Francesca Hong was supposed to win Wisconsin's Democratic gubernatorial primary by 20 points — she lost by less than one, and the polling industry has yet another corpse on its hands.
This isn't a glitch. It's the business model. The polling industry exists to tell establishment donors and media outlets what they want to hear, dampen populist energy, and manufacture consent for preferred candidates before a single ballot is cast. When the numbers are wrong, nobody in the industry loses their job. When they're right, they're hailed as oracles. Heads they win, tails you lose.
Consider the carnage. In Wisconsin, Hong, a democratic socialist, led pre-election surveys by nearly 20 points, according to CNBC. She lost to moderate Milwaukee County Executive David Crowley by less than a percentage point. In Michigan, Abdul El-Sayed won his Senate primary by just 1 point after polls showed him up double digits on Rep. Haley Stevens. In Minnesota, Lt. Gov. Peggy Flanagan crushed Rep. Angie Craig by nearly 20 points — despite polls showing a tight race. The errors cut in every direction, which means the problem isn't bias toward one faction. The problem is the industry itself.
And it gets worse. NBC News reports that Crowley had actually dropped out of the race on July 8. He quit. Five weeks later, after what NBC calls "a moment of panic set off an unlikely series of events," he was the nominee. The polling industry couldn't even accurately measure a race against a man who had already conceded. That's not a margin of error. That's a margin of fraud.
The prediction markets — the shiny new alternative the tech-savvy crowd keeps pitching — fared no better. Kalshi gave Hong a 95% chance of winning Monday night. Polymarket put her at 96% and called her a "near-lock" in a now-deleted post on X. After the results came in, a Polymarket spokesperson told CNBC the post was "flagged internally for its language" and deleted. Convenient. In Michigan, prediction markets showed El-Sayed with greater than 90% odds before polls closed. He barely survived.
The industry's defenders have their alibis ready. Kyle Kondik, managing editor of Sabato's Crystal Ball, told CNBC that white progressives respond to polls at higher rates, Wisconsin's open primaries make modeling harder, and there may have been late movement in the race. NBC's analysis points to undecided voters and the difficulty of predicting turnout. These are the same explanations offered after every cycle's misses. At some point, "we'll get it right next time" becomes a confession that you're not in the prediction business — you're in the narrative business.
CNBC notes that more money than ever is flowing into the electoral process, underscored by the emergence of prediction markets opening up widespread betting on elections. Follow the money. Polls don't exist to inform voters. They exist to guide donor dollars, set media coverage, and tell the public which candidates are viable before they've had a chance to decide for themselves.
The polling industry missed Wisconsin by 20 points and Michigan by double digits. It missed a race against a candidate who had already quit. It has no accountability, no consequences, and no reason to change. The only question left is why anyone still listens.
