Where the Money Disagrees: Markets, Polls and Models on Election Day

Some of that discount is scar tissue. Reuters' Morning Bid noted traders are wary after the Brazil miss, when heavily traded markets got a major election wrong. Set against that, academic work keeps finding the markets well calibrated at scale: a 2025 Vanderbilt study of nearly 2,500 markets on the 2024 election is among the research Reuters cited this morning. Tonight is a live rematch between those two findings, with more than $1.48 billion riding on the outcome.

District markets show the divergence in miniature. Vantage Predictions' Election Day snapshot, comparing this morning with Monday afternoon, ranked Pennsylvania's 1st district as the biggest overnight move on the House board: Democrat Harvie rose from 55 percent to 64.5 percent against Republican incumbent Brian Fitzpatrick, on the average of Kalshi and Polymarket prices, while Colorado's 3rd moved the other way, with Republican Jeff Hurd slipping from 62.5 percent to 55.8 percent. These are thin markets, and Vantage itself cautions that a single large order can move them several points. A nine-and-a-half-point overnight move is either information arriving or one trader's conviction, and the price alone cannot tell you which.

South Carolina's Senate race is the opposite case: the two venues agree with each other exactly and disagree with the polling. Graham is priced at 73 cents on both Polymarket and Kalshi, with Andrews at 28 cents on both, an exact cross-venue match, while recent polls run from a one-point Graham edge to a 48-48 tie in the New York Times/Siena survey (Action Network). A tied poll and a 73-cent price can both be honest: the market is pricing incumbency, fundraising and the state's lean on top of the survey snapshot. If the surveys are right and the race is truly even, 73 cents is the most expensive favorite on tonight's board. If the market is right, the polls missed the state's shape.

Alaska offers a third kind of gap, model against market. Sikt's AI forecast puts Mary Peltola at 53 percent against a market average of 72 percent (Sikt Intelligence). One of them is wrong by nearly twenty points, and tonight will grade both. Kansas, meanwhile, produced the week's biggest repricing: Democrat Adam Hamilton climbed to a record 63 percent on Kalshi in the Senate race, up 23 points in a week (NextPredict), a move large enough that either the race genuinely broke open or the market spent a week catching up to itself. Florida's governor race moved the other way, with Byron Donalds sliding 15 points in a week to 64 percent on Kalshi (DeFi Rate).

For all the disagreement at the edges, the venues rarely disagree with each other. Across 54 matched outcomes, Kalshi and Polymarket prices average a 1.2-point gap (PolymarketTrader). The real argument is not venue against venue. It is the combined market against the polls and models, and it is concentrated exactly where the money is thinnest: the Anti-Corruption Data Collective found 96 percent of congressional markets could be shifted five cents for $3,500 or less (Gambling Insider). Deep control markets are hard to move and have earned their calibration record. Thin district markets are easy to move and have not. Election night will test both kinds at once, and the honest read of today's divergences is that they live almost entirely in the second kind.