In-person schooling and community spread of Covid-19

Dan Goldhaber, Scott A. Imberman, Katharine O. Strunk, Bryant Hopkins, Nate Brown, Erica Harbatkin, Tara Kilbride 04 March 2021



As a consequence of the Covid-19 pandemic, virtually all public schools in the US closed their doors for in-person instruction for at least some period of time in spring 2020, and many remained closed in the 2020-21 school year. Whether schools are open for in-person (versus remote or hybrid) instruction has been a contentious policy issue. Indeed, the ability of students to attend school in person has become a political flashpoint in both Europe and the US. In the US, President Biden has committed to having the majority of K-8 public schools open for in-person instruction by the end of his first 100 days in office (Chalfant 2021).1 How to achieve that objective safely is obviously a health issue, but it is also a critical education and learning issue as numerous education policymakers, researchers, and stakeholders believe that online schooling is not working well for many students (Kuhfeld et al. 2020), particularly for those at the elementary level (Kuo and Nagel 2020, Wolfman-Arent 2021). 

In February, the US Centers for Disease Control and Prevention (CDC 2021) released guidelines and strategies for schools to consider in making decisions about whether to provide in-person schooling. The recommendations include consistent implementation of layered mitigation strategies (handwashing, physical distancing, contract tracing, etc.), tracking of community transmission indicators, and phased mitigation and learning modes based on levels of community transmission (e.g. cohorting and targeting earlier grades for return to in-person instruction). 

The CDC guidelines touch on how school openings may be connected to the level of existing Covid-19 cases in the communities where schools are located. For instance, they encourage administrators to “assess the level of risk in the community since the risk of introduction of a case in the school setting is dependent on the level of community transmission”. The CDC also notes that even if community rates are low, poor implementation of mitigation strategies in schools can increase risk of incidence and spread (CDC 2021). 

We studied the question of whether schools being open in-person (as defined by the instructional modality offered by schools, which we classify as either ‘in-person’, ‘hybrid’, or ‘remote’) appears to increase community spread of Covid-19 using data from Michigan and Washington (Goldhaber et al. 2021). 

In particular, we used information on instructional modality at the district level and Covid-19 cases at the county level to estimate statistical models relating Covid-19 case rates to prior school opening decisions. In these models, we account for a number of key factors — such as population density, reported mask wearing, daily mobility rates, and partisan vote share from the 2016 presidential elections — that might be related to compliance with social distancing and therefore the spread of Covid-19. We also allowed for the possibility that the effect of school modality on community spread of Covid-19 might depend on pre-existing local case rates, as noted in the CDC guidance and found in other research on mitigation strategies (Auger et al. 2020, Harris et al. 2021).

We find that when there are low levels of pre-existing Covid-19 cases in a community (measured in cases/day per 100,000 population), opening schools for in-person instruction does not appear to lead to additional community spread of Covid-19. However, when there are moderate or higher levels of pre-existing Covid-19 case rates in the community, in-person learning correlates with more case growth, even after accounting for prior case rates, mask wearing, and other community factors. 

Interestingly, what constitutes low, moderate, or high case rates varies by state as overall Covid-19 rates across the two states were different, and the Covid-19 levels at which we find that in-person schooling leads to additional spread also varied by state. This is illustrated in Figures 1 and 2. 

Figure 1 Predictions of Covid-19 cases per day per 100,000 in Michigan based on schools being in-person rather than remote

Figure 2 Predictions of Covid-19 cases per day per 100,000 in Washington based on schools being in-person (or hybrid) rather than remote

The figures display the estimated effects of school systems of being remote versus in-person in Michigan, and remote versus in-person or hybrid in Washington (there are too few districts in each category to get a reliable estimates, so we combine the two categories for Washington) on Covid-19 spread in the surrounding county. 

Taken together, the figures show that transitioning schools from remote to hybrid or in-person schooling has a limited impact on community Covid-19 spread when pre-existing county case rates are low. In Michigan, the results are only statistically indistinguishable from zero at the highest tails of the distribution (e.g. beginning at approximately 20 cases per 100,000 individuals per day). In Washington, the results become statistically significant at just under the 75th percentile of cases, or slightly fewer than five prior cases per 100,000 individuals. These findings are consistent with a similar study investigating school reopening and community hospitalisations. In fact, Harris et al. (2021) do not find a significant effect on Covid-19 hospitalisation rates when existing hospitalisation rates are at or below the 75th percentile.

It is difficult to pinpoint why we observe different thresholds at which school modality affects community spread. One possible explanation for this difference is that the state-level estimates depend on the contrasts in individuals’, schools’, and communities’ actions when students are and are not learning in person. In particular, when considering how in-person schooling may or may not contribute to community Covid-19 spread, one must consider both what is occurring inside school buildings and what is occurring when students are learning remotely. Remote learning environments may be at home, but students may also be learning in pods of several children, in day care centres, or interacting with children and adults who are not members of their immediate household. We cannot examine these conditions in our study, but it is possible and even likely that the conditions in both school buildings and in remote learning environments differ across states. Regardless of the reasons for the finding of differing thresholds, our results indicate that context appears to matter.

We also used school district estimates of the percentages of students attending schools in-person (as opposed to the modality of school offerings, as described above) to examine whether community spread is connected to in-person take-up. Here, too, our findings indicate that even as the share of students receiving in-person instruction grows, community spread does not appear to significantly increase unless pre-existing case rates are high in communities. Specifically, only when pre-existing cases are high (at the 75th percentile of pre-existing cases in the community), and when 76% of students or more are in-person (in the case of Washington) do we observe a statistically significant and positive trend associated with community spread. 

Our findings suggest that the link between community spread and school modality depends not only on whether schools are open for in-person instruction, but the overall Covid-19 rates in communities. Importantly, our findings are about community spread of Covid-19, not spread of Covid-19 in schools. Even in places with low case rates, it is likely that there is (and will be) some in-school spreading of Covid-19 when schools are in person. But this does not necessarily mean that having school buildings open is a less safe condition for communities given that if schools are closed, students and teachers may be engaging in other activities that contribute more to community spread than had they been physically in schools.


Auger, K A, S S Shah, T Richardson, D Hartley, M Hall, A Warniment, K Timmons, D Bosse, S A Ferris, P W Brady, A C Schondelmeyer and J E Thomson (2020), “Association between statewide school closure and COVID-19 Incidence and Mortality in the US”, JAMA 324(9): 859-870.

CDC (2021), “Operational Strategy for K-12 Schools through Phased Mitigation”, Centers for Disease Control and Prevention, 12 February.

Chalfant, M (2021), “Biden says goal is to reopen schools five days a week in first 100 days”, The Hill, 16 February.

CNN (2021), “Where schools are reopening in the US”, 11 February.

Goldhaber, D, S A Imberman, K O Strunk, B Hopkins, N Brown, E Harbartkin and T Kilbride (2021), “To What Extent Does In-Person Schooling Contribute to the Spread of COVID-19? Evidence from Michigan and Washington”, NBER Working Paper 28455.

Harris, D, E Ziedan and S Hassig (2021), “The effects of school reopenings on COVID-19 hospitalizations”, New Orleans, LA: National Center for Research on Education Access and Choice. 

Kuhfeld, M, J Soland, B Tarasawa, A Johnson, E Ruzek and K Lewis (2020), “How is COVID-19 affecting student learning? Initial findings from fall 2020”, Brown Center Chalkboard, 3 December.

Kuo, A and C Nagel (2020), “Remote school is putting kids under toxic stress Our pediatric patients are suffering. We need a way to open schools safely”, The Washington Post, 10 August.

Tate, E (2020), “As Districts Return to Remote Learning, Are They Any Better Prepared?”, EdSurge.

Wolfman-Arent, A (2021), “Philly’s youngest students are suffering the most academically, new data shows”, WHYY PBS, 5 February.


1 In early February, Burbio estimated that K-8 schools serving between 60% and 70% of US students were offering in-person schooling options. Burbio is a private company that has been tracking in-person, hybrid, and remote instruction throughout the 2020-21 school year ( Note also that Burbio is estimating school modality, not how individual students are electing to learn. Thus, the 60%-70% figure will overstate the percentage of students who are receiving instruction in-person as many schools that are in-person (or hybrid) also offer remote options (CNN 2021, Tate 2020).



Topics:  Covid-19 Education

Tags:  COVID-19, coronavirus, in-person schooling, school closure, school opening, community spread

Director, Center for Education Data & Research, University of Washington and Center for Analysis of Longitudinal Data in Education Research, American Institutes for Research

Professor of Economics and Education Policy, Michigan State University

Professor of Education Policy and Clifford E. Erickson Distinguished Chair in Education and the Director of the Education Policy Innovation Collaborative (EPIC) at Michigan State University

Post-doctoral Research Fellow, Education Policy Innovation Collaborative, Michigan State University

Research Manager, Center for Education Data & Research, University of Washington

Postdoctoral researcher, Education Policy Innovation Collaborative (EPIC), Michigan State University

Assistant Director for Research, Education Policy Innovation Collaborative


CEPR Policy Research