A Race Call Has an Author. A Market Price Doesn't.

Nine days before the midterms, Brazil ran the test American newsrooms needed to watch. On the eve of its October 25 runoff, Polymarket priced Flávio Bolsonaro at about 84 cents on the dollar, The Rio Times reported, and Lula won with about 51.8 percent of the valid vote. That miss alone proves little. An 84 percent favorite still loses about one time in six, and nine of the last ten polls had Bolsonaro ahead too. The evening itself deserves the attention. Lula's price crossed 50 cents more than an hour before the official count put him in front. The paper could not explain two earlier moves, one of them while polls stayed open, and it said it could neither prove nor rule out manipulation. At one point, about $41,000 of selling would have knocked 19 points off Bolsonaro's price, in a race where AFP had already counted about $185 million in contracts traded. Afterward, the Center for Democracy and Technology reports, prominent political figures pointed to those earlier odds as evidence of fraud.

American television entered this election tied to the same companies. Bloomberg's rundown lists business deals with Kalshi or Polymarket at CNN, CBS News, CNBC, Dow Jones and Fox Corp., and the terms vary widely. Bloomberg reported that odds would stay off CNN's Magic Wall and unmentioned on air that night. Fox said its Kalshi deal "excludes any political coverage." CBS named Kalshi its prediction market sponsor and planned to feature the data on a companion stream called The Nerd Show. CNBC discloses a commercial relationship that includes a minority investment. Kalshi says "our partners have full control over what and how they report," and I take that at face value. My concern sits elsewhere, in the gap between two kinds of figure now competing for one screen.

A race call has authors. The AP calls a contest only when the trailing candidate can no longer catch up, Reuters explained before the election, and Decision Desk HQ requires unanimous agreement from its team plus sign-off from a named data chief, Scott Tranter. Those systems sometimes fail, in public. The AP pulled a 2018 call in California's 21st District. DDHQ withdrew one in last year's Seattle mayor's race. That record counts in their favor, because somebody made each decision, somebody reversed it, and somebody can explain both. A price offers none of that. Traders stay anonymous, nobody publishes a method, and no one issues a correction. Chris Stirewalt of NewsNation calls the odds "lagging indicators, not leading indicators." Brazil suggests they can lead, and for a newsroom that makes life harder. An unexplained jump ahead of the official tally hands a fraud theory exactly what it wants, and the University of Washington's Center for an Informed Public warned that swings during counting "could be seen as evidence for election fraud."

A disclaimer will not fix this. In a survey experiment funded by the Brennan Center, with 1,994 registered voters and hypothetical candidates, a 67 percent Kalshi price made respondents 10 points more likely than a 50 percent price to name that candidate the likely winner, about the same pull as a poll, and warning them about the markets "had no impact." It measured expectations rather than votes, and one experiment should not carry more weight than that. Local races add a second problem. The Anti-Corruption Data Collective, a project of the Fund for Constitutional Government, read the order books in September and found that more than nine in ten congressional midterm markets could move ten cents for under $1,000. Those include the House seats local stations cover. My own work on cable news with Valerie Martinez-Ebers and Aida Ramusovic found networks moving with or against rivals depending on the story, so I would not assume any one network's wall outlasts a competitor's ratings.

The rule I would hand any news director ahead of the next election night, the local ones most of all, fits in a sentence. Before a figure reaches the screen, ask who stands behind it and who corrects it if it goes wrong. If you cannot name that person, it stays off. Keep prices out of results graphics, where viewers will read them as a count. When a swing becomes the story, report what moved it before you report that it moved. Put any commercial tie in the same breath as the price. CBS drew a usable version of that line at its local stations, where a DraftKings deal covers election night and its odds stay out of newscasts. In an October NORC poll for Medill, only 32 percent of Americans reported high confidence that news organizations would get the midterm results right. With that little trust in the bank, the name behind the call remains the one asset a station owns outright.