Interviewing sociologist Robert Wuthnow about his new book, “The Left Behind,” Vox’s Sean Illing quickly got to an interesting question.

“Most people assume that rural America means white America — is that a mistake?” he asked.

“The percentage that is African-American is relatively small, but it depends on what part of America we’re talking about,” Wuthnow replied. “There are many rural communities in the South that are almost entirely African-American, and the same is true of small towns in places like Minnesota. But I’d say something like 90 percent of rural America is white. Although the demographics are quickly changing as Hispanics and other immigrants settle in rural parts of the country.”

That’s interesting, because it highlights a tension between two demographic categories that stood at either end of the spectrum in 2016 voting. Rural voters strongly preferred Donald Trump, by a 27-point margin. Nonwhite voters overwhelmingly preferred Hillary Clinton — by 53 points, according to exit polls.

So who did rural nonwhite voters prefer? We pulled data from the Census Bureau to try to answer that question.

If you plot the percentage of the population in each county that lives in a rural area against the 2016 vote results, there’s a clear trend. Less-rural places voted more Democratic. More-rural places voted more Republican.

If we overlay race — the percentage of the population that’s not white (defined as white non-Hispanic) — the picture doesn’t get much clearer. In that line of counties at the top of the graph, places where the population is 100 percent rural, the circles on the left appear to be larger than the ones on the right — meaning that those counties appear to have larger nonwhite populations.

But things are clearer if we flip those two datapoints. Let’s plot the portion of the county that’s nonwhite on the vertical axis and scale the points by how rural the county is.

A pattern emerges.

This analysis obscures the preferences of nonwhite voters in more-white rural areas — meaning, for example, black voters in rural counties where the population is largely white. But the strong trend in that graph suggests that there may not be much of a difference.

What’s interesting is that the pattern shifts depending on which nonwhite group we examine.

If we consider counties with larger black populations (that is, black non-Hispanics), there’s basically a straight line where the county gets more Republican as the black population density drops. Even in more rural counties.

Among Hispanics, though? It’s much less clear. The pattern holds generally — more-Hispanic counties voted more heavily for Clinton — but there are also a number of counties that are more than half Hispanic in which Trump earned more votes. Those counties are mostly in Texas; remember that our demographic data deals with population, not registered voters.

The answer to our question seems to be that nonwhite voters in rural places voted a lot more like nonwhite voters nationally than they did like rural voters nationally.

Let’s just compare the two directly.

A nice little curve. But zooming in on the more-rural counties, the assessment above seems to be accurate.

Counties that are farther right on this graph were more likely to vote for Clinton.

The rest of Illing’s interview with Wuthnow is interesting, too, by the way. We just got a little hung up on the first question.