Tiebout bias on the demand for local public goods

Abstract

Empirical work estimating the demand for local public goods has been based upon the median voter model. It has generally proceeded with little consideration given to the possibility of Tiebout-type migration emphasized in the property value literature. In this paper we raise the question of how valid the typical cross section demand estimation procedures are when possible Tiebout-like migration among jurisdictions is taken into account. We find that such procedures are not in fact consistent with a model in which people vote with their feet as well as with ballots. We also show that, if a full Tiebout equilibrium has been achieved in a set of communities, there is a simple and obvious way to estimate demands for local public goods which is different from the median voter approach. This technique has not, to our knowledge, been used in existing studies, and, moreover, has some practical as well as theoretical advantages. If the workings of the Tiebout process are not taken into account, more is involved than just impracticality: we show that the estimated effects of personal characteristics will generally be biased, and that that direction of the bias can usually be determined.