The game theoretic study of coalitions focuses on settings in which commitment technologies are available to allow groups to coordinate their actions. Analyses of such settings focus on two questions. First, what are the implications of the ability to make commitments and form coalitions for how games are played? Second, given that coalitions can form, which coalitions should we expect to see forming? I examine classic cooperative and new non-cooperative game theoretic approaches to answering these questions. Classic approaches have focused especially on the first question and have produced powerful results. However, these approaches suffer from a number of weaknesses. New work attempts to address these shortcomings by modeling coalition formation as an explicitly non-cooperative process. This new research reintroduces the problem of coalitional instability characteristic of cooperative approaches, but in a dynamic setting. Although in some settings, classic solutions are recovered, in others this new work shows that outcomes are highly sensitive, not only to bargaining protocols, but also to the forms of commitment that can be externally enforced. This point of variation is largely ignored in empirical research on coalition formation. I close by describing new agendas in coalitional analysis that are being opened up by this new approach.
