Trump’s 2026 Midterm Warning Signs: Approval, Voter Enthusiasm and the Early-Vote Data

The polls are noisy, the early-vote data are immature and some of the loudest conclusions are plainly premature. There is still a pattern worth paying attention to.

Republicans could be forgiven for feeling relieved by the latest NBC News poll. Democrats led the generic congressional ballot 50% to 45%. Five points is hardly good news for the party in power, but after some considerably uglier numbers, D+5 has begun to look almost respectable.

That reaction says something by itself.

The NBC poll surveyed registered voters. That distinction matters, although perhaps not in the way partisans would like. A registered-voter poll is not "wrong" and a likely-voter poll is not automatically better. They are measuring somewhat different electorates, and every likely-voter screen contains assumptions about who will actually show up.

Still, there is an awkward detail for Republicans. The enthusiasm numbers are not helping them.

UMass Amherst found 42% of Democrats describing themselves as "very enthusiastic" about voting in the midterms, against 30% of Republicans. Include the somewhat enthusiastic and the gap narrows, but it remains: 67% to 59%. In the same survey, Democrats led the generic House ballot 42% to 34% once leaners were included.

That does not convert D+5 into D+10 by magic. It does make the registered-voter versus likely-voter distinction harder to dismiss.

There is a familiar human temptation here. Find the number that hurts least, call it the sensible number and declare the others hysterical. Markets do this. Corporate boards do it. Political movements do it with particular enthusiasm. The problem is that the least unpleasant datapoint is still a datapoint.

Five points down is five points down.

Iowa gets interesting

Then there is Iowa.

A Trafalgar survey of 1,089 likely voters put Republican Ashley Hinson at 44% and Democrat Josh Turek at 42.1%. Call it R+2 if you like. More precisely, it is a two-point race in a state Donald Trump carried comfortably two years ago.

One poll cannot tell us where Iowa will finish. It can tell us that the race is close enough to deserve attention.

The more interesting clue may be sitting at the petrol pump.

Hinson and fellow Iowa Republican Mariannette Miller-Meeks have been pressing the administration for relief from high diesel and gasoline prices. Proposals have included a temporary restriction on diesel exports and suspension of the federal gasoline tax. The politics are obvious. In an agricultural state, diesel is not an abstract component of CPI. It is an input cost. Farmers buy it. Truckers buy it. Eventually everybody buys some part of it through the prices of everything else.

It would be convenient to turn that into a grand theory about rural voters suddenly discovering economics. The evidence does not get us that far.

What it does show is more prosaic and probably more important. Policy feels different once it arrives as a bill.

People can ignore Washington theatre for months. They have more difficulty ignoring a fuel receipt.

That does not prove expensive diesel moved two points in an Iowa Senate poll. It tells us why Republican politicians in Iowa are suddenly treating fuel costs as urgent. Their behaviour is evidence too.

The Kennedy Center story is stranger than the caricature

The Kennedy Center affair is another case where exaggeration gets in the way of a perfectly serious criticism.

The centre does have financial problems. Internal documents obtained by The Washington Post showed projected fiscal-year revenue falling from roughly $220 million in the budget to about $124 million, with a deficit of around $23 million even after substantial spending cuts. Ticket sales and fundraising had weakened badly. Center officials blame inherited mismanagement and say Trump's involvement has helped attract money; the internal projections nevertheless show an institution under real financial strain.

So the clean story cannot be that a perfectly healthy institution was destroyed overnight because Donald Trump wanted his name on the wall.

The actual story is more peculiar.

Congress approved roughly $257 million for repairs. Trump has argued that his involvement and fundraising power are essential to the building's rescue. His allies on the board have repeatedly tried to attach his name to the Kennedy Center, while a federal judge has ruled that the board cannot rename the institution without congressional authority. The board has now voted to close the building for renovations, and Trump has publicly entertained much more radical reconstruction if necessary.

This is where governance starts to look odd even before anyone reaches for a partisan adjective.

A national cultural institution needs money. It needs repairs. Its revenue has deteriorated. Those are ordinary management problems, albeit serious ones.

Once access to political fundraising, the future of the building and recognition of the sitting president become tangled together, the problem changes character. The institution is no longer dealing only with its balance sheet and physical plant. It is dealing with the preferences of the person who temporarily controls the political machinery around it.

That is a much stronger criticism than calling the entire episode an act of personal vandalism, because it does not require guessing what Trump privately wants. The public record is strange enough.

Robert F. Kennedy Jr.'s role is awkward for a different reason. Trump discussed the naming dispute in Kennedy's presence and described putting the Kennedy and Trump names together as a symbol of bipartisanship. Kennedy did not publicly challenge him during that exchange.

Readers can decide for themselves what to make of that silence. Anything more would be mind-reading.

Austin needs fewer adjectives

The ICE shooting in Austin is the part of this story least improved by rhetorical excess.

An immigration officer shot 28-year-old Venezuelan Wilber Rafael Garces Perez on September 20. His lawyers say he entered the United States legally through the CBP One process in 2024, applied for asylum and had permission to work. Subsequent reporting says he missed an immigration hearing after notice was sent to the wrong address. DHS has given its own account of his status and the circumstances surrounding the encounter.

There were protests afterwards. Local reporting described dozens of protesters, not hundreds. Police used pepper-ball-like projectiles after some demonstrators threw rocks and bottles. Austin Mayor Kirk Watson called for an independent investigation.

Those facts do not need decorating.

Nor do they justify turning one incident into proof that every American police department is simply an arm of political repression. That conclusion is larger than the evidence. So is the claim that the shooting will cost Republicans some calculable number of Texas votes.

The narrower question is uncomfortable enough: when a federal enforcement action ends with a man shot in the street and the government and his lawyers giving materially different accounts of how he came to be there, who does the public trust?

Institutional credibility is an asset. It can be spent.

Once spent, it is expensive to rebuild.

About that 29%

American Research Group now has Trump's approval at 29%. It is an extraordinary number. It is also at the low end of the current polling range, which is exactly why it should not be treated as though every other poll has vanished.

Reuters/Ipsos has him at 32%. NBC has him at 41%. Different methodologies and survey populations produce different answers. What is harder to explain away is the weakness across several polls, particularly on the cost of living. Reuters/Ipsos found Trump's approval on that issue among Republicans falling from 70% to 44%, while his overall approval among Republicans fell from 86% in January to 73%.

That is more revealing than comparing him with George W. Bush by adding up wars, casualties, government expenditure and stock-market returns as though presidential approval were a quarterly earnings calculation.

People do not price presidents that way.

They notice what rent costs. They notice the grocery bill. They notice petrol. They notice whether the economy described at a podium resembles the economy they encounter on Tuesday afternoon.

Trump's political strategy makes those numbers particularly interesting because he is refusing to become background scenery.

In North Carolina he told voters, in effect, to treat him as though he were on the ticket. He has also attached a promised $5,000 payment to continued Republican control of Congress. Republican congressional candidates have been notably less enthusiastic about making that proposal the centre of their own campaigns, and House Speaker Mike Johnson has acknowledged that such payments would require congressional action.

There is an obvious tension here. Trump wants the midterms associated with Trump. Some Republican candidates seem to prefer a little more room to manoeuvre.

No sophisticated analysis needs to pretend that tension guarantees an outcome. It simply exists.

And then there are the early-vote charts

This is where everyone should calm down.

North Carolina has not started in-person early voting.

It begins on October 15.

What North Carolina has now is absentee-ballot activity, along with ballots for military and overseas voters. The large in-person early-vote pool that will eventually dominate much of the analysis does not yet exist.

So any chart purporting to tell us what North Carolina's full early electorate means for November is trying to extract a conclusion from data that have not arrived.

The same caution applies elsewhere. Application figures can be interesting. Return rates can be interesting. Racial composition can be interesting. None of them becomes a party result simply because somebody puts a red line and a blue line on a chart.

Georgia is a good example. A fall in the white share of absentee requests could matter. It could also reflect changes in voting method, age, geography or who has decided to vote by mail rather than in person. Race is not party registration, and an absentee application is not a vote.

Virginia is already voting in person, which makes its numbers more useful. Even there, Virginia does not register voters by party, so any Democratic or Republican classification involves inference from geography, primary history or modelling.

This is tedious. It is also how election analysis is supposed to work.

Early voting attracts bad analysis because the numbers arrive every day. Daily numbers create the illusion that something important must also be happening every day. Usually the denominator is the story.

A 20% increase from what?

A fall in white share compared with which election?

More Democratic-leaning voters than when?

Did those people vote early last time, or are they genuinely new voters?

Did one party expand its electorate, or merely persuade the same electorate to vote sooner?

Without those answers, the chart is decoration.

The broader picture in late September is less dramatic than either side would prefer.

Democrats have an enthusiasm advantage in the UMass survey. NBC has them five points ahead among registered voters. Trump's approval is weak across several current polls, although the precise number varies sharply by pollster. Iowa is close enough to be interesting. High fuel prices have become politically salient enough that vulnerable Republicans are asking Washington for relief. The Kennedy Center has genuine financial problems wrapped inside an extraordinary political and legal dispute. An ICE shooting in Austin has produced competing official and legal accounts that deserve scrutiny. And most of the early-voting data being waved around as destiny are nowhere near mature enough to carry that weight.

That picture is messy.

Good.

Politics usually is.

The useful discipline now is to resist the poll that feels best, the chart that arrives first and the explanation that makes everything fit. There will be plenty of time for certainty after the votes are counted.