So which is which?
The easiest way to tell is by looking at the Northeast. Much of New England votes reliably Democratic but is also densely White. So you can tell that Map B is the map of White Christians and Map A the map of 2020 election results.
The point, of course, is that it isn’t easy to differentiate between them. Looking at PRRI’s maps of the distribution of religious groups, the superficial similarity of White Christianity and Trump support is immediately obvious. But, of course, national maps of county-level data tend to obscure underlying trends, as anyone who has had a debate over how to depict presidential-vote results can attest.
So what do the actual data show? And how much of this is a function of the overlap between religion and race, and not race by itself? We know, for example, that there’s a correlation between how densely White a county is and how strongly it supported Trump. On the graph below, in which each dot depicts a county, you can see how the cloud of results runs from lower left to upper right: increasing margins for Trump as the density of Whites in a county increases.
You can often eyeball how strong correlation is by noticing how closely clustered the data points are. Here, the cloud gets blurry at the edges; there are a number of Trump-voting counties with lower densities of White people and some very White counties that voted Democratic (some of which are in New England).
PRRI was generous enough to share with The Post its county-level estimates of the percentage of White Christians in a county. When we introduce that second variable, look what happens to the relationship.
Statisticians measure how closely tied two series of data are using something called the “correlation coefficient.” A coefficient of 1 means that two sets of data are perfectly correlated. The density of Whites in a county correlates to Trump support in that county with a correlation coefficient of 0.52. The density of White Christians correlates to Trump support with a coefficient of 0.74.
We can show the importance of that overlap with religion in another way. Plotting the density of White Christians in a county against the density of Whites, we see, expectedly, a trend from lower left to upper right. More Whites means more White Christians, naturally.
Notice, though, that we colored the dots in the same way as the prior graphs: blue dots indicate counties that voted for President Biden and red ones counties that backed Trump. As the density of Whites increases, the counties that have a lower density of White Christians are more likely to be colored blue. In other words, the same density of Whites, but fewer White Christians — with a noticeable effect on vote results.
The graph below emphasizes that more clearly. The shaded areas approximate the range in which counties that backed Trump or Biden fall in terms of the density of White Christians. The red-shaded area is consistently higher than the blue-shaded area, indicating that counties that backed Trump have higher densities of White Christians, even as they have the same density of Whites overall.
The problem for Trump in 2020 was twofold. First, turnout was up substantially over 2016, particularly among groups that opposed his candidacy. Second, the correlation between the density of White Christians and Trump support actually dropped since 2016. While more White Christians meant more support for Trump, that wasn’t as true last year as it was when Trump was first elected. In most counties, the margin for Trump shifted downward, even in many densely White Christian ones.
(At lower right you can see Trump’s gains with heavily non-White counties, a shift not large enough to offset his overall declines.)
This is the story of the 2016 and 2020 elections. Trump had a key base of support that yielded him millions of votes. In 2020, though, it wasn’t enough.