INTRODUCTION
An enduring question for the social sciences is whether increasing contact and exposure between in-groups and out-groups enhances prospects for social tolerance and cooperation (Allport Reference Allport1954; Pettigrew and Tropp Reference Pettigrew and Tropp2006, Reference Pettigrew and Tropp2008 (meta-analyses); Williams Reference Williams1947). Although the contact hypothesis has been examined extensively across a range of observational and experimental studies with generally positive results, considerations of in-group/out-group exposure after violence are largely unexplored.Footnote 1 In addition, a well-established literature on group conflict and inter-group threat raises questions about adverse effects of contact in a potentially hostile environment with scarce resources (Blalock Reference Blalock1967; Bobo and Hutchings Reference Bobo and Hutchings1996; LeVine and Campbell Reference LeVine and Campbell1972; Riek et al. Reference Riek, Mania and Gaertner2006 (meta-analysis); Tajfel and Turner Reference Tajfel, Turner, Austin and Worchel1977). Drawing on these gaps and concerns in the literature, our research employs multi-locational experiments in post-war Kosovo to examine how different degrees of contact and separation affect in-group and out-group behavioral norms and preferences in the aftermath of violence.
Why study inter-group contact and norms after violence? Our research draws motivation from an ongoing debate in the conflict literature over whether groups can overcome bitter rivalries and legacies of violence to co-exist with one another or whether the partitioning of groups is the only viable solution to resolve problems of protracted violence. A major point of contention is whether separating groups is a more effective approach to building inter-group cooperation than forcing everyone into varying degrees of institutional and social integration. Some suggest that individual and group norms can recalibrate to changes in institutional and environmental contexts, and favor integrative approaches to promote lasting peace (e.g., Etzioni Reference Etzioni1992; Habyarimana et al. Reference Habyarima, Humphreys, Posner and Weinstein2008; Horowitz Reference Horowitz1985; Kumar Reference Kumar1997; Lijphart 1977; Sambanis Reference Sambanis2000; Sambanis and Schulhofer-Wohl Reference Sambanis and Schulhofer-Wohl2009). Others disagree on the malleability of norms and often advocate physical separation or partition as a viable solution to intractable group conflict (Kaufman Reference Kaufmann1996, Reference Kaufmann1998; Mearsheimer and Pape Reference Mearsheimer and Pape1993; Muller 2008; Posen Reference Posen1993). Our research considers the adaptability of norms to inter-group contact after violence using behavioral experiments.
Why use experiments to study norms? Behavioral experiments have been shown to be a useful tool for evaluating norms and preferences in many social contexts (Camerer Reference Camerer2003). Experiments may also offer advances over attitudinal methods by reducing social desirability and other cognitive/social demand effects on sensitive issues (Zizzo Reference Zizzo2010). However, scholars have only begun to apply behavioral experiments to the study of norms after violence–-often with counterintuitive results. For example, Voors et al. (Reference Voors, Nillesen, Verwimp, Bulte, Lensink and Van Soest2012) and Gilligan et al. (Reference Gilligan, Pasquale and Samii2013) find strong evidence of heightened individual pro-sociality and community-level solidarity following exposure to violence in Burundi and Nepal respectively. Alexander and Christia (Reference Alexander and Christia2011) offer evidence of how inter-group cooperation is enhanced by integrated environments and undermined by segregation in the case of post-war Bosnia. We also observed strong preferences for fairness across ethnicity in Bosnia, suggesting that social norms can rebound from episodes of conflict (Whitt Reference Whitt2014; Whitt and Wilson Reference Whitt and Wilson2007). These results are encouraging, but more needs to be done to identify effects of inter-group exposure on behavior after violence. With so few existing studies, we do not know the extent to which received wisdom to date is idiosyncratic to specific cases like Bosnia. Our research in Kosovo explores whether recent observations of pro-sociality are generalizable across post-conflict societies or conditional to ecological factors such as contact and exposure to former rivals.
RESEARCH DESIGN
To explore the effects of contact on out-group bridging norms, we test the hypothesis (H1) that increasing proximity between groups enhances pro-social behavior toward out-groups. To explore the impact of contact on in-group bonding, we test a second hypothesis (H2) that increasing proximity between groups enhances pro-social behavior within the group.
To evaluate pro-social behavior, we focus on the expression of a behavioral norm of altruism toward in-groups and out-groups. Altruism is just one of many types of behavioral norms but is believed to play a critical role in sustaining group cohesion (Bowles Reference Bowles2008; Choi and Bowles Reference Choi and Bowles2007; Fehr and Fischbacher Reference Fehr and Fischbacher2003). Moreover, in a recent meta-analysis, Pettrigew and Tropp (Reference Pettigrew and Tropp2008) find that one of the most important mechanisms through which contact can reduce prejudice is by building empathy toward out-groups. We utilize the concept of altruism as a behavioral measure of empathy for others.
We illicit preferences for altruism using a variation on the dictator experiment. The “dictator game” is a well-established instrument for evaluating social norms and preferences for altruism, charity, fairness, other regarding preferences, tastes for discrimination, and empathy (see Engel Reference Engel2010 for meta-analysis; Henrich et al. Reference Henrich, Boyd, Bowles, Camerer, Fehr, Gintis and McElreath2001). In a typical dictator experiment, subjects must decide how much of a sum of money to keep for themselves and how much to allocate to an anonymous recipient. We develop a variation on the dictator game to gauge how norms and preferences for self-interest and empathy toward others are affected by varying the in-group/out-group identity of the anonymous recipient in a within-subject design.
We conduct our research among ethnic Serbs in post-war Kosovo. Tensions between Serbs and Albanians in Kosovo have been well documented (Judah Reference Judah2002; King and Mason Reference King and Mason2006; Petersen Reference Petersen2011). Before 1989, Kosovo was a semi-autonomous province of Yugoslavia where rivalries between Albanians and Serbs were exploited if not stoked by Slobodan Milošević in his rise to power. Throughout the 1990s, Albanians increasingly challenged Milošević's authority in the region, culminating in an armed insurgency, crackdowns by Serbian forces, a massive refugee crisis, and ultimately NATO intervention in 1999. Although NATO provided a security umbrella to returning Albanian refugees, it faced major challenges protecting minority ethnic Serbs, many of whom remained clustered in enclaves, isolated villages, and mixed communities across Kosovo, from retaliatory attacks by Albanians.
Our study aims to exploit existing variation in the proximity of ethnic Serbs to Albanians in Kosovo to test hypotheses about effects of contact with former rivals and adversaries on basic social norms. Ethnic Serbs are largely concentrated in the north of Kosovo's territory along the Serbian border. The Ibar river and the divided city of Mitrovica serve as a symbolic ethnic borderline between majority Albanian and majority Serb populations (Figure A1 in Supplementary Appendix). There are virtually no Albanian settlements north of the Ibar, but substantial Serb communities remain inside the predominately Albanian-populated territory in the south. The current de facto autonomy of the northern Serbian enclaves, the incorporation of South Kosovo Serbs into a new, Albanian-dominated Kosovo state, and the tense status of the border territory offer a unique opportunity to examine how relative degrees of inclusion and separation, proximity and contact can affect group norms. We ask to what extent Kosovo Serbs with different levels of exposure to Albanians display heterogeneous in-group and out-group preferences as measured by a series of dictator games?
Finally, we acknowledge that our research design is more a measurement study than a true experiment in the framework of the Neyman–Rubin Causal Model (Neyman Reference Neyman1923; Rubin Reference Rubin1974). We use an experiment to measure a dependent variable of interest rather than to establish a causal relationship. Our independent variable of interest is not randomly assigned. We could face identification problems due to potential sorting and selection bias across our experimental “treatments”—in this case, different regions of Kosovo with different levels of ethnic contact and separation. However, we will attempt to show that selection into or out of a given region is primarily a function of displacement by violence (i.e., their homes were destroyed, severely damaged, or occupied by others during the fighting). In an environment of ethnic cleansing, everyone is a potential target, and internally displaced persons (IDPs) are not observably different in attitudes or behavior from non-IDPs in our data. This helps reassure us that our results are not driven by major sorting and selection effects resulting from inter-regional population transfers.
Despite these challenges of causal inference, the use of lab-in-the-field experiments advances on existing observational studies by measuring behavioral effects of contact after conflict in a multi-locational setting. To increase confidence in our results, we employ extended controls and matching techniques to address sorting and selection on observables in our sample.
DATA COLLECTION AND EXPERIMENTAL PROTOCOLS
Our sampling frame and experimental protocol are summarized as follows. Ethnic Serb subjects are recruited using a stratified random sampling method, where sampling units, in this case municipalities, are selected using a probability proportion to size method by local Serb recruiters supervised by one of the authors. Subjects are invited in groups of 20 to a location in their community (school or other location with a private room), which serves as the lab-in-the-field. All subjects receive instructions using a standard script read by a local administrator and engage in hypothetical examples of decision-making before the experiment begins. Subjects then receive a sheet of paper indicating the ethnicity of the recipient and 10 options ranging from 0 to 5 (in 0.5 euro increments). They must decide how much of the 5 euros to send to the recipient and how much to keep for themselves.Footnote 2 Whatever sum they select is matched to an anonymous recipient of random ethnicity at a future experimental session.Footnote 3 The experiment is then repeated for a within-subject design using random in-group/out-group ordering of ethnic treatments.Footnote 4 Subjects also complete a self-administered survey following the experiment.
In total, 158 Serbs (adults aged 18 years and over) completed the experiment, where 25% are sampled from inside the largely Albanian populated southern region of the country, an additional 25% are sampled from the north-south border region of North Mitrovica, and the remaining 50% are located inside the predominately ethnic Serb enclaves in the north (Figure A1 in Supplementary Appendix).Footnote 5 For simplicity, we identify Serbs from these three regional groups as “Primarily Albanian region,” “Border region,” and “Primarily Serb region” in order of decreasing proximity to Albanians.Footnote 6 We now consider whether Serb preferences in the Primarily Albanian region in the south differ from their co-ethnic counterparts in border territory and fortified Serb enclaves in the north.
RESULTS
Figure 1 indicates raw data box-whisker plots of the amount of money (ranging from 0–5 euros) sent by ethnic Serbs to other Serbs and Albanians across our three regions of interest. Comparison of means using t-tests in Table 1 indicates that Serbs in the Primarily Albanian region are the most altruistic, toward both other Serbs and Albanians, least likely to show in-group bias, and most likely to treat out-groups equally relative to in-group recipients. Tobit regression analysis in Table 2 indicates that observations of greater altruism among Serbs in the Primarily Albanian region are robust to controls for age, gender, education, employment (a proxy for income), urban vs. rural locations, victimization by violence, property damage due to violence, and displacement by violence (see Table A2 in Supplementary Appendix for a description of variable coding).
Altruism by Region (Mean Comparisons with Two-Sample t-tests Assuming Unequal Variance)
Note: ***p < 0.01, **p < 0.05.
In-group vs. Out-Group Dictator Giving by Location (Tobit Regression on Dictator Giving with Extended Controls)
Notes: Standard errors clustered by experimental location are in parentheses.
***p < 0.01, **p < 0.05, *p < 0.1.
See Appendix Table A2 for variable description.
Box-Whisker Plot of Serb Dictator Giving to In-group (vs.) Out-group Recipients by Location
We now consider whether our results could be sensitive to sorting effects and other selection problems. Table A1 in Supplementary Appendix shows that our sample is well balanced across regions on gender, age, education, employment, and other salient demographics. Refusal and turnout rates were also common across regions, perhaps in part due to financial incentives to participate in the study. We then question whether Serbs could have been selectively targeted or self-selected before, during, and after the war to different regions depending on their relative tolerance of Albanians. To address this concern, we look to see if subjects who are displaced by violence differ observably from others in our sample. In Supplementary Appendix Table A3 we show that displaced persons are not significantly more hostile toward Albanians based on a range of indicators, including out-group altruism, out-group trust, social distance, threat perception, and perceptions of fairness.Footnote 7 Also, our sample is well balanced in terms displacement and victimization by violence across regions. This helps mitigate concerns that hostile Serbs were purged or self-selected out of the Primarily Albanian region compared with other regions in our sample. In an environment of ethnic cleansing such as Kosovo, everyone within reach was a potential target, reducing the likelihood that our results are driven by regional selection into victimization and displacement.Footnote 8
Finally, signaling altruism is just one method through which a group might demonstrate a pro-social commitment to former rivals and to other in-group members. Additional manipulation checks and ecological validity tests in a supplementary online appendix indicate that Serbs in the Primarily Albanian region are more receptive of Albanians on a number of dimensions, including support for a common state, shared political parties and leadership, trust, threat perception, social distance, religious tolerance, and reconciliation indicators, including forgiveness for crimes of the past.
CONCLUSIONS
Using dictator experiments with ethnic Serbs in post-war Kosovo, we ask how varying degrees of exposure and proximity to ethnic Albanians affect in-group/out-group altruism in a within subject design. After controlling for a range of potential confounders and possible sorting and selection effects, we find that ethnic Serbs who live in close proximity to rival Albanians display greater altruistic norms (both greater inter-group bridging and intra-group bonding norms) than those who are increasingly removed from interaction with Albanians.
The results from our experiment lend further credence to the contact hypothesis and more broadly to integrationist perspectives for overcoming violence. In reviewing the contact literature, most research shows that contact “works” to reduce prejudices primarily by facilitating better information, reducing fear, and increasing empathy toward former adversaries (Pettigrew and Tropp Reference Pettigrew and Tropp2008). Our research speaks most clearly to the third mechanism. Behaviorally, we find that empathetic preferences are greater when a group is compelled to coexist with former rivals and adversaries than where they are left to their own self-governance, remain largely detached from one another in their political and economic affairs, and in every-day life. Serbs with greater exposure to Albanians are increasingly empathetic, as indicated by greater altruism in the dictator game. Data from our companion survey also indicate that Serbs living in greater proximity to Albanians are less fearful, more trusting, less socially distant, and more open to reconciliation. Our results provide insights for how peace could be sustained among former enemies when circumstances compel them to live together rather than apart. Contrary to assumptions of inter-group conflict and threat, individuals in the aftermath of violence appear willing to shift preferences toward accommodating out-group others when they must live in close proximity–as evidenced behaviorally by costly signaling in a dictator game and corroborated by observational data on support for political integration, social tolerance, and reconciliation after violence.Footnote 9
Supplementary Material
For supplementary material for this article, please visit Cambridge Journals Online. http://dx.doi.org/10.1017/xps.2014.18.
