We use Oportunidades, a conditional cash transfer to women, to show that standard demand models do not represent the sample's behavior: Oportunidades increases eligible households' food budget shares, despite food being a necessity; demand for food and high-protein food changes over time only in treatment areas; the treatment effects on food and high-protein food consumption are larger than the prediction from the Engel curves at baseline; and the curves do not change in eligible households with high baseline bargaining power for the transfer recipient. Thus, handing transfers to women is a likely determinant of the observed nutritional changes.