Résumé en anglais

The classic Working-Leser household Engel curve is unpacked to reveal individual budget allocations across commodities as a function of both individual and household total spending. Two main findings emerge on calibrating our model to an unusual sub-household dataset for Senegal. First, for all except education spending, our results are consistent with the separable structures found in two-stage bargaining and collective models of the household. Second, there are large biases in standard household Engel-curve estimates when compared to consistently aggregated sub-household estimates, though in differing degrees and directions depending on the type of commodity. The main source of bias is a household effect on sub-household spending behavior, though this is partially offset by a bias due to intra-household inequality, which emerges as a confounder in aggregating to the household level.