Deep dive into COVID contagion impact of Sturgis bike rally.

Here is the first serious attempt to quantify the contagion impact of the Sturgis bike rally. It's 60 pages and some pretty dense reading, but remarkably thorough and careful.

They used anonymous cell phone ping data to track the home counties of visitors, and looked for changes in post-event COVID rates relative to the population-adjusted number of attendees. They found a strong correlation between attendee density and post-event COVID rates, i.e. counties that had relatively more people travelling to Sturgis had much higher COVID rates after the event.

In total, they estimate that for the period 8/2 to 9/2 (covering the pre-event opening festivities), the rally resulted in a net increase of 267,000 COVID cases. That's 19% of nationwide cases during the period. If that sounds too huge to be possible, keep reading under the images below. I did some sanity checks and these numbers seem plausible.

(Note that this doesn't mean that half the attendees tested positive. It's much more likely that 5-10% did, and they infected workers and residents in SD, then family, friends and coworkers back home, who went on to infect others...)

The researchers were VERY thorough, investigating whether increased distancing by residents might offset the effects (one of the reasons the BLM protests didn't have much net impact, local populations were less mobile when protests were happening), checking bar and restaurant foot traffic, etc. My favorite snippet:

"...liquor store sales in Sturgis were up 27 percent from the prior year, which could be representative of a less risk averse population of attendees given that attendance was down roughly 7.5 percent from the prior year..."

Yep, they drank almost a third more than last year's attendees.

Meade County SD is in red, Sturgis is at the west end of the county. This map shows, adjusting for population, which counties had low and high rates attendees.

This was the core of their analysis. How did COVID results change post rally in counties with heavy attendance vs. light attendance?

Here's what happened in Meade County and the surrounding counties, relative to a "control" of similar counties that didn't have a lot of attendees.

I'm a bit disappointed that they use cumulative case counts rather than weekly new case counts, because this presentation makes changes less apparent. A steady upslope is a constant rate of infection, the jump in Meade county represents roughly a 1000% increase in infection rates.

Here they do look at the rate of new cases. Note that counties with low attendance to Sturgis (bottom right) show a continuation of our national trend of reducing COVID counts, but progressively heavier show increases of 50%-ish for moderate, 80%-ish for moderate-high, and over 150% for high inflow.

Translating: If your county had few people attend, you probably saw a 5% drop in COVID cases in the 24 days after Sturgis started. If you were in a county with a lot of attendees, your rate probably more than doubled.

Seriously, over a quarter million extra cases?

That was my first thought, so I did a sanity check. The light shaded area on the right is the period starting a week after Sturgis ended, enough time for symptoms to develop (6 days average) and people to start getting tested at home.

Note that the month-long downtrend in the US suddenly halt and is nearly flat for the most recent two weeks.

While other regions are flat to declining, the midwest - which had been flat for all of August - shows a sudden increase.

A scale of 19% of all US cases during the period is consistent with the flat US trajectory vs. the major decline we had going before Sturgis.

This event shouldn't change the slope of our trends long term - it should be a short blip to a new level, then a resumption of the underlying trend. It will be hard to tell, though, with Labor Day and resumption of in-person schooling happening concurrently.