Donald Trump speaks to the crowd at Ladd-Peebles Stadium in Mobile, Ala., on Dec. 17. (Dan Anderson/European Pressphoto Agency)

One of the recurring themes to the candidacy and presidency of Donald Trump has been that he’s enjoyed the robust and ongoing support of Americans without college degrees. This has often been simplified as “white working-class voters,” the core group that propelled Trump to the presidency.

Over at Monkey Cage, Duke’s Nicholas Carnes and Vanderbilt’s Noam Lupu make an important and apparently contradictory point: Trump’s supporters were actually fairly well-off.

“Trump’s voters weren’t overwhelmingly poor,” they write. “In the general election, like the primary, about two thirds of Trump supporters came from the better-off half of the economy.”

That seems like a contradiction from how Trump’s base has been described. It isn’t.

Terms like “working class” and “middle class” aren’t actually constrained by boundaries linked to income, particularly in the common usage of the terms. For example, a Wall Street Journal-NBC News survey from 2013 found that a greater percentage of those making over $100,000 described themselves as “middle income” than did Americans earning $40,000 to $50,000.

The biennial General Social Survey asked Americans directly what economic class they believed they belonged to. Among those earning incomes of $100,000 or more a year, only about a fifth said they were in the “upper class.” Ten percent described themselves as being “working class”; nearly 7 in 10 identified as “middle class.”

The reason for this is simple and immediately obvious: Economic class overlaps with a lot of social indicators that have nothing to do with income levels. We can readily conjure up stereotypical interests of someone in the working class or the upper class (NASCAR vs. lattes, etc.), and people actually use those identifiers to consciously embrace the identity with which they identify. When we say that Trump’s base was working-class white people, we may mean income — but we are often more likely to mean that they share a set of cultural values.

We can see how that self-identification worked for Trump in the GSS data. Among those earning $10,000 to $20,000 a year in income and who identified as “working class” or “middle class,” twice as many identified as Democrats or Democrat-leaning independents as Republicans or Republican-leaning independents. Among those earning $100,000 or more, though, the party split was nearly equivalent.

One big reason the graph above looks the way it does: Nonwhite Americans who earn lower incomes are much more likely to identify as Democrats. Wealthier Americans are generally more likely to be white and more likely to be Republican.

If we look only at college education, a lower percentage of those who earn $100,000 or more and who identify as Republican or Republican-leaning have college degrees than those in the same income group who identify as Democratic.

It is fair to say, then, that while most Trump voters were in the upper half of the economy, they may also be less likely to have a college degree and may identify as culturally working-class.

One part of the common descriptor of Trump’s support was clearly, true, though: They were mostly white.

The General Social Survey was funded primarily by the National Science Foundation and conducted through in-person interviews with a random national sample of roughly 1,900 adults in the spring of 2016. Overall results carry a margin of sampling error of roughly 2.5 percentage points; the error margin for subgroups is larger.