ABSTRACT
This chapter focuses on charter school studies that adopt one of two methods. The first approach involves comparing students who win and lose lotteries to attend charter schools. The second approach, known as value-added modeling, is not experimental, but takes into account a student's past academic achievement, unlike some of the weaker non-experimental approaches. The chapter focuses on studies of urban districts or on individual schools in urban areas. In all but one case the effect sizes are positive and statistically significant. Turning to variations by students, a key question is whether charter schools benefit students from some groups more than others. A warning about the subgroup estimates is that only a subset of studies have performed the analyses, meaning that they may not apply to the locations studied nationally. Charter schools operating as middle schools produce quite high predicted effects for both math and reading relative to charter schools in other grade spans.
