Dewey, Fahle, Kane et al., "From Learning Recession to Learning Recovery: Understanding the Sources of U.S. K-12 Improvement," Education Scorecard (May 2026)
Summary
Education Scorecard report From Learning Recession to Learning Recovery, dated May 2026, by Daniel C. Dewey, Erin Fahle, Thomas J. Kane, Sean Reardon and Douglas O. Staiger of Harvard, Stanford and Dartmouth. It argues the United States entered a learning recession in 2013, and that average annual reading losses in 2017–19 were as large as during the pandemic, with Grade 8 NAEP reading at its lowest point since 1990. Since 2022 the recovery has been U-shaped: middle income districts, with 30 to 70 percent of students receiving subsidized lunches, improved least. The authors attribute most of the recovery in the highest poverty districts to federal pandemic relief. They also examine absenteeism, federal school accountability and state early literacy reforms.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Dewey, Fahle, Kane et al., ‘From Learning Recession to Learning Recovery: Understanding the Sources of U.S. K-12 Improvement,’ Education Scorecard (May 2026).”
Original at educationscorecard.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/dewey-fahle-kane-et-al-from-learning-recession-to-learning-recovery-un-c21e4d31b0/
, SHA-256 7d1fededbdd06911e68080f78b5691f53e02a2828b9d9bac6f691b6a30f3ed9d.Full text
May 2026
From Learning Recession
to Learning Recovery:
Understanding the Sources of U.S.
K-12 Improvement
Daniel C. Dewey, Harvard University
Erin Fahle, Stanford University
Thomas J. Kane, Harvard University
Sean Reardon, Stanford University
Douglas O. Staiger, Dartmouth College
We would like to thank Emily Oster and Clare Halloran of the Education Data Center, Josh Bleiberg,
the Foundation for Excellence in Education (ExcelinEd), the National Assessment Governing
Board, the U.S. Department of Education, and Nat Malkus at the American Enterprise Institute
for providing data used in this report. For research assistance, we thank Derik Suria and Micaela
Keating Sharp at Harvard. Andrew Ho, Benjamin Shear, Jim Saliba, Jie Min, Jiyeon Shim and
Demetra Kalogrides contributed to the development of SEDA with research assistance from Sadie
Richardson, Sofia Wilson, Julia Paris, Ishita Panda, Amelia Bloom, Nahian Haque, Jackson Kinsella,
and Aditi Oak at Stanford University. Scott Sargrad and Adam Schott provided comments on an
earlier draft. We also want to thank our communications partners, Waypoint Education Partners
and Marathon Communications, for their efforts to amplify these findings. The research was
supported by grants from the Carnegie Corporation of New York, Bloomberg Philanthropies, Joyce
Foundation, and Kenneth C. Griffin and Citadel Catalyst. The Stanford Education Data Archive
received separate funding from the Gates Foundation. The opinions expressed here are ours and do
not represent views of NCES, the U.S. Department of Education, or any of the funders.
From Learning Recession to Learning Recovery | 1
I. Introduction
We launched the Education Recovery Scorecard in 2022 to track the pace of post-pandemic
recovery in local communities around the country. The hunger for comparable, timely estimates
of local achievement trends was evident in the more than 5,000 media hits that the Scorecard
received in its first four years. Last year, media outlets in 34 states wrote about the progress and
struggles of students in their own communities. Now that the federal relief dollars are gone, we
are dropping the “recovery” from our name, but we will continue to examine the national data to
shed light on contemporary policy debates while also highlighting local districts that are leading
the way forward.
In this report, we describe the “learning recession” that started in 2013, when math and reading
improvement stalled and U.S. students’ average test scores began declining. We explore factors
affecting the recovery since 2022: federal pandemic relief, lingering student absenteeism, and
federal school accountability. We also provide initial evidence on state-led early literacy reforms,
by comparing state reading gains against their implementation of evidence-based reading reforms.
Specifically, we find:
• The United States entered a “learning recession” in 2013, as student progress in math and
reading stalled and achievement began to decline. In reading, the average annual loss in
achievement immediately before the pandemic (2017–19) was just as large as during the
pandemic (2019–22). Grade 8 reading scores in NAEP are now at their lowest point since
1990 and Grade 4 is at pre-2003 levels.
• The slowdown in learning coincided with a dismantling of test-based accountability and a
rise in social media use. It is not clear which was most influential. At the least, the phasing
out of test-based accountability made the pre-pandemic losses harder to recognize.
• Since 2022, the recovery has been “U-shaped,” with larger improvements among the
highest and the lowest income school districts in the country. Middle income districts—with
between 30 and 70 percent of students receiving federally subsidized lunches—have seen
the least improvement on average. The recovery in achievement in the highest poverty
districts seems largely due to the federal pandemic relief. Without federal relief, the
average high-poverty district would have remained at 2022 levels of achievement.
• After the pandemic, math achievement rebounded immediately, with the annual rate of
improvement returning to pre-2013 levels in 2022–24. In reading, however, achievement
continued to decline through 2024. In 2025, we see the beginnings of a turnaround in
reading. The incipient recovery in reading appears to be related to state early-literacy
reforms. All of the states which improved in reading between 2022 and 2025 were
implementing comprehensive “science of reading” reforms (Maryland, Louisiana, Tennessee,
Kentucky, Indiana, Mississippi, Minnesota, District of Columbia). None of the states which
had eschewed literacy reforms as of January 2024 improved in reading between 2022
and 2025 (Massachusetts, California, Washington, New Hampshire, Georgia, Wisconsin,
2 | From Learning Recession to Learning Recovery
Rhode Island, New Jersey, Hawaii, South Dakota). Nevertheless, many states that were
implementing multiple elements of science of reading reforms have yet to turn around (e.g.,
Florida, Arizona, and Nebraska). Evidence-based reading reform may be a necessary but
insufficient path to improvement.
• High student absenteeism continues to be a headwind for academic recovery. If student
absences had returned to pre-pandemic levels, the recovery would have been meaningfully
larger (.03 to .05 grade equivalents) for districts at all income levels.
• There has been dramatic variation in the pace of recovery even among districts with
similar student characteristics. As a result, we have identified 108 districts with large
improvements in both subjects (and a total of 448 districts with large improvements in
either subject). We highlight these “districts on the rise” in the hope that they can serve as
examples of what is possible.
Recommendations
Our findings point to four recommendations for education leaders:
1. Now that the federal pandemic relief has expired, states should consider schools’
achievement losses since 2019 when identifying schools for comprehensive support and
improvement. The federal relief dollars were heavily concentrated in the lowest income
districts (with more than 70 percent of students receiving federally subsidized lunches),
while higher income districts (those with less than 30 percent of children receiving lunch
subsides) had greater financial and social capital to draw on. But many middle-income
school districts—those with 30–70 percent of students receiving federal lunch subsidies—
received little federal aid and remain far behind 2019 levels of achievement.
2. Lowering student absences remains a priority. If absence rates had returned to pre-
pandemic levels, recovery would have been meaningfully faster. Moreover, getting students
back into the habit of reliable attendance will continue to pay off for years into the future.
3. It is crucial to learn more about the factors that have led to declining scores and the
potential different strategies hold for improving student performance. To that end, the
federal government should support research in three areas:
a. The Role of Social Media: The decline in achievement which preceded the pandemic
was likely partially due to social media exposure. In the next year, researchers
around the country will be reporting on the impact of various cell phone bans. The
federal government should coordinate efforts to reach consensus and reconcile
any differences in findings. Early results suggest positive—but small—impacts on
student achievement. If further research confirms this, we should be evaluating new
approaches to reducing cell phone use in school and social media use outside of
school.
From Learning Recession to Learning Recovery | 3
b. Early Literacy Reforms: As of March 2026, 42 states and the District of Columbia
have passed laws or policies related to evidence-based reading instruction—often
referred to as the “science of reading” (Schwartz, 2026). But not every state’s
efforts are succeeding. The federal government should provide funding to evaluate
specific aspects of state policies—such as literacy coaches or third grade retention
policies—and share lessons learned.
c. Lowering Absenteeism: Although there is evidence on effective methods for
lowering absenteeism, most of those effects are small. The federal government
should support and evaluate new approaches to lowering absenteeism.
4. We hope that states will pair each of the 108 “Districts on the Rise” with one or more
similar districts in their state, to compare notes on recovery strategies. A small investment,
such as paying travel expenses for teams to visit in person in the fall, could yield large
dividends if districts successfully share ideas for improvement.
II. Data
Achievement data: The achievement data in this report come from the Stanford Education Data
Archive (Reardon et al., 2026a, 2026b), produced by the Educational Opportunity Project at Stanford
University. SEDA includes district-level test score data from 2009 through 2025 in math and
reading. The SEDA data for 2022-2025 rely on district-level achievement data collected by the
Education Data Center (Education Data Center, 2025). The estimates are constructed by converting
the shares of students in each state-defined proficiency category to estimates of the average test
score in each district. Because tests and proficiency definitions differ across states and frequently
shift within states over time (as states change their tests and proficiency definitions), SEDA
uses the federal National Assessment of Educational Progress (NAEP) to align state test scores
to a common scale. To make the scores more easily interpretable, they are rescaled to grade
level equivalents, using the 2019 NAEP as a referent.1 A one-grade level improvement is roughly
equivalent to the average amount of improvement students make from one grade to another
between 4th and 8th grade.
Unfortunately, the spring 2025 state test administrations used in this report fell between the 2024
and 2026 NAEP administrations. As a result, the 2025 state test scales cannot be linked to NAEP
if the tests changed between 2024 and 2025 (e.g., Illinois, Kansas). Moreover, high opt-out rates in
some states (e.g., New York, Colorado) make it impossible to compare the population of students
taking state tests to the sample used for NAEP estimates. Additionally, some states do not publish
sufficiently detailed data to enable the estimation of district-level average scores. Thus, the 2025
SEDA are limited to the 40 states where tests and proficiency levels were reportedly unchanged
from 2024 and where adequate data were available (for more on the SEDA methodology, please see
Fahle et al 2026a; 2026b). Next spring, when the 2026 NAEP is released, SEDA will add additional
1 Roughly speaking, we define one “grade level” as one-fourth the difference in average 4th and 8th grade NAEP scores
in 2019. See Fahle et al., 2025 for details.
4 | From Learning Recession to Learning Recovery
states and will include updated 2025 estimates to reflect the changes in state NAEP scores
between 2024 and 2026.
Average days absent: To measure student absences, we rely on district-level chronic absenteeism
rates collected by Nat Malkus at the American Enterprise Institute (AEI) for the school years 2016–
17 through 2024–25.
Using data from two states which report both chronic absence and average absence rates,
Massachusetts and Florida, we find a strikingly consistent linear relationship between the chronic
absence rates and average absence rates measures. The same relationship holds both before
and after the pandemic (see Appendix B for details). Each 10-point increase in chronic absence
corresponds to a 1.8-point increase in average absence rates. Therefore, we use the chronic
absence rates to estimate an average absence rate for each district. We also calculate the average
number of days absent by multiplying by 180.
Federal Pandemic Relief: The first package of federal pandemic relief for schools (ESSER I, $13.2
billion) was to be obligated by districts by the end of September 2022. Because we are investigating
achievement gains between spring 2022 and spring 2025, we focus on the second two packages,
ESSER II ($54 billion) and the American Rescue Plan or ESSER III ($122 billion). We use the total
district allocation under ESSER II/III as an estimate of the increase in spending in the district after
spring 2022 and before spring 2025. The proxy may overstate the increase in spending if some of the
ESSER II dollars may have been spent before spring 2022, or if local government offset some of the
federal relief by reducing local revenues for education. (It could also understate spending if some
districts had ESSER I dollars leftover in spring 2022.) The American Rescue Plan required states to
maintain their grants to school districts.
Federal accountability: We used data reported by the U.S. Department of Education for the number
of schools failing to make “adequate yearly progress” (AYP) under No Child Left Behind (NCLB)
for the years 2002–2003 through 2004–2005. For 2005–2006 through 2010–2011, we relied on AYP
data reported by the Center on Education Policy (CEP). CEP researchers originally collected this
data through state department of education websites, direct communication with state education
personnel, and State Consolidated Performance Reports submitted to the U.S. Department of
Education. Note that for school years ending in 2003, 2006, 2007, and 2008, the precise number of
schools failing to make AYP was not available. We instead estimate the number of schools failing
to make AYP using reported percentages. For years 2011–2012 through 2014–2015, we used school
improvement status data from EDFacts to get the number of schools failing to make AYP and
schools otherwise designated for improvement under federal waivers.
To determine the number of schools designated for “comprehensive support and improvement”
(CSI), “targeted support and improvement” (TSI), or “additional targeted support and improvement”
(ATSI) under the Every Student Succeeds (ESSA) Act, we used data from EDFacts for 2018–2019
and 2019–2020. Unfortunately, EDFacts only reported such data for these two school years. Josh
Bleiberg at the University of Pittsburgh generously provided us with data assembled from state
education agencies, identifying schools designated for CSI or TSI for 2017–2018. Bleiberg also
provided us with data on schools designated for CSI in 2021–2022 or 2022–2023, though data on
From Learning Recession to Learning Recovery | 5
schools designated as TSI was not available for these years (note that we combine data from the
2021–2022 school year into the 2022–2023 school year, as improvement statuses for these two years
were mutually exclusive). For data on the 2023–2024 school year, we received accountability data
from the U.S. Department of Education.
Federal accountability designations are made at the school level. As a result, we calculated
the percentage of all public school students in grades 3–8 who attend a school with the above
accountability designations and use this percentage to study the expected district-level effects.
State early literacy policies: The Foundation for Excellence in Education (ExcelinEd) tracks state
implementation of 18 elements of early literacy policies, such as training for K-3 teachers and
administrators in evidence-based reading instruction, universal screeners to identify students with
reading challenges, literacy coaches in elementary schools, and mandatory retention policies.
We used the Internet Archive to download ExcelinEd’s State-by-State Policy Implementation
Reports for each state for January 2024. As a proxy for the comprehensiveness of a state’s literacy
reforms, we used the count of the number of policy elements that were either in full or partial
implementation, as designated by ExcelinEd.
Pre-Pandemic Improvement and the Learning Recession
There is a widespread belief that student outcomes in the United States have been unchanged for decades.
However, that narrative is false. In Figure 1, we report the trend in scores on the NAEP in 4th and 8th grade
math and reading between 1990 and 2024.2 From the pre-No Child Left Behind Act (NCLB) minimum to the
post-NCLB peak, mean math achievement in 4th and 8th grade improved by the equivalent of 2.8 grade equiv-
alents and 2.2 grade equivalents, respectively. Many individual states, such as North Carolina, experienced
even larger improvements.
Reading achievement also improved, although less than in math. Reading achievement on the NAEP im-
proved by 9 points in 4th grade and 8 points in 8th grade, respectively, from the pre-NCLB period to the post-
NCLB period, equal to almost a full grade equivalent.
2 The NAEP also reports mean 12th grade achievement in math and reading, which has been flat. However, because of
the rise in high school graduation rates, more students who would have been dropouts in years past are taking the 12th
grade test in recent years. As a result, 12th grade achievement likely understates progress. Doty et al. (2025) find that
despite flat 12th grade scores, those born in states with large improvements in 8th grade math achievement enjoyed
higher earnings and educational attainment and lower teen motherhood, arrests, and incarceration.
6 | From Learning Recession to Learning Recovery
FIGURE 1. NATIONAL NAEP TRENDS, 1990 TO 2024
Note: Estimates are from National Center for Education Statistics. (n.d.). NAEP Data Explorer. U.S. Department of
Education, Institute of Education Sciences. We convert the difference in pre-NCLB minimum to post-NCLB maximum
using a grade equivalent of 10.25 for math and 10.625 for reading.
Skeptics have been concerned that the improvement in test scores in response to test-based
accountability was a result of “teaching to the test,” in which teachers focus on test-taking skills
and strategies rather than students’ content knowledge. However, such concerns would apply
primarily to state tests used for school accountability, not the NAEP. The NAEP is a low-stakes
test, and teachers generally do not know the content or format of NAEP tests (nor whether their
students will be sampled and assessed in any NAEP wave).
Moreover, Doty et al. (2025) find that those born in states with large increases in NAEP
achievement had larger increases in earnings and educational attainment and bigger declines
in teen motherhood and incarceration than those born in states with smaller increases. In fact,
the differences were nearly as large as one would expect based on the prior research on the
relationship between earnings and test scores. So, it would appear that improvements in NAEP
scores reflected real student learning that benefited students in the long run, not just score
inflation.
From Learning Recession to Learning Recovery | 7
Annual Rates of Improvement
In Figure 2, we report the annual rate of improvement at the district level during different periods.
Like a price index, the NAEP measures the level of student knowledge at a point in time. However,
the annual rate of improvement in achievement is more analogous to GDP growth. As with the
economy, when the rate of growth in learning shifts from positive growth to negative growth, it
implies we have entered a “learning recession.”
Between 2009–11 and 2011–13, for instance, math and reading achievement in grades 3–8 improved
by .05 to .1 grade equivalents per year on average. At that rate, students would have improved
between a half and a full grade level over a decade. However, as reported in Figure 2, there
was a sharp downturn in the rate of improvement after 2013. Between 2013 and 2015, the rate of
improvement became negative in mathematics and was essentially zero in reading. Although the
pandemic seemed to hasten the decline in math, the annual rate of decline in reading was similar
in the period before the pandemic (2017–19), during the pandemic (2019–22), and after the pandemic
(2022–24).
8 | From Learning Recession to Learning Recovery
FIGURE 2. ANNUAL RATE OF IMPROVEMENT IN ACHIEVEMENT 2009–2025
Note: District level average annual change in math and reading achievement using data from the Stanford Education
Data Archive.
From Learning Recession to Learning Recovery | 9
Explanations and Correlates of the Learning Recession
There are two potential explanations for the deceleration in achievement beginning in 2013: a
decline in test-based accountability and the rise in social media. We discuss each below.
The role of test-based accountability in pre-pandemic trends
Prior research (Carnoy and Loeb, 2002; Hanushek and Raymond, 2005; Dee and Jacob, 2011) has
found that achievement improved faster in states that had begun implementing test-based
accountability during the 1990s and early 2000s. The improvements continued after the federal No
Child Left Behind Act extended test-based accountability nationally in 2003.
In the top panel of Figure 3, we report the number of schools identified for intervention under
federal accountability requirements between 2003 and 2025. Between 2003 (the start of federal
accountability under NCLB) and 2007, about 25 percent of schools were identified annually as
failing to meet “adequate yearly progress” (AYP). Schools that failed to make AYP faced escalating
sanctions for each year that they remained in that status, the most extreme being replacement of
school leadership, conversion to a charter school, or state takeover.
NCLB envisioned that states would reach 100% proficiency by 2014 and most states backloaded
their improvement goals closer to 2014 (Center on Education Policy, 2008) . This led to the
thresholds for adequate progress rising over time. As a result, the percentage of schools failing to
make AYP rose from 25 percent in 2007 to over 40 percent in 2011.
Beginning with the 2012–13 school year, the Obama administration began granting waivers to states,
allowing them flexibility from NCLB accountability requirements in exchange for state adoption of
approved reform plans. Although the waivers were not official until 2012–13, many states stopped
identifying schools not making AYP in 2012, in anticipation of the waivers. Thirty-eight (38) states
received waivers for the 2012–13 school year, with six more states receiving waivers in the next two
years. As a result, the number of schools identified as failing to meet “adequate yearly progress”
declined to less than 10 percent of schools by 2014. Le Floch et al. (2025) showed that the lion’s
share of the decline was in the states receiving NCLB waivers. The states with waivers started
identifying low-performing schools as “focus” or “priority” schools, but the combination of schools
not making AYP or being designated as focus and priority schools settled at roughly 15 percent of
schools, far lower than in 2011 or before.
In December of 2015, U.S. Congress replaced the No Child Left Behind Act with the Every Student
Succeeds Act (ESSA). While AYP designations had been based on achievement scores alone, states
were required to design their own accountability measures , combining test-based measures with
non-test-based measures, such as chronic absenteeism, survey measures of school climate and
student engagement, and postsecondary enrollment.
10 | From Learning Recession to Learning Recovery
FIGURE 3: TRENDS IN SCHOOL ACCOUNTABILITY SANCTIONS AND MEDIA
COVERAGE
Note: Federal waivers from 2013–2015 allowed states to opt out of AYP requirements in exchange for alternative
accountability systems. Schools not meeting alternative requirements under waivers are included in this
percentage.
Note: Based on Factiva searches of 84 major news publications/sites. An example search term for NCLB is
“adequate yearly progress” and for ESSA is “comprehensive support and improvement.”
From Learning Recession to Learning Recovery | 11
States were not required to report accountability designations in the 2015–16 or 2016–17 school
years, as they transitioned from NCLB to ESSA. When accountability resumed in 2017–18, the
number of schools identified for the most serious interventions, “comprehensive support and
improvement” (CSI), was roughly 5 percent of schools, with another 10 percent designated for
“targeted support and improvement” (TSI).3 (Additionally, states were given more flexibility on the
consequences of being designated CSI or TSI) (Meyers et al., 2023).
During the beginning of the pandemic, every state suspended its school accountability systems as
state testing was cancelled in spring 2020. Thirty-one states resumed identifying CSI, TSI, and ATSI
schools in the 2021–22 school year (based on spring 2021 scores). The remaining states began their
second cycle of federal designations in the 2022–23 school year.4 About 5 percent of schools were
identified for CSI, while about 15 percent of schools were identified as TSI or ATSI by 2023.
The number of schools identified for intervention may have declined after the NCLB waivers began,
but the public salience of test-based accountability declined even more. In the bottom panel of
Figure 3, we report the number of news articles mentioning federal accountability designations
(AYP or CSI, TSI, or ATSI) in major U.S. news publications and websites. While the annual number of
schools identified for intervention under federal accountability declined by about 30% from 2003–
2012 to 2018–2024 (from 28,320 per year to 19,871 per year for years in which we have complete data),
the number of mentions in major news outlets declined by even more (97%) during that time period
(from about 304 per year to 11 per year).
Thus, the decline in student achievement in the United States corresponded with a decline in the
number of schools identified for intervention under federal test-based accountability as well as a
decline in the public prominence of test-based accountability, as reflected in major news media.
However, it is difficult to establish a direct connection between the decline in achievement and
the de-emphasis of test-based accountability. The decline in achievement after 2013 was similar
in the states that had not received waivers from NCLB as in the states receiving them. While the
waivers may have reduced pressure on schools to raise student achievement, accountability was
also breaking down in the states where NCLB remained in place. One reason for the breakdown
was that under NCLB, schools were expected to have 100 percent of students proficient in math
and reading by 2013–14. When nearly every school would be designated a failing school, failure had
lost its stigma.
Moreover, the most serious consequences of failing AYP did not start until schools had failed
for multiple years. Had expectations for schools remained at the same level as in 2010—and not
escalated to the levels required by No Child Left Behind—and if the Obama administration had not
granted waivers to states to reduce test-based accountability pressures, it is possible that the rate
of improvement would have continued beyond 2013. However, that hypothetical has never been
tested.
3 The five percent designated for CSI was no more nor less than ESSA required.
4 The U.S. Department of Education required all states to begin identifying CSI/TSI/ATSI schools in the 2022–23 school
year based on spring 2022 assessments. Forty-one states received waivers to delay or modify their accountability
designations in the 2021–22 school year (Ujifusa, 2021).
12 | From Learning Recession to Learning Recovery
The role of social media in pre-pandemic trends
Another hypothesis to account for the learning slowdown after 2013 is the rise of social media.
The most prominent proponent of the connection between social media use and academic
achievement is Jonathan Haidt of New York University. In his book The Anxious Generation
and in public presentations, he argues that the timing of the decline in achievement in 2013–15
corresponds with a sharp rise in teen social media use. According to the Pew Research Center, the
share of U.S. teens aged 13 to 17 reporting that they were online “almost constantly” increased from
24 percent in 2014–15 to 46 percent by 2022 (Pew Research Center, 2022).
Other than timing, there are at least two additional pieces of evidence which suggest that social
media may have played a role in the learning recession in the United States. First, the losses were
concentrated among lower achieving students, who are the heaviest users of social media. In
Figure 4, we report the average score on the Programme for International Student Assessment
(PISA) math exam for students reporting different amounts of time browsing social media per day.
For the 8 percent of U.S. students self-reporting to browse social media for more than seven hours
per day, mean achievement on the PISA test was 80 points below (.8 standard deviations) those
of students browsing for 1–3 hours per day. The relationship is similar in other PISA participating
countries. The fact that heavy users of social media are lower scoring is not necessarily causal.
However, if social media does have a negative impact, we would expect it to impact lower-
achieving students more. It is telling, then, that is where the decline occurred.
FIGURE 4. SOCIAL USE IS MORE PREVALENT AMONG LOWER-SCORING STUDENTS
Note: Based on authors’ tabulation of the 2022 Programme for International Student Assessment. In the PISA sample,
7% of students reported no time browsing social media, 20% less than an hour, 36% 1–3 hours, 21% 3–5 hours, 10% 5–7
hours and 7% more than 7 hours. The percentages were similar in the United States.
From Learning Recession to Learning Recovery | 13
Second, the gaps between 90th and 10th percentile students on the PISA exam started expanding
in other countries around the same time as in the United States. In Figure 5, we report the trend
in the gap between students at the 90th and 10th percentile in the United States and Canada. In the
United States, math achievement gaps were narrowing prior to 2012, while gaps were essentially
unchanged in Canada. In reading, the U.S. gap was narrowing through 2015, while growing slightly
in Canada. However, after 2015, achievement gaps were widening in both countries. Especially in
reading, the same widening between top and bottom scorers was occurring in many industrialized
countries, including Austria, Finland, Denmark, Germany, Korea, the Netherlands, and Sweden.
It would be hard to argue that the widening in Netherlands was due to the decline in test-based
accountability in the United States.
FIGURE 5. ACHIEVEMENT GAPS IN THE UNITED STATES AND CANADA WIDENED
AFTER 2013
Source: Programme for International Student Assessment.
Much of the research on exposure to social media has focused on student mental health outcomes
rather than test scores (Arenas-Arroyo et al., 2022; Braghieri et al., 2022; Churchill & Johnson, 2026;
Donati et al., 2025; Golin, 2022; McDool et al., 2020; Pugno, 2025). In reviewing research from the
United States and Europe, Pugno (2025) concludes that social media does have negative impacts
on mental health, especially among teenagers. It seems plausible that the decline in mental health
spilled over to a decline in academic achievement as well. In one of the few studies focusing on
achievement directly, Jain and Stember (2025) tracked academic achievement as 3G cellular rolled
out in 82 countries around the world and found that math achievement declined as children gained
access to broadband cellular services.
One way social media could impact student achievement is by distracting students in class.
Thus, if achievement were to rise sharply following a cell phone ban, it would imply that social
media had played a role in the original decline, operating through distractions in class. However,
thus far, we have little evidence on the effect of cell phone bans. Beland and Murphy (2016)
identified high schools in four English cities that had implemented cell phone bans and found a
modest improvement in achievement (about one-fifth of a grade equivalent) for those schools
14 | From Learning Recession to Learning Recovery
implementing bans, primarily among lower-scoring students. Using a similar research design
with high schools in Sweden, however, Kessel et al. (2020) found no impact of cell phone bans on
student achievement or GPA, even for lower-achieving students. Abrahamsson (2024) found that
cell phone bans in Norway had positive impacts on student grade point averages (GPA), but only for
girls. The impacts were generally insignificant for boys.
In the United States, at least 37 states and the District of Columbia have restricted the use of cell
phones in schools as of 2026. So far, there are two major studies of the impact of cell phone bans in
the United States. Figlio and Ozek (2025) found positive effects of a cell phone ban on student test
scores in a large Florida district. However, the effects were quite small—roughly one-tenth of a
grade equivalent—and did not appear until the second year after implementation.5 The authors also
reported an unintended consequence: a rise in disciplinary actions for African American students
in the first year of the ban. The second study’s estimates are largely consistent: Allcott et al. (2026)
found that disciplinary incidents rose when a national sample of schools implemented cell phone
bans using lockable pouches (Yondr bags). Although disciplinary incidents subsequently fell, they
found no impact on student achievement or attendance.
Thus, while the timing of the decline in achievement corresponded with a decline in test-based
accountability and a rise in social medial use, there is still no direct evidence that either was the
cause of the decline in U.S. achievement after 2013. Moreover, if social media did play a role, it
seems likely that the mechanism was not just through students being distracted by cell phones
during instructional time, but a combination of factors, including interrupting students’ sleep,
reducing reading time outside of school, increasing social anxiety, weakening executive function, et
cetera.
Poverty and Academic Recovery 2022–25: The U-Shaped Recovery
In Figure 6, we describe average pandemic achievement losses in math and reading by district
poverty status. Even in the districts with the lowest poverty rates, those with less than 10 percent
of students eligible for federal subsidized lunch, the average student lost .3 grade equivalents in
math and .1 grade equivalents in reading over the course of the pandemic. However, the losses
were roughly twice as large for the highest poverty districts. In districts with more than 80 percent
of students eligible for federal subsidized lunches, the average loss was more than .7 grade
equivalents in math and .4 grade equivalents in reading.
5 Figlio and Ozek (2025) report an effect of .622 percentiles on a nationally normed test of math and reading
achievement. We converted to standard deviation units by dividing by 21.06 (the factor used to convert standard
deviation units into normal curve equivalents). The result, .029 standard deviations, is roughly one tenth of a grade
equivalent.
From Learning Recession to Learning Recovery | 15
FIGURE 6. LOSS IN ACHIEVEMENT BY DISTRICT POVERTY: 2019–2022
Note: Mean change in achievement from the Stanford Education Data Archive, weighted by district enrollment size.
In previous reports, we found that losses were larger in districts with longer school closures
(especially when combined with high poverty rates) and in communities with larger disruptions in
economic and social activities (Fahle et al., 2025).
For the remainder of the paper, we focus on the factors shaping the academic recovery between
2022 and 2025. Figure 7 reports the average change in achievement between 2022 and 2025 in
math and reading by school poverty. In the average high-income district—those with fewer than 10
percent of students receiving federally subsidized lunches—average math achievement improved
by .25 grade equivalents between 2022 and 2025.6 Their reading achievement did fall between 2022
and 2025, but by less than middle income districts, only falling by .02 grade equivalents.
6 Since these are smaller in absolute value than the losses in Figure 4, the average high-income district would still be
below 2019 levels.
16 | From Learning Recession to Learning Recovery
FIGURE 7. CHANGE IN AVERAGE ACHIEVEMENT BY DISTRICT POVERTY, 2022–25 VS.
2022–24
Note: Mean change in achievement from the Stanford Education Data Archive, weighted by district enrollment size.
The changes were similar at the other end of the income distribution. The highest poverty
districts—those with more than 80 percent of students eligible for federal subsidized lunches—
improved by over .2 grade equivalents in math and lost only .05 grade equivalents in reading. The
middle poverty districts—districts between 30% and 70% FRPL—saw the least improvement in
both subjects. The same “U-shaped” pattern of recovery is evident in math and reading.
We investigate three potential drivers of the unusual pattern of recovery: federal pandemic
relief, student absenteeism, and the share of schools identified for comprehensive support and
improvement.
The federal government provided three packages of aid to school districts during the pandemic,
totaling $190 billion: Elementary and Secondary School Emergency Relief (ESSER) I and II and the
American Rescue Plan (sometimes referred to as ESSER III). School districts were required to
obligate funds from the first and smallest package, ESSER I ($13 billion), by September 2022. Since
most of those funds would have been spent prior to spring 2022 (the beginning of the period we are
studying), we focus on the second two packages, ESSER II and ARP.
In Figure 8, we report the variation in ESSER II and ARP grants per student by district poverty.
ESSER funding was highly targeted to the highest poverty districts. Among districts with more than
80 percent of students receiving federal lunch subsidies, the median grant was $6,128 per student
in ESSER II and ARP. This is more than three times larger than the median grant received by the
median district with 40–50 percent of students receiving federal subsidized lunch. Not only is the
median funding per student higher in highest poverty districts, but the variation in size of grants
From Learning Recession to Learning Recovery | 17
is much wider for the higher poverty districts. For example, in the highest poverty districts (those
with 80–100 percent of students eligible for federally subsidized lunches), the 10th percentile district
received $3,707 per student. However, the 90th percentile district in that group received roughly
three times more per student ($10,725).
FIGURE 8. FEDERAL PANDEMIC RELIEF PER STUDENT BY DISTRICT POVERTY
Note: ESSER II and ARP ESSER grant per student, weighted by district enrollment size.
Both the high degree of targeting based on income and the wider variation in grants per student
at the highest poverty levels is due to the way Congress distributed the ESSER funds: grants were
directly proportional to the amount of federal Title I aid the district received in Fiscal Year 2020.
Because the Title I program targets aid to districts with higher concentrations of poor children,
such districts received the lion’s share of the federal aid.
Moreover, the wide variation in ESSER grants among districts with similar levels of poverty is also a
reflection of the differing formulae for Title I in different states. As reported in Appendix Figure C.1,
some states provided larger Title I grants per student for districts with the same poverty level. The
disparity in state Title I funding formulas is widest among higher poverty districts.
In Figure 9, we report the trends in rates of chronic absenteeism by tertile of district poverty.
Before the pandemic, absence rates were fairly stable for all three groups of districts, albeit higher
among higher poverty districts than middle and lower poverty districts. However, in the wake of
the pandemic, rates of chronic absenteeism essentially doubled from pre-pandemic levels for
all three tertiles: rising from 10 percent in 2019 to 20 percent of students in 2022 in the lowest
poverty districts, and from 20 to 37 percent of students in the highest poverty districts. While the
18 | From Learning Recession to Learning Recovery
proportional increases were similar, the absolute increases in chronic absenteeism were larger in
the highest poverty districts because they were starting from a higher base.
FIGURE 9. CHRONIC ABSENTEEISM RATE 2017–25 BY DISTRICT POVERTY TERTILE
Note: Based on data from Nat Malkus at the American Enterprise Institute. Estimates are weighted by district
enrollment.
Thus, both the amount of federal pandemic relief received and the level of student absenteeism
post-pandemic were related to district poverty. For that reason, the simple correlation between
federal pandemic relief and academic recovery likely combines the effect of the federal pandemic
funding, higher absence rates, and the consequences of district poverty—all of which could either
offset (or add to) each other. In the following analysis, we attempt to isolate the effect of the
federal pandemic relief and absenteeism by studying the relationship with achievement growth
among districts with similar poverty rates. In our statistical model, we group districts by every 2
percentage points of their school-age population that were eligible for Title I in FY 2020. We look
within each of these groups of districts (those that were 0 to 2 percent poor, 2 to 4 percent poor, 4
to 6 percent poor and so on) at the relationship between federal relief per student and the increase
in achievement between 2022 and 2025. We also control for district enrollment, the share of
students in different race/ethnicity groups in 2022, and the change in those student characteristics
between 2022 and 2025.
We then ask whether the districts in each group which received more pandemic relief or had higher
absences in the post-pandemic years had faster or slower achievement growth. In doing so, we
are essentially taking the variation in federal ESSER grants among similar districts as a “natural
experiment” driven by the state Title I formulae. We are also assuming that the variation in days
absent within these groups is also as good as random, although we do not know the mechanism
From Learning Recession to Learning Recovery | 19
(like the state Title I formulae) driving those differences. (See Appendix D for details of the
statistical model.)
TABLE 1. FACTORS RELATED TO POST-PANDEMIC ACADEMIC RECOVERY
Change in Math Achievement Change In Reading Achievement
2022-2024 2022-2025 2022-2024 2022-2025
(1) (2) (3) (1) (2) (3)
ESSER II+III Allocation per 0.0195** 0.0313*** 0.0309*** 0.0313*** 0.0378*** 0.0378***
Student ($1,000s) (0.0080) (0.0098) (0.0099) (0.0071) (0.0110) (0.0112)
-0.0042*** -0.0055*** -0.0057*** -0.0037*** -0.0056*** -0.0056***
Days Absent
(0.0016) (0.0014) (0.0014) (0.0015) (0.0014) (0.0015)
0.0362 -0.0005
CSI Status SY 2022-23
(0.0801) (0.0894)
N 7408 5380 5380 7064 4988 4988
R2 0.234 0.257 0.257 0.204 0.302 0.302
Note: The outcome is measured in grade equivalent units. Controls include state fixed-effects, dummy variables for
every 10% free/reduced price lunch, log total enrollment, percent black, Hispanic, white, rural, town, suburb in 2022.
We also control for changes in demographics and enrollment from 2022 to 2025. Additional controls include % of the
year spent remote or hybrid in the 2020–2021 school year, the trend in achievement from 2015–2019, and dummies for
every 2% of formula-eligible children. Observations are weighted by grade 3–8 enrollment in 2022. We capped ESSER
allocation per student at $16,000 per student. Standard errors are robust without clustering.
We report the results of that analysis in Table 1 above. In column (1), we repeat our analysis from
last year’s report on the effect of federal relief and student absences on achievement growth
between 2022 and 2024 (Dewey et al., 2025). We find that each $1,000 per student is associated with
a .0195 additional grade equivalent in math and .0313 grade equivalents in reading.
In column 2, we re-estimate the relationship between spending and achievement growth between
2022 and 2025. In math, the coefficient increases by roughly 60 percent to .0313 grade equivalents
per $1,000 of pandemic relief from last year’s analysis. This could either be because districts had
some funds left over after the end of the 2024 school year (we assumed they expended it all
by then) or because districts invested the federal dollars in ways which would yield continuing
effects, such as teacher training, curriculum, or capital improvements. The estimated impact per
$1,000 on reading achievement also increased in 2022–25 compared to the earlier period, 2022–24,
though more modestly.
As we discussed in last year’s report, these estimates are in line with prior evidence on the effect
of general revenues on student achievement summarized by Jackson and Mackevicius (2024).7
7 The authors identified 32 quasi-experimental studies with the most plausible claims to estimating causal effects
of spending (as opposed to simple correlations). Their meta-analysis implied an average impact of .0316 standard
deviations per $1,000 annual increase in spending over 4 years. Dividing by four yields .0079 s.d. per $1,000 spent. A grade
equivalent is .287 s.d. in math and .268 s.d. in reading, implying a coefficient of .0275 in math and .0295 in reading.
20 | From Learning Recession to Learning Recovery
In column (1), we find that each additional day in student absences during the 2022–2024 period was
associated with a .0042 grade equivalent decline in achievement growth in math and .0037 decline
in reading. In column (2), we repeat the analysis estimating the effect of the number of days absent
between 2022 and 2025 on achievement growth between 2022 and 2025. We find a similar answer:
each day of absence is associated with a .0055 grade equivalent loss in both math and reading. We
see the same relationship between a day absent and achievement even though we are looking over
three years rather than two years.
However, our estimate of .0055 grade equivalents per day of absence is smaller than that found in
prior research. For example, three recent studies estimated how changes in an individual student’s
days absent from one school year to the next were associated with changes in their test scores
(by including student fixed effects) (Aucejo and Romano, 2016; Gershenson et al., 2017; Goodman,
2014). The authors found that every day of absence reduced achievement by between .005 and .008
standard deviations in math (.017 to .028 grade equivalents), and between .003 and .008 standard
deviations in reading (.011 to .030 grade equivalents). These estimates are 2–5 times larger than
ours.
On top of the impact on the students who are absent, prior research suggests that the negative
effects of absences spill over onto classmates. Two studies (Goodman, 2014; Gottfried, 2011)
estimate the impact of absenteeism on other students in a school, grade, or classroom. If teachers
are forced to use class time to reteach material when students return from absences, then the
collective effect of absences for a whole cohort of students would be larger than the effect on the
absent students.
That appears to be the case. For example, using data from Massachusetts, Goodman (2014) reports
that one additional day of absence is associated with .008 standard deviations in lost achievement
for the absent student. A one-day increase in peer absences is associated with a similar .008 loss
in achievement. Combining the two, a one day increase in average days absent in a cohort would
be associated with a .016 standard deviation loss in achievement (.056 grade equivalents). Gottfried
(2011) finds similar combined effects of an extra day of cohort absences (.062 grade equivalents in
math and .056 grade equivalents for reading).
Our estimate reflects the total effect of an extra day of absence both on the individual student and
on classmates. Our estimate implies that the marginal effect of a day of absence is equal to the
average effect of each day of attendance: In most states, the school year is 180 days. Multiplying
the coefficient on absences in column (2), .0055, by 180 yields an estimate of a full school year.
Following the pandemic, states resumed designating schools for “comprehensive” and “targeted”
support and improvement. Depending on the state, the ratings were announced in the 2021–22 or
2022–23 school years, based on school outcomes in the prior year (either 2020–21 or 2021–22). Both
sets of ratings were based on student outcomes preceding our baseline year of spring 2022, so they
are not a direct function of achievement growth over the period we are investigating.
Several recent papers have studied the effect of being designated for “comprehensive support
and improvement” under ESSA. For instance, Atchison et al. (2025) find no positive effects of CSI
designation on elementary or middle school math or reading achievement in Ohio. In contrast,
From Learning Recession to Learning Recovery | 21
Burns et al. (2023) find positive effects on math achievement for CSI schools in Michigan. Cullum
and Harbatkin (2023) find that Michigan schools designated for comprehensive improvement fared
better in math during the pandemic.
In column (3), we report the relationship between the share of students identified for
comprehensive support and improvement in the district and the growth between 2022 and 2025.
We would expect a positive effect if being designated for comprehensive support either boosted
districts’ incentives to remove schools from the list or reflected positive effects of additional
school improvement dollars. Unfortunately, we found no relationship between the share of schools
designated for comprehensive support and improvement and post-pandemic improvement. Being
designated as a school in need of comprehensive support and improvement under ESSA does not
seem to be generating improvement in student achievement.
The Net Effect of Federal Aid, Student Absences, and Federal
Accountability
In Figure 10, we use the coefficients from Table 1 to isolate the effect of federal pandemic relief
and absences on recovery. The solid lines portray the actual change in math and reading scores
between 2022 and 2025 by district poverty. As discussed above, the recovery was “U-shaped” with
larger recovery among the lowest poverty and highest poverty districts. For the long dashed lines,
we use the results in Table 1 to predict what each district’s academic recovery would have been
without the federal relief. For districts with more than 40 percent of students receiving free and
reduced-price lunch, we estimate that there would have been no recovery in math and a .3 grade
equivalent loss in reading. In both math and reading, the relationship is more “L-shaped” than
“U-shaped,” implying that much of the increase in achievement among higher poverty districts was
associated with the federal pandemic relief.
We did not do a similar exercise for measuring the impact of having schools designated for CSI
because the coefficient was small and statistically insignificant.
Nevertheless, our results are not cause for celebration. As reported in Appendix E, although high
poverty districts have recovered as much as low poverty districts since 2022, they remain behind
2019 levels. The highest poverty districts are still about half a grade equivalent behind their 2019
level of achievement (they lost .7 grade equivalents during the pandemic and recovered .2 grade
equivalents since 2022). In contrast, the lowest poverty districts are only an average of .1 grade
levels below their 2019 levels.
The federal dollars seemed to have an impact, but were they used effectively? The federal dollars
had a similar impact per dollar to a general revenue increase, such as from an increase in state
grants to districts. General revenue increases are used for instructional and non-instructional
purposes—everything from school buses to facilities improvements to administrator salaries.
The relief dollars would have had an even larger impact had they been fully invested in academic
recovery efforts, such as tutoring, summer learning, or extending the school year. For example,
Guryan et al. (2023) estimate that a specific high-dosage tutoring program had an impact per
dollar spent roughly 10 times larger than we are estimating. However, it would be inappropriate to
22 | From Learning Recession to Learning Recovery
blame all of that on district leadership. Part of the responsibility rests on the U.S. Congress, since
they only required districts to spend 20 percent of the American Rescue Plan grants on academic
recovery.
FIGURE 10. ACTUAL AND PREDICTED ACADEMIC RECOVERY BY DISTRICT POVERTY
Note: The solid line is mean change in achievement from the Stanford Education Data Archive, weighted by district
enrollment size. To calculate the estimated change in achievement without ESSER (the long dashed line) we
subtracted the product of the district’s ESSER II/ARP grant per student and the coefficient from column (3) in Table
1. To calculate the effect of absenteeism, we multiplied the change in average annual days absent from 2017–19 to
2023–25 by 3 to get the average days absent over the 3-year period. We then took this value and multiplied it by the
coefficient on days absent in column (3) in Table 1 and subtracted it from the actual change in achievement.
For the dashed line above the actual change, we subtract out the change in the average number
of absences since 2019 times the coefficient on absences in column (3) of Table 1. In effect, we
run the following thought experiment: “How much recovery would we have seen if absences rates
had remained at their 2019 levels?” We estimate that the average high poverty district would have
seen an extra .05 grade equivalent improvement in math and reading if absences had returned to
2019 levels. Recovery would have improved in the lowest poverty districts as well, but by a smaller
amount, given that their number of absences increased by less than in the highest poverty districts.
As discussed above, our estimates of the impact of a day of absence are considerably smaller than
prior research. Thus, if anything, our estimates of the consequences of rising absence rates are
understated.
From Learning Recession to Learning Recovery | 23
State Improvement in Reading and Math 2022 to 2025
In Figure 11, we report the change in achievement between 2022 and 2025 by state.8 Only seven
states (plus the District of Columbia) improved in reading achievement in the three years following
the pandemic: Louisiana, Maryland, Tennessee, Kentucky, Indiana, Mississippi, Minnesota, and D.C.
27 states followed up their pandemic declines with further declines in reading achievement.
Many of the states that improved in reading also had a larger than average improvement in math
achievement. However, four states lost additional ground in math in the years following the
pandemic: Iowa, Nebraska, Wyoming, and Idaho.
FIGURE 11. STATE ACADEMIC GROWTH BETWEEN 2022 AND 2025
Note: These are state improvements in math and reading achievement in grade level units between 2022 to 2025. State
improvements are based on overall state achievement which includes non-traditional public school districts.
8 We have done a similar analysis using the “adjusted state improvements” captured by the state fixed effects in Table
1. The results are similar using the state changes adjusted using district-level variation in demographics, absences, and
ESSER spending.
24 | From Learning Recession to Learning Recovery
Early Literacy Initiatives
As of March 2026, Education Week estimates that 42 states plus the District of Columbia have
passed laws or implemented policies related to evidence-based reading instruction—often
referred to as the “science of reading” (Schwartz, 2026). However, policies and implementation have
varied widely, with some states simply providing voluntary guidance on curricula while other states
implemented a full package of reforms, such as requiring all K-3 teachers and administrators to
be retrained, hiring literacy coaches for elementary schools, limiting district choices to approved
curricula, and retaining students who fail to show reading proficiency at the end of third grade.
In January 2024, the non-profit organization the Foundation for Excellence in Education (ExcelinEd)
reported on the implementation of 18 “science of reading” policy elements (including those above)
in every state: whether “full implementation,” “partial implementation,” “future implementation,”
or “not adopted.” Based on the ExcelinEd report, we created an index, counting each element in
full or partial implementation. Any policy element which was not adopted or was pending future
implementation was not counted. In Figure 12, we plot the change in state reading achievement
between 2022 and 2025 against the sum of the number of policy elements that had been
implemented or partially implemented as of January 2024.
FIGURE 12: IMPLEMENTATION OF SCIENCE OF READING AND GROWTH IN READING
2022–25
Note: The vertical axis is the achievement growth between 2022 and 2025. The horizontal axis is the number of “science
of reading” policy elements implemented fully or partially in each state according to the Foundation for Excellence in
Education in January 2024.
From Learning Recession to Learning Recovery | 25
There is not a strong correlation between the magnitude of state reading improvement and the
number of science of reading policy elements adopted. However, we did note a striking pattern:
none of the states which had implemented fewer than seven of the science of reading policy
elements as of January 2024 (South Dakota, Hawaii, Rhode Island, New Jersey, New Hampshire,
Georgia, Wisconsin, California, Washington and Massachusetts) saw achievement growth in
reading. (The upper left quadrant in Figure 12 is empty.) Reading achievement in these states
continued to fall following the pandemic.
Meanwhile, all of the states (plus the District of Columbia) with any improvement in reading
following the pandemic (the District of Columbia, Maryland, Louisiana, Kentucky, Tennessee,
Indiana, Mississippi, and Minnesota) had implemented more than seven of the policy elements
tracked by ExcelinEd. However, there were some states which had implemented more than 7 of
the policy elements who still had not improved in reading (notably, Florida, Arizona, and Nebraska).
While we see evidence suggesting that implementation of a comprehensive literacy strategy may
be a necessary condition for improvement, mere adoption is apparently not sufficient.
Districts on the Rise: Counting 108 Mississippis
Figure 13 illustrates the distribution of district-level improvements between 2022 and 2025 in math
and reading. Although the average district improved by .01 grade equivalents in math and lost .15
grade equivalents in reading, the actual amount of recovery varied widely from district to district.
For example, 18 percent of districts improved by more than half a grade level in math, while 9
percent declined by more than half a grade level.
FIGURE 13. DISTRIBUTION OF CHANGES IN ACHIEVEMENT 2022–25
Note: The distribution of changes in achievement between 2022 to 2025, weighted by district enrollment.
In the past couple of years, much has been written about the “Mississippi Miracle”—the large
improvement in both math and reading achievement seen in Mississippi since 2012. The progress
26 | From Learning Recession to Learning Recovery
in Mississippi has inspired state leaders around the country to take more significant action to
improve student achievement.
Using Mississippi as our example, we identify 108 “districts on the rise” (108 Mississippis, if you will)
that have shown unusual progress in both math and reading relative to similar districts in their own
states. A total of 448 districts showed large improvements relative to their peers in either math or
reading. To be considered a “district on the rise,” a district must have had enrollment of more than
1,200 students in grades 3–8, have achievement estimates available for 2019, 2022, and 2025, and
a comparison group of at least four similar districts, each with achievement estimates for 2019,
2022, and 2025. To ensure that any changes in achievement are not driven by demographic changes
within the district, the percentage change in grades 3–8 enrollment and the percentage change in
students eligible for federal free lunch from 2022 to 2025 must both be no more than 10 percentage
points different than the percent change among similar districts. We then selected the districts
which improved by more than .3 grade equivalents relative to their similar districts’ average
achievement from spring 2019 to spring 2025 and spring 2022 to spring 2025.9
The set of similar districts is provided by the SEDA data. Similar districts were selected based on
their similarity on the following district characteristics: the average enrollment in grades 3 to 8
between 2022 and 2025; the average SES of families in the district between 2022 and 2024; the
average percentage of students receiving federal subsidized lunches between 2022 and 2025; the
average proportions of students who identify as Asian, Black, Hispanic, and White in the district
between 2022 and 2025; and the proportions of students in urban schools, suburban schools, and
town schools from 2022–2025. The similar districts are the five districts in the same state closest in
these characteristics (as measured by Mahalanobis distance) to the focal district.
In Figure 14 below, we describe the trends in math achievement in six of those districts. For
instance, Birmingham City school district in Alabama, like its comparison districts, lost about
.4 grade equivalents in math achievement during the pandemic between 2019 and 2022. The
comparison districts have rebounded slightly, improving by roughly .2 grade equivalents since 2022.
However, Birmingham enjoyed an unusually robust recovery, improving by .7 grade equivalents
since 2022. Birmingham students are now scoring .3 grade equivalents above their predecessors in
2019.
Atlanta Public Schools offer another inspiring story. Like its comparison districts, math
achievement in Atlanta declined by .8 grade equivalents between 2019 and 2022. While the
comparison districts have continued to decline since the pandemic, Atlanta has made up more
than half of the loss, rising by .5 grade equivalents in math since 2022.
The other districts shown in Figure 14—Detroit, Michigan; Modesto, California; Cambridge,
Massachusetts; and Spring Branch, Texas—follow similar patterns. Although we cannot say
exactly what these districts were doing to help their students recover, we owe it to our children to
understand what they did and learn whether similar strategies are working elsewhere.
9 We also added districts in the top five in their state if their increase relative to comparison was less than .3 grade
equivalents.
From Learning Recession to Learning Recovery | 27
FIGURE 14. SIX DISTRICTS ON THE RISE
Note: The figure reports difference in math achievement for six districts relative to their within-state
comparisons. Both the districts and the comparisons are being compared to their own 2019 achievement.
28 | From Learning Recession to Learning Recovery
III. Conclusion
K-12 education in the United States has been far from stagnant. Like the economy, K-12 education
has been through ups and downs over the past four decades—but mostly ups. For nearly 25 years
between 1990 and 2013, math achievement in grades 4 and 8 rose steadily, improving by more than
two grade equivalents over that time period. The 4th grade students in 2013 were scoring at a similar
level to 6th graders in 1990. Reading achievement also rose, albeit by somewhat less than one
grade equivalent. Those prior improvements make the decline since 2013 appear more striking and
anomalous. The “learning recession” began before the pandemic and only turned around in 2022 in
math and in 2024 in reading. As with an economic recession, recovery is possible—but it requires
decisive action.
To capitalize on the energy behind the recent literacy reforms, we call for a national effort to
identify which efforts are most effective. For instance, states could use their own data systems
to monitor whether student achievement growth improves when schools switch curricula or
when schools receive a literacy coach. Payments to providers of literacy curricula could be made
contingent on students’ achievement growth.
A number of research teams (including ours) will be releasing studies on the impact of cell phone
bans in the coming months. Once those emerge, there should be a coordinated effort to resolve
any disparities in the findings and to reach a consensus on the impact of bans for students. The
research community does not have a great track record of reconciling differences in findings on
its own. New approaches to restricting cell phone use in schools are emerging which could avoid
the unintended consequences for student discipline, like geo-fencing cell phone access on school
sites. There should be a national effort to evaluate different approaches to cell phone bans.
But cell phone bans address only one of many possible mechanisms through which social media
could be impacting student achievement: by distracting students during instructional time. There
are other possible mechanisms which could be more important: displacing homework time out of
school, disrupting sleep, weakening executive function/attention management, promoting stress/
anxiety, etc. Failure to find large impacts of cell phone bans does not mean that the concern over
social media is misplaced. Rather, cell phone bans are only one vector for reducing potential harm
from social media exposure. We should also be piloting and evaluating efforts to reduce student
access to social media outside of school.
Absence rates remain a headwind for school improvement. Lowering student absenteeism offers
a substantial “bang for the buck.” Taxpayers are already paying for teacher salaries and classroom
space. If students were back in seats at rates similar to before the pandemic, the recovery would
have been more robust. Some may believe that attendance is a parental responsibility and that
if parents are willing to accept lower achievement for their own children as a consequence of
irregular school attendance, they should be able to make that choice. However, evidence suggests
that the costs of absenteeism spill over onto other students, when teachers have to reteach
material and learning slows for everyone. Thus, we all have a collective interest in lowering student
absenteeism.
From Learning Recession to Learning Recovery | 29
The most valuable contribution of the Education Scorecard is in making local recovery efforts—
both successful and unsuccessful—more visible. Without the SEDA data, it would be impossible to
compare achievement improvement across state lines, given the differences in state proficiency
definitions. Moreover, because states occasionally change their standards, it is also impossible to
compare a given community’s achievement over time.
We hope that states will use these data in two ways: first, states should consider continuing loss in
achievement since 2019 when designating schools for comprehensive support and improvement.
The federal pandemic relief is gone. If the schools that remain far behind 2019 levels are going
to recover, they will need additional support. States could target the federal dollars for school
improvement under Title I to the schools that remain furthest behind their own 2019 levels of
achievement. Second, states should offer to pay the cost for districts who have lost the most
ground to visit similar school districts in their state who have had much more success recovering.
To avoid overwhelming the successful districts, the state could pair one struggling district to each
“district on the rise”—and then see if the chosen districts outperform the other struggling districts
who were not chosen. It would be a low-cost way to see the power of peer learning.
30 | From Learning Recession to Learning Recovery
References
Abrahamsson, S. (2024). Smartphone bans, student outcomes, and mental health (Discussion Paper
No. 1). Department of Economics, Norwegian School of Economics. https://doi.org/10.2139/
ssrn.4735240
Allcott, H., Baron, E. J., Dee, T., Duckworth, A. L., Gentzkow, M., & Jacob, B. (2026). The effects of
school phone bans: National evidence from lockable pouches. (Working Paper No. 35132).
National Bureau of Economic Research. https://doi.org/10.3386/w35132
Arenas-Arroyo, E., Fernández-Kranz, D., & Nollenberger, N. (2022). High speed internet and the
widening gender gap in adolescent mental health. Journal of Health Economics, 102. https://doi.
org/10.1016/j.jhealeco.2025.103014
Atchison, D., Özek, U., Le Floch, K. C., & Hurlburt, S. (2025). Effects of Comprehensive Support
and Improvement school designation under the Every Student Succeeds Act on student
outcomes: Evidence from Ohio. Journal of Research on Educational Effectiveness. Advance online
publication. https://doi.org/10.1080/19345747.2025.2537131
Aucejo, E.M., & Romano, T. F. (2016). Assessing the effect of school days and absences on test
score performance. Economics of Education Review, 55, 70–87. https://doi.org/10.1016/j.
econedurev.2016.08.007
Beland, L.-P., & Murphy, R. (2016). Ill communication: Technology, distraction & student
performance. Labour Economics, 41, 61–76. https://doi.org/10.1016/j.labeco.2016.04.004
Bleiberg, J. (2026). The implications of No Child Left Behind’s waivers for contemporary
school improvement systems. Educational Policy. Advance online publication. https://doi.
org/10.1177/08959048251414160
Braghieri, L., Levy, R., & Makarin, A. (2022). Social media and mental health. American Economic
Review, 112(11), 3660–3693. https://doi.org/10.1257/aer.20211218
Carnoy, M., & Loeb, S. (2002). Does external accountability affect student outcomes? A cross-
state analysis. Educational Evaluation and Policy Analysis, 24(4), 305–331. https://doi.
org/10.3102/01623737024004305
Center on Education Policy (2008). Many states have taken a “backloaded” approach to No Child Left
Behind goal of all students scoring “proficient.” https://files.eric.ed.gov/fulltext/ED503796.pdf
Churchill, B. F., & Johnson, K. R. (2026). Broadband internet access and adolescent mental health
in the U.S. (Working Paper No. 34614). National Bureau of Economic Research. https://doi.
org/10.3386/w34614
Cullum, S., & Harbatkin, E. (2023). Student Achievement in the First Two Cohorts of Partnership
Schools. Education Policy Innovation Collaborative – Michigan State University. https://
epicedpolicy.org/wp-content/uploads/2023/04/Partnership_Student_Acheivement_Brief_
April2023.pdf
From Learning Recession to Learning Recovery | 31
Dee, T. S., & Jacob, B. A. (2011). The impact of No Child Left Behind on student achievement. Journal
of Policy Analysis and Management, 30(3), 418–446. https://doi.org/10.1002/pam.20586
Dewey, D. C., Fahle, E. M., Kane, T. J., Reardon, S. F., & Staiger, D. O. (2025). Pivoting from pandemic
recovery to long-term reform: A district-level analysis (Education Recovery Scorecard). Center
for Education Policy Research at Harvard University & The Educational Opportunity Project at
Stanford University. https://educationrecoveryscorecard.org/wp-content/uploads/2025/02/
Pivoting-from-Pandemic-Recovery-to-Long-Term-Reform-A-District-Level-Analysis.pdf
Donati, D., Durance, R., Sobbrio, F., & Zejcirovic, D. (2025). Lost in the net? Broadband internet
and youth mental health. Journal of Health Economics, 103. https://doi.org/10.1016/j.
jhealeco.2025.103017
Education Data Center. (2025). State assessment data repository (Version 3.0): Data documentation.
Retrieved from https://www.zelma.ai/data_documentation/EDC_technical_documentation_
v3.0.pdf
Fahle, E. M., Saliba, J., Reardon, S. F., Shear, B. R., Kalogrides, D., & Ho, A. D. (2026a). Stanford
Education Data Archive: Technical Documentation (Version 6.0). https://stacks.stanford.edu/
file/xh833nn4025/SEDA_documentation_6.0.pdf. Retrieved from https://purl.stanford.edu/
xh833nn4025.
Fahle, E. M., Saliba, J., Reardon, S. F., Shear, B. R., Min, J., & Shim, J. & Kalogrides, D., & Ho, A.
D. (2026b). Stanford Education Data Archive Technical Documentation: SEDA 2025.1. https://
stacks.stanford.edu/file/hm970gr1371/SEDA_documentation_2025.1.pdf. Retrieved from
https://purl.stanford.edu/hm970gr1371.
Fahle, E. M., Kane, T. J., Patterson, T., Reardon, S. F., Staiger, D. O., & Stuart, E. A. (2025). School
district and community factors associated with test score declines during the COVID‑19
pandemic. Teachers College Record: The Voice of Scholarship in Education, 127(6–7), 87–123.
https://doi.org/10.1177/01614681251369937
Fahle, E. M., Reardon, S. F., Shear, B. R., Ho, A. D., Saliba, J., Min, J., Shim, J. & Kalogrides, D.
(2025). Stanford Education Data Archive technical documentation: SEDA 2024. Stanford
University. https://purl.stanford.edu/pt329xg7054
Figlio, D. N., & Özek, U. (2025). The impact of cellphone bans in schools on student outcomes:
Evidence from Florida (Working Paper No. 34388). National Bureau of Economic Research.
https://doi.org/10.3386/w34388
Gershenson, S., Jacknowitz, A., & Brannegan, A. (2017). Are student absences worth the worry in
U.S. primary schools? Education Finance and Policy, MIT Press, 12(2), 137–165. https://ideas.
repec.org/a/tpr/edfpol/v12y2017i2p137-165.html
Golin, M. (2022). The effect of broadband internet on the gender gap in mental health: Evidence from
Germany. Health Economics, 31(S2), 6–21. https://doi.org/10.1002/hec.4570
Goodman, J. (2014). Flaking out: Student absences and snow days as disruptions of instructional
time. (Working Paper No. 20221). National Bureau of Economic Research. https://doi.
32 | From Learning Recession to Learning Recovery
org/10.3386/w20221
Haidt, J. (2024). The anxious generation: How the great rewiring of childhood is causing an epidemic
of mental illness. Penguin Press.
Hanushek, E. A., & Raymond, M. E. (2005). Does school accountability lead to improved student
performance? Journal of Policy Analysis and Management, 24(2), 297–327. https://doi.
org/10.1002/pam.20091
Jackson, C. K., & Mackevicius, C. L. (2024). What impacts can we expect from school spending
policy? Evidence from evaluations in the United States. American Economic Journal: Applied
Economics, 16(1), 412–446. https://doi.org/10.1257/app.20220279
Jain, R., & Stemper, S. (2025 ). 3G Internet and human capital development (Working Paper No. 453).
Department of Economics, University of Zurich.
Kessel, D., Hardardottir, H. L., & Tyrefors, B. (2020). The impact of banning mobile phones
in Swedish secondary schools. Economics of Education Review, 77, 102009. https://doi.
org/10.1016/j.econedurev.2020.102009
Le Floch, K. C., Hurlburt, S., Atchison, D., & Hyland, K. (2024). Identifying the nation’s lowest
performing schools: Shifts following the passage of the Every Student Succeeds Act (ESSA).
(NCEE 2025–001r). U.S. Department of Education, Institute of Education Sciences, National
Center for Education Evaluation and Regional Assistance. https://ies.ed.gov/use-work/
resource-library/report/evaluation-report/identifying-nations-lowest-performing-schools-
shifts-following-passage-every-student-succeeds-act
McDool, E., Powell, P., Roberts, J., & Taylor, K. (2020). The Internet and children’s psychological
well-being. Journal of Health Economics, 69, 102274. https://doi.org/10.1016/j.
jhealeco.2019.102274
Meyers, C. V., Brandt, W. C., & Van Gronigen, B. A. (2023). State ESSA plans and Comprehensive
Support and Improvement (CSI) status. Educational Policy, 37(5), 1277-1314. https://doi.
org/10.1177/08959048221087209
National Center for Education Statistics. (2023). Condition of education. U.S. Department of
Education, Institute of Education Sciences. Retrieved April 25, 2026, from https://nces.ed.gov/
programs/coe/indicator/cnc and https://nces.ed.gov/programs/coe/indicator/cnb/reading-
performance
Pew Research Center. (2022, August 10). Teens, social media and technology 2022. Pew Research
Center. https://www.pewresearch.org/internet/2022/08/10/teens-social-media-and-
technology-2022/
Pugno, M. (2025). Does social media harm young people’s well-being? A suggestion from
economic research. Academia Mental Health and Well-Being, 2(1). https://doi.org/10.20935/
MHealthWellB7581
Reardon, S. F., Ho, A. D., Shear, B. R., Fahle, E. M., Kalogrides, D., & Saliba, J. (2026a). Stanford
Education Data Archive (Version 6.0). https://purl.stanford.edu/xh833nn4025
From Learning Recession to Learning Recovery | 33
Reardon, S. F., Fahle, E. M., Ho, A. D., Shear, B. R., Saliba, J., Min, J., Shim, J., & Kalogrides, D.
(2026b). Stanford Education Data Archive (Version SEDA 2025.1). https://purl.stanford.edu/
hm970gr1371
Reber, S. & Gordon, N. (2023). A deep dive on how Title I funds are allocated. The Brookings
Institution. https://www.brookings.edu/articles/a-deep-dive-on-how-title-i-funds-are-
allocated/
Schwartz, S. (2022, July 20). Which states have passed “science of reading” laws? What’s in them?
Education Week. https://www.edweek.org/teaching-learning/which-states-have-passed-
science-of-reading-laws-whats-in-them/2022/07
Ujifusa, A. (2021, June 24). The feds offered waivers on ESSA accountability. Here’s where states
stand on getting them. Education Week. https://www.edweek.org/policy-politics/the-feds-
offered-waivers-on-essa-accountability-heres-where-states-stand-on-getting-them/2021/06
34 | From Learning Recession to Learning Recovery
Appendix A. Comparing the 2025 Sample
to National Sample
Nationally 38 States + DC Analysis Sample
2022-25 Change in Math 0.166 0.139 0.166
2022-25 Change in Reading -0.165 -0.166 -0.135
2019-25 Change in Math -0.396 -0.410 -0.360
2019-25 Change in Reading -0.445 -0.434 -0.428
2019-22 Change in Math -0.562 -0.550 -0.525
2019-22 Change in Reading -0.279 -0.267 -0.292
2017-2019 Chronic Absence Rate 14.8% 14.1% 13.8%
2022 Chronic Absence Rate 27.9% 27.2% 27.1%
2024 Chronic Absence Rate 23.0% 22.3% 22.2%
2025 Chronic Absence Rate 22.9% 22.0% 22.0%
% Black (2025) 14.8% 15.6% 15.7%
% Hispanic (2025) 30.0% 30.3% 30.0%
% Minority (2025) 57.1% 57.8% 57.8%
% Free Lunch (2025) 57.1% 57.3% 57.3%
Percent of 2020-21 SY in Remote Instruction 24.4% 25.0% 24.4%
Percent of 2020-21 SY in Hybrid Instruction 39.3% 36.8% 36.1%
ESSER I Dollars Allocated Per Student $261.96 $254.24 $226.60
ESSER II Dollars Allocated Per Student $1,077.74 $1,053.66 $938.90
ESSER III Dollars Allocated Per Student $2,420.45 $2,366.37 $2,095.10
ESSER Dollars Allocated Per Student $3,760.15 $3,674.28 $3,260.60
Number of Districts 17,952 13,656 5,680
Number of Students Grades K-12 48,999,325 41,136,507 34,802,337
Number of Students Grades 3-8 21,775,782 18,317,231 15,728,212
Average K-12 Enrollment (2025) 2,729 3,012 6,127
Average 3-8 Enrollment (2025) 1,213 1,341 2,769
From Learning Recession to Learning Recovery | 35
Appendix B: Imputing Average Absence
Rate from Chronic Absence Rate
The chronic absence rate is the share of students absent for more than 10 percent of a school
year, while the average absence rate refers to the percent of students missing on a particular day.
It is not possible to derive analytically the average absence rate from the chronic rate, because
absence rates will vary by student. However, it turns out that there’s a surprisingly robust empirical
relationship between the share of students chronically absent and average absence rates.
APPENDIX FIGURE B.1.
Note: Estimated with district level data on average absence rate and chronic absence rate in Massachusetts
and Florida in 2019 and 2022. If pooled across both time periods and two states, the estimated linear relationship
is , with standard errors on the intercept and slope of .03 and .002
respectively.
In Figure B.1, we plot districts’ average absence rate (on the vertical axis) against the share of
students chronically absent (on the horizontal axis) for districts in Florida and Massachusetts—
two states which happen to report both statistics. We estimate the relationship between the two
measures separately by state and by before the pandemic (2019) and after the pandemic (2022).
The relationship is strikingly consistent between states and over time: one percentage point rise in
chronic absenteeism implies a .18 percentage point rise in the average daily absence rate. We used
the pooled linear regression of average absence rate on chronic absence rate (Average Absence
Rate=3.12+.182*Chronic Rate) to infer average absence rate from the chronic absence rate.
36 | From Learning Recession to Learning Recovery
Appendix C. The Natural Experiment
Created by State Title I Formulae
APPENDIX FIGURE C.1 DIFFERENCE IN TITLE I ALLOCATIONS BY PERCENT OF
CHILDREN ELIGIBLE BY STATE
Notes: This figure was based on actual Title I allocations in FY 2020, fitted with a linear spline function, with knots at 2
percent, 5 percent and 15 percent eligible children, separately for each district. Although not a perfect fit (for instance,
large districts can qualify for a concentrated grant even if they have fewer than 15 percent poor children, as long as
they have more than 6,500 eligible children), the splines are a good summary, explaining 98 percent of the variance in
Title I grants per population for districts not subject to hold harmless provisions.
Figure C.1 reports the mean Title I allocation per population aged 5–17 by the percentage of school-
age children eligible for Title I for each of seven states.10 Title I allocations per child are typically
zero for the lowest poverty districts (those with fewer than two percent eligible children).
10 These figures were fitted with a linear spline function, with knots at 2 percent, 5 percent and 15 percent eligible.
Although not a perfect fit (for instance, large districts can qualify for a concentrated grant even if they have fewer than
15 percent poor children, as long as they have more than 6,500 eligible children), the fitted spline functions shown in
Figure 4 are a good summary. The fitted splines explain 98 percent of the variance in Title I grants per population for
those districts not subject to hold harmless provisions.
From Learning Recession to Learning Recovery | 37
The relationship between eligibility and allocations steepens at five and fifteen percent eligible
children, as districts become eligible for additional types of Title I grants.
At any given percentage of eligible children (equivalent to drawing a vertical line in Figure C.1), Title
I grants vary depending upon the state where the district is located. Those differences become
larger at higher poverty levels. For instance, a district with 40 percent of children meeting the
eligibility formula would have received $572 per child in Tennessee, $652 per child in Alabama, $742
in California, $769 in Ohio, $870 in Illinois, $957 in Massachusetts, $1,069 in South Dakota, and $1,602
in New Hampshire.11 In other words, for very poor districts, there is roughly a $1,000 difference in
Title I allocations per child for those in New Hampshire versus those in Alabama or Tennessee.
Such differences are driven by two primary factors: state average per pupil expenditures and
the minimum grants for small states.12 In general, when states increase their average per pupil
spending, districts will receive more Title I funding for each poor child. That is the primary reason
why districts in most states, like Massachusetts or Illinois, receive more Title I funding than
districts in Alabama and Tennessee with the same percentage of eligible children.
However, poor districts in small states benefit tremendously from the minimum state grant
provisions. The states with the largest Title I grants in Figure C.1 (New Hampshire and South
Dakota) are not particularly high spending states, but they do have small populations. Congress has
guaranteed that small states receive a minimum share of appropriations for Title I. But, in Figure
C.1, it is not just the average grant that is higher in South Dakota and New Hampshire; the slope is
steeper—meaning that poor districts especially benefit from the small state minimum. The reason
for the steeper slope is an indirect result of the fact that Title I is not fully funded. Based on the
formula, districts were eligible for seven times more funding under the Title I basic grant program
than Congress appropriated in FY 2021 (Gordon and Reber, 2023). When appropriations fall short of
authorized grants, Title I allocations are proportionally reduced within a state. Thus, when a state
benefits from the small state minimum, all the district allocations are increased proportionally,
meaning that districts in small states receive allocations closer to the full authorized amount. The
states that benefited from the small state minimums in FY 2020 were Vermont, Wyoming, North
Dakota, New Hampshire, DC, Alaska, South Dakota, and Montana (Gordon and Reber, 2023).
Because ESSER funding was more than 10 times larger than Title I funding in FY 2020, the pandemic
relief simply amplified by idiosyncrasies in the Title I formula. Total funding for the Title I program
was $16 billion in FY 2020, while the total funding for ESSER II and ARP was $175 billion—more
than 10 times larger. In effect, with ESSER II and ARP, Congress pushed $175 billion through pipes
that were designed to carry one-tenth the volume of funds. The relief packages essentially
multiplied the differences in Figure C.1 by 10: a $1,000 difference in Title I grant per student became
a $10,000 difference in federal pandemic relief per student. In our analysis, we use this variation to
investigate the impact of federal pandemic relief on student achievement.
11 For an excellent explanation of the Title I formula, see Gordon and Reber (2023).
12 On paper, there are other factors that matter as well, such as the “state equity factor” (based on the coefficient of
variation in expenditures per student across districts in the state) and the “state effort factor” (a function of the ratio
of education spending per child and per capita income). However, as Gordon and Reber (2023) show, those adjustments
have little effect on the state differences illustrated in Figure C.1.
38 | From Learning Recession to Learning Recovery
Appendix D. Statistical Model
We model the change in achievement between 2022 and 2025, , as a function of the
average number of days that students were absent, the federal pandemic relief per student, and
the share of schools identified for comprehensive support and improvement (CSI), as well as a set
of district characteristics, , and state fixed effects, .
We use the ESSER allocation per student as a proxy for the increase in expenditure per student.
Unfortunately, the data on local government revenues per student and actual expenditures per
student are not yet available for 2023–24 or 2024–25. Until such data are released by the federal
Department of Education, we assume that the federal aid did not crowd out spending from state
and local revenue. This is a reasonable assumption regarding state revenues, since maintenance
of effort provisions in the federal law required states to maintain their school funding.13 If state
governments tried to reduce their contribution, it would negatively impact their Title I revenues in
the future. However, the maintenance of effort provisions did not apply to contributions from local
revenue. If local governments did cut back their contributions, a $1,000 allocation per student in
ESSER dollars would have resulted in less than $1,000 in additional expenditures. Importantly, such
bias would lead us to understate (not overstate) the impact per federal dollar in aid.
The set of district characteristics included in equation (1) are student demographics in 2022 (the
percentage of students in each racial group, the log of total district enrollment, dummy variables
for every 10% free/reduced price lunch, percentage of students in rural, town, suburb, or urban
geographies), the change in these student demographics from 2022 to 2025, the percent of the
2020–21 school year that schools were either remote instruction and a hybrid of remote and in-
person (see our 2024 report for details on how this measure was calculated) and the trend in
district achievement between 2015 and 2019. In addition, we included fixed effects for every two
percentage points of student eligibility for Title I. By including such fixed effects, we attempt to
distinguish the effect of federal pandemic relief and student absences from the effect of district
poverty.
The state fixed effects in equation (1) reflect the achievement growth in each state, after adjusting
for the amount of pandemic relief states received as well as other components of : student
demographics, levels of poverty in the district, the share of the 2020–21 school year that schools
were remote or hybrid, and the trend in scores between 2015 and 2019.
13 The maintenance of effort requirements under ESSER did not apply to local governments. However, as the
Department of Education indicated in E-14 of ESSER-and-GEER-Use-of-Funds-FAQs-December-7-2022-Update-1.
pdf (ed.gov), any district that chose to replace state or local funds with federal ESSER funds risked failing to meet the
maintenance of effort requirement under the Title I program itself.
From Learning Recession to Learning Recovery | 39
Appendix E. Changes Since 2019 by
District Poverty (Without ESSER and
Absenteeism)
APPENDIX FIGURE E.1 ACTUAL AND PREDICTED ACHIEVEMENT CHANGES SINCE
2019 BY DISTRICT POVERTY
Note: The solid line is mean change in achievement from the Stanford Education Data Archive, weighted by district
enrollment size. To calculate the estimated change in achievement without ESSER (the long dashed line) we
subtracted the product of the district’s ESSER II/ARP grant per student and the coefficient from column (3) in Table
1. To calculate the effect of absenteeism, we multiplied the change in average annual days absent from 2017–19 to
2023–25 by 3 to get the average days absent over the 3-year period. We then took this value and multiplied it by the
coefficient on days absent in column (3) in Table 1 and subtracted it from the actual change in achievement.
40 | From Learning Recession to Learning Recovery
The Education Scorecard is a collaboration between two of the nation’s
leading education research centers: the Center for Education Policy
Research (CEPR) at Harvard University and the Educational Opportunity
Project (EOP) at Stanford University. Together, CEPR and EOP have built a
real-time, publicly accessible window into how students in local school
districts across America are learning and growing.
educationscorecard.org
File and source
- File
- education-scorecard-2026-05-learning-recession-to-learning-recovery.pdf
- Size
- 4,495,188 bytes
- SHA-256
- 7d1fededbdd06911e68080f78b5691f53e02a2828b9d9bac6f691b6a30f3ed9d
- Original
- educationscorecard.org