One of Donald Trump’s favorite White House decorations was a map of the 2016 presidential election results. He had the map framed and brought into the West Wing shortly after he was inaugurated. He would hand out versions of the map to reporters in an attempt to impress them. When he first faced impeachment in 2019, he tweeted a version of the map with the words “Impeach this” superimposed, which didn’t really make much sense.

But it was easy to understand the appeal to Trump. He’d lost the popular vote in that election by millions of ballots, but here was an image that showed the race as he saw it: a sea of dark red, reflecting all of the support he felt he’d earned. Nags like myself were quick to point out the political difference between winning 1,000 votes in an 20,000-square-mile area and winning 10,000 votes within 20 city blocks, but such appeals to reason were easily swept away by the grandeur he saw in those land-area maps.

Something that gets a bit lost in this, though, is that Trump’s map actually was a lot redder than maps in the past. In both 2016 and 2020, the number of counties that preferred Trump by a wide margin was substantially larger than the number of counties that had done so for candidates in any other election over the past 60 years.

In 2016, about 16 percent of counties preferred Trump by a margin of at least 50 points, up from about 10 percent of counties that preferred Mitt Romney by that margin four years prior. In 2020, the percentage increased to 33 percent — just shy of a third of counties.

We can visualize the 2020 results by plotting the percentage of counties that fell into any given five-point range of margins. It looks like this.

Again, that distribution is abnormal. We can develop a similar graph for each election going back to 1960 — and we did.

(Why so even in 1992? Because the presence of independent candidate Ross Perot siphoned off a lot of votes, reducing how wide the two-party margin could be.)

You’ll notice that the results jump around a bit, given that different presidential candidates do better or worse than others. In 1964, for example, counties were more likely to vote Democratic, reflecting Lyndon B. Johnson’s landslide win, while Richard M. Nixon’s blowout victory eight years later shifted counties to the right. (Even that year, though, only about a fifth of counties preferred Nixon by at least 50 points.)

We can adjust this to consider the relative result in each county: that is, the margin relative to the national margin. So if a county preferred Trump by two points this year, it was actually about 6.5 points more Republican than the country overall, since Democrat Joe Biden won by about 4.5 points.

If we do the same tally above for relative results, you see that a lot of the fluctuations even out. Until about 2000, when the shift to the right is obvious anyway.

There isn’t any equivalent shift to the left. About 2 percent of counties preferred Biden by at least 50 points, in line with the past few elections and up from essentially none in every election from 1968 to 1992.

It’s not a mystery why this is happening. America is increasingly polarized along an urban-rural axis. Younger Americans are moving to more urban areas, a shift referred to as the big sort. That leaves older, more conservative Americans in more rural areas, increasing the density of Republican votes. It’s a reason Biden won the presidency with far more votes than Trump had in 2016 despite flipping relatively few counties: those counties were often suburban ones with lots of votes that helped shift the results of the election in key states.

There is a lesson here, though. For all of the focus on how geographic maps poorly convey actual election results, which they do, those maps are probably getting worse at the task. The number of red counties isn’t changing much, and they are shifting more deeply red. Yet a Democrat was elected president by a wide national vote margin.