Discussant: Johan Rocha, Department of Comparative Human Development University of Chicago

Abstract: Does diversity lead to social fragmentation or integration? This question animates recent research on ethnic homophily – operationalized as the tendency to form social relationships with others of similar national origins – in friendship networks in diverse contexts. While national origins seems like a reasonable proxy for subjectively felt commonality underlying ethnic homophily, we find it implausible that friendship ties are shaped by such “passport homophily” in practice. According to historical-comparative work on nationhood and ethnicity, national origins subsume two major domains of culture that are likely to shape social relations: language and religion. The former is critical to coordination of social life, and the latter to its regulation through beliefs and norms. Based on data from hundreds of classroom friendship networks in four European countries, we find that national origins diversity is indeed entangled with linguistic and religious diversity, obscuring the potentially distinct impacts of each on relationship patterns. We fit exponential random graph models to friendship ties in classrooms across ethnic, linguistic and religious groups. Through a random-effects meta- regression analysis of coefficient estimates across classrooms, we show that ethnic and linguistic homophily similarly peak around moderate levels of ethnic and linguistic diversity, and decay at higher levels. Religious homophily, by contrast, increases linearly with religious diversity. Our estimates suggest that linguistic homophily peaks when about 60% of the classroom contains linguistic minorities, at which point shared language increases the probability of friendship by 6%. Religious homophily is highest when about 90% of the classroom is made up of religious minorities, at which point shared religion increases the probability of friendship by 8%. Additional analyses suggest that these patterns are partly driven by a subset of mixed minority status students who are highly sensitive to the diversity context, and dramatically increase their preference for in-group friendship as diversity increases. Overall, these results suggest patterns of relational fragmentation in diverse contexts, and point to a divergence in tie formation mechanisms embedded in elementary differences between major domains of cultural difference – language and religion – that have hitherto been ignored in scholarship on ethnic diversity and social cohesion.