A study by the Anti-Corruption Data Collective has put a number on how thin most political prediction markets are. As reported by the Telegram channel Podslushano v Gamble, the researchers examined 11,011 active markets on US congressional races across Polymarket, Polymarket US and Kalshi. By their calculation, as of 1 September, in 94% of those markets a sum below $1,000 was enough to move the quoted price by 10 cents — the equivalent of a ten-percentage-point swing in implied probability.
Where the liquidity actually is
The authors put the cause down to low activity rather than a flaw in the mechanism itself. On more than three quarters of the markets studied, total trading volume never reached $1,000. At the same time, a handful of high-profile races attracted tens of millions of dollars. The picture is a familiar one to anyone running a sportsbook: a few headline events carry the book, while the long tail of listings sees a trickle of trades.
Prices that did not come back
The report also describes hundreds of cases on Polymarket where a quote, after trades from one or two wallets, did not return to its previous level within 24 hours. The authors' concern is reputational as much as financial: such moves can create the appearance of rising support for a candidate, particularly if the shift is picked up and reported by media.
Polymarket and Kalshi disputed the conclusions. Their position, as reported by the same source, is that a distorted price is an opportunity for other traders to profit, which gives them an incentive to push the quote back towards a justified level. The study and the platforms are, in effect, describing the same mechanism and disagreeing about how reliably it operates in practice.
A product question, not just a political one
Prediction markets are increasingly pitched to the iGaming sector as a "wisdom of crowds" vertical. The study suggests that on most listed events there is no crowd to speak of. For an operator considering the vertical, that reframes the question. The headline number — 94% of markets movable for under $1,000 — is not an argument against the format; it is a measure of how much of a catalogue is genuinely price-discovered and how much is nominally listed.
The same logic applies to any long-tail catalogue, whether it is lower-league fixtures or niche casino content. Listing breadth looks good on a landing page, but a market nobody trades is a market whose price says little. Our note on game-mix optimisation covers how to read engagement across a wide lobby; the principle transfers directly to event catalogues.
What this means for operators
- If you are building or integrating a prediction vertical, treat liquidity on the long tail as a design problem from day one: decide which events you will actively make markets in and which you will list but not promote.
- Be careful about surfacing thin-market prices as "the odds" in marketing or content — a quote that one or two wallets can move is a weak signal to put in front of a customer.
- Expect regulators and the press to keep asking how prices are formed; documenting your market-making and monitoring policy is easier before launch than after a headline. Our guide to multi-jurisdiction product design covers building that documentation into the product from the start.