Responses to questions given to a random sample of Michigan households are used to estimate public spending demand functions. While income and price elasticities are similar to those obtained from aggregate data, positive income elasticities appear to arise because public services are distributed in a prorich manner. A relatively small variance in spending demands among urban and suburban communities in metropolitan areas with substantial public service variety suggests that the Tiebout mechanism works. This interpretation is supported by the fact that actual spending conforms substantially to desired levels in urban areas, but less so in rural areas with little public sector choice.

We devise and apply a new method for estimating demand for local public goods from survey data. Individuals' responses to questions about whether they wanted more, less, or the same amount of various local public goods are combined with observations of their incomes, tax rates, and the amounts of actual spending in their home communities. Parameter estimates turn out to be quite similar to those found with studies like Bergstrom and Goodman's study based on total expenditures across communities.