Search

When public expenditures for local services are determined by a referendum process, it is impossible to tell a priori whether increasing the proportion of nonusers increases or decreases the voting equilibrium [1,2]. For a given tax price and income nonusers have a lower demand than users for local services, but the tax price of local services to users is lowered by the presence of nonusers. Stigler [3] therefore asks an empirically interesting question when he seeks to ascertain the influence of Catholic parochial school parents on expenditures per public school pupil. This reply shows that Stigler's use of arithmetic averages for state-wide data probably biases his regression coefficients, and that an alternative specification provides a more powerful test of Stigler's model. Following Stigler let us assume that the demand function for expenditures per public school child is Cobb-Douglas and that all families within a given political jurisdiction pay the same taxes. Letting si denote the proportion of voting families who use the public schools in district i, Ti taxes per family desired by an "average" user, yi the income of an "average"user, and P the cost per "quality" unit of education (assumed constant across observations), Stigler shows that