Emily Oster: School infection rates are low even in areas with high COVID-19 rates, according to data on 550,000 US students
- Over the last four weeks, economist
Emily Osterbeen working on the COVID-19 School Response Dashboard, collecting data from US schoolson COVID-19 infections, reopening plans, and mitigation.
- In the most recent two weeks, the infection rates were 0.13% for students and 0.22% for staff. That's about 1.3 cases per 1000 students and 2.2 cases per 1000 staff over two weeks.
- The evidence suggests that we should expect more cases to show up in schools in areas where cases are higher in the community. But school rates here are low even in places with high community rates.
- While it's too early to say anything definitely, this suggests that schools are not super-spreaders. Even so, community spread clearly matters for schools.
Over the last four weeks, I've been working on the COVID-19 School Response Dashboard, a partnership between Qualtrics, the National Associations of School Superintendents and Principals, and me, along with a whole host of others. We have been collecting data from US schools on COVID-19, the disease caused by the coronavirus.
Our project starts at the school level, with baseline information on school enrollments, staffing, reopening plans and COVID-19 mitigation procedures. We return to the schools every other week to ask for information on COVID-19 cases in students and staff, and any updates on reopening plans.
Our latest data from the last two weeks of September covers about 550,000 enrolled students, of whom 175,000 are in person and 56,000 staff. We have data over this period from 47 of 50 states, plus Washington DC. Nearly all in-person schools in the data are doing some kinds of mitigation. Masks are the most common, but they are not universal. Only about 60% of schools report requirements of six feet distancing.
The initial focus of
While it's too early to say anything definitely, this also suggests that schools are not super-spreaders.
Looking at geographic variation is key
The top-line number is likely what grabs people first, but we are now trying to better understand the data. The first key to that is looking at geographic variation. How are the rates we are seeing in schools changing with the rates in the community?
It's worth pausing to ask what we would expect. As an example, let's start with the assumption that the prevalence of the virus in the community is 1%. One possibility is schools are just like any other random location in the community, but that there is little or no spread in school.
If this is the case, we'd expect the prevalence in people at school to be 1%. That is, we would still expect to see some COVID-19 in schools, coming in from outside.
Another possibility is that people in schools are positively selected as lower prevalence. This could be a result of general selection or, in the context of our data, it could reflect bias in who we recruit to our sample. In either case, we'd expect schools to be lower prevalence than the surrounding area, although still we'd expect higher school rates in areas with higher community rates.
If schools are doing a good job keeping the virus at bay — for example, by encouraging positive behavior in students and staff and by protecting people at school — you might expect the link between schools and community rates to be less strong. The community matters less if you have a good bubble.
Finally, if schools are a source of significant spread then we'd expect higher school rates than community rates. This is probably the major source of concern and it's notable that it's what we are seeing some in college-age populations. Universities have seen large spikes, outpacing the communities in which they are and contributing to overall community spread.
So which is it?
Schools have student COVID-19 infection rates in the same range as —but a bit lower than — local community rates
The images below show — by census division, since we don't yet have quite enough data to do each state — the student prevalence in schools (that's cases over a two week period) alongside the case rates in these census divisions over that period.
We're seeing that schools have student rates in the same range, but generally a bit lower, than their community rates. In New England, for example, our data show 24 cases per 100,000 students, and the community rate is 70 per 100,000. In the East South Central area, the student rate is 219, versus 267 per 100,000 in the population.
The staff cases are somewhat higher than student cases and are alternatively higher and lower than the community rates, although the numbers are small. We'll know more about this as the data evolves.
The staff rates are very similar for in-person and remote learning, an issue which needs more exploration.
The correlation with community rates is very high. In places with more spread, we see more school cases. In places with lower spread, we see less.
In the second set of maps we separate out by private and public schools. Private school rates are generally lower — this may reflect demographics, resources or mitigation strategies — but we also see they link less closely to community rates.
What does this tell us? On the one hand, community spread clearly matters for schools. This is a point made early on by Ashish Jha, the dean of the Brown University School of Public
On the other hand, the school rates here are low even in places with high community rates.
We will need more evidence, more time, and more data to understand this better.
Emily Oster is a Professor of Economics at Brown University and the author of Expecting Better and Cribsheet.
- Nifty closes just shy of 21K on gains in metal, IT shares; Sensex scales 70k-peak
- Online furniture brand Pepperfry registers Rs 188 cr losses in FY23
- Most searched movies in 2023 – Animal is not in the Top 10 and Adipurush beats Pathaan
- EPFO invested Rs 27,105 crore in exchange trade funds in Apr-Oct: Labour Min
- Cocktails to prasads: What Indians wanted to cook and brew in 2023