Survey finds Kalshi odds can shift voters’ expectations of election winners

A survey of 1,994 registered voters found that election prediction-market odds shaped perceptions of who would win even when they conflicted with polling. Candidates favored by Kalshi were 10 to 20 percentage points more likely to be seen as winners.

A Brennan Center-funded survey was conducted on Sept. 21, presenting respondents with hypothetical congressional candidates identified by race, age, party and occupation, alongside NBC News polling estimates and Kalshi odds. Researchers randomly varied the polling and market information to test reactions when the two signals diverged.

A candidate with 67% Kalshi odds was 10 percentage points more likely to be chosen as the likely winner than one with 50% odds. At 84% odds, the increase in perceived electability was 20 points. Conversely, candidates assigned 33% odds were 11 points less likely to be selected, and those at 16% odds were 19 points less likely.

The study found that market odds had an effect on perceived electability comparable to traditional polls and stronger than candidates' demographic, occupational and prior-experience details. Warning respondents about prediction markets did not reduce their reliance on the odds in judging the likely winner.

Respondents nevertheless expressed considerably less confidence in markets than in conventional polling. Nineteen percent called election prediction markets accurate, compared with 37% for traditional polls, while 29% said markets represented public opinion, against 55% for polls. Another 37% neither agreed nor disagreed that markets were accurate, and 29% were similarly undecided on whether they represented the public.

The survey also found broad support for intervention. Forty-eight percent favored stronger regulation, 23% supported banning some types of prediction markets and 16% favored a complete ban, while 14% favored no regulation at all. Seven in 10 Americans disapproved of betting on elections.

Market prices reflect the prices at which traders buy and sell outcome contracts, rather than answers from a sample of voters. This distinction is particularly consequential in state and down-ballot races, which are often not regularly polled and can have small-volume markets that are easier to move. The identities behind the money shaping market odds are not disclosed.

The findings arrive as election contracts have gained a wider audience. More than $220 million had been wagered on the 2026 midterms, and prediction-market odds have reached the public through partnerships with CNN, CNBC and Fox News.

Brian Schaffner of Tufts University has offered a more qualified view of the relationship between markets and polls, saying prediction markets are unlikely to replace polling and are probably informed by it more than anything else. He said markets can be useful where polling is sparse or unreliable, but fluctuate with factors including fundraising, candidate performance, insider information and scandals, and that people can misread an 80% probability as meaning a candidate will obviously win.