Election Day has come and gone, and once again, millions of people are sifting through the data and trying to make sense of it all.
One big question that remains unanswered: How accurate were the polls this time, and how well could they predict the outcome of the election compared to the prediction markets?
Polling errors in 2024
There’s almost always a polling error in one direction or another. In 2012, for instance, the polls overestimated Republican vote margins by 2.4 points, whereas in 2020, the polls overestimated Democrat vote margins by 3.9 points, according to data from the Pew Research Center. The error in 2020 was also notable as it was the highest in 40 years for the national popular vote and was “of an unusual magnitude,” per a 2022 report from the American Association for Public Opinion Research.
Experts say that they keep all this in mind when trying to determine what’s going to happen during a given election. This year, they say the errors don’t appear too severe. But they did, for the third time in a row, underestimate Donald Trump.
“The polling averages were reasonably accurate in the Sun Belt states,” says Dave Wasserman, a senior editor and elections analyst for The Cook Political Report with Amy Walter. “Most polls had Trump ahead in Georgia, North Carolina, and Arizona for much of the year.” However, “the common thread is that, once again, Trump was slightly underestimated.”
Election betting odds in 2024
While Trump may have been underestimated in the polls, which mostly showed a toss-up between him and Democratic candidate Kamala Harris, that wasn’t necessarily the case in the prediction markets, platforms that include Polymarket, Kalshi, PredictIt, and others.
In the days leading up to the election, those markets showed Trump with better odds to win at roughly 57%.
It’s hard, if not impossible, to ignore the fact that the prediction markets appear to have been more accurate at predicting Trump’s win than the polls were. Given the markets were more bullish on Trump than the polls and pundits, those platforms are now getting a new round of attention.
“Make no mistake, Polymarket single-handedly called the election before anything else,” wrote Shayne Coplan, the founder of Polymarket, in a post on X.
Wasserman acknowledges that betting markets have become a bigger element in elections today. “At this point, I think the betting markets are fairly ubiquitous in the discourse surrounding political analysis,” he says. “I’m not sure how it would inform a voter’s behavior to gauge chances on a betting market, but I do think it will continue to shape perceptions of how a race is moving that may not move nearly in line with the polls.”
Are prediction markets more accurate than polls?
While the prediction markets can be and have been manipulated in the past, if there is reason to think that the markets could be more accurate than the polls, it’s because they should conform to the efficient market hypothesis.
In the financial markets, the hypothesis states that assets are priced according to all available public information. The same logic could hold true for the odds of an election outcome, which differs from polls in that polls are generally only taking opinion and sentiment into account.
However, there have been studies that showed this to not necessarily be true. Others have shown that prediction markets tend to overstate the chances of a low-probability event actually occurring, and even more so when predicting political events. And during some elections, like 2022, the prediction markets have proven to be less accurate than the polls.
So why did the prediction markets seemingly outperform the polls in 2024? There may not be a clear answer yet—other than that sometimes, some people get it right, or they wanted to support their preferred candidate.
“Why did the prediction markets do better? Because people took a side,” says Tom Gruca, director of the Iowa Electronic Markets at the University of Iowa’s Tippie College of Business.
Gruca’s been running and studying an online political-prediction market since the late 1980s, which aims to predict the popular vote winner during presidential elections. Gruca says that, over the years, his model has had an average error of 1.3%—until this year.
“It predicted Harris would win the popular vote at 75%,” he said. “We’ve been doing it since 1992, and got it right every single time. Until Tuesday.”
His takeaway is simple: “The prediction markets don’t always get it right,” he says, noting that there is a lot of research showing exactly that. “Nobody’s going to go to investors and say that these markets are ‘pretty good,’ or that they’re ‘usually better than the polls.’ We know that’s not true,” he says.
As such, there may not be a clear lesson from 2024 relating to prediction markets and the polls. They both had the election as a coin flip, even if the prediction markets were a little more bullish on a Trump win. They may prove more accurate in the future, somehow, but 2024 may have just been a year when the chips fell in the right direction for those platforms.
Gruca says that with that in mind, don’t bet the farm on assuming the prediction markets are providing more insight into a potential outcome than the polls. “People who trade on these should do their homework because you’re speculating,” he says. “I wouldn’t use them instead of your own brain.”