The intergroup contact hypothesis posits that as people get to know members of an outgroup, they will reduce their prejudice towards the entire outgroup (Allport, 1954). A seminal meta-analysis found that intergroup contact, especially under conditions identified by Allport (1954), is reliably associated with reduced prejudice (Pettigrew & Tropp, 2006). This relationship is supported by subsequent reviews (Lemmer & Wagner, 2015; Paluck et al., 2019). Research in divided societies similarly finds a positive association between intergroup contact and improved outcomes, including reduced prejudice and anxiety, and increased openness to reconciliation (Hewstone et al., 2006), in contexts such as Israel (Faibish et al., 2023), Cyprus (Asimovic et al., 2023), South Africa (Fourie et al., 2022), and Northern Ireland (Reimer et al., 2022). More recent work, however, reassesses expectations about the magnitude of effects (Lowe, 2025), and highlights the need to better understand what constitutes effective intergroup contact in real-world settings.

Intergroup contact programs organized around shared interests, often facilitated by NGOs, are among the few opportunities that bring people from different groups together. These programs respond to a key challenge in divided societies: the numerous barriers to intergroup contact. Such barriers may include formal physical segregation, e.g. the checkpoints between the West Bank and Israel or the UN-patrolled buffer zone in Cyprus. But even in the absence of structural divisions, individuals may still choose to remain socially or spatially segregated (Dixon et al., 2020; McKeown et al., 2012). In these programs, volunteers meet regularly with members of the other group in a structured setting. Joining an intergroup contact program likely constitutes a major intergroup contact event (Shulman et al., 2025), as it may be the first time young participants encounter the outgroup in a social setting. Once enrolled, positive intergroup contact experiences are likely to accumulate over time. Initially, participants may remain within their own groups but gradually open themselves up to more meaningful engagement with the outgroup.

Unlike intergroup dialogue programs that focus on discussing the conflict (Nagda et al., 2018), intergroup contact programs often incorporate the encounter into appealing social settings. These programs typically follow a co-existence or joint projects model, which prioritizes shared experiences and highlights commonalities, rather than adopting the confrontational approach that often underlies intergroup dialogue programs (Maoz, 2011). Joint activities can include volunteering, arts or theater (Feuchte et al., 2020), with sports being particularly common (Kidd, 2008) due to its broad appeal and added physical and mental benefits. Despite their popularity, relatively little research within the broad field of intergroup contact research comes from such programs, and the existing evidence paints a mixed picture (Asimovic et al., 2024; Galily et al., 2013; Grady et al., 2023; Lowe, 2021; Mousa, 2020; Zhou & Lyall, 2025). Although these studies document varied outcomes, they generally do not examine the underlying mechanisms, leaving unanswered why certain interventions succeed while others fail. The current research moves beyond asking whether intergroup contact-based interventions are effective to instead examine how and why they work, thus advancing research on field interventions.

Simply bringing people together does not guarantee positive intergroup contact; in fact, exposure can worsen relations when the experience is negative (Enos, 2017). In divided societies with a history of group-based conflict, the quality of contact appears especially important: in fact, high-quality contact may be the only form capable of producing meaningful improvements in intergroup attitudes (McKeown & Psaltis, 2017). This underscores the need to better understand what constitutes meaningful engagement within structured intergroup programs, and the conditions under which it emerges. One form of meaningful engagement within structured intergroup programs is the formation of cross-group friendships, widely regarded as high-quality intergroup contact. The association between cross-group friendships and reduced prejudice is especially strong (Davies et al., 2011; Pettigrew & Tropp, 2006) and, longitudinally, having more cross-group friends in elementary schools predicts lower racial bias two years later (Gaias et al., 2018). Friendships seem to be particularly effective because they reduce intergroup anxiety (Page-Gould et al., 2008) and promote intergroup empathy through self-disclosure (Turner et al., 2007). In divided societies, where many institutions are segregated and relations are particularly strained, cross-group friendships are rare. When they do form through intergroup programs, they can be especially meaningful and potentially drive changes that extend beyond the immediate context of the program.

While friendships clearly matter for engagement with the outgroup, another type of experience may matter equally or more: feeling empowered by the outgroup. Intergroup contact experiences where individuals feel that they are respected for their group’s agency and competence have been referred to as empowering contact (Hässler et al., 2021). Empowering contact was first documented and is primarily studied among members of minoritized groups, where it is positively associated with the motivation to strive for social change within one’s ingroup (Hässler et al., 2021). Some evidence suggests that it may be relevant for everyone (Hässler et al., 2022), and this may be especially true in divided societies where there is some experience of victimization and submission on all sides (SimanTov-Nachlieli & Shnabel, 2014). The experience of victimization creates a need for empowerment, and if this need is fulfilled by an outgroup partner, it increases people’s psychological readiness for intergroup reconciliation (Shnabel et al., 2009).

Structured programs may be the only setting in divided societies where the conditions for cross-group friendships and empowering intergroup contact are present. However, it remains unclear whether and how such programs succeed in ensuring that members of all groups form meaningful connections and feel empowered, especially given that this is often not the case in the broader social context. Our survey examines the role of cross-group friendships and feeling empowered as potential channels through which these programs achieve their impact. Building on these findings, we use expert interviews to explore how programs foster cross-group friendships and empowerment.

Scholarship on intergroup contact mostly focuses on between-group processes, such as prejudice or the intention for future intergroup interactions (Pettigrew & Tropp, 2006). Given past research, our first hypothesis (H1) states that forming cross-group friendships and feeling empowered by the outgroup within a structured intergroup program will be associated with increased intergroup contact intentions. In addition to their focus on between-group processes, structured programs often include an additional focus on within-community transformation (Ditlmann et al., 2017). Within-community transformation can take many forms, including ingroup policing (Ditlmann & Samii, 2016) or in-group based advocacy (Dixon et al., 2020). Indeed, engaging ingroup members in difficult conversations holds significant promise in divided societies. By sharing counter-normative information about the outgroup with their ingroup, program participants can “de-freeze” conflict-supporting beliefs of their community members (Hameiri et al., 2014). This can help disrupt the conflict-escalation spiral, paving the way for conflict resolution and eventual reconciliation. We measure the extent to which participants intend to prepare their communities for reconciliation through constructive dialogue. Thus, our second hypothesis (H2) states that forming cross-group friendships and feeling empowered by the outgroup within a structured intergroup program will be associated with an increased intention to prepare the ingroup for reconciliation.

Recognizing the challenging roles participants are expected to play within their societies, structured intergroup contact programs in divided societies often emphasize personal development. An important psychological resource for youth engaging in reconciliation efforts – both with the outgroup and within the ingroup – is self-efficacy. Self-efficacy, the belief in one’s ability to successfully accomplish tasks and achieve goals has recently been identified as a key predictor of intergroup contact intentions (Bagci et al., 2023). Thus, our third hypothesis (H3) states that making cross-group friends and feeling empowered by the outgroup in a structured intergroup contact program will foster the belief in one’s ability to navigate difficult social situations outside of the program. This, in turn, can increase the likelihood that participants will engage in – rather than avoid – difficult situations in the future, thereby preparing them for further intergroup contact and successful dialogue within their own communities.

We collaborate with an international organization that implements structured intergroup contact programs in divided societies. The program facilitates intergroup contact mainly through sport complemented by other social activities (e.g., occasional joint retreats where the kids spend a weekend together). Sport is well-suited as a vehicle for intergroup contact because Allport (1954)’s optimal conditions are usually met: youths have to cooperate to reach a common goal, they have equal status as players in the game, and important authorities, such as coaches and program leaders, approve of the encounter. Leading with quality sports training also allows the program to appeal to a broad range of potential participants, who are interested in playing sports but maybe less inclined to engage with the outgroup.

While specific activities and schedules vary by location and age (more in Appendix A), the program generally includes three components. First, integrated practice and game sessions are central across all sites, with frequency depending on participant age and local factors such as security concerns. For example, in Israel, children aged 10–12 meet outgroup teams for joint practices with mixed teams about once a month. In Northern Ireland, where security is less of a concern and schools are often supportive, many children participate in weekly mixed training during or after school. Second, participants engage in segregated practice and peace education with their ingroup. The peace education component emphasizes shared humanity and helps children develop empathy and perspective-taking through sports-based exercises. An emphasis on commonalities aligns with the joint projects model in intergroup contact programs (Maoz, 2011); we return to strengths and limitations of this approach in the discussion. Third, the program cultivates psychological resources such as self-efficacy and conflict resolution skills throughout its activities. For example, successfully playing on a mixed team can foster a sense of mastery (Bandura, 1997), strengthening participants’ confidence in engaging with the outgroup. For older participants, formal leadership development components, which we refer to here as “league”, prepare them to become coaches, take on organizational roles, or act as community peacemakers. While implementation varies by site, integrated training remains the program’s core, with peace education preparing participants for contact and leadership training helping them apply what they have learned within their communities.

The research for this paper was conducted in four countries with a relatively recent history of large-scale or organized intergroup violence. Our conceptualization of divided societies largely follows Guelke’s (2012) definition, where “conflict exists along a well-entrenched fault line that is recurrent and endemic and that contains the potential for violence between the segments,” and where “the threat of political violence continually looms large in the concerns of its citizens.” The countries in our study – Northern Ireland, Cyprus, South Africa, and Israel – each have their own complex intergroup relations and distinctive historical contexts (described in detail in Appendix B). They differ in conflict salience, including the duration of the conflict, the recency of major violence, and the degree of structural barriers between groups. Despite these differences, the sites share certain similarities: all have suffered significant group-based violence within the past century, leaving behind enduring legacies and, in some cases, recurrent outbreaks of violence.

They are also characterized by pronounced segregation between main groups. This segregation manifests in various forms, from the tangible presence of checkpoints in Cyprus and Israel or peace walls in Northern Ireland to the less visible but deeply entrenched economic and social divides in South Africa. Participants in our programs navigate these divisions in their everyday lives, which are largely separated from those of the outgroup, and have little opportunities for high quality encounters with the outgroup outside of the program. The conflict dynamics within divided societies change frequently. This quantitative data was collected in 2022, so our results and descriptions are reflective of the events up to that date, whereas the qualitative data was collected in 2024.

As part of their Monitoring and Evaluation strategy, our partner organization recruited a sample of 431¹1

Data were collected from 449 participants but 19 were excluded because they had over 90% missing values as answers.

All questionnaires were translated into local languages by local teams, with team leaders consulted to ensure contextual accuracy. Most participants completed the survey electronically, with the exception of a small subset who did not have access to a device or reliable Wi-Fi connection and therefore completed the survey in person. The organization secured parental consent for program activities at the start of the year and informed consent from youth participants at the beginning of the surveys. The Hertie School Ethics Committee approved the research under the number 20231204-82.

We describe the three dependent variables and two mediators in Table 2 (Appendix C details the coding of the variables; Appendix D full survey questions).

Our analysis examines how time spent in the program shapes participants’ social distance, ingroup-oriented reconciliation intentions, and self-efficacy. We use structural equation modeling (SEM) to estimate associations between self-reported years in the program as our main predictor, representing opportunities for outgroup interaction, and each outcome, modeled as a latent variable to enhance measurement precision. While longer participation does not guarantee more or better-quality contact, it increases the likelihood of such experiences. SEM also enables us to assess theorized mediating mechanisms, namely cross-group friendship and empowerment. All models control for age, gender, country, majority status, and participation in the league, a higher-intensity version of the program. Alternative model specifications are presented in Appendix D as robustness checks.

We begin with a confirmatory factor analysis (CFA) to validate the structure of our latent constructs (Appendix E). The model shows good fit to the data: the Comparative Fit Index (CFI) is 0.956, the Tucker-Lewis Index (TLI) is 0.949, the Root Mean Square Error of Approximation (RMSEA) is 0.055 and the Standardized Root Mean Square Residual (SRMR) is 0.045. These values indicate acceptable to good model fit, suggesting that the latent variables are well represented by their observed indicators (Hu & Bentler, 1999).

We then estimate a structural equation model (SEM) to assess the direct associations between years in the program and three latent outcomes: social distance, ingroup-oriented reconciliation intentions, and self-efficacy. In the baseline model without covariates (Table E1), years in program are significantly associated with lower levels of social distance (b = -0.091, SE = 0.014, p < .001), higher reconciliation intentions (b = 0.072, SE = 0.018, p < .001), and greater self-efficacy (b = 0.105, SE = 0.018, p < .001). Substantively, participants with more years in the program report greater willingness to engage in intergroup contact with outgroup members. Additionally, they report a stronger inclination to prepare their ingroup for reconciliation through confronting ingroup members who insult the outgroup, sharing the perspective of the outgroup, defending their views, and remaining respectful during disagreements. Finally, they report higher self-efficacy, including the ability to converse with unfamiliar people, resolve conflicts, express opinions, and stand up for themselves. These associations remain robust after controlling for age, gender, country, majority status, and participation in the league, a higher-intensity version of the program (Table E2). We next examine the mechanisms underlying these associations by estimating a mediation model.

We test whether cross-group friendships and feelings of empowerment help explain the relationship between program duration and the three outcomes: social distance, self-efficacy, and ingroup-oriented reconciliation intentions. The model, estimated in R using the lavaan package on the combined sample, includes years in the program as the independent variable, number of friends and empowerment as mediators, and accounts for their interrelationship. To obtain more reliable estimates of the mediation paths, we computed bootstrapped standard errors and confidence intervals using 5,000 bootstrap samples. Age, gender, country, majority status, and participation in the league, a higher-intensity version of the program, are controlled for. We report the total, direct, and indirect effects for the full sample in Table 6.

The structural modeling results (Table 6) reveal a nuanced relationship between years in program, the number of cross-group friends formed, and feelings of empowerment. As predicted, there are indirect effects of program duration through cross-group friendships and feelings of empowerment on all three outcomes: social distance, reconciliation intentions, and self-efficacy. For social distance, these indirect effects account for a significant total effect of years in the program. For reconciliation intentions, while the total effect is not statistically significant, the significant indirect pathways suggest that cross-group friendships and empowerment are still channels through which program participation can contribute to reconciliation. In contrast, for self-efficacy, a significant direct effect of years in the program, along with additional significant indirect effects through the same mediators, accounts for a significant total effect of program duration.

We conduct robustness checks for the mediation model presented here (Figure 1) – including a quadratic term for years in the program (Table E6), country fixed effects (Table E4)²2

The effects on efficacy are less robust than for the other two outcomes, when controlling for country fixed effects.