This paper adapts the generalized axiom of revealed preference (GARP)
empirical method to the public goods problem to test whether observed
municipal public spending can be explained ''as if'' the city governme
nts maximize the utility of the median income voter. It applies the te
st procedure for medium-size municipal governments in five Midwest sta
tes. The data are consistent with GARP and reveal that the local gover
nments in the sample behave as if they maximize median voter utility o
nce we control for the state-specific effects, government management s
tructure, and population density.