Close Encounters of a Heterogeneous Kind: Understanding the Differential Impact of Social Distancing on COVID-19 Infections and Deaths

We investigate the relationship between social distancing, as measured by encounter rates using cellphone proximity data, and COVID-19 infections and deaths. Consistent with the existing literature on the effectiveness of non-pharmaceutical interventions, we find a positive and statistically significant relationship between the encounter rate and new infections and deaths. However, the magnitude of this effect is relatively weak. One explanation for this weak effect is that the effectiveness of social distancing varies across counties due to local population heterogeneity. To this end, we interact the encounter rate with county-level characteristics and find that several of these interaction terms are statistically significant. Furthermore, after controlling for these interaction terms, we find that the effect of social distancing is around 20 times larger than the effect size found in estimates without the interactions.