Weather, mobility, and COVID-19 infections: results from panel data analysis and an epidemiological model

Abstract:

The debate over the influence of environmental conditions on COVID-19 epidemiological dynamics, for instance the potential of warmer and wetter weather to slow the spread of COVID-19, is still unsettled. We empirically examine the interactions between weather, mobility, and infections and deaths from COVID-19. To do so, we combine several data sources of confirmed infections, recoveries, and deaths from COVID-19 with global high-resolution reanalysis weather data and datasets on observed mobility. We first explore the statistical associations between weather and COVID-19 infections using methods from machine-learning. We then estimate the effect of weather on COVID-19 infections using econometric methods for panel data. We finally plug our estimates into a full epidemiological model which we solve hierarchically using Bayesian techniques. We find that weather 4-8 days and 11-19 days before the reporting date is a good predictor of the growth of infections. While we find that mobility is highly sensitive to weather, mobility seems not to be the major channel by which weather affects the epidemiological dynamic. We take this as explanation of why we do not find significant effects of outdoor environmental conditions (precipitation, solar radiation, thermal comfort) on COVID-19 growth rates. Our results for absolute and relative humidity are ambiguous. Overall, we find that temperature is the best predictor of the growth of COVID-19, with a positive effect on growth rates at low temperatures. Our results therefore suggest potentially large seasonal components of COVID-19 epidemiological dynamics. From a policy perspective, our results could explain why reducing outdoor mobility might not be very effective and they support targeting indoor climatic conditions for COVID-19 mitigation.