We develop a revealed preference approach to analyse non-unitary household consumption behaviour that is not cooperative (or Pareto efficient). We derive global necessary and sufficient conditions for data consistency with the model. We show that the conditions can be verified by means of relatively straightforward mixed integer programming methods, which is particularly attractive in view of empirical analysis. Our framework extends to tests for separate spheres and joint contribution to public goods. An application to data drawn from the Russia Longitudinal Monitoring Survey demonstrates the empirical relevance of the non-cooperative consumption model.