Online School vs In-Person School: What the Post-COVID Research Actually Shows

Five years after the largest unplanned education experiment in history, researchers have enough data to move beyond headlines. The COVID-era shift to remote schooling affected more than 50 million K-12 students in the United States alone, creating a dataset of unprecedented scale. The conclusions that are emerging are specific enough to be useful — and they are not what either remote-school advocates or opponents tend to claim.

The aggregate picture is concerning. The nuanced picture is more complicated, and more interesting.

Key Takeaways

• Post-2020 research consistently shows K-12 remote schooling produced learning losses equivalent to roughly half a school year in reading and a full school year in math for the average student.
• The losses were not equally distributed: lower-income students, English language learners, and students in early elementary grades experienced the largest deficits.
• A meaningful minority of students — those with school anxiety, certain learning differences, bullying histories, or unsafe home environments — performed better in virtual formats.
• Socialization deficits from remote schooling are real and measurable, but recovery timelines vary by age and pre-existing social skills.
• Hybrid models produce heterogeneous results: well-designed hybrid programs can outperform full in-person for some populations; poorly implemented ones can produce losses similar to full remote.

What the Research Shows: Academic Achievement

The Scale of COVID-Era Learning Loss

The NWEA (Northwest Evaluation Association) tracked reading and math scores for millions of students before, during, and after pandemic-era school closures. Their 2023 report, “Pandemic Recovery in Context,” found that as of spring 2023:

• Students in grades 3–8 remained, on average, 0.40 grade levels behind in reading and 0.45 grade levels behind in math relative to pre-pandemic trajectories.
• Recovery was occurring, but unevenly — higher-income districts were recovering faster than lower-income ones.

Separate analysis from researchers at Harvard’s Center for Education Policy Research, led by Tom Kane, examined test score data from 2.1 million students across 10 states. Published in 2022 in the Journal of Policy Analysis and Management, their findings were more alarming: students in districts that remained fully remote the longest fell an average of 0.5 standard deviations behind in math relative to students in fully in-person districts — an effect the authors characterized as equivalent to roughly half a school year. Reading gaps were smaller but still statistically significant.

A McKinsey analysis from 2022 drew on NAEP (National Assessment of Educational Progress) data and estimated the average K-12 student lost four to five months of learning across subjects. The analysis found that the learning loss was about one and a half times larger in math than in reading — a pattern consistent with prior research on summer learning loss and the relative difficulty of building math skills through informal activity.

Why Math Suffered More

The reading-math asymmetry appears consistently across datasets and mirrors findings from summer learning loss research. Language is practiced continuously through conversation, media consumption, and daily reading. Mathematical reasoning requires deliberate, structured practice — the kind that school provides and informal home environments generally do not replace.

When schools shifted online, the instruction students received varied dramatically. Many students received synchronous video lessons; others received asynchronous packets. But even in the better-resourced cases, the informal practice, peer discussion, and teacher feedback loops that consolidate mathematical thinking were severely reduced.

The Equity Dimension

The aggregate learning loss figures understate the experience of the students who lost the most. Research from Stanford’s Educational Opportunity Project, released in 2022, found that test score gaps between high- and low-income districts widened by an average of 20% during the pandemic. Students in the lowest-income quartile experienced learning losses roughly twice as large as students in the highest-income quartile.

English language learners and students with disabilities showed similarly outsized losses in most datasets — though as discussed below, the picture for students with disabilities is more complex.

Virtual School Outcomes Research: Who Does Better Online

The aggregate learning loss narrative is accurate but incomplete. A consistent finding across multiple datasets is that a meaningful subgroup of students performed as well or better in virtual formats.

Students with School Anxiety and School Refusal

For students experiencing school refusal or significant school anxiety, the shift to remote learning sometimes dramatically improved both attendance and academic engagement. A 2022 analysis published in School Psychology Review examined case studies of students with anxiety-based school refusal and found that remote formats removed the primary anxiety trigger — the physical school environment — without removing instruction.

The implications for parents of children with school anxiety or school refusal are significant: for these students, the “worse on average” finding from population-level research may not apply. The question is whether the anxiety is school-environment-specific or generalized.

Students with Certain Learning Differences

The picture for students with learning differences is heterogeneous and does not support simple generalizations. Some findings:

• Students with sensory processing differences (including some autistic students) reported reduced sensory overload in home environments, which some families and teachers associated with improved focus and engagement.
• Students with ADHD showed mixed results: some benefited from parent-mediated structure at home; others lost the external scaffolding that school environments provide.
• Students with learning disabilities that affect reading showed among the largest losses — likely because these students most depend on specialized in-person intervention (reading specialists, small-group pull-out instruction) that was hardest to replicate virtually.

A 2021 survey study of 500 families of students with disabilities, published in Exceptional Children, found that parents’ assessments of remote schooling were highly polarized: roughly a third found it significantly worse, a third found it roughly equivalent, and a fifth found it better for their child’s specific needs.

Students in Unsafe Home Environments

Research here is limited by data availability — students in genuinely unsafe home environments are underrepresented in academic datasets. But qualitative research and social work literature have documented cases where remote schooling placed students full-time in abusive home situations that school had previously provided daily relief from. This is among the most troubling documented harms of extended school closures. Referrals to child protective services declined during school closures in many districts, not because abuse declined, but because mandatory reporters (teachers) had reduced contact with students.

The Socialization Gap

Academic outcomes are one dimension. Socialization is another, and the research here is consistent: extended remote schooling reduced social skill development and peer relationship quality for most students.

A 2023 study in Child Development by Orben, Tomova, and Blakemore examined adolescent social development during COVID restrictions and found that peer relationship quality declined significantly for 11–15-year-olds during extended school closures — a period the authors described as particularly sensitive for social brain development. The effects were not uniformly persistent: students who returned to in-person learning recovered much of the ground lost within 12–18 months, particularly those who had well-developed social skills before the pandemic.

For younger students (ages 5–8), the picture is more concerning. Early elementary school serves as a primary social skill development environment. Students who spent kindergarten or first grade fully remote showed measurable social skill delays in subsequent grades that were slower to recover than academic gaps, according to 2023 reports from several large urban school districts.

What Hybrid Models Look Like in the Data

“Hybrid” learning covers an enormous range of implementations: alternating-day in-person, synchronous online with in-person labs, self-paced online with weekly in-person check-ins, and many others. The research reflects this heterogeneity.

The most consistently positive findings for hybrid models come from implementations where:

• Students have choice in format based on demonstrated need
• In-person time is used for collaboration, labs, and activities that genuinely require physical presence
• Online time uses high-quality synchronous instruction, not just asynchronous packets
• Teachers receive specific training in managing hybrid classroom dynamics

Research from Brookings Institution’s 2022 analysis of hybrid models during COVID found that “hybrid” as implemented in most districts during 2020–21 was not a pedagogically designed hybrid approach — it was emergency remote with partial in-person access. Districts that piloted intentional hybrid models (pre-pandemic charter schools, some online schools with physical learning centers) showed more positive outcomes.

Virtual vs In-Person School Outcomes: Research Summary

Outcome Area	Full In-Person	Full Virtual	Hybrid	Key Research Source
Math achievement	Baseline	−0.4 to −1.0 grade levels avg	Mixed (implementation-dependent)	Kane et al., 2022; NWEA 2023
Reading achievement	Baseline	−0.2 to −0.5 grade levels avg	Slightly better than full virtual	McKinsey, 2022; NWEA 2023
Attendance	Higher average	Lower average (except school-refusers)	Variable	Attendance Works, 2022
Socialization	Baseline	Significantly below baseline	Below baseline, partial recovery	Orben et al., 2023
Students with school anxiety	Often poor	Often better	Dependent on child	School Psychology Review, 2022
Students with autism/sensory needs	Mixed	Mixed (some better)	Mixed	Exceptional Children, 2021
Students with reading disabilities	Baseline	Significantly worse	Below baseline	NWEA SWD analysis, 2022
Low-income students	Baseline	Largest losses	Smaller losses than full virtual	Stanford Ed Opportunity Project, 2022

What to Watch for Over the Next 3 Months

• NAEP 2026 results (due late 2026): The next National Assessment of Educational Progress round will provide the clearest longitudinal picture of whether the pandemic learning losses have closed, persisted, or widened.
• State virtual school enrollment trends: Several states expanded permanent virtual school options post-pandemic. Watch for their outcome reports — these are more rigorous than COVID-era comparisons because they represent students who voluntarily chose virtual formats.
• Hybrid model research from Florida, Arizona, and Georgia: These states have the largest populations of students in intentional hybrid programs and are publishing 3-year outcome data.
• Socialization recovery tracking: Long-term studies of the 2020 kindergarten cohort (now in 3rd–4th grade) will begin producing peer-reviewed findings on whether early social development gaps have closed.

Frequently Asked Questions

Is online school always worse than in-person school? No. The research shows that, on average, online schooling during the pandemic produced meaningful learning losses compared to in-person schooling. But “on average” hides important variation. Students with school anxiety, certain sensory or social needs, or unsafe home environments often performed better in virtual formats. The question parents should ask is not “which is better on average” but “which format works for my specific child’s specific circumstances.”

How large were the COVID-era learning losses really? The most rigorous estimates suggest average students in fully remote districts fell between 0.4 and 1.0 grade levels behind peers in in-person districts, with larger gaps in math than reading. These are averages — some students experienced little loss; others experienced much more. Research consistently shows losses were larger for lower-income students, English language learners, and younger children.

Did students recover from pandemic learning losses? Recovery has been ongoing but incomplete. NWEA’s 2023 data showed students were recovering at roughly one-third to one-half the rate needed to fully close gaps within a few years. Higher-income districts recovered faster. As of the most recent data, most students remain below pre-pandemic trajectories in math, though reading gaps have largely closed.

What does research say about virtual school for kids with ADHD? The findings are mixed and depend heavily on home environment. Some students with ADHD benefited from parent-provided structure and one-on-one attention at home. Others lost the external scaffolding and movement breaks that school provides. There is no single answer — it depends on the quality of support available in the home environment.

Are hybrid schools a good middle ground? Well-designed hybrid programs show promise, particularly for adolescents. But “hybrid” in research covers a wide range of implementations. The key variables are: how in-person time is used, whether synchronous online instruction is high-quality, and whether students have individual supports. Emergency-era hybrid implementations (which is what most COVID-era data reflects) are not a good proxy for intentionally designed hybrid programs.

What did the pandemic reveal about which students benefit from virtual school? Research and clinical experience converged on several groups that sometimes showed better outcomes in virtual formats: students with chronic school anxiety or refusal, students with sensory sensitivities that made classroom environments difficult, students who had experienced bullying and found physical school threatening, and some students with autism who performed better in highly controlled home environments.

How does class size matter in virtual vs in-person settings? Class size effects appear somewhat different in virtual settings. In-person research consistently shows smaller classes benefit younger and lower-income students most. In virtual settings, the limiting factor tends to be teacher-to-student interaction quality — smaller virtual classes allow more individual attention, but the relationship is less linear than in physical classrooms because breakout rooms and asynchronous tools can substitute for some direct interaction.

What should parents ask when choosing between online and in-person options? Useful questions: Does my child have anxiety about the physical school environment? Does my child have learning differences that require in-person specialized services (speech therapy, reading intervention)? What is my child’s capacity for self-directed work? What oversight can I provide during the school day? How important is the peer social development that physical school provides for my child’s current stage?

About the Author

About the author Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

Sources

• Kane, T. J., et al. (2022). “The Impact of COVID-19 on Student Achievement and What It May Mean for 2021–22.” Journal of Policy Analysis and Management. Center for Education Policy Research, Harvard University.
• NWEA. (2023). “Pandemic Recovery in Context: Fall 2022 MAP Growth.” nwea.org.
• McKinsey & Company. (2022). “Unfinished Business: The State of K-12 Student Learning Recovery.” mckinsey.com.
• Stanford Educational Opportunity Project. (2022). “The Widening Gap: COVID-19 and Socioeconomic Disparities in Learning Loss.” edopportunity.org.
• Orben, A., Tomova, L., & Blakemore, S.-J. (2020). “The effects of social deprivation on adolescent development and mental health.” The Lancet Child & Adolescent Health, 4(8), 634–640.
• Attendance Works. (2022). “Attendance in the COVID Era.” attendanceworks.org.
• Brookings Institution. (2022). “Hybrid Learning: Lessons From the Pandemic.” brookings.edu.