Does School Choice Lead to Sorting? Evidence from Tiebout Variation

Two issues dominate the school choice debate: whether competition would make schools more productive, and whether choice would result in sorting or stratification. In trying to answer these questions, economists have generally focused on the consequences of voucher systems, yet inter-district or Tiebout choice is, and is likely to remain, the main form of school choice in the United States. Surprisingly, we know comparatively little about its effects. This paper focuses on whether inter-district choice leads to sorting, an issue on which there is no consensus. In recent work, for instance, Charles T. Clotfelter (1999) argues that district availability does influence children’s peer groups, but Alberto Alesina et al. (2002) suggest this finding reflects a failure to consider within-district sorting via attendance areas. The issue is difficult to settle because district concentration may be correlated with unobserved factors that affect outcomes of interest. To address this problem, this study uses a research design that builds from an Alesina and Enrico Spolaore (1997)–type model. It starts with the observation that metropolitan areas in fact contain two education markets, one at the primary and one at the secondary level, and that these often contain different numbers of districts. It is natural to motivate such differences as reflecting the presence of higher fixed costs at the secondary level, and the paper presents substantial prima facie evidence consistent with this possibility. Differences in the number of primary and secondary districts make it possible to identify the effects of changes in district availability using between-level, within-area changes in district concentration. Two factors underlie this design’s contribution: (a) the use of within-area variation allows significant controls for unobserved heterogeneity across markets; and (b) the approach takes advantage of education level– specific data, which allows one to control for systematic and significant differences between the primary and secondary sectors. One must bear in mind, however, that both the presence and the magnitude of between-level differences are not randomly assigned, although these do seem to be significantly less correlated with observable metropolitan area (MA) characteristics than are measures of aggregate district concentration. The results suggest that increases in district availability do affect children’s districtand school-level peer groups (with respect to both race and parental education), and that they reduce private enrollment. Thus, district concentration seems to affect both the distribution and the composition of the students in the public sector. These findings are informative as regards some of the mechanisms that drive stratification in the United States. For instance, they are consistent with Steven G. Rivkin’s (1994) finding that inter-district choice has limited the success of past racial integration efforts. Additionally, they are relevant to a variety of analyses that relate variation in school district availability to educational outcomes. The paper proceeds as follows. Section I discusses some background and introduces the research design, and Section II describes the data used. Section III presents results, and Section IV concludes. * School of International and Public Affairs and Department of Economics, Columbia University, 1402B International Affairs Building, New York, NY 10027 (e-mail: [email protected]). For generous feedback, I am indebted to Timothy Besley, David Card, Kenneth Chay, Stephen Coate, Chang-Tai Hsieh, Darren Lubotsky, Randall Reback, Jesse Rothstein, David Stern, and three anonymous referees. Earlier versions of this paper circulated as “Demand Matters: School District Concentration, Composition, and Educational Expenditure,” and “School Choice and School Productivity: What Can Tiebout Really Reveal?” 1 For instance, Dennis Epple and Richard E. Romano (2003) illustrate how sorting is relevant to a number of analyses of school choice, particularly if peer effects exist.