Why it's difficult to predict the presidential race between Trump and Harris

The fear of the 2024 election is great. Both sides know that their candidate could lose – and they want surveys to show them how afraid they are.

Currently, surveys provide us with the clearest answers to these questions. But there is a problem: every year the polls become more difficult. We pollsters face three key challenges that threaten accuracy all political polls. No one has solved it, and it's not clear that anyone can.

Here's what we're dealing with:

Challenge #1: Almost no one wants to talk to pollsters — and those who do might be weirdos.

Polls are based on a simple idea: If we talk to a miniature representative version of a state or country, we can estimate what the entire state or country thinks. It's like taking a sample in an ice cream parlor: a well-mixed spoonful tells you what a whole ice cream cone tastes like.

But it is becoming increasingly difficult to reach the people needed to build this mini-country or mini-state. Response rates – how often people are willing to take a cold call or answer a text message from us – have been falling for decades. Response rates for the Pew Research Center's telephone surveys fell from 36% in 1997 to 6% in 2018. Nate Cohn of the New York Times reported a response rate for his surveys in 2022. And any pollster will tell you that Response rates are also low this year.

A lack of answers makes good data rare and expensive. Surveys are costly in part because we spend so much money sending unanswered text messages or making calls that go straight to voicemail. And as polls become more expensive, media organizations will either sponsor fewer polls or opt for polls that reduce costs through savings.

But even if a group can afford to take a survey, non-response creates huge potential data problems.

If only one in 100 people takes part in a survey, pollsters have to make statistical adjustments. Some – like finding the right demographic mix – are easy. When a pollster simply cannot reach enough Latino, working class, young or rural voters, they often give underrepresented voters the opportunity to vote for them did Add a little extra weight to your calculations. Weighted polls give each population group a more precise say, even if some groups were harder to reach.

Other adjustments are not so easy.

Suppose a pollster has the right demographic mix in their survey But For lack of a better description, nerdy rule followers are mostly interviewed. This pollster may be missing out on fickle, anti-establishment Trump voters — and risks undercounting Trump votes for the third straight year.

It is almost impossible to adapt to this type of problem directly. The census helps us calculate how many 18- to 34-year-olds should take part in a survey, but not how many weirdos and nerds. So pollsters have to get creative with the math – which leads to another problem.

Challenge #2: We have to be prepared for no one talking to us. That's risky.

The most common response to this problem — a lack of pro-Trump and anti-institutional Republicans in the 2020 polls — is to weight by “recalled votes.” Essentially, pollsters are asking people how they voted in 2020 and trying to include the right number of Trump and Biden voters in their sample.

Although I've used this tactic in a few surveys, there are downsides too. Respondents do not always correctly remember who they voted for. Any estimate of how many Trump or Biden voters will vote again in 2024 is just that – an estimate. The list goes on.

However, many reputable pollsters say that weighting based on recalled votes has improved the accuracy of previous polls. And pollsters who weight only by party — and not the votes recalled — may not fully address the problems that damaged the industry's credibility four years ago.

There is no right answer. Everyone uses math to adapt to the sad fact that normal people don't take surveys. And every pollster is nervous because if we make the slightest mistake, we will be punished for years.

Challenge #3: The elections are closer than ever, so “the polls” will almost certainly be “wrong.”

The last truly overwhelming presidential victory was the re-election of Ronald Reagan in 1984. The last 40 years have seen the most consistent presidential elections in living memory. That's bad news for polls – which are blunt instruments rather than accurate predictions.

When a pollster surveys the electorate at random, he or she may, through no fault of his or her own, accidentally select a few too many voters from one side or the other. When we try to vote for an upcoming election, we make (fallible) predictions about who will vote in the election and who won't. And there is a lot more uncertainty—from non-response, weighting decisions, and more—that is simply not easy to convey to the lay reader.

In a race this close, with Harris and Trump dying in poll after poll even in the swing states, a good pollster could get everything right but still miss the result by a point or two and face the ridicule of a huge pollster for years Audience facing readers.

But we pollsters can't look at these problems and scream, “That's not fair!” and go home. I have built election prediction models and have seen firsthand that polls are the best tool for predicting elections. More importantly, they are only Ability to ask the public their opinion on an issue in real time.

The problems with the survey are real. Perhaps the non-response or some other problem will eventually become intractable and lead to a catastrophic, industry-wide collapse. But if that doesn't happen, we have to keep doing the polls because nothing else is as good as polls.