The Centrality of Access

• In Event: The Centrality of Access in Charter School Debates

Abstract

Participant #4 has written extensively about the practices that some charter schools use to shape their enrollment and has also written about the legal framework within which charter schools operate. This participant’s work examines access-related practices as well as the incentives and disincentives—including rules, regulations, and constraints—that influence those practices. The work argues that, because of connections between access and the many other charter-school issues that the panelists of this symposium will be discussing, access is the cornerstone for understanding charter schools.
Charters interact with policies that reward exclusionary practices, this participant has explained. While some charter-school leaders have begun to address access barriers, they have done so in the face of substantial headwinds, due to the larger context and incentives. If the system continues to incentivize charter schools to act inequitably, the nation will continue to see only these sporadic attempts at access-increasing practices by individual actors. Larger-scale reform therefore entails changing funding incentives and test-based accountability incentives.
Drawing on this work, Participant #4 will bring to the symposium a perspective that illuminates the centrality of access. As one example, consider the validity challenges facing researchers comparing student outcomes.
There are two main types of charter-school-outcome studies: those using lotteries and those using longitudinal panel data. Studies based on lotteries have stronger internal validity but questionable external validity. For example, most charter schools in the Boston area are in high demand, so lotteries determine many admissions offers. This generates a random assignment of students in “intent to treat” and “control” categories. Although these categories often become muddied, they generally allow for well-grounded causal inferences about the outcomes of those over-enrolled charter schools. But they are not generalizable to charter schools that are not over-enrolled or to a population of students whose families did not participate in the choice process.
Studies using longitudinal data sets can have stronger external validity but greater threats to internal validity. These studies use OLS regression or matching (propensity score or “virtual twin”) approaches to compare charter-school students to students enrolled in local public schools. This approach is particularly susceptible to omitted variables, since charter-school students (and their parents/guardians) have actively opted into the choice process. In contrast, we can think of the group of comparison students as including two subgroups: (a) all students who did not engage with the choice process because their families lack the time, resources, or social capital to engage; and (b) students whose families have the time, resources, or social capital yet have nonetheless eschewed the charter school option. Students in this second subgroup are the best comparison, but they cannot be separated from the first.
Access issues can introduce systematic enrollment bias, amplifying these problems of non-random assignment. This is most obvious for charters that discourage or exclude students who lack the right “fit”. But even subtler enrollment-shaping choices involving, for example, location and advertising can result in systematic yet unmeasured differences between those families that apply for enrollment and those that do not.

Author

• Kevin G. Welner, University of Colorado Boulder