Partisanship and the Spread of COVID-19 in the United States

In this paper we use a Bayesian latent variable model to identify the effect of sociopolitical covariates on the historical COVID-19 infection rate among the 50 states. The model is calibrated using serology surveys issued by the Center for Disease Control. We show that as of July 14th, there are approximately 10 million people who have been infected with COVID-19 in the United States, and these people are concentrated in states that voted against President Donald Trump in 2016, are less concerned about COVID-19, are relatively unlikely to wear masks, and have fewer economic resources. Second, we find that increased mobility measured by Google cell phones in grocery stores and retail establishments has the highest correlation with subsequent COVID-19 spread, and that mobility is an important mediator of covariates and the spread of the disease. However, although support for President Trump correlates strongly with reduced COVID-19 infections, we find that this result does not come about via reduced mobility. Instead, it would appear more likely that conservative states were spared early outbreaks due to random or exogenous factors, and instead people may be inferring that partisanship has a causal effect on the disease when in fact it is likely a confounded association.