Pollsters say they are using a variety of methods to ensure they are not underestimating Donald Trump as poll after poll shows a heated race in various battleground states. are.
If the polls are correct, the battle between former President Harris and Vice President Harris could very well be decided.
But the numbers are viewed with skepticism in some quarters, given that pollsters missed how strong Trump's support was in the 2016 and 2020 campaigns. For Democrats, seeing the close numbers hits the panic button, fearing that Harris will need to lead by a few points to feel safe.
“We've done everything we had to do,” said Charles Franklin, head of polling at Wisconsin-based Marquette Law School. “We were very concerned about that. We made some changes to address this, but we won't really know until the vote counts come back in November.”
Opinion polls have been the standard for measuring presidential elections for decades, dating back to the 1930s.
However, the industry has been shaken by the last two presidential elections, 2016, when Trump defeated Hillary Clinton and was elected president, and 2020, when President Biden won by a narrower margin than some expected. did.
National polls leading up to the election gave Clinton a lead of several points, about the same margin as her popular vote victory, but Trump outperformed expectations for election victories in key states. And even though they predicted the correct winner, the voting averages diverged even further in the main battleground four years later.
But pollsters stress that every election cycle is different and people shouldn't simply assume that Trump will do better than current polls show because of his past. .
John Curvelius, deputy director of the Public Opinion Center at the University of Massachusetts Lowell, said pollsters constantly tweak their methods from cycle to cycle. He said pollsters need to respond to external factors, such as the fact that polls are becoming increasingly expensive, especially if they seek accuracy. At the same time, the country is becoming more closely divided and polarized.
Polling organizations have made some specific adjustments in response to the results of the past two presidential elections.
Curvelius said that in 2016, many pollsters didn't give much weight to results on educational attainment as representative of the population, and while this is not a problem unique to Trump, Republican candidates did well among voters with a college degree. The incident was said to be related to the election of the same year. .
He said the problem could have been an easier solution, but 2020 has shown how pollsters struggle with nonresponse from certain voters, many of whom supported Trump that year. I was doing it.
“The clearest story you can tell about 2020 is that there were certain Trump voters, not all Trump voters. Very specific Trump voters were less likely to answer the phone in 2020. But they were less likely to answer the phone. That's an obvious way to combat that,” Klubelius said.
Franklin said the biggest change in polling methodology since 2016 has been the move from all interviews being conducted over the phone to hybrid sampling, as people are less and less likely to answer the phone. said.
“By 2022, one person will be dialing 100 numbers to answer a phone call,” he said. “It just drove up the cost significantly. And it wasn't the people who refused to be interviewed. They couldn't get their foot in the door because they weren't doing well.”
Markets currently use lists of registered voters to learn if the voters they contact are registered, match them to phone numbers and email addresses, and pollsters can send text messages or text messages to voters. Allows you to send email. As a result, 80% of interviews were conducted online and only 20% were conducted over the phone, Franklin said.
Regarding the specifics of 2016 and 2020, he said pollsters are making a “special effort” to reach pro-Trump areas that have been underrepresented in previous years.
“It's difficult in the sense that you can call or email the right person, but you can't get them to respond. So that continues to be a challenge for the industry as a whole.” Mr. Franklin said.
But he added that samples are improving and that he believes Mr. Market is doing a better job of getting a representative sample.
Curvelius said other methods pollsters are using to address the issue of nonresponse include increasing the number of Republican voters, among others, and using a concept called “emphasizing recall voting.” He pointed out that there were things to do.
This method works by asking respondents how they voted in the last election and evaluating their voting results based on the margin by which one candidate won. But like any tactic, it has some flaws.
Nate Cohn, Chief Political Analyst, New York Times explained Last weekend, a Times/Siena College poll found that this method was not adopted because voters may not remember or misremember how to vote, which is why they lost in the last election. It has been raised that support for political parties may be overestimated.
But Cohn said more polling organizations are using this method, and he believes a more accurate method is possible now that there is more political engagement.
Curvelius said he believes the method is “fairly reliable” if properly weighted.
He said the University of Massachusetts Lowell weights the profile of registered voters in 2020 by their self-reported votes in 2020, but also by respondents' voting intentions, how closely they follow election news, and other factors. He said they also use a voter model based on how they answered the following questions: How often do they say they vote in presidential elections?
“At the end of the day, it's about trying to be as transparent as possible, treating the polls as the beginning of a conversation about the state of the race, not the end of the conversation,” he said.
Some polling organizations argue that the best course is to stay the course, especially those that had small margins of error in 2016 and 2020.
Jim Lee, president and CEO of Pennsylvania-based Susquehanna Polling & Research, said his company is leading Trump voters in polls and is not significantly underestimating them. He said he had made “few, if any” changes to his methodology as it was one of the In 2016, the company conducted polls only in Pennsylvania, but by 2020 it had begun polling nationally.
RealClearPolitics analysis Found Susquehanna polls had the second-lowest voting error on average in both the 2014-2022 and 2020 elections.
Lee said the idea that Trump voters were underrepresented was more valid in the past, when many people were uncomfortable telling pollsters they supported Trump. , claimed to believe it was a “special phenomenon”.
He said that weighting polls to increase the weight of Trump supporters could give Trump a better advantage than he actually does because the method requires pollsters to make “a lot of assumptions.” He said it could become a “Trojan horse'' to show that someone is in a position.
“That obviously puts Mr. Trump in a much better position to ask direct questions,” Lee said. “But there are other companies out there who, like me, haven't really experienced that problem firsthand, so we don't feel like our system is broken.”
Regardless of the outcome, Curvelius said he hopes this year will be an opportunity to have “lively conversations” about polling and methodology.
“We are thinking about democracy, which is the worst form of government apart from other forms of government,” he said. “Aside from everything else, polls are the worst way to measure population weight and public opinion. Learn more from polls.”
