October 21, 2021

Missing in the Margins 2021: Revisiting the COVID-19 Attendance Crisis

By Hailly T.N. Korman | Bonnie O’Keefe | Matt Repka

Share this article

Missing in the Margins 2020: Estimating the Scale of the COVID-19 Attendance Crisis

Hailly T.N. Korman
Bonnie O’Keefe
Matt Repka


  1. Introduction
  2. Why aren’t students attending school?
  3. What will happen if these students don’t get back to school?
  4. How did you estimate 3 million missing students?
  5. What should leaders do about these missing students?
  6. Appendix: Data Methods, Limitations, and Sources
  7. Acknowledgments & Credits


It’s 2020. Do you know where 3 million students are?



By mid-March 2020, most American schools had shut their doors, and about half remain fully or partially closed to in-person learning today. For approximately 3 million of the most educationally marginalized students in the country, March might have been the last time they experienced any formal education — virtual or in-person.  (To see how we estimated 3 million, click hereTo see a state-by-state breakdown of these estimates, click here. State-by-state estimates updated January 2021 to include Puerto Rico and the Bureau of Indian Education.)

Stories from across the country illustrate pieces of this crisis:

  • In Los Angeles, 15%-20% of English learners, students in foster care, students with disabilities, and homeless students didn’t access any of the district’s online educational materials from March through May.

  • In Washington, D.C., back-to-school family surveys found that 60% of students lacked the devices and 27% lacked the high-speed internet access needed to successfully participate in virtual school.

  • In Miami-Dade County, 16,000 fewer students enrolled this fall compared with last year.

Because COVID-19 has disrupted education for almost every student in the country, more severe effects on some groups of students can seem comparatively small. And in the ongoing chaos of the pandemic, reliable data is scarce.

We’ve attempted to estimate the total number of students experiencing the gravest consequences of school closure and the shift to distance learning by identifying the groups most at risk, and calculating a likely percentage of those groups not in school based on media reports and available data.

If even one in four students with disabilities, English learners, students in foster care, migrant students, and homeless students have been shut out of education for months, that adds up to over three million students, as if the entire school-aged population of the state of Florida dropped out of school. If only one in 20 of America’s most educationally marginalized students did not access any education in the midst of the ongoing pandemic, the number of students out of school would create the second largest school district in America, almost 620,000 students. The millions of students who face the greatest barriers to an education span age ranges and geography, but are more likely to live in urban areas.

The consequences for these students’ education and well-being are not marginal concerns: They are an emergency.

Why aren’t students attending school?

The reasons that as many as 3 million young people aren’t attending school are complex. But the scale is massive. It is well known that many students lack the devices and Wi-Fi access to fully participate in virtual instruction. National analyses suggest that before the pandemic, as many as 16.9 million children lacked home internet access, and 7.3 million children did not have a computer at home. These students disproportionately live in low-income households and are more likely to be Black, Latinx, or Native American. Gaps in technology access persist today even after efforts this spring in many school districts to directly distribute technology or reach students in other ways.

But the students we’re focusing on here need much more than a laptop. For example:

  • Many English learners and students with disabilities have no good options to receive the educational accommodations and services to which they are legally entitled, making learning inaccessible.

  • Children in foster care and children experiencing homelessness encountered barriers to education before the pandemic, and this year, instability, confusion, and poor communication are rampant for them.

  • Some young people have transitioned to work, both formal and informal, while others are providing full-time primary care and learning support for other children, like younger siblings, in their homes.

What will happen if these students don’t get back to school?

The long-term consequences of this crisis are difficult to estimate without seeming hyperbolic. Once a student leaves school, it is difficult to reenter. One study of a large, urban district found that two-thirds of high school dropouts never reenrolled, and among those who do, about half drop out again. Circumstances that might push a student out of school today are very different, but even if all of the currently missing students return to school as soon as they are allowed to do so, months of missed opportunities for learning could mean permanent setbacks.

We know that other, much shorter disruptions to learning can have long-term consequences on students’ knowledge and skills, achievement, path to college and career, and lifetime wages. Absent a remarkable and long overdue investment of resources and interventions, students who were already at an educational disadvantage will lag even further behind their peers who had access to education during this time. These educational effects of the pandemic will be compounded by the effects of increased job loss, housing instability, and adverse health consequences of COVID-19, all of which fall harder on low-income families. The sum total of this disruption will echo through generations and communities, and there is no clear end point in sight.

How did you estimate 3 million missing students?

The lack of any reliable attendance data to understand how many and which American students have functionally disappeared from school for the past seven months is itself evidence that this crisis is not well understood and needs greater attention.

We arrived at an estimated 3 million students using two broad methods. (Enlarge the infographic at right for more.)

First, we monitored media reports and district and state survey data for indications of which student groups were having difficulty engaging in or accessing education in spring and fall 2020. These sources led us to identify groups at highest risk of not having access to remote education and estimate a percentage range of students within those marginalized groups.

For example:

  • Detailed data on spring engagement released by the Los Angeles Unified School District showed that English learners, students with disabilities, students experiencing homelessness, and students in foster care in middle and high school were all less likely than their peers to log into the district school platform in the spring.

    • 6%-10% of students in these subgroups did not log in at all from March to May, and an additional 10%-15% logged in but did not view any educational materials or complete any assignments.

  • In Boston, reports suggest approximately 20% of students did not participate in virtual school in the spring.

  • In Memphis, data from Shelby County School District indicated that 3% of students had not logged into virtual school or claimed a device this fall.

  • In Chicago, 10%-16% of students missed the first few days of virtual school this fall.

  • Across the country, school districts are seeing markedly lower enrollment rates, especially in kindergarten. In Washington state, kindergarten enrollment dropped by 14%.

Next, we compiled existing federal data sources to estimate how many students fell into higher-risk groups in every state and nationally, using data sets from the U.S. Department of Education and the Department of Health and Human Services on the total student population and the following subgroups:

  • Students in foster care

  • Students experiencing homelessness

  • English learners

  • Students with disabilities (ages 6-21)

  • Students eligible for the Migrant Education Program

We did not include racial subgroups or students eligible for free and reduced-price lunch in our estimate. Although students of color and low-income students are undoubtedly more likely than their peers to experience educational disruptions during this time on top of the ongoing effects of classism and racism, we decided these groups were too large, and overlapped too much with the other groups of interest, to add meaningful estimates.

We also weighed the fact that a disproportionate number of students in higher-risk groups attend schools in urban districts that are less likely to offer in-person school options because of facilities challenges, lack of resources, or other reasons. Based on this information, we estimated that between 10% and 25% of students in most marginalized groups are likely to have had minimal or no educational access since schools shut down in March.

A raw sum of the super subgroup of students for which we have data is approximately 13.3 million students. But these groups overlap. For example, a student might be an English learner with a disability who is experiencing homelessness. No individual-level, deduplicated student data is available, so we accounted for estimated overlap between groups wherever we could to arrive at an estimate of approximately 12.4 million students. Ten to twenty-five percent of that 12.4 million would be between 1 and 3 million students.

To see estimates of marginalized students for all 50 states and D.C., click here. For deduplication methods, discussion of other known limitations of our methods, and data sources, see the appendix.

Our estimate of 3 million students includes both the missing — children who are offline and hard to find but would reengage in school if given the opportunity — and the gone — children and youth who have made a transition away from school engagement in ways that could be permanent.

What should leaders do about these missing students?

The answer is not to open school buildings indiscriminately, especially in communities with high levels of COVID-19 transmission and districts with insufficient resources, facilities, and processes to protect students, families, and staff.

While there are no easy solutions to this problem, there are obvious next steps.

First and foremost, public leaders at every level of government must develop, implement, and enforce coherent, science-based strategies to mitigate the effects of the COVID-19 crisis, stomp out community transmission, and eventually enable safe community reopenings, prioritizing a safe return to school for all. Schools, children, and families alike cannot solve the underlying problem that has disrupted education for every child in the country. We are education policy professionals — not doctors or epidemiologists — but it is clear that there are no true solutions to our current educational crisis without addressing the public health crisis constraining many of our options both in and out of school.

There are four additional things that public leaders can do to support the millions of students who face the greatest barriers to an education:

  1. School districts and states must collect and report disaggregated attendance data in real time and internally identify and follow up with individual students who went missing between spring and fall.

  2. Schools, districts, and communities must develop and implement attendance intervention strategies that start with an informed understanding of students’ unmet needs — and avoid punitive approaches that exacerbate those needs.

  3. Social service agencies, telecommunications companies providing Wi-Fi access, and community-based organizations must work with local and state education agencies to develop coherent and integrated plans that meet the unmet needs of each community’s most vulnerable children. Children and families in crisis often receive uncoordinated help from too many sources, leading to more frustration and gaps in services — and this crisis has been no different.

  4. State and federal government leaders need to provide guidance, funding, and resources for schools and other social services to support these plans. Safe returns to school, intensive educational and child welfare interventions, and coordination between schools, districts, and social service agencies in the current moment all require additional funds to implement effectively. Expecting schools and other public and nonprofit service agencies to do more with less could leave vulnerable students worse off than they already are.

See Bellwether’s Resources to Support Special Populations During COVID-19 for more.

There is not enough public recognition of the serious challenges facing America’s most vulnerable students at this moment or of the consequences if millions continue to be disconnected from schools and other support systems indefinitely. Not only are educational futures at stake, but in some severe cases, students’ basic safety and well-being are in jeopardy.

There remains tremendous work to be done to both fully understand the scope and scale of this problem and to take steps to intervene. We look forward to others’ efforts to complement or build on this estimate as well as alternative ways to conceptualize this problem. If you are interested in a deeper look at this question, please contact the authors.

Appendix: Data Methods, Limitations, and Sources

There are many known limitations to the data we used to estimate the total size of the most marginalized student groups nationally and in each state.

Our model includes homeless students, students in foster care, special education students, English learners, and students eligible for migrant education services. There are other student groups known to have high barriers to accessing education for whom there is no reliable data at the state and national levels. To start, there is no reliable single data set for young people who are pregnant or parenting, a life experience correlated with dropping out of school. We also could not identify the count of current high school dropouts: students aged 18 and below who are not enrolled in school and do not have a high school credential. Additionally, all data is from 2017-18, and student populations have grown since then.

We also could not find consistent, reliable estimates of school-aged children who were already hidden and who are some of the most vulnerable. For example, we could not find state-by-state data on school-aged children who are undocumented, nor could we get close enough to the number of school-aged children who have been detained, deported, or expelled at the border. In the past, many of those students would have been processed through our immigration systems and released to family or sponsors in the U.S. and then would have been entitled to enroll in local public schools. We also could not find any state-by-state estimates of runaway youth or unaccompanied homeless children 18 and younger, an experience that is more likely to be had by LGBTQ youth.

The exact formula we used based on available data to arrive at this estimate was: homeless students + (foster care, not SPED) + (foster care, YES SPED) + (SPED NOT EL, NOT foster care) + (SPED, YES EL) + (EL not sped not migrant) + (EL, migrant) + (migrant, not EL). The graphic below illustrates an example of this method applied to the state of Alabama:

Furthermore, we know that there are young people who have experienced or are experiencing multiple disruptive experiences, but estimates on the rates of overlap between these populations are at best a moving target. Some considerations:

  • One dimension of overlap, the percentage of students in foster care receiving special education services, is derived from a national estimate of overlap rather than state-level data. All other overlap estimates are drawn directly from federal datasets.

  • There is no data on the overlap between students experiencing homelessness and other student groups of interest.

  • The overlap of more than any two groups is also unknown, however, we believe those numbers are comparatively small.

  • The estimate of students for Wisconsin was calculated differently because the U.S. Department of Education has suppressed data on Wisconsin students in special education.

Data sources

Total student population National Center for Education Statistics. “Table 203.20. Enrollment in Public Elementary and Secondary Schools, by Region, State, and Jurisdiction: Selected Years, Fall 1990 Through Fall 2029.” Digest of Education Statistics. Accessed October 19, 2020.
Students in foster care The Annie E. Casey Foundation. “Children in Foster Care in the United States.” Kids Count Data Center. Updated March 2020.
Students experiencing homelessness United States Department of Education. “McKinney-Vento Act: Homeless Students Enrolled by State.” Ed Data Express. Accessed October 19, 2020.
Students eligible for the Migrant Education Program, 2018 United States Department of Education. “Eligible Migratory Students Served by Type of Service.” Ed Data Express. Accessed October 19, 2020.
English learners National Center for Education Statistics. “Table 204.20. English Language Learner (ELL) Students Enrolled in Public Elementary and Secondary Schools, by State: Selected Years, Fall 2000 Through Fall 2017.” Digest of Education Statistics. Accessed October 19, 2020.
Students with disabilities United States Department of Education. “IDEA Section 618 Data Products: State Level Data Files.” Office of Special Education and Rehabilitative Services. Accessed October 19, 2020.
Students with disabilities who are not English learners United States Department of Education. “IDEA Section 618 Data Products: State Level Data Files.” Office of Special Education and Rehabilitative Services. Accessed October 19, 2020.
Students with disabilities who are English learners United States Department of Education. “IDEA Section 618 Data Products: State Level Data Files.” Office of Special Education and Rehabilitative Services. Accessed October 19, 2020.
English learners who are not students with disabilities Estimated by subtracting the number of students with disabilities who are English learners from the total number of English learners.
Children in foster care receiving special education Estimated as 33% of the total foster care population based on the National Working Group on Foster Care and Education, which cites two studies estimating that between 36% and 47% of youth in foster care receive special education services.

Legal Center for Foster Care and Education. “National Working Group on Foster Care and Education.” National Working Group. Accessed October 19, 2020.

Children in foster care not receiving special education Estimated as 66% of the total foster care population based on the National Working Group on Foster Care and Education, which cites two studies estimating that between 36% and 47% of youth in foster care receive special education services.

Legal Center for Foster Care and Education. “National Working Group on Foster Care and Education.” National Working Group. Accessed October 19, 2020.

Students eligible for the Migrant Education Program who are English learners United States Department of Education. “Eligible Migratory Students Served by Type of Service.” Ed Data Express. Accessed October 19, 2020.
Students eligible for the Migrant Education Program who are not English learners United States Department of Education. “Eligible Migratory Students Served by Type of Service.” Ed Data Express. Accessed October 19, 2020.

Acknowledgments & Credits

The authors would like to thank ESC for infographics and video creation, Super Copy Editors for proofreading support, Michelle Lerner for outreach assistance, and Giant Rabbit for layout customization. A number of Bellwether colleagues made this work possible: Jennifer O’Neal Schiess, Andy Rotherham, Tanya Paperny, Alyssa Schwenk, and Lindsay Ferguson. Thank you additionally to Carnegie Corporation of New York for support. Any errors are the responsibility of the authors alone.

You can contact the authors of this work to offer feedback or suggestions here.

Missing in the Margins 2021: Revisiting the COVID-19 Attendance Crisis

Bonnie O’Keefe
Hailly T.N. Korman
Indira Dammu


  1. Introduction
  2. Data
  3. Recommendations
  4. Appendix
  5. Acknowledgments


Available data on 2020-21 enrollment, attendance, and engagement suggest massive missed learning opportunities for students.

2020-21 was a tumultuous school year for students, along with their families, educators, and school leaders. After many predictions and debates about the potential consequences of widespread pandemic-related disruptions, emerging evidence from spring 2021 assessments and other studies of academic achievement show cause for alarm: Students mastered less academic content over the course of the year, test scores went down, and course failure rates rose. There are caveats to this data and many questions still unanswered, but it is clear that disrupted education had a negative effect, especially for students with greater educational needs.

Three key indicators comprise this update analysis, following up on the focus on attendance and opportunities to learn highlighted in 2020’s “Missing in the Margins” analysis:

  • Enrollment: Nationwide enrollment in public pre-K through 12 schools dropped by more than 1.3 million students between 2018-19 and 2020-21, a decline of 2.7% from 2018-19 enrollment.
  • Attendance: Nearly all available attendance evidence suggests widespread reduced opportunities for learning, but changing definitions of attendance complicate the picture.
  • Engagement: Engagement is difficult to measure, but evidence suggests that reduced engagement during remote instruction was at least as big of an issue as attendance and enrollment.

[Click here for the data we looked at nationally, at the state level, and in select school districts.]

Unfortunately, more than a year and a half after schools began to close their physical doors in response to COVID-19, many fundamental questions about all three indicators are still unanswerable due to data limitations at the national, state, and local levels.

In the 2020-21 school year, common barriers to enrollment, attendance, and engagement included lack of access to technology and high-speed internet during periods of virtual school, lack of access to accommodations and services to support students with disabilities and English language learners, miscommunication with students and families, and competing responsibilities and life needs, such as health emergencies, child care, housing instability, or work.

As students and schools return to more “normal” instruction, students with severely disrupted education pathways remain in dire need of support in and out of school.

As in 2020, a focus on those students most deeply affected by the pandemic is essential, through partnerships between schools, social service agencies, and community organizations. Our recommendations focus on the policy, practice, and resource allocation decisions that will position students at the margins to fully achieve their goals in school and in life.

2. Data

Enrollment: Nationwide enrollment in public pre-K through 12 schools dropped by more than 1.3 million students between 2018-19 and 2020-21, a decline of 2.7% from 2018-19 enrollment.1

This is a stark change. Before the pandemic, pre-K through 12 public school enrollment across the country had been on a slight upward trajectory: Enrollment last grew by 3% over a much longer 10-year span from 2009 to 2019.

  • 79% of that decline, equivalent to about 1 million students, was in the elementary grades.
  • Enrollment drops were especially concentrated in kindergarten, which declined by more than 9%. Kindergarten enrollment in some districts and schools dropped by over 20%.
  • Nationwide, enrollment in grades 9-12 did not decline at all. This could still include significant student-level instances of dropouts, disruptions, and other difficulties.
  • White (-5%) and Black (-3%) student enrollment declined while Asian and (+0.2%) Hispanic (+0.43%) student enrollment held mostly steady.
    • Asian Americans and Hispanic Americans are the fastestgrowing demographic groups in the U.S., which may account for slight enrollment growth among these student populations.
    • Survey evidence suggests rapid growth in homeschooling, especially among Black families last year, although the definitional lines between homeschool, virtual school, and microschools/learning pods may have shifted over the past year with more formal and informal options to support intentional remote learning.
  • There is not yet national data available on enrollment changes specific to marginalized student groups, including English language learners, students with disabilities, students in foster care, students experiencing homelessness, and migrant students. Emerging state and local reports suggest enrollment declines were greater among these groups.

Enrollment drops were not evenly distributed among states and school districts. As shown in the graph below, states and territories with the greatest declines in enrollment spanned the country and included Puerto Rico, Oregon, and Mississippi.2 Only a handful of states saw no change or enrollment growth, including the District of Columbia, Utah, and North Dakota.

Bottom 5 and Top 5 State- and Territory-Level Student Enrollment Changes, 2018-19 to 2020-21

Bar graph of bottom 5 and top 5 state- and territory-level student enrollment changes, 2018-19 to 2020-21

Looking at this variation in the context of enrollment trends in the past five years, states and territories that were already losing enrollment before the pandemic generally declined at an even faster rate, states that had been holding steady or slightly growing also dropped, and states that had been growing quickly slowed or stopped their growth trajectory. Among the states and territories above with the largest enrollment drops, Puerto Rico, Mississippi, and West Virginia had all seen enrollment decline in the five years before the pandemic, but 2020-21 accelerated that trend significantly.

[Click here for more context on enrollment declines in Puerto Rico.]

At the school district level, districts offering only remote instruction in fall 2020 experienced relatively higher levels of dis-enrollment. In larger, urban school districts, which are more likely to serve marginalized populations of students, enrollment decreases tended to be even larger.

Dis-enrolled students went down a variety of paths with very different educational implications. These paths included private schools, learning pods, child care, homeschool, the workplace, or simply home. Many of them still had access to learning in different settings. Widespread enrollment drops at the state and national level suggest that large numbers of students were not simply moving from one district to another — they were moving out of the public school system entirely. Our colleagues estimated that approximately 600,000 students remain unaccounted for in combined estimates of homeschool, private school, and learning pod enrollment.

Attendance: Nearly all available attendance evidence suggests widespread reduced opportunities for learning and thus greater student learning needs going into 2021-22.

Enrollment is only one indicator of whether a student actually has access to learning and whether they had unmet educational needs throughout the year. For those students who remained enrolled in school, our next question should be: Did they attend school consistently, whether virtually, in person, or both?

Before the pandemic, research associated chronic absenteeism with lower academic achievement, increased high school dropout rates, and long-term adverse outcomes in health. Most states include measures of attendance and/or absenteeism as part of their state accountability framework and report on these metrics at least annually. But attendance data for 2020-21 is still sparse. Only 10 states from our scan of 50 states, D.C., Puerto Rico, and the Bureau of Indian Education reported attendance data as of September 2021.3

Connecticut stands out for extensive reporting on monthly student attendance, including by subgroup, and learning conditions in different districts over the course of 2020-21. Between 2019-20 and 2020-21, average overall attendance dropped about two percentage points, but attendance among “high need” students dropped by four percentage points. Chronic absenteeism was lower among Connecticut school districts offering predominantly in-person instruction and higher among English language learners, low-income students, Black students, Hispanic students, and students with a disability.

Where states aren’t reporting data, districts are still keeping track. In a representative survey of over 550 district leaders in 46 states, an analysis by the American Institutes for Research showed that 18% of districts reported that attendance was substantially lower in fall 2020 than in fall 2019. Rates of absenteeism were higher in early grades and higher among districts primarily offering remote instruction.

Another way to examine attendance is with data from outside vendors and support organizations contracting with schools across the country:

  • PowerSchool, a company that helps schools track grades and attendance, showed that 2020 attendance fell in about 75% of 2,700 school districts, dropping by 1.5% on average each month.
  • EveryDay Labs, a company that works with 2,000 schools to improve attendance, found a sevenfold increase in students missing more than half of the school year, a level of chronic absenteeism that is more than double the definition of “severe.” Absenteeism was greater among younger students, English language learners, and low-income students.

It’s important to note that attendance rates for 2020-21 may not be comparable to pre-pandemic in-person attendance because states and districts altered their definitions of attendance for remote learning. About 70% of U.S. school districts offering remote instruction defined attendance with some kind of once-per-day touch point with school, which could mean an interaction with a teacher, logging in to a learning management system, or submitting a daily assignment. Attendance rates even improved in a few places, potentially because the definition of attendance changed and could constitute as little as logging into email.

Engagement: Engagement is a critical measure of access to learning but is difficult to measure. We could find no states or districts with a non-attendance engagement metric that was comparable before the pandemic and for 2020-21; however, evidence suggests that it is at least as big of an issue as attendance and enrollment.

For example:

  • In January 2021, student participation in online math coursework was down more than 10% from January 2020. Participation among low-income students declined even more from 2020 to 2021.
  • In March 2021, a U.S. Census pulse-check survey of households found that 28% of respondents said their children spent much less time on learning activities compared to before the pandemic.
  • Some districts have collected and published engagement data from student surveys and online learning platforms.

Regrettably, these reports are too isolated to make pre-pandemic comparisons, but where they are available, they suggest that the most educationally marginalized students were less engaged than their peers in remote learning settings.

Although we don’t have as much information as we would need to have to make a more precise national estimate of students without access to education for long periods of time amid the pandemic, students cannot wait. The 2021-22 school year is well underway. Students need educators, policymakers, and other community-based practitioners to help them recover now.

3. Recommendations

Changes to practices, policies, and resource allocation can help support all students — especially those with limited access to learning opportunities in 2020-21.

Students who missed out on learning will still be expected to meet their state’s grade-level standards, graduation requirements, or college admissions threshold to ultimately achieve college and career readiness. The near-term consequences of COVID-19 school disruptions will have long-term impacts on students unless there is a more coordinated, concerted effort to identify and address these needs for students most severely affected.

The scale and severity of learning disruption is not a problem that individual families or schools can solve in isolation. Families and children are the ones experiencing this problem, not the ones who have created it nor the ones who should be tasked with fixing it. Leaders in policy, education, and social service systems must create environments where learning recovery and acceleration happens.

These leaders should begin with good practices, ensure that policies facilitate those good practices, and provide resource allocation to support change.

1. Implement practices at a local school and community level that can accelerate learning for students with differing needs and experiences over the past year and a half.

For the vast majority of students, the best course of action will be to return to school buildings for regular in-person learning, in classrooms equipped to provide academic acceleration and socio-emotional support.

Young people and families experiencing more significant missed learning are in the best position to share their needs and goals with a supportive adult with whom they have a trusting relationship. Identifying, listening to, and engaging the young people and families who experienced severe disruptions in 2020-21 should be a shared local responsibility among schools, other youth-serving agencies, and community organizations. A collaborative case-management approach for students with greater unmet needs would be resource intensive, but for those young people the costs of support now are far smaller than the long-term costs of a permanently disrupted education.

2. Change state and federal policies that constrain or facilitate those practices.

Many educators, schools, and districts understand their students’ needs and barriers to learning, but the national story is one of 13,000 fragmented district-level responses. And communities face barriers to implementing good practices from state and federal policies that constrain action or create unnecessary burdens.

First and foremost, reducing the spread of COVID-19 through vaccination, masks, and other public health measures will make learning more accessible for all. There is also an opportunity to create better educational opportunities for students who were already marginalized before COVID-19, which will require bolder, focused action in the following areas:

  • Accountability: More than a year’s worth of disrupted learning has created new tensions for leaders committed to maintaining high standards for all students and creating effective support and incentives to meet those standards. At the outset of the pandemic, many states paused policies on grade retention, high school graduation, and accountability systems. This was a temporary solution to what is now a multiyear problem that demands clear policy action. Now is the time for federal and state leaders to clearly communicate a set of goals for students in order to create predictability for schools and set a course for learning acceleration.
  • Systemic support for learning interventions: Over the course of the past year and a half, too many schools and districts have struggled to address students’ needs with relatively little support or guidance. State and federal policies should strengthen their focus on tiered systems of support in schools to make effective practices easier to implement. Importantly, policies should incentivize collaboration and partnership with other child-serving systems, organizations, agencies, and community partners.
  • Additional time for learning: One of the most important ways to make up for missed learning opportunities will be with additional learning time, when that time is used effectively. This might take the form of extended school days and years, summer and out-of-school tutoring and supplementary learning, or extending time in school for older students before a transition to postsecondary learning. All these ideas hold promise, but they will only be possible with enabling policies and a focus on quality in implementation and accountability.
  • Data collection and transparency: Most state and district policies related to enrollment, attendance, and engagement data have not kept up with the new demands of the COVID-19 era. Without more precise, current, and disaggregated data shared publicly, districts and states are stuck working in isolation from one another and from other child-serving systems. This inhibits collaboration and keeps the public in the dark.

3. Allocate resources to support the change that needs to happen.

Effective practices and policies to recover from learning disruption and to collaborate across agencies take money and time to implement. Federal stimulus funding has provided schools and communities with a large and temporary infusion of over $120 billion in funds that can support COVID-19 recovery and learning acceleration, targeted at schools serving low-income students.

This could be a game-changing funding boost. But it still might not be enough, or sustainable enough. Districts and states must create spending plans that include longer-term funding streams to equitably and effectively address student needs.

First, many of the practices for success are grounded in relationships and direct services. Those call for investments in new staff and staff training, which might be at risk if positions are supported by time-limited funds. Collaboration, especially with community-based organizations, could help expand schools’ capacity for non-instructional support without expanding in-school staff. While federal stimulus funds are flexible enough to allow for this kind of collaboration, other state, local, and federal funding streams are not. Policymakers must start thinking now about how to effectively fund and encourage deep collaboration between schools and communities in the long term.

Second, states can set districts up for longer-term success and sustainability by taking this moment to address inequities and inadequacies in their state education finance systems. State and local funds account for a much greater share of education funding than federal funds, with the potential for long-term stability and impact once stimulus dollars expire.

Finally, one of the most important resource allocation questions for policymakers to tackle is the long-term plan for student success. Even before the pandemic, too many young people didn’t have the systemic support they needed to achieve their educational goals — and one disruption could throw them off track through adulthood. Supporting all young people throughout their educational pathway so that they can thrive in adulthood must be a priority as we continue to learn more about student learning progress and unmet needs.


Click here for the data we looked at nationally, at the state level, and in select school districts.

We began our analysis by compiling enrollment data at the national, state, and district levels. Districts included in our analysis were the top 20 largest school districts by student enrollment, as well as districts participating in the NAEP Trial Urban District Assessment, 23 districts serving over 4.8 million students in total — about 10% of the nationwide K-12 population.

Enrollment: The primary source for enrollment data was the Common Core of Data (CCD) from the National Center for Education Statistics. We looked at three years of enrollment data — 2018-19, 2019-20, and 2020-21 — and analyzed enrollment figures based on student race/ethnicity and grade levels. All states and U.S. territories were included in national numbers. All states, plus D.C. and Puerto Rico, are included in state analysis for consistency with 2020 scope of analysis. The 2020-21 dataset is a preliminary estimate; the final dataset is expected to be published in spring 2022. The 2020-21 CCD dataset was missing Illinois data because the state did not meet the deadline for data submission. We used enrollment data as reported on the Illinois Department of Education website for the 2020-2021 academic year, as of October 1, 2020, which is similar to the CCD timeline. We could not find reliable estimates of state and national enrollment figures for other student subgroups — young people in foster care, English language learners, students experiencing homelessness, and migrant students.

Attendance: We explored any available data on attendance at the state and district level for 2018-2019 and 2020-2021. We looked at state department of education websites as well as school district websites to collect any publicly available data on average daily attendance, average daily membership, attendance rates, and chronic absenteeism. Most states did not report any attendance data publicly for 2020-2021. Some of the states that did report attendance data for 2019 and 2020 cautioned about its use in analysis because of COVID-19 and the impact of remote and/or hybrid instruction on attendance metrics and data collection. In addition to looking at state and district websites, we also explored any other estimates on attendance from national groups including Attendance Works, FutureEd, and the Center on Reinventing Public Education. The American Institutes for Research published additional results from a nationally representative survey of district leaders on attendance rates, attendance monitoring, outreach methods, and enrollment loss.

Engagement: We looked for publicly available data on student engagement at the state or district level. We were interested in any data about how schools were measuring student engagement in a virtual setting, including whether students were logging into virtual platforms. These data were rarely reported by states or districts. For example, New York City Public Schools and Los Angeles Unified published some data on student participation, but these districts were the exception rather than the rule, and there were no comparable metrics over time to see how engagement might have changed during the pandemic.

We supplemented these data with a scan of the available published research and reporting based on 2020-21 data on these topics and the student subgroups of greatest interest. This included district, parent, and school surveys; attendance data; curriculum or test vendor data; and/or case studies.

Acknowledgments & Credits:

The authors would like to thank Super Copy Editors for proofreading support, Michelle Lerner for outreach assistance, and Giant Rabbit for layout customization. A number of Bellwether colleagues made this work possible: Jennifer O’Neal Schiess, Alex Spurrier, and Amber Walker. Thank you additionally to Carnegie Corporation of New York for support. Any errors are the responsibility of the authors alone.

You can contact the authors of this work to offer feedback or suggestions here.


  1. See methods below for more detail on this estimate and other national estimates using June 2021 release of 2020-21 Common Core of Data (CCD) preliminary files. Go back
  2. We also examined data from the Bureau of Indian Education, which had the largest enrollment drops among the entities with available data, but we did not feel that this was fairly comparable to states. Go back
  3. Our scan prioritized chronic absenteeism, but we also looked for attendance rates, average daily attendance, average daily membership, or any other available attendance data in 2018-19 and 2020-21. Go back

More from this topic

Thank you! Your subscription has been confirmed. You'll hear from us soon.