The question of whether private schools provide better education than public schools is at the center of the current national debate over increasing choice in education. Much controversy surrounds the literature on the relative effectiveness of public and private schools. The controversy stems from the fact that selection into private schools is not random. In this project we will attack the selection problem for Catholic schools by taking advantage of the fact that proximity to Catholic schools has a strong effect on the probability of Catholic school attendance for Catholics but not for non-Catholics. This permits the use of the interaction between religious preference and proximity to Catholic schools to statistically identify Catholic school effects while controlling for both proximity and religious preference. We will implement the approach using two large panel data sets with precise geographical identifiers and information on the neighborhood and local public schools and on the location of Catholic schools. Our objective is to provide convincing estimates of Catholic school effects on test scores, teenage pregnancy, the probability of high school graduation, high school curriculum, attitudes toward education and work, achievement test scores, college attendance and completion, employment, and wage rates. We will investigate how Catholic school effects vary with neighborhood characteristics, school characteristics, and student characteristics such as race, gender, parental education, family income. We will develop new methods of accounting for unobserved heterogeneity in the effects of Catholic school attendance. Finally, we will provide evidence on the effects of individual characteristics, neighborhood characteristics, public school characteristics (including student body characteristics), and proximity to Catholic schools on Catholic attendance. As part of the project, we will differentiate between the effects of single-sex and coed Catholic schools on a variety of outcomes. A novel feature of the project is that we will use the interactions between the sex of the student, religion, and the distances from the nearest all boys, all girls, and coed Catholic schools to deal with bias that arises from nonrandomness in who chooses single-sex and coed schools.
