The Country Moved Right in 526 Counties

The sentence you will hear most between now and November is that the country moved right in 2024. It is true, and there is a county number that appears to confirm it: 2,756 of the 3,111 counties with a complete 2000–2024 series moved toward Republicans in raw margin. That is 88.6% of the map.

Now subtract what the country did that year. The national two-party margin went from D+4.52 in 2020 to R+1.52 in 2024, a move of 6.04 points toward Republicans. Hand every county that 6.04 points for free and ask which ones moved right beyond it. The answer is 526.

Those two numbers describe the same election, and the gap between them is the argument here. 2024 was not a broad rightward drift that happened to total six points nationally. It was six points of national tide arriving nearly everywhere at once, plus a real rightward move in a small, dense, identifiable set of places. Reading the first number as a fact about the map is the error we expect to see repeated all autumn, and it produces bad reasoning about a midterm.

The tide and the county

Lean is a county's two-party margin minus the nation's. Relative swing is the change in that lean between cycles — the county's own movement, with the national environment removed. A raw swing tells you what happened in a county. A relative swing tells you whether the county did it.

2016 is the control that makes 2024 legible. Its raw number looked like 2024's: 2,765 counties moved right, 88.9% of the map. But the national swing that year was only 1.80 points, so there was barely any tide to subtract, and the relative figure came out nearly identical — 2,664 counties, 85.6%. That was a map that genuinely moved. 2024 has the same raw headline and one-fifth of the relative one.

Where the six points came from

The 526 counties that moved right on their own cast 31.4% of the two-party vote — 47,702,180 two-party votes of 151,823,115, as summed from the county file. That is not a fringe. But their share of the change is far larger than their share of the vote.

Across all 3,111 counties the Democratic column fell by 6,225,455 votes between 2020 and 2024, and the Republican column rose by 3,089,438. Inside those 526 counties alone, Democratic votes fell by 4,313,450 — 69.3% of the entire national decline — while Republican votes rose by 1,455,036, or 47.1% of the national gain.

Thirty-one percent of the two-party vote produced roughly seven-tenths of the Democratic losses.

Splitting the national move by column gives the same shape. Summing these 3,111 counties puts the 2020 margin at D+4.51. Hold the Republican total there and let only the Democratic column change as it did, and the margin lands at D+0.51 — a move of 4.00 points. Hold the Democratic total and let only the Republican column change, and it lands at D+2.47, a move of 2.04 points. The columns are not strictly additive, so this is a decomposition by approximation rather than by identity. The ratio is not close enough for that to matter.

Ten counties

Rank counties by how much they actually pushed the national margin — each county's share of the two-party vote multiplied by its relative swing — and the top ten are these. They cast 7.69% of the national two-party vote. Between them, Democratic votes fell by 1,679,251 and Republican votes rose by 367,489: 27.0% of the country's entire Democratic decline and 11.9% of its Republican gain, out of one-thirteenth of the two-party vote.

Los Angeles is the largest and clearest case. Its Democratic total fell from 3,028,885 to 2,417,109, down 20.2%. Its Republican total rose from 1,145,530 to 1,189,862 — an increase of 44,332 votes against a Democratic loss of 611,776. Ballots cast fell from 4,264,365 to 3,728,427. Whatever happened there in 2024 was overwhelmingly subtraction from one column, not addition to the other.

Miami-Dade is where the same movement crossed a line: its lean went from D+2.87 to R+10.01 and the county changed sides, Democratic votes down 22.3% while Republican votes rose 13.7%. Bronx moved further against the nation than any of the ten, 16.50 points; Queens is second at 14.95.

What this does not show

• The 83.1% of counties that moved toward Democrats relative to the nation did not move toward Democrats. Republicans carried 2,269 of those 2,584 counties, and the Democratic vote total fell in 1,918 of them. Relative swing is defined net of the national move, so it cannot tell you which way the country went. Reading 83.1% as evidence the map moved left is the same error as reading 88.6% as evidence it moved right, pointed the other way.
• The county file records ballots, not voters. It cannot distinguish a 2020 Democratic voter who stayed home from one who voted Republican, and nothing here should be read as choosing between them. That distinction is the whole question about these ten counties and we do not have the data to settle it.
• The vote sums here — 74,747,402 Democratic and 77,075,713 Republican in 2024 — are summed from the county file across 3,111 counties. They are not the certified national result and should not be quoted as one.

What survives those three is still worth arguing about. The rightward move of 2024 was concentrated to a degree that the phrase "the country moved right" conceals, and the concentration sat in large, dense, heavily Democratic counties where one column emptied out rather than the other filling up.

We will not tell you what that means for November, and we will not publish a number that pretends to. We will say what would have to be true. If those 4.3 million missing Democratic votes were people who changed their minds, they are somebody else's voters now, and the 526 counties should hold their new lean in a lower-turnout year. If they were people who did not vote, the lean should snap back when the electorate changes shape. This file cannot settle that. It records presidential years only — 2000 through 2024 — so the next observation it offers is 2028, not a midterm. Whatever November shows, the answer will have to come from returns we do not currently hold. Anyone answering it in September is guessing.