There’s an interesting non-peer reviewed study that made the rounds earlier this week about the effect of school reopening on the spread of COVID-19. Here’s the abstract:

This paper examines the effect of fall 2020 school reopenings in Texas on county-level COVID-19 cases and fatalities. Previous evidence suggests that schools can be reopened safely if community spread is low and public health guidelines are followed. However, in Texas, reopenings often occurred alongside high community spread and at near capacity, making it difficult to meet social distancing recommendations. Using event-study models and hand-collected instruction modality and start dates for all school districts, we find robust evidence that reopening Texas schools gradually but substantially accelerated the community spread of COVID-19. Results from our preferred specification imply that school reopenings led to at least 43,000 additional COVID-19 cases and 800 additional fatalities within the first two months. We then use SafeGraph mobility data to provide evidence that spillovers to adults’ behaviors contributed to these large effects. Median time spent outside the home on a typical weekday increased substantially in neighborhoods with large numbers of school-age children, suggesting a return to in-person work or increased outside-of-home leisure activities among parents.

The authors used a ‘natural experiment’ that occurred in Texas*. Since different school districts opened at different times, they could assess the effects of opening. One key finding, based on mobility data, is that a considerable amount of the increase in COVID-19 was due to parents leaving the house because they didn’t have to stay home (they also might have believed things ‘were better’), and then contracting COVID-19 while they were out and about. To their credit, they actually reported the lower bound estimates (which are still not good).

Needless to say, people on both sides of the school opening debate used this to justify their previously held beliefs.

But one thing to keep in mind is that the data they used predate the arrival of B.1.1.7 to the U.S. (this isn’t BAD SCIENCETISMZ!!, since, by the time B.1.1.7 arrived in the U.S., all schools had been open). Had B.1.1.7 arrived, I think they would have seen a (relatively) larger effect of classroom and household transmission due to its increased ability to spread, including among children (when first discovered, B.1.1.7 was feared as a potential ‘child killer’ because it spread so rapidly within families; fortunately, this observation largely had to do with improved transmission, not disease severity).

Anyway, as usual, context matters here. In a pre-vaccine, post-B.1.1.7 world, opening schools, especially when prevalence is high and other precautions aren’t being taken, is probably not a good idea.

So get your fucking COVID-19 vaccination already.

*I think the statistical methods they used are fine, but I need to think about them some more, given the ‘clusteriness’ of COVID-19.