Media must stop covering polls it knows are unreliable

Last month's primaries handed the polling industry three separate humiliations in three separate states. Michigan's Senate primary broke sharply from what the surveys showed. In Wisconsin, Milwaukee County Executive David Crowley won the Democratic nomination for governor over Francesca Hong despite polling that favored her. And in Minnesota, the final KSTP/SurveyUSA poll before the primary showed Angie Craig with a two-point lead over Peggy Flanagan in the Democratic Senate primary. Flanagan won with 58.8 percent of the vote, not a narrow miss but a landslide the final poll never hinted at. The commentariat is now doing its usual hand-wringing about what went wrong. I would ask a different question: good, now what? The mechanics of the failure are not mysterious. Almost nobody answers a call from an unknown number anymore, which has gutted the random-sample telephone methodology pollsters relied on for decades. The pool of voters willing to talk to a stranger about their politics skews older, more institutionally trusting, and less representative with every cycle. Layer on a record share of the electorate that identifies as independent, which makes likely-voter modeling close to guesswork, and a market flooded with cheap online-panel polls competing to publish first rather than accurately, and you get exactly what Michigan, Wisconsin, and Minnesota just produced. Letter: Election results show Wisconsin turning from purple to blue I want to be precise about what is actually breaking here, because it matters. Survey research as a scientific method, the careful, well-funded, methodologically rigorous instruments political scientists like me have relied on for generations, is not what just failed. What failed is commercial horse-race polling: cheap tracking surveys commissioned by campaigns and newsrooms for one purpose, generating a number that can anchor a headline. That distinction should not let my own discipline off the hook entirely. Political science has leaned on survey data as its primary window into public opinion for decades, and a genuine crisis in survey methodology is a genuine problem for how we study politics, not just how campaigns cover it. But the discipline can adapt its methods. A media ecosystem built around the horse race cannot simply keep publishing numbers it knows are unreliable and call that journalism. Which is why I think the honest reaction to this month's failures is not mourning. Horse-race polling was never journalism about the stakes of an election. It was a cheap, easily quantified substitute for it, a number requiring no reporting, no candidate scrutiny, no explanation of what a Senate seat or a governorship would actually change in people's lives. If the numbers can no longer be trusted, newsrooms lose their excuse to lead with them, and are pushed back toward covering what campaigns are actually about. That does not mean we are left with nothing to measure. In two recent projects, one analyzing how consumer habits track voting behavior, the other how county-level health outcomes predict it, I found that revealed-preference data, what people actually buy, where they actually live, how their community actually fares, carries a political signal that self-reported survey data increasingly cannot. Polls were wrong: Wisconsin election results stun as humans wreck the machines | Opinion It is not because grocery receipts are magic. It is because this kind of data does not depend on a stranger picking up the phone, does not degrade as trust in institutions declines, and cannot be rushed out cheap to win a publishing race. It is a complement to polling's failures, not a replacement for polling's function, but it is a sturdier one. Michigan, Wisconsin, and Minnesota did not just embarrass a polling industry. They handed American political journalism a chance to stop outsourcing its coverage of elections to a number nobody can any longer defend. That is not a crisis. It is an opening, and one my own field would do well to take as seriously as the press should. David Schultz is the Winston Folkers Endowed Distinguished Faculty Chair, Hamline University Department of Political Science, Department of Legal Studies and Department of Environmental Studies. This article originally appeared on Milwaukee Journal Sentinel: Media must stop covering polls it knows are unreliable | Opinion