Midterm contracts cross $750 million as polling faces a stress test

Sometime in the summer of 2024, a television commentator watched Polymarket's numbers hold through election night and told a Fox Business audience that the platform had been right the whole time. That moment — more than the court ruling, more than the volume figures — is what changed the conversation. Credibility, once conferred publicly, compounds.

Two years on, the compounding is visible in the numbers. Notional volume on the 2026 midterm contracts had crossed $750 million by mid-October, against $92.4 million traded on all non-presidential contracts in 2024. The Anti-Corruption Data Collective puts active midterm offerings on Kalshi and Polymarket at more than 21,000. If the trajectory holds through election day, total volume on these races could reach $1.6 billion.

That is not just growth. It is a structural shift in where political information prices itself.

The consensus read is that this validates the 2024 story: prediction markets called it, the polls missed it, and now capital is following the track record. I don't think that's where this lands. The 2024 presidential race was a single high-salience event with enormous liquidity and a motivated base of sophisticated traders. What's being tested now is whether that same information quality survives when you distribute it across 21,000 contracts on state legislative races, county-level contests, and House seats that most national traders could not place on a map.

Thin markets do not aggregate wisdom. They aggregate the views of whoever is willing to trade on a contract that moves two dollars a day. I have watched this pattern before — a flagship market performs, and the lesson drawn is that the methodology scales. It does not always scale. The flagship performance reflects the flagship's liquidity, its trader base, and the volume of public information available on that one event. Strip those conditions away and you have a different instrument.

The S&P Global Kagan survey released this quarter puts the problem in useful relief. Seventy percent of prediction market users traded sports contracts. Thirty-three percent traded political outcomes. The audience that validated these platforms in 2024 is primarily a sports-betting audience that wandered into elections, not a political-information audience that found a better price-discovery tool. Whether that audience produces reliable signals on a Montana state senate race is a question the volume figures do not answer.

James Surowiecki, quoted in the Netflix documentary that has apparently become the industry's preferred origin myth, called these midterms "one of the biggest tests for prediction markets." He is right, but probably not in the way the promotional context implied. The test is not whether volume is large. The test is whether the markets on low-information, low-liquidity races outperform the polling aggregates that prediction market advocates spent two years dismissing. That result will not be known until November, and the answer will be legible only to people who are looking for it carefully rather than for a headline that confirms what they already believe.