U.S.-China Relations
NextFin News - With fewer than two weeks left before the Nov. 3 midterm election, the single most quoted number in American political forecasting is no longer a poll result. It is a price: the implied probability on a prediction market, where traders wager real money on who will win. More than $220 million has been bet on 2026 congressional races so far this cycle, compared with $92 million across the entire 2024 cycle, according to the Anti-Corruption Data Collective (ACDC). The number of markets on those races has ballooned to more than 21,000, up from 463 in 2024. Polls ask voters what they plan to do; prediction markets ask people what they think will actually happen, and put cash behind the answer.
The shift is not merely about volume. It is about who gets to define political reality. A landmark 2024 court victory for Kalshi Inc. reclassified election contracts as federally regulated financial swaps rather than state-regulated gambling, opening the door to a new information industry. Major media outlets now embed real-time market odds in their coverage, and several have struck distribution deals with the platforms. What began as a niche forecasting tool has, in two election cycles, moved from the fringe to the center of the political information ecosystem — and in the process upstaged the polling industry that dominated it for a century.
The 2024 Inflection Point: When the Market Called It First
The credibility of prediction markets rests on one election. In 2024, while poll-based forecast models gave Kamala Harris the edge, the betting platforms leaned the other way. On Oct. 7, 2024, Polymarket showed Donald Trump's odds of winning spiking to 53.3%, with Harris falling to 46.1%; the same day, a widely followed simulation model put Harris at 55%. The market moved first, and it turned out to be right.
The payoff was concrete and large. A single trader, identified in subsequent reporting as a French national named Théo, wagered more than $28 million on a Trump victory across four accounts and collected roughly $85 million when the result was certified. The "Trump whale" became the archetype of a new kind of political forecaster: not a pollster with a sampling frame, but a bettor with a concentrated position and skin in the game. By Nov. 5, 2024, more than $3.3 billion had been wagered on the presidential race alone, making it the most active market in the platform's history.
The legal foundation for this market was laid weeks earlier. On Sept. 6, 2024, a federal district judge in Washington granted Kalshi summary judgment against the Commodity Futures Trading Commission, vacating the regulator's ban on "Congressional Control Contracts." The court held that election contracts were financial swaps subject to federal oversight, not gambling. The CFTC appealed, but the ruling stood, and within months the industry had a regulatory green light. In October 2026, Kalshi won another federal ruling, this time against Illinois state regulators, after a judge held its wagering markets must be regulated as financial swaps under federal oversight rather than as gambling under state law. The company's chief executive has said the dispute could reach the Supreme Court.
Why Markets Can Outrun Polls
The mechanism is incentive design. A poll is a cheap-talk survey: respondents pay nothing for being wrong, and the sample can miss non-responsive voters — the "rural nonresponse" problem that has bedeviled pollsters in recent cycles. A prediction market is a wager. A trader who is wrong loses money; a trader with private information has a financial reason to act on it. The market price becomes an aggregate of dispersed information, including information that never appears in a survey.
That aggregation happens continuously. Polls are snapshots, released at intervals, each with a margin of error and a fieldwork window. Market odds update by the second, absorbing campaign events, fundraising reports, and — critics worry — non-public information. In 2024, the platforms reacted within minutes to the July 13 assassination attempt on Trump, repricing his odds upward before most polls could register the event. An academic analysis of the 2024 race concluded that Polymarket was superior to national polling data in predicting the presidential outcome, and that it captured swing-state dynamics that polls missed.
The product has also evolved. In 2024, 71% of election markets were about who would win a race. This cycle, only 16% are winner markets; 58% are about the margin of victory, plus new markets on voter turnout and candidate behavior. Traders are no longer just picking winners — they are pricing the shape of the outcome. That granularity is precisely what media consumers and campaign strategists want when polls offer only topline horse-race numbers.
The contrast is visible in real time. In the Maine Senate race, a recent survey of likely voters showed the Democrat, Troy Jackson, at 49% and the incumbent Republican, Susan Collins, at 46%. Polymarket, by contrast, put the likelihood of a Democratic pickup at 68% — a wider gap than the raw poll numbers alone would suggest, because traders are pricing in factors such as fundraising momentum and national partisan tides rather than a single head-to-head snapshot. In Texas, where Democrat James Talarico faces Republican Ken Paxton for an open seat, market odds have swung sharply as campaign events unfolded, and the race has become one of the most heavily traded congressional markets of the cycle.
Second-Order Effect: The Market Does Not Just Forecast Politics — It Changes It
The first-order story is that prediction markets forecast elections better than polls. The second-order story is more consequential: the markets are becoming part of the campaign machinery itself, and that changes the behavior of every actor in the system.
Campaigns now watch market odds as a real-time barometer. When a price moves against a candidate, donors react, volunteers reconsider, and opposing campaigns adjust their spending. The markets create a feedback loop: a falling price can depress fundraising, which in turn worsens the candidate's actual chances, which pushes the price lower still. This is the self-fulfilling dynamic that pollsters never had to contend with, because a poll is a measurement while a market is a participant.
The concentration of trading amplifies the effect. ACDC found that the top 1% of wallets on Polymarket, ranked by trading volume on congressional markets, account for 70% of all congressional betting volume this cycle. A market dominated by a handful of large traders is less a "wisdom of crowds" mechanism than a bet on whether a few whales are right. When one trader can move a price by placing a $28 million position, the market's claim to aggregate dispersed information weakens — and its power to move the underlying political reality strengthens.
There is also the insider-information problem. In July 2026, it was reported that a White House teleprompter operator placed bets on Kalshi, raising the question of whether government insiders can trade on non-public political information. Kalshi blocks election trades by politicians and campaign workers and says it refers suspicious cases to U.S. authorities; it has suspended a handful of candidates for betting on their own races. But surveillance is reactive, not preventive, and the definition of "inside information" in politics is far murkier than in securities markets.
This Is a Structural Shift, Not a One-Cycle Fluke
The question that determines the outlook is whether the upstaging of polls is cyclical — a one-off correction after pollsters' recent misses — or structural, a regime change that will not revert on its own. The evidence points to structural.
Three forces make reversion unlikely. First, the legal regime has changed: election contracts now have a federal regulatory classification, and an administration broadly supportive of novel financial products has given the industry room to test the limits of the "swap" definition, expanding into sports outcomes and player-style props. Second, distribution is locked in: major television networks and newspapers have signed partnerships to feature market odds, embedding the platforms into the daily flow of political coverage. Third, the talent and credibility have migrated: Nate Silver, the polling analyst who built FiveThirtyEight, became an advisor to Polymarket in 2024 — a symbolic passing of the forecasting torch from sampling to wagering.
Volume confirms the direction. The $220 million wagered on 2026 congressional markets through Sept. 30 already dwarfs the prior cycle's total, with the election still two weeks away. Trading volume is highly concentrated in a few races — most notably the Texas Senate contest, plus the Senate races in Maine, Michigan, and South Carolina — but the breadth of markets has expanded across the board. The platforms have the liquidity, the legal footing, and the media pipeline to remain central fixtures.
The Counter-Thesis: Markets Are Downstream, Concentrated, and Unproven
The strongest case against the "markets beat polls" narrative is that the two are not independent. Brian Schaffner, a political scientist at Tufts University and co-principal investigator of the Cooperative Election Study, argues that prediction markets are "downstream from polling" — informed by polls more than anything else. If market prices largely reflect poll releases, then the market is not a superior information source; it is a derivative of the same underlying data, with a layer of speculation added on top. On that view, the market's apparent outperformance in 2024 was not a triumph of aggregation but a bet on the same signals the polls contained, dressed in the language of probability.
There is also a concentration problem. A market dominated by a handful of large traders is less a "wisdom of crowds" mechanism than a bet on whether a few whales are right.
The accuracy record is thinner than the marketing suggests. Polymarket says its election markets have a 98% accuracy rate one day before an election, a figure supplied by the company itself and not independently audited. Accuracy one day out is also a low bar: by then, most competitive races have already been decided by fundamentals and polling. The harder test — calling tight races weeks in advance, when the forecast has real value — is less clearly won.
Then there is the manipulation risk. Silver has publicly warned that one vector of abuse is the publication of fake polls designed to move prediction-market prices, which can then be traded against. Campaign staff and even candidates have been caught betting on races they work on. But surveillance is reactive, not preventive.
Finally, the regulatory and public-opinion overhang is real. A bipartisan bill introduced in the Senate in March 2026 would restrict what entities like Kalshi and Polymarket can offer wagers on. State regulators continue to challenge the federal classification, and election officials in at least one state have warned voters that betting on an election could cost them their vote and carry felony charges. An Ipsos survey of 4,506 U.S. adults conducted Sept. 30 to Oct. 5 found 60% of Americans oppose using prediction markets to wager on elections, including 69% of Democrats and 61% of Republicans; only 13% support it. The industry has the courts and the volume; it does not yet have the public.
What to Watch: The Nov. 3 Test
The falsifying signal is specific and near: if aggregate polling averages for the 2026 House and Senate outcomes end up closer to the actual results than the prediction-market prices — particularly in the tight Senate races in Texas, Maine, Michigan, and South Carolina, where trading volume is most concentrated — then the "markets beat polls" thesis takes a material hit. Conversely, if the markets correctly price a result that polls miss, the structural shift narrative hardens.
Short term, the next two weeks will show whether market odds converge with late polling or diverge from it. Medium term, the Supreme Court question and the Senate bill will determine whether the federal-swaps classification survives. Long term, the structural case rests on whether the platforms can keep liquidity and credibility through a cycle in which they are wrong — because a forecasting industry that only works when it is right is not an information market at all.
"This data is bottom-up, not elites-driven," said a Kalshi spokesperson. "Prediction markets are a way to see what people are saying when there's money on the line."
The counterpoint, from the polling side, is just as sharp: markets may be measuring the mood of traders with money, not the intent of voters with ballots. On Nov. 3, only one of those groups actually decides the outcome.
Conclusion: A New Information Regime, With Old Risks
The upstaging of election polls by prediction markets is a structural change in how political information is produced, priced, and consumed. The beneficiaries are the platforms, the media companies that distribute their odds, and traders with information or conviction strong enough to back with capital. The exposed are pollsters, whose relevance now depends on being incorporated into market prices rather than standing alone, and voters, who may increasingly encounter a forecast that reflects betting flows as much as voter intent.
Base case: prediction markets remain a co-equal voice alongside polling, with the two feeding each other. Upside case for the platforms: a 2026 miss by the polling aggregates cements market odds as the default political forecast. Downside case: a high-profile market error, a successful regulatory clampdown, or a manipulation scandal slows the migration and restores polling's primacy.
The central judgment: prediction markets have not replaced polls, but they have displaced them as the default source of political certainty — and that displacement is structural, because it is built on legal change, distribution deals, and incentive design, not on a single election's surprise. The risk is that a market can be liquid, legal, and wrong all at once.
Explore more exclusive insights at nextfin.ai.
