The Polarization of Death

I’m continuing to update the covdata package in anticipation of a Data Visualization for Social Science course I’ll teach next semester. I revisited the Partisan Trajectories graph, as it seems there’s more that could be done with it. More on that in the future, I hope. For now, here’s an updated version using the 2020 Presidential election as the basis for the deciles, and more recent fatality data. As before, the idea is to take the time series of cumulative COVID-19 deaths and split it into deciles by a county-level quantity of interest. I look at how Republican the county is based on the two-party vote share for the 2020 Presidential election. Counties are cut into deciles by strength of support for Trump in 2020, we aggregate mortality counts to the deciles, and draw a line for each one, giving us an ecological picture of the relationship between deaths and political polarization. We see divergence at the very start for the 0th decile because New York City is in it, and it was hit the hardest by far early on. But then the polarization of death kicks in as COVID spreads everywhere and county-level outcomes start to become more heterogeneous.

Seeing as this image is getting circulated a lot, let me add: In case it’s not clear, the claim here is not that voting for Trump made people die of COVID-19. What I’m interested in is how the cumulative burden of the pandemic has been distributed in political terms. Think of the graph as showing how COVID-19 death rates are being distributed across “redder” and “bluer” counties, whatever the characteristics of the people in those counties who got sick and died.

The obvious thing to incorporate in to these estimates is an adjustment (i.e. a weight) for the varying age distribution of the counties, because older people are much more likely to die of COVID-19. The difficulty is that breakdowns of mortality by county and age aren’t straightforwardly available at present. Ideally, of course, what we want is full individual-level data on the outcomes and their covariates, but we’re not going to get that. Still, it’s hard not to think that these gaps are going to continue to widen, given where resistance to getting vaccinated is highest. There are a few ways we might do to at least get a sense of the structure of things, particularly as the age and racial/ethnic composition of counties is something we know.

The code and data are on GitHub.