Estimating the Effects of Private School Vouchers in Multidistrict Economies

This paper estimates a general equilibrium model of school quality and household residential and school choice for economies with multiple public school districts and private schools. In the model, households have heterogeneous religious preferences, and private schools can be either Catholic or non-Catholic. The model is estimated with 1990 data from large metropolitan areas in the United States, and the estimates are then used to simulate two large-scale private school voucher programs in the Chicago metropolitan area: universal (non-targeted) vouchers, and vouchers restricted to nonCatholic schools. The simulation results indicate that both programs increase private school enrollment, and affect household residential choice, with the largest school quality and welfare gains accruing to low-income households. Universal vouchers increase enrollment at both Catholic and non-Catholic private schools, yet when vouchers are restricted to non-Catholic schools, the overall private school enrollment expands less at moderate voucher levels, and enrollment in Catholic schools tends to decline, relative to the non-voucher equilibrium, as the dollar amount of the voucher increases. In addition, fewer households benefit from non-Catholic vouchers. This paper is based on my PhD dissertation at the University of Wisconsin-Madison. I wish to thank my advisors Derek Neal, Steve Durlauf and Phil Haile for their guidance and encouragement. I am also grateful to Peter Arcidiacono, Dennis Epple, Arthur Goldberger, John Kennan, Yuichi Kitamura, Bob Miller, Tom Nechyba, Richard Romano, Ananth Seshadri and Holger Sieg for many helpful ideas and suggestions. I owe special thanks to Grigory Kosenok for his help with the technical aspects of this project and for many insightful conversations. I have also benefited from conversations with Oleg Balashev and Hiroyuki Kasahara, and wish to thank seminar participants at Brown, Carnegie Mellon, Columbia Business School, Duke, Rutgers, U. of Illinois at UrbanaChampaign, Virginia, and Wisconsin, and participants at the 2002 CIRANO conference on education, at Camp Resources X, and at the Stanford Conference on Empirical Boundaries for IO, Public Economics and Environmental Economics, for their questions, comments and suggestions. I am grateful to the UW-Madison Condor Team from the Computer Science Department, and in particular to Peter Keller, whose help has been essential to carry out this project. Financial support from the UW Department of Economics to purchase data, and from the UW Robock Award in Empirical Economics are gratefully acknowledged. All errors are mine.