One of the metrics supporters of President Trump used in 2016 to connote the scale of his victory was his dominance at the county level. Trump did, in fact, win far more counties than Hillary Clinton; had each county been awarded an electoral vote, he’d have won an electoral college landslide.

But that’s not how it works. Most of those counties have very few residents, and many are in states that offer very few electoral votes. Clinton won nearly 3 million more votes than Trump, but Trump and his allies were eager to categorize his 306-electoral-vote victory as a landslide, so they used what numbers they could.

Four years later, Trump lost his bid for reelection after President-elect Joe Biden apparently won precisely the same number of electoral votes as Trump did in 2016. Unlike Trump, though, Biden did so while winning the popular vote by a wide (and widening) margin. But very much like Clinton, he did so while winning far fewer counties.

In fact, Biden’s victory came largely from running up margins in counties that Clinton won four years ago. In Arizona, for example, the current data suggest that Biden only flipped one county — Maricopa County, the state’s largest, which helped.

In Pennsylvania, a similar pattern: Biden only picked up two counties. One was Erie County, in the far northwest of the state — a county that flipped dramatically from Barack Obama in 2012 to Trump in 2016. This year, it flipped back.

Compare the number of counties that flipped over the past four years (as of writing) with those that flipped from 2012 to 2016. (On the maps below, the names in parentheses are the eventual winners, for reference.)

The colors (hopefully obviously) indicate the party that the counties flipped to. From 2012 to 2016, most counties that flipped went from backing Obama to backing Trump. That density in the Midwest was what helped Trump win Michigan, Pennsylvania and Wisconsin — and the electoral-vote majority he needed.

Notice, too, though how many more counties flipped four years ago. Over the past 50 years, it’s been much more common in presidential elections for a lot of counties to flip than for relatively few to have done so.

Since 1972, less than a third of counties haven’t flipped at all in a presidential contest.

What’s interesting about the change from 2012 to 2016, by the way, is how similar it looks to the map of changes from 2008 to 2012. In 2008, Obama easily defeated Republican presidential nominee John McCain in an actual landslide. In doing so, he flipped a lot of counties that George W. Bush had carried four years prior.

In 2012, though, many slipped back to the Republican nominee (then Mitt Romney). That change was particularly pronounced in the Upper Midwest — the region that would become problematic for Hillary Clinton in 2016.

There are 26 counties that flipped at least twice in a row. Four of those 26 counties flipped three times in a row. Three of those four went from McCain in 2008 to Obama in 2012 to Trump in 2016 to Biden this year. One — Kenedy County, Tex. — went from Obama to Romney to Clinton to Trump.

How counties have flipped in each election tells us a fair amount about the trends underlying the contest and American politics broadly. The shifts from 1972 to 1976, for example, depict the backlash to the Richard Nixon presidency and Jimmy Carter’s strength in the South. The shifts four years later reflect the rejection of Carter.

Here’s each year since 1976. (The flips are relative to the two-party vote.)

What the most recent map illustrates is how entrenched American politics tends to be. From 1972 to 1984, there were 39 counties which flipped three times in a row, mostly from Nixon to Carter to Reagan to Walter Mondale.

It didn’t help Mondale much. He lost by one of the largest margins in American history.