Prediction markets are betting big on the US midterms. But can they get it right?

As polls closed in Brazil last week, online bettors appeared confident that Flavio Bolsonaro would become the country's next president, putting his chances of winning at around 80 per cent. However, their prediction proved wrong. As the ballots were counted, it became increasingly clear that Luiz Inacio Lula da Silva would emerge victorious. The odds quickly changed as traders rushed to place new bets on the eventual winner.

This year, prediction markets are facing an unprecedented test of their influence and accuracy. Platforms such as Polymarket and Kalshi have been operating for several years, but their popularity has surged more recently. They now allow users to place bets on a wide range of outcomes, from World Series winners to consumer spending on Cyber Monday. The companies are looking to make the US midterm elections a defining moment for their businesses. They hope a series of accurate predictions will attract more users and demonstrate that their platforms can anticipate election outcomes more effectively than traditional polls and forecasting models.

To bring in new users and generate fees from betting activity, the companies have launched extensive marketing campaigns. Kalshi rented a massive billboard in Times Square displaying odds on which party will control the Senate. Meanwhile, Polymarket released a television advertisement showing a struggling political campaign with the tagline: “Watch the midterms move.” Accurate predictions could provide a significant boost to an industry that is increasingly facing scrutiny from lawmakers. Concerns include possible market manipulation and insider trading.

Research has hinted that prediction markets such as Polymarket and Kalshi can deliver highly accurate forecasts. However, that does not guarantee that the most heavily favored outcome will always materialize. Prediction markets allow participants to wager on the outcomes of specific events. Although much of the activity revolves around sports, elections have become an increasingly popular category. Campaigns typically unfold over several months, giving markets time to attract millions of dollars in wagers from both amateur and professional bettors.

Also Read: Women hold just 17.4% of Trump's cabinet-level posts, down from 52% under Biden: Report

As participants place bets on different outcomes, prices fluctuate. These changing prices create continuously updated odds that represent the market's assessment of how likely a particular event is to happen. A review examining more than one million markets, published in the Proceedings of the National Academy of Sciences this year, found that the odds were well calibrated. In other words, outcomes considered strong favorites generally occurred, although there were exceptions.

Researchers believe this year's election results could offer a deeper understanding of the predictive capabilities of these platforms. Earlier studies may not fully represent how prediction markets operate today, given the significant increase in users and the amount of money flowing through them. Dani Lever, a spokeswoman for Kalshi, compared prediction markets with the stock market, noting that investors have historically traded around elections. “By shifting the focus from what people want to happen to what they think will happen,” she said in an emailed statement, prediction markets can filter out social media noise and pundit bias.

According to a report titled “Gambling on Democracy,” published by the Anti-Corruption Data Collective, more than $221 million has entered markets connected to the midterm elections. That figure compares with $92 million recorded during the 2024 presidential race. The growing popularity of prediction markets has also attracted criticism, particularly over whether these platforms offer any meaningful predictive advantage. During the 2024 presidential election, prediction markets indicated much earlier than traditional polls that Donald J. Trump would defeat Kamala Harris.

However, prediction markets have also recorded some significant forecasting failures this year. During this year's primary elections, the platforms assigned Spencer Pratt, a conservative candidate for mayor of Los Angeles, a 75 per cent probability of making the Nov. 3 general election ballot. Instead, he finished third in the primary, falling nearly 30,000 votes behind and failing to advance.

High-quality polls conducted before the election offered mixed predictions. Half indicated that Pratt would finish second, while the other half predicted that Nithya Raman, who ultimately secured second place, would take that position. Pollsters and election forecasters have been particularly critical of prediction markets. They argue that these platforms primarily reflect the opinions of online gamblers and rely heavily on information already available through traditional polling.

Polls have indicated that the contest is evenly divided. Election modelers, including Morris, have favoured Josh Turek, the Democratic candidate. However, prediction markets such as Polymarket, Kalshi and PredictIt have placed Republican candidate Ashley Hinson ahead. In several other closely contested Senate races, prediction markets have favored Democrats. One example is Maine, where Republican incumbent Susan Collins, who has held the seat since 1997, is trailing Democrat Troy Jackson.

Collins has been behind in every high-quality poll conducted in October except one, which showed the candidates tied. The current polling picture is notable because Collins also trailed in every poll published during the final four months of her 2020 re-election campaign. Nevertheless, she won by more than 8 percentage points against her Democratic challenger on Election day. The result reflected a broader polling error that underestimated Republican support that year.