The danger of reading too much into individual polls

As time went on, polling averages (like that of 538) didn’t indicate much movement. And on Wednesday, the Times published the results of its new poll in Siena: Trump had gained ground, thanks in part to Republicans rallying around him.

The shift since the April national poll is modest. Then Trump polled by 1 point among likely voters; now he polls by 4 points. There are many caveats, including that we’re looking at likely voters and focusing on the head-to-head polling question rather than the one that includes third-party voters. But that’s part of the point: The story here is more about uncertainty than certainty.

Consider the shifts among different demographic groups since the Times poll in April. At the time, Biden’s soft support among younger voters was the focus of the Times’ reporting. But that 2-point Biden advantage among those under 30 is now 19 points — in line with the 24-point lead he had in 2020. Perhaps this is a function of the Biden team reaching younger voters, or perhaps it’s an indication that his base (as his team has promised) is coming home. Or – probably – it is a function of the way subgroup polls can fluctuate among relatively small populations.

Are the shifts significant among the youngest voters, ages 45 to 64, or among black voters? Do they show real movement? Or are they errors in a poll that the Times poll has suggested could be an outlier?

None of this is to say that the Times poll—or any other individual poll—isn’t useful and insightful. Instead, I want to point out that there are several ways in which there are incentives to elevate specific polls in ways that can be misleading. Trump’s celebration of that Daily Mail poll is one, picking up on a clearly suspect result that signals change. Extracting an interesting tidbit from a single poll is another. Tidbits from reliable polls are valid, but require caution and context. (In the interest of full disclosure, I must admit that I have sometimes been less cautious than perhaps warranted.)

With these caveats in mind, it’s useful to think about the ways in which the Times-Siena poll captures Trump’s enthusiasm advantage. There were several questions surrounding the fervor of support for each candidate, including favorability, concerns about age and whether each party’s presumptive candidate should actually be the nominee. Trump did better on every question, both overall and with his own party.

Trump’s overall numbers are, of course, determined in part by his fervent support. Trump voters make up about half of the respondents, and their enthusiasm therefore colors half of the sample. This is also not a surprising result. Biden’s support has always been less energetic, with voters in both the 2020 and 2024 polls often indicating that their planned votes for Biden largely reflected their opposition to Trump.

There’s another consistency in the new Times poll: A large share of Republicans and people who plan to vote for Trump in November believe he committed a serious federal crime. A poll earlier this year showed a similar result.

This is not a reflection of Republican indifference to Trump’s conviction. That is real; Nine in 10 Republicans said they viewed the charges against the former president as politically motivated. There’s an ironic “I’m voting for the felon!” a rallying cry that ripples through his base and reflects skepticism about the charges. But the above result is simply indifference to Trump’s actions, an acceptance among some supporters that he actually committed a crime.

So true do the racing mode? Poll averages tend to be a more reliable way to see how things have changed. The Washington Post has one of its own, which uses high-quality national and state-level polling to calculate who voters currently favor in the presidential race. Instead of looking at one poll, it looks at many polls, a better way to track trends and positions.

Not much has changed this year.

There is a lot of information included in individual polls, which provides many opportunities to detect apparent shifts and surprising results. But the state of the 2024 presidential race is what it has long been and what it will almost certainly remain into November: close and unpredictable.

Lenny Bronner contributed to this report.