Efficiency and Equity in the Provision of Public Education

Abstract

The paper focuses on the benefits derived from public elementary and secondary education; monetary values of households' perceived benefits from operating expenditures are estimated. Information concerning the preference mappings of citizens is obtained using a random utility model to analyze micro voting data. The estimated model in turn provides information concerning the efficiency and distributional consequences of public education expenditures. The conclusion is that operating expenditures exceed the estimated efficient level. Furthermore, it may be difficult to justify the inefficiently large expenditures on equity grounds since the estimated distribution of households' perceived benefits, net of local taxes, is pro-rich. G~ OVERNMENT provision of goods and I J~ services has the potential of greatly affecting the welfare of citizens. This paper focuses on the benefits derived from one such service, public elementary and secondary education; monetary values of households' perceived benefits from public education expenditures are estimated. Information concerning the preference mappings of citizens is obtained using a multinomial probit, random utility model to analyze micro voting data. The estimated model in turn provides information concerning the efficiency and distributional consequences of public education expenditures. The conclusion is that operating expenditures exceed the estimated efficient level. Furthermore, it may be difficult to justify the inefficiently large expenditures on equity grounds since the estimated distribution of households' perceived benefits, net of local taxes, is pro-rich. In contrast to this study, a number of studies employ ad hoc rules to measure citizens' benefits from public elementary and secondary education. Initially total benefits are assumed to equal the cost of educational services; this input cost measure of benefits is then allocated to households using some unproven rule of thumb.' When an ad hoc rule is used to allocate a questionable measure of total benefits, the results are even more suspect. Conceptually, it is quite odd to determine (assume) total benefits and then allocate the total to individuals or households. The correct method is first to obtain estimates of individuals' or households' benefits. These estimates can then be used to determine the magnitude of total benefits and analyze how the benefits are distributed across the population. Aaron and McGuire (1970) criticize ad hoc benefit measures and propose the income value measure. For X' units of the politically allocated good X this measure is IV = MRSj X' where MRSj is the marginal rate of substitution of X for a composite commodity evaluated at X' and one's aftertax income. This measure is comparable to measures of the income values of goods purchased in private markets. While Aaron and McGuire argue for preferencebased benefit measures, the lack of information concerning preference mappings necessitates that they assume the shape of the utility function. Even though the results by Maital (1973), based on a similar method, are slightly less restrictive, a number of restrictive assumptions are common to both studies. Not only are households assumed to have identical, separable utility functions, but the existing levels of politically allocated goods are assumed to be efficient, implying the total income value equals the input cost. Also, the allocation of benefits from mixed goods (e.g., public education) between public and private components is based on ad hoc rules. The major contribution of the current study is that the benefit estimates are based on information concerning the preference mappings of individuals as revealed through referenda voting.2 The estiReceived for publication March 9, 1983. Revision accepted for publication June 4, 1984. *State University of New York-Albany. This paper was improved as a result of the advice and encouragement of Kenneth Wertz, John Akin, David Guilkey and James Wilde, as well as the helpful comments of two referees. Financial support for this research was, in part, provided by a Dissertation Research Fellowship from the Brookings Institution. 1 An example is the 1967 Tax Foundation Study. 2NMicro voting data are also used by Rubinfeld (1977) to obtain information about citizens' preferences for public school expenditures; he uniquely estimates the parameters of a demand function up to an unknown multiplicative constant. A micro demand function for education is uniquely estimated by Bergstrom, Rubinfeld, and Shapiro (1982), using a different type of survey data. While the method used here is to directly estimate preference mappings, the information concerning such mappings embedded in such a demand function provides an