The highest-impact demographic splits from the 2020 election

Gender, age, race, education, etc. -- what were the deepest divides between Biden and Trump voters?

There are lots of demographic splits that matter in American politics – gender, age/generation, race, urban/rural, union membership, region, religion, church attendance, income, education, and other items. Various of these features have risen and fallen in importance in distinguishing the Democratic and Republican coalitions over the years.

What are the biggest deals currently? To give one perspective on this, using a large database I’ve put together of available survey data from 2020 to 2023 from CES, GSS, ANES, PRRI, and others, I looked at presidential voting in the 2020 election.

With Trump voters on the right, Biden voters on the left, and non-voters in the middle, I looked for the highest-impact initial demographic split. The winner was Black vs. non-Black.

I then took the largest subgroup (at this point, the non-Black group) and, again, identified the highest-impact demographic split within this subgroup. This time, there were lots of conceptually related top candidates involving religious identity and LGBT items. Having spent a lot of time with this kind of data, I know that a very sensible route is to combine various groups into a kind of Rebel Alliance that occupies political terrain that is most directly opposed to the political preferences of heterosexual evangelical Christians. This Rebel Alliance includes those who are explicitly not Christian – not the vague “nothing in particular” or “other” survey respondents, but the more definitive categories of atheist, agnostic, Jewish, Muslim, Buddhist, or Hindu. This Rebel Alliance also includes LGBT and non-binary folks – lesbians and gays, for example, are politically indistinguishable from atheists, particularly at higher education levels.

So that grounded the split I made in the non-Black subgroup – I put the Rebel Alliance into one group, Straight Evangelicals into another group, and all the other non-Black folks into a Religious Middle category (so this middle includes, e.g., non-evangelical Christians, “nothing in particular,” etc.).

I then took the biggest remaining group – at this point it was the non-Black Religious Middle – and looked for the highest-impact demographic split available for this sub-group. The winning split was to put gun owners and/or military veterans on one side and everyone else on the other.

Then I kept going with more splits until I had a pretty good number using a range of features. You can see the results in the graphic below. The blue-to-purple-to-red color shifts within each box represent the proportion of that group’s Biden vs. neither vs. Trump voters.

Lots of themes pop out. Primarily, notice both what’s present and what’s absent in these highest-impact splits. Present: Race, religion and sexual orientation, gun ownership and military service, education, religious attendance, urban/rural, and then age/generation shows up, but in a funny way: among both Blacks and Straight White Evangelicals, younger people do not show the uniformity of older people, such that younger Blacks did not as solidly prefer Biden as their elders did and younger Straight White Evangelicals did not as solidly prefer Trump as their elders did.

Some notable absences include: Gender (male vs. female) was never a big enough deal to make a split on that line, nor was union membership, income, home ownership, region, work status (full-time vs. part-time vs. not working), or marital status.

That’s not to say that these divides are unimportant in some absolute sense. Gender differences and regional differences in politics, for example, are real and interesting – they just weren’t among the central drivers of the Biden and Trump coalitions.

In general these days – both for partisan voting and for issue opinions – I find that the really big deals are race, religion, and sexual orientation. The moderately big deals are gun ownership, military service, and education level. Then we’ve got gender and urban-rural, which are both modestly impactful down the line. Some other items I’ve mentioned (e.g., union membership, home ownership, marital status) are small players – they can occasionally matter at the margins, but are almost never part of the central political story.

And then age/generation differences are complex. There are issues where younger folks are definitely more liberal (e.g., on immigration, and on expanding Medicare) but also are occasionally not at all more liberal and indeed perhaps a bit more conservative (e.g., on gun regulation). When it comes to partisan voting, younger people are less likely to vote and less likely to hold ideologically aligned left-right political opinions.

There’s lots more to unpack in these results, but I’ll leave it there for now. More to come!