Sign up for the daily CJR newsletter.
G. Elliott Morris has worked on political forecasting—the process of predicting election winners through polls and other information—for much of the past decade. Most recently, he led the now-defunct FiveThirtyEight. Now he runs his own polling and analysis site, Strength in Numbers. As news organizations strike partnerships with betting markets and use their odds to forecast events, Morris knows as well as anyone that it’s dangerous to claim to predict the future—especially during election season, when readers and journalists are especially susceptible to misinterpreting data. “When you give people probabilities for electoral outcomes, you’re giving them something to play with,” Morris told me. “We do degrade part of the political process when we do that.” Betting markets such as Polymarket and Kalshi, he said, increase that degradation. “I think it is incredibly irresponsible,” he said. “I find this uncritical wholesale acceptance of their numbers as the true probabilities for the world very bad.”
At one point, predicting election results seemed relatively straightforward. Nate Silver’s 2008 and 2012 predictions at FiveThirtyEight proved to be remarkably accurate. “Everybody was like, ‘We found the secret sauce to predicting elections,’” Nathaniel Rakich, the managing editor of Votebeat and a previous senior editor at FiveThirtyEight, told me. But Trump’s unexpected 2016 win triggered a massive crisis across the field—and a corresponding decrease in trust in pollsters, forecasters, and the media at large. Prediction markets’ partnerships with newsrooms are now claiming to offer certainty.
Usually, polls aim to measure public sentiment and political forecasts provide probabilities based on polls or other inputs, but betting markets claim to forecast the likelihood of events based on anonymous traders wagering real money. The relationship between responsible journalism and the still-emerging world of betting markets, however, is contentious. In a recent interview with CJR, Dustin Gouker, a gambling expert and writer of the Event Horizon newsletter, argued that prediction markets deserve a place in journalism, with caveats, as a way to renew interest in reporting and supplement traditional forecasting with real-time information. But trading on platforms like Polymarket and Kalshi poses problems unlike those that journalists—even ones accustomed to reporting on public opinion—have encountered before. “It’s not just conventional wisdom, it’s the conventional wisdom of the type of people who put bets on elections,” Rakich said. “That’s certainly not a representative sample, the way a poll is.”
Betting markets greatly differ from statistical predictions that journalists traditionally report on. In 2022, the Tow Center researched what came to be known as predictive journalism, which at the time included polls, forecasts, and other traditional ways to estimate outcomes. Our report asked, “How can journalists responsibly incorporate predictions into their work?” It argued that predictive journalism requires user-centered design, stringent editing, and an emphasis on interpreting uncertainty and probabilities. For example, polls or forecasts are often reported along with their margins of error (which Morris, in his book Strength in Numbers, says should be reported more often at the top of stories). These metrics aim to show the uncertainty inherent in making predictions about large groups of people. Prediction markets instead report a single probability. Reporters can dig into the total amount of money traded or suspicious trading activity, but as of now, there’s really no way to audit why markets say what they do.
There are also ways to probe weaknesses and potential biases with forecasts and polls: journalists can check whether firms have track records of supporting certain parties, accepting funding from certain groups, or using flimsy methods. News organizations have begun covering insider trading in betting markets, though it’s a much more opaque process.
Even traditional election forecasting is not a very useful tool for readers, although forecasts can help communicate the uncertainty inherent to polling, Morris said. Research shows that people have difficulty interpreting probability in election contexts, and journalists can be complicit in oversimplifying top-line numbers. “People see a seventy percent chance, and they round that up to a hundred,” Rakich said. Morris said that, thanks in part to problems like these, “this is something I wrestled with a lot in my forecasting work after 2020 and 2024.” Morris’s Substack analyzes and administers political polling, often to understand the state of politics and elections. The site notably lacks live probabilities that any candidate will win. (Another project cofounded by Morris, FiftyPlusOne, will have live election probabilities, but, he said, “we are de-emphasizing the odds.”)
Morris stressed that the business model of betting sites is based on gambling and engagement, primarily with people wagering on sports. (In 2025, sports markets accounted for almost 90 percent of the trading volume on Kalshi.) “We like the latest shiny toy; we like the appearance of objectivity from the data,” he said. “As an exercise in optimized marketing, this has been one of the most successful feats of advertising in human history.”
Journalists should remain vigilant about their coverage of anything related to predicting the future—especially if there’s money on the line. “There are responsible and irresponsible ways to cover it,” Rakich said. “A lot of it is on the journalist to be a smart consumer of it.” Morris had a starker opinion. “Absolutely no one has to be reporting this information,” Morris said. “We are all doing it because we are suckers.”
Has America ever needed a media defender more than now? Help us by joining CJR today.
