Wide-ranging education reforms in New York City Public Schools in the early 2000s under then-mayor Michael Bloomberg and schools chief Joel Klein ran the gamut from school closures and consolidations to revamping academic offerings and giving families additional choices within the district. The overarching goal was to boost long-languishing student achievement. A new working paper examines which reforms worked and extracts some vital lessons that other districts might heed.
A trio of economists divide the various reform efforts into categories derived from economics and computer science terms—“reallocations” and “resource augmentations”—that are precise but not necessarily intuitive. Reallocation reforms include changes to New York’s school assignment algorithms and increased information distributed to families regarding school options, intending to make it easier for students to find the best academic “fit.” Resource augmentation reforms include the creation of 200 small high schools in the city and the simultaneous closure of larger, more traditional schools that were underperforming. The clearest way to differentiate the interventions is that reallocation keeps most or all schools open but allows students to move around more freely, while resource augmentation closes schools to effectively eliminate bad choices and create better options. Students still move, but it is a less voluntary proposition and all remaining schools are, theoretically, better options.
The researchers’ first step is to see how the real-world reforms impacted students. Data come from the NYC DOE administrative information system and cover students enrolled from the 2003–04 through the 2012–13 school years. Specifically, they focus on four cohorts enrolled in eighth grade between 2003–04 and 2006–07. They follow each cohort through high school and examine middle school test scores, high school enrollment, New York Regents test scores, high school graduation, and college attendance within two years of their expected high school graduation date.
School quality is measured in two ways: (1) an ordinary least squares value-added model, and (2) an estimate of student preferences for specific schools based on ranked lists submitted as part of their high school applications. The first approach controls for observable student characteristics; the second accounts for unobserved factors that influence student choice such as school culture and ease of transportation. The researchers also create what they call counterfactual impact estimates, which simulate student outcomes under various idealized reform scenarios.
The report is wonky, but the bottom line is that the most ideal student–school matches achieved by reallocation generated only modest educational gains. For example, in a model designed to represent perfect choice—with each hypothetical child assigned to the ideal school for their needs—the impact of reallocation reform was just a 3.8 percentage point increase in both high school graduation and college attendance. That’s not much—and it’s also hypothetical. In the real world, such ideal student–school matches rarely occurred in the NYC data, largely because available seats at the most effective (and desired) schools are limited. In other words, making it easier in theory to access a coveted high school yields benefits for those who get in, but the reality is that very few additional students actually achieve such access.
By contrast, resource augmentation policies appear to provide as much or more positive impact for students with less systemic disruption. The terminology used makes these findings feel counterintuitive—how can closing a number of schools, opening a bunch of new ones, and making kids move be less disruptive?—but the concept makes more sense when the explanation is simplified. Closing lower-performing schools and making sure that impacted students move to higher-performing ones is a boon for student achievement. Whatever systemic disruptions do occur are more than offset by widespread student academic gains. In New York, closing 14 percent of the city’s underperforming high schools led to gains predicted to be comparable to the most idealized version of the perfect reallocation model.
Of course, getting kids into “replacement” schools that are better than the ones being closed is not an exact science. The researchers use a sophisticated value-added calculation in their impacts modeling, but also run similar analyses that suggest a simple look at graduation rates is nearly as effective in determining school quality. Whether graduation rates are an effective measure of student learning remains a question for another time (color all of Fordham skeptical on that score). Regardless of what criteria are agreed upon to determine school quality, there will axiomatically be a set of schools at the bottom of every ranking. By starting with the worst schools on the list when closures become the preferred reform, you come as close as possible to guaranteeing that impacted students will find a better option.
In the end, all this complexity boils down to a few simple, albeit highly generalized, points. One is that systemic reforms in school districts should be aimed at giving more students better access to the best possible schools, but shuffling students around is limited by the effectiveness of the lowest-performing options and by capacity restraints in better schools. Another is that if you don’t have enough space in existing high-performing schools, the easiest fix is to start closing the worst of the worst and investing in better ones.
SOURCE: Atila Abdulkadiroglu, Parag A. Pathak, and Christopher R. Walters, “Who Gets What in Education: Can School Matching Improve Student Achievement?” NBER Working Paper (March 2026).
