The Election Market Won in Court. Now It Has to Win the Argument.

One trader, one wrong price

Start with a single contract. A market lists a binary option on a public event, a Senate race, that pays one dollar if the Democrat wins and nothing if she loses. It trades at 62 cents. A trader who has read the state's early-vote returns thinks her real chance is closer to a coin flip. To him the contract is overpriced, so he sells it at 62 cents. If he is right, he keeps the gap.

His sale is not a private act. Selling pushes the quoted price down toward the number he believes, and everyone watching now sees a lower figure. That figure carries his information without him ever stating what he knows. The next trader who has read something the crowd has not does the same, from whichever side he thinks the price is wrong. The price becomes a running tally of everyone willing to back a disagreement with their own money, weighted by how much they staked and how sure they were.

This is what economists call information aggregation, and Wolfers and Zitzewitz state its strong form precisely. In a truly efficient prediction market, they write, "the market price will be the best predictor of the event, and no combination of available polls or other information can be used to improve on the market-generated forecasts." The claim does not require that everyone in the market be rational. It requires only that "the marginal trade in the market is motivated by rational traders."² The lazy money can stay; it just becomes the profit the informed trader collects, and the collecting is what moves the price.

Now set the pundit beside the trader. Hanson's founding complaint, on the personal page where he first laid out what he called Idea Futures, is that "our policy-makers and media rely too much on the 'expert' advice of a self-interested insider's club of pundits and big-shot academics."¹ The pundit who calls the race wrong pays nothing. He keeps his column and his fee, and next cycle he is asked again. The trader who calls it wrong is poorer by the spread. A price is honest for the same reason a wrong bet hurts: being mistaken has a cost, and telling a good story does not.

What the prices have actually done

The mechanism has a track record, and it should be measured against a named benchmark rather than praised on its own. In the week before the past four presidential elections, the Iowa Electronic Markets predicted the two-party vote shares with an average absolute error of about 1.5 percentage points. Over the same four elections the final Gallup poll erred by 2.1 points.² A point of accuracy on a vote share is not nothing, but it is not an oracle beating a coin-reader either.

The same authors supply the limit in the same breath. They note that stacking a market against "a mechanistic application of poll-based forecasting may not provide a particularly compelling comparison," because a careful forecaster does more with a poll than read it off the page.² And where the rival benchmark is strong, the market edge can vanish outright. On the monthly economic releases traded on economic-derivatives markets, market-implied forecasts and the consensus of roughly fifty professional forecasters showed "no statistically (or economically) meaningful differences in forecast performance."² A market is not magic. It is a good aggregator that ties or beats other good aggregators, and loses its shine against a benchmark that already pools many informed views.

The election-forecasting margin in particular is genuinely contested, and the case for the market there is far from settled.

Why the question is live tonight

For most of these markets' history the stakes were academic, because the markets were small by law. The Iowa Electronic Markets ran as a capped research exchange. PredictIt limited how many people could trade and how much each could stake. That changed in court. After the Commodity Futures Trading Commission moved to ban Kalshi from listing contracts on which party would control Congress, a district court vacated the ban, and in October 2024 the D.C. Circuit refused to freeze that ruling while the agency appealed, holding that "the Commission has failed at this time to demonstrate that it or the public will be irreparably injured absent a stay."³ By 2026 the Commission's own enforcement division was writing not to prohibit these markets but to police them, affirming that it retains "full authority to police illegal trading practices occurring on any" designated exchange.⁴

The scale is what makes 2026 different from the Iowa lab. Unlike PredictIt and the Iowa markets, which cap both the number of investors and the amount each can spend, Kalshi "places no cap on the number of investors and would allow individuals and entities to invest, in some cases, up to $100 million per contract."³ A deeper market is a better aggregator, because a single mistaken price now invites a hundred-million-dollar correction. It is also, as the counter will insist, a bigger target.

Hanson's ambition was always larger than forecasting. If a price can tell you what will happen, it can in principle tell you which policy would raise the outcome a society says it wants. That is his proposal of futarchy, in his own words.

One need not buy the whole governance scheme to see the pull of the smaller claim underneath it. Hanson's diagnosis is that ordinary institutions "frequently fail by ignoring the advice of relevant experts,"⁵ and a market is a machine for paying the relevant expert to speak through a price whether or not any committee thought to ask him. Whether that machine can be trusted at a hundred million dollars a contract is the fight the law has now handed to the rest of us.

The three strongest reasons to distrust the price

The case has three serious objections, and each is held by people who know these markets well. Take them at full strength.

First: the markets may add nothing a good poll does not. Robert Erikson and Christopher Wlezien, who have studied election forecasting for decades, reach a flat conclusion. "When we have both market prices and polls," they write, "prices add nothing to election prediction beyond polls." Early markets were "surprisingly good at extracting campaign information without scientific polling to guide them," but modern candidate prices largely follow the polls rather than lead them.⁶ If the price is just the poll wearing a dollar sign, the elaborate machinery buys nothing.

Second: a deep market is a manipulable one. Wolfers and Zitzewitz, the case's own witnesses, concede the hinge: "the extent to which markets are manipulable depends, at least in part, on how thin the markets are."² Kalshi's removal of the cap cuts both ways. In 2024 a single French trader raised his stake on a Trump win to about $80 million across several accounts roughly three weeks before Election Day, and cleared more than $80 million when it paid.⁷ A price one trader can move that far is a price that can be pushed, and a pushed price misleads everyone who reads it as information.

Third: some markets corrupt the thing they measure. The CFTC's stated worry was that election contracts "could threaten election integrity by, for example, creating monetary incentives for voters to support particular candidates or incentivizing the spread of misinformation."³ The moral hazard is not hypothetical. In 2025 the Commission's own enforcement record includes a political candidate penalized for trading contracts on his own candidacy, an outcome he could influence directly.⁴ Push the logic to its edge and you reach the markets on assassinations and attacks that sank the Pentagon's Policy Analysis Market in 2003: a contract that pays out when someone dies is an incentive for the death.

Each objection lands, and the answer to each comes from the same mechanism that built the case, not from around it.

On accuracy, concede the ground Erikson and Wlezien actually won. In a thick national race with abundant, well-understood polling, a debiased poll average is hard to beat, and the market's edge there can shrink to nothing. But that is the market's worst case, not its typical one. David Rothschild, debiasing both sides of the comparison across seventy-four presidential and Senate races in 2008, found that "early in the cycle and in not-certain races debiased prediction market-based forecasts provide more accurate probabilities of victory and more information than" debiased polls.⁸ The defensible claim is the modest one on the position card: at least as good as a good poll where polling is rich, and better where it is thin, early, or absent. Where a race is well polled, the market ties. Where it is not, the market is the only forecaster in the room.

On manipulation, follow the money the manipulator spends. Hanson and Ryan Oprea modeled exactly this trader, and their result is counterintuitive in the case's favor: a manipulator is a species of noise trader, and when other traders cannot read his target price, "increases in the variance of the target price can increase average price accuracy, by increasing the returns to informed trading and thereby incentives for traders to become informed."⁹ A trader who shoves a price away from the truth is posting a subsidy for everyone who knows better to bet against him. The 2024 whale is a clean test of the theory. An on-chain study of that Polymarket race found the market's vulnerability to price impact, its Kyle's lambda, fell from a spike of about 0.53 around the end of July to about 0.01 by October, and that the late capital flowed into both sides at once, "consistent with trading under heterogeneous beliefs … rather than … a single trader steering the price unopposed."¹⁰ The same study corrects the headline: the widely repeated figure inflated by share-minting overstates the real turnover, and the exchange-equivalent volume of the October Trump market was $391 million, not the billions usually cited.¹⁰ By October the price was deep enough that the whale's large mid-October order moved it very little.

On moral hazard, keep the concession honest and then draw the design line. Hanson wrote the post-mortem on the terrorism-futures panic himself, and the record he assembled cuts against the fear: "we know of no examples of anyone using financial markets to profit from such sabotage," and the studies of the September 11 attacks and the 1982 Tylenol poisonings turned up nothing.¹¹ The structural reason is thinness. Prediction markets are "typically very thin compared with most financial markets," which makes them "very poor places to speculate on sabotage."¹¹ That is also where Hanson concedes the real danger, and it is precise: "speculation on sabotage is possible when one person can acquire a large enough stake in an asset whose value they can directly enough influence."¹¹ Read that condition against the candidate the CFTC caught. He is the danger case exactly: a single actor, a large stake, an outcome he controls. The rule that follows is not to ban the instrument but to police the two danger conditions, thin markets and interested traders, which is what surveillance, position limits, and barring the interested party are for, and what the Commission now says it will do.⁴

What would change this paper's judgment is stated by the mechanism itself. If, in deep markets with debiased polls freely available, prices consistently added nothing that polls did not already contain, the strong version of the case would be dead, and only the thin-information version would survive. Erikson and Wlezien have shown something close to that for recent, heavily polled national races, which is why the claim on the card is bounded and not triumphal. But a bounded claim is still a claim. A price built from real money is at least as good a forecaster as the experts it competes with. Where no poll exists, it is the only forecaster there is. The failures its critics name are conditions to engineer against, not proof the machine is broken. The law was right to get out of its way. The vigilance the critics demand is the rent the mechanism pays for being trusted with a hundred million dollars a contract.