↖︎ Vishal Singh

Precinct returns · 2016 · 2020 · 2024

The Unpacking

For a generation the Big Sort ran one way: Americans piling into landslide neighborhoods where they never met the other side. Below the county line, between 2016 and 2024, it went briefly into reverse — and the Democratic city, the most packed political geography in America, did most of the unpacking.

The standard story of American political geography is a one-way ratchet. Bill Bishop called it the Big Sort: decade after decade, more of us live in landslide places, packed in among people who vote as we do. Jonathan Rodden gave it an electoral edge in Why Cities Lose — Democrats, crammed into hyper-blue urban precincts, "waste" votes by winning their own neighborhoods 90 to 10, while Republicans spread more efficiently across the map.

Three cycles of precinct returns — the scale of an actual polling place, far below the county lines most maps stop at — show the ratchet slipping. From 2016 to 2024 the country's neighborhoods grew less sorted, not more: more voters in competitive precincts, fewer in blowouts. And the single clearest movement is the one Rodden's thesis turns on. The Democratic packing premium — how much more lopsided the average Democratic vote's precinct is than the average Republican's — fell from nearly 12 points to 3. The deep-blue precinct started to thin out.

None of this undoes a half-century of sorting in eight years, and most of it traces to one realignment (more on that below). But it is a real, measurable loosening, and you can only see it from underneath the county.

11.6 → 3.0
Democratic packing premium (D−R gap in share of votes cast in 80–20 precincts), 2016 → 2024
24.8% → 15.6%
Share of Democratic votes cast in 80–20 landslide precincts — the packing that fell
15.8% → 18.7%
Voters in a competitive precinct (under 10 pts) — rising, against the Big Sort
−7.2 pts
Drop in Democratic packing inside large metro cores — vs −0.7 in rural counties
I · What happened

The packed party unpacked

Measure packing simply: of all the votes a party won, what share came from precincts it carried by 60 points or more — an 80–20 blowout. In 2016 the two parties were wildly asymmetric. 24.8% of Hillary Clinton's votes came from 80–20 precincts, against just 13.2% of Donald Trump's — a near-12-point gap, the statistical signature of Democrats stacked into landslide cities while Republicans spread thinner across suburb and country.

By 2024 the gap had all but closed. The Republican figure barely moved (13.2% to 12.6%); rural red precincts stayed as red as ever. But the Democratic figure collapsed to 15.6%. Kamala Harris won the same kinds of places Clinton did, but by smaller margins inside them — the 90–10 precinct became the 75–25 precinct. The packing premium fell from 11.6 points to 3.0.

The packing premium collapsed

Share of each party's two-party votes cast in precincts it won by 60+ points (an 80–20 landslide), by cycle. The shaded gap is the Democratic packing premium.

Mode-safe two-party precinct returns (MEDSL). The Republican share is roughly flat across the three cycles; the Democratic share falls by nine points, closing the gap from 11.6 to 3.0. "Packing" here is descriptive geography, not a redistricting claim — but it is the same quantity that makes Democratic votes electorally inefficient.
II · How it shows up

The distribution slid toward the center

Line up every vote by the margin of the precinct it was cast in, and the shape of the country redraws itself across the three cycles. The deep-Democratic tail on the right — the 80–20 and 90–10 precincts — visibly deflates from 2016 to 2024, and the mass piles back toward the middle. The Republican tail on the left holds far steadier. This is the asymmetry from the first chart, seen whole: the country didn't de-sort evenly, the blue end did the moving.

Where the votes piled up, three cycles

Share of the two-party vote cast in precincts of each margin, by year. Same election geography, viewed as a distribution.

Two-party margin = (Dem − Rep)/(Dem + Rep), vote-weighted; each curve integrates to 100%. The right (Democratic) tail thins markedly between 2016 and 2024 while the left (Republican) tail is comparatively stable — and the competitive center fills in. A naive read of the Big Sort would predict the opposite.
III · Where it happened

The cities converged inside themselves

Because precincts are redrawn every cycle, you cannot follow one precinct through time — but you can follow a county, and ask whether its precincts grew closer together or further apart. For each county with enough precincts in both elections, the map below shows the change from 2016 to 2024 in how spread out its neighborhoods are (the vote-weighted standard deviation of its precinct margins). Nationally that internal spread fell 2.5 points; the teal counties are where neighborhoods converged.

Change in how divided each county is inside, 2016→2024

Vote-weighted standard deviation of a county's precinct margins in 2024 minus 2016, in points. Teal = neighborhoods converged; orange = pulled further apart. Counties need 10+ reported precincts in both years; others are beige.

2,341 counties qualify. The convergence is broad but concentrated in and around metropolitan areas; rural counties, already homogeneous, had little room to converge. Individual county values are noisier than the national pattern because precinct boundaries are redrawn between cycles — read the color field, not single counties.
IV · Who unpacked

It was overwhelmingly a big-city story

Sort the same counties by how urban they are, and the unpacking is lopsided. In large metro cores, the share of the county's Democratic votes sitting in 80–20 precincts fell by 7.2 points between 2016 and 2024. In rural counties it barely moved — 0.7 points. The 2024 realignment that gets described in demographic terms — Democrats slipping with Hispanic, younger, and non-college voters who are concentrated in cities — lands, geographically, as the de-densification of the deep-blue precinct.

The unpacking was concentrated in the metro core

Change 2016→2024 in the share of a county's Democratic votes cast in its 80–20 precincts, vote-weighted mean by USDA urbanization class.

USDA Rural–Urban Continuum Codes (2024 county classification): "Large metro core" = code 1; "Other metro" = 2–3; "Small metro / micro" = 4–6; "Rural" = 7–9. Bars are negative everywhere — packing fell across the board — but the metro core fell roughly ten times as far as the countryside.

What this shows — and what it doesn't

This is a short window and a partial reversal, not the end of the Big Sort. The decades-long sort into landslide counties is real and largely intact; what these three cycles show is that beneath it, between 2016 and 2024, neighborhood-level sorting eased — and most of that is the 2024 realignment compressing Democratic margins in the places Democrats were most packed. Whether it persists depends on whether that realignment does. The measure is robust to one obvious worry: precincts were redrawn each cycle, and 2024 actually has more precincts than 2016, which should reveal more extremity, not less — so the de-sorting is, if anything, understated. But it is geography, not people: precincts are not a panel, so we describe how the distribution moved and which counties converged, never a precinct or a voter followed over time. The "who" here is a county type, not an individual — attaching neighborhood demographics would need precinct boundaries this analysis does not use. Everything is two-party (Trump vs the Democratic nominee); turnout and third parties are set aside.