Welfare economics relies on access to agents’ utility functions: we revisit classical questions in welfare economics, assuming access to data on agents’ past choices instead of their utilities. Our main result considers the existence of utilities that render a given allocation Pareto optimal. We show that a candidate allocation is efficient for some utilities consistent with the choice data if and only if it is efficient for an incomplete relation derived from the revealed preference relations and convexity. Similar ideas are used to make counterfactual choices for a single consumer, policy comparisons by the Kaldor criterion, and determining which allocations, and which prices, may be part of a Walrasian equilibrium.