US Midterms 2026

The US Midterm elections are upon us, and advocates for both major parties are framing this election as one which will enable or inhibit the Trump agenda for his remaining two years in the presidency.

House Rule

Analysis across the board expects the Democrats to re-take the house. My analysis unsurprisingly agrees. There are 435 voting members of the house, and the Republicans hold the balance here by a razor-thin margin of 215:210. If Democrats can flip three seats from their 2024 result, they’ll hold the majority. Several factors playing against the Democrats in 2024 (grievances against the Biden government, rapid installation of Harris without a primary or sufficient campaign time etc.) are no longer in play; conversely grievances against the Republicans have exploded as a result of the war in Iran, inflation and gas prices, perceptions of corruption and authoritarian-style government overreach among myriad other issues. At this stage if the Republicans only lost three seats, it would be a cause for celebration among the party.

In the analysis below, I will also include the non-voting delegates in the house representing American Samoa, D.C., Guam, the Northern Mariana Islands, the US Virgin Islands and Puerto Rico – although Puerto Rico’s Democrat representative is not actually up for election in 2026, so that’s already one seat of 441 secured on the board for the Dems.

Some other races are also foregone conclusions where a representative is running unopposed or one of the major parties is not contesting the seat at all. This lines up a further 18 seats for the Ds (Arizona 3; California 4, 7, 11, 12, 14, 29, 34 and 37; Florida 10; Massachusetts 1, 2, 5 and 7; New Jersey 8; Pennsylvania 3; Wisconsin 2; and the non-voting seat of US Virgina Islands) and two for the Rs (California 40 and non-voting Northern Mariana Islands).

Given the heavy fall in Republican support, I am comfortable also assuming no blue seats are turning red, except where these have been caught up as part of the controversial mid-decade redistricting of seats as part of a blatant Gerrymandering campaign by both parties.

Finally, in this first-pass prediction, I am assuming certain die-hard Republican seats are unwinnable for Democrats regardless of the political climate, but that Dems will win all typically Democrat seats. Using the Cook Partisan Voting Index, which measures each district’s deviation from the national voting average, I am assuming any seat with a Republican lean of R+15 or greater will vote red, while any Democrat lean (D+1 or higher) will vote blue. The threshold of R+15 was determined by (A) taking the the polled national vote split 43:52 which reveals a 9-point gap, (B) multiplying this by 150% for additional safety (rounded to 14) and (C) adding one. In other words, because the national vote has a 9-point lean to the Democrats, seats 9 points to the right of the national average are considered boarder-line and in-play, and a district needs a lean of over one-and-a-half times this to still be considered safe for Republicans.

Combining the foregone conclusions, re-election of all Democrats and the awarding of safe Republican districts, the first prediction looks as follows:

Prediction 1: Republican 74+1, Democrat 209+3, Undecided 152+2

One immediate peculiarity here is Florida – a strongly Republican state which has only one ‘R’ so far, and four ‘D’s. I alluded to the rash of out-of-season redistricting attempts ahead of this election; Florida was one of the states where this occurred. Because winning a district only requires one more than 50% of the vote, and winning 51% is as good as winning 99%, the goal in Gerrymandering is to spread your own supporters out to secure small margins of victory in many seats, while consolidating your opponents in a few incredibly strong seats. This almost certainly accounts for the many Florida seats falling under the R+15 threshold.

In the states where courts have allowed redistricting since 2024, I’ve assumed that the Gerrymandering in Republican controlled states was conducted competently to preserve all Republican incumbents (and new red seats), even if these maps were drawn prior to a substantial downtick in Republican support.

Pro-Democrat Gerrymandering has also occurred (e.g. in California) but because any seat with a nominal D+1 Cook Partisan Voting Index has already been given to the Dems in our prediction, the only seats that we are adding on at this stage are Republican – 64 in all (Alabama 2; Florida 2-8, 11-13, 15-19, 21, 22 and 26-28; North Carolina 3, 5-11, 13 and 14; Ohio 5, 7, 8 and 10; Tennessee 4-8; Texas 2-6, 8-10, 12, 14, 15, 17, 21-27, 31-32, 35, 36 and 38).

Prediction 2: Republican 138+1, Democrat 209+3, Undecided 88+2

This gives us a "short"-list of 90 districts for more granular analysis. Polling is available for many of these, aggregated here. Where I have subjectively considered the polling to favour one candidate or another, I’ve marked the board accordingly below. (This is necessarily subjective, because it involves weighing polls not only on how many found Republican or Democrats to be leading, but also by how strong the lead is and how recent these polls are). Republican-favouring polls from before the start of the Iran war (and subsequent collapse in Republican support) are given little weight. Some seats with only one poll might still be given to one or the other party if the poll is recent enough (especially given we’re running out of meaningful other data); other districts with two or three polls all from last year may be ignored.

From my judgement of the polls, Republicans pick up 14 more seats (Alaska at-large; Colorado 4 and 8; Indiana 5; Kansas 2; Minnesota 8; New Jersey 2; New York 21 and 23; Oklahoma 5; South Carolina 1; Tennessee 9; Virginia 5 and 6) and Democrats 24 (Arizona 1 and 6; California 6; Colorado 5; Florida 14 and 25; Iowa 1 and 2; Kentucky 6; Michigan 4 and 7; Minnesota 1; Missouri 2; Montana 1; New Jersey 7; Ohio 1 and 15; Pennsylvania 7, 8 and 10; Texas 28 and 34; Virginia 1 and 2).

There was some obvious overlap between the states for which polling was available and those we had not called yet. That said, some uncalled races had no polling for the general election. Perhaps naively, I assume the fact that no-one is watching those races closely is that they are considered pretty clear cut, which would favour the Republican incumbents in those 32 seats (Arizona 5 and 8; Arkansas 3; California 5; Georgia 1, 10 and 12; Illinois 16; Indiana 2; Kansas 4; Michigan 1 and 5; Minnesota 6; Mississippi 3; Missouri 3; New Jersey 4; New York 2, 11 and 24; Ohio 14; Oklahoma 1; Oregon 2; Pennsylvania 11 and 16; South Carolina 2, 4 and 5; Washington 4 and 5; Wisconsin 5, 7 and 8). Adding these in, along with a safe bet of a Republican returned in American Samoa and a Democrat in Guam, we get an updated prediction as follows: 

Prediction 3: Republican 185+2, Democrat 234+4, Undecided 16

And that’s really where the “serious” House prediction ends – but for the fun of calling every tossup, let’s assume the remaining 16 Electorates split evenly and sort them based on their Cook Partisan Voting Index:

Conclusion

Giving a final prediction of:

• Republicans 193 + American Samoa and Northern Mariana Islands
• Democrats 242 + D.C., Guam, US Virgin Islands and the continuing term in Puerto Rico

Send in the Senate

This is where things get interesting – while virtually every credible source has the Dems easily winning back the House, the Senate is a lot closer. In 2026 we have 35 senators up for election – 33 at the expiry of their six-year term and two special elections to replace Senators promoted to Trumps executive branch (to the roles of Vice President and Secretary of State).

The remaining 65 seats are shared between 31 Republicans, 32 Democrats and 2 Independents who caucus with the Democrats and, for balance of power purposes, are treated as Dems below.

The Republicans face a nominal disadvantage on this make-up, needing 20 seats to retain power vs the Dems’ 16. The Republicans currently hold 22 of the seats up for election, meaning they can only lose two and maintain an outright majority, or lose three and rely on their Vice-Presidential tie-break.

While the Republicans have a lot to lose in these elections (control of the Senate in addition to near-certain loss of the House), the stakes are also high for the Democrats. Despite consistently close polling in many races making this neck-and-neck, there is an expectation that the Democrats can and will win both houses – failure to do so will appear as a damning indictment regardless of how well they perform in absolute terms.

First let’s look at the 2026 senate races, 27 of which have been consistently rated across twelve separate publications as at least leaning to one party or another (14 to Republican, 13 to Democrat):

To this I would comfortably add South Carolina for the Republicans, as only Fox and RCP considered this a tossup, and Michigan for Dems, which has four publications considering it a tossup but I just don’t see a seat won by Dems in 2020’s Trump-returning-victory year falling to the Reps in 2026 given… everything.

That gives us six key races (Alaska, Iowa, Kansas, Maine, Ohio and Texas) of which the Republicans need to hold four to manage a tie in the Senate.

Alaska

In Alaska, polling consistently (though marginally) favours Democrat Mary Peltola, including every ranked choice poll (which is how the election will be conducted) as well as eleven of the fifteen first past the post polls since the primaries. (The remaining four include two ties, and two Republican wins, all from Republican-favouring polling houses.)

On this basis, Alaska seems like a Democrat win.

Iowa

In Iowa the polling is evenly divided. Two polling aggregators favour Republican Ashley Hinson and two favour Democrat Josh Turek. Eighteen individual polls favour the Republicans, and 18 favour the Democrats – though two Republican leaning polls were conducted before the primaries versus only one for the Dems.

This could easily go either way, but with a higher hill to climb and a Republican currently on-seat I’m nominally turning this one red.

Kansas

Despite not having won a Kansas senate race in almost a century, the Democrats lead the polling this year.

Kansas will be close, but I’m tipping the Democrats here.

Maine

Democrats are dominating the polling in Maine…

Ohio

… and in Ohio…

Texas

… and in Texas.

Conclusion

While US politics have taught me more than once to not trust blindly in polls, that brings us to 16 Republican wins and 19 Democrats, enough to flip the Senate into Democrat control. Even if some of these go against the polls, there is still a solid path to victory for the Dems.