Why Public Opinion Polls Don’t Necessarily Predict Votes

Public opinion polls can be great snapshots of what the public at large thinks about certain issues. However, a large chunk of the public doesn’t vote. Even in the contentious 2020 election, only 66.8% of eligible Americans voted. The US Census Bureau noted that this election had the “highest voter turnout of the 21st century.” 

In an interview with Stephen Colbert, Jake Tapper attributed the difference between polling and vote outcomes to the gap between the public and the voters. That’s born out in votes on various forms of gun control. The New York Times showed public polling that suggested 83-91% of the public’s support for background checks in four states in 2016. Votes on those measures ranged from 48-59%. Public polling consistently overestimates support for ballot measures that offer changes that the public supports.  

However, these disparities aren’t only about who shows up to vote and who stays home. The same New York Times story points out that voters can support an idea but not its implementation. Someone could support sports betting in Colorado but could’ve voted against its legalization because the initiative failed to properly fund problem gambling programs. It’s harder to get a group to agree on a specific solution than it is to get a group to oppose change.    

The gap between the public and the voters is part of the reason that public opinion polls should not be used for predictions about voting patterns. Instead, political bettors should take them into account to correct for overconfidence in their models. 

What Public Opinion Polls Can Be Good For 

If political bettors know that votes don’t match up with public opinion polls, they can get themselves ahead of the public that consumes these polls like sports updates. If political bettors can spot overconfidence in different markets, they can take advantage of it. 

PredictIt’s state and local markets are better places to account for the gap between public polling and voting outcomes. It’s easier to study the impact of targeted campaigning than to predict the outcomes of the midterms or presidential elections. State and local politics also have more potential to rise above partisanship, offering chances to bet against conventional wisdom. (They’re uncommon, but not so rare that studious bettors can’t find potential examples in key states.)

For ambitious political bettors, there’s no getting around building out a betting model flexible enough to make it from one election season to the next. Bettors will have to find their own betting systems. But they can at least avoid the mistake of taking public opinion at face value.