Pollsters are once again facing a familiar foe: Donald Trump.
Every year Trump has been on the ballot, the polling industry has miscalculated. Trump defeated Hillary Clinton to win the presidency in 2016. And, after four years of trying to figure out what went wrong, surveys in 2020 were even worse. Trump campaigned significantly more competitively against now-President Joe Biden than pre-election polls indicated.
After mostly nailing last year’s midterm elections, pollsters are breathing a sigh of relief. However, the Trump presidency has been a different affair.
And now, as Trump’s advantage over his GOP primary competitors grows, pollsters are concerned about a segment of the electorate that has made accurately measuring Trump’s support nearly hard.
“It’s looking a lot like Trump is going to be on the ballot” in November, Democratic pollster Andrew Baumann said. “So it’s all coming back with a vengeance.”
Trump isn’t some supernatural force. The issues are practical. In 2020, he attracted a sizable number of people who had rarely, if ever, voted and who were either not included in polls or refused to participate in them. Trump slammed polls that constantly showed him lagging Biden. This produced a feedback loop that made his supporters even less likely to reply, further skewing the polls.
Baumann was among the attendees and presenters at this week’s annual conference of the American Association for Public Opinion Research, a yearly gathering of pollsters from academia, media, and campaigns.
For decades, that organisation has been debating the future of political polling. Looking at the polls from the last two federal election cycles is enough to give you whiplash. By most accounts, last year’s midterm elections were the apex of election polling. According to FiveThirtyEight’s post-election analysis, polls were more accurate than in any year since 1998.
But that was two years after the previous presidential election, when national polls were the worst in 40 years and state polls were the worst in recorded history.
The threat of yet another polling disaster hangs over the industry’s ongoing efforts to modernise its processes.
CNN made some of the most substantial adjustments to their approach among public pollsters. The network and its polling provider SSRS randomly chose street addresses for its nationwide polls and mailed out solicitations to complete a poll online or by dialling a number, largely abandoning its long-standing technique of random-digit phone sampling.
Some people were polled from a file of registered voters and reached either by email or telephone, depending on the best contact information available, for much of their state polling, which necessitates faster turnaround times to poll statewide contests like those for governor or Senate. Others were drawn from SSRS’ existing panel of voters who indicated they were registered to vote.
The results were astounding. CNN polls accurately predicted the winner in eight of the nine key statewide contests they surveyed, with the exception of the Nevada Senate race, and half of the candidate vote shares were within a single percentage point of being correct.
“We were within the error margin on just about every [poll] we did,” Jennifer Agiesta, CNN’s director of polling and election analytics, said. “So I’m pleased with how these turned out.” That gives me some hope for the time between now and 2024.”
However, Agiesta believes it is too early to predict whether the same issues that hampered pollsters in 2020 will reappear.
“I don’t think that [Trump’s] comments on polling and the way he presented his views on polling to his supporters were helpful in terms of response rate in 2020,” said Agiesta, who also began a one-year term as president of the pollsters’ organisation at this week’s conference. “But I’m not sure if that will be the case in future elections.”