as measured by the public goods game,we personal relationship,not even as a coworker.The dif- first examine the distribution of preferences for exclu- ferences between levels of exclusion toward the Jewish sion and how social distance should be characterized. 11 We also checked for ordering effects of when the subject encoun. 12 Note that we are subsetting the data here to only those respon- dents who self-identify as either secular or ultra-Orthodox.We ac- tered the PCI player on the probability of cooperation and see no ef- count for the full range of Jewish religious identities in Israel-ultra- fects.Percent cooperating with PCI in first round is 34,second round Orthodox.religious,traditional,and secular-in later analyses.While is 30,third round is 35.The T-statistic for a test of difference of means we also included a category for "anti-religious,"since only four re between rounds 1 and 2,T=0.75;rounds 1 and 3,T=-0.18;and spondents identified as such,we merged this category with "secular' rounds 2 and 3,T=-0.97 in the analyses below. 748

Ryan D. Enos and Noam Gidron FIGURE 1. Social Distance by Group Share of respondents UO toward Secular UO toward Arabs relative friend neighbor coworker citizen visitor none 0.0 0.1 0.2 0.3 0.4 0.5 0.6 (a) ultra-Orthodox (UO) Share of respondents Secular toward UO Secular toward Arabs relative friend neighbor coworker citizen visitor none 0.0 0.1 0.2 0.3 0.4 0.5 0.6 (b) Secular player (N = 268) and 63% cooperated with the ultra￾Orthodox player (N = 277).11 After completing the experimental games, respon￾dents were asked a series of demographic survey questions. They were then asked for their opinions about intergroup relations in Israel, including how they would position different outgroup members— including PCI—along the social distance scale, using the following wording: “Below are some groups of peo￾ple in Israel. Look at each of them and say which is the closest relationship you would find acceptable for each group. For example, if you would accept some￾one from a group living on your street, but not as a close friend, then you would choose neighbors.” Re￾spondents were asked to choose from relative, friend, neighbor, coworker, citizen, visitor, and none. We also asked participants about the trustworthiness of differ￾ent groups in Israel, including PCI, using the following wording: “Below are some groups of people in Israel. For each, please mark how much you trust people from that group.” Possible responses were “none,” “little,” “some,” and “a lot.” RESULTS Before exploring the relationship between exclusion￾ary preferences, as measured by social distance and co￾operation, as measured by the public goods game, we first examine the distribution of preferences for exclu￾sion and how social distance should be characterized. 11 We also checked for ordering effects of when the subject encoun￾tered the PCI player on the probability of cooperation and see no ef￾fects. Percent cooperating with PCI in first round is 34, second round is 30, third round is 35. The T-statistic for a test of difference of means between rounds 1 and 2, T = 0.75; rounds 1 and 3, T = −0.18; and rounds 2 and 3, T = −0.97. Because social distance is rarely used in political sci￾ence literature, this will help us to interpret the mean￾ing of the variable and its relationship with coopera￾tion. In Figure 1, we show the distribution in our sam￾ple of social distance attitudes among ultra-Orthodox Jews toward secular Jews and PCI, and secular Jews toward ultra-Orthodox Jews and PCI.12 The distribu￾tion of Jews’ preferred social distance from PCI is strik￾ing: only a minority of Israeli Jews, either secular or ultra-Orthodox, expresses a willingness to have even minimal interpersonal relationships with PCI. The high share of ultra-Orthodox Jews who would prefer PCI to not be citizens of Israel is especially noteworthy: over 60% of ultra-Orthodox respondents would prefer not to admit PCI to Israel at all—not even as visitors. Less than 10% would event accept PCI as coworkers, and levels of acceptance for closer relationships are vanish￾ingly small. Notably, for both ultra-Orthodox and secular Jews, the distribution of exclusion toward a Jewish outgroup is starkly different. Even though many secular Jews have marked hostility toward ultra-Orthodox (Enos and Gidron 2016), secular Jews are far more accept￾ing of this group than of the PCI outgroup, with a ma￾jority willing to accept ultra-Orthodox as neighbor or closer but a majority not willing to accept PCI in any personal relationship, not even as a coworker. The dif￾ferences between levels of exclusion toward the Jewish 12 Note that we are subsetting the data here to only those respon￾dents who self-identify as either secular or ultra-Orthodox. We ac￾count for the full range of Jewish religious identities in Israel—ultra￾Orthodox, religious, traditional, and secular—in later analyses.While we also included a category for “anti-religious,” since only four re￾spondents identified as such, we merged this category with “secular” in the analyses below. 748 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Exclusion and Cooperation in Diverse Societies outgroup and toward PCI suggests that hostility does attitudes,on the other hand.is less clear.While there not necessarily result in exclusionary attitudes and that are prominent theories of ideology that posit that it is exclusionary attitudes are an independent and conse- a deeply rooted dispositional (Jost et al.2003)or even quential attitude,separate from other types of affect. physiological (Oxley et al.2008)trait,other prominent scholarship sees issue positions as more malleable and Exclusionary Preferences as Symbolic subject to elite influence (Zaller 1992).As such,we Attitudes treat the relationship between ideology and exclusion- We speculated that social distance,as an indicator of ary attitudes as weaker evidence for the symbolic na exclusionary attitudes,is a symbolic attitude.Given the ture of these attitudes centrality of the PCI in Israeli political discourse,this In Table 1,we present the results of OLS regres- might mean that attitudes toward PCI influences public sions,with social distance regressed on ideology,edu- opinion over a range of topics.A common quantitative cation,religiosity,and other demographic variables.3 test of whether an attitude is symbolic is to see if it is Political ideology,education,and religiosity all appear to be strongly related to social distance,with more 4 a significant predictor of attitudes and behaviors when included in multiple regression analysis with other con- right-wing,more religious,and less-educated subjects siderations(Sears et al.1997;Sears and Henry 2003). expressing more exclusionary preferences. We take this up below. In Model 2,we also include a measurement of self- 4 First,we test for another marker of symbolic at- reported social interactions with PCI.This variable titudes:whether social distance attitudes are related takes values of“daily,.”“weekly,”“monthly,”“yearly,” to slow-moving variables that tend to be socialized or"never."The results in this model suggest that those early in life.In an ideal study,we would test the sta- who interact with PCI more frequently are also likely to bility of these attitudes using longitudinal panel data report lower exclusionary attitudes.4 Of course,there over the course of a lifetime,starting with early adult is a question of endogeneity:it might be that those who are more accepting of PCI are more willing to inter- 4号 socialization-however.such data is unavailable.As such,we use a method common in the literature and act with them,rather than the other way around.Nev- turn to whether these preferences are predicted by ertheless,note that with the inclusion of this variable. 'asn traits that were likely established early in life.The liter- ideology and religion remain strong predictors of so- ature on attitudes toward outgroup minorities in gen- cial distance,suggesting that social distance is rooted, eral and in Israel in particular points to several such at least to some degree,in slow-moving individual-level factors: features.While exclusionary attitudes do appear to be influenced by social interactions,such variables do not explain all of the variation. Religiosity:Stronger religiosity may strengthen and The results in Table 1 are also consistent with the- reinforce ingroup identity and outgroup exclusion- ories that prejudice can spring from threats to status. ary preferences,especially when national dividing Ultra-Orthodox Jews,who share with the PCI several lines follow religious cleavages,as in Israel.Con- characteristics of low status,express strong social dis- versely,religious beliefs may generate a sense of tance from PCI. solidarity toward the less well-off,which under cer- We display predicted values of exclusionary prefer- tain circumstances may also encompass minority ences from Model 1 in Table 1 in Figure 2.In these outgroups(Ben-Nun Bloom,Arikan,and Courte- figures,we predict values for a secular,38-year-old manche 2015:Johnson.Rowatt.and LaBouff 2010: Sephardi male with average income,high-school edu- Knoll 2009) cation,and center-right political identity(5 on the 1-7 Political ideology:Right-wing ideological self- scale).Religiosity,ideology,and education vary in sub- identification.which is likely to be linked with figures a,b,and c,respectively.Note that in all these strong national sentiments,may be associated with figures,predicted values range only between coworker stronger exclusionary preferences (Golder 2003; and visitor,reflecting the high levels of exclusionary Semyonov,Raijman,and Gorodzeisky 2006) preferences in our sample of Israeli Jews. Education:Higher education is associated with Having established that social distance is predicted greater cultural openness(Stubager 2008,2009), by slow-moving and stable characteristics of a person, which may lead to greater openness toward out- groups.Indeed,Pedahzur,Halperin,and Canetti (2007)find that in Israel,higher education-more 13Ordered logit regression provides substantively similar results We measure levels of education on a four-level scale:primary educa- than other measures of socioeconomic status such tion (the lowest),high school education,undergraduate degree,and as employment-is associated with lower social graduate degree (highest level of education,which also serves as the distance from minority groups. reference category).Political ideology is measured on a continuous seven-point scale,ranging from 1(far left)to 7(far right),using the following question:"Many talk about left and right in politics.Use Education is,for most people,stable after adoles- the slider to show where you are on the left to right of politics.You cence or early adulthood.Religiosity also tends to be can put yourself all the way to the left,all the way to right,in the stable in the Israeli context(Coopermanm,Sahgal,and center,or somewhere else in-between."The other variables in the Schiller 2016).so it is reasonable to assume these vari- model are age,foreign born(0/1).income(categorically asisstandard ables are established prior to attitudes about exclusion om poteroto decreswrb rej The relationship between ideology and exclusionary dice,see Schroeder and Risen (2015). 749

Exclusion and Cooperation in Diverse Societies outgroup and toward PCI suggests that hostility does not necessarily result in exclusionary attitudes and that exclusionary attitudes are an independent and conse￾quential attitude, separate from other types of affect. Exclusionary Preferences as Symbolic Attitudes We speculated that social distance, as an indicator of exclusionary attitudes, is a symbolic attitude. Given the centrality of the PCI in Israeli political discourse, this might mean that attitudes toward PCI influences public opinion over a range of topics. A common quantitative test of whether an attitude is symbolic is to see if it is a significant predictor of attitudes and behaviors when included in multiple regression analysis with other con￾siderations (Sears et al. 1997; Sears and Henry 2003). We take this up below. First, we test for another marker of symbolic at￾titudes: whether social distance attitudes are related to slow-moving variables that tend to be socialized early in life. In an ideal study, we would test the sta￾bility of these attitudes using longitudinal panel data over the course of a lifetime, starting with early adult socialization—however, such data is unavailable. As such, we use a method common in the literature and turn to whether these preferences are predicted by traits that were likely established early in life. The liter￾ature on attitudes toward outgroup minorities in gen￾eral and in Israel in particular points to several such factors: Religiosity: Stronger religiosity may strengthen and reinforce ingroup identity and outgroup exclusion￾ary preferences, especially when national dividing lines follow religious cleavages, as in Israel. Con￾versely, religious beliefs may generate a sense of solidarity toward the less well-off, which under cer￾tain circumstances may also encompass minority outgroups (Ben-Nun Bloom, Arikan, and Courte￾manche 2015; Johnson, Rowatt, and LaBouff 2010; Knoll 2009). Political ideology: Right-wing ideological self￾identification, which is likely to be linked with strong national sentiments, may be associated with stronger exclusionary preferences (Golder 2003; Semyonov, Raijman, and Gorodzeisky 2006). Education: Higher education is associated with greater cultural openness (Stubager 2008, 2009), which may lead to greater openness toward out￾groups. Indeed, Pedahzur, Halperin, and Canetti (2007) find that in Israel, higher education—more than other measures of socioeconomic status such as employment—is associated with lower social distance from minority groups. Education is, for most people, stable after adoles￾cence or early adulthood. Religiosity also tends to be stable in the Israeli context (Coopermanm, Sahgal, and Schiller 2016), so it is reasonable to assume these vari￾ables are established prior to attitudes about exclusion. The relationship between ideology and exclusionary attitudes, on the other hand, is less clear. While there are prominent theories of ideology that posit that it is a deeply rooted dispositional (Jost et al. 2003) or even physiological (Oxley et al. 2008) trait, other prominent scholarship sees issue positions as more malleable and subject to elite influence (Zaller 1992). As such, we treat the relationship between ideology and exclusion￾ary attitudes as weaker evidence for the symbolic na￾ture of these attitudes. In Table 1, we present the results of OLS regres￾sions, with social distance regressed on ideology, edu￾cation, religiosity, and other demographic variables.13 Political ideology, education, and religiosity all appear to be strongly related to social distance, with more right-wing, more religious, and less-educated subjects expressing more exclusionary preferences. In Model 2, we also include a measurement of self￾reported social interactions with PCI. This variable takes values of “daily,” “weekly,” “monthly,” “yearly,” or “never.” The results in this model suggest that those who interact with PCI more frequently are also likely to report lower exclusionary attitudes.14 Of course, there is a question of endogeneity: it might be that those who are more accepting of PCI are more willing to inter￾act with them, rather than the other way around. Nev￾ertheless, note that with the inclusion of this variable, ideology and religion remain strong predictors of so￾cial distance, suggesting that social distance is rooted, at least to some degree,in slow-moving individual-level features. While exclusionary attitudes do appear to be influenced by social interactions, such variables do not explain all of the variation. The results in Table 1 are also consistent with the￾ories that prejudice can spring from threats to status. Ultra-Orthodox Jews, who share with the PCI several characteristics of low status, express strong social dis￾tance from PCI. We display predicted values of exclusionary prefer￾ences from Model 1 in Table 1 in Figure 2. In these figures, we predict values for a secular, 38-year-old Sephardi male with average income, high-school edu￾cation, and center-right political identity (5 on the 1–7 scale). Religiosity, ideology, and education vary in sub￾figures a, b, and c, respectively. Note that in all these figures, predicted values range only between coworker and visitor, reflecting the high levels of exclusionary preferences in our sample of Israeli Jews. Having established that social distance is predicted by slow-moving and stable characteristics of a person, 13 Ordered logit regression provides substantively similar results.We measure levels of education on a four-level scale: primary educa￾tion (the lowest), high school education, undergraduate degree, and graduate degree (highest level of education, which also serves as the reference category). Political ideology is measured on a continuous seven-point scale, ranging from 1 (far left) to 7 (far right), using the following question: “Many talk about left and right in politics. Use the slider to show where you are on the left to right of politics. You can put yourself all the way to the left, all the way to right, in the center, or somewhere else in-between.” The other variables in the model are age, foreign born (0/1),income (categorically as is standard in Israeli surveys), and ethnicity. 14 On the potential of interactions to decrease Jewish-Arab preju￾dice, see Schroeder and Risen (2015). 749 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Ryan D.Enos and Noam Gidron we next examine whether social distance is a strong TABLE 1.Social Distance with PCI predictor of policy preferences in the face of competing considerations,as would be predicted if social distance Dependent variable: 00！o is a symbolic attitude that dominates other attitudes Social Distance We focus on respondents'perceptions of whether PCI receive too much or too little from the government. (1) 2) Social distance overwhelmingly predicts perceptions of governmental spending on PCI,as shown in Table A.2 Ideology 0.314* 0.276* in the Online Appendix,even when accounting for de- (0.055) (0.052) mographic and other related factors such as political Religiosity (baseline:religious) Secular -0.425* -0.353 ideology.As is expected with symbolic attitudes,so- (0.256) (0.246) cial distance appears to dominate other considerations Traditional 0.236 0.287 when forming policy preferences. (0.285) (0.272) These results strongly suggest that social distance is Ultra-Orthodox 0.650* 0.495* a symbolic attitude.It is associated with slow-moving (0.241) (0.232) variables that are socialized early in life and domi- Education(baseline:graduate) nates other considerations in forming policy prefer- Primary school 1.013* 0.634 ences.This makes us suspect that exclusionary atti- (0.466) (0.458) tudes,as measured by social distance,influence polit- High school 0.368 0.250 (0.337 (0.321) ical behavior,including even costly behaviors.We now 0.570* turn to examine the relationship between exclusionary Undergrad 0.430 (0.340) (0.325) preferences and cooperative behavior and examine the Income (baseline:average) predictive strength of social distance in the face of mit- Very low income -0.002 0.048 igating factors. 4 (0.202) (0.195) Low income -0.096 -0.180 (0.216) (0.208) Preferences for Exclusion and Cooperation High income -0.474* -0.384 To examine the relationship between exclusionary (0.270) (0.260) preferences and cooperation,we created a dichoto- Very high income 0.151 -0.108 (0.402) mous variable of high and low exclusion based on (0.387) Age -0.003 -0.004 whether or not the subject would accept an outgroup (0.005) (0.005) member as a coworker or closer,meaning that respon- Foreign Born -0.326* -0.246 dents who would accept PCI as coworkers,neighbors, (0.197)） (0.188) friends,or relatives are coded as low exclusionary pref- Male -0.519** -0.376* erence and everyone else is coded as high exclusionary (0.149) (0.144) preference.15 Ethnicity(baseline:Ashkenazy) Looking at cooperation as a function this dichoto- Mixed -0.358 -0.478* mous variable,a t-test for difference of means yields (0.283) (0.269) Other 0.232 u=0.19,t=3.26,p <0.001,indicating that high exclu- -0.186 (0.456) sion subjects are significantly less likely to cooperate (0.438) Sephardic 0.132 0.137 with PCI(cooperation o =0.47,Cohen's D=0.40).To (0.163) (0.157) test whether this relationship will hold when control- Interactions (baseline:day) ling for other variables that may explain cooperation, Week 0.469 we use social distance as a predictor of cooperation in (0.311) multivariate regression.The coefficient estimates from Month 0.931* this logit regression are presented in Table 2.16 Strong (0.296) exclusionary attitudes are highly predictive of non- Year 1.166* cooperation,both with and without individual-level co- (0.312) Never 1.486* variates (Columns 1 and 2,respectively).Notably,in (0.259) Constant 3.846** 3.046* 15 As can be seen in Figure 2,key demographic variables are also (0.556) (0.560) related to moves across these two levels of social distance.Note that this dichotomy is roughly at the midpoint of the scale.The median servations 375 372 of the distribution is at visitor.If we chose to dichotomize our vari R2 0.289 0.370 able here,the results we report below would show an even stronger Adjusted R2 0.255 0.332 relationship between social distance and cooperation.In Table A4 in the Online Appendix,we report results of a regression with six cate- Note:*p<0.1;*tp<0.05:*p<0.01 gories of the social distance scale(we combine relative and neighbor L OLS regressions of social distance with PCI(range 1-7)on because less than 1%of respondents chose the relative category) individual-level variables(Column 1)and frequency of interac- The regression estimates from this model have similar substantive tions with PCI(Column 2) the mem sions with the missing values of covariates imputed and find no sig- nificant change in the results. 750

Ryan D. Enos and Noam Gidron TABLE 1. Social Distance with PCI Dependent variable: Social Distance (1) (2) Ideology 0.314∗∗∗ 0.276∗∗∗ (0.055) (0.052) Religiosity (baseline: religious) Secular − 0.425∗ − 0.353 (0.256) (0.246) Traditional 0.236 0.287 (0.285) (0.272) Ultra-Orthodox 0.650∗∗∗ 0.495∗∗ (0.241) (0.232) Education (baseline: graduate) Primary school 1.013∗∗ 0.634 (0.466) (0.458) High school 0.368 0.250 (0.337) (0.321) Undergrad 0.570∗ 0.430 (0.340) (0.325) Income (baseline: average) Very low income − 0.002 0.048 (0.202) (0.195) Low income − 0.096 − 0.180 (0.216) (0.208) High income − 0.474∗ − 0.384 (0.270) (0.260) Very high income 0.151 − 0.108 (0.402) (0.387) Age − 0.003 − 0.004 (0.005) (0.005) Foreign Born − 0.326∗ − 0.246 (0.197) (0.188) Male − 0.519∗∗∗ − 0.376∗∗∗ (0.149) (0.144) Ethnicity (baseline: Ashkenazy) Mixed − 0.358 − 0.478∗ (0.283) (0.269) Other 0.232 − 0.186 (0.456) (0.438) Sephardic 0.132 0.137 (0.163) (0.157) Interactions (baseline: day) Week 0.469 (0.311) Month 0.931∗∗∗ (0.296) Year 1.166∗∗∗ (0.312) Never 1.486∗∗∗ (0.259) Constant 3.846∗∗∗ 3.046∗∗∗ (0.556) (0.560) Observations 375 372 R2 0.289 0.370 Adjusted R2 0.255 0.332 Note: ∗p < 0.1; ∗∗p < 0.05; ∗∗∗p < 0.01 OLS regressions of social distance with PCI (range 1–7) on individual-level variables (Column 1) and frequency of interac￾tions with PCI (Column 2). we next examine whether social distance is a strong predictor of policy preferences in the face of competing considerations, as would be predicted if social distance is a symbolic attitude that dominates other attitudes. We focus on respondents’ perceptions of whether PCI receive too much or too little from the government. Social distance overwhelmingly predicts perceptions of governmental spending on PCI, as shown in Table A.2 in the Online Appendix, even when accounting for de￾mographic and other related factors such as political ideology. As is expected with symbolic attitudes, so￾cial distance appears to dominate other considerations when forming policy preferences. These results strongly suggest that social distance is a symbolic attitude. It is associated with slow-moving variables that are socialized early in life and domi￾nates other considerations in forming policy prefer￾ences. This makes us suspect that exclusionary atti￾tudes, as measured by social distance, influence polit￾ical behavior, including even costly behaviors. We now turn to examine the relationship between exclusionary preferences and cooperative behavior and examine the predictive strength of social distance in the face of mit￾igating factors. Preferences for Exclusion and Cooperation To examine the relationship between exclusionary preferences and cooperation, we created a dichoto￾mous variable of high and low exclusion based on whether or not the subject would accept an outgroup member as a coworker or closer, meaning that respon￾dents who would accept PCI as coworkers, neighbors, friends, or relatives are coded as low exclusionary pref￾erence and everyone else is coded as high exclusionary preference.15 Looking at cooperation as a function this dichoto￾mous variable, a t-test for difference of means yields μ = 0.19, t = 3.26, p < 0.001, indicating that high exclu￾sion subjects are significantly less likely to cooperate with PCI (cooperation σ = 0.47, Cohen’s D = 0.40). To test whether this relationship will hold when control￾ling for other variables that may explain cooperation, we use social distance as a predictor of cooperation in multivariate regression. The coefficient estimates from this logit regression are presented in Table 2. 16 Strong exclusionary attitudes are highly predictive of non￾cooperation, both with and without individual-level co￾variates (Columns 1 and 2, respectively). Notably, in 15 As can be seen in Figure 2, key demographic variables are also related to moves across these two levels of social distance. Note that this dichotomy is roughly at the midpoint of the scale. The median of the distribution is at visitor. If we chose to dichotomize our vari￾able here, the results we report below would show an even stronger relationship between social distance and cooperation. In Table A4 in the Online Appendix, we report results of a regression with six cate￾gories of the social distance scale (we combine relative and neighbor because less than 1% of respondents chose the relative category). The regression estimates from this model have similar substantive and statistical significance to the logit model reported in Table 2. 16 In Table A.3 in the Online Appendix, we estimate these regres￾sions with the missing values of covariates imputed and find no sig￾nificant change in the results. 750 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266

Exclusion and Cooperation in Diverse Societies FIGURE 2.Religiosity,ideology,education,and social distance -0001v00006/AL0LoL none none none visitor visitor visitor citizen citizen citizen coworker coworker coworker neighbor neighbor neighbor friend friend friend relative relative relative 4 cular s religious trad. e (a)Religiosity (b)Political Ideology (c)Education Subfigure (a)presents the predicted values for social distance by varying levels of religiosity:secular,traditional,religion,and ultra Orthodox(UO)respondents.Subfigure (b)presents the predicted values for social distance by ideology,ranging from far left(1) to far right(7).Subfigure (c)presents the predicted values for social distance by levels of education:primary school,high school, undergraduate degree,and graduate degree. these regressions,we include a number of variables lic goods game.8 For instance,Zussman (2013)finds 4号元 that might also influence economic decision-making, that Jews discriminate against PCI in the Israeli car including gender and income,and yet the influence of market due to concerns over trustworthiness.We there- this basic exclusionary attitude remains large and sta- fore include a binary variable for respondents'percep- tistically significant. tions of whether PCI can be trusted.with and with- Not only do exclusionary preferences powerfully out individual level covariates in Table 2(Columns 3 predict cooperation,suggesting that people who ad- and 4 respectively).19 Again,exclusion predicts a lack vocate for the exclusion of minority groups will not of cooperation,suggesting that the failure to cooper- subsequently come to cooperate,but a preference for ate is not rooted merely in statistical discrimination exclusion seems to dominate other attitudes,as is ex- based on outgroup stereotypes,but in exclusionary pected if exclusionary preferences are symbolic atti- preferences. tudes.Even with control variables,moving from low Second,we examine whether repeated interactions exclusion to high exclusion reduces the predicted prob- with outgroups may nudge individuals with a distaste ability of cooperation by 14.6 percentage points [95% for the outgroup toward cooperation,as argued by CI:-0.287:-0.003].Given that only 33%of subjects the important work of Axelrod (2006)on the evo- chose to cooperate with the PCI player,this represents lution of cooperation.By Axelrod's logic,the pay- an over 40%decrease in the average probability of co- offs for cooperation change as the frequency of inter- operation. action increases,making cooperation a more attrac- tive strategy.Thus,we should expect the influence of Potential Mitigating Factors in the exclusionary attitudes to become weaker as interac- Preferences-Behavior Link tions increase.To test this,we interact the binary so- cial distance variable with respondents'self-reported We now turn to investigate the power of exclusionary attitudes to predict behavior in the face of possibly mit- frequency of interaction with PCI.The results,pre- sented in Table A.6 in the Online Appendix,suggest igating covariates.If exclusionary preferences are sym that repeated interactions across groups do not sig- bolic attitudes,their relationship with behaviors should nificantly moderate the relationship between preju- not be affected by changes in other non-symbolic at- titudes or contexts.We focus on three such factors dice and lack of cooperation,again suggesting that,as would be expected of a symbolic attitude,preferences that may affect cooperation:stereotypes of the trust- eys worthiness of outgroup members,repeated interac- for exclusion are slow-moving and not immediately tions with outgroup members,and the local residential responsive to changes in other variables,even poten- environment.17 tially powerful influences such as interpersonal inter- First,it could be that lack of cooperation stems not actions.We note that measuring interpersonal contact through self-reports is standard in the literature (Islam from symbolic exclusionary preferences toward PCI but rather from general stereotypes that PCI are un- trustworthy and therefore likely to defect in the pub- 18 For an extended discussion of the role of general trust in cooper- ative games,see Yamagishi and Cook (1993). 17 The variablesexplored in this section correlate with social distance 19 We dichotomized"Trust in PCI"to be consistent with our treat- ment of social distance.It is coded one for those who have some or a at the following levels:trust =-0.36,frequency of interactions with lot trust in PCI,zero otherwise.In Tables A.4 and A.5 in the Online the outgroup =-0.37,and Arab segregation =0.04. Appendix,we present results with an ordinal coding of the variable. 751

Exclusion and Cooperation in Diverse Societies FIGURE 2. Religiosity, ideology, education, and social distance



relative friend neighbor coworker citizen visitor none secular trad. religious UO (a) Religiosity relative friend neighbor coworker citizen visitor none 1234567 (b) Political Ideology



relative friend neighbor coworker citizen visitor none prim. high under. grad (c) Education Subfigure (a) presents the predicted values for social distance by varying levels of religiosity: secular, traditional, religion, and ultra￾Orthodox (UO) respondents. Subfigure (b) presents the predicted values for social distance by ideology, ranging from far left (1) to far right (7). Subfigure (c) presents the predicted values for social distance by levels of education: primary school, high school, undergraduate degree, and graduate degree. these regressions, we include a number of variables that might also influence economic decision-making, including gender and income, and yet the influence of this basic exclusionary attitude remains large and sta￾tistically significant. Not only do exclusionary preferences powerfully predict cooperation, suggesting that people who ad￾vocate for the exclusion of minority groups will not subsequently come to cooperate, but a preference for exclusion seems to dominate other attitudes, as is ex￾pected if exclusionary preferences are symbolic atti￾tudes. Even with control variables, moving from low exclusion to high exclusion reduces the predicted prob￾ability of cooperation by 14.6 percentage points [95% CI: −0.287:−0.003]. Given that only 33% of subjects chose to cooperate with the PCI player, this represents an over 40% decrease in the average probability of co￾operation. Potential Mitigating Factors in the Preferences-Behavior Link We now turn to investigate the power of exclusionary attitudes to predict behavior in the face of possibly mit￾igating covariates. If exclusionary preferences are sym￾bolic attitudes, their relationship with behaviors should not be affected by changes in other non-symbolic at￾titudes or contexts. We focus on three such factors that may affect cooperation: stereotypes of the trust￾worthiness of outgroup members, repeated interac￾tions with outgroup members, and the local residential environment.17 First, it could be that lack of cooperation stems not from symbolic exclusionary preferences toward PCI but rather from general stereotypes that PCI are un￾trustworthy and therefore likely to defect in the pub- 17 The variables explored in this section correlate with social distance at the following levels: trust =−0.36, frequency of interactions with the outgroup =−0.37, and Arab segregation =0.04. lic goods game.18 For instance, Zussman (2013) finds that Jews discriminate against PCI in the Israeli car market due to concerns over trustworthiness.We there￾fore include a binary variable for respondents’ percep￾tions of whether PCI can be trusted, with and with￾out individual level covariates in Table 2 (Columns 3 and 4 respectively).19 Again, exclusion predicts a lack of cooperation, suggesting that the failure to cooper￾ate is not rooted merely in statistical discrimination based on outgroup stereotypes, but in exclusionary preferences. Second, we examine whether repeated interactions with outgroups may nudge individuals with a distaste for the outgroup toward cooperation, as argued by the important work of Axelrod (2006) on the evo￾lution of cooperation. By Axelrod’s logic, the pay￾offs for cooperation change as the frequency of inter￾action increases, making cooperation a more attrac￾tive strategy. Thus, we should expect the influence of exclusionary attitudes to become weaker as interac￾tions increase. To test this, we interact the binary so￾cial distance variable with respondents’ self-reported frequency of interaction with PCI. The results, pre￾sented in Table A.6 in the Online Appendix, suggest that repeated interactions across groups do not sig￾nificantly moderate the relationship between preju￾dice and lack of cooperation, again suggesting that, as would be expected of a symbolic attitude, preferences for exclusion are slow-moving and not immediately responsive to changes in other variables, even poten￾tially powerful influences such as interpersonal inter￾actions. We note that measuring interpersonal contact through self-reports is standard in the literature (Islam 18 For an extended discussion of the role of general trust in cooper￾ative games, see Yamagishi and Cook (1993). 19 We dichotomized “Trust in PCI” to be consistent with our treat￾ment of social distance. It is coded one for those who have some or a lot trust in PCI, zero otherwise. In Tables A.4 and A.5 in the Online Appendix, we present results with an ordinal coding of the variable. 751 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000266