Preference for Similarity and Information Collection in Groups

We then consider the group of peers (or friends) as an object of choice. We characterize the peer group's optimal composition for each individual in the population. We show that, for each individual, there is a large equivalence class of optimal groups, potentially with maximal variance of tastes. We also characterize the stable groups, that is, groups that are optimal for all participating individuals, and illustrate the similarity requirements such stable groupings entail. Our results help explain an array of applications, ranging from friendship formation, to selective exposure.