America Now Holds One Presidential Election, Not Three Thousand

Between 2020 and 2024 the country moved toward Donald Trump, and almost every place in it moved together. Of the 3,111 counties we can match across both elections, 86.8 percent shifted within five percentage points of the national shift. Hardly anything flipped: 86 counties changed sides, in a nation that still runs more than three thousand separate county tallies every four years.

Run the same test on 1972 to 1976 and the figure is 20.9 percent. Back then a presidential swing wasn't one number — it was thousands. Georgia counties lurched toward Jimmy Carter by enormous double-digit margins while pockets of the Mountain West barely moved or went the other way; 1,565 counties — roughly half the country's map — changed parties in a single cycle. The standard deviation of county swings was 26.0 points. By 2020–24 it had collapsed to 3.3.

That collapse is the cleanest single measurement we know of for the nationalization of American elections. Below, every presidential cycle pair since 1972, drawn the same way: each county's swing minus the national swing, so a county that moved exactly with the country sits at zero. Watch the distribution narrow from a smear to a needle.

The variance is gone

Three numbers, tracked across all thirteen cycle pairs, tell the same story. The dispersion of county swings — 26.0 points in 1972–76, still in the teens through 1984 — grinds down to 3.3 points by 2020–24 (vote-weighted, the fall is 20.0 to 4.3). The cycle-to-cycle correlation of county margins climbs from 0.27 in 1972–76 to 0.99 in each of the last two pairs: tell me a county's margin in the previous election and I can essentially tell you its margin in the next one. And the most concrete casualty, county party flips, falls off a cliff — 1,565 in 1972–76, still 829 as recently as 1996–2000, then 86 in 2020–24.

Two honest qualifiers about the starting point. The 1972–76 pair is partly mechanical: Nixon's forty-nine-state landslide had to revert, and the South was mid-realignment, so some of that smear is bounce-back rather than localism. And vote shares before 2000 are stored at two-decimal precision, which adds a little artificial noise — well under 1 point — to the early standard deviations. Neither comes close to explaining a fall from 26.0 to 3.3, and the downtrend holds across every subsequent pair.

It also holds at long horizons, which rules out the boring explanation that elections are simply quieter now. County margins sixteen years apart — four full presidencies, new candidates, new issues, new coalitions — correlated at 0.76 between 1972 and 1988. Between 2008 and 2024 they correlated at 0.87. The electoral map is not just stable from one election to the next; it is becoming stable across political generations.

2016 was the last local election

The descent is not monotone, and the one modern interruption is the most interesting thing in the series. The 2012–16 pair — the Trump realignment — spikes to a swing dispersion of 10.5 points, with only 23 percent of counties within five points of the national swing, dispersion the country hadn't produced since the 1980s. White, non-college, rural counties bolted right while educated metros edged left, and the smear briefly reappeared. Then it vanished: 2016–20 and 2020–24 are the two most uniform swings in the entire record. 2016 redrew the map; everything since has shaded it in together.

What happened to the places that did the redrawing? The 208 pivot counties — counties that voted for Barack Obama in both 2008 and 2012 and then for Trump in 2016, the most-watched "swing" geography in America — never swung back. Their average two-party margin ran +12.5 in 2008 and +8.7 in 2012, snapped to −12.2 in 2016, sat at −12.2 in 2020, and reached −17.1 in 2024, their reddest year yet (vote-weighted, −10.5). Twenty-six of them voted Democratic in 2020; twelve did in 2024.

The pivot counties sit inside a broader, slower realignment that the uniform-swing era has been quietly compounding. Measured from 2012 to 2024, 85.9 percent of counties got redder (the vote-weighted mean shift is −3.9 points, because big blue metros moved least). The sorting variable is education: vote-weighted, counties in the bottom quintile of college attainment shifted −22.9 points over those twelve years while the top quintile shifted +1.9 — a 24.8-point gradient. The single biggest mover among counties casting at least ten thousand votes is Starr County, Texas, on the Rio Grande: −89.85 points since 2012. Uniform swings, in other words, are not frozen politics. The map still drifts — it just drifts everywhere at once, along national demographic lines rather than local ones.

Twin elections, alien landscapes

Here is the neatest controlled comparison the data offers. Nationally, 1976 and 2024 are near-twins: Carter won the two-party vote by 1.8 points; Trump won it by 1.5. Hold the closeness of the race constant and look underneath.

In 1976, 548 counties — 17.6 percent, casting 22.4 percent of all votes — were decided by less than five points. In 2024, 155 counties were: 5.0 percent of counties, 12.9 percent of votes. The median county's winning margin tripled, from 14.3 points to 43.9. Counties decided by 40-plus points went from 331 (about 11 percent) to 1,722 — 55 percent of the national map, in an election the country split almost evenly.

This chart is a direct extension of Bill Bishop and Robert Cushing's Big Sort series. Their canonical numbers — landslide counties cast 26.8 percent of votes in 1976 and 48.3 percent in 2004 — replicate in our pipeline at 26.4 and 48.3 (we use two-party margins; they used raw pluralities). Carried forward, the series peaks at 60.8 percent in 2016 and stands at 57.4 percent in 2024. By population rather than votes, the landslide share runs 46 percent in 2000 to 58 percent in 2024. The deep-landslide end is growing fastest: counties decided by 50-plus points numbered 170 in 1976 and 126 in 1996; in 2024 there were 1,283 of them — a 10.2-fold increase since 1996.

Of 3,098 counties observed in every one of the fourteen cycles, the number that stayed within five points in all of them is zero. The competitive county is not declining. As a permanent feature of the landscape, it never existed — but the sometimes-competitive county is vanishing: 1,768 counties were within five points at least once from 1972 to 1996; only 911 were from 2000 to 2024.

One election, on a map

Color every county by its swing and the two eras barely look like the same country. The 1972–76 map is a patchwork — a continent of deep-blue Southern reversion shot through with counties moving every direction at every magnitude. The 2020–24 map is close to a monochrome wash: nearly everything a shade of light red, nearly everything the same shade of light red.