Purpose: One of the most important public policy debates over the past decade has been the appropriate role of choice in U.S. education policy. This debate stems in part from our lack of consistent evidence that schools of choice (charter schools, Catholic schools, other private schools, etc.) are more effective than traditional public schools. A major roadblock in this area is the problem of selection bias, which inhibits our abilities to infer a causal effect of choice schools on student outcomes. This study aims to overcome these difficulties to provide a more rigorous test of the following two research questions: first, how responsive is demand for private schooling to price (e.g., would voucher programs lead to greater private school attendance, and if so, what amount is required to lead to substantial increases?). Two, how do private schools affect student outcomes?

Project Activities: The project team will begin work by updating the Dynarski-Gruber Catholic Schools database, described below, with the most recent Catholic school tuition data. These data will be augmented with multiple secondary data sources, including the restricted-use versions of the decennial census (PUMS) and American Community Survey, the National Health Interview Survey, and the National Health and Nutrition Examination Survey. Using these data, the authors will estimate the price elasticity of demand for private schooling and assess the impact of private school attendance on student achievement and other student outcomes.

Products: The products of this project will be an update and expansion of the Dynarski-Gruber Catholic Schools database and published reports of the results of the secondary data analyses. The reports will expand our knowledge of the potential for school vouchers to impact private school attendance and the causal effect of private school attendance on student outcomes, including achievement.

Structured Abstract

Setting: The analytic sample consists of rural, urban, and suburban households across the United States. The secondary data sources include the Dynarski-Gruber Catholic Schools database, which first collected Catholic school tuition data in 2002. These data were collected from 76 dioceses across the U.S., representing 46 percent of Catholic school enrollment. Information collected included the zip code of the school's physical location, grades taught, total enrollment, enrollment of Catholic and non-Catholic students, tuition costs for a family's first child to attend, and costs for second and third children in a family, and other fees required for attendance. These data will be augmented by the restricted-use versions of the 1990 and 2000 decennial census (PUMS) and American Community Survey (2000 through 2003), which allow for block-group level analysis of student attendance at private schools. Restricted-use versions of The National Health Interview Survey (NHIS), and the National Health and Nutrition Examination Survey (NHANES III) are also used, which include multiple measures of academic and behavioral outcomes. While none of the datasets contain information on both private school attendance and student outcomes, the authors will combine the NHANES data with the decennial census using a two-sample instrumental variables strategy (TSIV).

Population: The analysis focuses on children of elementary school age, from five to 14 years old, who live within ten miles of a Catholic school for which tuition data is available. The sample contains 12 percent of U.S. households and 11 percent of all school-age children, and one-quarter of U.S. cities with populations over 100,000. The sample over-represents black and Hispanic students, since they tend to live near Catholic schools. While the sample contains 11 percent of all school-age children, it contains 18 percent of black non-Hispanic children and 16 percent of Hispanic children.

Intervention: The interventions to be evaluated are: (1) the offer of a tuition subsidy to a family, in the form of discounts for subsequent siblings, to encourage them to send their children to a private elementary school; and (2) resulting attendance at a private elementary school.

Research Design and Methods: This is a quasi-experimental study that relies on several secondary data sources in order to isolate the causal effect of private school attendance on a variety of student outcomes. In the first stage of the analysis, a difference-in-differences approach is used to estimate the effect of the tuition subsidies on demand for private schooling. In the second stage, an instrumental variable approach is used to account for student selection into private schools and estimate the impact of private school attendance on several student outcomes, including cognitive test scores.

Control Condition: The control group consists of children residing in the same neighborhood but whose families face a different price for Catholic schooling. The use of a difference-in-difference approach including fixed-effects for block group and the number of children in a family, plus local variation in Catholic school prices, isolates the effect of price on private school attendance in the first stage of the analysis. This exogenous source of variation in private school attendance then leads to an estimation of the impacts of private schooling on student outcomes that are free from selection bias.

Key Measures: This study assesses a wide range of student outcomes, including the following academic outcomes: WISC/WRAT cognitive test scores, grade repetition, and English fluency. Other measures of interest include behavioral problems, school suspension or expulsion, television use, health, smoking, drug use, physical and sexual activity, and social network formation. These variables are all included in the secondary data sources.

Data Analytic Strategy: A regression framework is used to estimate the price elasticity of private school attendance. Multiple-child discounts at the nearest Catholic school provide identifying variation in price. Neighborhood and family-size fixed effects absorb unobserved differences in the demand for private schooling. The causal impact of private school attendance on student outcomes is assessed using a two-stage least squares approach. In the first stage, the variation in price faced by families living in the vicinities of Catholic schools with different pricing structures allow for an unbiased estimate of the impact of tuition subsidies on the demand for private schooling via a difference-in-difference approach. Multiple-child discounts to the nearest Catholic school provide identifying variation in price, while neighborhood and family-size fixed effects absorb unobserved differences in the demand for private schooling. These demand equations form the first stage for a two-stage least squares analysis of the effect of private school attendance on academic and behavioral outcomes. Effects are estimated separately by race, ethnicity, gender, income, and parental education. The results are further used to simulate the effect of a voucher program on the distribution of different types of students across public and private schools.

Products and Publications

Working paper

Dynarski, S., Gruber, J., and Li, D. (2009). Cheaper by the Dozen: Using Sibling Discounts at Catholic Schools to Estimate the Price Elasticity of Private School Attendance (NBER 15461). Cambridge, MA: National Bureau of Economic Research Working Paper.

** This project was submitted to and funded under Education Policy, Finance, and Systems in FY 2008.