Kalshi Named VoteHub Its Source Agency. Then One of VoteHub's Pollsters Admitted Faking the Data.

Prediction markets don't fail on price. They fail on provenance.

Kalshi's September 2025 regulatory filing names VoteHub as the source agency for its recurring election contracts. On Oct. 20, polling firm Wedgewood admitted it fabricated most of its election surveys. VoteHub was one of the aggregators that carried those surveys. That's the whole chain. Three links, one rulebook, and a clause stating that revisions to the underlying data govern how contracts resolve.

So if the underlying data was fiction, what exactly is the contract settling against?

The Paper Trail

Start with the mechanics, because the mechanics are where this gets uncomfortable.

Polling aggregators don't generate numbers. They collect them, weight them, and republish. That means every aggregator is a trust-forwarding device. VoteHub forwards whatever its upstream pollsters hand it. Kalshi then forwards VoteHub's output into a contract spec. Each hop looks like a technicality. Each hop is actually a place where an invented number can enter a market that people settle with real money.

The September filing makes the dependency explicit. It identifies VoteHub as the source agency and addresses what happens when that source revises its underlying data. Read that clause twice. It's not a disclosure. It's the settlement mechanism. The source agency, not the exchange, holds the pen on what counts as truth.

That's a strange place to put the pen.

Kalshi is regulated. The CFTC signed off on its event contracts after a long court fight in 2024 that opened the door to US election markets. Regulators look at market conduct, surveillance, capital, and customer protection. They don't audit a pollster's raw crosstabs in Ohio. Nobody does, really. The polling industry runs on disclosure norms and reputation, not on chain of custody.

Which brings us to the part nobody wants to say out loud. Election contracts are the flagship product line for this entire category. Kalshi's 2024 election volume reportedly cleared $100 million in the final stretch of that cycle. That's real revenue built on top of a data supply chain that failed a basic integrity test.

The chart is the chart. But someone else drew it.

Steelmanning The Other Side

Here's the honest counterargument, and it's stronger than the bears admit.

Most Kalshi election contracts settle on certified results, not on polls. The winner-of-the-presidency market pays out on who gets inaugurated. Polls are an input to how traders price that market. They aren't the settlement oracle. So a fabricated poll distorts the price on the way in, but it doesn't change the payout on the way out.

Fair. And VoteHub itself didn't fabricate anything. Aggregators inherit upstream failures. That's the deal you accept when you source from a free, open polling feed instead of commissioning your own surveys. Building a proprietary polling operation would cost millions and still get things wrong.

Then there's scale. Wedgewood is one firm. Its surveys were a slice of a slice of the national polling diet. The 2026 midterms will be covered by a different roster of pollsters with different contracts and different reputations. Markets price around noisy data every single day. That's the job.

And yet.

Ask a simple question. Who audits the feed? Not the exchange. Not the regulator. Not the aggregator, which by design doesn't verify what it republishes. If the answer is nobody, then the revision clause isn't a technicality. It's a single point of failure wearing a lab coat.

The Verdict

I'm taking the skeptical side. Not because Kalshi did anything wrong. It didn't. It followed its own filing to the letter. I'm skeptical because the structure itself is fragile, and nobody has priced that fragility yet.

The invalidation point sits at the aggregator layer. Historically speaking, every market that scaled on top of unverified third-party data has eventually paid for it. Rating agencies in 2008. Index providers in 2015. The pattern is boring and it repeats. The venue gets big, the inputs get taken for granted, and then someone upstream revises the underlying into a version of reality that never happened.

So here's my first hot take. The most valuable company in prediction markets over the next 18 months won't be an exchange. It'll be whoever sells verified provenance. A firm that can hand a venue a signed, auditable chain from raw survey response to settled contract. That product doesn't exist at scale yet. It will within two cycles, and the first venue to license it will take the institutional flow that's currently sitting on the sidelines waiting for a reason to trust the number.

Second hot take. Kalshi should rewrite that revisions clause before the 2026 midterms, not after. Publish a full chain of custody for every source agency in every election contract. Name the pollsters behind the aggregator. Disclose the audit process and who runs it. Normalize the data providers to a fixed, pre-committed snapshot instead of a living feed that can move. None of this is expensive. All of it's optional today.

Optional is the problem. If a venue only fixes provenance after a scandal, then provenance is a marketing expense, not a risk control. And the traders who actually size positions for a living can smell the difference.

Does any of this kill prediction markets? No. The category is too useful and the regulatory runway is too clear now. Volume keeps climbing. Election contracts are the marquee product.

But the risk premium is real, and it's currently zero. That's never lasted.

The structure mirrors the 2020 setup in one specific way. Everyone assumed the plumbing worked because the lights stayed on. Then they didn't. Kalshi has an opportunity to prove its plumbing works before the next stress test. The calendar says it has about 12 months, and the midterms won't wait for a rewrite.