On Thursday, the Associated Press ran a story that’s become part of a well-established genre: revisiting places that voted overwhelmingly for Donald Trump to see how they felt about his presidency. In this case, the focus of the story was Sandy Hook, Ky., a town in Elliott County that had narrowly supported President Barack Obama over Mitt Romney in 2012 but, four years later, supported the Republican candidate by a more than 40-point margin.

Much of the reaction to the story was probably not what the AP was hoping for. That Trump’s base has been profiled so often was one thing. Others noted, though, that the emphasis on Elliott County, one of the poorest counties in the country, reinforced the idea that Trump’s support was largely made up of the economically disadvantaged. Why not look instead at Suffolk County, N.Y., asked journalism professor Jay Rosen, where the median household income was more than three times as high as in Elliott County. In fact, Suffolk County had the highest median income of any county Trump flipped; Elliott had the second-lowest.

At the Week, Ryan Cooper raised a similar question: Trump lost among poorer Americans but narrowly edged out Hillary Clinton in 2016 among those earning more than $50,000 a year. Why not interview those wealthier Americans?

Data from exit polling shows how the vote went by economic bracket:

There’s validity to the point, but it also highlights an interesting contrast in the 2016 election. One reason that coverage focuses on less well-off areas when assessing the election is that Trump did better in poorer areas even as Clinton did better with poorer voters.

We can look at exit poll data another way. More of Trump’s actual base of support was made up of wealthier Americans. He won a plurality of voters earning more than $50,000 a year, but nearly 7 in 10 of the people who actually voted for him were in that income group. (By contrast, only about 6 in 10 Clinton voters earned more than $50,000 a year.)

Trump voters were likely to have higher incomes than Clinton voters.

But county-level results don’t reflect that. The average median household income of counties won by Trump was about $47,200. The average median income in counties Clinton won was about $51,600. Counties that flipped from Obama in 2012 to Trump in 2016 had a slightly higher average median income of $48,200. Counties that flipped from Romney to Clinton — far fewer — had a much-higher average median income of $63,900.

Looking at the results of counties may make Trump feel better about 2016, but it’s deceptive given the uneven population distribution in counties. So we took the voting results in every county and grouped them by median household income, rounded to the nearest $5,000. More of Trump’s support came from poorer counties. In fact, voters in counties with a median annual income of more than $50,000 made up more than 64 percent of Clinton’s support — but only 54 percent of Trump’s. Forty-six percent of his support came from voters in poorer counties.

How is it possible that Trump’s voters could have higher incomes even if more of them came from places with lower median annual incomes? One possibility is much of Trump’s support in lower-income areas came from higher-income voters. (Higher-income voters are generally more likely to vote in part because they move less frequently, making it easier to maintain registration and know where to vote.)

This distinction, though, helps explain why so many of the Trump voters interviewed by media outlets are less economically well-off. Part of it, as FiveThirtyEight’s Nate Silver noted after the AP article, is certainly driven by concern that these voters were underrecognized by the national news media before the election. But part of it, too, is that, if you’re looking for a place that supported Trump strongly after backing Obama five years ago, those places are more likely to be poorer.

Not that it’s not still worth going to Suffolk County.