In the back-of-the-envelopment assessment earlier, I forgot to include tertiary and further layers of infections. Doing so increases the estimate of the overall number of eventual deaths significantly to 200-1100 deaths per day of protests. 1/12
Here, we can use a simple branching process simulation to calculate the distribution of the total downstream infections from an index case given a specific reproductive number. 2/12
If we assume an R0 of ~0.95, we get the following distribution of total downstream infections per index case. 3/12 
Many infections don't transmit onwards, but some infections spark substantial transmission chains (despite R0 < 1). Averaged across this large variation, each index case causes ~18 downstream infections. 4/12
This is the fundamental issue with having a barely controlled epidemic in the US. Each additional infection may result in a large number of downstream infections. 5/12
If we have 3000 infections sparked by protests each day, we can expect that these infections will eventually lead to ~54k downstream infections. 6/12
These infections will percolate through the community and be realized over the coming weeks. Assuming an IFR of between 0.5% and 1%, we would expect these ~54k infections to result in 270-540 eventual deaths per day of protests. 7/12
If we have 6000 infections sparked by protests each day, we would expect ~108k eventual infections and 540-1080 deaths per day of protests. 8/12
My (updated) best guess is that each day of protests involving 600k people will result in between 200 and 1100 eventual deaths. 9/12
Given the severity of COVID-19, even a modest increase of 3% to 6% on top of the current national epidemic has substantial downstream repercussions. 10/12
Racism is a critical public health issue, but societal benefit of continued protests must be weighed against substantial potential impacts to health. Again, these 200-1100 deaths per day of protests will be disproportionately among black individuals. 11/12
If the number of secondary infections in the protest setting is significantly higher than my guess of 1-2, we should start to see this as a spike in confirmed cases in the next few days in counties with large / ongoing protests. 12/12
(Technical note: I cut off the branching process at March 2021 to simulate introduction of vaccine, but this doesn't make a big difference in overall calculation. Without this, the average total number of downstream infections is 19.5 when R0 is 0.95.)
(Please read this thread for greater context that I've omitted here.)
Follow up in response to @mlipsitch's thoughtful reply ( ). I should have better considered how much choice of R0 impacts this analysis. Using R0 of 0.9 rather than 0.95 would reduce average downstream cases from ~18 to ~10.
Swapping R0 from 0.95 to 0.9 would decrease estimated downstream infections from 54k-108k to 30k-60k and estimated downstream deaths from 270-1080 to 150-600. Generally, these back-of-the-envelop calculations should be taken with a huge grain of salt. We don't yet have data.
And a Marc says, a very small improvement in our ability to control the epidemic through #TestTraceIsolate (or otherwise) results in a huge mitigation of infections engendered by protests as well as infections engendered by operating the economy and society.
