New York City’s mayoral primaries are an early test for prediction markets in the era of widely available election markets.
Prediction markets showed wild volatility around Zohran Mamdani and Andrew Cuomo‘s chances of winning the Democratic NYC Mayor primary in the wake of updated polling information in the lead-up to the day of the election. Commercial prediction market Kalshi saw Mamdani’s odds jump from 18% on June 18 up 35 points to 53% on June 23, the eve of the primary.
Throughout Monday and into Tuesday evening, Mamdani’s odds fluctuated between 41% and 58%. As of 6 p.m. ET, three hours before polls closed, Cuomo was back to being favored in the Kalshi market:
BREAKING: Cuomo regains the odds lead after being down 16% this morning pic.twitter.com/ny2VqrIvdO
— Kalshi (@Kalshi) June 24, 2025
But that lead did not last through the evening. While the market had Cuomo ahead for the majority of the final voting hours, Mamdani’s chances on Kalshi surged swiftly to 99% once reports of Mamdani’s projected win surfaced. Cuomo later officially announced he was conceding the Democratic primary for mayor.
The 20-point swings in odds so close to the release of influential polls in the run-up to election day raise questions about how predictive prediction markets are. Despite the industry’s triumphalism in the wake of the 2024 election, the results were actually mixed, showing some important nuance around the predictive power of event contract trading markets.
Prediction markets respond to new important information
One of the features prediction markets are most proud of is the live change in odds as new information becomes available to traders. On Saturday, Kalshi CEO Tarek Mansour posted that prediction markets “allow politicians and their followers” to track odds in real time:
The great thing about markets compared to polls is that they are live.
They allow politicians and their followers to track their odds of success in real-time and adjust their strategy accordingly.
Good luck to both candidates next week! https://t.co/4cZ8u5BbaR
— Tarek Mansour (@mansourtarek_) June 21, 2025
One of the lessons that the New York City mayoral race will teach is whether prediction markets are equally useful in all electoral races. After the 2024 election, economist and Columbia professor Rajiv Sethi compared 13 House race projections from crypto platform Polymarket and polls from FiveThirtyEight and The Economist.
Sethi found that FiveThirtyEight had the best projections, followed by Polymarket. However, The Economist performed better in certain races than did Polymarket.
So, why did the much-maligned polls outperform one of the largest prediction market platforms?
Sethi illustrated an important point about volatility in the 2024 presidential markets after Anne Selzer’s poll showed Kamala Harris leading by three points in Iowa, a state Trump won by about six points:
“On Polymarket, the likelihood of a Republican victory in the two most competitive Iowa districts fell sharply—from 58 to 25 percent in IA-01 and from 68 to 51 percent in IA-03. Changes in model forecasts were much more modest:”
Republicans ultimately won both races, showing the value of polls in a smaller, controlled environment. Sethi concluded that “prediction markets are a useful part of the forecasting ecosystem, especially in a rapidly changing world in which few historical regularities can be taken for granted.”
However, he also pointed out that they’re not “magic bullets” for divining the future. That leaves pollsters looking for a win over prediction markets a potential opportunity in New York City.
NYC race a potential victory for pollsters
Prediction markets aggregate more than just poll results. They also allow for speculation, for example, betting against polling accuracy, as the high-volume French trader on Polymarket did during the 2024 presidential election.
Prediction markets also overreacted to polls at certain points in the presidential race, but in local races, polls may be better equipped to capture the variables needed to predict the race’s likely winner. It may be more appropriate for prediction markets to follow polls in these races than the national ones that can buck historical trends.
If prediction markets are better suited for predicting outcomes of national races than local ones, then polls could outperform prediction markets in many local and House races. New York City’s primary is an early gauge of both the promise and limits of prediction markets.
