Polymarket makes Phil Scott an 84% favorite even as one local poll has him trailing. It backs Molly Gray too — and there, the polls agree.

Prediction-market prices in this story are as of the afternoon of Sept. 28 and change constantly.

by Compass Vermont

Last Thursday, a University of New Hampshire poll found Democrat Amanda Janoo ahead of Gov. Phil Scott, 49% to 43%, among likely voters, the people the pollster judged most likely to cast a ballot. UNH called the race a statistical tie, meaning the lead falls within the poll’s margin of error.

The betting odds moved, but Scott stayed a heavy favorite. A prediction market is a website where people buy and sell shares in the outcome of real events, from elections to wars. On two of the largest prediction-market platforms:

• Kalshi had Scott at 93% the day before the poll. The day it came out, his price fell as low as 82% on nearly five times as much trading as the previous two weeks combined. As of Monday afternoon, it sat at 83%.
• Polymarket had him at 84%.

Polymarket also gave Democrat Molly Gray an 83% chance of unseating Lt. Gov. John Rodgers. There, the polls agree:

• UNH has Gray ahead 54% to 35%.
• A Vermont Public poll released Monday has her ahead 43% to 35%.

That same Vermont Public poll put Scott ahead of Janoo 44% to 34% among likely voters, with 16% undecided. In the governor’s race, the betting markets line up with one Vermont poll and not the other.

That split raises a question bigger than who wins in November: what happens when betting odds become part of the news voters use to size up an election?

Long Before Polls, There Were the Odds

Prediction markets are not new to American politics. They are returning to it.

From 1868 to 1940, Americans bet on presidential elections in large, organized markets centered on Wall Street, and newspapers printed the odds the way they print polls today. The money was serious. In 1916, Americans bet about twice as much on the presidential race as the campaigns spent.

The markets were good at it. Economists Paul Rhode and Koleman Strumpf studied those markets and found that between 1884 and 1940, the mid-October betting favorite won 11 of 15 presidential elections, and the underdog won only once, in 1916. They concluded the markets faded largely because scientific polling arrived in the 1930s and other forms of gambling became legal.

For most of the century since, polls told voters where a race stood. Now the odds are back.

The Odds Go Mainstream

CNN has called 2026 the “first prediction-market election.” More than $750 million has been traded on this year’s election markets on Kalshi and Polymarket. Vermont’s governor race has drawn about $320,000 across the two sites, roughly $1 of every $2,300 traded on this year’s election markets.

The odds now show up inside the news itself:

• CNN and CNBC have exclusive data deals with Kalshi. CNBC discloses that its relationship includes a minority investment in Kalshi.
• Dow Jones, which owns The Wall Street Journal, Barron’s and MarketWatch, has an exclusive partnership with Polymarket.
• The Associated Press licenses its vote counts and race calls to Kalshi for the midterms.

Supporters see the odds as a real-time read on what large numbers of people expect. Dow Jones CEO Almar Latour called Polymarket “a rapidly growing source of real-time insight into collective beliefs about future events.” Status, a media-industry newsletter, argued that legacy news companies are signing these deals less out of editorial conviction than financial need.

What an 84% Does to a Voter

A poll and a market price measure different things:

• A poll asks voters who they would vote for today. Scott’s 43% means 43 of every 100 likely voters UNH surveyed picked him.
• A market price reflects what traders will pay on the outcome. An 84-cent price on Scott implies roughly an 84% chance he wins, or about one chance in six that he loses. It does not mean he gets 84% of the vote.

People react to those two kinds of numbers differently. Researchers Sean Westwood, Solomon Messing and Yphtach Lelkes found in experiments that a win probability made the outcome feel more certain than a poll showing one candidate a few points ahead. Participants were also less likely to spend the effort to “vote” in a simulated election. The researchers did not show that forecasts reduce turnout in real elections, and forecaster Nate Silver disputed how the findings were described.

Campaigns are paying attention either way. CNN reported that one candidate checks the market for their own race nearly every hour, and that donors watch the prices too.

Both Polls and Markets Miss

Vermont has recent examples of both getting it wrong.

The markets. In this year’s Democratic primary for governor:

• In May, Polymarket’s market didn’t list Janoo at all, and most of its money had been bet on Attorney General Charity Clark, who never ran, as Compass reported.
• In June, Polymarket priced Aly Richards at 66% and Janoo at 34%.
• UNH’s June and July polls both had Janoo ahead.
• Janoo won, 49.1% to 46.6%.

Elsewhere this year, heavily favored candidates lost primaries in Wisconsin, Hawaii and Texas.

The polls. UNH’s record on Scott shows the other kind of miss. Its early-October 2022 poll had Scott ahead by 17 points; he won by 47. Its final 2024 poll had him ahead by 39; he won by about 52.

Those past misses are information traders can weigh, along with Scott’s electoral history: no Vermont incumbent governor has lost reelection since 1962, as Compass reported in August. But a market price doesn’t tell us why any trader buys at a particular price.

The campaigns read the new poll exactly as you’d expect. Scott’s campaign manager, Dustin Degree, called it “statistical malpractice.” Janoo’s campaign told supporters a good poll was a good start, but plenty of Vermonters haven’t decided.

Whose Odds Are These?

Polymarket’s price on Scott comes from its international site, which blocks trading by anyone in the United States. Kalshi is different: it operates legally in the U.S. under federal regulation by the Commodity Futures Trading Commission. It has more than twice as much money traded on Vermont’s governor race, and it puts Scott in nearly the same place, at 83%. The high odds on Scott aren’t just an overseas view.

Vermonters, for their part, have told pollsters they’re wary of all this. In March, a UNH poll found:

• Most Vermonters had heard of prediction markets like Kalshi and Polymarket, but very few had used them.
• Vermonters said it should be illegal to use prediction markets to bet on elections, the economy or political events like war. They were split on betting on sports or cultural events.
• A majority support more federal regulation of prediction markets.
• Nearly half expect prediction markets to have a negative effect on the country over the next decade.

Rep. Tom Stevens, D-Waterbury, and 12 co-sponsors introduced a bill this year, H.913, that would have banned certain prediction-market contracts in Vermont, including bets on elections, as Compass reported. The House Government Operations and Military Affairs Committee heard an introduction and a walk-through of the bill in April but took no further action.

Polls ask Vermonters who they’ll vote for. Markets ask traders what they’d bet on the result. On Nov. 3, neither decides the election. The votes do.