Exclusion and Cooperation in Diverse Societies: ExperimentalEvidence from Israel

It 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.

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. 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.

CONCLUSION
 In this article, we have shown that social distance, a measure of exclusionary preferences, is strongly pre￾dictive of cooperation in a public goods game. Further￾more, we have shown that this tendency likely springs from basic prejudices, rooted in slow-moving features of a person’s socialization. Even within the same insti￾tutional context and when holding other variables con￾stant, those with more exclusionary attitudes were less likely to choose a cooperative strategy when facing an outgroup member.
 Our findings suggest potentially important future re￾search agendas. We have focused on cooperative be￾haviors because they are often argued to be an impor￾tant ingredient of successful diverse societies. But, of course, other consequential behaviors may be related to exclusionary attitudes. These may include more ba￾sic discriminatory behaviors, such as those reflected in other-regarding preferences (Habyarimana et al. 2007; Enos and Gidron 2016) or that manifest in socially consequential behaviors, such as hiring discrimination (Bertrand and Mullainathan 2004). Preferences for so￾cial exclusion may also affect other types of coopera￾tion,including cooperation found outside of the labora￾tory (Hjort 2014). Furthermore, exclusionary attitudes may be correlated with other prejudicial attitudes, such as specific stereotypes.
 There is also the question of the source of exclu￾sionary attitudes. We have established that responses to the social distance scale have the characteristics of a symbolic attitude.This suggests that these attitudes will arise early in life and be stable by adulthood. A natural question then is what causes the acquisition of these at￾titudes early in life (Sears 1993). In the American con￾text, the literature suggests several sources for symbolic attitudes and how these attitudes can be influenced by some later life experiences (see, for example, Sidanius et al. (2008)). There is a rich literature on political so￾cialization from colleges (Mendelberg, McCabe, and Thal 2017), schools, and families (Jennings and Niemi 2015; Clark and Clark 1950). Further research should investigate how these and other influences shape ex￾clusionary attitudes, especially as they relate to cooperative behaviors, and how the literature developed in the United States may or may not be applicable to the socialization of symbolic attitudes in other countries.
 What do our findings say about the ability of diverse societies to provide public goods? Our results suggest that exclusionary preferences are strongly implicated in inadequate cooperation that is related to poor pub￾lic goods provision. A large faction of individuals in many countries express a desire to exclude immigrants and other minorities, but nevertheless find themselves in increasingly diverse societies as the flow of immi￾grants is unabated. Our results suggest that these in￾dividuals may be unwilling to engage with these im￾migrants as they become fellow citizens. Indeed, this could be the phenomenon described by Putnam (2007), who argued that diversity causes people to “hunker down” and avoid pro-social activities. Our findings sug￾gest that many people, as a result of socialization, may have always preferred not to engage cooperatively with their neighbors who are different from them, and the hunkering down reflects a behavioral response to the increased exposure to an outgroup.
 We have also noted that the correlations between slow-moving demographic features of an individual and exclusionary preferences suggest that social dis￾tance is unlikely to change within an individual as other features of that individual or her context change. How￾ever, this does not mean that exclusionary preferences will not change within a society across time.Notably, Parrillo and Donoghue (2005) showed that average so￾cial distance toward a number of groups in the United States, including African Americans, has become less exclusionary in the over 80 years since the social dis￾tance scale was created and even significantly less ex￾clusionary in the last 40 years.22 While this does not signal that inclusion and harmony happens quickly, it does suggest that a society’s psychological barriers to cooperation can be lowered over time.
 Of course, this does not mean that aggregate exclu￾sionary preferences will necessarily lessen with time. In Israel, in particular, that the most exclusionary prefer￾ences are found among ultra-Orthodox reinforces the potential for aggregate exclusionary attitudes to shift  in a more exclusionary direction. Ultra-Orthodox rep￾resent the fastest-growing segment of the Jewish popu￾lation of Israel and aggregate preferences for exclusion among Jewish Israelis may increase as this population grows. If these attitudes are induced by threats to posi￾tion on a social hierarchy, as suggested by some scholar￾ship, then this speaks to the need for welfare-enhancing policies and institutions in promoting harmonious atti￾tudes and cooperation.

More generally, our findings suggest that diverse so￾cieties have to directly face the challenge of exclusion￾ary attitudes to overcome barriers to cooperation, as simply stressing the material incentives for cooperation or providing counter-stereotypical information will not suffice to enhance cooperation. Rather, if these preferences are rooted in symbolic attitudes that are social￾ized early in life, then the roots of these attitudes should be addressed. As our analyses of ultra-Orthodox/PCI non-cooperation suggests, this point is also especially relevant for understanding intergroup relations among low-status groups. Because low status groups exist in every society, including those diversifying via immigra￾tion, this issue is relevant for a large variety of contexts.

In Israel in particular, attitudes among ultra￾Orthodox show little hope of a mass-level coalition of low-status groups; even though both ultra-Orthodox Jews and PCI have lower socio-economic status com￾pared to secular Jewish Israelis, and both groups face institutional barriers for inclusion in mainstream Is￾raeli society and the labor market, our results show lit￾tle prospect for the type of cooperative behavior neces￾sary to build a bottom-up political coalition that is of￾ten pointed to as a potential source of political empow￾erment for low status groups in other contexts (Brown￾ing, Marshall, and Tabb 1984; Krochmal 2016). Our findings here indicate that, despite well-intentioned calls for such coalitions, the psychology of threats to status may be a stronger guide to behavior.

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.

---

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 Table2 .

---

Strong exclusionary attitudes are highly predictive of non- cooperation, both with and without individual-level co- variates (Columns 1 and 2, respectively). Notably, in 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.