by Robert E. Wright
The American Institute for Economic Research
Just before a major election, and scary Halloween, Doug Bernheim et al released a working paper (“The Effects of Large Group Meetings on the Spread of COVID-19: The Case of Trump Rallies,” 30 Oct.) that purports to show that 21 Trump rallies caused a total of 30,000 cases and 700 deaths. Those cases, they assert, were incremental, meaning that they would not otherwise have occurred.
Trump rallies, says this study, were superspreader events. For all I know, they might have been. The world is a complicated place. But we also know with absolute certainty that in order for this paper to cultivate its results – along with its definitive conclusion – the authors had to do a tremendous superspreading of statistical manure.
The authors question studies that show that individual Trump rallies, like the infamous indoor one in Tulsa, did not cause numerous infections (e.g., D.M. Dave et al, “Risk Aversion, Offsetting Community Effects, and COVID-19: Evidence from an Indoor Political Rally” NBER Working Paper 27522”) because “COVID-19 outcomes are highly variable” (p. 3). In other words, some rallies may have been superspreader events while others probably weren’t.