This paper contributes to a growing body of literature on the political economy of public ?nance in developing countries. Its main methodological interest is to demonstrate the usability of household-level data to study political economy features of public ?nances in developing countries that commonly escape empirical scrutiny due to poor data availability. The immediate empirical interest is in testing whether there is evidence for or against either of two competing models of political targeting of public sector spending in Zambia: the swing-voter versus the core-voter model, the proposition being that in “typical” neo-patrimonial regimes in sub-Saharan Africa, the core-voter model should prevail. I use data from Zambia’s Living Conditions Monitoring Survey (LCMS) to investigate whether there is evidence that the ruling Party in Zambia followed political motives in targeting public infrastructure spending at the turn of the millennium. I ?nd strong and robust evidence for the core-voter model applying to social infrastructure provision in Zambia. The ?ndings suggest that it is primarily the construction of new health and education facilities that is affected by political targeting, whereas there is no strong evidence for such targeting for the improvement and rehabilitation of existing infrastructure. For the roads sector, the evidence is less conclusive: although the estimates show the same pattern as in health and education, they are not as robust to modi?cations in the econometric speci?cation.