This dissertation studies the political economy of identity. What are the costs and benefits of integrating different groups in society? When different groups do interact, how can we increase benefits while decreasing costs? Is culturally-conditioned behavior malleable? In the first chapter, we study the costs and benefits of integrating different groups on social learning. A theoretical model suggests that integration is less beneficial as individual's become more biased against out-group members, and more beneficial when one group benefits to learn information that the other group knows. We find empirical support for these hypotheses in an experiment conducted in segregated high schools in Macedonia, and examine implications of our results on classroom organization in these schools. In the second chapter, we study how cooperative contact between different ethnic groups shapes cohesion and productivity, and how this is in turn affected by the presence of linguistic diversity. In segregated high schools in Macedonia, we find that when students are mixed with a different-ethnicity partner in a video game, their cohesion goes up, with no cost to productivity, but some non-pecuniary costs to participants. Importantly, these results seem to be driven by pairs that can speak a common language. In the third chapter, we study the malleability of cultural norms at the individual level. We conducted an online study with bilingual Latinos in the US, randomizing whether participants respond to questions in Spanish or English. We find that individuals exhibit cultural norms more similar to Americans in English compared to Spanish, indicating that language shapes selective usage of cultural norms and cultural flexibility even within the same individual.