Kalshi overstated primary favorites’ chances, accountability group finds

Kalshi’s election markets overstated the chances of perceived favorites in competitive 2026 primaries, according to a new analysis by Campaign for Accountability. Across 238 contests, the group found the platform was on average 7.3 percentage points too confident in the frontrunner, while underdogs won 74 races rather than the 57 upsets implied by market prices.

The review examined Kalshi odds from the day before 645 primary elections across all 50 states and Washington, D.C. It defined competitive races as those in which the favorite was assigned a 50% to 90% chance of victory. Actual upsets were therefore about 30% more frequent than the market forecast in that subset.

The discrepancy became sharper close to voting, the group said. Kalshi’s competitive markets were nearly perfectly calibrated a week before elections, but confidence in favorites rose during the final week. By election eve, the implied chance of the frontrunner was about six points above the share of favorites who ultimately won.

One conspicuous miss was Florida’s Democratic Senate primary, where Kalshi put Angie Nixon’s chance of defeating Alex Vindman at 5%. Nixon nevertheless won. Florida Politics had reported before the contest that 94% of investors expected Vindman to prevail, in a race with little polling.

Campaign for Accountability compared its results with Kalshi’s data and an August working paper published by the company, saying the paper appeared to show that election markets did not improve during the final week. The group also noted that a June Washington Post analysis had found near-perfect calibration in some subsets of Kalshi’s primary markets, while other parts of that review indicated weaker forecasting performance.

The findings arrive as prediction-market prices have become a visible input into voters’ judgments of electability. As covered earlier this month, a Brennan Center-funded survey of 1,994 registered voters found that a candidate favored by prediction markets was 10 to 20 percentage points more likely to be viewed as the likely winner, regardless of polling.

In that study, moving a candidate’s Kalshi odds from 50% to 67% increased the share of respondents selecting that candidate as the probable winner by 10 points. Odds of 84% raised perceived electability by 20 points, while odds of 33% and 16% reduced it by 11 and 19 points, respectively.

The survey also found considerable public skepticism. Only 19% of voters said election prediction markets were accurate, compared with 37% for traditional polls; 29% said markets represented public opinion, against 55% for polls. More than 80% favored either tighter regulation or some form of ban, while 14% said the markets should not be regulated.

Kalshi chief executive Tarek Mansour has described prediction markets as a source of truth free from bias. The Campaign for Accountability’s executive director, Michelle Kuppersmith, countered that the collective judgments of online bettors are not facts. The group’s researchers said election-market probabilities have not yet proven consistently superior to conventional election forecasting.