Existing work suggests that recipient governments selectively allocate their portfolio of aid projects to draw political support from their citizens. This paper introduces a formal model that allows citizens to hold heterogeneous preferences for aid projects along both cost and non-cost dimensions, including the type of project, the donor, and the level of recipient government involvement. The results of this model indicate the politically optimal share of aid allocated to each type of aid project given citizen demand. I then directly estimate the weight citizens place on the theoretically specified cost and non-cost dimensions of aid projects using a unique conjoint experiment from around 1,300 respondents in Northern Uganda. These experimental weights enter the formal model to estimate the demand for aid in Northern Uganda, the results of which imply substantial political returns to allocating social development projects, partnering with Western donors, and targeting projects with a large grant component.