Unitary models, assuming a single objective function and unified budget constraint, are traditionally used to model household behavior. Most empirical tests of unitary models rely on endogenous regressors. This paper uses an exogenous change in the intrahousehold distribution of income, provided by a change in U.K. Family Allowance policy. Two approaches address problems inherent in making inference about consumption from expenditure data. Expenditure shares and levels for a range of goods are estimated. Shifts in assignable goods suggest that children benefited at the expense of men when this policy change shifted income within households from men to women.