Gender and Performance: Evidence from School Assignment by Randomized Lottery

Abstract

School choice programs are intended to improve student achievement, by allowing for better matches between students and schools. It is not clear, however, that academic achievement will improve if parents make school choice decisions over both academic and nonacademic school attributes (Justine S. Hastings et al., 2005, 2006). Indeed, many randomized studies of impacts of school choice find little or no effect of school choice on academic outcomes. For example, initial evaluations of randomized voucher experiments in Milwaukee and New York City found modest academic impacts on eligible students (John F. Witte et al., 1995; Daniel P. Mayer et al., 2002). More recently, evaluations of public school choice lotteries in Chicago and Charlotte have found no difference between the average lottery winner and loser in academic outcomes such as test scores (Julie Cullen et al., 2003; Hastings et al., 2006). When parents are choosing schools for academic and nonacademic reasons, school choice may increase utility but not necessarily improve academic outcomes. There is growing evidence that educational interventions may have heterogeneous treatment effects by gender. Analysis of the Moving To Opportunity demonstration, in which parents were randomly given the opportunity to move to nonpoverty neighborhoods, found improvements in education, mental health, and criminal behavior for females, but negative effects on males (Jeffrey R. Kling and Jeffrey B. Liebman, 2004). Similarly, Michael Anderson (2005) reanalyzed data from three randomized trials of early childhood education and found that all of the long-term benefits accrued to girls and not to boys. We use data from a public school choice program, with school assignment by lottery, to estimate the impacts on academic outcomes by race and gender of attending a first-choice school. Our data come from the CharlotteMecklenburg school district (CMS) in North Carolina, which introduced district-wide public school choice in the fall of 2002 after a racebased busing plan was terminated by the courts. The data include students’ choices, lottery numbers, school assignments, demographics, and academic achievement for the years surrounding implementation of school choice. We compare outcomes for those making similar choices, whose school assignment was determined solely by lottery number. Overall, there was no gain in academic achievement for those winning the lottery. White females did experience significant improvements in test scores when randomized into their first choice school, however. White females were also more likely to choose academically focused magnets and, among those who won the lottery, reported significant increases in time spent on homework. Our evidence suggests that school choice programs may have heterogeneous treatment effects by gender, which are related to differences in the factors driving parental choices.