This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

trusted source

proofread

The presidential election is one year out. Why predictions are more reliable than polls when picking a winner

polling station
Credit: Pixabay/CC0 Public Domain

In 2015, Northeastern University associate professor of political science Nick Beauchamp predicted that Donald Trump had a 25% chance of winning the next year's presidential election.

"People I know who were not fans of Trump would be aghast and say 'That's ridiculous, how can you possibly say that?'" Beauchamp recalls, noting that Trump was then considered somewhat a "joke candidate" among the eight Republican and two Democratic candidates.

But Beauchamp's prediction—while prescient—was primarily based not on polling but odds. Trump was then leading the Republican primary, so he had a 50/50 chance of winning that … and then he would be in a two-way race, so again had a 50/50 chance to win the presidency.

A year out from the 2024 , Beauchamp is again looking at odds more than polling.

"Presidential polls are usually considered to be unpredictive this far in advance," Beauchamp says. "As you get closer to the , the accuracy of the polls goes up and the error goes down, and right now we're in the band where the error is equivalent to flipping a coin."

Costas Panagopoulos, distinguished professor of political science at Northeastern, concurred.

"Polls tend to get more accurate as elections approach, so it stands to reason that current polls are not necessarily determinative of what will happen on Election Day," Panagopoulos says.

Rather, he says that polling should be considered as a "useful snapshot in time."

"One of the more interesting features of polls is what they suggest about trends, or dynamics over time rather than the specific distributions of the poll," Panagopoulos continues. "And one of the more interesting features of polls is to examine what's happening over multiple polls, and how they suggest dynamics, how things might be changing, trends."

And David Lazer, distinguished professor of political science and computer sciences, says that "historically, this far out, it's just not predictive."

"Is it telling us zero? I wouldn't go that far," Lazer says of polling. "But it's not telling us much."

Political polling is "an art as much as a science," Panagopolous says, and much has been written about the hand-wringing among pollsters who didn't foresee Trump's victory in 2016 as well as the surprises from the 2020 presidential election. In fact, Panagopoulos found evidence that 2020's final national presidential pre-election polls were less accurate than in any cycle since 1996.

But while Beauchamp isn't paying a lot of attention to polls right now, there are some "fundamentals" that he is considering. Fundamentals—and their caveats—that make 2024 perhaps even more of a coin toss.

"Often the way it works is if you want to make a prediction a year in advance, it's better to use the fundamentals—as they say—than the polls," Beauchamp says.

The first fundamental?

"Right now, if you're going to make a prediction, the easiest is that incumbents usually win," Beauchamp says. "Although now we've had three times in recent memory when the incumbent has lost, if your recent memory goes all the way back to Jimmy Carter."

Beauchamp says other fundamentals include things like the economy and macroeconomic conditions, gross domestic product numbers, unemployment, consumer sentiment and general truisms such as that in midterm elections, the party out of the White House tends to do better.

Of course, there are caveats. For instance, a great economy amid economic pessimism, a 2022 "red wave" that fizzled, and special elections where Democrats have done better than expected… perhaps most notably in the fight for a Wisconsin Supreme Court seat and flipping the Pennsylvania House.

"The big question is whether these special elections are predictive," Beauchamp says. "Nobody really knows the answer to that, I think."

Lazer also notes that even these major fundamentals can change.

"This far ahead, one can look at structural things, the state of the economy and so on," Lazer says. "But the economy can change in a year's time."

Then there are recent trends that may or may not continue. Will the overturning of Roe v. Wade continue to motivate voters? What about the changing demographics of the Republican and Democratic electorates?

Add in other unknowns such as Election Day 2023's outcomes, and it makes sense that Beauchamp is looking to fundamentals rather than Ipsos or Gallup.

That being said, we pretty much know who will be at the top of the tickets. And with political polarization trends, are there really that many people who are swing voters? Then, what about Cornel West and RFK Jr.?

"I think minds are more made up this time than in recent history because this is the first time we have had a rematch in some time," Beauchamp says of the expected matchup between Trump and President Joe Biden. "Because it's a rematch and because everybody knows these two people pretty well, the polls are more predictive than usual."

Lazer concurs… to a point.

"Trump and Biden are more known quantities than is typical, so that's the only thing that might suggest it is more predictive," Lazer says. "But then again, people voted differently in 2020 then saying they are going to vote now."

Moreover, yes, the 2016 national polls missed Trump's victory; but they were only 1 percentage point off.

"Most polls by Election Day come very close to what ultimately happens," Panagopoulos says. "That's sometimes challenging when elections are so close, or even when polling data suggests a dead heat, but that doesn't make polls inaccurate—it means what ultimately happens on Election Day could shift on Election Day for all sorts of reasons."

Nevertheless, Beauchamp says pollsters have made changes in hopes of better accuracy.

"In recent elections it's been considerably better, even when they're getting it wrong, we've just got elections that are close," Beauchamp says.

And so with all these unknowns, caveats and trends, what are you to do when the latest poll comes out?

Panagopoulos points out a couple of things it's important to consider when looking at any recent poll.

"Americans consume polling information the way they consume other bits of political information—selectively or through the prism of their own political biases," Panagopoulos says. "Not everyone consuming poll information is a statistician; for most people it's just a piece of information that they take into account and people sometimes will dismiss polling information altogether if it is inconsistent with their political views."

Moreover, he says that polling methods differ among firms, which can be important.

"It's not just a matter of surveying a bunch of voters to see how they feel," Panagopoulos says. "You have to not only see what people feel, but who will vote on Election Day."

So, Panagopoulos advises looking at polls to find trends; not to find who's going to win.

"Take the estimates with a grain of salt, and focus more on trends than point estimates," Panagopoulos says. "Polls can be predictive, they are not necessarily determinative."

Meanwhile, Beauchamp is, again, thinking in odds.

"My guess is that the polls are not that far wrong, but that means it is basically a coin toss," Beauchamp says. "Because the polls say 'who knows?' They're both likely to win."

And, amid all this, Lazer notes that there's no precedent for a likely nominee facing charges in four criminal cases.

"It's likely there will be excessive weirdness in the American political system over the next year," Lazer says.

Citation: The presidential election is one year out. Why predictions are more reliable than polls when picking a winner (2023, November 7) retrieved 7 March 2024 from https://phys.org/news/2023-11-presidential-election-year-reliable-polls.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

Explore further

Pre-election polls in 2020 had the largest errors in 40 years

1 shares

Feedback to editors