What if they held an election and the polls didn’t blow their predictions? That just happened in the 2022 midterms.
House Minority Leader Kevin McCarthy’s hopes of becoming speaker are in jeopardy after a “red wave” predicted by some pundits and partisan firms — and not independent pollsters — went pfft. With results still being counted, the House of Representatives looks likely to turn Republican by only a narrow margin, and control of the Senate may once again hang on a runoff election in Georgia.
Maybe it shouldn’t have come as a surprise.
Independent pollsters were largely on target in key races in the last few weeks, offering some vindication for an industry vilified after failing to accurately predict the outcome of several recent election cycles — perhaps most notably, Donald Trump’s victory over former secretary of state Hillary Clinton in 2016. There were exceptions, such as in Florida, where polls underestimated Republicans’ winning margins. But this year, the bigger problem wasn’t the surveys: it was that many pundits and political experts failed to understand what those polls were saying.
“Polling was very much spot-on,” said Mike Noble, chief of research and managing partner of OH Predictive Insights. Independent polls from his group and others projected slim margins in Senate races in states like Arizona, Georgia and Pennsylvania. “The whole ‘red wave’ was manufactured by partisan polls and Super PACs. The independent polls were calling a lot of those races as dead even. And they were.”
Red wave wrong
Many political observers were stressing a red wave was coming in the week ahead of the election, a time when polls traditionally tighten across the board.
“The whole reason we have polls is to have data instead of anecdotes,” said Ashley Koning, director of the Eagleton Center for Public Interest Polling at Rutgers University. “We’re all so traumatized by past problems with polls that I think people were maybe too distrustful of them and put too much stock in anecdotes, which we do need to explain what happened in elections. But you really have to put the data first.”
The key breakwater against a red wave was that unlike past polling in recent elections, undercounts of expected Republican voters did not befuddle the independent polls. “Polling organizations took to heart the need to address the problems we saw in 2020, and many of them made changes to try to better represent Republicans and Trump supporters,” said Scott Keeter, senior survey adviser at the Pew Research Center. Votes are still being counted, “but the initial impression is encouraging.”
Reputable polling firms deliver their findings with transparency about how they weigh voters and with margin-of-error windows that are too often ignored in news coverage of polls, Koning added. “Polls are not meant to be predictive but to give a snapshot of where things stand when they were taken. We really should take more care in talking about them.”
Noble sees the growth in newer, more partisan polling firms attached to “super PAC” political action committees that need optimistic poll results to hit up donors for more cash as undercutting more reliable polls. Those polls, which can unrealistically weight their results to favor one party in his view, end up polluting the average of poll results reported by political websites like FiveThirtyEight or RealClearPolitics.
He pointed to Utah, a very conservative state where a Democratic PAC’s poll earlier suggested the Senate race between Republican incumbent Mike Lee and Independent candidate Evan McMullin was a toss-up. OH Predictive Insights ended up running a survey only a week before voting just to be sure — and it found Lee trouncing his independent opponent. Which is what happened.
“You really have to look at reputation of the polling firm and ask who they are working for in assessing a poll,” said Noble. (He compared complaints about polls to complaints about “the media,” which lump partisan news outlets in with serious ones.)
Columbia University statistician Andrew Gelman, who regularly analyzes polls and polling questions for problems, said 2022 looks more like a typical year rather than an unusual one, where voters tend to rebalance the government in midterms, penalizing the president’s party. With the end of a national right to abortion reminding voters that the Supreme Court is in the hands of Republicans, shifting Congress halfway toward Republicans looks more like old-fashioned midterm voting. (Kaiser Family Foundation surveys suggest about two-thirds of Democratic voters mentioned that abortion ruling as factoring into their turnout and vote.)
“It will take some time to figure out which pollsters did particularly well, but preelection polling averages that included a lot of partisan pollsters tended to be less accurate than those without the partisan polls,” said Keeter. Averaging polls can help tamp down idiosyncrasies across them, he said, a practice seen at FiveThirtyEight and the New York Times. “But this election shows that a good polling consumer needs to pay attention to what polls are included.”
That can be hard because some states see polling droughts ahead of elections, acknowledged Koning. Where traditional polls are not available, people should be very cautious about those states. “There is always room for improvement in polling, but it is really too difficult for people to sift through who is credible and who is not.”
Thanks to Alicia Benjamin for copy editing this article.