We test the unitary versus collective model of the household using specially designed data from Bangladesh, Ethiopia, Indonesia, and South Africa. Human capital and individual assets at the time of marriage are used as proxy measures for bargaining power. In all four countries, we reject the unitary model as a description of household behaviour, but fail to reject the hypothesis that households are Pareto-efficient. In Bangladesh and South Africa, women’s assets increase expenditure shares on education, while in Ethiopia it is men’s assets that have this effect. These increases have different implications for boys and girls across countries, however.