Pandemic Politics: Timing State-Level Social Distancing Responses to COVID-19

Christopher Adolph; Kenya Amano; Bree Bang-Jensen; Nancy Fullman; John Wilkerson

It is likely that governors will face this difficult decision again. Governors may choose to roll back social distancing at different points in time. Some experts contend that authorities will need to alternate between two months of social distancing and one month of restored freedoms to reduce the likelihood that health systems are overwhelmed by severe cases until a vaccine is developed, produced, and circulated. (3; 44) This implies state governments will be asked to impose difficult social distancing policies not once, but many times. Each instance will be a test of whether states can act promptly to prevent COVID-19 cases from peaking at unmanageable levels. If Republican governors and states with Republican majorities continue to lag behind, the cumulative impact on those states, and on the country as whole through spillovers, could be vast. Baseline hazards are stratified across the five pooled social distancing measures: recommendations and restrictions on gatherings, school closures, restaurant restrictions, non-essential business closures, and stay-at-home orders. Each row shows the hazard ratio for (the counterfactual change in) the covariate listed at the left. Counterfactual changes are chosen to simplify comparison of hazard ratios for covariates with different scales of measurement. Covariates with both 95% confidence limits below 1.0 significantly reduce the chance of adopting an additional social distancing measure. Standard errors used to compute confidence intervals are clustered by state. The concordance index shows the proportion of all pairs of states for which the model correctly predicts which state will adopt a given social distancing mandate first. Schoenfeld residuals show no evidence of violation of proportionality for any covariate. The Efron method is used to resolve ties.