Want to know who will win the US midterms? Forget the polls – and follow the money.
That’s the idea behind political prediction markets. Platforms such as Polymarket and Kalshi allow people to speculate on political outcomes, turning everything from control of Congress to individual races into tradeable events. And many of these bets are big business: Polymarket’s top market by volume in 2025 was the New York City mayoral race, which drew over $150 million in wagers from 70,000 traders.
Prediction markets have exploded since then. Monthly trading volume across major platforms jumped from roughly $1.2 billion in early 2025 to more than $20 billion by January 2026, according to crypto analytics firm TRM Labs – a more than sixteenfold increase.
This monetized “wisdom of crowds” has become attractive to political junkies, polling firms and news organizations. CNN, CNBC and The Wall Street Journal have all partnered with prediction-market companies to include their odds in political coverage.
As Americans head toward November’s midterms, the boom is raising a bigger question: can prediction markets rival – even replace – traditional polling?
The case for markets: accuracy, speed, and data. As far back as 2008, research found that earlier versions of prediction markets beat pollsters 74% of the time, and significantly outperformed polls when forecasting more than 100 days in advance.
Fast-forward to today, newer prediction market sites like Polymarket have proven to be accurate as well: a 2025 Vanderbilt University study comparing the platform with traditional polling during the 2024 presidential election found it forecast the outcome better, particularly in swing states. The researchers suggest this may be because Polymarket bettors represent a wider population than those sampled by polling firms. Prediction markets may also be valuable in races such as primaries, where many voters don’t pay as close attention until closer to Election Day.
Prediction markets also deliver faster results than polls. Polls can take days to conduct and tabulate, but markets move instantly. Events like debate disasters, candidate scandals, or major endorsements can alter prices within minutes, making markets into a “real-time” political barometer. The 2024 assassination attempt on Donald Trump, for example, caused an immediate surge in Polymarket bets predicting a Trump victory.
Traditional polling has other problems. Pollsters must determine not simply what people think, but which people are voting. Getting the composition of the electorate wrong can distort the result. This was one reason polls misjudged the 2016 presidential election: white voters without college degrees were underrepresented in many polls, causing surveys to underestimate Donald Trump’s level of support and overstate Hillary Clinton’s.
But polls and markets don’t always measure the same thing. Instead of asking whom voters intend to support, prediction markets ask a different question: Who do you think will win? A voter can tell a traditional pollster she supports Candidate A, even if she believes Candidate B will ultimately be victorious. But traders in a prediction market are interested in the second call – and in the hopes of a financial reward.
A poll also attempts to construct a representative sample of actual voters, but in a prediction market, anyone eligible to trade can participate, regardless of where they live. Prediction markets can show who people around the globe think will win, but a public opinion poll will show the inclinations of the electorate in a particular area – information which campaigns can then use to tailor their messaging or ground game.
Some observers say prediction markets won’t replace polling. Tufts political scientist Brian Schaffner sees them as “downstream from polling” because bettors may base their wagers on what polls are showing. While some have likened prediction markets to casinos, researchers found that bettors often factored in a host of data, including past voter turnout, fundraising numbers, breaking news, and traditional polls, to distill it into the probability of victory.
Money and manipulation. Prediction markets may not replace traditional polling because of their fundamental motivator: money. Someone betting $100,000 arguably has much greater influence than someone risking $10, because they will change the prediction and affect other bettors in the pool. A wealthy partisan who cares little about losing money, too, could theoretically bet enough to make his preferred candidate appear more likely to win, and the candidate can then tout those odds to donors, activists and journalists to give the illusion of momentum.
Election officials are increasingly worried prediction markets could impact public trust– and scrambling to establish rules. More than half of US states already restrict election betting in some form, and states are now wrestling with whether those rules cover prediction market contracts. Twenty states have taken prediction markets to court over whether they fall under state laws governing sports betting. Both Kalshi and Polymarket were contacted for this article but did not reply for comment.
Complement, not replace? Rather than the new polls, prediction markets may be more like a political futures market: fast, data-rich and potentially valuable, but also volatile, financially driven and vulnerable to manipulation. Each has its use: polls measure the electorate, while prediction markets measure expectations. The future of election forecasting may not belong to one or the other, but to analysts who can read both.
