The Shifting Map

Every county gets an arrow.

Each arrow starts at a county’s center. Its length is how far the county’s margin moved since the previous presidential election, in percentage points of all votes cast.

Toward the Republican.

Maverick County, Texas, on the Rio Grande, voted Biden +9.5 in 2020 and Trump +18.5 in 2024: a move of 28 points toward the Republican, so its arrow is 28 points long and leans right.

Toward the Democrat.

Henry County, Georgia, south of Atlanta, went from Biden +20.4 to Harris +29.7: 9.2 points toward the Democrat, one of the largest moves that way in 2024.

A margin can move without a winner changing.

The margin is the Democratic share of all votes minus the Republican share. A county that goes from 55–43 to 50–48 moved 10 points toward the Republican even though the Democrat won it both times. Length is points, not voters: a county of 900 and one of 2 million draw the same length for the same move. Thicker arrows mark bigger counties.

All twenty, side by side.

The same arrows, drawn larger so they read at this size. The number under each map is the national swing.

The six biggest swings came between 1952 and 1976.

Those six times, the national margin moved about 15 to 25 points from one election to the next.

Since 1996, shorter arrows.

No national swing since then has topped 10 points. 2016 stands out for direction, not size: small counties and big ones moved opposite ways. The last map, 2026, sets a midterm’s top races against a presidential year, a different kind of comparison.

Landslides, then knife-edges.

The popular-vote margin went from Johnson +22.6 in 1964 to Nixon +23.2 eight years later. Seven of these twenty elections were decided by less than 2.5 points, four of them since 2000.

The big swings came early.

From one election to the next, the national margin moved about 15 to 25 points six times, all between 1952 and 1976. Since 1996 it has never moved more than 10.

So did the big county moves.

Weighted by votes, the average county moved 17 to 28 points in those six elections. Since 1996 the figure has ranged from about 4 to 11.

48.7 million votes in 1948; 158.4 million in 2020.

The bars count the votes for president, or at a midterm for the House. The line is turnout, the share of eligible citizens who voted: from 36.7% at the 2014 midterm to 66.4% in 2020. The U.S. Elections Project’s early estimate for 2026 is 48.8%; the highest midterm turnout since 1948 was 2018’s 50.1%.

81.3 million votes, the most ever.

Each party’s nominee in every presidential election. Biden’s 81.3 million in 2020 is the most for any candidate; Trump’s 77.3 million in 2024 is the most for a Republican. The widest gap: Nixon’s 18.0 million over McGovern in 1972. Gray is everyone else, led by Perot’s 19.7 million in 1992 and Wallace’s 9.9 million in 1968.

The House vote, every two years.

Each party’s total across all 435 districts. By the Clerk’s count, the party with more votes won more seats every time but 1996 and 2012. In 2022 Republicans led, 54.2 million to 51.3 million. In 2026, with about 91% of the expected vote counted, Democrats had 61.1 million, more than any party at a midterm, and Republicans 47.6 million: a 12.2-point margin, the widest at a midterm since 1974.

The Senate vote depends on the map.

Only a third of the Senate is elected each time, so the national total depends on which states vote. In 2018 Democratic Senate candidates won 52.2 million votes to Republicans’ 34.8 million, yet Republicans gained two seats; California’s race that year, between two Democrats, put 11.1 million votes in the Democratic column and none in the Republican one. In 2026 so far: Democrats 39.8 million, Republicans 33.6 million.

103.5 million transactions, 26.2 million new.

Between the 2022 and 2024 elections, election offices processed 103.5 million voter registration transactions, by the Election Assistance Commission’s survey of election offices: 26.2 million new registrations, 59.1 million updates to existing records, 10.4 million duplicates and 2.8 million rejected as invalid. The rolls held 234.5 million registrations. The 2024 survey counts transactions, a broader unit than the forms it counted before.

A new registration isn’t always a new voter.

Texas reported 2.9 million new registrations, 56.5% of its transactions, close to its shares in 2020 and 2022 (54.8% and 54.6%). Under the survey’s rules, and Texas law, a voter who moves to another Texas county counts as a new registration there, and Texas reported no duplicates as a separate category. Per 100 registered voters, Texas had 15.6 new registrations, sixth among the states; the national figure was 11.2.

21.3 million removed.

In the same two years states removed 21.3 million registrations: 6.7 million after a mailed notice went unanswered and the voter missed two federal elections, as federal law allows; 6.5 million who had moved; 4.5 million who had died; 1.2 million duplicates. New applications are checked against motor-vehicle or Social Security records; in 2024 Social Security answered 16.2 million checks, 18.6% with no match, which by itself does not make anyone ineligible.

65 million checked, 28,635 flagged.

Since 2025 states can run their rolls through the federal SAVE database in bulk. In September 2026 the government told the Supreme Court it had checked more than 65 million voters and flagged 28,635 as potential noncitizens, under 0.05%. A flag starts a review: of 2,724 Texas voters flagged, 578 later showed they were citizens and 117 cases were referred for investigation. Georgia’s 2024 audit found 20 noncitizens among 8.2 million registrations.

Forty elections, the whole government.

After every November election from 1948 to 2026: which party held the presidency, and how the House, the Senate and the governorships divided. The dashed line in each band marks a majority.

Midterms swing back.

The midterm years are lit. At 18 of the 20 midterms since 1950, the president’s party lost House seats; it gained only in 1998 and 2002. It lost governorships at 16 of them.

Full control used to last.

Lit: every election that left one party holding the presidency, the House and the Senate. Twelve presidential elections did that. Through 1976, the winning party kept both chambers at the next midterm 4 of 5 times; since 1992 it has lost at least one 6 of 7 times, counting 2026.

The president’s party pays at midterms.

House seats won or lost by the president’s party at every midterm. From 1950 to 2022 the typical midterm cost it 18 seats. It lost 40 or more six times: 1958 (−48), 1966 (−47), 1974 (−48), 1994 (−54), 2010 (−64) and 2018 (−42). In 2026, with 25 races still uncalled, it has lost at least 37.

The Senate swings less.

Only a third of the Senate is elected each time. The president’s party lost Senate seats at 13 of the 20 midterms, gained at 5 and broke even at 2. The largest loss: 13 seats, for the Republicans in 1958.

A clean sweep, undone in two years.

Eisenhower won by R+10.5 and Republicans took the House and the Senate: they gained 22 House seats and 2 in the Senate. In 1954 they lost 18 House seats and 2 in the Senate, enough to lose both chambers. Democrats then held the House for 40 years.

A landslide, then 47 seats back.

Johnson won by D+22.6 and Democrats gained 36 House seats and 2 in the Senate, reaching 295 of 435 House seats, more than two-thirds. In 1966 they lost 47 House seats and 3 in the Senate but kept both chambers.

Forty-nine states, then Watergate.

Nixon won by R+23.2 and carried 49 states, while Republicans gained 12 House seats but lost 2 in the Senate. Nixon resigned in August 1974; that November, with Ford in the White House, Republicans lost 48 House seats and 4 in the Senate.

The Senate flips; the House swings back.

Reagan won by R+9.7. Republicans gained 34 House seats and 12 in the Senate, winning the Senate for the first time since 1952. In 1982 they lost 26 House seats and broke even in the Senate.

Full control, then 1994.

Clinton won a three-way race by D+5.6, and Democrats kept both chambers though they lost 9 House seats and broke even in the Senate. In 1994 they lost 54 House seats and 8 in the Senate, and Republicans won both chambers, the House for the first time since 1952.

A wave, then a bigger one.

Obama won by D+7.3; Democrats gained 24 House seats and 8 in the Senate. In 2010 they lost 64 House seats, the most at any midterm on this page, and 6 in the Senate; Republicans took the House.

Full control, then the House flips.

Trump won the Electoral College while losing the popular vote by 2.1 points; Republicans kept both chambers though they lost 6 House seats and 2 in the Senate. In 2018 they lost 42 House seats and the House majority but gained 2 in the Senate.

And now: 2024, then 2026.

Trump won by R+1.5, and Republicans kept the House and won the Senate. In 2026, with 25 House races still uncalled on Thursday morning, Democrats have gained at least 37 House seats, 7 Senate seats and 5 governorships, and won both chambers.

The swing back is often bigger.

In six of these eight pairs, the midterm cost the president’s party more House seats than its presidential year had won it; three times, it had won none. The Senate, a third of it elected each time, moved by single digits in all but one of these sixteen elections: 1980 (+12 for the Republicans).

The South traded places.

In 1948 the counties of the 11 former Confederate states, taken together, voted 19 points more Democratic than the nation. By 2000 they voted 11 points more Republican, and in every election since, between 10 and 14 points.

The Northeast and the West went the other way.

The Northeast leaned 5 points Republican in 1948 and 11 points Democratic in 2024. The West moved from about even to 12 points Democratic.

The Midwest stayed near the middle.

In twenty elections it never leaned more than 7 points either way. Its most Republican lean came in 2016 and 2020, at about 6.5 points.

Cities and the countryside once voted almost alike.

Sort counties by ballots cast per square mile, then cut the electorate into five equal fifths. In 1948 the densest fifth voted D+12.8 and the sparsest D+10.0: about 3 points apart.

Now they are 63 points apart.

In 2024 the densest fifth voted D+30.4 and the sparsest R+32.4. The gap widened in 10 of the 12 elections after 1976 and peaked at 68 points in 2020.

Fewer places sit near the middle.

The share of voters living in a county that voted at least 20 points away from the national margin: 24% in 1976, 58% in 2016, 56% in 2024.

And more live in lopsided places.

At 30 points or more from the national margin: 10% of voters in 1976, 41% in 2016 and 38% in 2024.

Three million-vote counties in 1948; nine in 2024.

Each circle is a county, sized by its votes for president. Maricopa County, Arizona, cast 79,000 votes in 1948 and 2.1 million in 2024. But 551 counties cast fewer votes in 2024 than in 1948, among them Brooklyn, the Bronx and the city of St. Louis.

Half the votes, 159 counties.

In 2024 the 159 counties that cast the most votes, out of 3,112, cast half of all the votes for president; the other 2,953 cast the other half. In 1948 it took 99. The vote has spread out: the 100 biggest counties cast 50% of it in 1948 and 40% in 2024.

Big counties lean Democratic, small ones Republican.

Each county in 2024, by votes cast (up) and margin (across). Harris won seven of the nine counties that cast a million votes or more; Trump won 2,309 of the 2,514 that cast fewer than 50,000. At the extremes: Roberts County, Texas, R+92.1, and the District of Columbia, D+83.8.

Counties won are not votes won.

Each bar is one party’s net votes from the counties its candidate carried. In 2024 Harris carried 452 counties, netting 19.1 million votes; Trump carried 2,660, netting 21.3 million. The difference is close to his national margin of 2.3 million; Alaska, which does not report by county, makes up most of the rest.

One county, 1.2 million votes.

Los Angeles County gave Harris a net 1.23 million votes in 2024, more than Trump’s ten largest county margins combined, 999,000. His largest: Montgomery County, Texas, 140,000. Since 1948 the largest county margin for a Republican is Orange County, California’s 429,000 for Reagan in 1984; for a Democrat, Los Angeles County’s 1.88 million for Biden in 2020.

The biggest moves, counted in votes.

In net votes rather than points, the largest one-election moves since 1948 came from the biggest counties: of the eight largest each way, all but one were in Los Angeles or Cook County. The largest toward a Republican: Los Angeles, 656,000 votes toward Trump in 2024. Toward a Democrat: Cook, 634,000 toward Kennedy in 1960.

Black voters moved first, and furthest.

About 7 in 10 Black voters backed Kennedy in 1960; in 1964 nearly all backed Johnson, and every exit poll since 1976 has shown Black voters backing the Democrat by roughly 65 to 90 points. Hispanic voters backed Democrats by about 30 to 55 points in most exit polls from 1988 to 2020, then came within about 5 points of even in 2024: 51–46 in the exit poll and 51–48 among Pew’s validated voters.

College graduates and everyone else traded places.

In the 1950s white college graduates were among the most Republican voters in the ANES surveys, about 72–28 in 1952. By 2020 they leaned Democratic, while white voters without a degree moved the other way: 32–67 in the 2020 exit poll and 32–66 in 2024.

Young voters pulled away, then drifted back.

From 1976 to 2000, voters under 30 and the oldest voters were usually within about 7 points of each other in the exit polls. In 2008, voters 18–29 backed Obama 66–32 while those 65 and older backed McCain 53–45. In 2024 the gap narrowed: 54–43 and 49–50.

The gender gap opened in 1980.

In the 1950s women leaned slightly more Republican than men. Since 1980, women have voted more Democratic than men in every exit poll, by about 5 to 25 points of margin.

Higher-income voters moved toward the Democrats.

In 1984, voters earning over $50,000 backed Reagan 69–31 in the exit poll. In 2024, voters earning $100,000 or more backed Harris 51–47, while those under $50,000 backed Trump 50–48. Brackets are in each year’s dollars, so the groups aren’t identical across years; the ANES income groups show the same narrowing.

In 1960, faith split the vote.

Catholic voters, closely divided in 1956, voted roughly 4 to 1 for Kennedy, the first Catholic nominee, in both the ANES and Gallup surveys. White evangelical and born-again Christians voted Republican by about 57 to 66 points in the 2004, 2012, 2016 and 2024 exit polls.

Union households drifted toward the middle.

Union households voted 83–17 for Johnson in 1964 in the ANES survey. In the 2024 exit poll, Harris won 53% of union households and 47% of non-union households.

White Southerners left the Democrats.

In the ANES surveys, white voters in the 11 former Confederate states gave the Democrat about half the two-party vote in the 1950s and a quarter to a third since 2000, while white voters elsewhere stayed near the middle. It is the same shift the county map shows, seen voter by voter.

If only one group had voted.

Each map colors a state by which candidate led among one group of its voters in 2024, in AP VoteCast’s survey of about 120,000 voters. The bar counts the electoral votes each would have won if only that group had voted. D.C. was not surveyed; dark states had too few of a group to report.

One electorate, eight different maps.

If only women had voted, Harris would have won 316 electoral votes to 219; if only men, Trump 426 to 105. Voters of color: Harris 497. White voters: Trump 358. White college graduates: Harris 314; white voters without degrees: Trump 516.

And every group keeps moving.

Each line is that group’s national margin in the exit polls since 1976, or since the polls first reported it. Voters of color, who favored Democrats by more than 60 points in most elections through 2012, did so by 33 in 2024. White college graduates went from R+4 in 2008 to D+8; white voters without degrees, from R+18 to R+35. Women have voted more Democratic than men in every election since 1980.

The widest gap: young women and young men.

In 2024 women under 30 voted D+23 and men under 30 R+1, a 24-point gap, wider than at any older age. Women 45 to 64 split almost evenly (R+1); men that age voted R+20.

Children and marriage split both sexes.

Fathers of children under 18 voted R+21 and men without children at home R+10; mothers D+4 and women without children D+11. Married women voted R+5 and unmarried women D+23; married men R+22, unmarried men Even. Unmarried women went from D+36 in 2012 to D+23.

In every race, men lean further right.

White men voted R+22, white women R+7; Black men D+56, Black women D+85; Latino men R+10, Latino women D+19. Latino men went from D+31 in 2008 to R+10; Black men went from D+90 in 2008 to D+56. No exit poll or AP VoteCast table splits these groups again by parenthood, so figures for groups like Black fathers or white women without children are not published.