ANALYSIS: Democrats lead by nine points, but ten redrawn maps already weigh on the House

52 to 43

6 a.m., September 25

A poll this morning delivered a scoreline.

On September 25, 2026, at 6 a.m., NBC News published its polling averages. For control of Congress, Democrats lead 52% to 43%, adjusted to likely voters. 9 points. Their biggest lead of the year.

A point in a poll is one person in a hundred leaning one way. In a 435-seat chamber, a point isn’t worth the same everywhere.

The number and the map

That is the whole subject of this analysis. What a point is worth on November 3, when it has to be won district by district.

Because while voters were drifting away from the president, ten states redrew their electoral maps. Some for one party. Others for the other.

The poll counts people. The map counts districts.

A poll point isn’t a seat yet.

The biggest lead of the year

An average, not a poll

NBC’s average is not a poll. It is a calculation.

It combines every public national poll that meets transparency standards since the start of Trump’s second term. The most recent weigh more. So do the most accurate. A pollster who publishes too often sees its older polls lose weight.

NBC writes it itself: its averages are not predictions.

Likely, registered, adults

Polls don’t all target the same population. Some survey all adults. Others, registered voters. Others still, likely voters.

Polls of all adults consistently give Trump lower approval than polls of registered or likely voters, NBC notes.

After Labor Day, NBC’s congressional average switches to a likely-voter basis. Both parties post higher shares there, because those voters have often already made up their minds.

At the start of the term, NBC recalls, the two parties were tied.

Tied in 2025. Nine points in September 2026.

An average smooths the noise. It doesn’t vote.

Marist says twelve

1,280 registered voters

An independent source puts the gap wider still.

The NPR/PBS News/Marist poll, conducted on September 14 and 15 among 1,280 registered voters, puts Democrats at 53% to 41% for Republicans. Twelve points. The margin of error is 3.8 points.

In May, the same poll gave Democrats a ten-point lead.

A vote against

Marist also asks another question. Is your vote for Congress mainly a vote for Trump, or against him?

45% of registered voters say against. 24%, for. 30% say Trump doesn’t weigh much in their choice.

Among independents, 54% describe a vote against the president.

On the issues, Marist gives Democrats the edge on the economy, 42% to 34%, on foreign policy, 42% to 36%, and on gas prices, 40% to 27%. On immigration, the gap narrows: 41% to 38%. In 2024, Republicans led there.

56% of voters say they prefer candidates with fresh ideas over a proven record. A climate hostile to incumbents, Marist writes.

Trump is asking voters to vote as if he were on the ballot, NPR reports. A plurality seems to have taken him at his word.

He isn’t on the ballot. He is in every voting booth.

A president at 39%

The president, in the average

In NBC’s average, Trump is approved by 39% of voters and disapproved by 59%. He began his second term at 50%.

Marist lands on the same 39%. But the detail matters. 27% strongly approve. 53% strongly disapprove.

A hardened electorate, Marist writes. Not an electorate waiting to be persuaded.

Among his own

Among Republicans, his approval hovered in the mid-to-high 80s for six months. It is at 79%, according to NBC.

Among independents, it falls to 26%.

On inflation and the cost of living, an NBC News Decision Desk poll conducted with SurveyMonkey measured 40% approval in April 2025. In April 2026, 32%.

This piece argues that the 9-point figure is, first of all, this: a real lead, but a national one, which will have to cross maps drawn to slow it down.

Nine points ahead. Two and a half already drawn away.

A president at 39% is not yet a lost Congress.

Independents have moved

From six to sixteen

Partisans barely move. Each camp keeps its own.

The movement comes from independents. At the start of the second term, they preferred a Democratic Congress by about 6 points, according to NBC. Today, by 16 points.

In the spring, the gap had even reached about twenty points.

What tipped them

NBC names three causes. The handling of the economy. The war in Iran. The deaths of two Americans at the hands of federal immigration agents.

Marist measures the same slope. Independents there choose the Democratic candidate 57% to 32%.

Sixteen points at NBC, twenty-five at Marist. The two don’t measure the same population or in the same way. They point in the same direction.

Independents don’t vote for a party. They vote against a moment.

A dozen steps to the side

Hinson, Rogers, Salazar

That same September 25, NBC News reported that at least a dozen Republican candidates had distanced themselves from Trump in two weeks.

NPR gives names. Ashley Hinson, running for the Senate in Iowa, and Mike Rogers, in Michigan, have come out against the war in Iran and are calling for help with rising prices.

In Florida, Representative María Elvira Salazar is running an ad that says Trump’s immigration policy has gone too far.

Seven votes in the House

In the House, seven Republicans joined Democrats last week on a resolution to withdraw U.S. forces from Iran, according to NPR. There were four this summer.

Even outside the campaign, cracks are showing. Utah Republican Senator John Curtis, who isn’t up for reelection until 2030, asked the Senate Judiciary Committee to investigate Donald Trump Jr., after a ProPublica article, according to NPR.

Republican strategist Sarah Longwell, a longtime critic of the president, sees it as the start of an acknowledgment: Trump no longer has the same strength.

Republican strategist Ron Bonjean tells NPR that those in real danger at the polls have to set themselves apart.

And yet Salazar also says she had a fantastic conversation with the president. The distance has its limits.

They step away from the president, not from his voters.

Ten states, ten maps

The wave that began in Texas

Here is what the national number doesn’t show.

In July 2025, Trump suggested Texas revisit its map to add Republican districts, Ballotpedia recalls. The U.S. Supreme Court then allowed the new Texas map to be used in 2026. Potential effect, according to Ballotpedia: five more Republican seats.

California answered. On November 4, 2025, its voters approved Proposition 50 by 65% to 35%. Potential effect: five more Democratic seats.

Ballotpedia’s count

Then the others.

A map for Florida, signed by Ron DeSantis on May 4, 2026: four potential Republican seats. A map for Ohio, approved by its commission on October 31, 2025: two. A map for North Carolina, passed on October 22, 2025: one. Tennessee, signed May 7, 2026: one. Louisiana, signed May 29, 2026: one. Utah, a court-approved map: one Democratic seat.

Alabama adds one potential Republican seat, after a green light from the U.S. Supreme Court.

Before 2025, Ballotpedia notes, only two states had voluntarily redrawn mid-decade since 1970.

In all, Ballotpedia puts the potential effect of the new maps at 10 seats more for Republicans.

That total includes a seat in Missouri.

It shouldn’t anymore.

Before the first ballot, seats had already changed hands on paper.

Missouri, between two maps

September 3, a unanimous ruling

On September 3, 2026, the Missouri Supreme Court issued a unanimous ruling. The new map adopted in 2025 never took effect.

A referendum will put it to voters on November 3. Until then, Judge Ginger Gooch writes, the map drawn in 2022 remains fully in force for the general election.

Six Republicans, two Democrats. Not seven to one.

A primary on one map, an election on another

It all started with a petition. The group People Not Politicians gathered more than 300,000 signatures to put the map to a vote. On August 4, 2026, the secretary of state had rejected it. The state Supreme Court overturned that refusal.

Yet the August primaries were held on the new map.

August, one map. November, another.

On September 10, the website Political.org reports, the U.S. Supreme Court stayed a conflicting federal order. Secretary of State Denny Hoskins then told local authorities to use the 2022 map.

The scene is anything but hypothetical. It applies to every polling place in Missouri. A voter chose a candidate in August in one district. He will vote in November in another.

The same name on the envelope. A different line on the map.

And if the November referendum approves the 2025 map…

On Ballotpedia’s paper, the balance therefore drops from ten to nine potential Republican seats.

In Missouri, the map changed between two votes by the same voter.

A 2.5-point tilt

Nate Cohn’s estimate

How much are these maps worth in points? There is one estimate, and only one I rely on.

According to an analysis by Nate Cohn of The New York Times, cited on September 2 by SciLine, Republicans could in May 2026 lose the popular vote by 2.5 points and still keep the House, thanks to redistricting.

That estimate predates the Missouri ruling. It is not a seat forecast. It is a tilt.

A House that had turned neutral again

The tilt is recent. In March 2025, three researchers, Nicholas Stephanopoulos, Eric McGhee and Christopher Warshaw, wrote that the House had shown no overall partisan bias in 2022 and 2024.

They added that this neutrality might not last, if states redrew their maps mid-decade.

It happened.

The field was level. Someone tilted it.

2024, as a yardstick

The Clerk’s count

We need a yardstick. The Clerk of the House provides one.

In 2024, Republican House candidates received 74,390,864 votes. Democrats, 70,571,330. Out of just over 149.5 million votes cast.

About 49.7% to 47.2%. A gap of about 2.5 points.

220 to 215

The result: 220 Republican seats, 215 Democratic, at the start of the 119th Congress.

A two-and-a-half-point lead in votes gave a five-seat lead.

Two and a half points is also, by coincidence, the tilt Cohn attributed to the new maps in May.

In other words: in 2026, a party could win the national vote by the margin that was enough for Republicans in 2024, and lose the House.

The same gap can give a majority or take it away.

An average is not a vote

Trump’s districts

Now for what tempers the Democratic reading. And it has to be written.

The average measures a national preference before candidates and local races. NBC says so. It adds that many of the decisive races are in districts and states Trump won in 2024, some by double digits.

A national poll doesn’t see those districts one by one. It doesn’t see the candidates. It doesn’t see the maps.

So who does?

What voters believe

Marist asked who will win in each district. 52% of registered voters think the Democratic candidate will prevail. 44% bet on the Republican.

An advantage, not a predicted landslide.

And yet nine points remain well above the estimated 2.5-point tilt. Even reduced, the gap leaves a margin.

The question isn’t whether the map erases the lead. It is how much of the lead survives.

A national lead gets spent district by district.

Both parties held the scissors

Two can play

It would be easy to make this a story with a single culprit. The record doesn’t allow it.

Texas redrew at Trump’s request. California’s Democratic governor, Gavin Newsom, answered in July 2025 that two could play that game, Ballotpedia recalls. And he played.

In Virginia and Maryland, Democratic attempts failed, politically or in court, according to SciLine. In Georgia, Indiana and South Carolina, Republican efforts failed too.

The same move, two signatures

The same move in Texas, at the president’s request. The same move in California, by referendum. The same move in Missouri, suspended by a court. And in Virginia, where it failed.

Choosing your voters before they choose you.

SciLine recalls why the wave was possible. The U.S. Supreme Court made it much harder to challenge partisan maps in federal court. And after the Louisiana v. Callais ruling, several Southern states dismantled majority-Black districts.

In New York, by contrast, the Supreme Court on March 2, 2026, blocked a decision ordering the map redrawn, according to Ballotpedia. The existing map will be used.

The only difference lies in the balance. According to Ballotpedia, it tilts toward Republicans.

I admit this passage troubles me more than the others. I don’t know how to defend one party that redraws against another party that redraws.

I only know that the voter held neither pencil.

Two parties hold the scissors. One voter loses points.

What nine points are worth

The proportion

So here is the proportion, as the sources allow it.

Nine points in NBC’s likely-voter average. Nine points resting on a swing among independents. Nine points that coexist with a president at 79% among Republicans. Points that will have to pass through districts Trump often won. Points facing a 2.5-point tilt.

Twelve at Marist among registered voters. A 2.5-point tilt estimated in May by Cohn, before Missouri removed a potential Republican seat from the count.

I don’t convert this into seats. None of the sources published here does so for September.

How many seats, exactly? Nobody publishes it.

What remains to be won

What I can say is simpler. The Democratic lead is bigger than the tilt. It isn’t infinite.

A September poll is not a November vote. A map, though, is already printed.

And yet the less visible of the two is the one that weighs most surely.

The average moves every week. The map no longer moves.

November 3, district by district

Thirty-nine days

A little over a month remains before November 3.

Thirty-nine days. Ten maps. One vote per voter.

The average will say more. Republican candidates may step further away. The president may stay at 39%.

But in ten states, the lines are drawn. In Missouri, they have just been redrawn the old way.

The point and the pencil

The booth curtain pulled shut. The pencil on its string. The small sound of the machine swallowing the ballot.

On November 3, in those ten states, who will have the last word: the ballot, or the pencil that drew the map?

On November 3, every point will be counted. Not everywhere at the same price.

Nine points in the booth. Ten maps drawn before it.

Sources:

Primary Sources:

• Marist Poll — NPR/PBS News/Marist poll on the midterm elections, September 23, 2026
• Clerk of the House of Representatives — statistics for the November 5, 2024, election, March 10, 2025
• Missouri Supreme Court — Von Glahn v. Hoskins ruling on the electoral map, September 3, 2026

Secondary Sources:

• NBC News — polling averages, Democrats at 52 to 43, September 25, 2026
• Ballotpedia — redistricting ahead of the 2026 elections, September 2026
• SciLine — guide to redistricting and Nate Cohn’s estimate, September 2, 2026