Thread by @SDSUCHEPS

Yesterday the @nberpubs released our Sturgis COVID-19 paper. Dhaval Dave, @FriedsonAndrew, @Drew_McNichols, and Joe Sabia are grateful for constructive comments. Below the authors offer responses to some important questions. [1/9]

Link: nber.org/papers/w27813

While local counties may be contaminated by the Sturgis Rally, our synthetic control results hold if we require geographically proximate (within SD and border states) donor counties with no or low Sturgis inflows. [2/9]

Moreover, if we add census division-specific time effects to our “dose response“ diff-in-diff models, which account for unmeasured regional shocks (in addition to county time trends) and require within census division county comparisons, our main findings are unchanged. [3/9]

Our findings are unlikely to be due to differences in risk preferences of county residents (linked to inflows). Event studies show pre-Rally trends in COVID-19 were similar in treatment & control counties. If risky counties see faster *post*-Rally growth, may be a mechanism [4/9]

Do our results generally hold if we replace our dependent variable with the log of daily cases (essentially a first derivative of our current model)? Yes. What about testing? Both our synthetic control and diff-in-diff estimates control for testing rates. [5/9]

In terms of plausibility, our results are in the ballpark of other studies of mass gatherings. Our findings of a 7.0 to 12.5% increase in cases is similar in magnitude to Ahammer et al. (2020) who show that an additional NBA game increased local cases by 8.3 percent. [6/9]

Wing et al. (2020) find a 10.8% increase in COVID-19 from NCAA games. While comparisons not apples-to-apples (Sturgis is larger, longer, attracts nationwide attendees), 7% not unreasonable. [7/9]

Ahammer et al: bit.ly/3mj5lBP

Wing et al: bit.ly/32ujp3r

Finally, how do we reconcile our findings w/ an AP report of only 300 COVID-19 cases linked to Sturgis? Our results, in fact, show the difficulty of accurately measuring superspread w/ incomplete contact tracing, non-universal testing, non-compliance & political opposition. [8/9]

We thank our colleagues for useful comments and suggestions on our study, which shows strong evidence of a superspreading event in Sturgis, SD. We look forward to peer review by qualified economists. [9/9]

NBER: nber.org/papers/w27813