We estimate and test the restrictions of a collective model of household consumption, using z-conditional demands, in the context of a large conditional cash transfer program in rural Mexico. The model can explain the impacts of the program on the structure of food consumption. We use two plausible and novel distribution factors: the random allocation of a cash transfer to women and the relative size and wealth of the husband’s and wife’s family networks. Our structure does better at predicting the effect of exogenous increases in household income than an alternative, unitary, structure. We cannot reject efficiency of household decisions.