A version of this news analysis originally appeared in the Inside Politics newsletter, out Fridays. Sign up here to get it first to your inbox.
The 2016 presidential election is remembered as a polling bust.
Polls showed Hillary Clinton was ahead, and she lost. Everyone was shocked, perhaps including Donald Trump himself.
But the reality is that 2016 wasn’t that bad of a polling miss. Clinton won the popular vote by 2.1 percentage points, after the final average of polls had her winning it by 3.2%.
What’s forgotten is what a disaster 2020 was.
Joe Biden won the popular vote by 4.5% — but the final polling average had him winning by 7.2%. There were plenty of highly regarded pollsters who were even further off in their final surveys:
Quinnipiac: Biden +11
NBC News/Wall Street Journal: Biden +10
CNN: Biden +12
Ouch.
One polling firm that nailed 2020 was Emerson College. It predicted Biden would win by 5 — 50% to 45%. It’s possible there was some luck involved, of course, but overall Emerson is a highly regarded pollster.
I spoke recently with Matt Taglia, Emerson’s senior director of polling. We talked about how Emerson did well in 2020 and what changes other pollsters are making for 2024.
First off, two items:
- If you do not like a poll, do not criticize it by saying “They are only calling landlines. They aren’t calling cell phones.” Almost all pollsters are calling cellphones, or using cellphones to reach people.
Find a better line of attack.
- Taglia said there is a problem with a Democratic bias in most polls. It’s not that the pollsters are biased, but that it’s selection bias, with Democrats being “more willing to take 15 mins and trust the process.”
Now, on to the discussion.
Taglia said that 2016 was, for many pollsters, “the Stone Age.” He said they were still using random-digit dialing — having a computer generating random phone numbers until they found someone who answered who was a registered voter.
“No one does that now,” he said.
He chalks Emerson’s success in 2020 to what he calls “mixed mode” polling.
His firm doesn’t use live-caller polling for its polls released to the public. That’s too expensive.
Instead, it’s a combination of Interactive Voice Response phone calls, or “robocalls.” Emerson also sends text messages and relies on online panels. The latter is where people are reached by computer.
By law, pollsters can only robocall landlines. That means robocalls are the best ways to reach older voters, though Taglia cautions that in 2024 that doesn’t necessarily equal Trump voters.
He said texting and online panels are Emerson’s “bread and butter.” That allows the group to reach hard-to-find voters: The young. Minorities. Trump supporters.
He said Emerson has been willing to experiment with online surveys –- even if it doesn’t conform to historical best practices.
“We use things that don’t conform to what social science says is valid,” he said. “A lot of that dogma was holding people back. We’re not running some sort of double-blind study. We are conducting surveys for the media.”
Emerson’s most recent poll has Kamala Harris leading Trump 50% to 46%.
Taglia said it’s looking like a repeat of 2020.
“The race is looking very similar,” he said. “We see changes within different demographics that are balancing each other out right now. Trump might be doing a little better among these minority voters, especially Black voters. That’s canceled out by Harris going better with white voters.”