All too often when their work goes awry, election pollsters turn to feeble excuses such as “polls really aren’t predictions. They’re just snapshots in time.” That is nonsense, because no pollster takes on the hard work of conducting election surveys with the expectation that those surveys will be embarrassingly wrong.
A refreshing departure from the blame-dodging tendency came the other day, following the outcome of the New Jersey governor’s race in which the incumbent Democrat, Phil Murphy, narrowly defeated Republican Jack Ciattarelli. The result represented another setback for pollsters, whose pre-election surveys collectively estimated Murphy’s lead at nearly 8 percentage points, according to RealClear Politics. The Monmouth University poll pegged Murphy’s advantage at 11 points.
Murphy won by fewer than 3 points.
Two days after the election, Patrick Murray, director of the Monmouth poll, publicly embraced his survey’s error, resorting to none of the grudging defensiveness that’s common when pollsters address their failings. That made Murray’s acknowledgement of responsibility as remarkable as it was candid. “I blew it,” he declared in a commentary posted at NJ.com. “The final Monmouth University Poll margin did not provide an accurate picture of the state of the governor’s race.”
Murray apologized in the next paragraph to the Murphy and Ciattarelli campaigns, noting that “inaccurate public polling can have an impact on fundraising and voter mobilization efforts. But most of all,” he added, “I owe an apology to the voters of New Jersey for information that was at the very least misleading.”
It was a statement of regret of a kind the polling industry seldom sees. In a way, Murray’s comments evoked the admonition of Warren Mitofsky, the innovative director of surveys for CBS News who died in 2006. “There’s a lot of room for humility in polling,” Mitofsky said in 1998, at the 50th anniversary of election polling’s greatest debacle, President Harry Truman’s stunning victory over Republican Thomas E. Dewey.
“Every time you get cocky, you lose,” Mitofsky added.
Cockiness was rampant in 2016, when polling-based forecasts anticipated Hillary Clinton’s easy victory over Donald Trump. One model, for example, said Clinton had a 98.2 percent chance of winning the presidency, while declaring that Trump had “essentially no path to an Electoral College victory.”
Polls in key Great Lakes states that year largely failed to detect surging support for Trump’s candidacy, and surprise victories in Pennsylvania, Michigan and Wisconsin became his path to an Electoral College victory.
National polls last year collectively turned in their worst performance in 40 years, systematically understating support for Trump. He lost the popular vote to Joe Biden, but not by the blowout margins that several prominent polls had anticipated.
Murray’s Monmouth poll was not the only survey to underestimate Ciattarelli’s support in New Jersey’s gubernatorial race. The Fairleigh Dickinson University poll estimated Murphy’s advantage at 9 points as the campaign neared an end. The Rutgers University’s Eagleton Center poll estimated Murphy’s lead at 8 points late in the campaign.
The New Jersey outcome is the latest sign of a profession in trouble. Election polling may have become too unstable to be consistently accurate, in part because response rates among would-be respondents are very low.
In his post-election commentary, Murray noted that prominent polling organizations such as the Gallup Organization and Pew Research have given up election surveys in recent years.
“Perhaps that is a wise move,” Murray wrote. “If we cannot be certain that these polling misses are anomalies then we have a responsibility to consider whether releasing horse race [poll] numbers in close proximity to an election is making a positive or negative contribution to the political discourse.”
It is an interesting perspective, suggesting that polling organizations have a social obligation to accuracy. Even so, it is difficult to imagine a large-scale retreat from election polling.
After all, the threshold to entry is fairly low, given that online polling in 2020 was the technique most employed by pollsters, as described in a report by a task force of the American Association for Public Opinion Research. “It’s a lot easier for people to do polling … using online methods,” Joshua Clinton, the task force chair, said during a recent online presentation. “Anyone with a thousand bucks can be a pollster now and do an online poll with a thousand respondents. … And then if you get it right, maybe you’re a superstar. Or not.”
Another reason to believe election polling will not fade away, despite its problems and predictive failures, is that its pedigree is so extensive. Reports about crude pre-election polls were published in U.S. newspapers as long ago as 1824. Since at least the 1940s, news organizations have relied on polling to shape and drive their campaign coverage. Their troubles notwithstanding, election polls remain central to how journalists interpret a campaign’s prospects and its ebb and flow.
But when polls misfire, media narratives about the election likely are in error as well.
W. Joseph Campbell is a professor of communication at American University and the author of seven books, including, “Lost in a Gallup: Polling Failure in U.S. Presidential Elections.” Follow him on Twitter @wjosephcampbell.