Could Prediction Markets Replace Election Polling?

Many eyes are on Maine and a race that could determine which party controls the U.S. Senate after the midterms. 

For an idea of who will win this key seat, you could look at a recent poll of likely voters that showed 49% would pick Troy Jackson, the Democrat, while 46% said they would vote for incumbent Republican Susan Collins.

Or you could look at Polymarket, where bettors have put the likelihood of a Democrat taking the seat at 68%.

Thanks to relaxed legal barriers and convenient apps, prediction markets including Polymarket and Kalshi have made placing wagers on election outcomes as easy as sports betting. Proponents of the markets claim that collecting this wisdom of crowds—asking people who they think will win, rather than who they will vote for—gives more accurate prognostication than traditional polls and pundits. 

News organizations are certainly taking notice, with CNBC, CNN, and The Wall Street Journal all signing partnerships with the markets to take advantage of their forecasts.

Yet Brian Schaffner, Newhouse Professor of Civic Studies in the Department of Political Science and Tisch College, predicts that prediction markets won’t supplant polls as we know them.

As the co-principal investigator of the Cooperative Election Study, Schaffner has spent years capturing and analyzing public opinion. He told Tufts Now that from what he knows about the American electorate, prediction markets aren’t as in sync with voter minds as some believe.

Do you think prediction markets could make polls obsolete?

I don’t think they are going to remove the use of polling at all. These markets are probably informed by polls more than anything else. I see them as downstream from polling. 

Why would market predictions on elections differ from what polls show? 

People who play in these things often think they know better than the pollsters, and that’s fair enough, because polls aren’t perfect. Bettors might be looking at things like fundraising or how candidates are doing in particular states or insider information from Trump’s teleprompter guy. Or if a big scandal comes out, they may try to get ahead of the polling and bet that the scandal is going to hurt that person’s standing. That’s why the markets fluctuate much more than polling does.

But those investors probably overestimate the significance of any given news story on everyday voters. Bettors can get swept up in that.

John Phillips, CEO of betting platform PredictIt, said on Politico that traders are “incentivized with cold, hard cash to separate the emotion, to make a bet with their head rather than the heart,” making markets more accurate than polls. Does that ring true to you?

Some people are just playing around investing $20. Other people are investing lots of money. If the argument is that people who bet large have a lot of expertise, I would say that many people overestimate their own expertise. 

There was a famous case in 2018 where a person who ran a polling firm on the Democratic side was basically gambling against Democrats in a bunch of races. He bet tens of thousands of dollars in these prediction markets and lost a lot of money. So here’s someone whose experience in polling should be really informative, but they probably overestimated their own ability over what their polls were telling them. 

Where in politics can prediction markets be helpful?

I think where the markets are useful is in races that aren’t being heavily polled, or where polling is not as reliable. For example, primary polling is often much more error prone, in part because people don’t pay attention to primary candidates until very late, and it’s harder to know who exactly is going to vote in the primaries. In those cases, I suspect that prediction markets probably can do better than polling.

If we don’t have polling on something, I would rather have Harry Enten on CNN pointing to prediction market movement than hear some random political consultant’s opinion. 

Media organizations are increasingly including market predictions in their political coverage. What should readers know when they see those statistics?

Prediction markets show their data as probabilities—the bets placed on a race may give a candidate an 80% chance of winning, for example. But people struggle to understand probabilities in a balanced way. 

If you tell someone one candidate’s leading another candidate by 5 percentage points, as a poll might show, that seems like a fairly close race. 

But a 5-point margin across a bunch of polls could give that candidate a really high probability of winning, say an 80% chance. When people see that, they think this candidate is obviously going to win. However, if you present the race in terms of margins, they’re going to tamp down their expectations a bit.

Do you think prediction markets could actually influence elections? If, say, a handful of big bettors make it seem a candidate is highly favored to win, voters might then figure that candidate has merit?

I’m pretty doubtful. How many people really are paying attention to prediction markets? I think they are the people who are super interested in politics—the political hobbyists who are following politics all day long. These are people who are going to vote and already know who they are going to vote for. The people who don’t pay that much attention to politics, including people who are on the fence or unsure who to vote for, are probably not paying any attention to prediction markets either.