Given all the recent criticism of election polling, it may be hard to believe that in the 2020 presidential election, polls correctly predicted the outcomes in 48 states.
In a sports or business setting, that kind of success rate would likely earn endless praise. But in a polarized political climate, two misses out of 50 opens the door wide enough for endless criticism — which is why some in the polling industry are calling for a change.
“We have to really rethink what we're doing because it's just simply increasing the public mistrust,” Monmouth University Polling Institute director Patrick Murray told NBCLX.
Faith in polling has fallen in recent years, with former President Trump routinely deriding their accuracy. But Democrats haven’t exactly championed polling either, and Murray says politicians of all stripes have bashed the science when the findings are unpopular, further undermining trust in polls.
In 2016, public polling averages predicted Hillary Clinton would beat Donald Trump by three points in the national popular vote. She won by two points but lost the Electoral College. And for the 2018 midterms, polling did exceptionally well to predict the Democratic gains in Congress.
Murray is frustrated by news organizations boiling complex surveys down into simple headlines, with only limited context or explanation of the poll’s limitations, such as its built-in margin of error.
“They take all this uncertainty out there and make it seem like there's more certainty by doing that,” he said. “My question as a pollster is: What am I contributing to the dialog here, when I ask 15 questions and you only report one of them?”
"The least precise thing" about election polls
Monmouth was one of the only major institutions to conduct polling ahead of the 2021 gubernatorial race in New Jersey, a race that polls suggested would be won handily by the Democratic incumbent, Phil Murphy. But Murphy won by just three points, and not the 11 points Monmouth projected a week earlier.
“I blew it,” Murray wrote in an op-ed for NJ.com after the election, blaming his “likely voter” modeling, which tries to predict who will turn out in an election, a necessary step in order to predict the outcome of an election. It's a challenge that has tested pollsters across the country the last few presidential cycles.
Murray says in order to restore trust in public opinion polling, it may be time to stop publishing election polling, also known as “horse race” polls, which try to determine which candidate has a lead in a given race.
Those polls are often sensationalized and oversimplified, he says, due to a political landscape that treats every development as a win or a loss for Democrats and Republicans.
“The media just really obsesses on that margin between candidate A and candidate B, and that is the least precise thing that polling ever does,” Murray said.
He says public opinion polling still remains quite accurate, helping economists understand America’s consumer habits, helping governments track labor trends and helping politicians understand what policies voters want them to prioritize.
But when consumers rely on news aggregation sites that share polling headlines without delving into the poll’s context, as well as its limitations, the trust of the entire industry suffers.
“Most public pollsters are committed to making sure our profession counters rather than deepens the pervasive cynicism in our society,” Murray concluded in his November op-ed. “If election polling only serves to feed that cynicism, then it may be time to rethink the value of issuing horse race poll numbers as the electorate prepares to vote.”
Noah Pransky is NBCLX’s National Political Editor. He covers Washington and state politics for NBCLX, and his investigative work has been honored with national Murrow, Polk, duPont, and Cronkite awards. You can contact him confidentially at noah.pransky@nbcuni.com or on Facebook, Instagram, or Twitter.