In 2016, back in the pre-COVID 19 era, the result of the presidential election came as a shock to seasoned political observers and to the two major party candidates themselves. That’s not to mention the 65,844,954 American voters who selected Hillary Clinton, and the 62,979,879 for Donald Trump.
After all, didn’t almost all of the 2016 pre-election polls promise a victory for Clinton and a loss for Trump?
Today, on the eve of election 2020 with polls pointing to victory for Joe Biden, are we collectively heading down that same rabbit hole?
As it turns out, the polls in 2016 got things right for the most part, say a trio of LMU experts. They have confidence as well that the 2020 polls will hold up. But, there are some caveats.
In 2016, “the polls were not inaccurate,” says Fernando Guerra, professor of political science and international relations and Chicana/o Latina/o studies and director of the Thomas and Dorothy Leavey Center for the Study of Los Angeles (StudyLA). “The polls continued to indicate that Hillary Clinton was ahead by four percentage points nationwide, with an error margin of three,” says Guerra. “That meant that she could actually have been ahead only by one, or by as much as seven.” Clinton won the popular vote by 2.1 percent.
“I think the critiques of polling were a little overboard in 2016,” says Richard Fox, professor and associate chair of political science and international relations. When polls went astray, it happened locally, in a few select places. But that affected overall polling. “They didn’t really have enough state-level quality polls,” says Fox.
Just as the vote in states like Wisconsin, Michigan and Pennsylvania shaped the Clinton-Trump outcome, polling in those states affected the national pre-election picture. And, in pivotal swing states, polling is all the more important.
“In 2016 there was no accurate or consistent polling in Wisconsin, Michigan and Pennsylvania,” says Guerra. “Why? Because they were always considered part of the ‘Blue Wall.’ These were not considered battleground states,” so there wasn’t a major investment in polling. Says Fox: “[Pollsters] got those three of the 50 states wrong. Those states happened to decide the election.”
Brianne Gilbert says that perceptions about the accuracy of polling sometimes involve the reporting of polls. Gilbert is associate director of StudyLA and a principal investigator on StudyLA’s annual public opinion survey of L.A. residents known as LA Votes. (Due to the pandemic, that survey, an exit poll, will not take place this year.)
Gilbert views the 2016 polls not so much as being wrong. “But what we saw a lot of, from an academic research perspective,” she says, “is that there was a difference in how the results were reported.”
“In a press release,” Gilbert explains, “we may say that a race is too close to call if it’s 52 percent–48 percent, with a margin of error of 4 percent. But the media could pick that up and say, ‘So and so is in the lead, 52 percent–48 percent.’ No. We said it’s too close to call.”
Gilbert doesn’t blame the media. Instead, she compares the process of information getting from pollsters to voters as “almost like a game of Telephone.” After 2016, she says, members of the American Association for Public Opinion Research asked themselves, “How could things have been done differently?”
“A definite topic was how can we explain when we’re talking with reporters, in a clear way, what the data means?” she says. AAPOR has since continued webinars for journalists, including its Transparency Initiative, which aims to heighten public confidence in polling and shine light on research methodology.
As of the evening of Oct. 30, 2020, FiveThirtyEight pollster-guru Nate Silver forecasts Biden having an 89 percent chance of victory and Trump 10 percent, with a tie at 1 percent. In 2016, Silver gave Clinton a 71 percent chance of winning. This time around, there are more and presumably better state polls. While voting during a pandemic is an unexpected variable that hasn’t happened since the 1918 mid-terms, Guerra believes pollsters have experience dealing with nonpolitical events that could shape an outcome. He also notes that an unprecedented number of people have already voted, reducing the number of people whose leanings may have gone uncaptured in polls.
It’s true that likely voters whose views do not get registered in polls is a concern when it comes to accuracy. In the Clinton-Trump contest, says Fox, there was a late surge of rural voters in some states that the modeling hadn’t captured. In 2020, for example, it’s possible that if turnout is huge among young voters or is depressed among Black voters, polls might not pick those changes up. There’s also chatter about the potential effect of voter suppression, violence and intimidation on turnout. In addition, “shy voters” — those who favor Trump but may not want that known — may not get picked up when polled. Generally speaking, though, Fox is not especially worried. “I think the challenge this year is probably exaggerated,” he says. Guerra goes further: “What I’m suggesting is that Biden is actually even further ahead than what the polls are saying.”
Oh, and what if the U.S. Supreme Court decides the presidential election on the fly, as they did in 2000? That would not be an outcome that went unpredicted because of polling.
Guerra suggests thinking of polling as a movie — a movie that begins when candidates announce they’re running and ends when the votes are counted. “All a poll is,” Guerra says, “is one frame of a movie. If you think about it that way, it’s not telling you the end. It’s telling you where we are right now.”
Jeremy Rosenberg, a frequent contributor to LMU Magazine, is a Los Angeles-based writer, editor and consultant. His “Under Spring, Voices + Art + Los Angeles” received the first California Historical Society Book Award in 2013. Rosenberg’s writing has appeared in the Los Angeles Times, the OC Weekly, at KCET.org and elsewhere. Follow him @LosJeremy.