D+12–14
Widest midterm margin since 1974. Final polls said about D+9.
By the politic.design data desk · Published 5 Nov 2026 · Updated 6 Nov 2026
Midterms 2026 · Data report
The national margin is the balance point of every group's vote. In two years it moved 15 points. Five data stories on who moved it, why, where it won seats, and who filled the stands.
Preliminary. Seats and margins from network projections on 4–5 Nov; exit-poll figures are NBC News/SSRS estimates. Sources at the end.
Five questions about this week's result, each answered with one chart and one sourced figure. Counts are preliminary and update until states certify in December.
| # | Story | The question | Form |
|---|---|---|---|
| 1 | Weight Transfer | Who moved furthest, and who actually moved the national balance? | 2.5D |
| 2 | The Receipt | What were voters billing the president for, and what did the news cover? | 2D |
| 3 | The Blue Wave | How did the swing become 40+ seats, and did the new maps backfire? | 3D |
| 4 | The Borderlands | How did South Texas go from Trump's showcase to the Democrats' biggest gain? | 3D |
| 5 | Who Won the Convention | Was it persuasion or turnout, and what did $11.2bn buy? | 3D |
Young and Hispanic voters moved furthest. The national balance moved mostly because white voters, men and people in their 30s and 40s moved, because there are more of them.
The national margin is a balance point. Put every group on a ruler at its margin, give each a weight equal to its share of voters, and the point where the ruler balances is the country's margin. That is the whole chart. Between 2024 and this week every weight slid left, and the balance point moved about 15 points, from R+2 to D+13.
Seen that way, the size of a swing and its effect come apart. Under-30s moved 33 points but carry a ninth of the weight, so they pushed the balance about 3.6 points. White voters moved 12 points, but at 73% of the electorate that pushed it nearly 9. Men moved the balance more than women (8.5 points against 6.9), and voters aged 30 to 44 more than any other age group.
In shortA see-saw with its pivot at a tied vote. Each group is a weight: bigger groups weigh more, and groups sitting far from the middle press harder, the way a child at the end of a see-saw does. The beam tips toward the party the country chose; the amber diamond is where it would balance. Hover or tap a weight.
In shortThe Democrats' gain since 2024, counted as filled bubbles. Each bubble is half a point of the national margin, and its colour is the group that filled it. A long row is a group that moved the country, not just one that moved a lot.
Voters handed the president an itemised bill: tariffs, Iran, prices, enforcement. The news covered the items in a different order from the one voters ranked them in.
Exit polls read like a receipt. Two-thirds of voters said tariffs hurt the economy, and they voted Democratic 81–18. Nearly two-thirds called military action against Iran the wrong decision (83–15). Four in ten said their family was worse off than two years ago (75–23). Only 37% approved of the president.
The second half of the story is Brantopic's. We ran its news collector over 90 days of US coverage (GDELT, 5 August to 4 November) for each item on the bill, and its anomaly gate, the same one that tells a client when coverage of their brand is unusual, flagged the days each issue broke out. That puts newsroom attention beside voter priority. The test: did coverage track what voters say decided their vote, or what campaigns paid to put on air?
The kerb chart covers the persuadable middle: the groups still within six points of even. In a 13-point national win, the close groups were men, people over 45, Catholics, voters with some college, and voters whose finances hadn't changed. These are the groups 2028 will be fought over.
In shortWhat each issue cost or earned Republicans, in votes. Voters named their top issue; each line is that issue's share of voters times how lopsidedly they voted. Issues that cost Republicans are charges, issues that won them votes are credits, and the total is the Democrats' winning margin.
In shortEach dot is a group of voters; the further left, the more Democratic. Groups inside the shaded strip were within six points of a tie, so they're still up for grabs in 2028. Bigger dots are bigger groups.
In shortOne cover per issue; hover or tap a cover to open its full chart. The line behind each cover is how much of the US news mentioned it each day from August to election day; amber dots are days Brantopic flagged as unusually busy. The line under the title is how much voters said the issue mattered.
Brantopic flagged these spikes the day they happened. Get the same alert when coverage of your name jumps past its normal range. First alert free.
Democrats took every Republican seat Trump won by less than ten points, and a few the tide should never have reached. Most of those few were in states that had just redrawn their maps to add Republican seats.
Line up the 220 Republican-held districts by how much Trump won them by in 2024, on the lines used this year, and the result reads like a tide chart. The national swing since 2024 is about 15 points, which on a uniform swing puts 54 Republican seats under water. Democrats have taken all 28 that Trump won by less than ten points (or that Harris carried), 7 of the 26 he won by 10 to 14, and none of the safer seats except eight.
The seats that beat the tide are the story. Of the eight flips in districts Trump won by 15 or more, six are in North Carolina and Texas, both of which redrew their maps in 2025 to add Republican seats: North Carolina's 6th, 9th, 13th and 14th, and Texas's 15th and 23rd. Spreading Republican voters thinly across more seats leaves each with a smaller cushion, so a big enough wave goes over all of them at once. Across the four states that redrew for the GOP (Texas, North Carolina, Ohio and Missouri), Democrats made 11 of their 43 gains.
In shortEach tower is a Republican-held House seat. The taller it is, the more Trump won it by in 2024. The sea is the national swing toward Democrats: anything above the water stays Republican red, towers under water are seats that swing could flip, and blue towers are the ones that actually flipped. Amber roofs are states that redrew their maps for the GOP.
Texas swung 26 points in two years and every one of its 254 counties moved the same way. The border moved furthest.
Trump won Texas by 13.7 points in 2024. On Tuesday James Talarico won its Senate seat by 12.1, by Decision Desk HQ's count, the first Democratic Senate win there since 1988. All 254 counties moved towards the Democrats. In Starr County the margin went from Trump +16 to Talarico +56, a 72-point swing. The four Rio Grande Valley counties went from R+4 to D+44 between them, and cast 41% more ballots than in the 2022 midterm.
It wasn't only the border. Twenty-eight counties flipped from Trump to Talarico, including Tarrant, Collin and Denton in the Dallas–Fort Worth suburbs and Brazoria and Jefferson on the Gulf coast. The smallest swings, six to eight points, came in the Panhandle. The open question is durability: the same valley swung hard the other way in 2024. With certified files we'll extend the relief back to 2016, so readers see a cycle rather than a single turn.
In shortEach county is lifted by how far it moved toward Democrats between the 2024 presidential vote and this week's Senate vote. Taller means a bigger swing; colour shows who won it this time. Every county rose. Drag to turn; hover or tap a county.
About six points of the Democrats' margin came from who showed up. The rest came from voters who changed sides or were voting for the first time.
Turnout was 48.8% of eligible voters, short only of 2018 in more than a century. Maine (65.8%), Minnesota and Michigan led; West Virginia (33.3%) trailed. The exit poll asked people how they voted in 2024. Among this year's voters, Harris voters outnumbered Trump voters 49 to 44, even though Trump won the 2024 popular vote. That gap is the turnout story.
We split the margin in two. With the 2024 mix of Harris and Trump voters and this year's loyalty rates, Democrats would have won by about 6. With the actual 2026 mix they won by about 12 in this sample. So roughly six points came from turnout, and the rest from switchers (7% of Trump voters backed a Democrat) and from 2024 non-voters, who broke 70–25.
Then the money. A record $11.2bn in ads, with Republicans spending $410m more, and Texas's Senate race the most expensive non-presidential contest on record at over $560m. The spending didn't fill the stands for the side that spent more.
In shortEvery voter in one hall, 990 seats, so each seat is about one in a thousand voters. Sections are how people voted in 2024; seat and flag colour is how they voted this week. Cards pop up from random seats to say who sits there. "Only switchers" lights the people who changed sides or were new. Drag to turn it.
Brantopic's Truth module checks what a brand's story says against the evidence. Here the "brand" is the 2026 midterm itself. Its messages are what the press said it meant; the evidence is the data behind this page.
In shortWe took twelve things outlets reported about the election, paraphrased and linked, and asked Brantopic to match each one against the fifteen facts behind our charts. A claim stands only if a fact supports it with confidence above 0.6. What survives is the election's truth.
Brantopic matches what is being said about you against the evidence, and keeps a receipt for every verdict. Free snapshot.
Each chart, its data, what we did to it and where it can mislead. Every dataset is stored with a content hash, its receipt, so any figure can be recomputed from the same rows.
| Chart | Data | Method | Limits |
|---|---|---|---|
| Tipping scale and ballot | Group margins for 2020 and 2024 (presidential) and 2026 (House) as published in a widely shared reference chart; group shares of 2026 voters from the NBC News/SSRS exit poll. | Margins read from pixel positions (about ±1 point). A group's push on the national margin = its share × its swing. The balance point is the share-weighted average margin; the tilt is the turning force around Even, capped at 6°. | Exit polls are estimates; subgroup error runs to several points. Shares are held at their 2026 values, so the balance ignores changes in who turned out. |
| The bill (receipt) | NBC News/SSRS exit poll, top-issue question and issue items; 122.7m ballots (state tallies compiled by Al Jazeera). | Net votes per issue = share naming it × (D% − R%) × ballots. Because the top-issue shares cover every voter, the lines add up to the national margin. | Says how issue voters split, not what caused their vote. "Also on the tab" items overlap and are not added. |
| The kerb | NBC News/SSRS exit poll crosstabs. | Margin = D% − R% per group; the band marks ±6 points. | Groups overlap; small groups (4–5% of voters) carry wide error. |
| Covers | GDELT DOC 2.0, US English-language sources, 5 Aug–4 Nov 2026, collected by Brantopic. | Daily matching articles ÷ all articles GDELT read that day. Brantopic's quasi-Poisson gate flags days above a 0.1% false-alarm bound, with a 28-day minimum baseline. Cover lines are headlines from each issue's busiest day. | Keyword queries miss some stories and catch some off-topic ones. GDELT covers what it crawls, not all news. |
| Blue wave city | The Downballot's 2024 presidential results on the 2026 district lines; CBS News's list of flipped seats (5 Nov). | Uniform swing: a seat is under water when the national shift exceeds Trump's 2024 margin there. Tower layout (shore to inland by margin) is for legibility only. | Dozens of races were uncalled on 5 Nov. Incumbency and candidates make real seats depart from uniform swing. |
| Texas relief | Decision Desk HQ county results (2026 Senate, 2024 President, 2022 turnout); us-atlas county shapes. | Swing = 2026 Senate margin − 2024 presidential margin. Height ∝ swing; colour = 2026 margin or ballots against 2022. | Senate against President, so part of the swing is the candidates. 41 counties were still counting on 5–6 Nov. |
| Convention hall | NBC News/SSRS exit poll: recalled 2024 vote × 2026 House vote. | 990 seats, each 0.1% of voters. Turnout effect = margin with the actual 2026 mix of Harris and Trump voters minus the margin with the 2024 mix, holding loyalty rates fixed. | People misremember their past vote, usually toward the winner. To be checked against voter files (Catalist, L2) in 2027. |
| Topline | Network projections (NBC, CBS, CNN, Yahoo/Newsweek), Al Jazeera turnout, AdImpact spending. | Taken as reported on 4–6 Nov. | Preliminary until states certify in December. |
One chart and one sourced finding from politic.design, in your inbox.