Heading into 2020, pollsters are still grappling with how to wash down the bitter aftertaste of the last presidential election, when many state-level surveys failed to pick up on President Trump’s late surge and the industry — fairly or unfairly — took the blame for getting it wrong.
They’ve made adjustments to their methodologies, offered numerous disclaimers, and trod carefully on using their data to make bold predictions, while also laboring to feed a fixation on the “horse race” aspects of the White House contest from both the media and the presidential campaigns themselves.
News coverage of polls has suggested “they should be more precise than they could possibly ever be,” said Patrick Murray, director of the Monmouth University Polling Institute. “It’s like if you’re trying to put a screw in the wall and you’re using a hammer. You can get it in the wall, it’ll stay there for a while, but it’s not the right tool to get the job done as well as you wanted to.”
In August, Mr. Murray took the unusual step of publicly stating that a survey was an “outlier” when it showed former Vice President Joseph R. Biden, then a solid front-runner in national polls, falling behind Sens. Elizabeth Warren of Massachusetts and Bernard Sanders of Vermont.
“Polling is an imprecise measurement and it’s always going to be an imprecise measurement,” Mr. Murray told The Washington Times. “Polling is not predictive. It just is explanatory about what the general state of the race is at any particular time. And that’s what we have to do a better job of — conveying what polling is actually capable of telling you.”
Mr. Murray and other pollsters are paying much closer attention this time to better matching the education levels of the expected electorate in their models after noncollege graduates broke hard for President Trump over Hillary Clinton in 2016.
Jim McLaughlin, a pollster for Mr. Trump’s 2020 reelection campaign, said traditional public polls are still getting it wrong.
“A lot of the media pollsters, unfortunately, don’t use voter lists,” Mr. McLaughlin said. “So they’re calling all these people that aren’t necessarily going to vote. They claim they’re going to vote, or they’re just a registered voter. They don’t do a ’likely voter’ model.”
He said many of the polls were wrong in 2016 because a lot of the modeling was based on the 2012 match-up between President Barack Obama and Republican nominee Mitt Romney.
“You had the big surge in turnout among young people, you had the big surge in turnout among African Americans and Hispanics, too,” he said. “And you had a depressed turnout, quite frankly, among quite a few white voters — different groups and whatnot. So they were using the wrong model.”
John Couvillon, a Louisiana-based pollster, said the science of polling is “a continuous improvement cycle.”
“I think that right now, the focus in terms of trying to avoid the 2016-type disaster would be to focus on education level,” he said.
Pollsters are also dealing with diminishing response rates, both from landline and cellphones, which has the potential to skew results.
The Pew Research Center said recently that most of their surveys during the 2020 election cycle will be conducted online using a “randomly chosen panel of respondents.”
Michael Dimock, president of the group, said that “high-quality surveys” remain “a trustworthy tool for measuring public opinion.”
“While some individual state polls did fall short in 2016, it was not an industry-wide failure, and rigorous national-level surveys turned out to be quite accurate by historical standards,” Mr. Dimock wrote.
Indeed, Mrs. Clinton did win the U.S. popular vote by about two percentage points over Mr. Trump — not far from what national polls were showing in the run-up to election day.
But as Mr. Dimock indicated, surveys were off in individual states like Wisconsin that were crucial to Mr. Trump’s win. For example, the highly regarded Marquette University Law School poll had Mrs. Clinton leading Mr. Trump by 6 points in Wisconsin right before the election, and Mr. Trump ended up narrowly carrying the state.
One approach that some outfits have adopted is to re-interview the same respondents on a rolling basis to better track and reflect swings in public opinion and what’s driving the changes.
“We can track specific people, how they change their mind, and where they’re going, which is something you can’t do with just sort of fresh interviews every month,” said Dave Peterson, a political science professor at Iowa State University who helps oversee surveys ISU conducts with Civiqs.
He said, for example, that 40% of Biden supporters in September ended up picking someone else the next month, and that much of an uptick for Mayor Pete Buttigieg of South Bend, Indiana, came from Biden and Warren supporters.
“We interview the same people whether or not they said they were going to participate,” he said. “Buttigieg and Klobuchar are the two who really seem to benefit from the people changing their mind about participating. So that’s a question that if you’re just asking fresh [voters], you can’t quite track.”
The Democratic National Committee has used both national and early state polling to determine which candidates qualify for the DNC presidential debates, making polling an active part of the election process regardless of the predictive value of the surveys.
That’s left some candidates now on the outside looking in getting exasperated about their significance.
“Polling is such bull——,” Sen. Cory A. Booker of New Jersey said in a recent appearance on “The Breakfast Club” radio show.
“There has never been a point in the Democratic Party in our lifetime where somebody who was leading in the polls went to the White House this far out.”
Michael Miller, a political science professor at Barnard College, said there’s little utility in parsing over national polls in a presidential election at this point in the cycle.
“We’ve created this house of cards — so it’s people on our side of it, the political scientists have bought in, journalists have bought in, campaigns have all bought in — to focusing on the ’who’s winning?’ aspects of polling,” he said. “I think there are dangers here.”
Mr. Couvillon, though, said the 2016 post-mortem shouldn’t be taken to the extreme.
“The whole thing about ’because Trump won, that means all polls are bad’ — I can’t tell you how many times I’ve had to deal with that false narrative. And it’s almost like people don’t want to know the truth,” he said. “You don’t really have a good counternarrative challenging that false, lazy assumption.”
• David Sherfinski can be reached at dsherfinski@washingtontimes.com.
Please read our comment policy before commenting.