House Control Odds 2026: Who Wins, And When It Gets Called

Kalshi has Democrats at about 85% to win control of the House in the 2026 midterms, with Republicans near 15%. That is a heavy favorite against a party that actually holds the chamber, and the price rests on two facts rather than on any poll: the president's party almost always loses House seats in a midterm, and the Republican majority is thin enough that a net gain of three seats flips it. Kalshi also runs a second board on the same election, pricing what night the Associated Press calls it, and that is where the market and our model panel disagree by nearly 50 points.

The Board

Event CONTROLH-2026, pulled from the live Kalshi book on August 12, 2026. A Yes contract pays $1.00 if the outcome happens, so the price in cents reads as an implied probability. The contract settles early on a consensus of media calls, otherwise on the party of the Speaker of the House on February 1, 2027. The cheaper Republican side has traded more contracts than the favorite; that shows where traders are active, not which way the crowd leans, since every open contract has a Yes holder and a No holder. Not a recommendation to trade.

Why Democrats Are Priced So High

The president's party has lost House seats in all but two midterms since World War II, and the recent editions have not been gentle.

The row that matters is 2022, the smallest loss on the table, because even that mild result would be more than enough this cycle. Republicans, the party holding the White House under President Trump, came out of the 2024 elections defending a certified 220-215 majority, one of the thinnest in modern history, so Democrats need a net gain of three seats to control the chamber.

In every postwar midterm where the president's party lost ground, it lost at least four seats. That base rate, applied to a majority with almost no cushion, produces a big Democratic number before anyone reads a district poll. Resignations and vacancies have nudged the live count since 2024, but the flip math runs off the elected baseline.

The polling points the same direction. Public generic-ballot averages in late July and early August 2026 mostly sat between Democrats plus-4 and plus-6, with at least one likely-voter average closer to plus-8, depending on the aggregator and its screen. That range sits between 2022, when Republicans won the national House vote narrowly and squeaked out a majority, and 2018, when Democrats won it by about 8.6 points and gained about 40 seats.

What Could Pull The Price Down

The one thing a national poll cannot see is that the district map moved mid-decade. Texas redrew its congressional lines in 2025 with the stated aim of adding Republican-leaning seats, California voters answered that November by passing Proposition 50 to add Democratic-leaning ones, and litigation over both maps and their imitators is still running.

A redrawn map changes how efficiently a national margin converts into seats. District-by-district raters like Cook Political Report and Inside Elections still score only a few dozen of the 435 seats as competitive, so a court ruling that redraws three to five likely seats moves this market more than a one-point generic-ballot wobble ever could.

No seat-level model we would put our name on is on this page yet; when our panel prices this market, the verdict will sit next to the live number. My own read, stated as a read: the base rate has earned most of that 85 cents, but a mid-decade map shuffle is exactly the kind of variance the postwar table has never had to price, and I would want wider tails than a one-cent spread implies.

When Does The House Get Called?

Everything above prices which party ends up holding 218 seats. Kalshi's second board on the same election asks when the Associated Press makes its first definitive call of a House majority, and for anyone planning to be awake on November 3 it may be the more useful of the two. Eight contracts, one clock. Each rung asks whether the call lands before its own deadline, so every later rung contains every earlier one; the numbers are a running total, and adding them together produces a figure that means nothing.

The record it has to price cuts both ways. The last time the House was settled on election night was 2018, when ABC News's live coverage timestamped its projection at 10:58 p.m. ET, two minutes short of this board's first deadline. In the three federal elections since, AP has needed a week or more.

Event KXAPCALLHOUSE-26AUG20, pulled from the live Kalshi book on August 24, 2026, at 5:45 a.m. ET. Panel figures are the median of eight models' estimates, each produced without ever seeing a price; the panel's card carried no 2026 polling and no verified race-ratings table, so treat its margin views as arguments and the call-date record below as the evidence. The two longest rungs have never traded, so their quotes are market-maker screens rather than transaction evidence.oddsshopper.com/articles/prediction-markets/model-verdict-scoreboard) once the market settles.

Read plainly, the board says there is a better-than-seven-in-ten chance the House is called before dawn on November 4, and only a 13.5% chance the question is still open a week later. The panel puts the overnight call at 28% and its central date a week out, and every rung except the last is priced above what the panel thinks it is worth.

Why The Panel Sits A Week Behind The Board

The disagreement is about the size of the win. A wire service can call a chamber early only when the races it has already called, plus the races where the outstanding ballots cannot change the leader, add up to 218 on their own. In a 40-seat wave that arithmetic clears while polls are barely shut. In a two-seat result the last races to 218 are, by construction, the closest in the country, and the closest races need every ballot counted.

Sources, in order: ABC News's 2018 live coverage, PBS NewsHour reporting AP's 2020 call, NPR on 2022, and Al Jazeera on 2024. This board settles on AP specifically; the 2018 minute is ABC News's, the only one we could source, and the three later dates are AP's own.

The one election-night call is also the one blowout. Then there is the part that is written law. California counts a mail ballot if it is postmarked on or before Election Day and arrives within seven days, and its counties run a 30-day canvass, per the Secretary of State. A close California race cannot be resolved on election night, because ballots that will legally count have not arrived yet.

That window is recent. AB 37, signed in September 2021, replaced a three-day deadline, so it has run exactly two federal elections, and in both AP called the House the day after California stopped receiving mail.

The third cycle keeps this honest. In 2020 an emergency measure stretched the window to 17 days, and AP called the House on November 10 anyway, because by then the outstanding vote could no longer change 218. AP waits for the arithmetic to settle; the deadline only binds when a close race keeps it in doubt.

Under today's law the equivalent date for the 2026 election is November 10. That is where the panel's central estimate sits, and it is why nearly every model named California without being asked.

Arizona used to count the same way. Through 2024 every early ballot's signature had to be verified before the envelope was opened, voters got a five-day window to cure a questionable signature, and ballots dropped at the polls on Election Day, up to 350,000 in Maricopa County alone, waited until afterwards.

Arizona changed that for this cycle. Under legislation passed in 2024 and in effect for the 2026 midterms, a voter dropping a mail ballot at a voting location can show a driver's licence or passport instead; if the name and address match the registration record, the ballot bypasses signature verification. It does not alter who may vote or when ballots are due. Votebeat walked through the new process before it was first used.

It has now been used. In Arizona's August 2026 primary roughly 23,000 Maricopa County voters, about 30% of those who dropped a ballot on Election Day, took the ID option, and the county finished counting its mail ballots markedly faster than usual. That is one primary of evidence, and it is the strongest argument on the market's side of this board: one of the two states that produced the deciding seat in the last two cycles has legislated itself faster, and the panel was handed that fact before it answered.

Which seats decide it? The most reliable evidence is which seat actually reached 218. In 2022 it was California's 27th, Mike Garcia's, called on November 16. In 2024 it was Arizona's 6th, Juan Ciscomani's, called on November 13. Both sit in the two slowest-counting large states, and one of them has now rewritten the rule that made it slow.

AP Is Not A Neutral Clock

One more input belongs in the open. At 2:50 a.m. ET on November 4, 2020, AP joined Fox News in calling Arizona for Joe Biden while every other major outlet held off, and the Trump campaign attacked the projection publicly, as Al Jazeera reported at the time and the Washington Post revisited four years later. The call held; Arizona finished 10,457 votes apart, the closest presidential result in the state's history.

The tempting inference, that AP was burned once and has been slower ever since, is the one to resist. The 2022 and 2024 dates are explained just as well by two- and four-seat margins as by institutional caution, and most of the panel priced the margin rather than AP's temperament.

Where The Panel Changed Its Mind

The revision round, in which each model read the other seven anonymised and revised, produced two discoveries, both from models reading clocks and calendars rather than politics.

The first concerns the first rung. California's polls close at 11 p.m. Eastern, per the Secretary of State, the exact minute of that rung's deadline. Democrats won 43 of California's 52 districts in 2024, and none of them can enter a tally before those polls shut. So the first contract is really asking whether AP can reach 218 out of eastern and central seats alone.

"Pacific polls close at 11:00 PM ET so no West Coast seat can enter AP's tally before rung 1 expires, meaning a sub-11PM call requires clearing 218 out of eastern/central seats alone... a 2018-size wave more likely resolves between 11PM and 1AM once the safe Pacific seats flood the tally than literally before 11PM..." — Kimi

The second is a calendar detail with money attached. The pattern in the last two cycles is a call landing the day after California's receipt deadline. In 2026 that deadline is November 10, which sits inside the "before Nov 11" rung; the day after it does not. Several models had been treating a week-long call as one bucket and discovered mid-round that the board splits it.

Between them the two arguments pulled the extremes toward each other, and the result was convergence rather than drift: the panel's spread on the Nov 11 rung tightened from 47-87 to 63-78 with the median almost unmoved.

What Would Change The Panel's Mind

• The generic ballot settling at Democrats plus-8 or better this autumn, or a run of large special-election overperformances. That is wave territory, the one condition under which the overnight rungs look cheap rather than dear.
• Arizona's ID opt-in rate in the general. The primary ran at about 30%; a materially higher November figure pulls the close-race branch a day or two earlier, and the counties publish it.
• Where the tipping-point seat sits. Late ratings putting it in a fast-counting eastern or midwestern state pull the call earlier; a California Proposition 50 seat or an Arizona toss-up pushes it later.
• Final rulings in the Texas and California map litigation, or a recount in a decisive district. A post-election contest is the only realistic route to the rungs asking whether the House is still uncalled in December.

Election Day is Tuesday, November 3, 2026. California's mail-ballot window closes November 10, all eight timing contracts expire January 8, 2027, and the panel is re-scored when one of the triggers above fires rather than on a calendar. Who-wins prices, volume and open interest fetched from the live Kalshi book on August 12, 2026, event CONTROLH-2026; the call-timing board above, event KXAPCALLHOUSE-26AUG20, was pulled separately on August 24, 2026. Polling figures reference public generic-ballot averages as of early August 2026.

More on this: California Governor Election Odds · Texas Senate Election Odds · North Carolina Senate Election Odds · 2028 Presidential Election Odds · Justice Alito Retirement Odds