Often in error but still seductive: Why we can’t quit election polls

Polls showed Joe Biden, right, holding double-digit leads over Donald Trump, left, in the run-up to the 2020 election, but he won election by only 4.5 percentage points.
AP Photo/Patrick Semansky, File

Their record is uneven. They misfired in one way or another in the past three presidential elections. And yet the prevalence of election polls is undiminished. Thirteen months before the 2024 election, polls are many – and inescapable.

Why is that? What explains polling’s abiding appeal despite its performance record?

The reasons go beyond facile analogies that election polls are akin to weather forecasts in offering a fluid, if sometimes contradictory, sense of what lies ahead.

Polls and poll-based forecasts are not always in error, as I noted in my 2020 book, “Lost in a Gallup: Polling Failure in U.S. Presidential Elections.” Their enduring appeal rests in part in offering a sense of data-based certainty that can be irresistible, especially to journalists who tend to value precision in a field awash with ambiguity.

It is hardly surprising, then, that news organizations also are polling operations. Ties to election polling live deep in the media’s DNA.

A cartoon published two weeks before the 1948 election, in which Dewey was expected to win the presidency by a wide margin – but didn’t.
Clifford Kennedy Berryman, Artist/National Archives, Records of the U.S. Senate, 1789 – 2015

A deep affinity for polls

The Literary Digest, an American news magazine published weekly from 1890 to 1938, conducted massive mail-in polls early in the 20th century and built what was widely called an “uncanny” record for predicting presidential elections accurately – until it failed utterly in 1936.

The magazine’s poll that year was based on returns of more than 2.3 million postcard ballots and estimated an easy victory for Republican Alf Landon over incumbent President Franklin D. Roosevelt, a Democrat.

The Literary Digest miscalled the election by nearly 20 percentage points. Landon won just two states in one of the most lopsided presidential elections in American history.

Within two years, the Literary Digest was absorbed by Time magazine.

Major news organizations these days figure prominently in election surveying. CNN, The Economist, Fox News, NBC News, The New York Times, The Wall Street Journal, The Washington Post and Yahoo News – among others – all conduct or commission preelection polls. Results of media-sponsored polls are shared with the built-in large audiences of the respective news outlets, exerting whatpolling expert David Moore has called “a major influence” on public perceptions of an election.

Indeed, media polls – notably CNN’s and the final preelection survey conducted jointly for NBC News and The Wall Street Journal – showed Joe Biden holding double-digit leads over Donald Trump in the run-up to the 2020 election, encouraging expectations of a landslide. But the outcome was no rout. Biden won the election by 4.5 percentage points.

Forgetting history

Another reason why election polls endure is that memories among both the public and pollsters about past failures tend to be fleeting.

By nature, opinion research and journalism are forward-looking pursuits. Their practitioners tend not to dwell on, or often recall, how poorly election polls fared in 2020, for example, by collectively overstating Biden’s prospects of victory.

Recollections have likewise dimmed about how polls in key states in 2016 mostly failed to detect decisive, late-campaign shifts in Trump’s favor. Or how in 2012 the venerable Gallup Poll erred in consistently giving Mitt Romney the advantage in preelection matchups with President Barack Obama.

Memories, inevitably, are even dimmer about the polling failures of 1936, 1948, 1952 and 1980.

Failure notwithstanding, polls long ago became central features of the theater of American politics. Their results help sharpen and give dimension to the competitive drama of national elections. A tightening race or a building landslide can be at least mildly suspenseful and diverting.

Polls command attention because they “give the public and the candidates some inkling of the electorate’s thinking,” as media critic Jack Shafer has noted.

Moreover, polls have become useful both to Democrats and Republicans in winnowing bloated fields of presidential candidates, most of whom have no chance of being nominated. Both parties in recent years have imposed thresholds of polling support for candidates to qualify for party-sponsored debates early in the cycle.

The 2024 presidential election campaign has offered a historically unusual reason to pay attention to the polls, even ones this early.

Not since 1912, when Theodore Roosevelt split the Republican Party in a failed attempt to return to the White House four years after he had left the office, has a former president campaigned to reclaim the office. Trump’s candidacy for reelection no doubt has encouraged closer than customary attention to the polls, especially as they have indicated he maintains enormous leads over rivals for the Republican nomination.

The Literary Digest, an American general interest magazine, conducted massive mail-in polls early in the 20th century.
Screenshot, eBay

What else is there?

Significantly, election polling benefits from a what-else-is-there attitude among journalists, opinion researchers, historians of public polling and politicians that no other reasonably accurate options exist in sampling the public’s views and attitudes.

Extensive interviewing by political journalists, an earnest technique called “shoe-leather” reporting, occasionally has been tried by news organizations seeking an alternative to reliance on polls. But such experiments have produced little success.

A notable advocate of shoe-leather journalism was Haynes Johnson, a political reporter for The Washington Post who died in 2013. Johnson was a poll-basher, asserting that they represent “no substitute for hard reporting” and bluntly acknowledging, “I hate the polls.”

In the weeks before the 1980 election, when President Jimmy Carter sought a second term, Johnson wrote lengthy preelection articles based on many hours of interviews with Americans in places as diverse as Boston, San Diego and Youngstown, Ohio. As the campaign neared its end, Johnson was asked on the PBS program “Washington Week in Review” who was positioned to win. “I think somehow that Carter is going to slip through” and be reelected, he replied.

A few days later, Carter lost to Ronald Reagan in a landslide.

So, if you want clues to what the electorate is thinking in an election campaign – and almost everyone seems interested in such knowledge, from voters, journalists and pundits to donors, campaign workers and candidates – “shoe-leather” journalism won’t cut it. Counting candidates’ yard signs won’t cut it, either. Estimating crowd size at campaign rallies is seldom very revealing. Odds produced by betting markets may be intriguing but aren’t consistently reliable.

For want of a better alternative, society is stuck with sample surveys, a popular, familiar, but often-fallible way of estimating outcomes in high-stakes elections.

W. Joseph Campbell does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.