Does electoral accountability discipline public spending? After the fall of Suharto, Indonesia held local elections for the first time in decades. I study how this democratization affected the spatial allocation of public investment in healthcare infrastructure. Using the staggered rollout of new hospitals, clinics, and subclinics over time, I estimate a spatial demand system for healthcare that allows me to quantify the surplus generated by any given allocation of facilities. I find that the actual allocation generates only 60% of achievable surplus. Facilities do not go to areas that need them most, especially prior to democratization. To understand why, I model the spatial allocation decision as a dynamic discrete choice problem, and I estimate the government’s objective function by revealed preference. On one hand, I find that Suharto-era biases toward certain areas, such as those within the patronage network, are substantially lower after democratization. On the other hand, spillover effects are less internalized as districts become more focused on their own constituents. The first effect dominates, and democratization decreases misallocation overall.