Abstract: I conduct an artefactual experiment in which married individuals make several sets of dichotomous choices between payments given to their spouses and payments given to themselves. Some individuals are revealed to value their spouses earnings and own earnings equivalently, some prefer their own earnings, while others prefer their spouses’ earnings. The different household types revealed experimentally explain part of the differences in real-world household income, which mainly comes from household production in this context. Controlling for household productive asset levels and other covariates, households where either spouse equivalently values the spouse’s earnings and her own earnings have a significantly higher conditional income than the remaining households. Such a result is consistent with predictions from a non-cooperative Cournot-Nash model of household production, where individuals’ marginal rates of substitution of their spouses’ production profits for their own profits, if different from one, drive a wedge between their individual utility-maximizing allocation decisions and the Pareto efficient production factor allocation within the household. The estimates imply that about 8.2 percent of total household income loss is associated with intrahousehold productive inefficiency.