Intimate partner violence (IPV) is a major health and human rights problem globally. However, empirical findings on the predictors of IPV cross- culturally are highly inconsistent, and the theory of IPV is underdeveloped. We propose a new analytical framework based on cooperative game theory in which IPV is a function of the power relations of the dyadic relationship, not simply the actors involved. Using data from the 2005 Colombian Demographic and Health Survey, we test the hypothesis that IPV is predicted by large asymmetries in dyadic power using a hierarchical generalized linear model. Results suggest that education, urban residence, age at sexual debut, whether the woman has other sexual partners, and the age difference between spouses have strong effects on the log-odds of a woman experiencing IPV. Cooperative game theory and social network analysis offer a general approach to the problem of intimate partner interactions which can be applied broadly cross-culturally.