Evaluating the Impact of Nontraditional School Choices

Abstract

In my first paper, I address the question of whether the test score “bump” for charter schools signals real improvements in mastery of the domain being assessed, or whether it is an artifact of score inflation specific to one high-stakes test. I disaggregate high-stakes mathematics test results from New York into subgroups of items and investigate whether the charter-school performance differential is localized in ways that indicate instructional reallocation or coaching on particular item types or content. My findings suggest that charter schools do focus more on frequently-tested curricular strands and standards, compared to district-run schools serving similar student populations. I also find that charter students have larger score differences on constructed-response vs. multiple-choice items, suggesting that charter schools are coaching to the rubric used to score constructed-response items more, or more effectively, than are district-run schools. My second and third papers estimate the academic impacts of a long-standing inter-district integration program. I exploit exogenous variation in the resident enrollment of the METCO suburban districts to estimate the program’s causal effect on the average standardized test scores of applicants. I find that a 2.5% increase in resident enrollment causes a decrease of about .06 SD in applicants’ later average writing scores, but has no significant effect on math or English Language Arts (ELA) scores. In descriptive results, I show that 3rd-8th grade students in the program have average performance levels above their demographically similar peers in the sending district’s schools in ELA and writing but not in math, and that they perform.4 SD lower on average in math than their peers in the city’s charter schools. I also investigate impacts on students’ later outcomes by comparing program applicants who received a referral to a suburban district to those who did not, using student-level covariates to adjust for nonrandom selection. I find that referred students had an adjusted high-school graduation rate 18 percentage points higher than students who did not, while their college enrollment rate was 17 points higher. The latter difference is due to enrollment in four-year, not two-year, institutions and is statistically equivalent for girls and boys.