This paper develops a method for using data for a large cross-section of municipalities relating expenditures on specific local public goods, median income, median house value, total assessed valuation, and population to estimate demand functions for local public goods. The key idea is to make the assumption that the quantity chosen in any municipality is the median of the preferred quantities of its citizens. The method is applied to cities with population exceeding 10,000 in several states. Seemingly plausible estimates of income and price elasticity are found. The estimated crowding parameter suggests that most local public goods are congestible in the sense that utility functions depend on the per capita quantity of public goods.