Trump Approval Rating Summer 2026: What the Polls Actually Show

Two of the most-followed public trackers of Trump’s job approval covered summer 2026 and came away with different stories. One has him at the lowest point of his presidency. The other had him improving.

Both readings are defensible. The reason they differ is more interesting than either number.

What is Trump’s approval rating right now?

35.3% approve and 58.6% disapprove, on the Civiqs daily trendline as of 27 August 2026 — a net of −23.3. That is within two-tenths of a point of the lowest reading of his second term.

The most recent reading from the other source here is older and higher: Marquette’s national survey of adults, interviewed 22–29 July, put it at 40% approve and 60% disapprove.

Those are five points apart, and about half of that gap is a methodology difference that never goes away. The rest is a month of calendar.

Did Trump’s approval change over the summer?

Barely, and then decisively at the end. On the Civiqs series the whole of summer fits inside 2.3 points — it opened at 36.5% on 1 June, peaked at 37.4% on 16 June, and closed at 35.3% on 27 August.

For scale, that same tracker has covered 10.8 points across the whole second term. Three months moved a fifth of that.

But the floor of those three months is also the floor of the presidency: 35.1% on 15 August, the lowest single day of the term. A summer that went almost nowhere still finished somewhere the number had never been.

The monthly averages show it was a slide rather than an event:

August was the steepest month of 2026 so far, tied with April. Nothing in the shape of the line suggests a single moment; it is a drift that accelerated.

Why do two trackers show the summer differently?

Because they are measuring different people, asking in a different format, and — decisively — stopping on different dates.

Set each Marquette survey against the average of the Civiqs line across that survey’s own interview dates, and the picture resolves:

The Civiqs column is that tracker’s average across each Marquette survey’s own interview dates, not a headline figure.

Across all eleven aligned windows the correlation is r = 0.95 and the average gap is +2.4 points. They are the same descent, drawn at two heights.

Three things account for the difference:

Registered voters versus all adults. Civiqs polls registered voters; Marquette polls adults. These are different populations and there is no fixed conversion between them.

What happens to people with no opinion. This is the biggest single factor and it is pure arithmetic. Civiqs’ latest reading is 35.3 plus 58.6, which leaves 6.1% expressing no opinion. Marquette’s July toplines are 40 and 60 — they sum to exactly 100, so effectively nobody is left out. A survey that pushes undecided respondents onto one side or the other will read higher for whichever side they lean toward. That is not a flaw in either; it is a design choice with a predictable consequence.

Marquette stopped measuring on 29 July. This is the part most coverage misses. The entire August decline, including the 15 August term low, falls after Marquette’s last interview. The two sources are not disagreeing about August. Only one of them has been there.

That last point cuts both ways. Marquette’s next national survey will be the test: if it comes back at 38 or lower, the August fall was real and Marquette simply had not caught it. If it comes back at 40 again, the Civiqs dip was noise in a modelled series.

What does “strongly approve” show that the headline number doesn’t?

That the July improvement was hollow. In the same survey where Marquette’s headline rose two points, strong approval fell to 15% — the lowest of the term.

Here is the June-to-July move broken out:

Every point of the gain, and one point more, came from “somewhat approve.” The committed end of his support kept shrinking through the month his headline number rose.

Over the term as a whole the trend is unbroken. Marquette’s published table has strong approval at 26% in March 2025 falling to 15% in July 2026, while somewhat-approve rose from 20% to 25%. The release’s own text puts strong approval at 29% at the very start of the term, in a January 2025 survey that predates the published table — which would make the fall roughly half of where it started.

Marquette attributes the July uptick to improving economic perceptions, specifically views of gas and grocery prices. Its own issue numbers put a ceiling on how much that helped: approval of his handling of inflation, the economy, the Iran war and gas prices all sat at 30% or below in the same survey.

How does this compare with the rest of his second term?

It is the bottom of a long, fairly smooth descent rather than a collapse. Both sources show the same arc from a start in the mid-to-high 40s.

Marquette’s own framing is that approval “declined steadily throughout his second term until this improvement in July,” with the number falling into the high 30s from April onward. The Civiqs monthly series says the same thing in more granularity: nineteen months, one direction, no reversal lasting more than a few weeks.

What is unusual is not the level but the stability of the disapproval side. Civiqs has disapproval between 57 and 59 for most of 2026. The movement in the net figure comes mostly from approval leaking away, not from disapproval hardening further — there is not much room left for it to.

What can these numbers not tell you?

Quite a lot, and it is worth being specific about which things.

They are not a forecast. Approval measures the present. There is a relationship between presidential approval and midterm outcomes, but it is loose, it is mediated by turnout and candidate quality, and translating one into the other is a separate exercise with its own error. We keep that on the 2026 midterm House forecast rather than smuggling it in here.

A modelled daily line is not a poll. The Civiqs trendline is a smoothed estimate, not a single survey with a single margin of error, and its most recent days are the least settled — they get revised as more interviews arrive. Our own state-level page found exactly that: reading the same date two days later moved some states by more than two points, with no systematic direction .

Two sources is two, not many. These are the two that publish enough underlying detail to be checked, which is why they are the two used here. A broader average would include houses whose methods cannot be inspected.

Sub-2-point moves are not news. Marquette’s July survey has a stated margin of error of ±3.2 points on 1,076 adults. Its two-point rise does not clear that on its own — what makes it worth reporting is not the size but the composition, which is where the strong-approval finding comes from.

Sources