Three Networks Are About to Put a Number On Screen That 39,820 Wallets Made Up

Editor's note: TrueEdge builds odds tools and earns affiliate commissions from licensed sportsbooks. This piece argues a position on prediction-market regulation. Weigh it accordingly.

Reuters and Ipsos asked 4,506 American adults whether people should be able to wager money on the outcome of an election. Sixty percent said no. Thirteen percent said yes. Twenty-four percent did not know and three percent skipped the question. The poll ran September 30 to October 5 on Ipsos's probability-based KnowledgePanel and published Monday.

The margin of error depends on who is telling you. Fortune, which was given the data, says plus or minus 1.5 points. The Reuters wire says "about two percentage points." The Hill says 2.3. Those are three different numbers for the same survey of the same 4,506 people, and we do not know which is the questionnaire's own figure. Nothing in this piece turns on the difference, but a poll's margin of error is the one thing everyone quoting it should get from the same place.

The number is bipartisan in a way almost nothing is: 61 percent of Republicans against, and 69 or 70 percent of Democrats depending on which write-up you read. Seventy-five percent call prediction-market wagering gambling; 8 percent say it is not. Seven percent of respondents have ever used one. Opposition to wagering on US military action came in at 65 percent and on government policy announcements at 63 percent, against 8 and 9 percent in favour.

Two further numbers from the wire version deserve more attention than they got. Asked about wagering on professional sport, the same sample split 43 percent opposed to 29 percent in favour — still a plurality against, but a different country from the election figure. And asked whether it is appropriate for members of Congress or their staff to bet on political events, 93 percent of Democrats and 89 percent of Republicans said no. That is the most unanimous thing in American politics, and it is the question nobody regulating this has answered.

Set all of that aside for a moment, because public disapproval is a weak argument and we are not going to lean on it. Majorities have been wrong about plenty, and "most people find this distasteful" is not a regulatory standard. The real problem with putting Kalshi's prices into an election broadcast is not that voters dislike betting. It is that the prices may not be measuring what a viewer will think they are measuring.

What a probability on a chyron implies, and what produced it

CBS News said on September 30 that it would run Kalshi's real-time contract prices alongside its own Battleground Tracking in midterm coverage. Kalshi has similar arrangements with CNN and CNBC, and it licenses vote counts and race calls from the Associated Press. CBS News and CNN now share an owner: Skydance closed its purchase of CNN's parent, Warner Bros. Discovery, on October 6. At announcement, Kalshi's markets had Democrats at 63 percent to take the Senate and 91 percent to take the House.

A viewer who sees "91%" next to a network's logo reasonably infers a crowd. The Anti-Corruption Data Collective went and counted the crowd. In an August 17 analysis covering 7,466 markets on Kalshi, Polymarket and Polymarket US, with data through August 10, the group found that just 39,820 Polymarket wallets had traded any 2026 congressional race at all. Eighty percent of those markets had fewer than 100 wallets in them. Exactly ten markets in the entire cycle had crossed 1,000 participants — roughly the sample size of an ordinary political poll.

Concentration on top of that thinness: the top 1 percent of wallets accounted for 68 percent of volume, and ten wallets alone for 17 percent. Those ten had positions in 426 of the 470 seats on the ballot, which is 91 percent of the races in the country. ACDC's David Szakonyi put the obvious risk plainly — concentrated betting "poses a number of threats to election integrity," manipulation and insider trading on things like endorsements among them. His colleague Michelle Kendler-Kretsch made the narrower point that matters for broadcast: the combination of low participation and concentrated volume "makes prediction markets hard to trust" as a signal.

Total 2026 midterm volume had already passed the entire 2024 cycle by August 10 — $133 million against $92.4 million — with ACDC projecting $1.6 billion for the year. Volume is going up. The number of people is not going up nearly as fast, which is another way of saying the average position is getting bigger.

Kalshi's objection, which is half right

Kalshi told Fortune the poll is rigged. "It's not hard to bias a poll and this poll is biased," a spokesperson said. "Ask leading questions about whether people support 'betting on war,' and you'll get the answers this poll arrived at. Kalshi doesn't even offer war markets." The company asked for research that "measures familiarity, uses neutral wording, and reports all responses."

Fortune checked the questionnaire. The word "war" does not appear in it. The instrument described prediction markets as letting people "wager money on the outcome of real-world events" and then asked separately about elections, military actions, government policy decisions and sport. So the specific accusation is wrong, and it is wrong in a way the company could have avoided by reading the thing.

The general complaint lands, though, and we will say so. The poll did not measure familiarity, and familiarity is low: a March Ipsos survey for the American Institute for Boys and Men found 21 percent of Americans were very or somewhat familiar with prediction markets. Asking four-fifths of a sample to render judgment on a product they cannot describe produces a measurement of instinct, not of informed opinion. "Reports all responses" is a fair ask too. Ipsos did not respond to the criticism before Fortune published.

Here is why the objection does not rescue the position. Kalshi's case for being on television is that its prices are better information than polls — more current, financially disciplined, harder to fool. That claim does not depend on the public liking it, and it survives the poll being imperfect. It does not survive 80 percent of the relevant markets having fewer than a hundred traders. The exchange is asking to be treated as a measuring instrument. An instrument's credential is its calibration, not its popularity, and on the 2026 congressional markets specifically, nobody has shown the calibration.

The part that is genuinely in Kalshi's favour

Dustin Gouker, who writes The Event Horizon, gave Fortune the most useful framing anyone has offered on why election markets are small while sports markets are enormous. "I don't think it's necessarily that election trading isn't popular," he said. "It's more that it doesn't feature some of the things that we see in these markets that are seeing more volume." A poll moves an election market occasionally. A touchdown moves a sports market in four seconds. Combined monthly volume on Kalshi and Polymarket went from about $26 billion in May to about $53 billion in July, per a September Pew analysis, and sport drove nearly all of it.

That is a structural explanation, not an excuse, and it cuts in Kalshi's direction on the question of intent: thin election markets are a product of the calendar, not of anyone hiding the ball. Polymarket, for its part, says more than 90 percent of its visits are people looking at markets rather than trading them, and frames the product as information rather than wagering.

Gouker also offered the correct test and the correct warning in the same breath. The test: "When a market says something should happen, it happens roughly the amount of times we would expect it to." The warning, delivered after he pointed to Abdul El-Sayed's Michigan primary — where markets had him an overwhelming favourite near the close of voting and he ended up leading by just under a point when the AP called it — is that "you have to look at them as probabilities and not as these crystal balls."

Nobody is going to say that on air at 9 p.m. on election night.

What we would actually ask for

Not a ban. A label, and a specific one.

If a network is going to render a contract price as a probability on screen, it should render the two numbers that tell a viewer whether to believe it: the number of distinct accounts holding a position in that contract, and the share of volume held by the largest ten. Both are computable from exchange data that Kalshi and Polymarket already have, and both would have told a CBS viewer in August that most congressional markets were being priced by fewer people than fit in a lecture hall. A network that will put a market's output on screen and not its participation count is doing the one thing it was built to avoid, which is sourcing a number it cannot characterise.

We would say the same about a sportsbook's closing line, and we say it about our own tools. The difference is that nobody watching an NFL broadcast mistakes the spread for a measurement of the public will.

One more number, because it belongs in this story and almost nobody has put it there. At the foot of CNBC's October 9 piece on Adam Silver sits the network's own disclosure: "Kalshi and CNBC have a commercial relationship that includes a minority investment and customer acquisition." Three of the outlets carrying Kalshi's election prices this autumn are CBS News, CNN and CNBC. Two of those now share an owner. One of them owns equity in the exchange. The disclosure is there, in small type, at the bottom of the article — which is more than we would have got five years ago, and considerably less than a viewer needs.