Intergroup Behavioral Strategies as Contextually
Determined: Experimental Evidence from Israel 1
Ryan D. Enos2
Noam Gidron3
1 Thanks
to Amir Gidron, Eric Nelson and the Harvard Center for Jewish Studies, Harvard
Academy for International and Area Studies, Alex Mintz, Yalli Cananni and Geocartography, Julia
Hutchinson and Harvard Global Services, Marika Beaton, Michael Gribben, Lee Cahaner, Ori
Barzel, Yotam Kaplan and Alexander Sahn
2 Department of Government, Harvard University; renos@gov.harvard.edu
3 Department of Government, Harvard University; gidron@fas.harvard.edu
Abstract
Why are the negative effects of social diversity more pronounced in some places but not
others? What are the mechanisms underlying the relationship between diversity and socially
inefficient outcomes, and why do they vary in prevalence and strength across places? Recent
experimental work in comparative politics has made significant advances in examining these
questions by testing for differences in behavioral strategies when the identity of a strategic
opponent is varied. At the same time, research in American politics, and increasingly in other
places as well, has demonstrated that individual strategies are a function of context. In this
paper, we unite these two lines of research and argue that the relationship between diversity
and socially inefficient outcomes can be understood as a basic psychological process, which
is a function of both the identity of individuals and the environment in which individuals
reside. We argue that attitudes and behaviors are shaped by the context/prejudice link:
context shapes prejudice, which shapes behaviors, which lead to social inefficiencies. What
matters about context is the ways in which groups are organized in space, and especially
the interaction between diversity and segregation, two universally applicable variables in the
study of intergroup conflict. We examine these claims using original data compiled through
multi-site lab-in-the-field experiments and survey responses collected across 20 locations in
Israel. We demonstrate that attitudes and behavioral strategies of Ultra-Orthodox and nonUltra Orthodox Jews are strongly related to prejudice and vary systematically with the local
size and segregation of social groups.
Across and within countries, social diversity is positively correlated with between-group
discriminatory attitudes and behaviors (Forbes 1997), poor governance (Easterly and Levine
1997), reduced social capital (Putnam 2007), inefficient resource distribution (Habyarimana,
Humphreys, Posner, and Weinstein 2007), lack of democratic consensus (Couzin, Ioannou,
Demirel, Gross, Torney, Hartnett, Conradt, Levin, and Leonard 2011), and violent conflict
(Lim, Metzler, and Bar-Yam 2007, Bhavnani, Donnay, Miodownik, Mor, and Helbing 2013,
Vanhanen 1999). Innovative work has examined the mechanism underlying this relationship,
from both theoretical and empirical perspectives (Alesina, Baqir, and Easterly 1999, Habyarimana et al. 2007, Habyarimana, Humphreys, Posner, and Weinstein 2009, Lieberman and
McClendon 2013). These investigations focus on behavioral strategies that are posited to be
a function of individual-level variables, such as the ethnicity of opposing players. However,
long-standing evidence from the United States (e.g. Green, Strolovitch, and Wong (1998)),
and increasingly also other countries (e.g. Kasara (2013)) suggests the relationship between
diversity and inter-group behavioral strategies and attitudes should vary, not only by the
identity of the opposing players, but also by local context.
Why are the negative consequences of diversity more pronounced in some places but not
others? What is it about the local context that affects behavior? In this paper we propose
a common theoretical framework for understanding the mechanisms linking diversity and
social mechanisms and predicting how their prevalence and strength are expected to vary
systematically across geographic units of analysis. We argue that inter-group behaviors and
attitudes can be explained as part of a class of psychological phenomena that are shaped
by the ways in which groups are organized in space. More specifically, we argue that the
interaction between diversity and residential segregation, rather than diversity per se, plays
an important role in determining intergroup affect, attitudes, and behavioral strategies, with
different consequences for minority and majority groups. Using original experimental and
survey data from Israel, we demonstrate that behavioral strategies and attitudes toward
outgroups are, at least in part, a function of the social environment in which people live.
1
The empirical section of this paper is based on a large scale multi-site lab-in-the-field
research conducted in 20 locations across Israel, focusing on the growing religious tensions
within the Jewish population. This research design combines two, usually distinct, research
strategies: survey data that bring the advantage of contextual variation and laboratory
studies that precisely measure individual behavioral strategies. This allows for a rigorous
measurement of intergroup behaviors, the psychological processes underlying these associations, and their contextual variation. The multiplicity of cross-cutting social cleavages, along
with the availability of locations with great contextual variation within relatively small geographical distance, make Israel uniquely suited for studying the relations between residential
patterns and inter-group relations.
With the findings in this paper, we make several contributions: 1) we establish a theory
of geographical context, prejudice, and intergroup behavioral strategies. We extend previous
work to model contextual variation in the mechanisms through which the correlation between
diversity and inefficient social outcomes operates. By doing so, we demonstrate that individual behavioral strategies are not only a function of the players, but the context in which
the game is played. 2) We suggest that the class of mechanisms contributing to this relationship is broader than shown in previous research. 3) We demonstrate that the mechanisms
underlying diversity and inefficient outcomes vary systematically by context, which should
serve as a caution and guide for future research in the area. 4) We provide novel methods for
controlling for self-selection when contextual variables are included in a theoretical model.
5) Finally, we provide some of the first systematic evidence for intra-Jewish relations in the
geopolitically important state of Israel. We also argue that our results have important implications for social policy in diversifying Western nations. We point to residential segregation
as an important component of intergroup relations and social efficiency.
The rest of this paper is organized as follows: first we discuss why behavior is a function
of context and how this informs the relationship between diversity and social inefficiency.
We then describe the aspects of intergroup relations in Israel that are relevant to our study.
2
We move to describe our research design and the empirical results and close by discussing
the implications of our findings for scholarship and policy.
1
The Negative Outcomes of Social Diversity
Research on the negative social, political, and economic effects of demographic diversity has
flourished in recent years. This body of literature ranges across a wide variety of theoretical approaches and empirical research designs. In addition to the topics mentioned above,
various works establish a correlation between demographic heterogeneity and a bundle of negative outcomes, such as lower generalized trust (Alesina and La Ferrara 2002, Delhey and
Newton 2005), poor distribution of public goods (Jackson 2013, Miguel and Gugerty 2005)
and low economic development (Easterly and Levine 1997, Collier 2000, Alesina, Baqir, and
Easterly 1999). In this paper, we collectively refer to these variables as socially inefficient
outcomes. These correlations are shown to hold not only cross-nationally, but also when
comparing sub-national units, as with the case of cities in the United States (Alesina and
Ferrara 2004, Costa and Kahn 2003). The relationship between diversity and socially inefficient outcomes is generally described as causal; the implicit counter-factual claim in these
studies is that a similar area with a homogeneous population should have better public goods
provision, higher levels of trust, etc.
While many studies on social diversity concentrate on ethnicity, others focus on other
types of identities. In this paper, we explore a religious cleavage and we argue that our
findings are generalizable to other salient social identities and conflict between politically
relevant groups. In discussing the relevant populations in this study, we therefore adopt the
general terminology of ingroup and outgroup.
3
1.1
Mechanisms Linking Diversity and Social Inefficiency
Our central claim is that the behavioral mechanisms underlying the relationship between
diversity and social inefficiency are a function of attitudes toward the outgroup and these
attitudes are a function of context. As such, behavior too will vary by context. In this
section we review the theoretical foundations of this claim.
We begin with the mechanisms underlying the relationship between social diversity and
social inefficiency. The literature has suggested several mechanisms underlying this relationship. Scholars have grouped these into three categories: preferences, technology, and strategy
selection (Habyarimana et al. 2009).1
Preferences can be divided into commonality of tastes and other-regarding preferences.
Commonality of tastes, in the context of governance, means tastes for public policies, such
as preferences for expenditures on types of public goods. These tastes are thought to vary
across social groups, making consensus on policy hard to achieve and, therefore, inaction by
governments and other institutions designed to overcome collective action problems. Otherregarding preferences is a function of the psychological utility derived from the welfare of
another person. With this mechanism, individuals derive more utility from the welfare of
an ingroup member than an outgroup member, making actions undertaken to benefit the
ingroup more common.
Within the technology category there are many sub-mechanisms, most relevant among
them for our study is efficacy. Efficacy is the degree to which it is easier for ingroup members
to work together, perhaps because of shared language or customs, which lowers the incentive
for cooperation across social groups.2
Finally, strategy selection also consists of a large number of sub-mechanisms, all of which
involve individuals choosing different strategies when interacting with ingroup and outgroup
1
These mechanisms were usefully described by (Habyarimana et al. 2007) and (Habyarimana et al. 2009)
and the description in this section is largely taken from these works.
2
Habyarimana et al. (2007) also discuss and test findability. Findability posits that the density of socialnetworks among ingroup members makes cooperative strategies available between ingroup members that are
not available to the outgroup. We do not examine findability in this paper.
4
members. In the language of collective action, this involves the greater tendency for ingroup
members than outgroup members to arrive at a cooperative equilibrium in a public goods
game, thereby leading to less efficient outcomes when the game is played between ingroup
and outgroup members.
Our theoretical contribution begins by noting that these mechanisms can all result from a
common psychological process. This process can generally be described as a distaste, affect,
aversion, or prejudice against the outgroup. The exact form of this process is beyond the
scope of this paper, so we adopt the general term prejudice, which describes a preference of
the ingroup over the outgroup. Scholars have already noted that the psychological processes
behind prejudice may cause the behaviors noted as mechanisms for diversity and inefficient
outcomes. For example Habyarimana et al. (2007) appeal to social-cognitive theory of Social
Identity Theory (SIT) (Tajfel 1969, Tajfel, Billig, Bundy, and Flament 1971, Tajfel et al.
1971) to explain why other-regarding preferences may favor ingroup over outgroup members.
We note that prejudice can easily be extended to explain the presence of other mechanisms. Generally, this follows from increasing evidence that affect plays a role in all political
behavior: as Lodge and Taber (2013, p.43) put it “all socio-political concepts are affectladen.” The connection between affect and political behavior likely arises because of the
complexity of political decisions, which compels people to rely on affectively based heuristics
(Sniderman, Brody, and Tetlock 1993). Importantly, Lodge and Taber (2013) and others
have demonstrated that group-based attitudes are central to these affect-laden heuristics.
This basic cognitive function can be extended to explain variation in all the mechanisms reviewed above. The tendency for individuals perceive more efficacy to working with ingroup
members can be attributed to threat induced prejudice against the outgroup, which results in
the attribution of negative stereotypes (Fiske, Cuddy, Glick, and Xu 2002) or negative utility
from associating with the member of the outgroup. Strategy selection too, as measured by
cooperative behavior, is dependent on assumptions about the behavior of the other player,
which is conditioned by stereotypes attached to the other player, i.e. because decisions such
5
as whether or not to cooperate are complex, subjects rely on heuristics such as their feelings
toward the other player. Finally, even the mechanism of tastes can spring from prejudice
because aversion to outgroup members makes a person less likely to interact and acquire a
commonality of tastes with the outgroup.
1.2
The Psychology of Geography and Intergroup Behavior
Under what conditions does diversity lead to socially inefficient outcomes? Some scholars
focus on types of diversity, distinguishing between economic and cultural differences (Baldwin
and Huber 2010) or between groups with different degrees of political relevance (Posner
2004a). Others stress the role of political-institutional arrangements, suggesting that the
effect of diversity varies between democratic and non-democratic regimes (Collier 2000) or
small and large states (Posner 2004b). Other contextual variables have been proposed as
important moderators of behavior, for example disparities in social status (Blumer 1958)
and income (Gay 2006) between groups.
Related to this scholarship, but largely working on different cases and within a different
theoretical tradition, is the large literature on the contextual determinants of intergroup
attitudes. This body of scholarship involves theory and evidence that closely resembles
the literature on diversity and inefficient outcomes: aggregate levels of the presence of an
outgroup are shown to be correlated with behavioral outcomes that are thought to operate
through intergroup prejudice. These include overt and implicit racial attitudes (Wright 1977,
Fossett and Kiecolt 1989, Fitzpatrick and Hwang 1992, Sigelman and Welch 1993, Quillian
1995, Bobo and Hutchings 1996, Taylor 1998, Oliver and Mendelberg 2000, Welch, Sigelman,
Bledsoe, and Combs 2001, Oliver and Wong 2003, Gay 2006, Oliver 2010), policy positions
(Glaser 1994, Hopkins 2010), social participation (Campbell 2006, Putnam 2007, Wright
2011), voter turnout (Key 1949, Matthews and Prothro 1963, Hill and Leighley 1999, Leighley
and Vedlitz 1999), and candidate support (Giles and Buckner 1993, Voss 1996, Carsey 1995,
Enos 2010, Spence and McClerking 2010). The predictions generated by this literature are
6
very similar to the diversity and inefficient outcomes literature: an area that is diverse will
be associated with different behaviors than a more homogeneous area. However there is one
key difference: the literature on the contextual determinants of intergroup relations models
individual behavior as not only a function of the group membership of the opposite player,
but also as a function of the context in which the game is played. In general terms, certain
contextual variables (which we explore below) can stimulate prejudice toward the outgroup.3
Here we bring context into the discussion on the behavioral mechanisms connecting diversity to social inefficiencies by making the general case that prejudicial attitudes affect
behavior toward the outgroup. While not all attitudes predict behavior, the relationship is
generally quite strong. For example, Kraus (1995) reports the results of a meta-analysis,
concluding that “attitudes significantly and substantially predict future behavior”. This
leads to the prediction that individual behavior is shaped by the the context/prejudice link:
Context shapes prejudice, which shapes behaviors, which lead to the social inefficiencies
documented in the literature.
In this paper, we begin by assessing the effect of prejudice on behaviors. Once we established this relationship, we move on to analyze the linkage between behavior and context.
We focus on two broad, geographically-based contextual variables: outgroup size and residential segregation. We build a theory from these variables that connects geography to
prejudicial attitudes and subsequently to behavior. These are attractive variables because
from which to build a general theory because they are universally applicable: segregation
and group proportions are phenomenon that are applicable to any society. These variables
are also increasingly measurable in a variety of contexts due to improvements in Geographic
Information Systems and other technologies, allowing the phenomenon to be widely studied
3
The mechanisms underlying the connection between contextual factors and outgroup affect are also much
studied and disputed. In the context of American racial politics, these include: rational responses to material
threat (Bobo 1983), competition over descriptive representation (Spence and McClerking 2010), increased
group salience increasing perceptions of intergroup difference (Enos 2013), stimulation of old-fashioned racial
stereotypes (Giles and Buckner 1993), manipulation of fear by interested elites (Key 1949), and preservation
of “white power” (Voss 1996). While a fascinating topic of research, which could also benefit from an
experimental approach, sorting through these mechanisms is unnecessary for our study.
7
(for example Ichino and Nathan (2012)).
There is general agreement that the size or relative proportion of the outgroup in a locality
causes increased prejudice, all else equal (Pettigrew and Tropp 2006).4 However, importantly
for our analysis, it has also been argued in the American context that the relationship between
outgroup proportion and prejudice is quadratic (Blalock 1967, Taylor 1998, Huffman and
Cohen 2004, Enos 2010, Stephens-Davidowitz 2013), so that when the outgroup reaches
large proportions, animosity tends to diminish.
Additionally, recent scholarship has also begun to focus on the effects of spatial segregation, independent of group size, as a variable that directly affects prejudice, in both
the U.S. (Baybeck 2006, Enos 2010, Uslaner 2012, Zingher and Steen forthcoming) and
comparative contexts (Kasara 2012, Bhavnani et al. 2013, Koopmans and Veit 2013). The
prediction in this literature is that the spatial segregation of groups (measured as residential segregation) increases prejudice. A range of mechanisms has been proposed to explain
this relationship, including a lack of interaction with the outgroup leading to reduced trust
(Uslaner 2012, Kasara 2012) or segregation serving as a heuristic for intergroup difference
that leads to stereotyping (Enos 2011). The relationship between segregation and prejudice predicts that the discriminatory behavioral strategies proposed as mechanisms linking
diversity and social inefficiency should operate more strongly in more segregated localities.
These insights from the literature on contextual attitudes offer a key prediction for the
study of the mechanisms linking diversity and inefficient social outcomes: the operation
of the relevant mechanisms should vary predictably with the proportion of the outgroup
and segregation in the local population. As we discussed above, the behaviors toward the
outgroup can be thought of as a function of prejudice. If this is the case, then many, if
not all, of the relevant mechanisms should be contextually dependent. For instance, as
we demonstrate below, in some contexts we find very strong evidence that subjects prefer
working with ingroup members on a task. However, this holds only for some locations but
4
We note that the finer details of this point are far from settled in the literature (Oliver 2010, Enos 2013)
8
not others. This pattern is predicted by the relationship between context and prejudice that
will cause a simple aversion to working with outgroup members. And we find the greatest
levels of ingroup preference in the areas predicted to have the highest levels of anti-outgroup
prejudice. This contextual variation holds true for tests of all the classes of mechanisms
proposed in the literature.
Of course, segregation can affect the operation of mechanisms independently of individual
psychology. For example, segregated groups have greater potential for acquiring differentiated tastes (Massey and Denton 1993), which implies that shared tastes should also vary
with the contextual level of intergroup segregation. And segregation may also reinforce the
accumulation of ingroup technology advantages that comes with residential segregation, such
as shared customs, language, and habits, which makes the efficacy of the choice contextually
dependent. Measuring the contribution of psychological and non-psychological processes to
the operation of these mechanisms is beyond the scope of this paper. Here we note that they
both can and should be operating in a diverse context.
1.3
Hypotheses
Our theory of context and intergroup behavior generates several observable implications.
We expect that increases in the size and segregation of a minority outgroup should increase
intergroup prejudice and that this should be reflected in intergroup behaviors that have
been proposed and tested as mechanisms linking diversity with negative social outcomes. In
particular, we argue that size and segregation should be most visibly related to mechanisms
most closely linked to affect: other-regarding behavior and strategy selection. Geographiccontextual variables should also have an impact on other mechanisms, but less directly and,
therefore, perhaps more weakly.
To summarize our expectations:
1. We hypothesize that the operation of taste for ingroup bias in other-regarding behavior,
ingroup preference in task efficacy, and ingroup bias in cooperation will be greater in
9
individuals with higher outgroup prejudice.
2. We hypothesize that these same ingroup-biases will be systematically related to context. Ingroup bias should be greater in localities of higher segregation.
3. The local proportion of the outgroup should also significantly interact with segregation, but the relationship between local presence of the outgroup and ingroup bias will
be quadratic, so that at lower levels of the outgroup population, increases in the outgroup proportion will lead to increases in prejudice. At higher levels of the outgroup
proportion, it will lead to decreases in prejudice. While detecting the inflection point
of this quadratic relation it outside the scope of our analysis, this theoretical insight
produces two main predictions:
(a) For small minority groups, increases in the outgroup are related to decreases in
discriminatory behavior.
(b) For large majority groups, increases in the outgroup are related to increases in
discriminatory behavior.
2
2.1
The Intra-Jewish Cleavage in Israeli Society
Demographic Background
Due to its social and geographic characteristics, Israel is well suited for studying the effects of
residential patterns on intergroup attitudes and behaviors. Israeli society shows high levels
of ethnic fractionalization, especially compared to other OECD countries (Fearon 2003). As
can be seen in Figure A.1 in the Appendix, Israel generally fits with the expected relations
between social fractionalization and public spending (Alesina, Devleeschauwer, Easterly,
Kurlat, and Wacziarg 2003, Fearon 2003). Due to the small geographic size of the country,
within a relatively small area there are cities with vast variations in their demographic
10
composition and degree of segregation. This includes both cities with mixed Jewish (religious
and non-religious) residents and inter-ethnic population (Arabs and Jews).
Israeli society is rife with social cleavages. Much attention had been given to national
and religious tensions between Arabs from Jews. Within the Israeli-Jewish society, the
ethnic cleavage between Ashkenazi (with European origins) and Sephardic (with origins in
Muslim countries) Jews has been analyzed from both economic (Fershtman and Gneezy 2001)
and sociological (Mizrachi and Herzog 2012) perspectives. In addition, the intra-Jewish
religious cleavage between religious and non-religious Jews is also highly salient in Israel
(Ben-Porat 2013).5
Our study focuses on attitudes and behaviors between Ultra Orthodox [UO] and non-UO
Jews. The UO are not defined by ethnic boundaries and are in fact deeply divided along
ethnic lines (Ashkenazi and Sephardic origins). At the same time, the UO are clearly a
“politically relevant group”, which Posner (2004a) identifies as a necessary condition for
intergroup conflict. A politically relevant group is defined as a sub-group of the population
that is a source of social identification and has direct impact on important political outcomes
(Posner 2004a). We should therefore expect that theories regarding the effects and underlying
mechanisms of inter-group relations should be also apparent within the intra-Jewish cleavages
in Israel.
According to Israel’s Central Bureau of Statistics (CBS), the Israeli population in 2009
was estimated at 7.5 million citizens. Around 75% of the population is Jewish, and the
rest consist of mostly Muslim Arabs.6 In 2009, the UO population was estimated at around
750,000 people, approximately 14% of the Jewish population. The UO are rapidly growing due to high rates of reproduction. According to a recent CBS report, in 2012 around
23% of the students in Jewish schools attended institutions that are under UO supervision,
5
As can be expected, these cleavages intersect in multiple ways: “secular Israeli tend to be more educated,
Ashkenazi, and politically identified with the left, and tend to describe themselves as upper-middle class.
The [religious] ‘traditional’ category is popular among Mizrahim [Sephardic Jews]. Furthermore, religiosity
strongly correlates with a rightist political orientation” (Ben-Porat 2013, p. 47)
6
This count does not include Arabs living in the Palestinian territories.
11
indicating a rapidly growing population. Figure 1 presents the CBS’s projections for demographic changes in 2009-2059 for Arabs, UO and non-UO Jews. The two minority groups Arabs and UO - are growing faster than the majority group of non-UO Jews. This pattern
of population shift fits with a condition identified by scholars as associated with increased
intergroup conflict (Hopkins 2010, Newman 2012).
Figure 1: Projected demographic changes
200
300
All
UO
Non−UO Jews
Arabs
0
100
Percentage of expected change
400
(a) Projected Rate of Population Change
2010
2020
2030
2040
2050
2060
Years
Figures generated from authors calculations based on Israeli Central Bureau of Statistics data.
In Hebrew, the UO are called “Haredim”, which translates “those in awe of divine power”.
What distinguishes the UO from secular or other religious Jewish groups is their strict
adherence to a very traditional interpretation of the sacred texts, which shapes their ways
of living. They tend to live in segregated communities (as discussed below), attend an
independent education system, wear distinctive clothing, pay special attention to issues
of modesty and separation between men and women, and some prefer Yiddish to Hebrew
12
(Rubin 2012, pp. 161-163). Their beliefs and conduct of living have their origins in late 18th
and early 19th century Eastern Europe: the UO “today practice a tradition that preserves
to a remarkable degree the lifestyle of their villages (shtetls) in central and eastern Europe
in the nineteenth century” (Berman 2000, p. 910). This includes also their style of dressing,
which is unique in the Israeli landscape and makes them a highly visible minority.
Our sample contains Israeli Jews across the full spectrum of religious adherence. It is
important to state clearly that Israeli society contains a wide-range of religious devotion and
cultural distinctiveness, including Jews that maintain religious customs, while not following
the distinctive UO customs. It is therefore crucial to recognize that the Israeli Jewish society
is not dichotomously divided between the non-UO and UO. Rather, the religious dimension
is a continuum, with many traditional and Orthodox Jews situated in between (Rubin 2012,
Ben-Porat, Levy, Mizrahi, Naor, and Tzfadia 2008). We took this point seriously in shaping
the research instrument and consider it when analyzing the results, as discussed below.
Because we were aware of the ambiguities in religious identity, we measured it in several
ways, both by self-reported identity and the classification provided by the field workers. In
the results reported here, we use classifications provided by the field workers to group our
subjects, but the results using other classifications are similar and our conclusions would be
unchanged.
2.2
The Politics of the Intra-Jewish Religious Cleavage
The evolution of social tensions between UO and secular Jews in Israel is closely related
to developments in the partisan arena. The Israeli party system is characterized by high
levels of fragmentation. The UO are represented in the Knesset (the Israeli Parliament) by
the Ashkenazi Agudat Israel (Union of Israel) party, which merged with other groups and
is called today Yahadut HaTura (The Judaism of the Torah), and the Shas, the Mizrahi
UO party. Shas’ success is at least partially due to its ability to attract also traditional
and religious supporters, who do not see themselves as strictly UO (Ben-Porat 2013). The
13
influence of the UO parties increased as they participated in most governments since then
(Sandler and Kampinsky 2009, p. 77). Because they have governed with both the Right
and the Left, the UO parties play a pivotal role in a highly fragmented multi-party system
and are well-positioned to serve as the king-makers of Israeli politics. In the last 30 years,
the UO parties raised their demands to obtain material resources from the state for their
community in the form of funds for the UO education system, generous childcare subsidies,
governmental support for religious institutions, etc. These generous public funds allow the
existence of a UO ‘society of learners’, exempt from military service and many times also
from work. In light of their very low rate of participation in the labor market, government
subsidies serves them as a crucial source of income (Doron and Kook 1999).
However, the growing impact of the UO parties, coupled with projections over their
demographic ascendance, has also created a backlash response, as the secular public felt
coerced to accept more and more religious norms (Gordon 1989). According to Ben Porat et
al., “in recent surveys a majority of respondents described the relationship between religious
and secular as negative and as a great threat to Israeli democracy” (Ben-Porat et al. 2008,
p. 31).
2.3
The Conflicting Dynamics of Integration and Segregation
Changes in residential patterns constitute a major source of inter-group tensions in Israel.
The communal aspect of the UO unique customs (code of dressing, educational institutions,
Kosher regulations, gender segregation, etc.) is easier to preserve within an environment that
consist mostly of UO. Traditionally, the UO preferred living in segregated neighborhoods in
Jerusalem, or in the city of Bnei-Brak. As the size of the UO grew exponentially, there
was a need to address their demand for affordable housing. During the 1960s, UO groups
began to migrate into cities in the vicinity of Jerusalem and Bnei-Brak. They later moved
farther into the periphery, where housing is cheaper (Cahaner 2009). In Jerusalem, UO have
expanded from their traditional neighborhoods into the secular parts of the city. In some
14
cases, this process of UO-ization was accompanied with growing social tensions between
groups. According to one observer (Efron 2003, p. 99), “The most primal fear of many
secular Israelis is that Haredim [UO] are taking over our cities and neighborhoods”. By the
late 1990s, non-UO residents of Jerusalem showed more concern regarding the intra-Jewish
cleavage (secular-UO) than the inter-ethnic cleavage (Jews-Arabs) (Hasson and Gonen 1997).
3
Research Sites and Experimental Framework
Previous works examining inter-group attitudes and behavioral strategies tend to rely on
surveys (e.g. Tesler and Sears (2010)), laboratory (e.g. Fershtman and Gneezy (2001)) or
lab-in-the-field (e.g. Habyarimana et al. (2007)) designs. Lab-in-the-field design has the
advantage of moving beyond students in Western universities as the main population of interest in experimental social science research. Lab-in-the-field design also approximates a
more natural environment for the respondents and thus increases the external validity of the
experimental results (Grossman 2011). At the same time, as it stresses the importance of
the local environment, this type of experiments also assumes that behavioral strategies vary
across contexts, suggesting that lab-in-the-field experiments in a single area may suffer from
external validity problems similar to those attributed to labs in traditional university settings. As with traditional lab experiments, it is sometimes hard to infer the scope conditions,
or degree of generalizability, of the findings. For instance, it could be that degree of discrimination between Ashkenazi and Sephardic Jews differs across cities or neighborhoods in
Israel (Fershtman and Gneezy 2001), or that behavioral strategies of French-natives toward
Muslim immigrants vary across regions with more or less foreign-born population (Adida,
Laitin, and Valfort 2012). And while surveys can be more easily deployed across multiple
locations to measure scope conditions, it is often more difficult to manipulate conditions and
precisely measure behavioral responses in a survey environment than in a lab.
Previous works that used similar experimental games in a variety of different contexts
15
show that the answers ones finds depend on where one is looking (Henrich, Boyd, Bowles,
Camerer, Fehr, Gintis, and McElreath 2001, Koopmans and Veit 2013). A comparative
setting, with clear expectations for how and why behaviors should vary across sites, seems
as the next logical step in experimental research design in the social sciences. While the
execution of an elaborate multi-site research strategy can be costly and requires intimate
knowledge with a host of contextual factors, it could at the same time extend researchers’
understanding of variations in the operation of key social mechanisms. Such a multi-site
experimental design thus marries the comparative advantages of survey research (through
close attention to contextual variations) and lab research (by measuring nuanced changes in
respondents’ behavior).
Prior to beginning our research we examined survey data on intergroup relations in
Israel and found spatial variation in responses which were consistent with our hypotheses.
We display these results in Table A.1 in the Appendix. We then designed and piloted our
experiments on Israelis living in the U.S. We also used a web-based survey to pilot our
questions about residency on a convenience sample of several dozen Israelis (living in Israel).
3.1
Case Selection
Our locations include 16 cities across Israel and 4 neighborhoods in Jerusalem, which were
selected based on variation in degrees of diversity and residential segregation. Table A.2 in
the Appendix provides descriptive statistics on the key features of these locations. In mixed
cities and neighborhoods, the sample consists of approximately half UO and half non-UO
Jews, regardless of the distribution of the population in this location. This means that
in each location, the minority group is over-represented, which allows sufficient statistical
power to measure their behavior. This results in UO comprising a larger share in our sample
compared to their relative size in the population.
16
3.2
Sampling method
Fieldwork in Israel was conducted during late July and early August 2013 by a local professional survey firm, which has a unique expertise in working with the UO sector. We
supervised approximately 20% of the field work in order to insure the integrity of the process.
Our sample consists of 456 individuals spread across 20 locations, with an average of
23 respondents in each location. The number of respondents in each location ranges from
21 to 27 (see table A.3 in the Appendix). Quotas for gender and age groups were used to
ensure a balanced sample of the population in each location. We limited participation to
Hebrew speakers. Participants in each location were randomly selected to participate in the
research, based on a random walk strategy. Fieldworkers followed a strict protocol in choosing
potential participants. Walking from door to door in the locations to which they were
assigned, fieldworkers presented the research as dealing with “Israeli society” in the broadest
terms and explained that compensation will depend on participants’ decisions in a series
of decision-making games, with a guaranteed minimum of 20 NIS (the equivalent of about
$6). Participation took around 40 minutes, in which respondents worked independently on a
laptop provided by us. The response rate for those invited to participate was approximately
17%, which compares favorably to survey response rates in the U.S. (Schoeni, Stafford,
Mcgonagle, and Andreski 2013)7 With this method, we obtained a sample that is similar
to a nationally representative Jewish sample obtained in other surveys (see Table A.3).
However, because we oversample the UO population, which tends to be poor and less well
educated than the typical Israeli Jew, our sample is over-represented on these dimensions.
This method of data collection provides us with three main advantages. First, whereas
7
Considering the high costs of participation (letting fieldworkers into participants’ homes for more than
half an hour), this response rate seems actually quite high. One thousand fifteen individuals immediately
refused to participate, without opening the door. This includes children who are not allowed to bring in
strangers, immigrants who did not speak Hebrew, etc. 1225 individuals refused to participate after learning
about the research topic and expected compensation. In some cases, due to language or technological
limitations, fieldworkers provided participants with basic technical help in responding.
17
random sampling is standard in collecting survey data, it is more rarely used in experiments.
By bringing the lab to participants’ homes, rather than the other way around, we were able
to achieve a broader and more representative sample of the population. Second, since we
are interested in the effects of local residential patterns on attitudes and perceptions, it was
important that participation takes place within the natural environment where individuals
live and interact with others (rather than in a lab outside their neighborhood). Lastly,
fieldworkers were instructed to record the exact location of the interview. This allowed us
to match participants with geo-coded data on their immediate area of living, providing us
with rich contextual information about their local environment.
3.3
Key tests
Our subjects engaged in a series of games and survey questions to test the mechanisms
of diversity and public goods. We designed our tests to closely resemble the design of
Habyarimana et al. (2007). These tests took the following form:
1. Respondents expressed their preference for policy priorities to assess differences in
tastes. They were asked which of the following four issues, which are commonly debated
in Israeli politics, they believed to be the most important for the government to spend
money on: defense, education, inequality, labor, or other.
2. Respondents played a dictator game to assess differences in other-regarding behavior.
In this game, they were allocated 20 NIS (approximately $6) and allowed to allocate
as much as they wanted between two other players or keep any amount for themselves.
The other players were always either a UO and a non-UO, a UO and an Arab, or
an Arab and a non-UO. Each subject played three rounds, with each combination
presented in random order. It was also mentioned that all of the opposing players
reside in Jerusalem, where both Arabs, UO and secular Jews live. At the end of this
section, we documented whether respondents could recognize the social group of each
18
of the players.
3. Respondents played a public goods game to assess strategy selection differences when
playing against the ingroup or outgroup. This was a Prisoner’s Dilemma where the
subject was given 20 NIS and allowed to cooperate by sharing all of it or defect by
keeping the 20 NIS for themselves. Only after they played were respondents informed
on how the opposing player acted. Payoffs were multiplied in the standard format
of a Prisoner’s Dilemma: the sum of money in the public pot was multiplied by 1.5
and then equally divided between the two participants. Subjects played three rounds,
facing a UO, non-UO, or Arab in each.
4. Respondents were asked to choose a partner for a complex task to assess perceived
technology advantages from working with the ingroup. Next to photo of a large Lego
box, they were asked the following question: “Imagine you were given a complicated
LEGO set similar to the one shown above, complete with instructions. If you had a
limited time to complete it and you could choose a partner out of the people below,
which one would you choose to work with?”. Respondents were given the option of
choosing a UO, non-UO, or Arab (represented by photos of the same individuals from
the previous games).
Subjects were paid the winnings from a randomly determined round from one of their
games. To recruit players as opponents for economic games, we contacted local residents of
Jerusalem (Arab, Jewish and UO men at the ages of 20-30), who agreed to serve as opposing
players in the dictator and public good games. We collected their basic demographic data
and photo, and recorded their decision in a theoretical public-good game.
Before each game, players were given detailed instructions. The group memberships of
the other player was not explicitly identified; only the first name of the player and a photo
were displayed. The difference between Arab and Jewish names and unique dress of the UO
were intended to serve as clear identifiers of group membership. After the games, we asked
19
respondents to judge the group membership of the opposing players. A large majority of our
respondents did so correctly.
To test for prejudice, we use a modified Bogardus Social Distance Scale (Bogardus 1925),
which is a cumulative scale designed to measure affect between two groups. Meta-analysis
of this type of scale has demonstrated that they are predictive of behavior (Kraus 1995).
We presented respondents with different outgroups and asked them to indicate the closest
relationship for which they would accept that person with the options of “none”, “visitor
to my country”, “citizen of my country”, “coworker”, “neighbor”, “friend”, and “relative”.
Accepting a person into none of the relationships is maximum negative prejudice, while
acceptance as a relative represents minimal negative prejudice. We also measured prejudice
using a number of other theoretically similar measures, which yield similar results in the
analysis to follow
3.4
Measurement of Ultra Orthodox population
Using contextual data derived from the Israeli census of 2008, we construct measures of both
demographic homogeneity and segregation. Segregation scores were computed using the
Dissimilarity Index, a common measure of spatial segregation. We measure these variables
at the Statistical Area (SA) level. These are the smallest unit of analysis used by the CBS
(roughly equivalent to a Census Tract in the United States). In large cities, this unit stands
for a neighborhood, yet in rural parts of the country a statistical area may include the whole
town. In most cases, SA include around 1,000 to 10,000 inhabitants, though some of them
are significantly smaller. In the Appendix (Section A.4) we offer a longer description of
measurement of this population and some alternative measurements that yield consistent
results.
20
4
Results
4.1
Presence of mechanisms nationwide
We display the basic nationwide results for the preference, technology, and strategy mechanisms found in the literature on diversity and social efficiency in Table 1. In the Appendix
(Section A.5) we more fully explore the distribution of each of these variables and report
differences in behavior for UO and non-UO samples.
Table 1: Nationwide Means of Behavioral Strategies
Dictator Ingroup Preference
UO - non-UO Education Preference
UO - non-UO Defense Preference
Public Goods Ingroup - Outgroup Cooperation Strategy
Ingroup - Outgroup Task Efficacy
mean
t statistic
p value
2.91
0.12
-0.06
0.20
0.63
7.69
2.59
-1.94
6.63
12.77
0.00
0.01
0.05
0.00
0.00
Mean behavioral responses for ingroup preference in Dictator Game, ingroup cooperation minus outgroup
cooperation in Public Goods Game, ingroup selection for task, and UO minus non-UO proportions selecting
two policy priorities for government spending. N = 440
The difference between giving to the ingroup and outgroup in the dictator game shows
a strong tendency to favor the ingroup (Row 1). We operationalize the level of otherregarding behavior by measuring the amount distributed to ingroup members minus the
amount distributed to outgroup members. The median difference was 0, which reflects that
a large portion of players gave no money to either player, rather choosing to keep it for
themselves. On the other hand, over 10% of players awarded all 20 NIS to their ingroup
member. The mean differences is 2.91 NIS. A T-test to reject the Null Hypothesis of no
difference between ingroup and outgroup giving can very confidently be rejected (t = 7.69).
Policy preferences for UO and non-UO are also different, further indicating differences in
tastes between groups. Rows 2 and 3 display the proportion of UO minus the proportion of
non-UO that favor spending priority for the government. Consistent with their large families,
UO favor educational spending, while the non-UO, who serve in the military, have a stronger
preference for defense spending. T-tests for a Null hypothesis of no difference in preference
21
on defense spending and education spending allow us to reject both nulls (t = 2.59 and
t = −1.94, respectively).
Cooperation in the Public Goods Game (Row 4) was much more likely with the ingroup
(90%) than the outgroup (70%). These differences allow us to easily reject a Null Hypothesis
of no difference in cooperation between ingroup and outgroup members (t = 6.63). These
differences indicate strongly different strategy selection depending on the player.8
In the selection of partner for a complicated task, we also see a strong ingroup preference
(Row 5). In our sample, 62.9% of respondents indicated they would prefer to complete
the task with their ingroup member. These differences also allow us to easily reject a Null
Hypothesis of no difference in cooperation between ingroup and outgroup members (t =
12.62).9 These differences indicate that a technology mechanism is operating.
In summary, in our sample, nationwide we find support for all three of the categories of
mechanisms linking diversity to inefficient social outcomes. This basic result differs slightly
from previous findings, such as those of Habyarimana et al. (2007) in Uganda. And, consistent with scholars such as Lieberman and McClendon (2013), this suggests that there may
be cross-national variation in the operation of these mechanisms. Of course, some crossnational variation is naturally to be expected and it would be a much larger under-taking
to assess the extent of this variation. Instead, our focus in this study is to demonstrate that
there is strong systematic variation within a single country, suggesting that the operation
of these mechanisms are subject to the same psychological forces observed in the contextual
determinants of intergroup relations. We therefore turn now to examining variation across
our study locations next.
8
Consistent with Habyarimana et al. (2007), the results reported here are for “non-egoists”. Egoists
are players that opted not to cooperate regardless of the identity of the opposing player. Full results for
non-egoists are reported in Table A.5 and for all players in Table A.4 in the Appendix.
9
Respondents were given three alternatives: a non-UO Jew, an UO or an Arab. If they were choosing
randomly, we would expect the ingroup member to be chosen in 1/3 of selections, a T-test against this
alternative hypothesis yields t = 12.62, allowing the hypothesis to be easily rejected. If the Jewish players
are expected to pass over the Arab but are indifferent between the two Jewish alternatives, we should expect
them to select the ingroup 50% of the time. A t-test on this alternative yields t = 5.50 also allowing us to
strongly reject this hypothesis.
22
In the analysis above and all the analysis to follow, in unreported results, we also subset
the analysis by UO and subjects which consider themselves as secular (as opposed to all
non-UO) and this also yields consistent results. We also performed the analysis using a
different classification scheme for religiosity, in which self-reported classifications were used.
This too yields substantively similar results.
4.2
The Relationship Between Prejudice and Behavioral Strategies
We have proposed that intergroup behavioral strategies can usefully be understood as a
general psychological phenomenon that operates through prejudice against members of the
outgroup, perhaps based on negative stereotypes that serve as heuristics or other such mechanisms.
We display the distribution of prejudice, as measured by social distance, in Figure 2. The
modal response is that a subject would accept an outgroup member as a neighbor (29.5%).
Only 31% of respondents would accept an outgroup members as friend or relative and 24.4%
would only accept an outgroup member as a fellow citizen of their country or more distant,
meaning they would not want to live in the same neighborhood or work with outgroup
members.
We created a dichotomous variable based on whether or not the respondent would accept
an outgroup member as a neighbor or closer, meaning the sample includes people that
would accept the outgroup as friends or relatives.10 We looked at behavioral strategies as
a function of this variable as well. Table 2 displays results of T-tests for the differences
in outgroup discrimination for subjects who would not accept an outgroup member as a
neighbor and those would would accept an outgroup member as a neighbor (or closer).
This table demonstrates that social distance is strongly related with these discriminatory
10
We also created a variable based on acceptance as a relative member which yields substantively similar
results in the analysis that follows. Similarly, measuring affect through a seven-point scale of trust of the
outgroup yields substantively similar results.
23
Figure 2: Distribution of Social Distance
0.35
0.3
0.30
0.25
0.2
0.20
0.2
0.15
0.15
0.11
0.10
0.05
0.03
0.02
visitor
nothing
0.00
relative
friend
neighbor
coworker
citizen
Distribution of responses on Social Distance Scale. N = 413
behaviors. Subjects with high social distance have a 2.88 NIS greater ingroup bias in the
Dictator Game than those with low social distance, this is equal to .37 of a standard deviation
of the variable. They also cooperate with the ingroup over the outgroup 17 percentage points
more than those with low social distance. And they are 12 percentage points more likely
to choose an ingroup member for a task. All these differences yield p < .05 for a Null
Hypothesis of no difference between the groups.
We use this as non-parametric demonstration of association between these variables, indicating that negative affect toward the outgroup is positively associated with discriminatory
behavior. Definitively demonstrating that this relationship is causal - that prejudice causes
these behavioral strategies - would require further experiments. However, as reviewed in the
introduction to this paper, we argue that there is a strong theoretical and intuitive plausibility that discriminatory behavior is driven by prejudice and, as such, should vary with context
in a manner similar to prejudice. Having demonstrated that the discriminatory behavior is
24
strongly predicted by prejudice, we turn to demonstrating that these same discriminatory
behaviors are strongly predicted by context. We then demonstrate and discuss some of the
implications of this relationship for the study of diversity and social efficiency.11
Table 2: Behavioral Strategies by Social Distance
Dictator Ingroup Preference
Public Goods Ingroup - Outgroup Cooperation Strategy
Ingroup - Outgroup Task Efficacy
mean
t statistic
p value
2.88
0.17
0.12
3.79
2.58
2.60
0.00
0.01
0.01
Mean behavioral responses for ingroup preference in Dictator Game, ingroup cooperation minus outgroup
cooperation in Public Goods Game, and ingroup selection for task for subjects rejecting the outgroup as a
neighbor minus subjects that would accept the outgroup as a neighbor. N = 422
4.3
Contextual Influences on Prejudice and Behavioral Strategies
We have argued that the relationship between individual strategies and social inefficient
outcomes can be understood as part of a general psychological phenomenon of anti-outgroup
prejudice. And we argue that this implies that individual behavioral strategies are not only
a function of the outgroup status of other players but of the context in which the games are
played.
We can assess this relationship by regressing behaviorally revealed ingroup preferences
on the interaction of local area segregation and homogeneity of the outgroup. As dependent
variables, we use the results of main experiments: other-regarding behavior as measured
by the Dictator Game, outgroup cooperation as measured from a Prisoners’ Dilemma, and
preferences for ingroup members in a complex task. In linear regression models, we include
individual level covariates and cluster the standard errors at the local level. We regress
these variables separately for UO and non-UO subjects because the range of homogeneity is
significantly different across the two groups. This is because UO, the minority within the
11
In addition to definitively demonstrating the causal relationship between prejudice and discriminatory
behaviors, a direction for further research should be to decompose the direct effects of context on behavior,
perhaps through network or social sanctioning effects, and the direct effect of context on behavior, as
mediated by prejudice.
25
Jewish population, rarely lives in cities where outgroup is not a majority. As stated in hypothesis 2, this leads to a potential nuance in the expected results. The relationship between
outgroup group-size and prejudice is often thought to be quadratic, so that at low-levels of
the outgroup population, as the outgroup becomes larger they are more of threat and induce
more prejudice. However, at the higher-ends, the relationship may actually negative as an
outgroup becomes a large majority of the population, individuals living in these locations
may feel particularly at ease with the outgroup. This leads to different expectations for the
relationship between outgroup size and discrimination for the UO and non-UO in the analysis
below. For the UO, where the outgroup usually comprises over 80% of the local population,
increases in the outgroup should be related to decreases in discriminatory behavior. For the
non-UO, where the outgroup is usually comprises under 20% of the population, increases in
the outgroup should be related to increases in discriminatory behavior.
In the regressions below we included only subjects that correctly identified the religiosity
of the opposing player. Including all subjects does not substantively change the results. We
also tested the relationship using a hierarchical model with varying intercepts for locality,
which produces the same substantive results. Below we focus on the results from the Dictator
Game. In the Appendix, we report the results of the same regressions using task ingroup
preference in task efficacy and ingroup bias in cooperation in Tables A.6 and A.7, which
follow the same pattern as the results for ingroup bias in other-regarding behavior.
The regression coefficients are reported in Table 3. Our main variables of interest are
Segregation, Outgroup Percent, and the interaction of the two. For both the UO and nonUO, respondents are very responsive to changes in these variables, with large coefficients.
This is true however especially for the non-UO, where the variables are large and achieve
conventional levels of statistical significance. Furthermore, for non-UO, the interaction term
is especially strong, demonstrating the sensitivity of individual behavior to contextual influences. Remember that UO and non-UO are operating on opposite ends of the distribution
of Outgroup Proportion, which results in different signs on the coefficients, even though the
26
Table 3: Regressions of ingroup bias in other regarding behavior on contextual variables
UO
Intercept
Segregation
Outgroup Proportion
Male
Age
Mixed Ethnicity
Other Ethnicity
Sephardic Ethnicity
Political Ideology
High Income
Low Income
Arab Population
Jerusalem
Segregation x Outgroup Percent
N
R2
adj. R2
Resid. sd
−6.55
(13.66)
31.26
(32.29)
4.91
(16.49)
0.10
(0.92)
0.04
(0.03)
−2.12
(1.50)
−2.85∗
(1.36)
−1.30
(1.03)
−0.17
(0.37)
1.81
(2.32)
2.14
(1.57)
31.29∗
(13.09)
−0.84
(1.86)
−31.68
(37.73)
200
0.09
0.03
7.13
Non UO
11.69∗
(5.56)
−13.88∗
(6.03)
−41.89∗
(8.89)
−0.47
(1.20)
−0.05
(0.03)
−3.56∗
(1.21)
2.89
(3.42)
−2.94∗
(1.09)
0.16
(0.83)
−1.23
(1.65)
−1.10
(1.53)
7.15
(16.00)
−1.76∗
(0.87)
95.40∗
(25.76)
189
0.08
0.01
8.17
Regression of ingroup bias in dictator game for UO (column 1) and non-UO (column 2) on contextual and
individual level variables. Ingroup bias in the Dictator Game is the amount in NIS contributed to ingroup
members minus the amount contributed to outgroup members. In both cases, higher numbers represent greater
discrimination against the outgroup. Clustered standard errors in parentheses. ∗ indicates significance at
p < 0.05
behavior of individuals in both groups follows the theoretical prediction. To demonstrate this
pattern, we display the predicted values of ingroup preference in the dictator game in Figure
3. In these figures, the predicted amount of ingroup bias is shown while moving between
the inner 90% range of levels of segregation found in Tel Aviv (.29) and Jerusalem (.65) and
proportion of UO in Tel Aviv (.01) and Jerusalem (.33), while holding other variables at
27
mean values. The solid lines represent the mean values and the dotted lines represent the
95% confidence intervals of these predictions. As predicted, greater segregation (Figures a
and c) results in greater ingroup bias and outgroup proportion follows a curve-linear pattern,
with lower levels of the outgroup (Figure b), seen among the non-UO population, leading
to more ingroup bias. As outgroup levels reach higher levels (Figure d), as with the UO
population, ingroup bias decreases. We choose not to extrapolate either of these populations
across the entire range of possible outgroup levels because we simply do not have the data
to support inference for the lowest outgroup levels for the UO or the highest outgrip levels
for the non-UO. However, in the areas we do predict values, the data neatly fits theoretical expectations. Moving across the range of segregation for non-UO (a) takes the ingroup
bias from about 0 to almost 4 NIS and the range of Outgroup proportion (b) increases the
ingroup bias by over 500%. Similarly, although more weakly, the point estimates on the
associated changes for the UO are over 75% increase when segregation increases (c) and an
85% decrease when out group proportion increases (d).
Another way to look at the powerful influence of context, especially segregation, is displayed in Figure 4. In this Figure, we demonstrate the effect of segregation by predicting
the changes in discriminatory behavioral strategies when moving between Tel Aviv and
Jerusalem levels of segregation, while at the population proportions projected for 2059, as
discussed in the previous section. These changes indicate the potential strength of segregation to shape intergroup behavior in a more diverse society. The relationship between
discriminatory behavior and segregation shows that as the minority population of Israel
grows, decisions about the spatial integration of groups can have powerful consequences. In
Figure 4 we see that moving from Tel Aviv levels of segregation to Jerusalem levels increases
the level of discriminatory behavior. We report increases in discriminatory behavior in the
Dictator Game, ingroup preference for a task, and non-cooperation in the Public Goods game
for non-UO (white circles) and UO (black circles). Ninety-five percent confidence intervals
are displayed with dotted lines. We standardize the results of the Dictator Game to a 0 to
28
Figure 3: Predicted ingroup preference by levels of segregation and homogeneity
(a) Segregation, non-UO
(b) Outgroup Proportion, non-UO
10
Ingroup Preference
Ingroup Preference
10
5
0
−5
5
0
−5
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.00
0.05
0.10
Segregation
0.20
0.25
0.30
Outgroup Proportion
(c) Segregation, UO
(d) Outgroup Proportion, UO
10
Ingroup Preference
10
Ingroup Preference
0.15
5
0
−5
5
0
−5
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.70
Segregation
0.75
0.80
0.85
0.90
0.95
1.00
Outgroup Proportion
Predicted level of ingroup - outgroup allocation in dictator game in NIS with varying levels of segregation (a)
and homogeneity (b) for secular and segregation (c) and homogeneity (d) for UO. All other variables held at
mean levels. Dotted lines represent 95% confidence intervals.
1 range.
This figure demonstrates the potentially powerful effects of segregation to shape inter-
29
Figure 4: Predicted changes in discriminatory behavioral strategies, moving from Tel Aviv
to Jerusalem levels of segregation with 2059 population projections
Dictator
●
●
Task
●
●
Public Goods
●
●
−0.5
0.0
0.5
1.0
Expected Change from Tel Aviv to Jerusalem
Predicted changes in discriminatory behavioral strategies for non-UO (white circles) and UO (black circles)
moving from Tel Aviv levels of discrimination to Jerusalem levels at 2059 population projections. Dotted
lines are the 95% confidence intervals. All other variables held at mean levels.
group relations as population patterns change. If Israeli society develops in a highly segregated, such as the current Jerusalem residential patter, rather than in a more integrated
pattern, such as that of the current Tel Aviv situation, intergroup behavioral discrimination
can be expected to greatly increase, resulting in more ingroup preference and more perceived
ingroup efficacy in tasks, and less cooperation. We further explore the implications of this
relationship for Israel and other societies in the Conclusions Section.
30
4.4
Variation in Mechanisms Across Locations
Our findings have implications for the understanding of the mechanisms underlying the
effects of social diversity and socially inefficient outcomes. Previous work, most prominently
Habyarimana et al. (2007), has explored these mechanisms by testing for the presence of
discriminatory behavior in different types of games or decision tasks. We have demonstrated
that the levels of these behaviors vary systematically with context. This should mean that
the operation of these mechanisms will be contextually dependent, so while, for example,
preference mechanisms may be most prominent in a certain setting, technology mechanisms
may also be found prominently elsewhere. In this section, we demonstrate how different
conclusions would be reached by a scholar deciding to locate their study in different locations,
even within the same country. We believe this point has important implications for the study
of comparative politics. Because behavior is a function of context, as well as the players,
scholars may observe the operation of different mechanisms depending on the location chosen
for research.
In Figure 5 we display findings that illustrate this point. This figure shows four locations where we conducted our experiments: two cities (Kiryat Gat and bnei Brak) and two
neighborhoods in Jerusalem (Kiryat Yovel, Jerusalem Center). Pairs of these locations have
similar levels of UO population: 4%, 6%, 55%, and 65% respectively (as a percent of the
Jewish population), but different levels of segregation: .50, .33, .53, and .34, respectively.12
For each location, we display four quantities of interest: 1) differences in tastes for
spending priority on defense between UO and non-UO (T); 2) standardized differences in
other regarding preferences for the ingroup over the outgroup (R); 3) perceived efficacy
advantage from working with the ingroup over the outgroup fin the Lego test (E); and 4)
probability of cooperation with an ingroup member minus probability of cooperation with an
ingroup member (C). For each of these measures, positive values mean greater levels of the
12
bnei Brak is a city in the Tel Aviv metropolitan area. In Israel, it is often considered a strongly UO
location, but official statistics reveal that the population is more religiously mixed than might be popularly
perceived. Reflecting this reality, our sample from bnei Brak includes a mix of UO and non-UO.
31
ingroup favoring behavior. For each quantity, we display the mean result and 95% confidence
intervals. In each individual city, we have only small samples, so this discussion is for the
purposes of demonstration, rather than making firm inferences.
According to our theory, we should expect higher outgroup prejudice in the higher segregation areas. This prediction is largely supported in this example: Kiryat Gat and Kiryat
Yovel have similar population proportions, but Kiryat Gat is more segregated. If a researcher
chose to locate their laboratory in Kiryat Gat, rather than Kriyat Yovel, they might conclude
that all four of the proposed mechanisms were operating, while if they located their laboratory in Kiryat Yovel, they might conclude that only two of the mechanisms were operating
- and at weaker levels than in Kiryat Gat. Similarly, Jerusalem Center and bnei Brak have
similar population proportions, but Jerusalem is much more segregated. Consistent with our
expectation, all of the mechanisms proposed in the literature can be found strongly operating
in Jerusalem Center. On the other hand, if researchers chose to study the less segregated city
of bnei Brak they might conclude that none of the mechanisms are operating, or that they
are operating weakly. In summary, depending on which locality a researcher chose to explore
public goods provision, a researcher may arrive at substantially different conclusions about
the mechanisms underlying this phenomenon that has been observed on a global scale.13
Can researchers be expected to test for differences in behavior across every location in a
country or even a reasonable random sample? In most cases, the answer is likely no. How
then can research on individual mechanisms underlying global phenomenon be informative?
Our answer is that by drawing on theories of contextual variation in behavior we can form
expectations about variation in individual behavior across locations that will inform case
selection and interpretation of results.
13
Of course, testing across many locations, with relatively small sample sizes in each, we might expect
that in some locations false positives or false negatives would arise by chance. However, using a p = .05 level
of significance for testing a null hypothesis of 0 difference in other-regarding behavior, we can reject the null
in 11 of 20 locations - a far greater number than would be expected by chance.
32
Figure 5: Levels of differences in tastes, other regarding preferences, perceived efficacy, and
cooperation by city
T
R
Kiryat Gat
E
C
T
R
Kiryat Yovel
E
C
T
R
Jerusalem Center
E
C
T
R
Bnei Brak
E
C
−0.5
−0.3
−0.1
0.1
0.3
0.5
0.7
For each location, we display four quantities of interest: 1) differences in tastes for spending priority on
defense between secular and Ultra Orthodox (T); 2) standardized differences in other regarding preferences
for the ingroup over the outgroup for all subjects (R); 3) perceived efficacy advantage from working with the
ingroup over the outgroup for all subjects (E); and 4) probability of cooperation with an outgroup member
(C). Ninety-five percent confidence intervals are the horizontal bars.
33
4.5
Is the relationship between context and behavior causal?
In the Introduction, we discussed why context may affect the operation of the mechanisms
linking diversity and socially inefficient outcomes. Whether the relationship is causal is not
directly relevant for demonstrating that behaviors do vary by context. However, understanding why the mechanisms vary by context does call for understanding whether segregation and
outgroup size directly affect behavior, are spuriously related, or if the causality is possibly
reversed with people with strong affect selecting into segregated areas. The most obvious
objection is that individuals with strong anti-outgroup bias may select into the most segregated areas in order to avoid the outgroup, thereby reversing the proposed relationship
between attitudes and segregation. It is unlikely that a person with strong anti-outgroup
attitudes would select into areas with a large presence of the outgroup. However, this
relationship could still be driven by other unmeasured variables, making the relationship
between attitudes and outgroup proportion spurious. A causal relationship in this area of
study is difficult to establish (Sampson 2008) due to the inability randomize context. This
literature includes experimental (Enos 2014) and quasi-experimental (Enos 2013) designs
demonstrating a causal link between context and attitudes, such as prejudice.
In this paper, while estimating the effect of context on ingroup preference, we controlled
for individual-level variables that might also drive the relationship and found that the prediction is robust to the inclusion of these controls. However, because we collected a unique
dataset that has been designed specifically to test this relationship, we are able to directly
test for the effects of selection in a manner not usually possible with survey data. We collected a rich-set of variables that allow us to directly measure the abilities and attitudes
related to selecting into segregated and homogeneous localities. We do not doubt that selection is responsible for some of the relationship between context and attitudes, but we also
believe that context can have a direct effect on attitudes and behaviors. Our unique survey
design allows us to test for this relationship.
Using open-ended responses to the survey, we can understand the prevalence of this type
34
of selection. We asked respondents whether, when and why they have moved into a new
address. Respondents were asked whether they had been living in the same address for
their whole life; those who answered negatively, were asked a series of questions about the
circumstances of moving and choosing a new place of living. When asked “What was a
factor in your decision to move out to another place?”, most respondents mentioned the
size of their house and family reasons (getting married or divorce), as well as the cost of
living and educational options for their kids. Only 30 respondents (less than 7%) mentioned
the people around them. Those respondents who mentioned multiple reasons to leave were
asked to choose the most important reason. Only four of these signaled “people” as the most
important reason for moving. When asked what was it about the people around them that
pushed them to move, only 12 respondents mentioned degree of religiosity.
While not many non-UO respondents in our sample opted out of a location because the
people around them were seen as too religious, UO opt into places based on the perceived
religiosity of the neighborhood. When asked about the factors that made them choose the
city and neighborhood to which they moved, around 20% of the respondents mentioned “the
people who live there”. When asked what was it about the people around them which they
most care about, the most common answer was “their degree of religiosity”.
In order to further inquire this point, we subset our data using questions that were
specifically designed to test for the influences of selection on our results. We use variation
in the means of individuals relocate as variables on which to subset and we also measure expressed and revealed preference for selection away from the outgroup. We subset individuals
with reduced ability to change location as a strategy for reducing the likelihood of selection
by subsetting on low-income individuals and individuals that report that they have little
freedom to choose where they would like to live. Then we subset individuals that express
attitudes making them unlikely to select into segregated areas. We subset on individuals that
explicitly report they would not want to live in a segregated community, people that are not
satisfied with where they currently live (reducing the potential that they have selected into
35
their preferred neighborhood), and on individuals that do not report the religion of others
as a factor in their choice of city or neighborhood. We also gave our subjects the ability to
construct their ideal neighborhood by choosing the occupants of each of 10 houses neighboring their own. Respondents could choose from UO, non UO, or Arab (see Figure A.4 in
Appendix) with the explicit instruction that their choices of neighbors would not affect other
qualities of the neighborhood. We subset to individuals that did not construct exclusively
homogeneous neighborhoods. We also collected data that precisely geocodes the prior and
current residence of our subjects. So we can look for individuals that moved to an area that
had a higher proportion of ingroup members and subset on those that did not relocate to a
more heavily ingroup area. Controlling for means, this is a measure of revealed preference
for moving to a homogeneous area. With each of these groups, we reanalyze the data with
the same strategy above. We find substantively the same results and remarkably similar
coefficient estimates, indicating that our findings are not driven by individuals selecting into
highly segregated areas.
We estimated the same model reported in Table 3, except using the subsets described
above. We display the coefficients on the contextual variables from these regressions in Figure
6. Coefficients are for outgroup proportion (circle), segregation (square), and the interaction
of the two (triangle) from regression of ingroup preference in dictator on contextual and
individual level variables for non-UO subjects (a) and UO subjects (b). The coefficients for
the full sample are displayed at the top of the graph with vertical lines at these estimates.
The results for all subsets are very consistent with the results for the full sample. In the
Appendix (A.7), we report the full coefficient tables for the regression of all behavioral
measures on all these subsets. The robustness of our results indicates that segregation and
homogeneity likely have a direct effect on attitudes and intergroup behaviors, as predicted by
prominent theories. Our findings provide further evidence that, despite movement causing
concerns about selection biasing results, theories about context strongly affecting attitudes
can be analytically supported.
36
5
Conclusion
In this paper, we demonstrated that mechanisms associated with diversity and social inefficiency will systematically vary with levels of segregation and demographic diversity. We
suggested that in highly segregated localities, general outgroup prejudice should be expected,
leading to a general breakdown of socio-political relations and, accordingly, differentiated
policy tastes, diminished task efficiency, ingroup preference, and a general lack of cooperation between groups. All of these mechanisms are expected to lead to socially inefficient
outcomes. We have quantified the severe strain on relations within the Israeli Jewish community, where aversion between these two particular groups appears to be reflected in the
prominence of all three classes of mechanisms characterizing diversity and social inefficiency.
We have also demonstrated that this relationship is likely causal and that inferences about
the mechanism underlying this relationship are highly sensitive to context.
5.1
Implications of our findings for research design
By demonstrating how the strength of the relationship varies with context and that the
relationship may be causal, our findings also can serve as a guide for designing future research
around the mechanisms connecting diversity and social inefficiency. Researchers should be
aware that their answers may vary significantly based on the levels of these variables in the
context they select. First, by articulating a model linking context to behavior, scholars can
better understand where their findings are likely to fit in the distribution of behaviors within
or across countries.
When scholars only test a single location, it is difficult to know how behaviorally representative a particular location is of the rest of the country. Theory then is very useful
in understanding how findings collected in single location can help to predict behavior in
other regions. By establishing an expected relationship between behavior and variables that
can potentially be measured in any locality - segregation and diversity - we hope to offer
37
scholars a tool for benchmarking their findings in the future. For example, was the research
conducted in an area with low segregation? If so, the findings probably reflect the low-end
of intergroup animosity and estimates of levels of discriminatory behavior should be considered as downwardly biased if imputed to the entire country. The opposite would be true of
research conducted in highly segregated areas.
Of course, the ideal research would be a design that drew a representative sample of
individuals from a representative sample of localities. This is only possible in rare circumstances, which is why scholars, such as Habyarimana et al. (2007), carefully selected locations
that is believed to be an adequate microcosm of the entire country. However, our findings
demonstrate that a scholar must consider multiple dimensions when selecting locations. For
example, a scholar cannot just choose a research location based on ethnic fractionalization,
because variables besides diversity, i.e. segregation, can affect the inferences that are drawn
from this location.
5.2
Implications for policy and understanding of diversity and social inefficiency
One implication of our findings is, of course, that not all diverse countries are the same and
that not all actors within a diverse country are the same. Depending on the organization
of their social and political space, countries will meet the predictions that diversity causes
social inefficiency more or less well. Countries that concentrate power in the hands of populations particularly subject to outgroup pressure may have particularly inefficient public
goods provision.
A prime example of this is the United States when anti-Black Southerners gained disproportional power in the U.S. national government and leveraged it to restrict social spending
agendas that they saw as benefiting African Americans. These Southerners, according to
classic theories in political science (Key 1949), and consistent with our findings here, were
anti-Black because of the high presence of blacks in their localities. Another example may
38
possibly be found in Israel itself, where UO have gained increasing influence due to rapid
growth and alliances in governing coalitions with right-wing parties. Our research demonstrates that the UO maintain higher levels of anti-outgroup attitudes than other groups,
which is consistent with a highly segregated population, experiencing close contact from
large outgroup populations of secular Jews (and in some cases, perhaps also Muslim Arabs).
These attitudes, combined with the disproportional influence of UO may be reflected in
inefficient public goods provision in Israel, where revenues are inefficiently directed toward
institutions catering to the small UO population, rather than projects serving a larger population.
One practical implications of our study is that housing policies have social implications.
Governments play a role in organizing their citizens in space in various ways, from the design of programs for public housing to subsidies that encourage groups in the population to
reside in specific neighborhoods or cities. In Israel, governmental agencies have debated how
to allocate subsidies for housing and whether to encourage greater integration between the
secular and UO or to concentrate the UO in their own segregated communities. In the US,
there are different opinions about whether programs of public housing should try and bring
closer together low- and middle-income families. These questions have normative implications and they deal with different visions of a liberal, plural society; however, our research
provides empirical evidence that should inform policy makers and the public discussion on
these issues.
The demonstration of the effects of a society organized as Tel Aviv or Jerusalem in
Figure 4 illustrated the policy dilemmas facing a country such as Israel with a rapidly
growing, culturally distinct population. This is a situation faced by many countries in the
Western world due to international migration and it is also similar to the situation faced in
the U.S. during large waves of internal migration in the 20th Century. As the population
grows, should segregation be allowed to occur as the result of the seemingly non-prejudiced
accumulated choices of individuals (Schelling 1971), as some people might prefer (Massey
39
and Denton 1993) - or should governments intervene to shape housing patterns? The choices
are difficult because segregation may in fact be efficient in the short-term but inefficient in
the long-term (Alesina, Baqir, and Easterly 1999). As noted by Uslaner, “While ‘solving’ the
problem of segregation is not easy [...] changing housing patterns is considerably less difficult
than changing who lives in a country.” (Uslaner 2011, p. 244). As population movements,
brought on by technological and economic change, spur rapid diversification of countries,
policy makers may be able to strongly influence intergroup relations and subsequent social
efficiency by shaping patterns of residential segregation.
40
Figure 6: Regression of contextual variables on other-regarding preferences, subsetting to
control for selection
(a) non-UO
(b) UO
All Respondents (232)
All Respondents (208)
●
●
ideal neighborhood not homogenous (112)
ideal neighborhood not homogenous (149)
●
●
did not select into high−ingroup area (91)
did not select into high−ingroup area (179)
●
●
religion not a factor in neighborhood choice (167)
religion not a factor in neighborhood choice (196)
●
●
religion not a factor in city choice (173)
religion not a factor in city choice (197)
●
●
low income (165)
low income (126)
●
●
not free to move (168)
not free to move (163)
●
●
not satisfied with location (140)
not satisfied with location (128)
●
●
do not prefer homogenous neighborhood (64)
do not prefer homogenous neighborhood (118)
●
−50
●
0
50
100
150
200
−150
Regression Coefficients
−100
−50
0
50
100
150
Regression Coefficients
Regression coefficients generated for outgroup proportion (circle), segregation (square), and the interaction
of the two (triangle) from regression of in-group preference in dictator on contextual and individual level
variables for non-UO subjects (a) and UO subjects (b) for subgroups listed above the coefficient points.
Ninety-five percent confidence intervals of the regression are indicated by dotted horizontal lines. Dotted
vertical lines represent the coefficient estimates from the regression with all respondents. In-group preference
in the Dictator Game is the amount in NIS contributed to ingroup members minus the amount contributed
to outgroup members.
41
References
Adida, Claire, David Laitin, and Marie-Anne Valfort. 2012. “Muslims in France: Identifying
a discriminatory equilibrium.”.
Alesina, Alberto, and Eliana La Ferrara. 2002. “Who trusts others?” Journal of public
economics 85(2): 207–234.
Alesina, Alberto, and Eliana La Ferrara. 2004. Ethnic diversity and economic performance.
Technical report National Bureau of Economic Research.
Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain
Wacziarg. 2003. “Fractionalization.” Journal of Economic growth 8(2): 155–194.
Alesina, Alberto, Reza Baqir, and William Easterly. 1999. “Public goods and ethnic divisions.” The Quarterly Journal of Economics 114(4): 1243–1284.
Baldwin, Kate, and John D Huber. 2010. “Economic versus cultural differences: Forms of
ethnic diversity and public goods provision.” American Political Science Review 104(4):
644–662.
Baybeck, Brady. 2006. “Sorting Out the Competing Effects of Racial Context.” Journal of
Politics 68: 386–396.
Ben-Porat, Guy. 2013. Between State and Synagogue: The Secularization of Contemporary
Israel. Cambridge: Cambridge University Press.
Ben-Porat, Guy, Yagil Levy, Shlomo Mizrahi, Arye Naor, and Erez Tzfadia. 2008. Israel
since 1980. Cambridge: Cambridge University Press.
Berman, Eli. 2000. “Sect, subsidy, and sacrifice: an economist’s view of ultra-orthodox
Jews.” The Quarterly Journal of Economics 115(3): 905–953.
Bhavnani, Ravi, Karsten Donnay, Dan Miodownik, Maayan Mor, and Dirk Helbing. 2013.
“Group Segregation and Urban Violence.” American Journal of Political Science .
Blalock, Hubert M. 1967. Toward a theory of minority-group relations. New York: Wiley.
Blumer, Herbert. 1958. “Race Prejudice as a Sense of Group Position.” Pacific Sociological
Review 1(1): 3–7.
Bobo, Lawrence. 1983. “Whites’ Opposition to Busing: Symbolic Racism or Realistic Group
Conflict?” Journal of Personality and Social Psychology 45(6): 1196–1210.
Bobo, Lawrence, and Vincent Hutchings. 1996. “Perceptions of Racial Group Competition: Extending Blumers Theory of Group Position to a Multiracial Social Context.”
American Sociological Review 61(6): 951–72.
Bogardus, Emory S. 1925. “Measuring social distance.” Journal of applied sociology 9(2):
299–308.
42
Cahaner, Lee. 2009. The Development of the Spatial and Hierarchic Structure of The UltraOrthodox Jewish Population in Israel [in Hebrew]. Haifa University: Dissertation.
Campbell, David E. 2006. Why We Vote: How Schools and Communities Shape Our Civic
Life. Princeton, NJ: Princeton.
Carsey, Thomas M. 1995. “The Contextual Effects of Race on White Voter Behavior: The
1989 New York City Mayoral Election.” Journal of Politics 57(Febuary): 221–28.
Collier, Paul. 2000. “Ethnicity, politics and economic performance.” Economics & Politics
12(3): 225–245.
Costa, Dora L, and Matthew E Kahn. 2003. “Civic engagement and community heterogeneity: An economist’s perspective.” Perspective on Politics 1(1): 103–111.
Couzin, Iain D, Christos C Ioannou, Güven Demirel, Thilo Gross, Colin J Torney, Andrew
Hartnett, Larissa Conradt, Simon A Levin, and Naomi E Leonard. 2011. “Uninformed
individuals promote democratic consensus in animal groups.” Science 334.
Delhey, Jan, and Kenneth Newton. 2005. “Predicting cross-national levels of social trust:
global pattern or Nordic exceptionalism?” European Sociological Review 21(4): 311–327.
Doron, Gideon, and Rebecca Kook. 1999. “Religion and the Politics of Inclusion: The Success
of the Ultra Orthodox Parties.” In The Elections in Israel 1999, ed. Ahser Arian, and
Michal Shamir. Albany: State University of New York, pp. 67–84.
Easterly, William, and Ross Levine. 1997. “Africa’s growth tragedy: policies and ethnic
divisions.” The Quarterly Journal of Economics 112(4): 1203–1250.
Efron, Noah J. 2003. Real Jews: Secular versus ultra-orthodox and the struggle for Jewish
identity in Israel. New York: Basic Books.
Enos, Ryan D. 2010. The Persistence of Racial Threat: Evidence from the 2008 Election.
American Political Science Association, Annual Meeting Washington, DC: .
Enos, Ryan D. 2011. “Testing the elusive: a field experiment on racial threat.”.
Enos, Ryan D. 2013. “What Tearing Down Public Housing Projects Teaches Us About the
Effect of Racial Threat on Political Participation.”.
Enos, Ryan D. 2014. “The Causal Effect of Intergroup Contact on Exclusionary Attitudes.”
Proceedings of the National Academy of Sciences of the United States of America .
Fearon, James D. 2003. “Ethnic and cultural diversity by country*.” Journal of Economic
Growth 8(2): 195–222.
Fershtman, Chaim, and Uri Gneezy. 2001. “Discrimination in a segmented society: An
experimental approach.” The Quarterly Journal of Economics 116(1): 351–377.
43
Fiske, Susan T, Amy JC Cuddy, Peter Glick, and Jun Xu. 2002. “A model of (often mixed)
stereotype content: competence and warmth respectively follow from perceived status
and competition.” Journal of personality and social psychology 82(6): 878.
Fitzpatrick, Kevin, and Sean Shong Hwang. 1992. “The Effects of Community Structure
on Opportunities for Interracial Contact: Extending Blaus Macrostructural Theory.”
Sociological Quarterly 33(1): 51–61.
Forbes, Hugh Donald. 1997. Ethnic conflict: commerce, culture, and the contact hypothesis.
New Haven: Yale University Press.
Fossett, Mark A., and K. Jill Kiecolt. 1989. “The Relative Size of Minority Populations and
White Racial Attitudes.” Social Science Quarterly 70(4): 820–35.
Gay, Claudine. 2006. “Seeing Difference: The Effect of Economic Disparity on Black Attitudes Toward Latinos.” American Journal of Political Science 50(October): 982–997.
Giles, Micheal W., and Melanie A. Buckner. 1993. “David Duke and Black Threat: An Old
Hypothesis Revisited.” Journal of Politics 55: 702–13.
Glaser, James M. 1994. “Back to the Black Belt: Racial Environment and White Racial
Attitudes in the South.” The Journal of Politics 56(1): 21–41.
Gordon, Carol. 1989. “Mutual perceptions of religious and secular Jews in Israel.” Journal
of Conflict Resolution 33(4): 632–651.
Green, Donald P., Dara Z. Strolovitch, and Janelle S. Wong. 1998. “Defended Neighborhoods,
Integration, and Racially Motivated Crime.” American Journal of Sociology 104(2).
Grossman, Guy. 2011. “Lab-in-the-field Experiments.” Newsletter of the APSA Experimental
Section 2: 13–19.
Habyarimana, James, Macartan Humphreys, Daniel N. Posner, and Jeremy M. Weinstein.
2007. “Why Does Ethnic Diversity Undermine Public Goods Provision?” American
Political Science Review 101(4).
Habyarimana, James, Macartan Humphreys, Daniel N. Posner, and Jeremy M. Weinstein.
2009. Coethnicity: Diversity and the Dilemmas of Collective Action. New York: Russell
Sage.
Hasson, Shlomo, and Amiram Gonen. 1997. The Cultural Tension within Jerusalems Jewish
Population. Jerusalem: The Floersheimer Institute for Policy Studies Publications on
Reli- gion, Society, and State in Israel.
Henrich, Joseph, Robert Boyd, Samuel Bowles, Colin Camerer, Ernst Fehr, Herbert Gintis,
and Richard McElreath. 2001. “In search of homo economicus: behavioral experiments
in 15 small-scale societies.” The American Economic Review 91(2): 73–78.
Hill, Kim Quaile, and Jan E. Leighley. 1999. “Racial Diversity, Voter Turnout, and Mobilizing Institutions in the United States.” American Politics Research 27(July): 275–295.
44
Hopkins, Daniel J. 2010. “Politicized Places: Explaining Where and When Immigrants
Provoke Local Opposition.” American Political Science Review 104(1): 40–60.
Huffman, Matt L, and Philip N Cohen. 2004. “Racial Wage Inequality: Job Segregation
and Devaluation across US Labor Markets1.” American Journal of Sociology 109(4):
902–936.
Ichino, Nahomi, and Noah L. Nathan. 2012. “Crossing the Line: Local Ethnic Geography
and Voting in Ghana.” American Political Science Review pp. 1–18.
Jackson, Ken. 2013. “Diversity and the Distribution of Public Goods in Sub-Saharan Africa.”
Journal of African Economies 22(3): 437–462.
Kasara, Kimuli. 2012. “Local Ethnic Segregation and Violence.” Unpublished working paper,
Columbia University .
Kasara, Kimuli. 2013. “Separate and Suspicious: The Social Environment and Inter-Ethnic
Trust in Kenya.” The Journal of Politics .
Key, V.O. 1949. Southern Politics in State and Nation. New York: Knopf.
Koopmans, Ruud, and Susanne Veit. 2013. “Cooperation in Ethnically Diverse Neighborhoods: A Lost-Letter Experiment.” Political Psychology .
Kraus, Stephen J. 1995. “Attitudes and the prediction of behavior: A meta-analysis of the
empirical literature.” Personality and social psychology bulletin 21(1): 58–75.
Leighley, Jan E., and Arnold Vedlitz. 1999. “Race, Ethnicity, and Political Participation:
Competing Models and Contrasting Explanations.” The Journal of Politics 61(4): 1092–
1114.
Lieberman, Evan S, and Gwyneth H McClendon. 2013. “The Ethnicity–Policy Preference
Link in Sub-Saharan Africa.” Comparative Political Studies 46(5): 574–602.
Lim, May, Richard Metzler, and Yaneer Bar-Yam. 2007. “Global pattern formation and
ethnic/cultural violence.” Science 317(5844): 1540–1544.
Lodge, Milton, and Charles S Taber. 2013. The rationalizing voter. Cambridge University
Press.
Massey, Douglas S., and Nancy A. Denton. 1993. American Apartheid: Segregation and the
Making of the Underclass. Cambridge, MA: Havard University Press.
Matthews, Donald R., and James W. Prothro. 1963. “Social and Economic Factors and
Negro Voter Registration in the South.” The American Political Science Review 57(1):
24–44.
Miguel, Edward, and Mary Kay Gugerty. 2005. “Ethnic diversity, social sanctions, and
public goods in Kenya.” Journal of Public Economics 89(11): 2325–2368.
45
Mizrachi, Nissim, and Hanna Herzog. 2012. “Participatory destigmatization strategies among
Palestinian citizens, Ethiopian Jews and Mizrahi Jews in Israel.” Ethnic and Racial
Studies 35(3): 418–435.
Newman, Benjamin J. 2012. “Acculturating Contexts and Anglo Opposition to Immigration
in the United States.” American Journal of Political Science .
Oliver, J. Eric. 2010. The Paradoxes of Integration: Race, Neighborhood, and Civic Life in
Multiethnic America. Chicago: University of Chicago Press.
Oliver, J. Eric, and Janelle Wong. 2003. “Intergroup Prejudice in Multiethnic Settings.”
American Journal of Political Science 47(4): 567–82.
Oliver, J. Eric, and Tali Mendelberg. 2000. “Reconsidering the Environmental Determinants
of White Racial Attitudes.” American Journal of Political Science 44(3): 574–89.
Pettigrew, Thomas, and Linda Tropp. 2006. “A Meta-Analytic Test of Intergroup Contact
Theory.” Journal of Personality and Social Psychology 90(5).
Posner, Daniel N. 2004a. “Measuring ethnic fractionalization in Africa.” American Journal
of Political Science 48(4): 849–863.
Posner, Daniel N. 2004b. “The political salience of cultural difference: why Chewas and
Tumbukas are allies in Zambia and adversaries in Malawi.” American Political Science
Review 98(4).
Putnam, Robert D. 2007. “E pluribus unum: Diversity and community in the twenty-first
century the 2006 Johan Skytte Prize Lecture.” Scandinavian political studies 30(2):
137–174.
Quillian, Lincoln. 1995. “Prejudice as a Response to Perceived Group Threat: Population
Composition and Anti-immigrant and Racial Prejudice in Europe.” American Sociological Review 60(4): 586–611.
Rubin, Barry. 2012. Israel: An Introduction. New Haven: Yale University Press.
Sampson, Robert J. 2008. “Moving to Inequality: Neighborhood Effects and Experiments
Meet Social Structure.” American Journal of Sociology 114(1): 189–231.
Sandler, Shmuel, and Aaron Kampinsky. 2009. “Israels Religious Parties.” In Contemporary
Israel, ed. Robert O. Freedman. Philadelphia: Westview Press pp. 77–95.
Schelling, Thomas C. 1971. “Dynamic Models of Segregation.” The Journal of Mathematical
Sociology 1.
Schoeni, Robert F., Frank Stafford, Katherine A. Mcgonagle, and Patricia Andreski. 2013.
“Response Rates in National Panel Surveys.” The Annals of the American Academy of
Political and Social Science 645(1): 60–87.
46
Sigelman, Lee, and Susan Welch. 1993. “The Contact Hypothesis Revisited: Black-White
Interaction and Positive Social Attitudes.” Social Forces 71(3): 781–95.
Sniderman, Paul M, Richard A Brody, and Phillip E Tetlock. 1993. Reasoning and choice:
Explorations in political psychology. Cambridge University Press.
Spence, Lester K., and Harwood McClerking. 2010. “Context, Black Empowerment, and
African American Political Participation.” American Politics Research 38(5): 909–930.
Stephens-Davidowitz, Seth. 2013. “The Effects of Racial Animus on a Black Presidential
Candidate: Using Google Search Data to Find What Surveys Miss.”.
Tajfel, Henri. 1969. “Cognitive Aspects of Prejudice.” Journal of Social Issues 25: 29–97.
Tajfel, Henri, M. G. Billig, R. P. Bundy, and Claude Flament. 1971. “Social categorization
and intergroup behaviour.” European Journal of Social Psychology 1(2): 149–78.
Taylor, Marylee. 1998. “How White Attitudes Vary with the Racial Composition of Local
Populations: Numbers Count.” American Sociological Review 63(4): 512–35.
Tesler, Michael, and David O. Sears. 2010. Obama’s Race. Chicago: University of Chicago
Press.
Uslaner, Eric M. 2011. “Trust, Diversity, and Segregation in the United States and the
United Kingdom.” Comparative Sociology 10(2): 221–247.
Uslaner, Eric M. 2012. Segregation and Mistrust: Diversity, Isolation, and Social Cohesion.
Cambridge University Press.
Vanhanen, Tatu. 1999. “Domestic Ethnic Conflict and Ethnic Nepotism: A Comparative
Analysis.” Journal of Peace Research 36(1).
Voss, D. Stephen. 1996. “Beyond Racial Threat: Failure of an Old Hypothesis in the New
South.” The Journal of Politics 58(4): 1156–70.
Welch, Susan, Lee Sigelman, Timothy Bledsoe, and Michael Combs. 2001. Race and Place:
Race Relations in an American City. New York: Cambridge University Press.
Wright, Gerald. 1977. “Contextual Models of Electoral Behavior: The Southern Wallace
Vote.” American Political Science Review 71(2): 497–508.
Wright, Matthew. 2011. “Diversity and the Imagined Community: Immigrant Diversity and
Conceptions of National Identity.” Political Psychology 32(5): 837–862.
Zingher, Josh, and Thom Steen. forthcoming. “The Spatial and Demographic Determinants
of Racial Threat: A Demonstration from Louisiana.” Social Science Quarterly .
47
A
Appendix
A.1
Relationship between Ethnic Fractionalization and Spending
50
40
10
20
30
Govt. Spending % GDP
60
70
Figure A.1: Israel, Ethnic Fractionalization and Spending as % of GDP
0.0
0.2
0.4
0.6
Ethnic Fractionalization
Data taken from (Fearon 2003) and the World Bank 2010 dataset
48
0.8
A.2
Observational Data
Prior to undertaking our on the ground experimental research, we used existing observational
data in order to establish the systematic variation in inter-group attitudes across cities in
Israel. We relied on the Guttman-Avichai survey from 2009, which focuses on beliefs, values
and social tensions within the Israeli-Jewish population. Respondents were presented with
the following question: “In your opinion, how good are the relations between the religious
and nonreligious?”, and asked to choose from four possible answers: very good, fairly good,
not so good and not good at all.14
With 2,192 observations taken from 41 towns in Israel, the observational data supported
our initial intuitions. We display these results in Table A.1. As expected, a more uneven
distribution of the Jewish population corresponds with worse perceptions of relations between
the two groups. This holds true for both the UO and secular respondents. When UO
population is a minority in a mostly secular city, as is usually the case in mixed cities,
the interaction of diversity and segregation leads to more negative perceptions of relations
between religious and non-religious Jews.
In order to further test our intuitions, we conducted a short Internet-based survey (N=98)
and a lab-based pilot experiment (N=15), concentrating on perceptions of intergroup relations in Israel and the determinants of residential patterns. Based on these preliminary
explorations, we designed the research instrument to focus on the following: the mechanisms of public goods provision (through experimental decision-making games), attitudes
toward and interpersonal contact with different groups in Israeli society, perceptions of discrimination of groups by the government, demographics and other identity questions, and
residential preferences and movement. We now turn to discuss to analyze the results of the
lab-in-the-field experiments we conducted.
14
In order to examine the robustness of our results, we examined also responses to the following question:
“to what extent would you consent for your children to attend an elementary school that enrolls pupils from
both religious and secular families?”. The results are substantively identical, suggesting that both questions
capture general perceptions of the intra-Jewish religious cleavage in Israel.
49
Table A.1: Observational Data
Relations Between Religious and Non-Religious:
1=Very good; 4=Not at all good
UO
Non-UO
Secular
−0.459
(0.231)
−0.033
(0.420)
−0.221
(0.585)
0.042
(0.102)
0.478∗∗∗
(0.098)
0.246∗∗∗
(0.105)
0.025
(0.138)
0.633∗∗∗
(0.148)
0.259∗
(0.142)
0.149
(0.207)
76.12∗∗∗
(18.304)
27.097∗∗∗
(6.938)
−5.602∗∗∗
(1.394)
−0.009
(0.014)
−80.702∗∗∗
(20.161)
147
−0.216∗∗
(0.091)
1.894
(1.361)
3.503
(3.864)
0.854
(0.771)
−0.003
(0.010)
−9.817
(11.554)
2,045
−0.182
(0.132)
4.296∗∗∗
(1.651)
8.672∗
(4.722)
1.454
(1.026)
0.015
(0.015)
−26.649∗
(14.102)
950
Gender (female = 1)
Average income
Low income
Baseline: High income
Academic Degree (1 = yes)
Segregation
Outgroup Proportion
Jerusalem
Arab Population
Segregation x Outgroup Percent
N
Standard errors are clustered at the city level.
A.3
Descriptive Statistics of Experimental Locations and Subject
Sample
Table A.2 describes the experimental locations for our study, including Jerusalem, which
we separated into four “Quarter” as is done by the Central Bureau of Statistics. Table
compares or sample to a nationally representative Jewish sample included in the GuttmanAvichai survey from 2009.
50
Table A.2: Descriptions of Experimental Locations
Locations
Number of residents
Share of UO
Dissimilarity
score
Arab population
Ashdod
Kiryat Malachi
Elad
Arad
Bet Shemesh
Kiryat Gat
Haifa
B’nei Brak
Tveria
Safed
Rehovot
Zichron Yakkov
Ofakim
Netivot
Modi’in-Makabim
Tel Aviv
Jerusalem
204,300
20,500
33,900
23,400
72,700
47,500
264,300
151,800
41,600
29,600
111,100
18,300
24,000
26,100
69,300
402,600
759,700
15%
7%
88%
15%
46%
10%
4%
95%
13%
21%
8%
12%
24%
44%
0
4%
31%
0.91
0.87
0
0.86
0.71
0.92
0.94
0.37
0.63
0.57
0.89
0.79
0.92
0.72
0
0.9
0.82
No
No
No
No
No
No
Around 20%
No
Less than 10%
Less than 10%
No
No
No
No
No
Around 10%
Around 35%
Jerusalem Quarters
Type of Neighborhood
Dissimilarity
score
Arab population
Neve Yaakov
Ramat Shlomo
City Center
Kityat Yovel
Mixed
Mostly UO
Mixed
Mostly secular
0.81
0.57
0.68
0
6%
No
No
6%
Table A.3: Sample Description
Male
Academic education
Ethnic origin: Sephardic
Income above average
Income below average
Average age
Average left-right self-placement
Born in Israel
N
Full sample
UO
Non-UO
Gutmann Sample
51%
35%
49%
15%
68%
40
5.2
78%
549
45%
40%
41%
9%
76%
39
5.3
87%
216
56%
30%
56%
20%
61%
42
5
70%
243
48%
34%
48%
39%
34%
42
4.7
64%
2437
51
A.4
Measurement of the UO Population
Our primary tests look at behavioral differences as a function of size and segregation of UO
and secular populations. In order to assess the number of UO, the Israel Central Bureau
of Statistics uses different measurements. This measurement challenge reflects a deeper
problem of classification. The definition of being part of the UO is mostly subjective and
its meaning may change over time and place or across individuals. Furthermore, the UO
are far from a cohesive social group. There are multiple religious strands within the UO
community. In addition, the intra-Jewish ethnic cleavage between Ashkenazi and Mizrahi
UO has remained especially salient among the UO over the years, and has been a continuous
source of tension within this group.
The CBS deals with these challenges by providing different measurements for the UO population, among them the percent of voters casting ballots for parties known to be supported
by Ultra-Orthodox voters and the percent of male youth enrolled in schools for learning
traditional Jewish religious texts (Yeshivas, which are attended almost exclusively by Ultra
Orthodox). We also obtained city-level measures of Ultra Orthodox population based on expert assessments by Israeli academics (Cahaner 2009). All three of these measures are highly
correlated and yield similar results in the analysis below. We report our findings based on
segregation and group size constructed from the percent of youth in Yeshivas because we
believe it is a more reliable measure than voting percentages, which may be affected by
unknown processes.
52
A.5
Distribution of Responses to Key Tests
Figure A.2: Spending priority distributions of UO and Secular
UO Spending Priorities
0.46
0.5
0.4
0.32
0.3
0.2
0.1
0.08
0.06
0.04
labor
other
0.0
defense
education
inequality
Secular Spending Priorities
0.5
0.39
0.4
0.34
0.3
0.2
0.14
0.06
0.1
0.03
0.0
defense
education
inequality
labor
other
Responses of UO and secular populations to question “Which of the following is the most important item on
which the government should spend money”?
53
Table A.4: Play in the Public Goods Game
Outgroup
Ingroup Cooperate
Defect
Cooperate
Defect
206 (47.8)
42 (9.7)
248 (57.5)
108 (25.1) (25.1)
102 (23.7) (23.7)
210 (48.7)
Play of all UO and-UO in public goods game with option of cooperating and contribution 20 NIS or defecting
and keeping 20 NIS.
Table A.5: Play in the Public Goods Game
Outgroup
Ingroup Cooperate
Defect
Cooperate
Defect
197 (59.9)
33 (10)
230 (69.9)
99 (30.1) (30.1)
0 (0)
(0)
99 (30.1) (30.1)
Play of UO and non-UO in public goods game with option of cooperating and contribution 20 NIS or defecting
and keeping 20 NIS. This table is only “non-egoists”. Egoists are players that opted not to cooperate regardless
of the identity of the opposing player.
A.6
Contextual Variation on Behavioral Strategies
54
Table A.6: Regressions of ingroup preference in task efficacy on contextual variables
UO
Intercept
Segregation
Outgroup Proportion
Male
Age
Mixed Ethnicity
Other Ethnicity
Sephardic Ethnicity
Political Ideology
High Income
Low Income
Arab Population
Jerusalem
Segregation x Outgroup Percent
N
R2
adj. R2
Resid. sd
−1.24
(1.08)
3.55
(2.97)
1.60
(1.34)
−0.02
(0.08)
0.00∗
(0.00)
−0.24∗
(0.12)
−0.06
(0.22)
−0.15
(0.08)
0.03
(0.03)
−0.06
(0.12)
0.21∗
(0.10)
0.49
(1.08)
−0.34
(0.19)
−3.99
(3.43)
200
0.14
0.08
0.48
Non UO
0.95∗
(0.26)
−0.07
(0.37)
−1.38∗
(0.65)
0.01
(0.09)
−0.00
(0.00)
0.06
(0.08)
0.35∗
(0.10)
−0.07
(0.08)
−0.02
(0.03)
−0.03
(0.10)
0.10
(0.06)
−0.41
(1.05)
0.07
(0.07)
2.32
(1.53)
189
0.07
−0.00
0.44
Regression of ingroup preference in task efficacy for UO (column 1) and non-UO (column 2) on contextual
and individual level variables. Ingroup preference in task efficacy is the tendency to choose the ingroup
member in the Lego task. Higher numbers represent greater discrimination against the outgroup. Clustered
standard errors in parentheses. ∗ indicates significance at p < 0.05
A.7
Results of Regressions Subsetting for Selection
Figure A.4 is an example of the “design your neighborhood” task. Subjects were allowed
to hypothetically choose their own neighbors, while keeping everything else about their
neighborhood constant. The instructions in Figure A.4 (a) are:
Many people are pleased with where they live. But were you to to move into a
new place, imagine you would buy a house you like and that you can afford. We
want to know in which neighborhood you would have liked to buy your house.
Which people would you have liked to live in the area? Imagine that your house
55
Table A.7: Regressions of ingroup bias in public goods cooperation on contextual variables
UO
Non UO
∗
Intercept
Segregation
Outgroup Proportion
Male
Age
Mixed Ethnicity
Other Ethnicity
Sephardic Ethnicity
Political Ideology
High Income
Low Income
Arab Population
Jerusalem
Segregation x Outgroup Percent
N
R2
adj. R2
Resid. sd
−1.46
(0.53)
4.22∗
(1.51)
2.04∗
(0.63)
0.08
(0.06)
0.00
(0.00)
0.06
(0.15)
−0.14
(0.10)
0.10
(0.08)
−0.02
(0.03)
0.08
(0.12)
0.14
(0.08)
1.38∗
(0.67)
−0.33∗
(0.10)
−5.43∗
(1.81)
164
0.12
0.04
0.44
0.70∗
(0.21)
−0.69
(0.41)
−1.88∗
(0.62)
−0.07
(0.06)
−0.00
(0.00)
−0.00
(0.14)
0.36
(0.19)
0.06
(0.09)
0.03
(0.03)
0.08
(0.15)
−0.05
(0.14)
−0.90
(0.67)
−0.20∗
(0.06)
2.80
(1.55)
142
0.13
0.05
0.46
Regression of ingroup bias in public goods game for UO (column 1) and non-UO (column 2) on contextual
and individual level variables. Ingroup bias in the public goods games is decision to cooperate with ingroup
member and not outgroup member. Higher numbers represent greater discrimination against the outgroup.
Clustered standard errors in parentheses. ∗ indicates significance at p < 0.05
is not affected by the people who live next to you. Look at the neighborhood
below, and imagine you are buying the house in the middle of the neighborhood.
Next to you there are 10 empty houses. Click on the box next to each house in
order to choose which family will reside in it. Please choose residents for each of
the 10 houses in the neighborhood.
56
Figure A.3: Levels of differences in tastes, other regarding preferences, perceived efficacy,
and cooperation by city
T
Ashdod
R
E
C
T
Neve Yaakov
R
E
C
T
Ofakim
R
E
C
T
Bet Shemesh
E
T
Zichron Yaakov
R
C
R
E C
T
Jerusalem Center
R
E
C
T
Ramat Shlomo
R
E
C
TR
Kiryat Gat
E
C
T
Netivot
R
EC
T
Tveria
R
E
C
T
Rehovot
R
E
C
T
Haifa
R
CE
R
Arad
EC
T
R
Kiryat Malachi
T
E
T
Bnei Brak
C
TR
E
Elad
R
CE
T
Safed
C
T
Kiryat Yovel
C
R
E
R
E
E
T
Tel Aviv
C
R
Modiin Makabim Reut
C
R
E
C
−0.5
−0.3
−0.1
0.1
0.3
0.5
0.7
Cities are arranged by levels of segregation. For each city, we display four quantities of interest: 1) differences in tastes for spending priority on defense between secular and Ultra Orthodox (T); 2) standardized
differences in other regarding preferences for the ingroup over the outgroup for all subjects (R); 3) perceived
efficacy advantage from working with the ingroup over the outgroup for all subjects (E); and 4) probability of
cooperation with an outgroup member (C). Ninety-five percent confidence intervals are the horizontal bars.
57
Figure A.4: “Design Your Neighborhood” task
(a)
(b)
Example of the “design your neighborhood” task used as a direct measure of neighborhood preference. (a) is
the task with a blank neighborhood. (b) shows the interactive task as the neighborhood begins to change.
58
59
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
−13.49
−15.35
−7.17
−8.09
2.28
−2.75
−11.24
−5.95
−6.55
(12.09)
(13.56)
(18.31)
(10.77)
(11.76)
(11.77)
(14.46)
(16.82)
(13.66)
39.48
51.96
30.08
41.22
15.43
31.17
45.08
31.02
31.26
(25.85)
(38.73)
(36.40)
(25.55)
(29.44)
(25.41)
(31.89)
(30.40)
(32.29)
15.34
14.12
4.69
10.52
−1.81
3.56
8.89
0.06
4.91
(12.01)
(18.40)
(20.59)
(13.62)
(15.99)
(13.95)
(16.37)
(16.24)
(16.49)
−0.32
−0.57
0.84
0.18
−0.37
−0.32
0.25
2.30
0.10
(1.36)
(1.20)
(1.10)
(1.20)
(1.12)
(1.10)
(2.10)
(1.40)
(0.92)
−0.02
0.11∗
0.05
0.04
0.02
0.05
−0.01
0.08
0.04
(0.06)
(0.04)
(0.03)
(0.04)
(0.03)
(0.03)
(0.04)
(0.07)
(0.03)
−0.48
−1.22
−0.04
−1.00
−2.07
−2.71
−2.93
−1.66
−2.12
(1.87)
(2.39)
(2.05)
(1.85)
(1.81)
(1.57)
(2.84)
(1.63)
(1.50)
0.44
−1.94
−3.15
−2.23
−2.00
−2.85∗
(1.80)
(1.27)
(1.60)
(1.33)
(1.44)
(1.36)
−0.19
−0.99
−0.98
0.32
−2.37
−0.79
0.96
−2.48
−1.30
(1.71)
(1.36)
(1.62)
(1.07)
(1.36)
(1.51)
(1.39)
(1.91)
(1.03)
0.37
−0.19
0.07
−0.32
−0.43
−0.37
0.90
−0.77
−0.17
(0.72)
(0.63)
(0.56)
(0.48)
(0.33)
(0.45)
(0.55)
(0.72)
(0.37)
2.35
1.33
0.47
2.07
−0.71
−2.45
1.48
1.81
(2.58)
(4.51)
(2.92)
(2.84)
(2.52)
(2.64)
(2.00)
(2.32)
1.70
2.26
1.88
1.98
0.44
−0.09
3.91∗
2.14
(1.38)
(1.94)
(1.57)
(1.64)
(1.57)
(2.68)
(1.81)
(1.57)
1.05
28.09
29.32∗
46.20∗
17.27
43.78∗
97.24∗
25.14
31.29∗
(22.91)
(17.62)
(12.52)
(14.91)
(17.52)
(16.45)
(20.52)
(19.67)
(13.09)
0.19
−0.90
−0.62
−1.52
−1.38
−0.59
−0.28
−0.59
−0.84
(0.93)
(2.17)
(1.46)
(1.30)
(1.90)
(1.29)
(1.92)
(1.18)
(1.86)
−51.85
−59.94
−32.86
−48.33
−10.46
−35.00
−62.49
−24.46
−31.68
(33.48)
(46.59)
(42.93)
(29.88)
(37.20)
(30.35)
(36.70)
(36.18)
(37.73)
51
118
140
151
147
141
74
83
200
0.24
0.17
0.10
0.10
0.07
0.14
0.28
0.25
0.09
−0.02
0.08
0.01
0.02
−0.02
0.05
0.13
0.11
0.03
4.98
6.87
6.91
6.98
7.25
6.77
6.77
6.61
7.13
Regression of in-group bias in dictator game for Ultra Orthodox on contextual and individual level variables. In-group preference in the Dictator Game
is the amount in NIS contributed to ingroup members minus the amount contributed to outgroup members. In both cases, higher numbers represent
greater discrimination against the outgroup. Clustered standard errors in parentheses. Columns represent the following subsets of subjects: 1) people
that do not report preferring to live exclusively with the ingroup, if they had the choice, 2) people that are not satisfied with where they live, 3) people
that do not feel that they are free to move if they wanted to, 4) low income people, 5) people who do not list religious people as a reason affecting their
choice of city of residence, 6) people who do not list religious people as a reason affecting their choice of neighborhood of residence, 7) people who
did not relocate to an area with less of the outgroup, 8) people that did not design a homogeneous ideal neighborhood, 9) all respondents. ∗ indicates
significance at p < 0.05.
N
R2
adj. R2
Resid. sd
Segregation x Outgroup Percent
Jerusalem
Arab Population
Low Income
High Income
Political Ideology
Sephardic Ethnicity
Other Ethnicity
Mixed Ethnicity
Age
Male
Outgroup Proportion
Segregation
Intercept
Table A.8: Regressions of ingroup bias in other regarding behavior on contextual variables, subsetting to control for selection,
for UO
60
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
9.97
12.39∗
15.17∗
8.93
12.75∗
11.33∗
8.56
4.81
11.69∗
(6.70)
(4.97)
(4.79)
(8.24)
(5.53)
(5.65)
(5.45)
(6.83)
(5.56)
−18.30∗ −13.33∗ −10.85∗ −15.15
−13.00∗ −14.13∗ −12.77
−9.32
−13.88∗
(8.64)
(6.29)
(5.30)
(7.99)
(6.02)
(6.20)
(7.18)
(5.45)
(6.03)
−44.48∗ −26.76
−20.38∗ −31.52∗ −31.03∗ −42.35∗ −40.48∗ −30.06∗ −41.89∗
(13.54)
(14.79)
(8.31)
(11.39)
(9.20)
(9.28)
(8.87)
(9.82)
(8.89)
−1.22
−2.40
−0.94
−1.55
−1.35
−0.17
−0.09
0.35
−0.47
(2.00)
(1.21)
(1.52)
(1.73)
(1.20)
(1.15)
(1.41)
(1.86)
(1.20)
0.02
−0.03
−0.06
−0.01
−0.06
−0.05
−0.06
−0.04
−0.05
(0.05)
(0.04)
(0.04)
(0.05)
(0.03)
(0.03)
(0.04)
(0.05)
(0.03)
−3.63
−1.50
−2.42
−5.29∗
−3.52∗
−3.38∗
−3.41∗
−1.98
−3.56∗
(3.02)
(2.25)
(1.49)
(2.55)
(1.12)
(1.25)
(1.49)
(1.72)
(1.21)
0.43
0.91
1.63
0.57
1.98
3.42
3.63
1.82
2.89
(5.44)
(3.75)
(4.05)
(6.19)
(3.68)
(3.57)
(3.95)
(5.42)
(3.42)
−2.66
−2.53
−1.47
−2.69
−2.43∗
−2.76∗
−2.49∗
−1.68
−2.94∗
(2.20)
(1.39)
(1.30)
(1.79)
(1.18)
(1.15)
(1.25)
(1.66)
(1.09)
0.46
0.17
−0.24
0.35
0.28
0.19
0.61
0.19
0.16
(1.17)
(0.83)
(0.68)
(1.07)
(0.87)
(0.85)
(0.83)
(1.03)
(0.83)
0.94
−2.95
−4.04∗
−2.25
−1.56
−1.80
0.00
−1.23
(2.35)
(2.44)
(1.55)
(1.57)
(1.71)
(1.80)
(2.04)
(1.65)
−1.88
−2.84∗
−2.72
−2.31
−1.05
−1.30
−1.42
−1.10
(2.71)
(1.38)
(1.55)
(1.25)
(1.53)
(1.85)
(1.67)
(1.53)
2.79
−10.45
−5.44
3.89
4.37
13.51
23.29
17.40
7.15
(24.90)
(21.71)
(11.05)
(18.68)
(15.24)
(13.00)
(12.28)
(19.84)
(16.00)
−2.05
−1.94
−2.17∗
−1.58
−1.54∗
−1.69∗
−1.54∗
−0.05
−1.76∗
(2.11)
(2.04)
(0.83)
(1.30)
(0.59)
(0.82)
(0.78)
(1.80)
(0.87)
97.88∗
78.57
49.75∗
80.89∗
69.61∗
93.33∗
90.15∗
75.98∗
95.40∗
(37.65)
(40.66)
(23.39)
(33.20)
(25.00)
(26.60)
(25.85)
(26.81)
(25.76)
103
113
145
116
178
177
162
118
189
0.09
0.14
0.09
0.06
0.08
0.08
0.10
0.07
0.08
−0.04
0.02
−0.01
−0.03
0.01
0.00
0.02
−0.05
0.01
9.34
7.52
7.80
8.73
7.93
8.20
8.03
8.02
8.17
Regression of in-group bias in dictator game for Non-UO on contextual and individual level variables. In-group preference in the Dictator Game
is the amount in NIS contributed to ingroup members minus the amount contributed to outgroup members. In both cases, higher numbers represent
greater discrimination against the outgroup. Clustered standard errors in parentheses. Columns represent the same subsets listed in A.9. ∗ indicates
significance at p < 0.05.
N
R2
adj. R2
Resid. sd
Segregation x Outgroup Percent
Jerusalem
Arab Population
Low Income
High Income
Political Ideology
Sephardic Ethnicity
Other Ethnicity
Mixed Ethnicity
Age
Male
Outgroup Proportion
Segregation
Intercept
Table A.9: Regressions of ingroup bias in other regarding behavior on contextual variables, subsetting to control for selection,
for Non-UO
61
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
0.20
−1.86
−1.43
−1.50
−0.92
−1.57
−3.18∗
−1.05
−1.24
(0.71)
(1.17)
(1.16)
(1.25)
(1.19)
(1.16)
(0.88)
(1.26)
(1.08)
1.97
4.32
4.04
4.25
3.21
5.06
8.08∗
4.00
3.55
(1.31)
(3.40)
(3.26)
(3.34)
(3.17)
(3.19)
(2.32)
(2.84)
(2.97)
0.67
1.90
1.91
1.98
1.46
2.33
3.70∗
1.70
1.60
(0.61)
(1.52)
(1.43)
(1.49)
(1.45)
(1.43)
(1.04)
(1.53)
(1.34)
−0.01
−0.07
0.00
−0.07
−0.07
−0.11
−0.19
−0.11
−0.02
(0.13)
(0.08)
(0.08)
(0.08)
(0.09)
(0.09)
(0.11)
(0.15)
(0.08)
−0.00
0.01∗
0.00
0.01∗
0.00
0.00
0.01∗
0.00
0.00∗
(0.01)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
0.03
−0.18
−0.19
−0.20
−0.11
−0.26∗
−0.31
−0.11
−0.24∗
(0.17)
(0.14)
(0.15)
(0.14)
(0.16)
(0.12)
(0.18)
(0.24)
(0.12)
0.58∗
−0.07
−0.03
0.02
0.43∗
−0.06
(0.17)
(0.19)
(0.22)
(0.20)
(0.17)
(0.22)
−0.13
−0.12
−0.07
−0.15
−0.17∗
−0.15
0.06
−0.08
−0.15
(0.15)
(0.09)
(0.08)
(0.10)
(0.08)
(0.12)
(0.12)
(0.16)
(0.08)
−0.05
0.04
0.01
0.04
0.01
0.01
0.01
0.02
0.03
(0.07)
(0.04)
(0.03)
(0.04)
(0.04)
(0.03)
(0.04)
(0.03)
(0.03)
0.08
0.03
−0.12
−0.08
−0.12
0.21
0.19
−0.06
(0.36)
(0.17)
(0.15)
(0.17)
(0.12)
(0.26)
(0.34)
(0.12)
0.16
0.29∗
0.21∗
0.20
0.14
0.13
0.10
0.21∗
(0.23)
(0.13)
(0.10)
(0.14)
(0.11)
(0.15)
(0.18)
(0.10)
0.56
0.95
1.45
0.65
0.07
0.23
1.65
1.52
0.49
(1.93)
(1.68)
(1.85)
(0.93)
(0.93)
(1.20)
(1.37)
(0.94)
(1.08)
−0.42∗
−0.29
−0.37
−0.35
−0.36
−0.40∗
−0.54∗
−0.46∗
−0.34
(0.07)
(0.23)
(0.22)
(0.21)
(0.20)
(0.18)
(0.14)
(0.11)
(0.19)
−2.41
−4.66
−4.75
−4.63
−3.55
−5.98
−9.03∗
−5.00
−3.99
(1.77)
(3.92)
(3.72)
(3.82)
(3.76)
(3.70)
(2.60)
(3.49)
(3.43)
51
118
140
151
147
141
74
83
200
0.26
0.21
0.19
0.15
0.12
0.17
0.38
0.18
0.14
0.00
0.12
0.11
0.08
0.04
0.08
0.26
0.03
0.08
0.48
0.47
0.47
0.47
0.49
0.48
0.43
0.49
0.48
Regression of ingroup preference in task efficacy for UO on contextual and individual level variables. Ingroup preference in task efficacy is the tendency
to choose the ingroup member in the Lego task. Higher numbers represent greater discrimination against the outgroup. Clustered standard errors in
parentheses. Clustered standard errors in parentheses. Columns represent the same subsets listed in A.9. ∗ indicates significance at p < 0.05.
N
R2
adj. R2
Resid. sd
Segregation x Outgroup Percent
Jerusalem
Arab Population
Low Income
High Income
Political Ideology
Sephardic Ethnicity
Other Ethnicity
Mixed Ethnicity
Age
Male
Outgroup Proportion
Segregation
Intercept
Table A.10: Regressions of ingroup preference in task efficacy on contextual variables, subsetting to control for selection, for
UO
62
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
0.65
0.92∗
1.07∗
1.09∗
0.92∗
0.86∗
0.98∗
0.72∗
0.95∗
(0.38)
(0.36)
(0.31)
(0.32)
(0.29)
(0.29)
(0.27)
(0.36)
(0.26)
0.34
−0.04
−0.07
−0.26
0.06
0.02
0.07
0.23
−0.07
(0.59)
(0.49)
(0.45)
(0.54)
(0.40)
(0.39)
(0.39)
(0.61)
(0.37)
−1.06
−0.25
−0.22
−1.01
−1.30
−1.01
−1.14
−1.02
−1.38∗
(0.77)
(0.99)
(0.84)
(0.86)
(0.84)
(0.71)
(0.63)
(0.93)
(0.65)
0.01
0.01
−0.00
−0.05
−0.01
0.03
−0.03
−0.07
0.01
(0.13)
(0.10)
(0.10)
(0.12)
(0.09)
(0.10)
(0.10)
(0.09)
(0.09)
−0.00
−0.00
0.00
0.00
−0.00
−0.00
−0.00
−0.00
−0.01∗
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
0.15
0.01
−0.00
0.03
0.08
0.07
0.05
0.09
0.06
(0.12)
(0.13)
(0.11)
(0.13)
(0.07)
(0.08)
(0.06)
(0.08)
(0.08)
0.47∗
0.13
0.20∗
0.24∗
0.37∗
0.36∗
0.29∗
0.27∗
0.35∗
(0.14)
(0.12)
(0.08)
(0.09)
(0.12)
(0.09)
(0.14)
(0.11)
(0.10)
−0.07
−0.08
−0.05
−0.05
−0.04
−0.06
−0.10
−0.04
−0.07
(0.11)
(0.09)
(0.09)
(0.10)
(0.10)
(0.08)
(0.08)
(0.10)
(0.08)
0.00
−0.01
−0.03
−0.02
−0.02
−0.01
−0.00
0.02
−0.02
(0.05)
(0.04)
(0.03)
(0.03)
(0.03)
(0.03)
(0.03)
(0.04)
(0.03)
−0.06
0.07
−0.18
−0.03
−0.03
−0.06
−0.04
−0.03
(0.15)
(0.15)
(0.13)
(0.12)
(0.12)
(0.13)
(0.11)
(0.10)
0.17
0.02
−0.05
0.06
0.10
0.10
0.17∗
0.10
(0.10)
(0.10)
(0.07)
(0.07)
(0.07)
(0.08)
(0.08)
(0.06)
−2.25
−1.85
−0.75
−0.17
−0.47
−0.28
−0.90
−0.30
−0.41
(1.78)
(1.75)
(1.47)
(1.38)
(1.08)
(1.19)
(1.41)
(1.01)
(1.05)
0.18∗
0.05
0.08
0.12
0.08
0.09
0.01
0.29∗
0.07
(0.06)
(0.10)
(0.07)
(0.10)
(0.07)
(0.07)
(0.08)
(0.10)
(0.07)
1.35
0.43
−0.18
1.92
1.92
1.47
1.54
1.16
2.32
(1.81)
(2.08)
(1.86)
(2.02)
(1.80)
(1.67)
(1.40)
(2.11)
(1.53)
103
113
145
116
178
177
162
118
189
0.15
0.08
0.05
0.04
0.06
0.06
0.09
0.14
0.07
0.02
−0.04
−0.04
−0.06
−0.02
−0.02
0.01
0.03
−0.00
0.47
0.43
0.41
0.44
0.44
0.44
0.42
0.45
0.44
Regression of ingroup preference in task efficacy for non-UO on contextual and individual level variables. Ingroup preference in task efficacy is the
tendency to choose the ingroup member in the Lego task. Higher numbers represent greater discrimination against the outgroup. Clustered standard
errors in parentheses. Clustered standard errors in parentheses. Columns represent the same subsets listed in A.9. ∗ indicates significance at p < 0.05.
N
R2
adj. R2
Resid. sd
Segregation x Outgroup Percent
Jerusalem
Arab Population
Low Income
High Income
Political Ideology
Sephardic Ethnicity
Other Ethnicity
Mixed Ethnicity
Age
Male
Outgroup Proportion
Segregation
Intercept
Table A.11: Regressions of ingroup preference in task efficacy on contextual variables, subsetting to control for selection, for
Non-UO
63
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
−0.16
−1.93∗
−1.26∗
−1.31∗
−1.24∗
−1.03
−3.40∗
−0.05
−1.46∗
(1.69)
(0.63)
(0.62)
(0.63)
(0.49)
(0.57)
(1.41)
(0.98)
(0.53)
2.89
5.01∗
4.09∗
4.35∗
3.89∗
3.24∗
8.85∗
1.61
4.22∗
(3.25)
(1.37)
(1.55)
(1.72)
(1.32)
(1.34)
(2.81)
(2.54)
(1.51)
0.85
2.25∗
2.05∗
2.16∗
1.77∗
1.76∗
4.27∗
0.73
2.04∗
(1.37)
(0.65)
(0.66)
(0.72)
(0.65)
(0.62)
(1.40)
(1.13)
(0.63)
0.17
0.09
0.02
0.07
−0.01
0.05
0.18
0.05
0.08
(0.12)
(0.09)
(0.08)
(0.07)
(0.08)
(0.09)
(0.12)
(0.12)
(0.06)
−0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
0.21
0.11
0.07
0.10
0.07
−0.00
−0.08
0.20
0.06
(0.18)
(0.20)
(0.18)
(0.16)
(0.19)
(0.13)
(0.22)
(0.25)
(0.15)
0.04
−0.13
−0.14
−0.19
−0.22
−0.14
(0.09)
(0.10)
(0.12)
(0.12)
(0.18)
(0.10)
0.03
0.15
0.22∗
0.14
0.04
0.03
0.12
−0.06
0.10
(0.20)
(0.12)
(0.10)
(0.10)
(0.08)
(0.10)
(0.14)
(0.09)
(0.08)
−0.07
0.01
−0.05
−0.02
−0.02
−0.03
0.01
−0.06
−0.02
(0.10)
(0.03)
(0.03)
(0.03)
(0.03)
(0.03)
(0.04)
(0.05)
(0.03)
0.09
0.03
−0.03
0.17
0.02
0.17
0.27
0.08
(0.23)
(0.17)
(0.12)
(0.14)
(0.14)
(0.27)
(0.24)
(0.12)
0.16
0.22
0.20∗
0.16
0.15
0.08
0.08
0.14
(0.15)
(0.11)
(0.09)
(0.10)
(0.11)
(0.21)
(0.14)
(0.08)
1.55
2.93∗
0.37
1.35
1.69∗
0.65
2.04
−0.86
1.38∗
(1.62)
(1.09)
(0.71)
(0.83)
(0.71)
(0.55)
(1.36)
(1.14)
(0.67)
−0.31
−0.38∗
−0.30∗
−0.31∗
−0.33∗
−0.33∗
−0.59∗
−0.39∗
−0.33∗
(0.23)
(0.08)
(0.11)
(0.11)
(0.10)
(0.08)
(0.14)
(0.19)
(0.10)
−3.74
−6.48∗
−5.43∗
−5.79∗
−5.03∗
−4.36∗ −11.33∗
−2.28
−5.43∗
(3.92)
(1.72)
(1.81)
(2.05)
(1.66)
(1.60)
(3.43)
(2.98)
(1.81)
43
97
117
123
120
114
63
66
164
0.36
0.17
0.16
0.11
0.11
0.13
0.31
0.28
0.12
0.08
0.05
0.07
0.02
−0.00
0.02
0.14
0.10
0.04
0.38
0.46
0.44
0.46
0.45
0.44
0.42
0.41
0.44
Regression of ingroup bias in public goods game for UO on contextual and individual level variables. Ingroup bias in the public goods games is decision
to cooperate with ingroup member and not outgroup member. Clustered standard errors in parentheses. Clustered standard errors in parentheses.
Columns represent the same subsets listed in A.9. ∗ indicates significance at p < 0.05.
N
R2
adj. R2
Resid. sd
Segregation x Outgroup Percent
Jerusalem
Arab Population
Low Income
High Income
Political Ideology
Sephardic Ethnicity
Other Ethnicity
Mixed Ethnicity
Age
Male
Outgroup Proportion
Segregation
Intercept
Table A.12: Regressions of ingroup bias in public goods cooperation on contextual variables, subsetting to control for selection,
for UO
64
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9
0.85∗
0.77∗
0.72∗
0.82∗
0.77∗
0.75∗
0.64∗
0.44
0.70∗
(0.28)
(0.23)
(0.27)
(0.29)
(0.22)
(0.25)
(0.21)
(0.34)
(0.21)
−0.77
−0.71
−0.63
−1.02
−0.79
−0.91∗
−0.49
−0.74
−0.69
(0.52)
(0.45)
(0.44)
(0.71)
(0.43)
(0.41)
(0.47)
(0.48)
(0.41)
−2.20∗
−1.75
−1.91∗
−1.69
−2.07∗
−2.30∗
−2.01∗
−0.85
−1.88∗
(0.73)
(1.01)
(0.80)
(1.18)
(0.86)
(0.69)
(0.69)
(0.69)
(0.62)
−0.07
−0.28∗
−0.05
−0.13
−0.07
−0.04
−0.06
0.02
−0.07
(0.09)
(0.09)
(0.06)
(0.07)
(0.06)
(0.07)
(0.07)
(0.10)
(0.06)
0.00
−0.00
−0.00
0.00
0.00
0.00
−0.00
0.00
−0.00
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
(0.00)
0.04
−0.00
−0.01
0.10
−0.04
0.02
0.02
0.12
−0.00
(0.19)
(0.13)
(0.13)
(0.17)
(0.15)
(0.14)
(0.15)
(0.19)
(0.14)
0.37
0.48∗
0.55∗
0.32
0.36
0.32
1.05∗
0.36
(0.25)
(0.19)
(0.12)
(0.20)
(0.19)
(0.17)
(0.15)
(0.19)
−0.08
0.12
0.06
0.16
0.02
0.07
0.04
0.13
0.06
(0.14)
(0.09)
(0.08)
(0.11)
(0.10)
(0.09)
(0.09)
(0.13)
(0.09)
0.03
0.03
0.03
0.03
0.03
0.04
0.04
−0.00
0.03
(0.04)
(0.04)
(0.04)
(0.04)
(0.03)
(0.03)
(0.03)
(0.06)
(0.03)
−0.10
−0.07
−0.00
0.04
0.07
0.11
0.21
0.08
(0.19)
(0.21)
(0.17)
(0.14)
(0.15)
(0.17)
(0.17)
(0.15)
−0.06
−0.08
−0.10
−0.07
−0.09
−0.07
0.03
−0.05
(0.15)
(0.14)
(0.15)
(0.14)
(0.14)
(0.16)
(0.10)
(0.14)
−1.11
−0.03
−0.88
−2.10∗
−1.12
−0.76
−0.93
−0.42
−0.90
(1.08)
(0.98)
(1.00)
(0.87)
(0.75)
(0.78)
(0.63)
(0.79)
(0.67)
−0.23∗
−0.00
−0.16
−0.12
−0.23∗
−0.24∗
−0.28∗
−0.07
−0.20∗
(0.09)
(0.12)
(0.11)
(0.15)
(0.07)
(0.07)
(0.08)
(0.08)
(0.06)
3.37∗
2.85
3.02
2.47
3.20
3.81∗
3.09
1.08
2.80
(1.51)
(2.57)
(2.06)
(2.99)
(1.99)
(1.71)
(1.75)
(1.46)
(1.55)
77
85
114
88
132
133
125
90
142
0.18
0.17
0.13
0.14
0.13
0.15
0.15
0.15
0.13
0.02
0.02
0.01
0.03
0.04
0.06
0.05
0.01
0.05
0.46
0.46
0.47
0.46
0.47
0.47
0.47
0.42
0.46
Regression of ingroup bias in public goods game for non-UO on contextual and individual level variables. Ingroup bias in the public goods games is
decision to cooperate with ingroup member and not outgroup member. Higher numbers represent greater discrimination against the outgroup. Clustered
standard errors in parentheses. Clustered standard errors in parentheses. Columns represent the same subsets listed in A.9. ∗ indicates significance
at p < 0.05.
N
R2
adj. R2
Resid. sd
Segregation x Outgroup Percent
Jerusalem
Arab Population
Low Income
High Income
Political Ideology
Sephardic Ethnicity
Other Ethnicity
Mixed Ethnicity
Age
Male
Outgroup Proportion
Segregation
Intercept
Table A.13: Regressions of ingroup bias in public goods cooperation on contextual variables, subsetting to control for selection,
for Non-UO