TODAY'S PAPER
Good Evening
Good Evening
OpinionColumnistsMichael Dobie

When polls get it wrong ... again

Patrick Murray has been around a poll or two. He has been the director of the well-regarded Monmouth University Polling Institute since its founding in New Jersey in 2005.

On Thursday, in the wake of an eye-opening off-year election, Murray did something unusual.

He apologized.

"I blew it," he wrote in a refreshingly candid op-ed for NJ.com. And he said sorry to the campaign of Republican gubernatorial candidate Jack Ciattarelli and to voters for getting the race so wrong.

Monmouth's final poll had Democratic incumbent Gov. Phil Murphy ahead by 11 points when, as we all know now, he apparently is going to squeak by Ciattarelli in a count that continues.

Monmouth's Murray floated the idea of dispensing with election polling late in campaigns "if we cannot be certain that these polling issues are anomalies." He cited decisions by the estimable Pew Research Center and Gallup Poll, among others, to abandon election polling for the way it distracts from their public interest polling, which gauges people's opinions about issues.

Holding a funeral for election polling has merit. That wasn't the only race pollsters got wrong this year, nor the only year polls diverged from reality. The Nassau County executive race was supposed to be a lock for incumbent Laura Curran, Glenn Youngkin was not expected to win Virginia's gubernatorial contest by 2.5 points, and pollsters consistently underestimated Donald Trump's support in 2016 and 2020, and Republican support in general this year.

In polling, trust is essential — people must believe in the veracity of your poll or you're done. It's easy to say the results don't inspire confidence. Pinpointing why is harder.

Certainly, we live in a time when truth has become malleable and truth-telling optional; does that affect what people are willing to tell pollsters?

A survey by online market research firm CloudResearch last year found 11.7% of Republicans and 10.5% of independents said they "would not report their true opinions" in phone surveys about which presidential candidate they preferred. For Democrats, the figure was 5.4%. That's assuming they all were telling the truth about their true intentions. You see where this could lead. A 2020 Cato Institute national survey found that 62% of Americans say they won't say things they believe, fearing others will find them offensive.

Further complicating the pollsters' dilemma are the great shifts in voter preferences underway across America. They are taking place along traditional fault lines of race and ethnicity, fueled by issues like education and culture, and polling has not yet been able to — and might not be able to — adjust.

In an excellent analysis on Substack of New York City's mayoral contest, writer-researcher Matthew Thomas documents a 14-point swing in Asian-dominated precincts and a 30-point swing in Hispanic-dominated precincts toward Republicans in Queens — a process accelerating since he noted it in the 2020 presidential election. A similar dynamic in the Virginia gubernatorial saw Democrat Terry McAuliffe lose more ground in Black and Hispanic-dominated precincts than in white-dominated precincts, compared to Joe Biden's 2020 performance.

Now add in increasing difficulties in modeling expected turnout (i.e., predicting accurately which groups are more motivated to vote) and it might be time to play taps.

It also might be that candidates can get a better feel for races by talking to voters themselves. Suffolk County Legis. Kevin McCaffrey, of Lindenhurst, said he could sense the Republican tide that swamped the county during door-to-door campaigning.

"People were asking, 'How do I get a gun permit?'," McCaffrey told Newsday. "People never asked me that before."

Columnist Michael Dobie's opinions are his own.

Correction: Patrick Murray‘s last name was incorrect in an earlier version of this column.

Columns