Trump's approval is near the lowest of any midterm since 1946

Donald Trump goes into the 3 November midterms with a job approval of 35% on our average of every pollster, and history has no close match for it. Put on the scale Gallup used for every president since Harry Truman, the figure is 31.5%. That is lower than any president's approval in the run-up to any midterm since 1946.

Our smoothed average stood at 42% on 1 January and has stayed below 40% since early May. It has slipped again this month, from 36.5% on 1 October to 34.7% on 13 October. Most of the newest polls in our table are lower still: Ipsos at 32%, Pew Research Center at 30% and AP-NORC at 31%, with Rasmussen Reports at 41% among the highest. Newsweek reported that AP-NORC's reading, taken from 24 to 28 September, was the lowest of either Trump term.

Why the scale matters

Gallup tracked presidential approval for 88 years and stopped publishing it this year. CNN reported in February that the firm would no longer publish assessments of individual political figures. Its last reading of Trump dates from December 2025. That leaves a problem for anyone comparing today with the past, because the history is Gallup's and the present is everyone else's.

In 2025, the last year both existed, Gallup ran 3.8 points below the other pollsters in the same months. Our research table applies that offset to this year's readings: each pollster's latest figure averages 35.3%, which becomes 31.5% on Gallup's scale. The adjustment is an estimate, but it does not change the ranking much. On Gallup's scale Trump is below the weakest midterm president in the record, Truman in October 1946 at 33%. Even on the unadjusted figure, only Truman ranks lower.

What low approval has cost before

Our history table holds all 20 midterms from 1946 to 2022. The president's party lost House seats in 18 of them, and the average loss was 25.6 seats. It lost Senate seats in 14, by 3.7 seats on average. The two exceptions in the House, 1998 and 2002, came with presidents at 64% and 67%.

Approval sorts the rest. In the 11 midterms where the president stood below 50% in October, his party lost 34.8 House seats on average. In the nine where he stood at 50% or higher, it lost 14.3. Only three midterms came with the president below 40%: Truman in 1946 (33%, 55 seats lost), Truman again in 1950 (38.5%, 27 seats) and George W. Bush in 2006 (38.8%, 30 seats).

A straight line through the 20 midterms gives about 1.1 House seats for every point of October approval. Run at 31.5%, it implies a Republican loss of about 47 seats. That line has a typical error of 19 seats when each midterm is predicted with that year left out, and approval explains only about a third of the variation in seat losses. Ford in 1974 lost 48 seats with approval of 55.5%, after Watergate. Joe Biden in 2022 lost only nine with approval of 40%.

Where our forecast sits

Our forecast does not use approval. It builds the House from the generic ballot, district by district, and today gives the Democrats 231 seats to the Republicans' 204. That is a Republican loss of 16 from the 220 seats they won in 2024, with a Democratic majority in 78% of our simulations. The 80% range runs from 210 to 270 Democratic seats, so a loss on the scale the approval line suggests sits inside our range but toward its Democratic end.

The two methods agree more closely on votes than on seats. The same history, fitted to the national House vote, says a president at Trump's September level costs his party about six points of its two-party share. That would take Republicans from 51.2% in 2024 to about 45%. Our generic-ballot polling average is D+8.8, which works out to a Republican two-party share a little under 46%. After our correction for how generic-ballot polls have missed before, our reading is D+6.7.

So the difference lies in how votes become seats. The approval line rests mostly on midterms from an era with far more competitive districts, when a modest swing moved many more seats. It knows nothing about this year's map, which includes new lines in nine states. Our district model does. We would treat the approval line as a sign that the risk to Republicans runs toward a larger loss, not as a second forecast.

Our "this day" comparison points the same way. In the five midterms since 2006, the House vote ended an average of 2.0 points worse for the president's party than CNN's generic-ballot reading at this stage, though in two of those years it ended better. Five cycles are too few to build that into the forecast, and we do not.

The Senate is a different calculation

Approval says much less about the Senate, where only a third of seats are up and which ones matters most. Fitted on the same midterms, approval explains about an eighth of the variation in Senate losses. Taken at face value, it points to a Republican loss of about six seats.

This year Republicans defend 22 of the 35 seats on the ballot and Democrats 13. Our projection is a 52-48 Democratic Senate, a net Republican loss of five, with Democratic control in 57% of our simulations. Three races sit at toss-up.

The latest

The news this month has added to the pressure without changing the picture. Reuters reported on 8 October that its latest Reuters/Ipsos poll put approval at 32%, level with its record low, and that 30% of Trump's 2024 voters called themselves very enthusiastic about voting next month, against 48% of Kamala Harris's voters. Newsweek reported that Pew found Trump's approval among his own 2024 voters at 67%, down from 75% in July.

At a rally in Clarksville, Tennessee, on 10 October, the president told supporters, as Newsweek reported, that Republicans would win "because I am popular." The polls in our table do not show that. A president at 35% has no close precedent in a midterm year, and the history that does exist points to a heavy loss. How heavy depends on a map that history did not have to deal with.