This paper showing that the Sturgis Motorcycle Rally was a super spreader event is getting a lot of attention. Given it is in my wheelhouse, I decided to give it a careful read.
My take: reasonable paper, but the effects are almost surely way smaller than reported. https://twitter.com/SDSUCHEPS/status/1302480031638147074

I would argue that this is the key figure in the paper. It shows an increase in Covid cases in South Dakota overall, but especially in Meade County (where the Rally took place from 8/7-8/16). So cases were trending up before the Rally, but appear to intensify 1-2 weeks after.

The authors find this case increase statistically significant compared to a synthetic control (see figure below for example). But what is this synthetic control? It includes areas around the country that match various South Dakota observables before the Rally.

The graph looks impressive because the synthetic control is exactly on top of the South Dakota line pre-treatment. But it is important to remember that similar pre-treatment trends occur by construction. So of course the graph is impressive leading up to 8/3.

Here is where I start to have a problem with the paper. The synthetic control areas come from various locations (Hawaii, Texas, etc.). But, the authors exclude border states to South Dakota (Iowa, North Dakota, etc.) because they worry about spillover effects.

The problem, however, is Covid is very regional. So if the Midwest just happened to have an increase in cases relative to Hawaii, Texas, etc. anytime in mid to late August, the authors would find an effect.

If you go to the New York Times and look at where cases are "high and staying higher", the first three states they mention are North Dakota, South Dakota, and Iowa.

Either the Sturgis Rally had huge spillover effects in states near South Dakota (unlikely given statistics in the paper on where Rally goers are from) or there was an independent increase in cases that was happening in the Midwest for which the synthetic control does not capture.

Overall, I think the "Sturgis Effect" that the authors document is in large part just a Midwest surge that took place during this time period. There is likely still a small Sturgis Effect (for example, Meade county looks especially bad), but the results are likely biased upward.

Of course, this is just my opinion. Further peer review would be useful. Overall, I think the paper is reasonable and tries to do a careful job. But, I would suggest caution in terms of running with numbers like the Rally had "public health costs of $12.2B."