Research on pandemic learning loss has settled into an uncomfortable consensus: U.S. students lost ground measured in months of schooling, recovery began when buildings reopened, and by the 2023-24 school year, the most recent with full national assessment data available before mid-2026, progress had largely stalled, leaving average achievement below pre-2020 levels with the gaps widest for students in high-poverty schools. The findings come mainly from NWEA's MAP Growth assessments covering millions of students and from the Education Recovery Scorecard maintained by researchers at Harvard and Stanford.
This site publishes information about education research, not advice on tutoring or school choices for individual children.
How large were the losses?
Large by the standards of the field. The Education Recovery Scorecard analyses published in 2023 found the average student had lost roughly half a year of math learning and a quarter of a year in reading relative to the pre-pandemic trajectory, with districts in high-poverty communities losing a full year or more. The Department of Education's national assessment, NAEP, documented corresponding drops in its 2022 and 2023 releases, the largest declines in the math assessment's history at the national level. Because these are population averages, they conceal even wider spread between districts and between student groups.
What happened during recovery?
Recovery came in two phases. In 2021-22 and partially into 2022-23, students regained ground at above-normal rates as in-person instruction resumed, shrinking a meaningful share of the gap. Then the pace slowed. NWEA's analyses of 2022-23 and 2023-24 data, released in 2023 and 2024, found that improvements had largely flattened and estimated that the average student needed several additional months of schooling, in the range of four to five, to catch up in math and reading, with middle-grade students, whose foundational years coincided with closures, showing the least recovery.
Federal relief money, roughly 190 billion dollars across three pandemic aid packages, ended in September 2024, before the shortfall closed.
Which interventions actually worked?
Studies conducted as the money flowed point toward intensity and dosage rather than program labels. The strongest evidence in education research before the pandemic already favored high-dosage tutoring, multiple sessions weekly in small groups, and a 2021 meta-analysis by Nickow, Oreopoulos and Quan estimated large positive effects for tutoring programs of that shape. Summer learning, extended school years and targeted small-group support showed positive but smaller average effects. A prominent caution came from evaluations of software-based self-paced remediation, which generally underperformed expectations. The honest limitation: districts deployed aid during a labor shortage and enrollment volatility, so few programs were evaluated with randomized designs, and much of the confidence rests on prior evidence rather than fresh causal estimates of pandemic-era implementations.
How do we know any of this?
Three data systems make this unusually well-measured for education. NWEA administers adaptive MAP Growth tests to millions of students several times a year, allowing researchers to compare cohorts against 2019 baselines; its public reports through 2024 are the source of the stall findings above. The Education Recovery Scorecard, from the Center for Education Policy Research at Harvard and Stanford's Sean Reardon's group, harmonizes district-level test data to map recovery geographically. NAEP provides the slow, nationally representative anchor. Each has limits: NWEA's sample is students who test, and testing participation shifted after 2020; NAEP reported that lower-performing students were less likely to participate in its 2022 assessment, which muddies trend lines; score data say nothing directly about well-being or dropout, outcomes other studies found were also affected. Convergent findings across all three systems, despite different samples, is what gives the consensus its sturdiness.
Who is still furthest behind?
Every dataset tells the same story about distribution. Students in high-poverty districts, which were more likely to spend longer in remote instruction, lost the most and recovered the least, so inequality grew on both ends of the period. The Harvard-Stanford scorecard work in 2024 found that gaps between the highest- and lowest-income districts had widened relative to 2019, and NAEP's 2022-23 releases showed score declines concentrated among lower-performing students. Notably, a handful of districts, including some that intensified tutoring early, outperformed their pre-pandemic trajectories, a fact researchers have used to argue the losses were not inevitable.
What should be watched next?
By mid-2026, the outstanding questions are whether the 2024-25 assessment cycle, when published, shows any resumption of catch-up now that relief funds have ended, and what happens to the middle-school cohorts whose closures hit during foundational skill acquisition. Chronic absenteeism, which rose sharply and remained elevated in federal data through 2023-24, is a leading candidate explanation for the stall, though attribution is difficult. What the research has already established is more than enough: the losses were real and unequal, recovery was real and incomplete, and the tools with the best evidence were known before the emergency began.
Did subjects and grades recover equally?
No, and the pattern is consistent across datasets. Math, which fell furthest, initially recovered fastest, because math achievement responds strongly to instructional time and the subject sat at the center of many districts' tutoring investments. Reading recovery was slower and flatter, and the reading data carried a worrying cohort signature: the youngest students, those learning to read during closures, showed persistent deficits that later schooling had not closed as of the 2023-24 data. Middle school deserves separate mention, since adolescents experienced closures during a developmental window when absenteeism and disengagement compound. Subject-by-grade detail matters for interpreting averages, because a national mean that looks flat can conceal one grade band recovering while another slips, which is exactly what the district-level scorecard data showed.
For more context, read How dual enrollment changes college pathways.
For more context, read community colleges stem.
For more context, read Why science education emphasizes inquiry now.