Midterms 2026: What the Markets Say (and Where They Disagree With the Polls)

As of August 31, 2026, prediction markets make Democrats heavy favorites to take the House — about 86 cents on the dollar on Kalshi, about 87 on Polymarket — and call Senate control a toss-up that has tilted slightly Republican on Kalshi (Democrats 47%, Republicans 53%) while Polymarket still prints 50–50. The polls point the same direction: generic-ballot averages have Democrats ahead by six to seven points. They part ways on the Senate, on how fast prices absorb news, and — most usefully — on what kind of number each produces. This hub updates weekly through Election Day, November 3. Prices are approximate and move continuously; all figures verified August 31, 2026.

The board, as of August 31, 2026

Market	Kalshi	Polymarket	The polling picture
Democrats win House control	≈86%	≈87%	Generic ballot: Democrats +6.8 (RealClearPolitics average, 48.6–41.8)
Democrats win Senate control	≈47%	≈50%	No single poll analogue; Democrats must net four seats on a 53–47 chamber
Democrats sweep both chambers	—	≈51%	—

The money behind the numbers is real but modest: nearly $200 million traded on midterm outcomes across both venues by mid-August, per an NBC News analysis, with Kalshi’s House board alone past 22 million contracts. For scale, Polymarket’s single 2024 presidential market handled over $3.6 billion. The midterms are an order of magnitude thinner than the last cycle’s main event.

What moved this summer

The House has been the boring end of the board: both venues sit in the mid-to-high 80s for Democrats, within a point or two of each other — about as close to consensus as this industry gets.

The Senate is where summer actually happened. Earlier in the cycle, Kalshi priced Democrats below 20 cents to take the chamber, and Republicans still held a modest edge — mid-50s — into early August. Then, in the second week of August, the market snapped to even. The trigger was Maine: when Democratic nominee Graham Platner suspended his campaign, clearing the way for a replacement, traders repriced Democrats to roughly 65 percent for the seat held by Republican Susan Collins — and chamber control moved to 50-50 with it. No fresh Maine polling existed when the price moved. That is the signature difference: polls have field dates, markets have seconds.

The polls drifted the same way on their own clock: Decision Desk HQ’s generic-ballot average ebbed to roughly four points in July, then recovered to about six by mid-August.

Where the markets and the polls actually disagree

The House isn’t a disagreement — it’s a unit conversion

How can markets say 86 percent when polls only say plus-six? Because they answer different questions. A generic-ballot average measures the national margin; a market prices the probability that the margin lands on the winning side of the seat math. A six-to-seven-point lead is far beyond the House’s tipping point, so converting it into a win probability legitimately produces a number in the 80s. The market isn’t more bullish than the polls — it’s the polls, restated in a different unit.

Polls measure margins. Markets price outcomes. Most of their famous disagreements are unit conversions.

The Senate is the real argument

Here the market says something polls can’t. Democrats need to net four seats. On Kalshi’s seat-by-seat boards, only two Republican-held seats trade as clear Democratic pickups: North Carolina, where Roy Cooper trades around 92 percent against Michael Whatley for the open seat, and Maine at about 65 percent; Jon Ossoff trades near 92 percent to hold Georgia. The priced-in map delivers two flips — yet the chamber still trades near even money (Kalshi 47% Democratic / 53% Republican; Polymarket 50–50).

That gap isn’t an arithmetic error; it’s a bet on correlation. Seats three and four run through states like Texas, Ohio, Iowa or Michigan — races Kalshi’s boards still list as genuinely in play — and the environments that deliver them tip several races at once. Polymarket’s balance-of-power board makes the same statement from another angle: a full Democratic sweep trades around 51 percent, nearly the entire Senate probability, because any world where Democrats take the Senate is one where they’ve already taken the House. Race-by-race polling averages can’t express that if-one-then-many structure. Markets price it natively.

Speed is the third difference

Maine repriced within hours of the Platner news, ahead of any polling. That cuts both ways: markets absorb real information instantly, and rumors just as fast. “It’s the wild west right now,” Decision Desk HQ chief elections analyst Geoffrey Skelley told The Hill — and the wildness lives mostly in the thin, fast end of the board, not the deep chamber-control markets.

How to read an election price

A refresher, because midterm markets punish sloppy reading. A contract at 86 cents claims the event happens about 86 percent of the time — the market fully expects the other outcome one cycle in seven. An 86 percent favorite losing isn’t the market being “wrong”; it’s the 14 percent happening. The full mechanics — spreads, why yes and no don’t quite sum to a dollar, what fees do to the math — are in how to read a prediction market price, and the instrument itself in what is a prediction market.

Two midterm-specific habits: check volume before you believe a number, because a thinly traded seat market can be pushed around in a way the chamber markets can’t; and treat the chamber price, not any single race, as the market’s considered opinion. None of this is a recommendation to trade — it’s how to read a public forecast that now moves faster than the people covering it.

What happened in 2024, the reference case

The reason anyone takes these prices seriously is the last cycle. Through late October 2024, Polymarket priced Donald Trump in the 60s while national polling averages sat near even — a genuine, loud divergence, not a unit conversion. A single French trader, “Théo,” had staked more than $28 million on Trump across four accounts, CNBC reported, and for two weeks nobody could say whether the price was information or one man’s opinion. Then Trump swept all seven swing states, the trader collected roughly $50 million, and the markets-beat-the-polls narrative was born.

The honest version is more modest. The whale wasn’t manipulating; he was acting on a thesis — polls were underestimating Trump again — that happened to be right. One cycle is one data point, and pollsters push back: Ipsos public affairs president Cliff Young called betting markets “about as good as the polls.” What 2024 settled is legality and relevance — Kalshi beat the CFTC in federal court that fall — and 2026 is the first full cycle where regulated U.S. election markets run from the start.

The worry: manipulation, insiders, and election night

Election officials spent the summer preparing for the downside. Their concerns, laid out in a Votebeat report, run in three directions. Trust: Los Angeles County registrar Dean Logan described early returns in LA’s mayoral primary drawing suspicion about routine ballot processing because they diverged from what markets had priced. Insiders: some officials know results before the public does — Delaware County, Pennsylvania’s election director Jim Allen barred his poll workers from the markets, comparing the alternative to “referees betting on the outcome of a basketball game.” And incentives: money riding on outcomes creates reasons to interfere — or to appear to.

The institutions are responding in real time. The Senate voted unanimously in April to bar its own members and staff from trading on prediction markets. Rep. Ritchie Torres introduced a bill to ban federal officials holding nonpublic information from trading event contracts, after what his office described as a suspicious $400,000 Polymarket position opened just before a U.S. operation against Venezuela’s Maduro government became public; the Congressional Research Service has examined how insider-trading law applies to event contracts at all. Kalshi says it has flagged manipulation and insider-trading cases to the CFTC and suspended three political candidates — including one Senate hopeful — for trading on their own races. And Wisconsin officials issued an advisory citing a 19th-century state law against voters betting on races they vote in, which a Kalshi staffer publicly called “active voter suppression.” Every one of these fights is live; none is resolved.

What to watch through November 3

• The Senate chamber price versus the seat prices. Either the second-tier races start trading closer to toss-ups, or the 50-50 chamber price drifts back toward Republicans. The boards can’t stay in tension forever.
• The generic ballot against the House price. They agree today. If polls tighten while the market holds in the 80s — or the reverse — that’s a 2024-style divergence worth taking seriously.
• Candidate shocks. Maine just showed a single withdrawal can reprice a chamber. Health, scandal, and replacement fights are what markets react to fastest.
• Election night itself. Markets will effectively “call” races off partial counts, hours before officials do. Where mail ballots count slowly, expect price-versus-count gaps — and expect them to be portrayed as evidence of something. Officials are bracing for exactly that.

Quick answers

What are the 2026 midterm odds right now? As of August 31, 2026: Democrats about 86 percent on Kalshi and about 88–89 percent on Polymarket to win the House, and Senate control a toss-up — 47 percent Democratic on Kalshi, 50 percent on Polymarket. Prices move continuously; this hub updates weekly.

Why do markets say 86 percent when polls only show a 6–7 point lead? Different units. Polls estimate the national vote margin; markets price the probability that the margin produces a majority. Six to seven points is comfortably past the House tipping point, so a probability in the 80s is consistent with the polling.

Are prediction markets more accurate than polls? They beat the polling consensus in 2024, pricing Trump far above even when averages were tied — and he won. But that’s one cycle, and pollsters note markets have historically been about as good as polls, not clearly better. The fair reading: a market is a fast, money-weighted summary of all available information, polls included.

Is it legal to trade on the midterms? Contracts on CFTC-regulated exchanges have been federally permitted since Kalshi’s 2024 court win; Polymarket operates in the U.S. through a licensed venue. Restrictions target specific people: senators and their staff are banned by Senate rule, and some counties bar poll workers. Whether this counts as investing or gambling is its own question; nothing here is a recommendation to trade.

Could someone manipulate the midterm odds? Thinner markets are easier to move — roughly $200 million on the midterms so far versus $3.6 billion on Polymarket’s 2024 presidential market alone. The 2024 whale proved one trader can dominate a price — he turned out to be informed, but nobody knew until the votes were counted. Deep chamber markets are hard to distort for long; thin seat markets are another story.

Last verified August 31, 2026.