Will Pollsters Get It Wrong (Again) in 2022?

People holding ballots in blue and red

Getty Images

Here’s what they’re saying about the midterms — and a look at whether we should trust them.

All eyes are on November’s midterm elections amid a flurry of hotly contested races from Georgia to Nevada, some of which are still too close to call. These razor-thin margins are, of course, putting even more pressure on pollsters to get it right — but could their gauge of public sentiment be another swing and a miss? 

After all, the country’s faith in polling just hasn’t been the same since former President Donald Trump’s shocking upset against Hillary Clinton in 2016, when pollsters systematically underestimated Republican support. Four years later, the final polls in the 2020 presidential election did correctly predict Joe Biden’s victory — but they also overstated his strength. This topsy-turvy success rate may feel like a recent development, but then again, pollsters say accurately measuring Americans’ political opinions has always been a bit like the Wild West. 

“We’ve always been shooting at a moving target,” says Lee Miringoff, who’s the director of the Marist College Institute for Public Opinion. “The campaigns don’t stop when you do a poll.” 

So where does polling stand now, and have there been any improvements? Well, the answer might be a bit more nuanced than you think. 

What are polls saying ahead of the 2022 midterm elections?

As campaigns hit the home stretch, Republicans are still expected to take over the House, while it’s a “dead heat” for the Senate, according to FiveThirtyEight’s “lite” prediction model, which is based on the latest polling data.

Meanwhile, a New York Times/Siena College poll released Monday found the GOP with a slight advantage overall in the last few weeks leading up to the midterms. The findings show that 49 percent of likely voters said they planned to vote Republican in the fall, compared to 45 percent who said they planned to swing Democrat.

Of course, these predictions are far from certain. As The New York Times has noted, Democrats’ strongest numbers have tended to be in states that saw the greatest polling misses in past elections.

Wisconsin’s Senate race is a prime example: Republican incumbent Ron Johnson was favored to win re-election from the get-go, but back in August, a Marquette Law School poll showed his Democratic challenger Mandela Barnes with a seven-point lead — almost the same spread that polls inaccurately claimed Biden enjoyed over Trump in the state in 2020. Biden did ultimately win Wisconsin, but the actual vote was much tighter than projected, and his victory came with a lead of less than one percentage point. And in the last two months, polling in the Senate race has readjusted to better match that outcome, with Johnson in the lead with 52 percent support, compared to Barnes’ 46 percent. 

What issues are currently driving voters?

Democrats have been betting big on the likelihood that threats against abortion rights will motivate their base on Election Day, but the reality is the economy and rising inflation have become a much more potent political issue. Overall, 82 percent of Americans named inflation as an extremely or very important issue, compared with 56 percent who ranked abortion as a top worry, according to an October poll by Monmouth University

This is even true among some women, who accounted for the biggest shift towards Republicans in the latest New York Times/Siena poll. Currently, Independent women favor Republicans by 18 points, compared to backing Democrats by 14 points in September. Robin Ackerman, a 37-year-old Democrat in Delaware, told the Times that while she disagrees “1,000 percent” with the Supreme Court’s decision to overturn Roe v. Wade, she’s leaning Republican because “they’re more geared towards business.” 

Pollster Don Levy tells us this thinking reveals “a thick partisan prism” through which issues like abortion and the economy are seen among many Americans.  

“Democrats haven’t solved the Rubik’s Cube of this belief across America that Republicans are better on economic issues, and the economy continues to be the number one issue to most Americans,” says Levy, the director of the Siena College Research Institute, which conducts regional, statewide and national surveys. 

Are pollsters making the same mistakes?

If the polls are wrong again, it may not come as a surprise, and that’s because the polling industry hasn’t been able to figure out where it exactly has gone wrong in recent elections.

The American Association for Public Opinion Research, which is comprised of more than 2,000 public opinion professionals, said it was “impossible” to determine why Biden’s support was overstated in the 2020 election, largely due to the lack of reliable information on key data like demographics, opinions, and vote choice of those not included in polls (either because they were excluded or chose not to participate).

Even now, some pollsters have said they’re seeing the familiar signs of “nonresponse bias” — an insider term for when people don’t respond to a poll — creeping back into their work. For example, a survey sent out on a new app may only reach younger people while overlooking older voters, or it might ask potentially embarrassing questions about private information that make many people unwilling to respond. Some Americans are also less inclined to take surveys due to their ideology, as seen with the “hidden Trump voters” who didn’t openly express their support for the Republican candidate in 2016. Miringoff points out that this group tended to have less education, something that also threw off that year’s polling, which focused heavily on those with college degrees.

But Levy says pollsters have become much more aware of this pool of voters, and the Siena College Research Institute has taken certain steps in their post-data collection analysis to accommodate for this group, though he admits they pose an “interesting challenge.”

How do pollsters measure their findings?

On top of everything else, pollsters are still relying on outdated modes of conducting their surveys, such as calling landline phones. Though Levy says this was once considered the “gold standard” of gauging public opinion, only about 37 percent of American households still use landlines. But he adds that pollsters have been trying to diversify their toolbox by turning to online surveys, text messaging, and calling voters’ cell phones, but even these methods aren’t without challenges. 

“Many people are reluctant to answer their cell phones from a number that they’re not familiar with, and some have software installed on their phone that will block a call from someone unless that person is in their contacts,” says Levy. 

What else do people often get wrong about polling?

Pivoting their methods isn’t the only thing on pollsters’ agenda. They’re also contending with a long-held misunderstanding among the general public about what polling is actually supposed to do.

While Miringoff says most people think of polls as a way of predicting the future, they’re meant to be “snapshots in time.” So polling can’t predict what ultimately affects an election’s outcome, especially when it comes to things like turnout or an unexpected scandal, like the one that rocked GOP candidate Herschel Walker’s Senate campaign in Georgia.

Many people also confuse forecasting with polling, but there’s a key difference. Forecasting predicts the future based on an aggregate of the latest polls — think RealClearPolitics and FiveThirtyEight. So if a certain poll within that collection is leaning hard a certain way, Miringoff says it could throw off the prediction.

Polling, on the other hand, measures a single data set — and while it does account for factors like voters’ education, it captures voters’ anticipated behavior at a fixed point in time. The trouble, according to both pollsters, is that the narrative in 2016 was primarily driven by forecasting, a model that was more optimistic about Clinton’s chances than polling. (The latter had her likelihood of victory around 80 percent, leaving the door open for the Trump win that did come to pass.)

Then there’s the margin of error, an important factor that’s commonly overlooked in discussions about polling. The larger that margin of error is, the less confidence one should have that a poll result would reflect the census of the entire population.

The bottom line

It’s clear that polling is imperfect, but the people who do the work find value in it beyond the predictive aspect that fuels headlines in the media. At the end of the day, Levy hopes that polling sheds some light on the political landscape and fosters a better understanding of one another amid bitter political divisions that can hinder conversation.

“One of the great mistakes that most people make is they tend to believe that everybody else believes what they believe,” says Levy. “By talking to people, we share the beliefs, hopes, dreams, behaviors, and attitudes of the broad swatch of America.”