After Donald Trump won the Nevada caucuses in February 2016, he gave a celebratory speech in which he rattled off his performance among demographic groups as delineated by exit polling.

“So we won the Evangelicals. We won with young. We won with old,” Trump said. “We won with highly educated. We won with poorly educated. I love the poorly educated!”

For a number of observers, this admittedly unusual statement constituted something of a tell: Trump did do well among voters with lower educational attainment, and his awkward celebration of them made obvious the incongruity between himself and those voters.

The comment also helped to cement the association between less-educated voters and Trump supporters, which itself blurred into facile dismissal of Trump’s base as dumb. That was probably fine with Trump, given that his campaign was predicated on being the voice of normal Americans against the hyper-educated coastal elite. But it was also never a fair or particularly useful assessment of his supporters.

All of this was elevated again in the wake of the violence at the Capitol last week. Dueling essays at the Atlantic, for example, offered two portraits of the rioters that overran the building. In one, by Caitlin Flanagan, the rioters were depicted as dumb slobs. In the other, by Adam Serwer, they were cast instead as everyday members of the American middle-class — if not the upper-middle-class.

It’s a tension we’ve seen over and over during the past five years, this idea that Trump’s base is both less educated and somehow less impoverished than America as a whole. Normally education and income move hand in hand, so how could those two descriptors (however loosely accurate they might be) overlap?

Polling in the wake of Trump’s election loss and the violence in D.C. helps shed some light on the question.

On Thursday, Axios released the results of a poll conducted on its behalf by Ipsos. Among the questions it included was one in which respondents were asked to classify themselves into one of five categories: progressive or liberal, moderate Democrat, political independent, traditional Republican or Trump supporter. A third of Republicans chose “Trump supporter,” with nearly 6 in 10 picking “traditional Republican.”

When we look at the demographic breakdowns on the question, though, we see the pattern outlined above. Less educated Americans are more likely to call themselves “Trump supporters” than “traditional Republicans” (though a plurality of the least-educated group identifies as “moderate Democrat”).

There’s another pattern that emerges: Older Americans are also more likely to call themselves Trump supporters above all else.

This, in fact, is an underrecognized demographic correlation. When former Wisconsin governor Scott Walker was exploring a run for president in 2016, it was noted that he didn’t have a college degree. That would have made him unusual for a president — but was far from unusual for an American citizen. Census Bureau data released in March show that two-thirds of Americans over the age of 18 don’t have a college degree, with degrees being most common among those ages 35 to 54.

Among those ages 30 to 39, more than 4 in 10 have a college degree. Among those 70 and up, only about 3 in 10 do.

As any millennial will be quick to tell you, that college degrees are less common among older Americans does not mean that older Americans have less income. One of the starkest shifts in the United States over the past several decades has been that degrees are increasingly required for higher-paying jobs. The erosion of labor unions and decrease in employment like manufacturing has meant fewer opportunities for good-paying jobs that don’t require a college education. Many older Americans were able to build wealth in a period where such jobs were more readily available.

As Axios’s Chris Jackson noted in an email to The Washington Post, Trump supporters also tend to be more likely to live in rural areas than are Republicans overall. That, too, correlates with lower educational attainment.

In other words, given that the rioters at the Capitol were overwhelmingly people who would identify as Trump supporters, it is likely true that they overall tended to have fewer academic credentials but may also have been better positioned economically. In the Axios-Ipsos poll, there wasn’t a large difference in identification as Trump supporters between the highest and lowest income groups.

It’s hard to argue — particularly when considering the results in 2018 and in the Georgia runoff elections last week — that Trump brings out voters who might not cast a ballot if it doesn’t feature his name. His coalition of support didn’t overlap with the support seen by traditional Republican presidential candidates, though he benefited from the support of traditional Republican voters. Trump supporters are, in fact, a group that doesn’t fit neatly into the sorts of political categories that were used before 2015, leading to differing interpretations of who they were.

Trump had good reason to love the poorly educated. They helped make him president.