School reopenings in the wake of the coronavirus pandemic are an ongoing, imperfect work in progress, and the science that school leaders and parents are being encouraged to follow is evolving. In mid-February 2021, NBER released a study conducted by seven scholars, including Dan Goldhaber from the University of Washington and Katharine Strunk and Scott Imberman from Michigan State University, that provided up-to-that-minute data on the extent to which in-person schooling contributes to the spread of Covid-19. It is a large, detailed, and impressive piece of work which adds important pieces to this vital puzzle.
The research team uses data on reported Covid cases in Michigan and Washington State as collected by the CDC and the respective state health agencies, as well as district-level information on learning modalities as reported by each state’s department of education via monthly surveys administered to districts. The latter data include not only the modality offered (fully in-person, hybrid, or remote-only schooling) but the approximate share of students receiving each of these different modes of instruction. Their first model—the “sparse model,” since it uses few controls—shows that in-person modalities (fully in-person and hybrid) are positively correlated with increased Covid cases—more so in Michigan than Washington.
The researchers then use a wide variety of other community variables to build more complex models designed to account for non-schooling risk factors for Covid-19 spread. Among them are the share of individuals in a county who report “always” wearing face masks in public as of July 2020, the share of people who stay at home as calculated by the U.S. Department of Transportation, the share of the 2016 presidential election vote for Donald Trump in each county (a measure of political leaning), county unemployment rates, the share of the population that attend county public schools, the share who are age sixty-five or older, the share who live in a nursing home, and so on. Once they allow the effects of modality to differ according to the level of community spread and account for the factors above, in-person schooling is not associated with increased spread of Covid—at least in counties reporting low levels of pre-existing Covid cases.
That result changes in counties reporting moderate to high pre-existing Covid rates. For example, Michigan districts offering in-person learning show increased Covid-19 spread for daily average case counts over twenty-one cases per 100,000 population. Still, the researchers are cautious because their district and county fixed-effect models can’t pick up every unobserved factor that may be affecting modality offerings or the choice to attend in person that are also related to Covid spread in communities.
They show, too, that the amount of uptake matters. When examining data on the proportion of students actually attending school in person, they find that the results in Washington are driven mainly by school systems with relatively high proportions of students—over 75 percent—attending in person. Yet in Michigan, compared to the fully-remote option, they find that Covid spread may be reduced even more in districts where small percentages of kids are in person (25 percent or less). This feels counter-intuitive, but could be an indication that there are simply too many out-of-school variables to collect and model. To put a finer point on it, the estimated effect of in-person schooling on Covid spread may depend on what students and staff in the counterfactual of no in-person schooling are doing. We can’t assume fully-remote learning is safer, since some remote families are participating in learning pods, others may be hiring babysitters and nannies who watch several kids, or families may be involving kids in other social activities because they are missing school. Any of these activities could serve to expose kids and adults to Covid as a direct result of the imposition of a learning modality meant to keep them safe.
Their bottom line finding—that in person schooling has a significant impact only at higher pre-existing levels of spread—seems in line with earlier guidance. But we also know that case rates are a function of testing as well as incidence. And as of late, some of the guidance is saying that it is safe to attend schools even when spread is higher in a community assuming other safety protocols are in place. While many early fears regarding Covid-19 spread in schools—kids won’t wear masks, social distancing is impossible in the lunch room, etc.—have not come to pass, others persist. This research is fairly positive regarding the safety of reopening schools to in-person learning. Its data, however, predate widespread vaccination, which could add even more to the evolving picture. Still, given what we know at this juncture, a combination of moving forward on reopening while exercising abundant caution seems to be the way forward.
SOURCE: Dan Goldhaber, Scott A. Imberman, Katharine O. Strunk, Bryant Hopkins, Nate Brown, Erica Harbatkin, and Tara Kilbride, “To What Extent Does In-Person Schooling Contribute to the Spread of COVID-19? Evidence from Michigan and Washington,” NBER Working Paper #28455 (February 2021).
