The Media Today

Polling panic is back

September 25, 2023
Jim Bourg/Pool via AP

Yesterday, a pair of polls drove a great deal of discourse in political media. One, from NBC News, wasn’t especially remarkable in isolation—it showed deep voter misgivings about both President Biden and his likely 2024 presidential challenger, Donald Trump, and found that the pair would be locked in a dead heat among registered voters if the election were held today, 46 to 46 percent. The other poll, from the Washington Post and ABC News, also found voter misgivings about Biden. But its head-to-head numbers were astonishing, showing Trump leading Biden by ten points—52 to 42 percent—among registered voters. (ABC led its coverage of the poll with a slightly lower margin51 to 42 percent—reflecting the responses of all adults.)

ABC’s main story about the poll splashed the head-to-head numbers in its headline. The Post’s topline framing was more circumspect: the paper’s headline focused on criticism of Biden’s performance, and while the subhead mentioned the head-to-head finding, it noted that it “does not match other recent polling… suggesting it is an outlier.” The Post reiterated this caveat in the third paragraph of its story on the poll, and again further down, adding that “outlier results occasionally occur in polls due to random error and nonresponse issues, although the political composition of the poll is typical on other metrics.” Despite its headline, ABC also caveated the numbers in an article and on air, noting that the order in which it asked its questions and/or voters’ desire to send a message of dissatisfaction with Biden may have skewed the results. (Nearly a fifth of respondents who said that Trump should be constitutionally disqualified from the presidency said that they would vote for him over Biden.) As the poll reverberated across the rest of the media landscape, a similar split played out: some headlines centered the jaw-dropping numbers, while others centered all the caution around them.

In some quarters, the poll was heavily criticized, as were the Post and ABC for running with it. Simon Rosenberg, a Democratic commentator, said that it should not be taken seriously; Larry Sabato, a political scientist at the University of Virginia, called the poll a “ridiculous outlier” and asked how one could “even publish a poll so absurd on its face,” describing doing so as “a lingering embarrassment.” By contrast, Nate Cohn, the data maven at the New York Times, credited the decision to publish. But he also argued that the two outlets can’t claim the poll is an outlier given that a survey they published earlier this year also showed a significant lead for Trump over Biden (seven points, on that occasion). “If it happens twice in a row in the same race, it’s clear that this is the result of some element of your approach,” Cohn wrote. “You either need to decide you’re good with it and defend it or you need to go home.”

The new Post/ABC poll, and to a lesser extent NBC’s, reinvigorated the endless, if sometimes comparatively dormant, debate about polling in America. This debate is, essentially, twofold, revolving around the methodology and reliability of polls themselves but also the extent to which they should drive the news agenda—considerations that are, for obvious reasons, linked, but not inextricably so, since many critics would like to see even perfect polls take a backseat in media coverage of politics. I am one such critic, as I’ve written often in this newsletter and won’t recapitulate here. But the view that the political press over-covers polls is not to say that they don’t matter at all—they can be a potent, if limited, tool for measuring public sentiment at scale. We thus all have at least something of an interest in the wonkish debates that follow in the wake of polls like the latest from the Post and ABC. And we certainly all have an interest in how the press talks about polls, which matters at least as much as the results of polls themselves.

The polling debate, of course, has raged with particular intensity since the 2016 presidential election, when a narrative hardened firstly that Hillary Clinton would handily beat Trump, and then that the polls had failed to see Trump coming. This was always, at minimum, an oversimplification: Nate Silver, then of FiveThirtyEight, wrote later that national-level polls in 2016 were actually “pretty good” and that reports of their abject failure were “mostly bullshit”; writing for CJR in 2020, Sam Wang, a professor at Princeton, made a similar point, arguing that journalists and others “may have turned an emotional experience with polling into a lasting trauma” and thus learned the wrong lessons. In the 2018 midterms, pollsters performed well; in 2020, they did worse (if not as badly as many pundits concluded), underestimating support both for Trump and Republicans down ballot. Last year, the midterm polls were pretty good again, despite a narrative forming that they falsely forecast a Republican “red wave.” (“It was the pundits who made the red wave narrative,” Silver wrote afterward, “not the data.”) Attending a conference of pollsters earlier this year, Politico’s Steven Shepard found satisfaction with their midterms performance—but also trepidation that 2024 could prove challenging given the past difficulties in accurately measuring support for Trump. Which brings us back to yesterday.

If my historical sketch here is contestable or oversimplified, that’s partially the point: there are polls and the narratives journalists weave from those polls—and polling postmortems and the narratives journalists weave from those postmortems—and these things often do not align. This should serve as a reminder of something I’ve written before: that polling is neither a precise science (as too many media observers treated it in 2016) nor an inscrutable branch of mysticism (as too many media observers, burned by 2016, treated it in 2020). It’s also a reminder that polls are hard to separate from the expectations the media has of them and the ways in which we talk about them—and about politics as a whole. Which, again, brings us back to yesterday.

Sign up for CJR's daily email

One strain of criticism of the Post and ABC held that polls do not merely reflect some abstract notion of public sentiment but rather exist within a media feedback loop: our coverage shapes people’s perceptions of the world, which they then feed back to us through polls, which we then (often) cover as if we played no role in shaping them. If Biden is underwater, such critics argue, that’s because his coverage is currently—and many might argue unduly—awash in negativity, not least around his age and the impeachment probe targeting him over his son Hunter’s business dealings, as I wrote last week. (This time, ABC itself conceded that such coverage may have had an impact on its poll.) This argument has merit—the media does have the power to shape public perception, albeit, clearly, not in isolation. We should be honest about this; either way, we should interrogate whether our coverage of politics is proportionate and substantive. If we have the power to shape perception, though, this is true regardless of the proportionality and substantiveness of our coverage. So the question of how we present polls still applies.

In some ways, the Post (and to a lesser extent ABC) deserves credit for stressing the potential limits of its latest poll, an expression of the type of humility that media critics often demand of major outlets. But ABC, at least, ought to have gestured to the caveats in its online headline, which gave those who didn’t click through no indication that the poll was an outlier. And both outlets afforded significant prominence to the poll—the Post on its print front page; ABC at the top of its Sunday-morning show This Week—which, as various critics pointed out, sent its own message, intentional or not, as to how they viewed the worth and reliability of its findings.

Indeed, the idea of prominence gets to the core of the wider problem here. A blanket refusal to publish outlier polls strikes me as untenable; such a move would arguably itself constitute a lack of humility. Deciding not to publish them on a case-by-case basis is more defensible; reasonable observers can disagree as to whether the Post and ABC should have pocketed their latest poll or not, and I, for one, lack the expertise to adjudicate among them. But a better solution—and here I must, sadly, break my earlier promise not to recapitulate—would be to reduce the weight that we in the media put on polls across the board, a move that wouldn’t just lower the stakes of a given poll turning out to be unreliable, but would be a good first step in weaning political journalism off of the horserace as its central organizing principle, as I’ve so often encouraged in this newsletter. Again, this isn’t to say that we shouldn’t cover polls—or the horserace—at all. But we could de-emphasize both. And we could uncouple them. “A hypothetical vote-preference question 14 months before an election is predictive of nothing; it’s best seen as an opportunity for the public to express its like or dislike of the candidates,” ABC wrote yesterday. But we have other means of divining the latter, without the distracting headline focus on the former.

And timing matters here, perhaps above all else. I argued last week that flooding the news cycle with 2024 chatter is inappropriate when Biden still has a third of his term to go, for multiple reasons. When it comes to obsessing over polls, we can add to those that however good a pollster’s methodology, they can’t—and don’t claim to—predict all the things that might change in the fourteen months before people actually vote. Yesterday, Silver noted that it’s “insane” to sweat every general election poll right now, adding that “it’s going to be a long time before any poll ought to meaningfully move you off whatever priors you have.” At roughly this stage in the 1984 election cycle, polls showed John Glenn, the astronaut turned senator and Democratic presidential candidate, besting Ronald Reagan, while Walter Mondale was tied with Reagan. Glenn didn’t win the nomination. Mondale, the eventual nominee, won a single state.

Other notable stories:

ICYMI: Is Murdoch sticking around or bowing out? Yes.

Jon Allsop is a freelance journalist whose work has appeared in the New York Review of Books, Foreign Policy, and The Nation, among other outlets. He writes CJR’s newsletter The Media Today. Find him on Twitter @Jon_Allsop.