U.S. student achievement was falling long before pandemic, study finds
A new analysis of student test scores reveals that American schools were in a “learning recession” for seven years before the COVID-19 pandemic, with student test scores in math and reading on a steady decline since 2013.
This reversal ended two decades of progress, according to Sean Reardon, the Professor of Poverty and Inequality at Stanford Graduate School of Education, whose data forms the backbone of the new research.
“From the early 1990s through 2013, public elementary and middle school students’ math and reading skills improved dramatically—by more than two grade levels in math, for example,” said Reardon, faculty director of The Educational Opportunity Project. “That shows that we can improve our public schools and equalize educational opportunity. But we haven’t been doing much of that for the last decade.”
This stall and reversal is one of the findings of the new report, “From Learning Recession to Learning Recovery: Understanding the Sources of U.S. K-12 Improvement,”released May 13 by researchers at Stanford, Harvard, and Dartmouth. The study reframes the narrative of pandemic-era learning loss, arguing that the crisis of the last few years was an acceleration of a problem that was already underway.
“The pandemic was the mudslide that followed seven years of erosion in student achievement,” said Professor Tom Kane, faculty director of the Center for Education Policy Research at Harvard University, and a lead author of the report. “The ‘learning recession’ started a decade ago... The recovery of U.S. education has begun. But it’s up to the rest of us to spread it.”
The study found that the slowdown in learning coincided with two major shifts in American childhood and education policy: the widespread dismantling of test-based accountability systems that defined the No Child Left Behind era and the rise of social media use among young people. Reading scores, in particular, suffered consistently, with the average annual loss in the years just before the pandemic being just as large as the loss during it.
This long-term decline challenges the notion that a return to 2019 performance levels is the ultimate goal, as the data suggests those levels were already part of a downward trend. Today, 8th-grade reading scores on national assessments are at their lowest point since 1990. Compounding the problem, chronic student absenteeism remains a major obstacle to improving learning. Though down from its pandemic peak, 23% of students were chronically absent in the 2024-25 school year, far above the pre-pandemic rate of 15%.
The findings are part of the fourth annual Education Scorecard. The report’s power comes from Stanford Education Data Archive (SEDA), a massive database constructed by the Educational Opportunity Project that links the test results of roughly 70 million students in grades 3 through 8 in 2009 through 2025 to a common national scale, allowing for direct, apples-to-apples comparisons of district performance across state lines.
U-shaped recovery
As schools work to climb out of this decade-long trough, the report reveals an unequal recovery since 2022. The most significant gains have been concentrated at the economic extremes – in the wealthiest and the poorest school districts.
The researchers suggest this "U-shaped" pattern is due to a divergence in resources. The highest-poverty districts, which suffered the most severe learning losses during the pandemic, were targeted with a massive infusion of federal relief funds. The researchers estimate these funds were the primary driver of their recovery. Without that aid, the average high-poverty district would have seen no academic improvement since 2022. Meanwhile, the wealthiest districts were able to draw on their own financial and social capital to support students.
Caught in the center are the nation’s middle-income districts—those where 30% to 70% of students receive federally subsidized lunches. These communities, often suburban or in smaller cities, had neither the concentrated poverty to qualify for the largest federal grants nor the deep private wealth to fill the gaps. As a result, they have experienced the slowest recovery, creating a new group of students at risk.
However, the report adds a crucial layer of context to the recovery in low-income areas. Despite making faster progress since 2022, the highest-poverty districts remain the furthest from their own pre-pandemic achievement levels, which were already far below those in more affluent districts. These high-poverty districts experienced catastrophic losses between 2019 and 2022, with students in the poorest schools falling an average of 0.7 grade levels in math. Their subsequent recovery has only made up a fraction of that loss.
Amid the challenging data, the report highlights signs of hope. One promising signal of a national turnaround in reading is linked to state-led "Science of Reading" initiatives. Every state that saw reading improvement between 2022 and 2025 – including Mississippi, Tennessee, and Louisiana – was implementing comprehensive, evidence-based literacy reforms. And in every state that had not systematically adopted such reforms, reading scores declined from 2022 to 2025.
Nonetheless, the researchers caution that this is far from conclusive evidence that simply adopting “Science of Reading” policies will lead to improvements in reading skills. “There’s no silver bullet,” noted Reardon. “We need more research to understand what specific teacher training, coaching, and reading instruction practices are most effective, and for whom, and under what conditions. And we need the same for math.”
The researchers also identified 108 “Districts on the Rise,” school systems of all income levels that are outpacing their peers.
“Our hope is that people will learn from these states and districts and use them as models for improving our schools,” Reardon said.
Top photo by Jacob Wackerhausen/iStock
Video by Ranya Adkinson
Reardon is also a faculty affiliate of the Stanford Accelerator for Learning and a senior fellow at the Stanford Institute for Economic Policy Research (SIEPR). Daniel C. Dewey of Harvard University, Erin Fahle of Stanford, and Douglas O. Staiger of Dartmouth College, are co-authors of the report.
The Education Scorecard receives philanthropic support from Citadel founder and CEO Ken Griffin and Griffin Catalyst, Bloomberg Philanthropies, Carnegie Corporation of New York, and the Joyce Foundation. The data used in the analyses come from the Stanford Education Data Archive (SEDA). SEDA is funded by the Gates Foundation. Some of the raw data used in constructing the SEDA files were provided by the Education Data Center (EDC), the National Center for Education Statistics (NCES), and the National Assessment Governing Board (NAGB).
The findings and opinions expressed in the research reported here are those of the authors and do not represent views of EDC, NCES, NAGB or any of the aforementioned funders.
Faculty mentioned in this article: sean reardon
