Approximately 7.5 million U.S. students receive special education services in public schools—more than one in seven pupils nationwide. Its services cost roughly twice as much per pupil as general education and account for about one-third of federal appropriations for elementary and secondary education. Despite its scale, there has long been uncertainty about special education’s causal impact on student outcomes due to selection bias and methodological challenges. A new working paper from Boston University’s Wheelock Center (2026) helps fill this gap by applying rigorous causal methods to longitudinal administrative data to examine the impact of special education classification on student learning and educational progress.
The researchers used data on public school students in grades 3–8 from three states: Massachusetts, Indiana, and Connecticut. The analysis focuses on students first classified in grades 3–8 and primarily examines common disability categories such as specific learning disabilities, other health impairments, emotional disturbance, and speech or language impairments, rather than students with the most significant cognitive disabilities. The Indiana data spanned the 2011–12 through 2017–18 school years; the Connecticut data spanned the 2013–14 through 2019–20 school years; and the Massachusetts data spanned the 2007–08 through 2017–18 school years. Each data system used unique student identifiers to track students over time, and student demographic characteristics such as race/ethnicity, gender, and free or reduced-price lunch eligibility were observed in each state. The researchers employed an event-study design, comparing students to themselves before and after their special education classification, supplemented by a difference-in-differences approach that leveraged variation in the timing of classification.
The study revealed several notable findings. Most importantly, across all three states, students’ achievement trajectories declined prior to special education classification and rose sharply after they began receiving special education services. Gains in standardized math and English language arts test scores reached approximately 0.2 to 0.4 standard deviations within three years, suggesting substantial effects on learning. These effects were similar across disability types and student demographics. Taken together, this pattern suggests that special education services do not merely support students at a fixed level of performance, but instead help reverse ongoing academic decline.
Additionally, student gains did not occur all at once but increased over multiple years. This gradual pattern was consistent with sustained improvements in underlying skill development rather than short-term test score gains. The findings indicate that special education enhances the productivity of instruction over time, not just short-term performance.
The researchers also found that the benefits of these services were broad and not driven by testing accommodations. While critics have argued that accommodations create unfair advantages for high test scores, the evidence suggests otherwise. In Massachusetts—where data on testing accommodations are available—the researchers controlled for supports such as read-aloud and calculator use and found that the estimated gains remained nearly unchanged. This indicates that students demonstrated real improvements in learning, not just higher scores due to testing supports.
Importantly, the study’s methodological approach yielded more credible estimates than prior research by better accounting for pre-existing declines in student achievement. Traditional methods underestimated the relationship between these, potentially understating the effects of special education. By accounting for these trends in using the event-study design, the researchers were able to isolate the causal impact of special education services and show meaningful gains following classification.
Overall, these findings suggest that special education can produce meaningful learning gains when instruction and supports are tailored to students’ needs. However, because the study focuses on students identified in grades 3–8, the results may not generalize to students identified earlier. And while the analysis demonstrates substantial learning gains, it does not examine the costs relative to these benefits, leaving open questions about the return on investment. As a working paper, the findings would benefit from further peer review and replication, given mixed evidence in prior research.
From a policy perspective, the results suggest the importance of accurate and timely identification, expanded access to high-quality services, and more consistent service delivery so that qualifying students can benefit from these supports. At the same time, policymakers should recognize ongoing concerns about both over- and under-identification across student groups, including the possibility that some students who could benefit from services are not being identified.
SOURCE: Stephanie Coffey, Joshua Goodman, Amy Ellen Schwartz, Leanna Stiefel, Marcus A. Winters, and Yunee H. Yoon, “Special Education Substantially Improves Learning: Evidence from Three States,” Boston University, Wheelock Educational Policy Center (2026).
