Prediction markets and polls aligned in Ohio's 2026 midterms, but timing differed
Kalshi and traditional forecasting models correctly identified winners in three key Ohio races, though the markets showed higher confidence earlier in the campaign cycle.
By Ricardo Alvarez ·
Traditional polling and prediction markets reached the same conclusions about the winners in Ohio’s three most competitive races during the November midterm elections, but they arrived at those answers on different timelines.
The divergence in timing was most visible in the U.S. Senate race between Democrat Sherrod Brown and Republican Jon Husted. Kalshi, a major prediction platform, showed Brown with a 51.1 percent chance of victory on Sept. 10. That figure climbed steadily to 70.3 percent by Nov. 1, marking a gain of nearly 20 points.
In contrast, the forecasting model from Decision Desk HQ never assigned Brown a higher probability of winning than 57 percent during that same period. RealClearPolling data indicated that Brown was the favorite in eight of the final ten polls conducted before Election Day.
The race for Ohio’s 7th Congressional District showed a wider gap in when the models flipped. Incumbent Republican Rep. Max Miller was favored by Kalshi in late July, with a 63.5 percent chance of reelection. His odds dropped after a Mother Jones report detailed domestic violence allegations made by his ex-wife, Emily Moreno, in May. The story received widespread media coverage at the end of July.
Kalshi listed Democrat Brian Poindexter as the favorite for the first time on July 29. He maintained that lead for the rest of the campaign, with his winning probability peaking at 89 percent on Oct. 23. Decision Desk HQ continued to favor Miller until Oct. 21, when it reached a 50-50 split. The model did not name Poindexter the favorite until Oct. 26, at which point Kalshi already gave him an 86.1 percent chance of winning by eight points.
In the gubernatorial race, the two systems mirrored each other more closely. Democrat Dr. Amy Acton first led Republican Vivek Ramaswamy on Kalshi on Aug. 13. Following a series of polls in September, Decision Desk HQ placed Acton ahead of Ramaswamy on Sept. 17. Both platforms maintained her lead through Election Day.
Dr. Justin Buchler, an associate professor of political science at Case Western Reserve University, noted that predicting seat swings in Senate elections is difficult because only a third of the chamber is up for election in any given year. He cited the death of U.S. Sen. Lindsey Graham and the underperformance of his sister, Republican candidate Darline Graham, as examples of idiosyncratic factors that complicate forecasts.
Dr. Robert Alexander, a political science professor at Bowling Green State University, observed that increased media coverage of incumbents can negatively impact their standing in prediction markets. He stated that while he cannot pinpoint the exact cause of the shifts in the 7th District, the correlation between media attention and odds movement makes sense.
Alexander added that the overall performance of prediction markets and polls was similar. "Where they were successful, both were successful, polling and the prediction markets," Alexander said. "And where they failed, they both failed."
The popularity of these platforms has grown sharply in 2026. Pew Research Center analysis of data from The Block showed that monthly global trading volume for companies like Kalshi and Polymarket rose from $26 million in May to $53 billion in July. Much of that early growth was driven by sports trading, but election interest has since expanded the user base.
Fidelity defines prediction markets as marketplaces where participants buy and sell contracts tied to the outcome of future events. Each outcome is tradable, and contracts typically settle at $1 per contract. With the 2028 presidential election approaching, Alexander believes this form of political prognostication will continue to gain traction.
