Crowdedness as the Missing Link between Shelter-In-Place and the Spread of COVID-19
35 Pages Posted: 1 Jul 2020 Last revised: 29 Jul 2020
Date Written: July 27, 2020
As of the end of May 2020, 44 of the 51 U.S. states and District of Columbia have implemented shelter-in-place (SIP) policies at some point during the initial outbreak of the novel coronavirus. Despite the widespread adoption of SIP policies, empirical researchers are unable to reconcile SIP policies' apparent causal impact on improvements in public health outcomes with the lack of evidence of a mediating social distancing (SD) mechanism. We argue that these contradictory econometric results can be reconciled if researchers (1) use historical SD data to account for heterogeneous seasonality patterns across regions and (2) adopt our proposed crowdedness measure of SD that captures traffic density over non-residential square footage. Using weather variables to instrument for SD behaviors, we show that a 10% decrease in crowdedness leads to a 1-1.5 percentage point decrease in the confirmed case growth rates while changes to the percent of time at home, a standard SD measure, has a negligible impact on case growth rates. Furthermore, we adopt a triple differences approach, comparing differences between pre- and post-intervention (1st) counties' year over year changes in SD behavior (2nd) with the corresponding metrics of counties without such interventions over the same timeframes (3rd), to demonstrate that SIP policies decreased crowdedness by nearly 10%. This magnitude of decrease in crowdedness fully mediates our estimated direct impact of SIP on case growth rates. Informed by our findings, policy makers should consider allocating limited resources to tracking and controlling crowdedness rather than attempting to enforce blanket SIP orders.
Note: Funding: None to declare
Declaration of Interest: None to declare
Keywords: COVID-19, Coronavirus, social distancing, shelter in place, causal inference, econometrics
Suggested Citation: Suggested Citation