The Verdict on the 'First Prediction-Market Election': Markets Called the Democratic Sweep but Underestimated It
Democrats won control of both chambers of Congress in the November 3 midterms, an outcome Kalshi and Polymarket had priced as the favourite. Both platforms claimed roughly 98% directional accuracy across resolved races. But the headline figures carry caveats: much of that accuracy came from non-competitive races, independent reviewers judged the markets roughly comparable to conventional forecasts, and analysts noted the markets underestimated the size of the Democratic wave. The New York Times found prediction markets fared about the same as election forecasts; one economist gave them a 'B.' Kalshi's midterm volume reached $2.5 billion, nearly double the 2024 cycle.
Democrats won control of both the Senate and the House of Representatives in the November 3, 2026 US midterm elections, delivering the outcome prediction markets had favoured throughout the autumn. On the eve of the vote, Kalshi gave Democrats a 74% chance of winning the Senate and 94% for the House; Polymarket priced the Senate at 73%. The markets had been directionally correct for months, moving from a Republican-favoured position when the 2026 cycle opened in late 2024 to a solidly Democratic one by autumn, driven by rising energy prices and declining presidential approval. The result validated the markets' top-line forecast: they called the Democratic sweep, earlier and more confidently than much traditional commentary.
Both platforms made strong accuracy claims. Kalshi said its favoured candidates won or led in roughly 98% of resolved races; Polymarket put its figure at 98.4% of settled contests, measured against the day before the election. Earlier in the count Kalshi's figures were lower — around 95% for House races and 94% for Senate races — rising as more races were called. For a sector defending its legitimacy against state enforcement and arguing its products serve a genuine informational purpose, a successful midterm forecast is valuable evidence. The platforms have used the numbers to argue that prediction markets aggregate information more effectively than polls or pundits.
The headline figures require context that complicates that framing. Much of the 98% accuracy came from non-competitive races that any method — including a simple partisan-lean model — would have called correctly. The meaningful test is how well a tool prices the close races, and there the markets' performance was more ordinary. The New York Times found prediction markets fared about the same as conventional forecasts through most of the campaign, and an economist who studies them gave a grade of 'B.' More pointedly, analysts noted the markets underestimated the size of the Democratic wave: they correctly identified Democrats as favourites but priced a narrower result than materialised, missing the magnitude even as they got the direction right. A forecast that calls the winner but misprices the margin is useful but not oracular.
The accuracy debate bears directly on the sector's central legitimacy argument. The platforms contend prediction markets are valuable because they produce accurate, real-time probability estimates that serve the public interest — an argument distinguishing them from gambling and supporting federal financial regulation over state gambling regulation. If the markets are genuinely more accurate than alternatives, that argument is strong; if they are roughly comparable, the informational-value claim weakens and the markets look more like a betting product with a forecasting veneer. The 2026 midterms, the first major US election conducted with prediction markets at full scale, produced a mixed verdict: the markets worked, but not dramatically better than what existed. Kalshi's midterm volume reached $2.5 billion, nearly double the 2024 cycle's $1.3 billion, and Polymarket recorded nearly $700 million — confirming the commercial success of election markets even as the forecasting-superiority claim got only partial support. The 'first prediction-market election' showed the markets are accurate enough to be taken seriously and popular enough to be commercially significant, but not so superior to conventional forecasting that the informational-value argument settles the regulatory debate on its own.
Operators mentioned in this article
Recent updates
Kalshi's Parlays Drive Back-to-Back $5B Days — and Hand Critics Their Strongest Argument
Kalshi traded a record $5.33 billion on November 7 and $5.36 billion on November 8, with multi-leg parlay-style 'combos' making up roughly 75% of both days' volume. Its 30-day volume topped $100 billion. But the parlay dominance cuts against the platform's legal defence: parlays are a quintessential sportsbook product, and the American Gaming Association has told the Second Circuit that parlays make up about two-thirds of Kalshi's volume and should be treated as sports bets. The implied house edge on 10-plus-leg combos is around 25%, undercutting the argument that these are neutral financial instruments rather than gambling.
Cruz Bill Would Require State or Tribal Approval for Sports Contracts — a Bipartisan Legislative Threat Emerges
Senator Ted Cruz, who chairs the Senate Commerce Committee, is preparing a bill that would bar federally regulated prediction markets from offering gambling-tied event contracts, including sports, unless a state or tribal government authorises them. It would not ban the contracts nationwide but would require platforms to obtain approval jurisdiction by jurisdiction, resembling state-by-state sportsbook regulation. The proposal arrives as Democrats take control of both chambers after the midterms, creating a potential bipartisan reckoning for a sector that had relied on federal protection.
CFTC Tells Platforms to Stop Geofencing States — Directly Colliding With State Court Orders
The CFTC issued an interpretation on November 5 of its 'impartial access' core principle, telling regulated exchanges including Kalshi and Polymarket to drop state-by-state geographic blocks within 30 days or explain why they cannot. The demand collides head-on with existing state court orders requiring the platforms to geofence: Washington ordered Kalshi to geofence in August, and Michigan ordered it to use licensed geolocation to block the state in September, with a $500,000 daily fine for violations. Chairman Michael Selig framed the interpretation as a uniformity requirement. The platforms are now caught between a federal demand to unblock states and state court orders to block them.