Trump approval at 30% or 42%? Why national polls disagree so much
Dated editorial analysis; figures in this article are not recalculated automatically. Visit the linked race pages for current data.
Based on national polls in the Midterm Results Live ledger, Sept. 27 to Oct. 5, 2026
If you read only one poll this month, you might think President Trump's approval rating is 30%. If you read a different one, you might think it is 42%. Both numbers come from established pollsters, and both were taken within about a week of each other.
The same thing is happening with the generic congressional ballot, the question that asks voters which party they plan to support for the House. Recent polls show Democrats ahead by anywhere from 2 to 16 points.
This kind of spread confuses a lot of readers. It is also completely normal. This piece explains where the differences come from and how to make sense of them.
The numbers
Here are the seven presidential approval polls in our ledger from the three weeks before October 8:
| Pollster | End date | Approve | Disapprove |
|---|---|---|---|
| Reuters/Ipsos | Oct 5 | 32% | 65% |
| Pew Research Center | Oct 4 | 30% | 68% |
| AP-NORC | Sep 28 | 31% | 69% |
| Economist/YouGov | Sep 28 | 36% | 62% |
| Harvard CAPS/Harris | Sep 28 | 42% | not shown |
| Quinnipiac | Sep 27 | 34% | 61% |
| Morning Consult | Sep 27 | 42% | 56% |
The simple average of approval across all seven is about 35%. But the polls cluster in two groups: four between 30% and 34%, and two at 42%, with YouGov in between.
And here are the five generic ballot polls:
| Pollster | End date | Democrats | Republicans | Margin |
|---|---|---|---|---|
| Economist/YouGov | Oct 5 | 52 | 40 | D+12 |
| Pew Research Center | Oct 4 | 44 | 36 | D+8 |
| Economist/YouGov | Sep 28 | — | — | D+16 |
| Harvard CAPS/Harris | Sep 28 | — | — | D+2 |
| Quinnipiac | Sep 27 | 51 | 39 | D+12 |
The average margin across these five is about D+10. Again, one poll, Harvard CAPS/Harris, sits well apart from the others.
Why polls of the same country give different answers
Who they ask
Pollsters choose different groups of people to measure:
- All adults: everyone 18 and older, whether or not they are registered to vote.
- Registered voters: people who say they are registered.
- Likely voters: registered voters the pollster expects to actually vote, based on screening questions or past behavior.
Each group gives a slightly different result. In midterm years, the people who show up tend to be older and more politically engaged than the adult population, and the out-party's voters are often more motivated. Pew's October release, for example, was headlined around Democratic voters being more engaged in the midterms. A poll of likely voters and a poll of all adults can therefore produce different numbers without either one being wrong.
YouGov's October release made this point directly: it reported the generic ballot under two different definitions of a likely voter, and Democrats led by double digits under both.
How they reach people
Some pollsters call landlines and cell phones with live interviewers. Some send text messages. Some use online panels of people who have agreed to take surveys. Some, like Pew and AP-NORC, recruit panels by mail using random samples of addresses. Each method reaches a slightly different mix of people, and each pollster adjusts its data, or "weights" it, to correct for that in a different way.
How they ask the question
Small wording differences matter. Some polls give respondents a "not sure" or "no opinion" option, and some push them to choose. Pew's generic ballot, at 44% to 36%, adds up to only 80%, which means about one in five respondents did not pick either party. A poll that pushes those people to choose will show higher numbers for both parties, and the margin can change as well.
When they ask
These polls ended on different days, between September 27 and October 5. In a stable race, a week does not change much. But news events can move numbers quickly, and two polls a week apart are not measuring exactly the same moment.
"House effects"
Over time, some pollsters consistently produce results that are a little more favorable to one party than the average. Analysts call this a "house effect." It is not necessarily bias. It usually comes from the methodological choices above. Harvard CAPS/Harris and Morning Consult both put approval at 42% here, which is well above the other five polls. That does not make them wrong, but it does mean you should compare a pollster's new number with its own previous numbers, not with a different pollster's.
How to read the spread
Three rules of thumb help:
- Look at the cluster, not the extremes. Five of the seven approval polls put the president between 30% and 36%. That is the most useful single takeaway.
- Compare a pollster with itself. If YouGov moves from D+16 to D+12 in a week, that tells you more than comparing YouGov's D+12 with Harris's D+2.
- Do not over-read one poll. A single new survey that is far from the others is more likely to be noise than a sudden change in public opinion.
Why it matters for November
The generic ballot and presidential approval are not direct forecasts. But historically, the president's party has usually lost House seats in midterm elections, and the losses have tended to be larger when the president is unpopular. Even the most favorable numbers here, 42% approval and a two-point Democratic lead, are not strong numbers for the party in power. The less favorable numbers point to a difficult environment for Republicans in close House and Senate races.
We show these national indicators, and the sources for each poll, on our home page and on latest polls. Every poll links to its original release so you can check the numbers yourself.