- Polling firm Wedgewood has admitted fabricating most of its election surveys, and VoteHub, an aggregator that carried those results, is the named data source for recurring Kalshi contracts.
- Kalshi’s September 2025 filing says revisions made after expiration will not be counted in determining the expiration value, which raises questions about settled contracts built on figures later withdrawn.
- The public record does not show whether any settlement relied on tainted data, and the rulebook governing the October series still needs to be confirmed.
A polling firm has conceded that most of the election surveys it published were fabricated, putting scrutiny on the aggregators that distributed them. The episode matters for prediction markets because VoteHub serves as the reference source for Kalshi contracts that settle on its polling averages. The central question is what happens to a settled contract when the number behind it is later shown to be false.
What Happened
On Oct. 20, Wedgewood acknowledged that it had fabricated most of its election surveys. The admission followed a review by the research group FiftyPlusOne, which requested raw data from Wedgewood’s Mississippi Senate survey and found inconsistencies in respondent identifiers and transaction records. To check the findings, the group commissioned a comparison survey through the same provider, Pollfish, and its records looked materially different. Wedgewood then admitted the fabrication but described one Nebraska poll as legitimate.
VoteHub carried Wedgewood’s results. Its head of data science, Zachary Donnini, said he spoke with a Wedgewood representative before including the firm’s polling. He also said VoteHub had graded Wedgewood a D and incorporated its data anyway. The Guardian reported that VoteHub was among the polling databases that carried the results.
What Kalshi’s Filing Says
Kalshi’s certification for its VOTEHUBAPPROVE contracts, dated Sept. 12, 2025, names VoteHub’s time-adjusted polling average as the underlying measure and fixes the settlement value at a set observation time. The filing states that revisions to the underlying made after expiration will not be accounted for in determining the expiration value. A separate fallback provision applies if the relevant VoteHub average is missing: the contract then uses the most recent eligible average before the specified date. How a later retraction would be treated under that fallback is left to the rulebook.
Kalshi’s Oct. 20 generic congressional ballot contract, KXVHGCB-26OCT20, resolves Yes when VoteHub’s time-weighted Democratic-minus-Republican margin matches a specified percentage at the observation time. The contract concerns what the measure reads at a precise moment, not who wins. Holders are therefore exposed to a third party’s methodology, publication schedule, and correction procedures.
The filing and the October contract page also differ on timing. The 2025 document refers to the first available 10 a.m. Eastern reading after the specified date, while the October page specifies 5:05 p.m. Eastern the same day. The current-series rulebook is therefore the document that needs to be checked. For now, the filing confirms that the post-expiration revision exclusion exists, but how it applies to the October series has not been confirmed.
Why It Matters
Fixed observation times are standard in event contracts because traders need a definitive reference point. That design, however, ties settlement to the quality of source data available at that moment. A contract that resolves on a figure later revealed as fabricated creates tension between settlement certainty and accuracy. Traders take on exposure not only to the election itself but also to the reliability of the aggregator’s inputs.
Consider a hypothetical contract that resolves Yes if the polling average reads 8% at the observation time. If the source later revised that reading to 7.8%, the post-expiration exclusion would, as written, keep the original settlement. The tension is sharper when the underlying figure was fabricated rather than simply updated under an ordinary methodology change.
The wider market is also affected by how these inputs are constructed. VoteHub’s 35-page methodology describes a forecasting model that combines polling, fundamentals, expert assessments, and Kalshi prices, with market weight varying by race and subject to a cap. A separate legacy model runs on polling and fundamentals. The document acknowledges that feeding market prices into forecasts consumed by traders can create feedback loops. Because the settlement reference is a polling average and the forecast is a different product, the methodology keeps settlement inputs separate from market-derived estimates. The practical question is whether polling observations, model forecasts, and tradable probabilities stay clearly distinct when all three circulate together.
Market Context
Prediction markets have drawn attention for how heavily they depend on third-party data. In August, the Associated Press reported that a Kalshi contract tied to a candidate moved about two cents within 15 minutes of a fabricated Los Angeles mayoral poll from Median Strategies. A cluster of Polymarket accounts traded in the same direction shortly afterward. AP’s reporting did not determine who owned those accounts or whether any profits resulted. That case involved a contract on an election outcome rather than settlement against a polling average, but it illustrates how quickly a fabricated number can move prices.
The scope of the Wedgewood episode, as far as the public record shows, is narrower. The identified fabricated surveys covered state races, while the example contract tracks a national average. Whether Wedgewood data affected that average or any payout has not been documented.
What to Watch Next
- Which rulebook governs the October series, and how the 10 a.m. and 5:05 p.m. Eastern timing references are reconciled.
- Whether any contract has settled on a figure that was later withdrawn or found false, and whether Kalshi allows settlement to be reopened once fabrication is confirmed.
- Whether VoteHub publishes which Wedgewood surveys entered its averages and when they were removed.
- How election contracts are handled as Nov. 3 approaches, since each depends on third-party data screened before settlement becomes final.
Source: original report