President Biden on Wednesday renewed his calls for Americans to be vaccinated against the coronavirus. In doing so, he directly addressed a key divide on willingness to get vaccinated: partisanship. While 9 in 10 Democrats in polling conducted by the Kaiser Family Foundation said they have gotten a shot or will as soon as possible, only half of Republicans said the same. Three percent of Democrats said they would definitely not get a vaccine; more than a quarter of Republicans said the same.

“We were elected to be president and vice president for all Americans, and I don’t want to see the country, that is already too divided, become divided in a new way: between places where people live free from fear of covid and places where, when the fall arrives, death and severe illnesses return,” Biden said during a speech in Washington.

“Getting the vaccine is not a partisan act,” he added. “The science was done under Democratic and Republican administrations, and, in fact, the first vaccines were authorized under a Republican president and widely deployed by a Democratic president. All over the world, people are desperate to get a shot that every American can get at their neighborhood drugstore at no cost with no wait.”

Fair points, save one. Getting the vaccine is a partisan act, it turns out, even if it shouldn’t be.

In fact, a review of data on election results, vaccination rates and case totals at the county level shows that the results of the 2020 presidential election in a county are as good a predictor of the per-capita change in new coronavirus cases over the past month as is the density of the vaccinated population.

The correlations here aren’t that robust, to be fair. While the rollout of vaccines nationally has unquestionably pushed the number of new infections lower, a lot of counties didn’t see large changes in the per-capita number of new infections since the beginning of May. So we see a relatively weak overall correlation between more vaccinations and a drop in cases over the month.

The populous county at the center of the graph is Los Angeles, by the way. Its data includes a day of anomalous data (3,700 new cases reported on May 30) that is artificially increasing its per-capita totals. Since it’s only one county, though, it doesn’t affect the analysis below to a significant degree.

When we compare the change in cases with the result of the 2020 presidential vote in the county, the correlation shifts. Places that voted more heavily for Biden last year were more likely to see larger declines in their new case totals over the past month than counties that preferred President Donald Trump.

If we break out the counties into deciles (that is, 10 groups of equal size), we see how the correlations work. The more vaccinated the county, the bigger the average drop in new cases per capita relative to a month ago. The more Trump-supportive, the smaller the drop.

Why? Almost certainly because of the link between vaccinations and partisanship.

You’ll notice that our figures here look at completed vaccinations. Even those who’ve received only one dose of a two-dose vaccine see some protective benefit, and the correlation between those who’ve received one dose and the 2020 vote is even stronger.

In other words, vaccinations are partisan. Perhaps it’s a fairly weak link, but it’s a link nonetheless. Biden postulated that the country could see a new divide, between places that see coronavirus outbreaks in the fall and those that don’t — another divide in a country already riven by them. It probably isn’t much consolation that this new sick-healthy divide overlaps heavily with the partisan divide he was lamenting.