Chapter 11.1

Precincts within counties

In 2024, 65.0% of the two-party presidential votes in the sample were cast in precincts decided by 20 points or more, against 57.2% in counties decided by that margin and 56.6% in a simulation that assigns voters to precincts at random.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 30 September 2026

DraftFigures and numbers may change as the analysis is rebuilt.what is still being checked

The 64 Louisiana parishes are missing because the 2024 precinct file holds about half of their votes. Absentee, early-vote and provisional buckets are kept as reporting units and checked in a side calculation. The file has no precinct boundaries, so the chapter cannot map neighborhoods; that belongs to the planned Neighborhoods explorer.

In 2024, 65.0% of the 149.0 million two-party presidential votes in this chapter's sample were cast in precincts decided by 20 points or more, against 57.2% cast in counties decided by that margin. At the other end, 18.5% of votes were cast in precincts decided by under 10 points, against 23.1% in counties. A county's margin is the average of its precinct margins weighted by their votes, and this chapter measures how far the precinct margins vary around that average and which counties vary most. The standard deviation of precinct margins inside a county, called its spread here, has a mean across counties of 23.3 (95% CI 22.5 to 24.1) points when each county counts in proportion to its votes. It is higher in metro counties and in counties that lean Democratic, and it varies widely among counties that finished near a tie, with a margin under 5 points and at least 20 precincts.

A two-party margin is the Democratic vote minus the Republican vote over their sum, in points, and a positive margin is a Democratic lead. A margin of 20 points gives the winner 60% of the two-party vote, and a margin of 10 points gives the winner 55%. The spread is weighted by each precinct's two-party votes and stated in the same points. A county that cast equal numbers of votes in precincts at Democratic +30 and at Republican +30 would have a margin of zero and a spread of 30, which is 15 points of Democratic vote share. Third-party votes are left out.

Two Pennsylvania counties that finished nearly even show what spread measures. Figure 1 gives the share of each county's votes by the margin of the precinct where they were cast. Bucks County finished at a margin of −0.1 points, with 304 precincts and a spread of 14.8, and 50.0% of its votes were cast in precincts decided by under 10 points. Its most Democratic precinct had a margin of 42.0 points and its most Republican a Trump lead of 33.7. Centre County finished at +2.9 with 87 precincts and a spread of 36.3, and only 5.5% of its votes were cast in precincts that close. The most Democratic precinct in Centre County had a margin of 65.0 points and the most Republican a Trump lead of 75.2. The Neighborhoods explorer will map these counties and others below the precinct level.

Figure 1

Bucks County precincts cluster near an even split, while Centre County precincts form two groups, one on each side

Share of each county's 2024 two-party votes cast in precincts in each 10-point band of margin

Share of the county's votes (%)

Show data

Data for Bucks County precincts cluster near an even split, while Centre County precincts form two groups, one on each side
county	n	share	x0	x1
Bucks County, PA	0	0	−100	−90
Bucks County, PA	0	0	−90	−80
Bucks County, PA	0	0	−80	−70
Bucks County, PA	0	0	−70	−60
Bucks County, PA	0	0	−60	−50
Bucks County, PA	0	0	−50	−40
Bucks County, PA	4	1.05	−40	−30
Bucks County, PA	25	8.35	−30	−20
Bucks County, PA	54	17.71	−20	−10
Bucks County, PA	83	27.62	−10	0
Bucks County, PA	68	22.48	0	10
Bucks County, PA	40	13.77	10	20
Bucks County, PA	15	5.06	20	30
Bucks County, PA	13	3.63	30	40
Bucks County, PA	2	0.315	40	50
Bucks County, PA	0	0	50	60
Bucks County, PA	0	0	60	70
Bucks County, PA	0	0	70	80
Bucks County, PA	0	0	80	90
Bucks County, PA	0	0	90	100
Centre County, PA	0	0	−100	−90
Centre County, PA	0	0	−90	−80
Centre County, PA	1	0.333	−80	−70
Centre County, PA	5	4.17	−70	−60
Centre County, PA	8	6.12	−60	−50
Centre County, PA	12	6.15	−50	−40
Centre County, PA	7	7.06	−40	−30
Centre County, PA	8	10.56	−30	−20
Centre County, PA	5	5.43	−20	−10
Centre County, PA	3	4.12	−10	0
Centre County, PA	2	1.34	0	10
Centre County, PA	7	11.96	10	20
Centre County, PA	8	14.10	20	30
Centre County, PA	7	12.99	30	40
Centre County, PA	7	9.71	40	50
Centre County, PA	3	2.83	50	60
Centre County, PA	4	3.13	60	70
Centre County, PA	0	0	70	80
Centre County, PA	0	0	80	90
Centre County, PA	0	0	90	100

Bucks precincts cluster around an even split; Centre's form two groups, one on each side, with few votes near even. Each bar is the share of the county's two-party votes cast in precincts in one band of margin. The spread is 14.8 points in Bucks and 36.3 in Centre.Note: The number in the tooltip is the number of precincts in the band. The two counties were chosen for the contrast between them, not at random.Note: Full counts of votes: there is no sampling, so the shares carry no confidence interval.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

How the gap between precincts and counties changes across scales, from states to blocks, is the subject of Chapter 11.2, Sorting by geographic scale, 2016 to 2024 (planned). This chapter asks what the precinct distribution looks like inside each county, which is the information that a county color leaves out. County-level maps supported the argument that Americans sort into like-minded places [1], a claim that others contested [2], and address-level data show that many voters live with almost no exposure to voters of the other party [3]. Precinct returns sit between the two scales.

The sample covers 175,345 reporting units in 3,043 counties in 48 states and the District of Columbia; Alaska and Louisiana are absent.The precinct file from the MIT Election Data and Science Lab reports votes by whatever unit each state or county election office uses, so a precinct here is a reporting unit and differs in size and kind by state. Alaska is left out because its 2024 county codes are election districts. So are the 64 Louisiana parishes, where the precinct rows hold 51.5% of the certified vote, and 6 other counties whose precinct rows do not add to the certified county total. The 70 excluded counties hold 2.0% of certified votes, and Alaska holds a further 0.2%. 190 counties in the sample have too few reporting units for a spread. A county needs at least 5 reporting units to have a spread, which leaves 2,853 counties. The returns are full counts, so shares and spreads carry no sampling error; a statistic across counties, such as a mean or a correlation, carries a 95% interval from resampling counties 2,000 times.

Precincts are more lopsided than the counties that contain them

Figure 2 counts the same votes by the margin of the precinct they were cast in, by the margin of the county, and by the margin of a simulated precinct. Precincts hold more of the votes in landslides and fewer in close contests. Precincts decided by 40 points or more held 36.4% of votes, against 25.4% for counties decided by that margin.

Figure 2

Precincts are more often lopsided, and less often close, than counties or than random assignment of voters would make them

Share of 2024 two-party presidential votes cast in units decided by each margin, in percent of votes

Precincts
Counties
Simulated precincts

Show data

Data for Precincts are more often lopsided, and less often close, than counties or than random assignment of voters would make them
band	share	unit
Under 10 points	18.52	Precincts
Under 10 points	23.13	Counties
Under 10 points	23.32	Simulated precincts
20 points or more	64.97	Precincts
20 points or more	57.25	Counties
20 points or more	56.58	Simulated precincts
40 points or more	36.37	Precincts
40 points or more	25.38	Counties
40 points or more	24.96	Simulated precincts

Compare the precinct bar with the other two in each group: precincts hold more of the votes in landslides and fewer in close contests. A margin of 20 points gives the winner 60% of the two-party vote, 40 points or more is very lopsided, and under 10 points gives the winner under 55%. In the simulation each precinct keeps its number of votes and its county's Democratic share, its Democratic votes are drawn at random, and the bar is the mean of 200 runs.Note: Full counts of votes: there is no sampling, so the shares carry no confidence interval.Note: The simulated bars are means over 200 runs and move by at most 0.4 points across runs.Note: The sample is 175,345 reporting units in 3,043 counties: Alaska, the 64 Louisiana parishes and 6 other counties whose precinct rows do not match the certified total are left out.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

Because each county margin is the average of its precinct margins weighted by their votes, precinct margins are at least as spread out as county margins, counting each vote once. That alone does not put more votes in landslide precincts, since a county decided by a landslide can split into precincts of which some are not, so the simulation tests whether chance would. In each of 200 runs, every precinct keeps its number of votes and its county's Democratic share, and its Democratic votes are drawn at random. The simulated precincts hold 56.6% of votes in landslide precincts, 23.3% in close ones and 25.0% in very lopsided ones, within about 0.7 points of the county shares, and each share moves by at most 0.4 points from run to run. Small units alone therefore do not produce the gap between precincts and counties. Precinct margins inside a county vary more than random assignment of voters would make them.

Random assignment still leaves some spread. If the voters of a county were assigned to its precincts at random, the spread would be 3.1 points on average across counties, weighted by votes, and 3.4 points in the median near-tied county.

Spread is widest in Mississippi and Alabama

Across the 2,853 counties with at least 5 reporting units, the median county, counted once, has a spread of 15.6 (95% CI 15.0 to 16.0) points. That is lower than the vote-weighted mean because counties with more votes tend to have wider spreads, as the next sections show. A spread of 30 points or more occurs in 368 counties, which hold 18.7% of the votes cast in the 2,853 counties. Figure 3 maps the spread, and its second view maps the county margin.

Figure 3

Spread inside counties is widest in Mississippi and Alabama and scattered across the Southwest and Plains

Vote-weighted standard deviation of precinct margins inside each county, 2024, in points of two-party margin; counties with at least 5 reporting units

Loading county values

Spread of precinct margins inside the county (points)

Values outside 0 to 60 take the end colors.

No data (3,143 counties)

Select a class to show only those places.

Find a county

Hover or tap a county for its value, or search by name. Zoom with the buttons on the map, then drag to move it.

Alaska draws as no data. Its election returns are reported by election district, and the district codes cannot be matched to boroughs and census areas.

Darker counties have precinct margins that vary more. Of the 33 counties with a spread of 50 points or more, 20 are in Mississippi or Alabama, and the rest include counties in Arkansas, Arizona, South Dakota, Utah, Montana and Nebraska. Spread is the standard deviation of a county's precinct margins, weighted by each precinct's two-party votes, across the 2,853 counties with a spread. Use the switch to see the counties by margin.Note: Counties with fewer than 5 reporting units (190 of the 3,043 in the sample), the 64 Louisiana parishes, 6 other counties whose precinct rows do not match the certified total, Alaska and Kansas City, Missouri are drawn as no data.Note: A county near the floor of 5 reporting units has a spread that one unit can move a long way.Note: Full counts of votes: there is no sampling, so county margins and spreads carry no sampling error.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

The vote-weighted mean spread is 38.2 (95% CI 34.9 to 41.5) points in Mississippi and 32.2 (95% CI 25.2 to 38.3) in Alabama, the two highest of the 48 states and the District of Columbia. The units that election offices report differ in size by state. The median precinct holds 227 two-party votes in Kansas and 2,083 in Connecticut, and California reports 578 precincts per county on average while North Dakota reports 7. Smaller units can show more spread, so a comparison of spread across states mixes the electorate with the way votes are counted. Counties with few reporting units also have spreads that one unit can move a long way: 4 of the 33 counties with a spread of 50 points or more have fewer than 10 reporting units. The state controls in Figure 6 address part of that mix.

Among the 318 counties with at least 100,000 two-party votes, the widest spreads are in Kings County, New York (54.9 points), Jefferson County, Alabama (53.5) and Mobile County, Alabama (50.2). The narrowest are in Arlington County, Virginia (6.5) and the District of Columbia (5.9), whose margins are +59.9 and +86.6 points.

Spread falls from metro to less urban counties

Figure 4 groups counties by Rural-Urban Continuum Code. The vote-weighted mean spread falls from 24.8 (95% CI 23.5 to 26.1) points in the largest metro counties to 22.7 in smaller metro counties, 19.4 in the larger nonmetro counties and 15.5 (95% CI 14.6 to 16.5) in the least urban ones. The two ends of the scale also differ in lean: the vote-weighted mean margin is +11.7 (95% CI 7.3 to 15.7) points in the largest metro counties and −45.2 (95% CI −47.6 to −42.7) in the least urban, so the next section separates lean from size. The demographics panel has no code for 10 counties, which are left out of the bars: the Connecticut counties, Kansas City, Missouri, and Oglala Lakota County, South Dakota.

Figure 4

Spread of precinct margins falls from the largest metro counties to the least urban ones

Vote-weighted mean of county spreads by Rural-Urban Continuum Code group, 2024, in points of two-party margin

Show data

Data for Spread of precinct margins falls from the largest metro counties to the least urban ones
Place type	Description	Rural-Urban Continuum Code	Mean spread (points)	Interval, low	Interval, high	Counties
Largest metro	Metro area of 1 million or more	1	24.8	23.5	26.1	419
Smaller metro	Smaller metro area	2 to 3	22.7	21.9	23.5	690
Larger nonmetro	Nonmetro, larger towns	4 to 7	19.4	18.7	20.1	861
Least urban	Nonmetro, smaller towns and open country	8 to 9	15.5	14.6	16.5	873

Read the bars from top to bottom: the mean spread falls at each step away from the largest metro counties. Each bar averages county spreads with the county's two-party votes as weights, and whiskers are 95% intervals from a county bootstrap. Counties are grouped by their 2023 Rural-Urban Continuum Code; the data table lists the codes in each group.Note: The number in the tooltip is the number of counties in the group. 2,853 counties have at least 5 reporting units; 10 of them have no code in the demographics panel (the Connecticut counties, Kansas City, Missouri, and Oglala Lakota County, South Dakota) and are left out of the bars.Note: The returns are full counts with no sampling error. The whiskers show how much a mean would move with a different set of counties, from a county bootstrap.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated; county demographics panel (ACS 2024, Rural-Urban Continuum Code 2023)

Spread rises with Democratic lean, partly through county size

Figure 5 plots each county's spread against its margin. Weighted by votes, the correlation is +0.41 (95% CI 0.33 to 0.49) across 2,853 counties. The binned means rise from 6.7 (95% CI 5.2 to 8.0) points in counties near Trump +82 to 25.4 (95% CI 23.4 to 27.5) in counties near Trump +14. The 6 bins from there to Harris +34 have means between 24.1 and 28.4. The highest bin, 31.1 (95% CI 24.7 to 36.8) near Harris +43, has an interval that overlaps each of them, and the interval for its difference from each of them includes zero except for the lowest, near Trump +4, where the difference is 6.9 (95% CI 0.7 to 12.9) points. The data show a rise and then a plateau, with the highest bin distinguishable from only the lowest of the bins before it. The 12 counties near Harris +66 average 19.3 (95% CI 15.3 to 23.9), which is 11.8 (95% CI 3.9 to 18.9) points below the highest bin, and few counties are that blue.

Figure 5

Spread rises from the reddest counties to about Trump +14, then levels off

Vote-weighted standard deviation of precinct margins inside each county against the county two-party margin, 2024; one dot per county, sized by two-party votes

Dot area: Two-party votes360,0001,300,0003,600,000

Show data

Data for Spread rises from the reddest counties to about Trump +14, then levels off
County	Margin (points)	Spread (points)	Two-party votes	Precincts
Autauga, AL	−46.7	17.4	27,923	21
Baldwin, AL	−58.7	16.3	120,732	61
Barbour, AL	−14.8	32.1	9,764	18
Bibb, AL	−64.8	24.6	9,191	10
Blount, AL	−81.6	6.7	27,930	26
Bullock, AL	+46.1	42.8	4,084	10
Butler, AL	−22.8	41.2	8,423	23
Calhoun, AL	−45.1	37.0	48,106	44
Chambers, AL	−23.4	30.5	14,116	18
Cherokee, AL	−75.9	9.7	12,911	25
Chilton, AL	−72.5	15.5	19,618	20
Choctaw, AL	−24.0	55.1	6,618	34
Clarke, AL	−17.1	41.6	11,892	31
Clay, AL	−70.5	19.7	6,727	17
Cleburne, AL	−84.1	8.2	7,593	16
Coffee, AL	−58.4	26.2	22,096	29
Colbert, AL	−46.8	26.4	26,851	36
Conecuh, AL	−14.0	52.0	6,003	28
Coosa, AL	−43.5	48.4	5,236	14
Covington, AL	−72.8	15.1	16,991	27
Crenshaw, AL	−54.9	31.7	6,457	17
Cullman, AL	−81.1	9.6	42,743	51
Dale, AL	−52.7	20.9	18,960	20
Dallas, AL	+32.7	47.3	15,426	31
Dekalb, AL	−74.4	10.2	29,391	46
Elmore, AL	−52.5	22.7	41,148	31
Escambia, AL	−46.6	39.9	14,848	31
Etowah, AL	−56.1	37.1	45,680	33
Fayette, AL	−72.5	21.8	8,300	26
Franklin, AL	−73.8	17.8	11,985	26
Geneva, AL	−77.4	10.6	12,320	23
Greene, AL	+55.9	21.4	4,018	16
Hale, AL	+6.9	37.0	7,237	16
Henry, AL	−51.1	24.4	9,252	16
Houston, AL	−48.2	31.8	43,821	30
Jackson, AL	−71.9	8.8	23,349	32
Jefferson, AL	+10.6	53.5	293,235	179
Lamar, AL	−76.4	15.6	6,839	27
Lauderdale, AL	−52.0	27.6	43,034	33
Lawrence, AL	−62.7	37.2	16,007	31
Lee, AL	−28.2	28.7	71,818	27
Limestone, AL	−44.4	24.2	52,468	27
Lowndes, AL	+37.5	44.6	5,625	14
Macon, AL	+56.7	38.4	7,766	16
Madison, AL	−9.1	31.9	193,254	81
Marengo, AL	−3.8	38.7	9,626	20
Marion, AL	−82.2	4.1	13,442	17
Marshall, AL	−72.2	7.1	39,987	31
Mobile, AL	−16.6	50.2	172,814	82
Monroe, AL	−23.3	50.6	9,747	29
Montgomery, AL	+31.1	40.0	88,423	53
Morgan, AL	−53.1	31.9	52,841	41
Perry, AL	+42.9	46.9	4,443	14
Pickens, AL	−23.5	47.9	8,853	21
Pike, AL	−25.3	41.8	13,123	36
Randolph, AL	−65.2	23.0	11,022	22
Russell, AL	+1.7	40.2	20,500	20
St. Clair, AL	−64.6	14.0	43,141	32
Shelby, AL	−41.3	16.8	112,753	40
Sumter, AL	+41.4	23.8	5,267	15
Talladega, AL	−33.9	30.2	32,998	28
Tallapoosa, AL	−49.9	37.2	19,859	28
Tuscaloosa, AL	−20.6	47.1	84,123	56
Walker, AL	−72.3	16.2	29,566	47
Washington, AL	−55.6	35.8	8,397	21
Wilcox, AL	+31.6	48.9	5,242	21
Winston, AL	−84.0	4.8	11,075	20
Apache, AZ	+19.2	57.9	31,667	44
Cochise, AZ	−23.4	26.6	58,232	54
Coconino, AZ	+20.2	34.4	69,080	75
Gila, AZ	−37.9	28.2	27,405	39
Graham, AZ	−48.6	34.3	15,044	23
Greenlee, AZ	−41.5	17.5	3,262	6
La Paz, AZ	−44.5	22.5	7,571	11
Maricopa, AZ	−3.5	24.2	2,031,547	907
Mohave, AZ	−56.1	9.4	109,725	24
Navajo, AZ	−17.4	52.7	50,234	17
Pima, AZ	+15.3	24.6	507,119	265
Pinal, AZ	−22.3	17.9	207,582	108
Santa Cruz, AZ	+18.8	19.2	18,950	31
Yavapai, AZ	−34.0	20.9	147,083	54
Yuma, AZ	−20.6	24.2	67,568	44
Arkansas, AR	−44.9	39.5	5,454	30
Ashley, AR	−48.7	40.0	6,919	30
Baxter, AR	−57.8	5.8	20,594	11
Benton, AR	−27.7	20.6	124,948	426
Boone, AR	−66.1	9.1	16,822	12
Bradley, AR	−39.4	44.8	3,174	14
Calhoun, AR	−63.1	18.1	2,053	8
Carroll, AR	−32.4	33.2	11,282	11
Chicot, AR	+4.0	49.0	3,454	7
Clark, AR	−20.9	35.2	7,485	20
Clay, AR	−62.8	9.1	4,875	17
Cleburne, AR	−69.0	9.7	12,536	34
Cleveland, AR	−68.5	22.5	3,328	17
Columbia, AR	−37.0	35.1	7,833	14
Conway, AR	−41.3	25.2	8,342	31
Craighead, AR	−38.3	27.2	36,362	53
Crawford, AR	−59.3	7.5	23,368	13
Crittenden, AR	+2.3	41.8	14,390	17
Cross, AR	−48.6	40.6	6,395	14
Dallas, AR	−30.0	47.2	2,280	9
Desha, AR	−4.9	45.7	3,443	21
Drew, AR	−34.4	29.2	6,253	17
Faulkner, AR	−33.1	29.8	53,109	51
Franklin, AR	−63.8	9.4	6,814	39
Fulton, AR	−63.4	9.0	4,946	10
Garland, AR	−37.1	20.5	41,374	52
Grant, AR	−70.0	9.4	7,947	18
Greene, AR	−62.3	8.9	15,552	11
Hempstead, AR	−40.5	41.1	5,969	24
Hot Spring, AR	−53.2	29.0	12,044	32
Howard, AR	−47.4	34.8	4,404	9
Independence, AR	−60.8	17.1	13,712	34
Izard, AR	−67.3	9.2	5,803	45
Jackson, AR	−49.6	30.7	4,692	45
Jefferson, AR	+20.4	59.6	21,270	169
Johnson, AR	−52.5	12.6	8,873	11
Lafayette, AR	−39.0	42.9	2,287	9
Lawrence, AR	−65.4	9.2	5,573	28
Lee, AR	+3.6	43.2	2,451	9
Lincoln, AR	−51.0	51.2	3,315	26
Little River, AR	−55.1	27.9	4,828	20
Logan, AR	−63.5	9.4	8,031	36
Lonoke, AR	−54.8	20.0	30,015	55
Madison, AR	−59.6	6.2	7,376	10
Marion, AR	−61.2	9.2	7,731	29
Miller, AR	−51.7	35.1	15,611	19
Mississippi, AR	−32.2	45.1	10,537	16
Monroe, AR	−16.0	48.4	2,387	24
Montgomery, AR	−64.5	9.0	3,632	16
Nevada, AR	−40.4	43.2	2,851	20
Newton, AR	−65.3	18.2	3,707	19
Ouachita, AR	−19.4	38.3	8,468	18
Perry, AR	−58.8	9.1	4,482	17
Phillips, AR	+13.5	39.5	4,852	10
Pike, AR	−74.0	14.3	4,306	11
Poinsett, AR	−64.5	9.4	6,966	12
Polk, AR	−71.8	7.5	8,132	17
Pope, AR	−53.5	18.9	23,605	55
Prairie, AR	−66.8	16.1	3,152	24
Pulaski, AR	+22.7	35.3	150,015	132
Randolph, AR	−65.0	10.6	6,505	29
St Francis, AR	+0.8	53.8	5,862	15
Saline, AR	−41.0	20.1	56,345	42
Scott, AR	−74.5	10.5	3,338	26
Searcy, AR	−73.2	8.1	3,816	15
Sebastian, AR	−38.5	27.2	44,371	180
Sevier, AR	−62.8	18.8	4,634	9
Sharp, AR	−63.9	10.4	7,294	25
Stone, AR	−63.1	7.6	5,897	18
Union, AR	−34.0	41.0	15,215	28
Van Buren, AR	−61.5	11.9	7,460	21
Washington, AR	−6.9	33.1	94,022	224
White, AR	−62.6	14.1	30,155	92
Woodruff, AR	−33.1	45.0	2,273	20
Yell, AR	−61.9	16.1	6,360	28
Alameda, CA	+56.0	26.1	640,028	822
Alpine, CA	+32.7	20.6	722	5
Amador, CA	−28.5	20.3	21,742	79
Butte, CA	−3.1	31.4	91,203	212
Calaveras, CA	−28.8	25.8	25,806	50
Colusa, CA	−29.0	20.6	6,845	19
Contra Costa, CA	+39.3	24.3	510,211	1,202
Del Norte, CA	−16.8	18.1	10,257	36
El Dorado, CA	−12.3	25.0	108,695	260
Fresno, CA	−4.5	27.6	316,996	684
Glenn, CA	−35.7	27.4	9,928	113
Humboldt, CA	+29.8	30.0	61,284	164
Imperial, CA	−0.7	23.3	52,228	215
Inyo, CA	−2.9	19.6	8,650	28
Kern, CA	−21.6	26.4	275,461	1,082
Kings, CA	−23.4	21.6	40,435	174
Lake, CA	−1.4	14.4	25,945	73
Lassen, CA	−55.2	16.0	10,972	65
Los Angeles, CA	+34.0	29.4	3,605,749	4,808
Madera, CA	−21.3	21.1	53,240	112
Marin, CA	+65.7	14.9	140,102	217
Mariposa, CA	−21.5	25.8	9,159	32
Mendocino, CA	+28.1	25.1	37,197	240
Merced, CA	−3.9	28.4	82,772	621
Modoc, CA	−48.1	14.3	3,874	30
Mono, CA	+21.1	31.4	5,808	22
Monterey, CA	+30.8	20.8	142,286	403
Napa, CA	+36.1	17.7	63,325	336
Nevada, CA	+12.7	31.3	59,904	180
Orange, CA	+2.8	25.4	1,346,071	2,899
Placer, CA	−8.7	23.9	227,596	468
Plumas, CA	−17.5	20.0	9,745	28
Riverside, CA	−1.3	24.9	914,400	1,711
Sacramento, CA	+20.4	27.2	633,476	1,249
San Benito, CA	+12.9	22.8	26,881	80
San Bernardino, CA	−2.1	24.1	738,081	2,738
San Diego, CA	+17.3	26.6	1,433,911	3,315
San Francisco, CA	+67.6	18.0	386,298	1,014
San Joaquin, CA	−0.9	25.2	255,429	437
San Luis Obispo, CA	+11.2	28.3	146,169	251
San Mateo, CA	+52.1	15.9	319,349	523
Santa Barbara, CA	+27.6	31.9	178,735	424
Santa Clara, CA	+41.6	20.7	721,284	1,311
Santa Cruz, CA	+56.6	17.8	128,801	282
Shasta, CA	−37.4	16.0	86,661	128
Sierra, CA	−24.9	22.6	1,707	19
Siskiyou, CA	−19.9	29.1	20,790	105
Solano, CA	+23.8	24.9	184,102	366
Sonoma, CA	+47.8	14.5	242,395	755
Stanislaus, CA	−11.2	22.6	192,155	469
Sutter, CA	−32.2	14.8	38,388	48
Tehama, CA	−42.7	16.5	25,869	68
Trinity, CA	−9.8	13.5	5,428	14
Tulare, CA	−21.2	24.3	135,075	584
Tuolumne, CA	−22.4	17.8	28,098	78
Ventura, CA	+15.6	24.4	374,949	1,715
Yolo, CA	+37.6	31.3	89,249	95
Yuba, CA	−26.6	20.2	29,216	63
Adams, CO	+9.3	20.1	227,080	275
Alamosa, CO	−11.1	20.4	7,301	8
Arapahoe, CO	+20.7	16.7	316,036	394
Archuleta, CO	−14.4	9.9	9,122	8
Baca, CO	−72.0	7.4	1,961	9
Bent, CO	−39.7	24.7	2,141	5
Boulder, CO	+57.3	17.6	190,907	190
Broomfield, CO	+29.4	10.9	45,497	58
Chaffee, CO	+14.0	20.3	14,026	11
Cheyenne, CO	−79.2	10.3	1,038	5
Clear Creek, CO	+17.1	9.3	5,916	9
Conejos, CO	−18.3	27.8	3,985	10
Costilla, CO	+15.2	23.8	2,005	8
Crowley, CO	−48.9	4.9	1,653	6
Delta, CO	−36.5	23.3	18,978	20
Denver, CO	+57.7	16.6	353,399	301
Douglas, CO	−7.2	14.3	237,859	204
Eagle, CO	+25.1	17.1	27,091	34
Elbert, CO	−52.3	8.5	19,977	18
El Paso, CO	−10.1	21.0	370,530	324
Fremont, CO	−39.4	9.3	24,839	20
Garfield, CO	+2.1	38.2	29,621	27
Gilpin, CO	+13.2	16.6	3,983	6
Grand, CO	−0.8	28.9	9,565	12
Gunnison, CO	+30.2	31.2	10,576	15
Huerfano, CO	−8.8	9.5	4,311	5
Jefferson, CO	+20.0	14.1	351,380	296
Kit Carson, CO	−69.4	7.6	3,639	13
Lake, CO	+17.1	7.9	3,756	6
La Plata, CO	+19.2	33.6	34,701	33
Larimer, CO	+18.1	25.8	219,056	155
Las Animas, CO	−13.8	21.7	7,604	7
Lincoln, CO	−65.7	7.5	2,522	13
Logan, CO	−57.8	15.9	9,953	16
Mesa, CO	−24.9	19.2	89,412	67
Moffat, CO	−64.2	9.5	6,525	13
Montezuma, CO	−21.5	21.8	14,765	12
Montrose, CO	−33.3	17.0	25,058	21
Morgan, CO	−49.8	15.6	13,121	18
Otero, CO	−27.1	24.2	8,684	17
Ouray, CO	+21.6	23.8	4,015	5
Park, CO	−17.0	17.1	11,669	13
Phillips, CO	−64.2	8.7	2,300	8
Pitkin, CO	+45.2	5.2	10,924	12
Prowers, CO	−51.5	11.6	5,060	5
Pueblo, CO	−5.3	24.3	83,016	144
Rio Blanco, CO	−64.9	9.8	3,625	5
Rio Grande, CO	−23.8	18.5	6,049	15
Routt, CO	+28.5	26.9	15,884	19
Saguache, CO	+3.9	34.0	3,281	7
San Miguel, CO	+50.7	30.0	4,683	6
Sedgwick, CO	−57.5	7.5	1,331	5
Summit, CO	+38.3	8.9	17,006	20
Teller, CO	−36.4	5.6	15,921	14
Washington, CO	−76.2	4.3	2,765	6
Weld, CO	−21.5	21.9	175,221	184
Yuma, CO	−66.8	8.3	4,565	14
Fairfield, CT	+19.9	17.6	445,282	190
Hartford, CT	+22.9	23.7	421,938	167
Litchfield, CT	−8.2	20.7	104,392	48
Middlesex, CT	+13.1	13.5	95,827	33
New Haven, CT	+12.2	27.5	390,416	204
New London, CT	+12.8	17.2	135,048	54
Tolland, CT	+8.2	18.5	80,084	28
Windham, CT	−10.4	17.3	55,984	32
Kent, DE	+2.0	30.3	86,680	98
New Castle, DE	+33.1	23.3	271,568	312
Sussex, DE	−11.1	23.7	145,861	119
District of Columbia, DC	+86.6	5.9	315,261	144
Alachua, FL	+21.3	29.3	134,517	63
Baker, FL	−73.4	12.3	14,908	9
Bay, FL	−47.9	17.0	96,698	29
Bradford, FL	−57.5	28.2	13,866	13
Brevard, FL	−21.0	13.4	357,766	150
Broward, FL	+17.1	26.3	866,280	358
Calhoun, FL	−68.0	22.5	6,388	10
Charlotte, FL	−34.2	10.4	122,930	81
Citrus, FL	−46.2	7.7	97,632	32
Clay, FL	−39.6	24.6	125,637	48
Collier, FL	−33.3	12.6	214,987	67
Columbia, FL	−50.5	21.0	33,358	26
Desoto, FL	−43.2	28.7	12,413	15
Dixie, FL	−70.8	13.2	8,103	10
Duval, FL	−1.5	31.3	465,650	160
Escambia, FL	−19.8	28.8	161,008	80
Flagler, FL	−28.4	9.5	79,445	21
Franklin, FL	−44.2	28.6	6,701	8
Gadsden, FL	+30.9	39.8	21,698	25
Gilchrist, FL	−68.6	7.2	10,593	10
Glades, FL	−53.5	24.5	5,256	13
Gulf, FL	−54.5	24.1	8,654	10
Hamilton, FL	−39.3	30.8	5,691	8
Hardee, FL	−56.7	23.6	8,088	12
Hendry, FL	−38.6	29.0	13,349	10
Hernando, FL	−37.3	9.7	109,877	32
Highlands, FL	−41.0	17.3	51,609	25
Hillsborough, FL	−3.1	24.2	663,472	448
Holmes, FL	−80.6	4.5	9,075	8
Indian River, FL	−27.5	10.9	98,391	34
Jackson, FL	−46.4	25.2	21,966	14
Jefferson, FL	−18.7	20.3	8,440	16
Lafayette, FL	−76.4	15.5	3,737	5
Lake, FL	−24.9	17.6	225,046	92
Lee, FL	−28.6	13.8	389,901	99
Leon, FL	+22.0	29.8	154,917	132
Levy, FL	−50.5	14.5	24,239	13
Liberty, FL	−67.3	30.8	3,464	8
Madison, FL	−29.0	34.1	9,105	11
Manatee, FL	−23.7	15.2	227,160	71
Marion, FL	−31.9	17.7	212,609	106
Martin, FL	−31.3	10.8	97,660	28
Miami-Dade, FL	−11.5	34.0	1,085,945	721
Monroe, FL	−18.5	22.9	43,997	33
Nassau, FL	−47.3	22.6	65,088	18
Okaloosa, FL	−42.9	15.4	112,383	52
Okeechobee, FL	−54.1	10.9	15,986	11
Orange, FL	+13.8	21.1	599,086	257
Osceola, FL	−1.5	20.2	170,918	127
Palm Beach, FL	+0.8	23.9	739,348	781
Pasco, FL	−25.5	12.8	315,229	120
Pinellas, FL	−5.3	22.8	511,924	286
Polk, FL	−20.9	23.0	345,923	172
Putnam, FL	−48.1	27.6	36,054	37
St. Johns, FL	−31.6	11.1	195,621	53
St. Lucie, FL	−9.1	18.9	183,810	61
Santa Rosa, FL	−51.4	11.7	111,349	41
Sarasota, FL	−18.3	16.8	275,887	103
Seminole, FL	−3.6	14.8	250,452	82
Sumter, FL	−37.8	8.9	104,685	30
Suwannee, FL	−61.3	18.6	21,778	14
Taylor, FL	−60.0	35.5	9,945	14
Union, FL	−68.7	7.0	6,195	8
Volusia, FL	−21.9	19.0	307,823	143
Wakulla, FL	−44.7	7.9	19,687	12
Walton, FL	−58.2	9.9	49,257	21
Washington, FL	−65.8	15.0	12,510	12
Appling, GA	−62.5	36.4	8,321	9
Baker, GA	−19.9	24.0	1,473	5
Baldwin, GA	−2.2	35.8	18,733	14
Banks, GA	−78.3	3.8	10,494	13
Barrow, GA	−40.7	11.2	43,679	8
Bartow, GA	−51.3	13.9	57,213	17
Berrien, GA	−70.0	7.4	8,050	5
Bibb, GA	+22.5	40.6	68,830	31
Brooks, GA	−26.9	30.5	7,189	7
Bryan, GA	−36.5	16.0	24,517	10
Bulloch, GA	−29.1	28.5	32,499	16
Burke, GA	−9.4	24.6	11,021	16
Calhoun, GA	+12.3	32.5	2,053	5
Camden, GA	−35.9	12.4	26,224	11
Carroll, GA	−41.4	25.7	60,170	28
Catoosa, GA	−55.8	6.7	34,854	11
Charlton, GA	−56.4	25.1	4,614	6
Chatham, GA	+18.1	38.5	140,094	87
Chattooga, GA	−64.4	9.1	10,665	7
Cherokee, GA	−39.3	15.3	160,980	42
Clarke, GA	+38.7	16.1	52,346	24
Clay, GA	+7.5	32.0	1,434	5
Clayton, GA	+69.6	13.5	111,080	70
Clinch, GA	−51.6	14.2	2,903	5
Cobb, GA	+15.0	26.6	397,083	148
Coffee, GA	−45.2	15.7	15,683	6
Colquitt, GA	−50.3	35.0	16,565	19
Columbia, GA	−25.8	21.5	85,281	46
Cook, GA	−46.6	35.4	7,330	8
Coweta, GA	−34.1	22.6	85,315	24
Crawford, GA	−50.0	29.3	6,324	5
Crisp, GA	−26.0	39.3	8,092	5
Dade, GA	−67.0	9.1	8,147	7
Decatur, GA	−24.0	27.9	11,512	10
Dekalb, GA	+65.4	24.7	362,252	191
Dodge, GA	−50.0	21.6	8,330	16
Dooly, GA	−7.7	13.6	4,164	5
Dougherty, GA	+41.3	36.0	33,735	25
Douglas, GA	+31.7	32.1	70,236	25
Early, GA	−11.5	37.5	4,876	5
Effingham, GA	−49.3	12.0	36,087	20
Elbert, GA	−43.5	23.5	9,560	11
Emanuel, GA	−44.3	23.7	9,592	11
Fannin, GA	−65.0	5.0	16,039	12
Fayette, GA	−3.2	32.7	73,999	36
Floyd, GA	−42.2	25.2	44,493	19
Forsyth, GA	−33.5	16.9	136,790	29
Franklin, GA	−72.8	8.4	12,106	7
Fulton, GA	+45.4	38.7	529,389	450
Gilmer, GA	−62.9	4.9	18,389	13
Glynn, GA	−26.1	29.2	43,702	17
Gordon, GA	−63.7	13.3	27,477	12
Grady, GA	−38.4	40.4	10,675	13
Greene, GA	−29.1	25.6	12,729	5
Gwinnett, GA	+16.7	21.3	415,548	156
Habersham, GA	−65.2	6.0	23,178	6
Hall, GA	−44.1	22.0	101,338	31
Hancock, GA	+35.5	33.4	4,228	10
Haralson, GA	−74.7	6.9	16,304	12
Harris, GA	−46.3	12.1	22,259	12
Hart, GA	−55.0	12.1	14,274	7
Heard, GA	−72.3	8.4	6,194	6
Henry, GA	+29.8	31.3	128,235	37
Houston, GA	−11.3	23.5	80,997	17
Jeff Davis, GA	−68.5	25.6	5,859	9
Jefferson, GA	−1.2	26.3	7,439	8
Jenkins, GA	−30.6	37.0	3,396	5
Jones, GA	−38.2	25.0	16,038	10
Laurens, GA	−32.8	36.7	23,280	15
Lee, GA	−43.7	11.7	17,612	10
Liberty, GA	+17.5	25.7	22,900	12
Lincoln, GA	−45.0	23.6	4,910	7
Long, GA	−29.6	17.9	7,033	7
Lowndes, GA	−18.1	35.3	47,568	18
McDuffie, GA	−25.0	41.4	10,499	7
McIntosh, GA	−28.7	25.4	7,375	5
Macon, GA	+18.0	24.0	4,671	5
Madison, GA	−55.1	18.5	16,704	12
Marion, GA	−30.4	44.3	3,601	5
Meriwether, GA	−25.6	25.4	11,748	14
Mitchell, GA	−16.4	39.9	8,851	11
Monroe, GA	−46.8	27.1	17,643	14
Morgan, GA	−46.2	28.5	13,122	5
Murray, GA	−71.8	5.0	17,424	7
Muscogee, GA	+23.5	40.3	80,029	25
Newton, GA	+15.2	39.9	58,732	19
Oconee, GA	−36.3	11.8	27,044	8
Paulding, GA	−24.3	21.1	94,571	21
Peach, GA	−6.1	36.1	13,397	7
Pickens, GA	−66.1	9.0	20,803	12
Pierce, GA	−77.6	9.0	9,744	8
Pike, GA	−73.7	7.9	12,512	8
Polk, GA	−60.7	14.7	19,101	7
Putnam, GA	−42.4	25.2	12,832	5
Randolph, GA	+7.7	28.5	2,974	6
Richmond, GA	+36.3	38.6	83,129	68
Rockdale, GA	+47.8	20.9	44,876	16
Screven, GA	−25.3	28.7	6,906	7
Seminole, GA	−40.5	46.9	4,002	5
Spalding, GA	−16.8	38.4	32,863	17
Sumter, GA	+2.2	38.7	12,005	11
Talbot, GA	+12.0	27.3	3,371	7
Tattnall, GA	−53.6	13.7	8,482	8
Telfair, GA	−39.4	24.1	4,204	6
Terrell, GA	+4.1	42.3	4,328	6
Thomas, GA	−24.2	39.7	22,017	20
Tift, GA	−35.8	39.4	16,934	12
Toombs, GA	−50.9	17.8	10,882	5
Troup, GA	−24.5	36.5	31,149	13
Twiggs, GA	−14.7	34.9	4,444	5
Union, GA	−62.8	4.1	17,786	11
Walker, GA	−59.6	10.7	31,898	11
Walton, GA	−46.2	20.7	58,012	19
Ware, GA	−43.3	46.6	14,347	13
Washington, GA	−1.9	20.9	9,467	8
Wayne, GA	−59.9	29.0	13,519	12
White, GA	−68.8	8.9	16,745	10
Whitfield, GA	−44.7	23.6	39,608	23
Wilcox, GA	−49.3	34.2	3,340	6
Wilkes, GA	−16.9	33.8	5,083	7
Wilkinson, GA	−17.9	42.4	4,900	9
Worth, GA	−50.5	36.8	9,291	12
Hawaii, HI	+28.9	19.7	80,911	64
Honolulu, HI	+22.0	23.7	334,790	302
Kauai, HI	+19.9	21.2	29,478	31
Maui, HI	+26.5	19.3	61,526	70
Ada, ID	−10.6	34.1	259,875	197
Adams, ID	−55.9	15.1	2,614	5
Bannock, ID	−25.9	21.6	38,635	57
Bear Lake, ID	−78.7	8.2	3,254	15
Benewah, ID	−62.8	12.8	5,029	11
Bingham, ID	−61.3	22.4	20,066	27
Blaine, ID	+32.6	27.8	12,705	16
Boise, ID	−55.6	12.5	4,792	7
Bonner, ID	−47.2	25.2	29,002	30
Bonneville, ID	−47.0	16.8	54,511	51
Boundary, ID	−67.0	14.7	6,939	7
Canyon, ID	−47.8	15.2	98,424	51
Caribou, ID	−73.4	6.6	3,351	9
Cassia, ID	−70.8	11.7	9,318	24
Clearwater, ID	−64.2	7.7	4,324	14
Custer, ID	−54.6	28.4	2,584	9
Elmore, ID	−49.8	6.5	10,402	17
Franklin, ID	−79.9	4.6	6,982	18
Fremont, ID	−71.1	12.3	6,599	13
Gem, ID	−67.3	8.8	10,406	14
Gooding, ID	−61.9	8.7	5,776	6
Idaho, ID	−69.6	10.9	9,608	28
Jefferson, ID	−75.4	5.2	15,372	21

Showing the first 500 of 2853 rows.

The line rises across the reddest counties and levels off from about Trump +14 to about Harris +43; the two bluest bins fall back but hold few counties. The line follows the vote-weighted mean spread in each 10-point bin of county margin, whiskers are 95% intervals from a county bootstrap, and bins with fewer than 10 counties are left out. Each dot is one of 2,853 counties with at least 5 reporting units.Note: The number in the tooltip is the county's number of reporting units. Kings County, New York, and the two Pennsylvania counties named in the text are labeled.Note: The returns are full counts with no sampling error. The whiskers show how much a mean would move with a different set of counties, from a county bootstrap.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

Bluer counties are also larger. The correlation of county margin with the log of two-party votes (the natural logarithm, so that each doubling of a county's votes adds the same amount) is +0.64 (95% CI 0.59 to 0.69), and the correlation of spread with the log of votes is +0.40 (95% CI 0.33 to 0.46), in line with the concentration of Democratic voters in dense places [4]. Larger counties have more precincts, and with more precincts there is more room for very different ones. Figure 6 holds fixed the log of a county's votes and the log of its number of precincts, and the correlation of lean with spread falls from +0.41 to +0.21 (95% CI 0.12 to 0.30). It then adds state fixed effects, which compare counties within the same state and so absorb each state's own precinct conventions, and the correlation is +0.28 (95% CI 0.19 to 0.37). With both controls, a county 10 points more Democratic has a spread 0.9 (95% CI 0.6 to 1.3) points wider, against 1.1 (95% CI 0.8 to 1.4) points without them. The correlation is also positive when each county counts once, so its sign does not depend on the vote weights.

Figure 6

Spread rises with Democratic lean, with or without controls for county size and state

Correlation between the spread of precinct margins and the county two-party margin, Democratic positive, across 2,853 counties, 2024, with 95% intervals

Vote-weighted
Unweighted

Vote-weighted

Unweighted

Correlation with the spread of precinct margins

Show data

Data for Spread rises with Democratic lean, with or without controls for county size and state
Weighting	Controls	Correlation	Interval, low	Interval, high	Counties
Vote-weighted	No controls	+0.41	+0.33	+0.49	2,853
Vote-weighted	Size fixed	+0.21	+0.12	+0.30	2,853
Vote-weighted	Size and state fixed	+0.28	+0.19	+0.37	2,853
Unweighted	No controls	+0.54	+0.51	+0.56	2,853
Unweighted	Size fixed	+0.46	+0.43	+0.50	2,853
Unweighted	Size and state fixed	+0.47	+0.43	+0.51	2,853

Every interval is above zero, weighted or not, and controls for county size lower the correlation. Each dot is a correlation across counties. Vote-weighted counts each county in proportion to its two-party votes; unweighted counts each county once. Size fixed removes the linear effect of the log of two-party votes and the log of the number of precincts; state fixed compares counties within the same state. Intervals come from a county bootstrap.Note: Each interval resamples counties 2,000 times; a full count has no sampling error, so the interval says how much the correlation would move with a different set of counties.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

Two county characteristics measured outside the vote tell a similar story. Across 2,843 counties with demographic data, which leaves out the Connecticut counties, spread correlates +0.43 (95% CI 0.36 to 0.50) with the nonwhite share of residents (the share who are not non-Hispanic white, American Community Survey, 2024) and −0.36 (95% CI −0.41 to −0.30) with the rural share (the share living outside urban areas as the 2020 Census defines them). With size and state held fixed the two are +0.31 (95% CI 0.22 to 0.40) and −0.13 (95% CI −0.19 to −0.08), so the tie to the nonwhite share weakens less than the tie to the rural share does.

Spread falls with distance from an even split

Distance from an even split is the size of a county's margin, whichever party leads. One might expect the counties closest to a tie to be the most divided, because a near-tied county needs votes from precincts on both sides. Across all counties, the correlation of spread with distance from an even split is −0.29 (95% CI −0.39 to −0.20) when weighted by votes, and it stays negative with size and state held fixed, at −0.19 and −0.25. Counting each county once, it is −0.55 (95% CI −0.57 to −0.52). Spread is lower in counties that are farther from an even split.

Among the 1,725 counties with a margin of 50 points or less, the vote-weighted correlation is −0.04 (95% CI −0.18 to 0.08), an interval that includes zero, against −0.26 (95% CI −0.30 to −0.22) when each county counts once. The vote-weighted result depends on the cutoff, as Figure 7 shows: its interval excludes zero only from 60 points up, where the correlation is −0.15. The absence of a link is specific to the vote-weighted comparison of counties within 50 points or less, and it is not a finding about counties in general.

Figure 7

Spread falls with distance from an even split, except in vote-weighted comparisons of counties within 50 points

Correlation between the spread of precinct margins and the size of the county margin, among counties with a margin at or below each cutoff, 2024, with 95% intervals

Correlation of spread with distance from an even split

Show data

Data for Spread falls with distance from an even split, except in vote-weighted comparisons of counties within 50 points
Cutoff (points of margin)	Weighting	Counties	No controls	Size fixed	Size and state fixed
30	Vote-weighted	954	−0.02 [−0.13, +0.08]	+0.01 [−0.09, +0.10]	−0.07 [−0.18, +0.02]
30	Unweighted	954	−0.12 [−0.18, −0.06]	−0.13 [−0.19, −0.07]	−0.17 [−0.23, −0.11]
40	Vote-weighted	1,321	−0.07 [−0.18, +0.04]	−0.03 [−0.13, +0.05]	−0.08 [−0.20, 0.00]
40	Unweighted	1,321	−0.20 [−0.25, −0.14]	−0.21 [−0.26, −0.15]	−0.25 [−0.30, −0.20]
50	Vote-weighted	1,725	−0.04 [−0.18, +0.08]	+0.01 [−0.12, +0.12]	−0.04 [−0.19, +0.06]
50	Unweighted	1,725	−0.26 [−0.30, −0.22]	−0.26 [−0.30, −0.21]	−0.31 [−0.36, −0.27]
60	Vote-weighted	2,156	−0.15 [−0.26, −0.04]	−0.08 [−0.18, +0.03]	−0.11 [−0.24, −0.02]
60	Unweighted	2,156	−0.38 [−0.41, −0.34]	−0.36 [−0.39, −0.32]	−0.39 [−0.44, −0.35]
70	Vote-weighted	2,593	−0.24 [−0.34, −0.13]	−0.15 [−0.25, −0.04]	−0.20 [−0.33, −0.10]
70	Unweighted	2,593	−0.48 [−0.51, −0.45]	−0.44 [−0.47, −0.40]	−0.48 [−0.52, −0.45]
100	Vote-weighted	2,853	−0.29 [−0.39, −0.20]	−0.19 [−0.30, −0.09]	−0.25 [−0.37, −0.16]
100	Unweighted	2,853	−0.55 [−0.57, −0.52]	−0.49 [−0.52, −0.46]	−0.55 [−0.58, −0.52]

Compare the two lines: the unweighted line stays below zero at every cutoff, while the vote-weighted line has an interval that includes zero up to a cutoff of 50 points and excludes it from 60 points up. A county enters a cutoff when the size of its margin is at or below it, so the last point uses every county. Vote-weighted counts each county in proportion to its two-party votes; unweighted counts each county once. The data table adds the correlations with county size fixed and with size and state fixed.Note: The last cutoff, 100 points, includes every county, because no margin can be larger. The number in the tooltip is the number of counties in the row. Each interval resamples the counties of the row 2,000 times; a full count has no sampling error, so the interval says how much the correlation would move with a different set of counties.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

The negative correlation grows as the tails are added. Precinct margins cannot pass 100 points in either direction, so a county with a margin near the limit has little room for spread. Among the 260 counties with a margin beyond 70 points, however, the widest spread is 34.7 points, against a limit of 71.4 at a margin of 70 points, and the highest ratio of a county's spread to the limit at its own margin is 49%.The largest spread a county can have is the square root of 100 squared minus its margin squared, so it falls to zero as the margin approaches 100. Among the 260 counties beyond 70 points, the mean ratio of a county's spread to the limit at its own margin is 12%. The limit therefore does not explain the low spread in lopsided counties, and the data cannot say what does. One candidate is that the losing party has few precincts in those counties, and another is that margins are compressed near one side of the scale well short of the mathematical limit. This chapter reports the pattern and does not read the negative correlation as a finding about competitiveness. The link of spread to Democratic lean does not come from the tails. Within 50 points the correlation of spread with lean is +0.41 (95% CI 0.32 to 0.49), against +0.41 (95% CI 0.33 to 0.49) across all counties.

Near-tied counties differ widely in spread

A near-tied county has a margin under 5 points and at least 20 precincts. The group has 111 counties with 18.9 million two-party votes. Figure 8 shows their spreads. The median is 23.9 (95% CI 22.8 to 24.9) points, against 3.4 for the median county if voters were assigned to precincts at random, and the range runs from 10.1 to 58.1. A spread of 30 points or more occurs in 29 counties, 26.1% of the group.

Figure 8

Near-tied counties differ widely in how much their precinct margins vary

Share of the near-tied counties in each 5-point bin of spread of precinct margins, 2024; margin under 5 points and at least 20 precincts

Spread of precinct margins inside the county (points of margin)

Show data

Data for Near-tied counties differ widely in how much their precinct margins vary
n	share	x0	x1
11	9.91	10	15
21	18.92	15	20
34	30.63	20	25
16	14.41	25	30
11	9.91	30	35
11	9.91	35	40
5	4.50	40	45
1	0.901	45	50
0	0	50	55
1	0.901	55	60

Most near-tied counties (64.0% of the 111) have a spread between 15 and 30 points, and a tail runs past 30, the spread of a county split evenly between Democratic +30 and Republican +30 precincts. The median is 23.9 points. Bucks County, with a spread of 14.8, is in the lowest bin, and Centre County (36.3) is past the dashed line. A county whose voters were assigned to precincts at random would show a spread of about 3.4 points in the median case.Note: The number in the tooltip is the number of counties in the bin.Note: Full counts of votes: there is no sampling, so the shares carry no confidence interval.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

The median near-tied spread is above the 15.6 of all counties, but near-tied counties are larger, with a median of 78,269 two-party votes against 13,707 for the median county with a spread, and larger counties have wider spreads. The comparison of the two medians therefore mixes closeness with size.

Their votes are spread across the range of precinct margins. Near-tied counties cast 35.6% of their votes in precincts decided by under 10 points, against 18.5% in all precincts, and 38.2% in precincts decided by 20 points or more, against 65.0%. The landslide share divides into 17.5% in Democratic precincts and 20.7% in Republican ones, and precincts decided by 40 points or more held 10.5%. A county built from two landslides would have at least 20% of its votes in Democratic landslide precincts and at least 20% in Republican ones, and 39 of the 111 counties, 35.1%, meet that test. The near-tied counties as a group hold more of their votes in close precincts than the country does and fewer in landslide precincts, and they differ from one another in how the votes divide.

The counties in Figure 1 fall at different points of this range. Bucks County, with a spread of 14.8, is in the lowest bin of Figure 8, and Centre County, at 36.3, is past the 30-point line.

What these data cannot show

The chapter describes one election. It says nothing about whether precinct margins are spreading or narrowing, and it makes no comparison with earlier years. Chapter 11.2, Sorting by geographic scale, 2016 to 2024 (planned) compares precincts across elections, and Chapter 10.1, Nationalization of the county vote measures how closely county results move together from one election to the next.

A precinct is a reporting unit and not a neighborhood. The file has no boundaries, so the data cannot say which precincts are adjacent, and a county with a wide spread need not have opposed precincts next to each other. Units also differ in kind. Units labelled absentee, provisional, early voting, mail, failsafe, military or overseas, and units labelled as statistical adjustments, hold 1.5% of votes across 870 units, and they are buckets of ballots and not places. Figure 9 repeats the main statistics without them and without the counties where they hold 20% of votes or more, which leaves 2,721 counties with a spread and 106 near-tied counties, against 111 in the full sample.

Figure 9

Dropping ballot-bucket units moves each correlation by at most 0.02 and each vote share by at most 0.3 points

Main statistics in the full sample and in a side check that drops ballot-bucket units, 2024 two-party presidential returns

Statistic	Full sample	Without ballot-bucket units
Counties with a spread	2,853	2,721
Near-tied counties	111	106
Median spread of near-tied counties (points)	23.9	24.0
Correlation of spread with Democratic lean	+0.41	+0.40
Same, with county size and state fixed	+0.28	+0.27
Correlation of spread with distance from an even split, counties within 50 points	−0.04	−0.04
Votes in precincts decided by 20 points or more	65.0%	64.8%
Votes in precincts decided by under 10 points	18.5%	18.6%

Show data

Data for Dropping ballot-bucket units moves each correlation by at most 0.02 and each vote share by at most 0.3 points
full	side	statistic
2,853	2,721	Counties with a spread
111	106	Near-tied counties
23.9	24.0	Median spread of near-tied counties (points)
+0.41	+0.40	Correlation of spread with Democratic lean
+0.28	+0.27	Same, with county size and state fixed
−0.04	−0.04	Correlation of spread with distance from an even split, counties within 50 points
65.0%	64.8%	Votes in precincts decided by 20 points or more
18.5%	18.6%	Votes in precincts decided by under 10 points

Read each row across the two columns. The side check drops units labelled as ballot buckets and the counties where they hold 20% of votes or more. Correlations are weighted by two-party votes.Note: The changes in the title are the largest differences before rounding, rounded up to the decimals printed.Source: MIT Election Data and Science Lab, 2024 precinct-level presidential returns; county totals summed from the precincts; full counts of votes, not survey estimates, so nothing is weighted except by two-party votes where stated

A precinct margin describes a place and not the voters in it. A precinct that gave one party 60% or more still holds voters of the other. Ecological inference, which moves from places to individuals, can fail, because a pattern across places can differ from the pattern among individuals [5]. The results also depend on the units. Shares and spreads change with the size and shape of the areas, a problem known as the modifiable areal unit problem [6], and this chapter uses the units that election offices happened to report. The 64 Louisiana parishes, the 6 other counties and Alaska that are left out are not a random set. The cutoffs for near-tied counties and for a wide spread are choices, and so are the vote weights and the 50-point cutoff behind the correlation of spread with distance from an even split, which is near zero only in the vote-weighted comparisons within 50 points or less. A single spread cannot tell a county with two humps from a county with a long tail. The random-assignment benchmark holds each unit's size and its county's margin fixed and does not model why voters differ across precincts. Nothing here identifies why spread is wider in some counties. It goes with lean, size, nonwhite share and place type, and the controls are few and linear.

Data sources

Precinct presidential returnsMIT Election Data and Science Lab, 2024 presidential returns by precinct (Harvard Dataverse), two-party votes per reporting unitYears: 2016, 2020, 2024 · Weights: none (full counts); statistics weighted by two-party votes · License: See the provider's terms of use.
County election returnsCertified 2024 county totals from the repaired county file, used to check each county's precinct rowsYears: President 1868-2024 (40 elections); House and Senate 1990-2024 · Weights: none (full counts) · License: The Heideman file is CC BY 4.0 (doi:10.5281/zenodo.22976553), but the state records behind it keep their own terms, and the sites of Florida, Indiana, Colorado and Wisconsin restrict reuse.
County demographics panelCounty panel: ACS 2024 nonwhite share, rural share and Rural-Urban Continuum Code 2023Years: 2000-2024 (election years, interpolated between census and survey anchors) · Weights: none (county values; counties weighted by two-party votes in the correlations) · License: US government source data, public domain. The interpolated panel is derived work.

All data sources

Methods

Sample: the 2024 presidential returns in the MIT Election Data and Science Lab precinct file, two-party votes per reporting unit (Democratic: Harris; Republican: Trump; write-ins and other candidates left out) from b3.load.precinct_two_party, which takes a precinct's TOTAL row where the state reports one and otherwise adds the rows for each voting mode. A 'precinct' is the unit the state or county election office reports, so it varies in size and kind by state. Rules applied in order: units with no two-party vote or a negative count are dropped (15,602 units, three of them Michigan statistical adjustments); Alaska is dropped (523 units; its 2024 county codes are election districts); 4 units with no county code are dropped; 123 Kansas City units are moved from code 36000 to 2938000, the code of the county file; a county is kept only when its precinct rows add to between 95% and 105% of its certified two-party vote in the repaired county file, which removes 70 counties (64 Louisiana parishes, where the precinct rows hold about 52% of the certified vote, and 6 others, four of them with rows that appear twice, in Bergen and Gloucester, NJ, and Hendricks and St. Joseph, IN). The analysis sample is 175,345 units, 148,953,406 two-party votes and 3,043 counties in 48 states and the District of Columbia; the 70 excluded counties hold 2.0% of the certified votes and Alaska a further 0.2%. Estimands: a precinct or county margin is 100 times the Democratic minus the Republican votes over their sum; the spread of a county is the standard deviation of its precinct margins about the county margin, weighted by each precinct's two-party votes (population divisor), and needs at least 5 reporting units; the spread expected from chance is 200 times the square root of p(1 - p) times the number of units over the votes, with p the Democratic two-party share, which is the spread if each voter were assigned to a unit at random. Near-tied counties have a margin below 5 points in absolute value and at least 20 precincts. Shares of votes in bands use exact tests on vote counts (a margin of 20 points or more is 5 |D - R| >= D + R). Random-assignment benchmark for the shares in bands: 200 simulated assignments (seed 12345) in which each precinct keeps its two-party votes and its county's Democratic share and its Democratic votes are a binomial draw; the figure shows the mean of the runs, and the largest range of a share across runs is 0.4 points. County correlations are weighted by two-party votes unless the text says each county is counted once; holding county size fixed removes the linear effect of the log of two-party votes and the log of the number of units, and holding state fixed removes each state's weighted mean from every variable (equal to a regression with state dummies). The slope of spread on margin is the coefficient of a weighted regression with the same controls, per 10 points of margin. Counties within 50 points have a margin of 50 points or less in absolute value; the correlation of spread with distance from an even split is also computed at cutoffs of 30, 40, 60 and 70 points and across all counties, weighted and counting each county once, each with no controls, with county size fixed and with size and state fixed; the correlation of spread with lean is computed weighted and counting each county once, with the same three sets of controls. Binned means use 10-point bins of county margin and need 10 counties. Place types use the USDA Rural-Urban Continuum Code, 2023 edition (1; 2 and 3; 4 to 7; 8 and 9); nonwhite share is 100 minus the non-Hispanic white share (ACS 2024) and rural share is the share of residents outside 2020 census urban areas, both from the county demographics panel, which has no values for the Connecticut counties, Kansas City, Missouri, and Oglala Lakota County, South Dakota. Weights: none from a survey, because the returns are full counts; statistics are weighted by two-party votes as stated. Intervals: the returns have no sampling error, so shares, counts and county spreads carry no interval; a correlation, a slope, a mean or a median across a set of counties has a 95% percentile interval from a county bootstrap (2,000 replicates, seed 12345); the interval for the difference between two bins of county margin resamples the counties of each bin separately, and a state mean resamples the counties of the state. Side check: dropping the units labelled as absentee, provisional, early voting, mail, failsafe, military, overseas and statistical-adjustment units (870 units, found by these words in the label: absen(tee), abs, provisional, prov, early voting, early vote, earlyvote, early on its own, EV, mail, statistical adjustments, military, overseas, failsafe) and the counties where they hold 20% of the votes or more leaves 2,721 counties with a spread; the results are printed beside the full-sample values in the section on what the data cannot show. Suppression: no county with fewer than 5 reporting units enters a spread statistic; no other cell is suppressed. Everything is a 2024 cross-section.

References

Bishop, Bill; Cushing, Robert G. (2008). The Big Sort: Why the Clustering of Like-Minded America Is Tearing Us Apart. Houghton Mifflin.
Abrams, Samuel J.; Fiorina, Morris P. (2012). "The Big Sort" That Wasn't: A Skeptical Reexamination. PS: Political Science and Politics 45(2): 203-210. doi:10.1017/S1049096512000017
Brown, Jacob R.; Enos, Ryan D. (2021). The Measurement of Partisan Sorting for 180 Million Voters. Nature Human Behaviour 5: 998-1008. doi:10.1038/s41562-021-01066-z
Rodden, Jonathan A. (2019). Why Cities Lose: The Deep Roots of the Urban-Rural Political Divide. Basic Books.
Robinson, W. S. (1950). Ecological Correlations and the Behavior of Individuals. American Sociological Review 15(3): 351-357.
Openshaw, Stan (1984). The Modifiable Areal Unit Problem. Concepts and Techniques in Modern Geography 38, Norwich: Geo Books.

How to cite

@misc{singh2026precinctswithincounties,
  author = {Singh, Vishal},
  title = {Precincts within counties},
  year = {2026},
  url = {https://vishalsingh.org/politics/precincts-within-counties},
  note = {In The Public Mind. Published 30 September 2026.}
}

Precincts are more lopsided than the counties that contain them

Spread is widest in Mississippi and Alabama

Spread falls from metro to less urban counties

Spread rises with Democratic lean, partly through county size

Spread falls with distance from an even split

Near-tied counties differ widely in spread

What these data cannot show

Data sources

Methods

References

How to cite

Reading paths