Over the last two decades, US public school districts have rapidly expanded school choice, providing students with options such as magnet schools, dual-language programs, and theme-based academies, all within their own districts. These reforms are aimed at combatting stark educational disparities—within the same district, students at the lowest-performing schools fall nearly two-thirds of a grade level behind their peers each year—and at responding to competitive pressure from charter schools.
Yet these reforms pose a design challenge: how should districts graft new choice options onto a system built around neighborhood schools? In this paper, the authors examine the structure of public school choice systems and their impact on educational opportunities.
The National Landscape
The authors begin by collecting original data from the nation’s 150 largest school districts to document how choice systems are structured. The data reveal the following:
- Enrollment in non-neighborhood public schools of choice has more than doubled over the past two decades, with enrollment shares now exceeding those of the charter sector and serving more than 14% of public school students.
- Among these districts, 64% use centralized assignment algorithms to manage student placement. Yet only 9% require all students to participate, as in New York and Boston, while the rest rely on opt-in systems, where parents must proactively submit applications.
- Compared to mandatory systems, opt-in districts are harder to navigate, with less clear information and more complex applications. As a result, economically disadvantaged students are less likely to enroll in choice schools in these districts.
These patterns raise a concern. Voluntary participation may segment public education by allowing the most advantaged students to leave their less advantaged peers behind. Building on these findings, the authors turn to the nation’s largest opt-in system to study how design shapes achievement and inequality.
Who Opts In?
If opt-in systems make participation voluntary, which families choose to participate? The authors turn to nearly two decades of data from the Los Angeles Unified School District (LAUSD), the largest opt-in system in the country. The data cover students’ applications to hundreds of magnet programs, affiliated charters, dual-language, and theme-based schools, linked to achievement and other outcomes.
Using these data, the authors document the following characteristics of applicants:
- Choice schools in LAUSD attract a student body that is more racially diverse but also more economically and academically advantaged. Applicants have substantially stronger academic preparation—the gap in baseline scores is equivalent to roughly two and a half grade levels—and are far less likely to be low-income or English-learners.
- Geographic distance is a primary driver of participation, as students who live closer to a choice program are much more likely to apply. Leveraging the rapid expansion of LAUSD’s choice sector, the authors show that when new programs opened nearby, neighborhood application rates rose by roughly 20%, without changing the observable characteristics of who applied.
These findings suggest that while advantaged students naturally sort into choice programs, supply-side policy, such as program expansion or reduced travel costs, can broaden access.
Who Benefits?
Building on these results, the authors turn to nearly two decades of randomized lottery data to examine which students stand to gain the most from attending a choice school. By comparing lottery winners and losers, the authors uncover a phenomenon known as “inverse-Roy” sorting: a pattern in which the people most likely to participate in a program are not those who benefit most from it, the opposite of self-selection based on comparative advantage , or negative selection on gains. Their findings include:
- Students who live closest to choice programs—a proxy for weak demand—benefit the most from attendance. The difference in impact between the nearest and farthest distance quintiles is equivalent to roughly one-third of a grade level of learning.
- Students from subgroups least likely to apply, such as low-income students and English-learners, benefit the most when they enroll.
Taken together, these results suggest that opt-in designs create allocative inefficiency: a situation in which resources, in this case seats at effective schools, are not allocated to the people who would benefit most from them . The students who would gain the most are precisely those the system screens out.
What If the System Were Redesigned?
To evaluate how alternative designs would perform, including for the 80% of students who never apply, the authors build a structural model using lottery offers and quasi-experimental variation from school openings.
The model first confirms that choice schools are genuinely effective, generating average achievement gains of roughly one full grade level in math and slightly less in ELA. The problem is not school quality, but rather it is that opt-in design prevents these benefits from reaching the students who need them most.
The authors then simulate a range of alternative policies including information interventions, busing-type subsidies to reduce travel costs, a decentralized market allowing multiple applications, and fully centralized assignment with mandatory participation. Their findings:
- Information interventions and reduced travel costs modestly increase the number of families who participate, but produce limited gains in student achievement across the district. This is because lowering barriers helps, but as long as participation remains voluntary, the same pattern of selection persists.
- Mandatory centralized assignment, or requiring all families to submit preferences through a clearinghouse, delivers the largest improvements. It eliminates two inefficiencies present in opt-in systems: negative sorting on gains (whereby students who would benefit most self-select out) and substantial slack (whereby seats at effective schools go unfilled). For students who enroll in choice schools, achievement impacts roughly double compared to the status quo.
- To benchmark these reforms, the authors simulate an achievement-maximizing allocation in which students are directly assigned to schools in a way that maximizes district-wide test scores. Centralization alone captures nearly 50% of these potential gains, and combining it with busing captures over 55%. The remaining gap reflects a fundamental limitation: Even with universal participation, families’ preferences do not perfectly align with school effectiveness.
This research offers two policy lessons. First, district-run choice programs, often created to compete with charter schools, are genuinely effective at raising student achievement, evidence that districts respond competitively when the K-12 landscape shifts. Second, layering voluntary choice onto neighborhood assignment deepens inequality. High application costs screen out students who would gain the most, while effective schools operate with substantial unused capacity. Expanding families’ options represents real progress, but that promise will be realized only when assignment mechanisms are designed to distribute opportunities equitably.