The causal effects of chronic air pollution on the intensity of COVID-19 disease: Some answers are blowing in the wind

The threats posed by COVID-19 have catalyzed a search by researchers across multiple disciplines for policy-relevant findings about critical risk factors. We contribute to this effort by providing causal estimates of the link between increased chronic ambient pollutant concentrations and the intensity of COVID-19 disease, as measured by deaths and hospitalizations in New York City from March through August, 2020. Given concerns about unobservable characteristics that contribute to both ambient air pollutant concentrations and the impacts of COVID-19 disease, we instrument for pollutant concentrations using the time spent downwind of nearby highways and estimate key causal relationships using two-stage least squares models. The causal links between increases in concentrations of our traffic-related air pollutants (PM2.5, NO2, and NO) and COVID-19 deaths are much larger than the correlations presented in recent observational studies. We find that a 0.16 g/m3 increase in average ambient PM2.5 concentration leads to an approximate 30% increase in COVID-19 deaths. This is the change in concentration associated with being downwind of a nearby highway. We see that this effect is mostly driven by residents with at least 75 years of age. In addition to emphasizing the importance of searching for causal relationships, our analysis highlights the value of increasing the density of pollution-monitoring networks and suggests potential benefits of further tightening of Clean Air Act amendments, as our estimated effects occur at concentrations well below thresholds set by the National Ambient Air Quality Standards.

Quarantine fatigue thins fat-tailed coronavirus impacts in U.S. cities by making epidemics inevitable

We use detailed location data to show that contacts between individuals in most U.S. cities and counties are fat tailed, suggesting that the fat tails documented in a small number of superspreading clusters are widespread. We integrate these results into a stochastic compartmental model to show that COVID-19 cases were also fat tailed for many U.S. cities for several weeks in the spring and summer. Due to epidemiological thresholds, fat-tailed cases would have been more prevalent if not for the gradual increase in contact rates throughout the summer that made outbreaks more certain.