bs_bs_banner

Political Psychology, Vol. 38, No. 3, 2017 doi: 10.1111/pops.12340

Reciprocity and Discrimination: An Experiment of Hindu-Muslim Cooperation in Indian Slums Andrej Tusicisny Columbia University

This article shows that indirect positive reciprocity triggered by experiencing short and superficial cooperation with outgroups’ individual members reduces discrimination of other members of the same group. The field research combined a lab-in-the-field experiment and a survey conducted in the slums of Mumbai, an Indian city notorious for Hindu-Muslim violence. After the treatment manipulated expectations of cooperative behavior, ethnically heterogeneous groups produced as much public goods in a public goods game as homogeneous groups. This positive experience radically reduced Hindu subjects’ discriminatory attitudes towards the Muslim minority after the experiment. The effect was equally strong among voters of two extremist parties implicated in ethnic riots. The survey compared reciprocity with alternative explanations of why people discriminate against some, but not other ethnic groups. Indirect positive reciprocity and intergroup contact are associated with less, and relative size of the outgroup with more discriminatory attitudes. KEY WORDS: ethnic discrimination, contact hypothesis, ethnic politics, India, experiment, survey

Why might the same person discriminate against members of some ethnic groups and not others? Whether it is selective distribution of patronage, voting along racial lines, or ethnic civil wars, many important political phenomena involve discrimination based on collective identities. At the same time, not all outsiders are subject to the same treatment. For example, Black Africans’ ethnic identification in South Africa has been associated with negative attitudes towards Afrikaans Whites, but not towards English Whites (Duckitt & Mphuthing, 1998). Much research has been done on the role of conflicting interests and values (Posner, 2005). Another strand of literature relates to intergroup conflict as arising from “natural” ingroup favoritism (Horowitz, 1985) or particular personality traits (Sidanius, 1993). Drawing on insights from social psychology, evolutionary biology, and behavioral economics, this article highlights a different causal factor contributing to discrimination—one that has been largely overlooked by the prejudicereduction literature. It argues that people discriminate against groups they do not trust to cooperate. Stereotypes about cooperation are subject to bounded rationality and can be updated by observing the behavior of outgroup members. Indirect positive reciprocity reduces discrimination against groups to which the cooperative individuals belong. The proposed argument provides a potential causal mechanism for the celebrated contact hypothesis (Allport, 1954). I tested observable implications in ethnically heterogeneous slums of Mumbai, India. Mumbai is a uniquely valuable test environment for theories of intergroup relations. In the slums, provision of 409 C 2016 International Society of Political Psychology 0162-895X V Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, 9600 Garsington Road, Oxford, OX4 2DQ, and PO Box 378 Carlton South, 3053 Victoria, Australia

410

Tusicisny

public goods requires bottom-up contributions channeled through civil society organizations that have to transcend various religious, linguistic, and caste boundaries. At the same time, Mumbai is wellknown for communal violence, including ethnic riots and terrorist attacks. Against this background, my original survey demonstrates that indirect positive reciprocity explains a great deal of variation in discrimination against different ethnic minorities. Experiencing cooperation with individual Muslims during a lab-in-the-field experiment also radically reduced ethnic discrimination of the generally disliked Muslim minority, when measured after the experiment. Indirect Reciprocity I propose reciprocity as an explanation of why people discriminate against some but not other groups and as a means to reduce discrimination. Experimental research shows that most people are conditional cooperators; they reward cooperation and punish defection (Chaudhuri, 2011; Fischbacher, G€achter, & Fehr, 2001; Ostrom, 2000). Possibly evolved as the society’s “immune response” against free riding (Nowak & Sigmund, 2000, p. 819), reciprocal cooperation has become one of the universal social norms, present in most if not all moral codes (Gouldner, 1960). Violations of this norm are routinely punished both in laboratory experiments and in real life (Bowles & Gintis, 2011). Punishment often targets not only the defector herself, but also the group of which she is a member. For example, in a recent experiment “subjects who were harmed by a partner’s uncooperative action reacted by harming other members of the partner’s group” (Hugh-Jones & Leroch, 2013, p. 1). This indirect negative reciprocity—punishment of a different group member—may be caused by attribution of negative emotions, such as anger and contempt (Fehr & G€achter, 2000; Sanfey, Rilling, Aronson, Nystrom, & Cohen, 2003), to a whole group (Lawler, 2001). Punishment of strangers for what other members of their group had allegedly done is quite common in ethnic conflict. According to Horowitz (2001), anger projected onto the target group precipitated ethnic riots in Burma, South Africa, Uzbekistan, and elsewhere. Petersen (2002) elucidated the role of group-centered negative emotions, such as resentment, in ethnic violence in Eastern Europe. One example of indirect negative reciprocity is statistical discrimination. Statistical discrimination arises when a difference in average experiences with members of different groups leads a person to use observable characteristics as a proxy for unobservable characteristics. For instance, a taxi driver in New York justified racial profiling of potential customers by arguing: “I have had a lot of bad experiences with black people, so whenever I see a black person in the street I look the other way” (Gambetta & Hamill, 2005, p. 165). In this case, he is using race as a proxy for customer behavior. Just as indirect negative reciprocity leads to discrimination of third parties because of their group membership, indirect positive reciprocity should reduce discrimination. I define indirect positive reciprocity by the following sequence of events:

1. Persons A and B are members of the same group. 2. Person A engages in a behavior beneficial to person C. 3. This leads person C to engage in a behavior beneficial to person B or to refrain from behavior harmful to B. I broadly define “group” as any social category in which an individual may be a member. Ingroup is a group in which an individual is eligible to be a member, and outgroup is a group in which an individual is not eligible to be a member. Although the proposed argument is general enough to be applicable to any group, the empirical test presented here focuses on ethnic groups. Why should indirect positive reciprocity work? I argue that there are two major causal channels. First, successful social exchange between persons A and C produces a positive emotion of gratitude in

Reciprocity and Discrimination

411

the same way as failed social exchange causes anger. If C attributes these positive feelings to A’s group, C is less likely to discriminate against other members of the same group (such as person B). The second channel is cognitive: C’s positive experience with A updates prior beliefs about A’s group. C’s more positive beliefs about the group prevent statistical discrimination of the group’s member B. It is a normal cognitive function to classify people into social categories based on visible characteristics, such as race or gender, and to view groups formed thus as homogenous in terms of personality traits or behavioral patterns (Fiske, 2000; Yzerbyt & Demoulin, 2010). Many different triggers can activate stereotypes about groups and lead to prejudice and discrimination. According to the theory of symbolic politics, people develop a limited set of highly stable symbolic predispositions, including partisanship, ideology, and ethnic prejudices, early in their lives (Sears & Funk, 1999). Stereotypes may also be formed by the mass media—though empirical evidence is mixed (Green & Seher, 2003)—and by direct experience (Bassili, 2008). Whatever their origins are, prejudices sometimes change in a dramatic way. For example, the number of the General Social Survey respondents stating that “sexual relations between two adults of the same sex” are “always wrong” decreased from 71% to 54% in just six years (P. R. Brewer, 2003). According to the influential contact hypothesis, people may lose their prejudices through intergroup contact under the optimal conditions of equal status, common goals, cooperative interdependence, and institutional support (Allport, 1954; Pettigrew, 1971). Positive contact also should be voluntary, nonsuperficial, and sustained (Pettigrew, 2008). However, even indirect contact—through mass media or having a friend with an outgroup friend—may reduce prejudice (Pettigrew, Tropp, Wagner, & Christ, 2011). Pettigrew and Tropp (2006) conducted a widely cited meta-analysis that included 713 separate samples from 515 studies on the contact hypothesis. They concluded that intergroup contact typically did reduce prejudice. This article highlights the importance of the information content transmitted during intergroup contact. I hypothesize that observing individual behavior can change group stereotypes about cooperative behavior. According to Weber and Crocker (1983), people respond to information that deviates from their stereotypes by updating the beliefs. Every new piece of evidence causes a minor change in the stereotype. Such change is in line with the so-called bookkeeping model (Rothbart, 1981). The subtyping model proposes a modified prediction: An individual disconfirming the stereotype is perceived as an exception to the rule unless she is representative enough of her group (Taylor, 1981). Taking one more step in the causal chain proposed in this article, Peffley, Hurwitz, and Sniderman (1997) found that information contradicting preexisting strong racial stereotypes changed discriminatory political attitudes. Without discarding other potential sources of information, I focus on direct experience because it generates stronger attitude changes (Bassili, 2008). However, in contrast to the contact hypothesis, I argue that the interaction does not need to be long or profound for updating to work. To bring this all together, people who have benefited from a cooperative act attributed to a member of an outgroup should be less likely to discriminate against this outgroup. Therefore, we should observe:

1) A negative correlation between discriminatory attitudes towards outgroups and help received from members of these groups in the past. 2) Recued discriminatory attitudes towards an outgroup as a result of benefiting from a cooperative act attributed to a member of that outgroup. To test these two hypotheses, I conducted a survey and a lab-in-the-field experiment respectively in ethnically heterogeneous slums in Mumbai.

412

Tusicisny

Intergroup Relations in Mumbai India’s biggest city has a very diverse population. Life in densely populated slums necessitates everyday interactions with various ethnic groups. These interactions vary from cooperation to exploitation to open ethnic conflict. Communal violence between Hindus and Muslims has been a major issue in Indian politics and Mumbai is India’s historically most riot-prone city, with 1,137 deaths in 1950–95 (Varshney, 2002). The latest recorded riot between Muslims and the mostly Hindu police occurred in August 2012. The single worst episode of the perennial ethnic conflict in Mumbai followed the destruction of the Babri Mosque in Ayodhya by Hindu extremists in December 1992. In that case, ethnic riots to a great degree orchestrated by the local extreme-nationalist party Shiv Sena, with the complicity of the police, left 900 dead (Srikrishna, 1998). Rioters razed nearly 10,000 houses and an estimated 200,000 people fled the city (Engineer, 1993). Large numbers of Muslims previously living among the Hindu majority left their scattered enclaves and moved to a few overcrowded, but more defensible, ghettos (Hansen, 2001; Shaban, 2010). During my fieldwork 18 years later, every conversation about ethnic relations still inevitably drifted to the topic of the 1990s riots. Although ethnic riots have become more of a terrifying memory than a serious threat, frequent terrorist attacks attributed to Islamist terrorist organizations regularly rekindle the simmering conflict between Hindus and Muslims living in Mumbai. The first large-scale terrorist attack came as a response of the Muslim criminal underworld to the 1993 riots. Ten powerful bombs exploded at major landmarks and killed 251 people. Incidentally, the most recent terrorist attack in Mumbai happened just after I finished my field research there. On July 13, 2011, bombs exploded at several crowded locations across the city; they killed 26 and injured 141 people. Hopefully, this study of people living in a violence-prone environment will help remedy the current state of our field, in which “there is little sustained experimental evaluation of conflict negotiation and reduction for the many millions of ordinary citizens living in conflict or postconflict settings” (Paluck & Green, 2009, p. 359). Among all ethnic groups living in Mumbai, the politically dominant Marathi-speaking Hindus (colloquially also called Maharashtrians) have a particularly salient social identity. This is the only ethnic group that often insists on outsiders using its language instead of the usual Bambaiya Hindi pidgin of Hindi, English, and Marathi. Many of them support the nationalist political parties Shiv Sena and the MNS (Palshikar & Deshpande, 1999; Shaban, 2010). The Shiv Sena movement was founded in Bombay in 1966 as a vehicle for the interests of the Marathi-speaking middle class against skilled South Indians. Over the years, Shiv Sena’s targets changed from South Indians to Communists to the city elite to Indian Muslims (Hansen, 2001). The field research focused on Maharashtrians because of their political importance. I sampled only Marathi-speaking men. The core electoral base of the nationalistic alliance of the BJP and Shiv Sena in Maharashtra came from the young males (Palshikar & Deshpande, 1999), and participants in ethnic riots are overwhelmingly male (Horowitz, 2001). A higher propensity for violence (Pinker, 2011) and discrimination (Fershtman & Gneezy, 2001) makes men a particularly interesting target group for discrimination-reducing interventions. Although there are no reliable demographic data collected at the level of neighborhoods, local experts helped me identify three predominantly Marathi slums: Bhoiwada, Magathane, and Shivaji Nagar. As is often the case in Mumbai—and in other ethnically diverse parts of the world—more or less homogenous buildings or streets lie in close proximity to the zones inhabited by other ethnic groups. For example, although all the buildings in Shivaji Nagar are the same corrugated iron huts, ubiquitous saffron or green flags visibly mark alternating “Hindu” and “Muslim” streets. The spatial proximity of different groups creates a quirky mix of intergroup competition and cooperation. Local violent nonstate actors, whether it be gangs in Muslim slums or extremist political parties in Hindu slums, enjoy impunity in their respective areas (Shaban, 2010). At the same time, common

Reciprocity and Discrimination

413

civic institutions are securing communal peace in and between these neighborhoods. After the deadly riots in 1993, each police station set up a mohalla committee consisting of local community leaders, who work to diffuse intergroup tensions at times of crisis. Similarly, provision of public goods in Mumbai’s slums requires bottom-up contributions by hundreds of residents belonging to different castes, religions, and linguistic groups. For example, slums located on hillsides do not get their water directly from the authorities, but through their own community-based organizations (CBO) that maintain pumps, collect money from members, and pay bills to the authorities. Survey Survey Design The survey tests whether indirect positive reciprocity is associated with weaker discriminatory attitudes towards different ethnic groups (Hypothesis 1). A “random route” technique was used to sample randomly 402 adult male Marathi-speaking slum dwellers—134 in each neighborhood. Among them, 210 (70 per neighborhood) were interviewed for the survey. The remaining 192 subjects (64 in each slum) participated in the experiment described in the next section. The questionnaire asked questions about four minorities: Muslims, Gujaratis, Biharis, and Parsis. As the same person could discriminate in different ways against these four ethnic groups, the unit of observation is the respondent-outgroup dyad. Each respondent is thus responsible for four observations in the dataset. Accordingly, standard errors in the analysis below are clustered at the level of individual respondents. I used two questions from Bogardus (1925) to measure discriminatory attitudes: “If it was up to you, would you accept a [group] as a neighbor?” and “If it was up to you, would you accept a [group] to close kinship by marriage?” The average of the two produced a composite indicator of discrimination. Similar questions have been used by Sniderman, Hagendoorn, and Prior (2004), Tausch, Hewstone, and Roy (2009), and the World Values Survey. To capture the effect of indirect positive reciprocity, the survey asked: “In the past 12 months, which people, if any, have helped you directly by giving you money or some of their time?” A separate dichotomous variable for each ethnic group was coded 1 if the respondent had received help from a member of that particular group. This question was inspired by Coleman (1988) and Phan, Blumer, and Demaiter (2009). Negative reciprocity is measured by the question “How often have you been taken advantage of by [group]?” Tausch et al. (2009) adopted this question from Stephan et al. (2002). This article measures intergroup contact by the question: “How often do you have informal talks with [group]?” It is based on Islam and Hewstone (1993) and has been used to study Hindu-Muslim relations in India by Tausch et al. (2009). Since we already measure reciprocity by separate variables, the contact variable captures the effect of intergroup contact on discrimination mediated by other factors apart from reciprocity: increased liking, reduced anxiety, or perceived outgroup variability (Pettigrew & Tropp, 2006). According to a view associated with realistic group conflict theory, competition for scarce resources drives intergroup conflict (Sherif, Harvey, White, Hood, & Sherif, 1961). Consequently, Posner (2004) predicted that relations should be worse between groups that are large enough to form a minimumwinning coalition required to seize control of the contested resource. A dominant group, such as Marathispeakers, should feel threatened more as the relative size of an ethnic minority increases (Blalock, 1967). Relative size of the outgroup is measured by the outgroup’s proportion of the total population of Mumbai. Muslims and Gujaratis represent 19% of the population each, Biharis 1.5%, and Parsis 0.5%. As perception of collective threat is ultimately subjective, I asked two questions from the ISSP National Identity Survey measuring the perceived threat to the economic well-being (“[Group] take

414

Tusicisny

jobs away from people like me”) and safety (“[Group] increase crime rates”). These are two of the strongest manifestations of intergroup anxiety (Blalock, 1967; Hardin, 1995; Quillian, 1995; Sniderman et al., 2004; Stephan & Stephan, 1985, 2000), which “stems mainly from the expectation of negative consequences for oneself during contact” (Islam & Hewstone, 1993, p. 701). Empirical research in psychology reviewed by Hewstone, Rubin, and Willis (2002) and Brown (2000) shows that high-status groups express more group bias than low-status groups. Their bias is especially strong if they perceive the status gap to be closing. Mirroring these laboratory results, Petersen (2002) argued that status reversals triggered ethnic violence in twentieth-century Eastern Europe. Relative status in this article is the difference between the perceived social status of the outgroup and that of Maharashtrians, both measured on the MacArthur Scale of Subjective Social Status (Goodman et al., 2001). Reciprocity is also compared against the common ingroup identity model (Gaertner, Mann, Murrell, & Dovidio, 1989). It claims that people discriminate less against those outgroups that share some superordinate identity with them. A superordinate identity is an identity held by the members of otherwise distinct subgroups. As most Gujaratis and Biharis share a common Hindu identity with Maharashtrians, the indicator is coded 1 for them and 0 for Muslims and Parsis. However, some Gujaratis and Biharis are not Hindus. Dropping the imperfect indicator of a superordinate identity from the analysis does not change any of the substantive results. Although the study does not ask which person discriminates against outgroups, but rather which outgroups the person’s discrimination targets, the model also includes individual-level correlates of discrimination, such as salience of the Maharashtrian identity, education, and age. Analysis Table 1 shows that the effect of reciprocity is remarkably stable across various regression models. One model specification (3) includes individual-level fixed effects to control for unmeasured individual-level confounders; model (2) does the same for confounders at the group level; and model (4) controls for both individual- and group-level fixed effects. As a robustness check, I also replicated the analysis using a proportional odds ordinal logistic regression with the Huber-White method to correct for heteroskedasticity and clustering. However, the ordinal logistic regression did not change any substantive conclusions drawn from the generalized linear model presented here. In all four models, receiving help from outgroup members in the past year lowers the willingness to discriminate against the said outgroup by 6 to 10 percentage points. This relationship remains statistically significant after controlling for intergroup contact and other alternative explanations. Interestingly, being taken advantage by an outgroup member in the past is not associated with more discrimination. One plausible explanation is that slum dwellers are simply used to being taken advantage by others. A vast majority of slum dwellers find temporary jobs in the informal sector with little security of contract enforcement. People negotiate the price of nearly everything, from vegetables to bribes. In an environment characterized by constant bargaining, it is easy to feel that one has been taken advantage of by the other side. Therefore, positive correlation between this variable and intergroup contact (r 5 0.31) should not come as a surprise. Even when asked about being taken advantage of by another Maharashtrian, 30% answered that it happens very often. Unmeasured individual attributes explain almost half of the variation in discriminatory attitudes, and only a small fraction of variance is explained by characteristics of outgroups themselves. This result indicates that many of the respondents simply preferred the ingroup over all four outgroups. Among group-level variables, relative group size is the strongest factor. All other things being equal, the number of Maharashtrians willing to discriminate against two largest minorities (Muslims and Gujaratis) is higher by 4 to 5 percentage points than the number of Maharashtrians discriminating

Reciprocity and Discrimination

415

Table 1. Generalized Linear Model Predicting Discrimination Model Positive Reciprocity Intergroup Contact Negative Reciprocity Economic Threat Safety Threat Relative Group Size Relative Group Status Shared Hindu Identity Identity Salience Education Age Hindu Magathane Shivaji Nagar Intercept Fixed Effects N Adjusted R2

(1)

(2)

(3)

(4)

20.103** (0.037) 20.038* (0.018) 20.003 (0.019) 20.010 (0.013) 0.000 (0.014) 0.269* (0.120) 0.000 (0.008) 20.026 (0.018) 20.040* (0.020) 20.006 (0.009) 20.000 (0.002) 0.146* (0.063) 20.133** (0.046) 0.009 (0.039) 0.791*** No 778 0.09

20.099** (0.035) 20.039* (0.016) 20.014 (0.017) 20.009 (0.013) 0.002 (0.014)

20.073* (0.028) 20.042** (0.013) 20.030. (0.017) 20.004 (0.009) 0.005 (0.010) 0.232* (0.092) 0.006 (0.006) 20.029. (0.016)

20.058* (0.026) 20.056*** (0.011) 20.019 (0.014) 20.006 (0.008) 20.006 (0.009)

20.037* (0.019) 20.011 (0.008) 20.001 (0.002) 0.170** (0.061) 20.064 (0.048) 0.070. (0.040) 0.698*** Group 838 0.09

0.977*** Individual 778 0.47

0.442*** Both 838 0.52

Note. Cluster robust standard errors (for the models [1] and [2]) and robust standard errors (for the models [3] and [4]) in parentheses. † p < .10, *p < .05, **p < .01, ***p < .001.

against the smaller minorities of Biharis and Parsis. Size of the ethnic minority matters in India as much as it does in Sub-Saharan Africa (Posner, 2004, 2005) and Western Europe (Quillian, 1995). However, there is no evidence that intergroup anxiety mediates the relationship between group size and discrimination. Self-assessed perceived threat is not a statistically significant predictor of discriminatory attitudes in any of the regression models. Moreover, it did not become significant even after the relative group size was dropped from the regression equation (results not reported here). These surprising results contradict an earlier study conducted among students in the Indian state of Orissa by Tausch et al. (2009). At first glance, this insignificance is surprising in a city in which major political parties routinely accuse minorities of stealing jobs from the Marathi-speaking “sons of the soil.” Even before Shiv Sena launched its campaign of intimidation to secure jobs for Maharashtrians, some industrial strikes escalated into ethnic violence after employers tried to recruit replacement workers from a caste or religion different from that of the striking employees (Noronha, 2005). However, Marathi slum dwellers in contemporary Mumbai know that their principal competitors on the job market are not Muslims, but rather other Maharashtrians. Sixty-two percent of the survey respondents strongly agreed with the statement: “Maharashtrians take jobs away from people like me.” Only 10% strongly agreed with the similar statement about Muslims. Insignificance of the perceived threat to safety is more counterintuitive. In a city generally considered India’s capital of organized crime, sensationalist media pay a lot of attention to the crimes committed by Muslims and almost automatically link them to Islamist terrorist networks (Shaban, 2010). Despite this dominant narrative, more respondents blamed Maharashtrians (76%) rather than Muslims (60%) for crime. This finding is reminiscent of Sniderman et al. (2004), which found that perceived threat to safety was the least important predictor of hostility towards immigrants in the Netherlands. Although reducing anxiety is considered one of the main mediators between intergroup contact and prejudice reduction (Pettigrew & Tropp, 2008), intergroup anxiety is less important in Mumbai than expected.

416

Tusicisny

Experiment Experimental Design The main goal of the experiment was to find out whether indirect positive reciprocity has a causal effect on discriminatory attitudes (Hypothesis 2). Psychology experiments are run typically in university laboratories with student convenience samples, and it is usually problematic to generalize results to any meaningful population (Henry, 2008; Sears, 1986). This experiment used a sample representative of a politically relevant population of mostly Hindu Marathi-speaking male slum dwellers, many of whom support extreme nationalist parties implicated in ethnic riots. The descriptive statistics for the sample can be found in Table S.2 in the online supporting information. Each subject learned that he would play a simple computer game with people in two other areas of Mumbai: Bhendi Bazar (a typical Muslim enclave) and Dadar (a well-known Hindu Marathi neighborhood). Although ethnic affiliation was never explicitly mentioned during the experiment, the computer screen showed a photograph, the first name, and the neighborhood of the other player—three unobtrusive cues of the partner’s membership in the Hindu ingroup or the Muslim outgroup. A failure to associate these cues with a correct ethnic group would make the results noisier, but not biased. Players participated in 10 rounds of a public goods game: first with five distinct ingroup members and then with five distinct outgroup members. Since players knew they would face the same partner only once and they could not see their partner’s previous record, any reciprocity was indirect and could not be explained by reputation building. I employed the public goods game to randomly assign cooperative experience with ingroup and outgroup members. Unlike in other experiments that use this behavioral game merely to measure cooperation, here, it served to manipulate the treatment variable in this experiment. In each round, the player received 10 Rupees and could decide whether to keep them or to invest them in a common project with the partner on the screen. The partner made the same decision simultaneously, without any communication. If they both invested, each of them received 20 Rupees (CC). If the player invested, but the partner kept his money, the player lost his investment (CD). If both the player and the partner kept their money, each got the payoff of 10 Rupees (DD). If the partner invested, but the player kept his money, the player’s payoff was 10 Rupees, while the partner’s payoff was 0 (DC). The best strategy depended on whether the player expected his partner to cooperate or not. The payoff structure of the game corresponds to the stag hunt game, which has been used to model public goods provision since David Hume. A discussion of this game choice is in the online supporting information. At the beginning of each session, a random number generator assigned the subject to one of four treatment groups. Unlike in many other laboratory experiments, subjects were not assigned to treatment in clusters (e.g., everyone in the same session assigned to the same treatment group). Subjects in the Generalized Reciprocity group always faced cooperators: first five Hindus, then five Muslims. Subjects in the Cooperative Hindus group played five rounds with cooperating Hindus and then five rounds with defecting Muslims. The Cooperative Muslims group was paired with uncooperative Hindus and cooperative Muslims. In the No Reciprocity group, all 10 partners defected. The manipulated frequency of cooperation created low or high expectations of cooperative behavior from the two groups. These four ideal types cover the whole spectrum of possible intergroup interactions. Consistent behavior within the group should also make the subtyping process less likely. The experiment involved deception. Subjects believed they were playing a game with real people, while their partners were in fact generated by a computer. The experiment aims to measure the causal effect of the treatment variable, not just to measure its preexisting levels and the resulting equilibrium behavior in the sample. If I merely observed evolving unrestricted behavior among real pairs of people instead of randomizing cooperative experience, I might not have been able to identify the average treatment effect (ATE) at all for the reasons explained in Green and Tusicisny (2012). Another reason

Reciprocity and Discrimination

417

Figure 1. Cooperation rate in the public goods game.

for deception was—paradoxically—ethical. After the debriefing, subjects knew that noncooperative behavior was generated by a computer, and it was not a real characteristic of the Muslim minority. Had they encountered real uncooperative Muslims during the experiment, the research itself could contribute to negative stereotyping and produce anger against Muslims. Deception was explained during the debriefing, and all subjects were given the maximum payout possible. None of the 192 participants expressed disbelief or doubts about the existence of other players during the experiment or when asked about it during the debriefing. In order to prevent deception from polluting the sample, every day we sampled participants from a new part of the neighborhood. Subjects filled in a questionnaire asking about their real-life discriminatory attitudes in the time between the experiment and the debriefing. The questions matched those from the survey described in the previous section. Although a within-subject design with a pretreatment survey and a posttreatment survey could establish a treatment effect against baseline attitudes, a pretreatment survey would also make the real purpose of the experiment obvious to the participants. That is why I chose a betweensubject design with a posttreatment survey. The online supplement discusses why spillover effects, attrition, social desirability bias, and other factors are unlikely to bias the treatment effect on the discriminatory attitudes measured by the posttreatment survey. Analysis Figure 1 indicates that subjects understood the logic of the public goods game and updated their expectations of cooperation based on experience in previous rounds. In the first round, about 77% of participants invested money in a common project with a stranger on a computer screen. In comparison, the initial cooperation rate in Western laboratories is typically between 40 and 60% (Ostrom, 2000). That said, most public goods experiments in Western laboratories follow the prisoner’s dilemma payoff structure, where the best strategy is to defect 100% of the time. The game used here had a mixed strategy equilibrium of cooperating in 50% rounds. By the fifth round, the cooperation rate in the two groups that encountered Hindu defectors (Cooperative Muslims and No Reciprocity) dropped by half, while the two other groups retained a high cooperation rate. When ethnicity of the partner changed suddenly in the middle of the experiment from the last Hindu partner in the fifth round to the first Muslim partner in the sixth round, the resulting drop in the

418

Tusicisny

cooperation rate was rather small: 6–10 percentage points. In the No Reciprocity treatment group, the cooperation rate in fact increased by 4 percentage points. None of these differences was statistically significant. Clearly, cooperation across ethnic lines in the slums of Mumbai is possible. Willingness to invest money in a transaction with a Muslim during the game reflects the level of market integration in Mumbai. Autorickshaws, the ubiquitous three-wheelers of South Asian roads, offer an illustration of the existing Hindu-Muslim economic ties. The mostly Marathi owners rent their autorickshaws out for exorbitant amounts of money to Muslims and immigrants from North India, who actually drive the vehicles. The same Marathi owners typically support the political parties promoting hatred of Muslims (Shiv Sena) and of North Indian immigrants (MNS). Although a principally exploitative relationship between owners and drivers does not necessarily improve intergroup relations, familiarity with intergroup economic ties may explain the lack of ethnic bias in the laboratory game. One-third of the Cooperative Hindus and No Reciprocity groups kept contributing despite their partners’ defections in the second half of the experiment. Unlike conditional cooperators in the rest of the sample, these participants were probably motivated by altruism or confusion (Andreoni, 1995) instead of reciprocity. The cooperation rate in the Cooperative Muslims group eventually more than doubled by the tenth round (from 33 to 69%). However, it never converged at the same high level of cooperation as the Generalized Reciprocity group (88%). Conditioned by past behavior, ethnically heterogeneous pairs produced as much public goods as ethnically homogenous pairs in the public goods game. Regardless of ethnicity, participants cooperated with those whom they trusted to reciprocate. But did a positive interaction with Muslims change attitudes towards Muslims as a social category? The hypothesis is that indirect positive reciprocity changes discriminatory attitudes towards Muslims as a group. If this is true, we should observe less discrimination against Muslims in the treatment groups that interacted with cooperative Muslim players. We can identify the ATE by comparing the treatment groups that shared the same history of play with ingroup members and differed in nothing but the behavior of outgroup members: (1) Generalized Reciprocity versus Cooperative Hindus and (2) Cooperative Muslims versus No Reciprocity. Let us look first at the survey question whether the respondent would accept a Muslim as a neighbor (top two graphs in Figure 2). The proportion of people who would not accept a Muslim neighbor dropped from 54% in the No Reciprocity condition to 41% in the Cooperative Muslims group. In the predicted direction, the treatment effect was not statistically significant in this case (p-value 5 0.13).1 The difference of 24 percentage points between the Generalized Reciprocity and Cooperative Hindus groups was, however, significant (p-value 5 0.01). As these two groups differed only in whether Muslim partners cooperated, the large effect can be attributed to the treatment and not some other difference between the two groups. The effect of reciprocity remains negative, strong, and statistically significant even after controlling for covariates (models 2 and 4 in Table 2). Although experimenters rarely ask subjects about their motivations, I was interested in understanding why the subjects answered the survey questions as they did. Eight interviews with randomly selected subjects support the story that arose from the quantitative analysis. Only one respondent did not mind Muslims as neighbors. His explanation illustrates the causal argument proposed in this study: “We allow the Muslim people to be our neighbors because they are very honest and help us all the time in all the situations.” In other words, the subject was motivated by positive reciprocity. Other 1

The p-values in all comparisons of means across treatment groups in this study are obtained using randomization inference under the sharp null hypothesis of no difference between the two groups. Unlike a t-test or other parametric tests, randomization inference does not assume normality and independence. The only assumption is that assignment of treatment is random—which it is. Fisher (1935) developed randomization inference as the ideal nonparametric test for experimental data. An important advantage is that randomization inference will work with any sample size and any scale of the outcome variable (Gerber & Green, 2012). As the theoretical argument predicts the direction of the effects, p-values are one-tailed.

Reciprocity and Discrimination

419

Figure 2. Avoidance of Muslims by treatment group, percentage with 95% confidence intervals.

respondents did not desire Muslim neighbors. According to one respondent, Muslims “always quarrel with their neighbors.” Another said: “I do not accept the Muslim people as my neighbors because it is very difficult to survive beside them in the time of riots.” The interviews show clearly that participants had their real living conditions in slums in mind while answering this survey question. This increases confidence that the dependent variable is valid. Surprisingly, the question whether the participants would “accept a Muslim to close kinship by marriage” produced a similar pattern. Usually arranged by parents, most marriages in India occur not only within the same religion, but also within the same caste. The survey conducted in parallel to the experiment showed a strong preference for ethnic endogamy in the population: 170 out of 210 respondents (81%) said they would not accept a Muslim marrying into their family, while 209 out of 210 respondents would accept a Maharashtrian. Given the importance of ethnicity in the selection of a marriage partner, I did not expect any effect of the experimental treatment on attitudes concerning intermarriage. Nevertheless, as Figure 2 demonstrates, the experimental treatments reduced the number of people who would not accept a

0.214 95

21.002 * (0.429)

(1)

0.409 0.439 20.641 14.815 20.080 0.509 0.312 21.314 213.548

(0.612) (0.588) (0.501) *** (0.890) (0.319) (0.553) (0.678) * (0.592) *** 91

21.427 ** (0.538)

(2)

0.167 97

20.539 (0.411)

(3)

21.064 * (0.523) 1.493 * (0.647) 0.668 (0.662) 0.839 (0.527) 20.391 (1.259) 0.355 (0.267) 20.637 (0.489) 0.175 (0.654) 20.155 (0.519) 23.463 92

(4)

Cooperative Muslims vs. No Reciprocity

Note. Design-based robust standard errors in parentheses. Two-tailed test. † p < .10, *p < .05, **p < .01, ***p < .001.

Cooperative Muslims Magathane Shivaji Nagar Tap Water Marathi Language Log Income Extremist Voter Dadar Bhendi Bazar Intercept N

Generalized Reciprocity

Model

Generalized Reciprocity vs. Cooperative Hindus

Would Not Accept a Muslim Neighbor

Table 2. Logistic Regression Predicting Discriminatory Attitudes

1.186 *** 95

21.186 ** (0.448)

(5)

20.145 (0.714) 20.331 (0.720) 20.541 (0.524) 0.355 (0.998) 0.048 (0.377) 20.095 (0.550) 21.524 (0.937) 21.200 . (0.618) 2.795 91

21.589 ** (0.578)

(6)

Generalized Reciprocity vs. Cooperative Hindus

0.990 ** 97

20.703 (0.436)

(7)

20.522 (0.523) 1.144 . (0.676) 1.348 . (0.731) 20.669 (0.564) 1.000 (1.131) 0.412 (0.277) 20.010 (0.534) 0.751 (0.744) 20.672 (0.580) 24.653 92

(8)

Cooperative Muslims vs. No Reciprocity

Would Not Accept Marriage with a Muslim

420 Tusicisny

Reciprocity and Discrimination

421

Muslim to marry into their family by 27 (from Cooperative Hindus to Generalized Reciprocity) and 16 (from No Reciprocity to Cooperative Muslims) percentage points. The p-values are 0.01 and 0.08, respectively. Table 2 shows that the negative sign of the coefficients does not change after covariate adjustment. To put these numbers in perspective, Dugar, Bhattacharya, and Reiley (2012) observed an effect of a similar size on the number of inquiries from Middle Caste families to matrimonial advertisements posted by fake, potential Low-Caste grooms when the experimenters manipulated the groom’s advertised income. There is a good reason to believe that the treatment effect is so strong because it was measured immediately after the experiment. Some of it should probably be attributed to affect, priming, or other fleeting factors. An attempt to measure how quickly the effect decays in time would raise ethical concerns about artificially induced negative stereotypes influencing subjects outside the laboratory. One can imagine a follow-up study, in which measurement of hostile behavioral intentions—or of behavior itself—occurs in various randomized intervals ranging from hours to months after the experiment and the experiment itself only involves a positive intervention. Some persistence of the treatment effect in time is very likely: Olson and Fazio (2006) detected a reduction in racial prejudice two days after they induced it in a laboratory; Malhotra and Liyanage (2005) were able to measure an effect of a four-day workshop on attitudes and behavior towards an outgroup one year after the workshop. Alternative Comparisons By necessity, the treatment groups used for pairwise comparisons differed not only in whether Muslim partners cooperated, but also in the total number of cooperative partners. For example, subjects in the Generalized Reciprocity group faced five more cooperators overall than subjects in the Cooperative Hindus group. Experiencing cooperation regardless of ethnicity may have led to less discrimination due to improved mood, satisfaction with a higher payout, or some other reason. In order to disentangle the effect of experiencing cooperation with a Muslim from the effect of experiencing cooperation in general, I compared the Cooperative Hindus and Cooperative Muslims groups. Despite the same number of cooperative partners and similar average earnings during the game (Rs. 130 and Rs. 115 respectively), discrimination against Muslims in the Cooperative Muslims group was lower by 20 percentage points (p 5 0.04) for intermarriage and by 15 percentage points for the acceptance of neighbor (p 5 0.11). Other comparisons offer additional evidence that interacting with cooperative Muslims affected discrimination against Muslims above and beyond the effect of interacting with cooperative partners in general. The subjects in the Generalized Reciprocity group faced five more (Hindu) cooperators than the subjects in the Cooperative Muslims group. However, discrimination against Muslims was not significantly higher in the Generalized Reciprocity group. Similarly, for people, who have faced Muslim defectors (No Reciprocity and Cooperative Hindus), the effect of interacting with five Hindu cooperators (Cooperative Hindus) on discrimination of Muslims was close to zero. The treatment in this experiment affected discriminatory attitudes as predicted across a variety of specifications (comparisons of means, regressions, two different measures of social distance). Subjects generalized their experience with Muslim partners to the ethnic group as a whole. The treatment effect is specific to experiencing cooperation with Muslims and cannot be attributed to experiencing cooperation in general. All four treatments satisfied Allport’s (1954) optimal conditions of equal status, common goals, cooperative interdependence, and institutional support. However, intergroup contact did not automatically lead to lower discrimination in all treatment groups. What mattered was whether outgroup members were cooperative. Contradicting much of the contact hypothesis literature, the experiment also shows that even a very short, superficial, and involuntary interaction can improve antagonistic relations between groups if it allows group members to display cooperative behavior. Allport merely

422

Tusicisny

hinted at the possible causal mechanism behind the contact hypothesis: “Contacts that bring knowledge and acquaintance are likely to engender sounder beliefs concerning minority groups, and for this reason contribute to the reduction of prejudice” (p. 268). Since then, other scholars have focused mostly on two causal channels: reducing anxiety and increasing empathy (Pettigrew & Tropp, 2008). Increasing low expectations of cooperative behavior may be another mechanism through which intergroup contact reduces discrimination. Table S.3 in the online supporting information shows that the signs of coefficients do not change if we exclude self-declared Buddhists from the analysis. As Table S.4 shows, voters of extreme nationalist parties Shiv Sena and the MNS responded to a cooperative experience in the same way as moderates. Neither was there any sign of a stronger (or weaker) treatment effect among those who had been in less frequent contact with Muslims (Table S.5). Absence of a heterogeneous effect across these subgroups suggests that it is possible to update even strong stereotypes. The fact that supporters of extremist political parties responded to the treatment equally well as the moderates also has encouraging policy implications.

Conclusion The survey showed that indirect reciprocity adds more explanatory power to models of selective discrimination based on relative group size. Therefore, group stereotypes about cooperative behavior combined with the lack of direct positive experience can help explain discrimination of ethnic groups that are too small, weak, or politically disenfranchised to be considered serious contenders by the dominant group. Examples of such groups include not only Muslims and Christians in many parts of India, but also Roma and immigrants in Europe. The experiment showed that experiencing cooperation in a behavioral game reduced discriminatory intentions in completely unrelated domains: ethnic segregation of housing and intermarriage. This finding indicates that the effect of reciprocity spills over not only from the individual to the group, but also across situations. The treatment produced a robust causal effect in the predicted direction across different specifications of the model. Although the experimental intervention reduced the preexisting ethnic bias substantially, it did not erase it. Even in the Generalized Reciprocity group, acceptance of Muslims as potential neighbors (69%) was still much lower than acceptance of fellow Maharashtrians (98%). The two methodological approaches complement each other. The experiment allows us to identify a causal relationship between indirect positive reciprocity and discriminatory attitudes without an endogeneity bias, but its ecological validity is limited. The survey identifies the same association between real-life reciprocity and discrimination in relations to a larger number of ethnic minorities. Of course, it is yet to be seen whether the results from Indian slums can be generalized to other, more peaceful and economically more developed environments. The optimistic conclusion of this article naturally leads to the question of why we see so much ingroup bias in real life. One possible explanation is consistent with the observation that research participants updated their beliefs quickly: The effect of indirect positive reciprocity may be short-lived, soon to be offset by news articles, conversations with more prejudiced ingroup members, and other experiences. Another explanation is that people may have too few opportunities for intergroup cooperation. If social sanctioning facilitates cooperation with ingroup members (Fearon & Laitin, 1996; Habyarimana, Humphreys, Posner, & Weinstein, 2007, 2009; Yamagishi & Mifune, 2008), it will cause more positive reciprocity within the ingroup in the long run. This tendency will lead to a situation similar to the Cooperative Hindus condition. A higher cooperation rate in a well-bounded group due to a more

Reciprocity and Discrimination

423

effective enforcement increases expectations of future cooperation, which convinces more members to cooperate, which leads to more positive reciprocity. Unfortunately, vicious cycles of conflict are as likely as virtuous cycles of cooperation. For example, a dearth of prior intergroup contact may increase intergroup anxiety, which leads to more avoidance and stereotyping (Stephan & Stephan, 1985). Varshney (2002) provided an example of such virtuous and vicious cycles: 90% of Hindu and Muslim families in the peaceful city of Calicut, but only 42% in conflict-prone Aligarh, reported that their children played together. Aligarhs of our world may never improve without introducing more opportunities for interethnic cooperation. Once reciprocal cooperation becomes institutionalized, however, it can protect the community from ethnic violence for centuries—as Jha (2013) suggested takes place in the case of Hindu-Muslim relations in former medieval ports in India. There are not many known ways to overcome ethnocentrism. Social psychologists admit that their usual prescriptions to manipulate social identities are rarely applicable outside scientific laboratories (M. B. Brewer, 1997). Positive intergroup contact seems to work better, but its four optimal conditions are impractically restrictive. This study suggests that we should focus on one particular aspect of intergroup contact, which is indirect positive reciprocity. If we build institutions encouraging positive reciprocity between ethnic groups, we may be able to reduce discrimination. Although concrete policy recommendations would require more empirical evidence, expectations of cooperative behavior can be raised through better dissemination of information across group boundaries, socialization in the norms of fairness, and credible enforcement of the rule of law. Focus on these areas may help prevent conflict and increase intergroup cooperation in multiethnic societies. ACKNOWLEDGMENTS In different stages, this research benefited greatly from critical comments and suggestions made by Donald Green, Jack Snyder, Macartan Humphreys, Neelanjan Sircar, John Huber, Alexandra Scacco, Kanchan Chandra, Thomas Blom Hansen, Michael J. Donnelly, as well as participants in the South Asia Workshop in New Delhi and the Sixth Annual NYU-CESS Conference on Experimental Political Science, and seminar participants at Columbia University, the European University Institute, and the Carlos III-Juan March Institute. I would also like to thank R. N. Sharma for his guidance during my field research in India and to my two tireless research assistants, Aditya Padwal and Priyesh Gohil, for being willing to work with me 12 hours a day. The project was supported by a Columbia University GSAS International Traveling Fellowship. Correspondence concerning this article should be addressed to Andrej Tusicisny, Google Inc., 1600 Amphitheatre Parkway, Mountain View, CA 94043. E-mail: [email protected] REFERENCES Allport, G. W. (1954). The nature of prejudice. Cambridge, UK: Addison-Wesley. Andreoni, J. (1995). Cooperation in public-goods experiments: Kindness or confusion? American Economic Review, 85(4), 891–904. Bassili, J. N. (2008). Attitude strength. In W. D. Crano & R. Prislin (Eds.), Attitudes and attitude change (pp. 237–260). New York, NY: Psychology Press. Blalock, H. M. (1967). Toward a theory of minority-group relations. New York, NY: Wiley. Bogardus, E. S. (1925). Measuring social distance. Sociology and Social Research, 9, 299–308. Bowles, S., & Gintis, H. (2011). A cooperative species: Human reciprocity and its evolution. Princeton, NJ: Princeton University Press. Brewer, M. B. (1997). The social psychology of intergroup relations: Can research inform practice? Journal of Social Issues, 53(1), 197–211.

424

Tusicisny

Brewer, P. R. (2003). The shifting foundations of public opinion about gay rights. Journal of Politics, 65(4), 1208–1220. Brown, R. (2000). Social identity theory: Past achievements, current problems and future challenges. European Journal of Social Psychology, 30(6), 745–778. Chaudhuri, A. (2011). Sustaining cooperation in laboratory public goods experiments: A selective survey of the literature. Experimental Economics, 14(1), 47–83. Coleman, J. S. (1988). Social capital in the creation of human capital. American journal of sociology, 94(1), 95–120. Duckitt, J., & Mphuthing, T. (1998). Group identification and intergroup attitudes: A longitudinal analysis in South Africa. Journal of Personality and Social Psychology, 74(1), 80–85. Dugar, S., Bhattacharya, H., & Reiley, D. (2012). Can’t buy me love? A field experiment exploring the trade-off between income and caste—status in an Indian matrimonial market. Economic Inquiry, 50(2), 534–550. Engineer, A. A. (1993). Bombay riots: Second phase. Economic and Political Weekly, 28(12/13), 505–508. Fearon, J. D., & Laitin, D. D. (1996). Explaining interethnic cooperation. American Political Science Review, 90(4), 715–733. Fehr, E., & G€achter, S. (2000). Cooperation and punishment in public goods experiments. American Economic Review, 90(4), 980–994. Fershtman, C., & Gneezy, U. (2001). Discrimination in a segmented society: An experimental approach. Quarterly Journal of Economics, 116(1), 351–377. Fischbacher, U., G€achter, S., & Fehr, E. (2001). Are people conditionally cooperative? Evidence from a public goods experiment. Economics Letters, 71(3), 397–404. Fisher, R. A. (1935). The design of experiments. Edinburgh, Scotland: Oliver and Boyd. Fiske, S. T. (2000). Stereotyping, prejudice, and discrimination at the seam between the centuries: Evolution, culture, mind, and brain. European Journal of Social Psychology, 30(3), 299–322. Gaertner, S. L., Mann, J., Murrell, A., & Dovidio, J. F. (1989). Reducing intergroup bias: The benefits of recategorization. Journal of Personality and Social Psychology, 57, 239–249. Gambetta, D., & Hamill, H. (2005). Streetwise: How taxi drivers establish their customers’ trustworthiness. New York, NY: Russell Sage. Gerber, A. S., & Green, D. P. (2012). Field experiments: Design, analysis, and interpretation. New York, NY: W.W. Norton. Goodman, E., Adler, N. E., Kawachi, I., Frazier, A. L., Huang, B., & Colditz, G. A. (2001). Adolescents’ perceptions of social status: Development and evaluation of a new indicator. Pediatrics, 108(2), e31–e31. Gouldner, A. W. (1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25(2), 161–178. Green, D. P., & Seher, R. L. (2003). What role does prejudice play in ethnic conflict? Annual Review of Political Science, 6(1), 509–531. Green, D. P., & Tusicisny, A. (2012). Statistical analysis of results from laboratory studies in experimental economics: A critique of current practice. Paper presented at the 2012 North American Economic Science Association Conference, Tucson, AZ. Habyarimana, J., Humphreys, M., Posner, D. N., & Weinstein, J. M. (2007). Why does ethnic diversity undermine public goods provision? American Political Science Review, 101(4), 709–725. doi: 10.1017.S0003055407070499 Habyarimana, J., Humphreys, M., Posner, D. N., & Weinstein, J. M. (2009). Coethnicity: Diversity and the dilemmas of collective action. New York, NY: Russell Sage. Hansen, T. B. (2001). Wages of violence: Naming and identity in postcolonial Bombay. Princeton, NJ: Princeton University Press. Hardin, R. (1995). One for all: The logic of group conflict. Princeton, NJ: Princeton University Press. Henry, P. J. (2008). College sophomores in the laboratory redux: Influences of a narrow data base on social psychology’s view of the nature of prejudice. Psychological Inquiry, 19(2), 49–71. Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup bias. Annual Review of Psychology, 53(1), 575–604. Horowitz, D. L. (1985). Ethnic groups in conflict. Berkeley: University of California Press. Horowitz, D. L. (2001). The deadly ethnic riot. Berkeley: University of California Press. Hugh-Jones, D., & Leroch, M. A. (2013). Intergroup revenge: A laboratory experiment on the causes. Retrieved from SSRN 2275173. Islam, M. R., & Hewstone, M. (1993). Dimensions of contact as predictors of intergroup anxiety, perceived out-group variability, and out-group attitude: An integrative model. Personality and Social Psychology Bulletin, 19(6), 700– 710. Jha, S. (2013). Trade, institutions and religious tolerance: Evidence from South Asia. American Political Science Review, 107(4), 806–832.

Reciprocity and Discrimination

425

Lawler, E. J. (2001). An affect theory of social exchange. American Journal of Sociology, 107(2), 321–352. Malhotra, D., & Liyanage, S. (2005). Long-term effects of peace workshops in protracted conflicts. Journal of Conflict Resolution, 49(6), 908–924. Noronha, E. (2005). Ethnicity in industrial organizations: Case of two organizations in Mumbai. Jaipur, India: Rawat. Nowak, M. A., & Sigmund, K. (2000). Shrewd investments. Science, 288(5467), 819–820. Olson, M. A., & Fazio, R. H. (2006). Reducing automatically activated racial prejudice through implicit evaluative conditioning. Personality and Social Psychology Bulletin, 32(4), 421–433. Ostrom, E. (2000). Collective action and the evolution of social norms. Journal of Economic Perspectives, 14(3), 137– 158 Palshikar, S., & Deshpande, R. (1999). Electoral competition and structures of domination in Maharashtra. Economic and Political Weekly, 34(34/35), 2409–2422. Paluck, E. L., & Green, D. P. (2009). Prejudice reduction: What works? A review and assessment of research and practice. Annual Review of Psychology, 60, 339–367. Peffley, M., Hurwitz, J., & Sniderman, P. M. (1997). Racial stereotypes and Whites’ political views of Blacks in the context of welfare and crime. American Journal of Political Science, 41(1), 30–60. Petersen, R. D. (2002). Understanding ethnic violence: Fear, hatred, and resentment in twentieth-century Eastern Europe. Cambridge, UK: Cambridge University Press. Pettigrew, T. F. (1971). Racially separate or together? New York, NY: McGraw-Hill. Pettigrew, T. F. (2008). Future directions for intergroup contact theory and research. International Journal of Intercultural Relations, 32(3), 187–199. Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90(5), 751–783. Pettigrew, T. F., & Tropp, L. R. (2008). How does intergroup contact reduce prejudice? Meta-analytic tests of three mediators. European Journal of Social Psychology, 38(6), 922–934. Pettigrew, T. F., Tropp, L. R., Wagner, U., & Christ, O. (2011). Recent advances in intergroup contact theory. International Journal of Intercultural Relations, 35(3), 271–280. Phan, M. B., Blumer, N., & Demaiter, E. I. (2009). Helping hands: Neighborhood diversity, deprivation, and reciprocity of support in non-kin networks. Journal of Social and Personal Relationships, 26(6–7), 899–918. Pinker, S. (2011). The better angels of our nature: Why violence has declined. New York, NY: Viking. Posner, D. N. (2004). The political salience of cultural difference: Why Chewas and Tumbukas are allies in Zambia and adversaries in Malawi. American Political Science Review, 98(4), 529–545. Posner, D. N. (2005). Institutions and ethnic politics in Africa. New York, NY: Cambridge University Press. Quillian, L. (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. Rothbart, M. (1981). Memory processes and social beliefs. In D. L. Hamilton (Ed.), Cognitive processes in stereotyping and intergroup behavior (pp. 145–181). Hillsdale, NJ: Lawrence Erlbaum. Sanfey, A. G., Rilling, J. K., Aronson, J. A., Nystrom, L. E., & Cohen, J. D. (2003). The neural basis of economic decision-making in the ultimatum game. Science, 300(5626), 1755–1758. Sears, D. O. (1986). College sophomores in the laboratory: Influences of a narrow data base on social psychology’s view of human nature. Journal of Personality and Social Psychology, 51(3), 515–530. Sears, D. O., & Funk, C. L. (1999). Evidence of the long-term persistence of adults’ political predispositions. Journal of Politics, 61(1), 1–28. Shaban, A. (2010). Mumbai: Political economy of crime and space. New Delhi, India: Orient BlackSwan. Sherif, M., Harvey, O. J., White, B. J., Hood, W. R., & Sherif, C. W. (1961). Intergroup conflict and cooperation: The robbers cave experiment. Norman: University of Oklahoma. Sidanius, J. (1993). The psychology of group conflict and the dynamics of oppression. In S. Iyengar & W. J. McGuire (Eds.), Explorations in political psychology (pp. 183–219). Durham, NC: Duke University Press. Sniderman, P. M., Hagendoorn, L., & Prior, M. (2004). Predisposing factors and situational triggers: Exclusionary reactions to immigrant minorities. American Political Science Review, 98(01), 35–49. Srikrishna, B. N. (1998). Report of the Srikrishna Commission appointed for inquiry into the riots at Mumbai during December 1992 and January 1993. Government of Maharashtra. Stephan, W. G., Boniecki, K. A., Ybarra, O., Bettencourt, A., Ervin, K. S., Jackson, L. A., . . . Renfro, C. L. (2002). The role of threats in the racial attitudes of Blacks and Whites. Personality and Social Psychology Bulletin, 28(9), 1242–1254. Stephan, W. G., & Stephan, C. W. (1985). Intergroup anxiety. Journal of Social Issues, 41(3), 157–175.

426

Tusicisny

Stephan, W. G., & Stephan, C. W. (2000). An integrated threat theory of prejudice. In S. Oskamp (Ed.), Reducing prejudice and discrimination (pp. 23–45). Mahwah, NJ: Lawrence Erlbaum. Tausch, N., Hewstone, M., & Roy, R. (2009). The relationships between contact, status and prejudice: An integrated threat theory analysis of Hindu–Muslim relations in India. Journal of Community & Applied Social Psychology, 19, 83–94. Taylor, S. E. (1981). A categorization approach to stereotyping. In D. L. Hamilton (Ed.), Cognitive processes in stereotyping and intergroup behavior (pp. 145–182). Hillsdale, NJ: Lawrence Erlbaum. Varshney, A. (2002). Ethnic conflict and civic life: Hindus and Muslims in India. New Haven, CT: Yale University Press. Weber, R., & Crocker, J. (1983). Cognitive processes in the revision of stereotypic beliefs. Journal of Personality and Social Psychology, 45(5), 961–977. World Values Survey. (Multiple Years). Official data files. World Values Survey Association. Retrieved from www. worldvaluessurvey.org Yamagishi, T., & Mifune, N. (2008). Does shared group membership promote altruism? Rationality and Society, 20(1), 5–30. Yzerbyt, V., & Demoulin, S. (2010). Intergroup relations. In S. T. Fiske, D. T. Gilbert, & G. Lindzey (Eds.), Handbook of social psychology (5th ed., Vol. 2, pp. 1024–1083). New York, NY: Wiley.

Supporting Information Additional supporting information may be found in the online version of this article at the publisher’s website: Table S.1: Descriptive Statistics for the Survey Data Table S.2: Observable Characteristics of the Treatment Groups Table S.3: Replication for only Hindus Table S.4: Test of Heterogeneous Treatment Effects, Extremist Voting Table S.5: Test of Heterogeneous Treatment Effects, Frequency of Contact Discussion of the Stag Hunt Game Used in the Experiment Potential Biases in the Experiment Table S.6: Multinomial Logistic Regression Predicting Treatment Assignment Questionnaire Used in India