United States Presidential election7 Nov 2028in 717 days Senate D 99% House D 58% Calendar Track record Research

Senate incumbents have lost almost all their polling edge

Sitting senators poll about 1 point better than open-seat candidates of their party this year, down from 20 points in 2006 and 9 in the 2010s.

Written in house from our poll data and our forecast.

Vice President Kamala Harris swears in Senator Chris Van Hollen for a second term.jpg
Photo: Office of Senator Chris Van Hollen

Sitting senators running again this year poll about 1 point better than open-seat candidates of their own party in states of similar partisanship. On the same measure the figure was 20 points in 2006 and about 9 through the 2010s. It fell to between 3 and 4 points in each of the three elections since 2020. On the current polls it is close to nothing.

Governors appear to have held on to more of it. Incumbent governors poll 7 points ahead on the same measure. That figure rests on three people, though, and without them it disappears too.

How we measured it

For every election year we took each Senate and governor race with independent polling. For 2006 to 2024 we used the average of nonpartisan polls in the final three weeks; for 2026, our current adjusted poll average, three weeks before election day. We then set each race's polled margin against the state's partisan lean and against whether an incumbent was on the ballot. The incumbency figure is the extra margin a sitting senator or governor polls compared with an open-seat candidate of the same party, in a state with the same lean, in the same year. Our history begins in 2006, the first year our records mark incumbents.

The 2026 Senate figure covers 26 races. It leaves out Idaho, Nebraska and South Dakota, where an independent rather than a Democrat is the main challenger, and six seats with no polls. The estimate is 1.1 points, with a standard error of 2.4. Counting the three independent races moves it to 0.1.

Year by year, the Senate figure reads 20 points in 2006, 16 in 2008, 8 in 2010, 10 in 2012, 9 in 2014, 10 in 2016 and 8 in 2018. Then it drops to 3.4 in 2020, 3.7 in 2022 and 3.6 in 2024. The polls tracked the results. Measured on the actual vote the figure was also 20 in 2006, and 3.8, 5.5 and 5.6 in the three most recent elections.

We tried to break the 2026 number. Dropping any single race leaves it between minus 0.6 and 2.0. A regression that damps outliers gives 1.5. Keeping only states within 20 points of the national lean gives 3.1, against 7 to 20 points in every election from 2006 to 2018. Leaving out the three appointed senators, Ashley Moody in Florida, Jon Husted in Ohio and Darline Graham in South Carolina, lifts it only to 2.0. Timing does not explain it either: in past cycles the measure taken a month out was close to the final reading, at 4.6 points 30 days out in 2024 against 2.8 two weeks out.

Republican senators are polling below their party

The near-zero average hides a split by party. Democratic incumbents poll 9 points ahead of open-seat Democrats in comparable states. Republican incumbents poll 3.1 points behind open-seat Republicans.

Set against what open-seat Republicans are polling in states with similar leans this year, Tom Cotton in Arkansas runs about 12 points behind, Roger Marshall in Kansas 10, Graham 9, Husted 8 and Dan Sullivan in Alaska 6. In Kansas, a state with a lean of R+16 on our measure, our average has Adam Hamilton 1.2 points ahead of Marshall. In Ohio, Sherrod Brown leads Husted by 3.6. Sullivan is 0.3 points ahead of Mary Peltola. Roll Call's September list of the most vulnerable senators put Husted first, citing a Marist poll with Brown 8 points ahead, followed by Susan Collins, Sullivan and Marshall. Graham was appointed in July after the death of her brother, Lindsey Graham, Ballotpedia reported, and won the Republican nomination in an August runoff against Ralph Norman.

The three races against independents sit outside the measure but point the same way. Pete Ricketts leads Dan Osborn by 0.6 points in Nebraska, a state with a lean of R+20. Mike Rounds leads Brian Bengs by 4.9 in South Dakota, lean R+28.

Collins is the exception among Republicans. She polls about 9 points better than an open-seat Republican would in Maine, yet still trails Troy Jackson by 1.6 in our average, because the state leans 7.5 points Democratic. On the Democratic side Mark Warner in Virginia runs 12 points ahead of the open-seat expectation and Jon Ossoff in Georgia 7.

Governors: three outliers carry the average

Incumbent governors poll 7.1 points ahead of open-seat candidates, with a standard error of 2.4. Phil Scott, the Republican governor of Vermont, leads by 6.3 points in a state with a lean of D+32.5, about 36 points better than an open-seat Republican would expect there. Kelly Ayotte leads by 11.1 in New Hampshire, which leans D+3.9, about 20 points better. Josh Shapiro leads by 22.6 in Pennsylvania, which leans R+1.0, about 17 points better.

Take out Scott and the governor figure falls to 3.8. Take out all three and it is minus 0.9. The middle incumbent governor runs 1.5 points ahead of expectation. That is new: in earlier cycles, removing the three strongest incumbents still left 3 to 8 points. At the other end, Tina Kotek in Oregon trails Christine Drazan by 2.1 points in our average, in a state with a lean of D+14.7, about 19 points worse than an open-seat Democrat would poll.

What the research says

Political scientists have tracked this decline for a decade. Gary Jacobson, writing in the Journal of Politics in 2015, found that the electoral advantage of House incumbents had fallen to levels not seen since the 1950s, and tied it to rising party loyalty and straight-ticket voting shaped by presidential politics. Joel Sievert of Texas Tech, writing for the Legislative Branch Capacity Working Group in 2018, found the same in the Senate: incumbents who once could expect a 6 to 8 point boost were doing little better than non-incumbents.

Split Ticket measured how closely Senate results follow the presidential vote. The share of the variation in the Republican Senate vote explained by the presidential vote rose from about 22% in 2000 to 93% in 2020. The number of states that voted for different parties for president and Senate fell from 10 in 2000 to none in 2016 and one, Maine, in 2020. In 2024 there were four, Arizona, Michigan, Nevada and Wisconsin, all Democratic Senate wins in states Donald Trump carried. Split Ticket also noted that incumbency and challenger strength go together, because parties rarely field strong candidates against senators they think cannot be beaten. Several 2026 Republicans face unusually well-known opponents, a former senator in Brown and a former member of Congress in Peltola. Our data cannot separate the two effects.

What it means for our forecast, and the limits

Our forecast's fundamentals, fitted on results from 2000 to 2024, credit a Senate incumbent with 9.6 points and a governor with 15.1. The polls now say far less. Because polls carry most of the weight wherever there are several of them, the difference moves our projections by about a point in contested races. In Kansas the fundamentals alone say R+15.0; the polls say D+1.2; our projection is R+0.5.

There are good reasons not to treat the polls as the last word. In 2020 and 2022 the results gave Republican Senate incumbents 3 to 4 points more than the final polls did: 4.6 against 1.7 in 2020, and 10.3 against 6.4 in 2022. Final polls in strongly Republican states have also overstated Democrats in recent cycles. If that repeats, part of the Republican incumbents' deficit is a polling miss rather than lost incumbency. The samples are small, at 14 Senate incumbents, and several races rest on one or two polls, among them Bill Hagerty in Tennessee and Warner in Virginia. A standard error of 2.4 points means the true figure could plausibly be as high as 5. Even that would be in line with 2020 to 2024 and below every year before 2020.

Polling Forecast reads every poll in ten countries, corrects for how pollsters have missed before, and turns them into seats. Our forecasts, the polls and our methods are free to explore.

Our projection and every poll →

Elsewhere on Polling Forecast

Figures as of . Published by Polling Forecast.