Prediction markets and polls generally matched across most key races CoinDesk tracked in the 2026 midterms, and both missed a few. Races in Mississippi and South Carolina drew little data from either.
Across most of the key races CoinDesk tracked, forecasts from both sources also generally matched the eventual results, with a few exceptions. After the election, a debate on X asked whether platforms like Kalshi and Polymarket are more accurate than polling models.
Three races that departed from the forecasts
In Wisconsin's First House District, Rep. Bryan Steil lost to Democrat Mitchell Berman. Polls and prediction markets both had that race in a dead heat the Friday before.
In Alabama's Second House District, Rep. Shomari Figures defeated challenger Rhett Marques. Both polls and prediction markets showed Marques leading in the redrawn district.
The Iowa Senate race split the two tools. State Rep. Josh Turek defeated Rep. Ashley Hinson and flipped the seat, even though prediction markets had given Hinson the edge and averaged polls showed a 50/50 response.
Thin data on Mississippi and South Carolina
Other races showed up little in either polling or prediction markets. The Polymarket contract for the Mississippi Senate race favored incumbent Republican Cindy Hyde-Smith until vote results started coming in on Tuesday night, and Kalshi had no contract for that contest.
Neither Polymarket nor Kalshi had a contract on South Carolina's Senate race, though Polymarket did offer a parlay on the winners of the Montana, South Carolina and Mississippi Senate races. Spokespeople for both companies did not respond to questions about these races.
Both races became upsets as Democrats Scott Colom and Annie Andrews won their seats. Polls offered little more help: according to The New York Times' archive, the most recent South Carolina race had Senator Graham up by 1, and the Times page listed no Mississippi polls.
Source: CoinDesk
Trading involves risk.
Risk Disclaimer
