Will voters halt Harris? Dump Trump? Survey says...
Toss a coin? Toss your lunch?Tomorrow's election might be the closest ever. Or not. If you're looking for clues don't look to the polls.
Good news folks—soon we’ll be back to being annoyed by prescription drug ads instead of political ones. Actors and actresses (am I allowed to say that?) will again be singing the praises of Jardiance and lowering their A1C, instead of political candidates singing their own and lowering our expectations. So, who will be the next President of the United States? If you’re watching the polls for a clue—the truth is they don’t have a clue either.
“The polls,” writes New York Times chief political correspondent Nate Cohn, “show one of the closest presidential elections in the history of American politics. Across the key battlegrounds collectively or nationwide, neither Kamala Harris nor Donald J. Trump leads by more than a single percentage point. Neither candidate holds a meaningful edge in enough states to win 270 electoral votes.”
The graphic below shows state-level polling from the 6 “battleground” states that will likely determine our next President. I’ve also included the National polls (“NA”). Some pollsters and election analysts are also including polls from Arizona, but Trump has been holding a steady 4-5 point lead there—so I think it’s a safe state for Trump. Both nationally, and in the battlegrounds, they show a virtual dead-heat.
A poll is like a snapshot, a moment in time, and more, or less, out of focus. Every poll, if it uses a random sample, has error built into it—even the best poll. That built in error is what pollsters report as the margin-of-error. Other polling errors, or bias, come from the polling firm’s methodology, just plain bad luck, or both.
So, no one should put too much stock in any one poll, even from a reputable pollster. Reputable polls with larger and better constructed samples usually show a margin-of-error of around 3%. A poll showing a 50-50 tie, for example, could actually be 53-47 either way, or anywhere in between. And research has shown that pollster’s calculations of the margin-of-error are often underestimates. Even the best poll, with a solid sample and good methodology, has a 1 in 10 chance of being off by more than the margin-of-error.
Don’t underestimate Donald Trump!
Added to the normal challenges of predicting an election based on taking random samples of say a thousand or so voters from the millions who will vote, Donald Trump has proved to be as vexing for pollsters as he has for many voters.
Polls in 2016 and 2020 consistently underestimated support for Trump. Either the polls were systematically failing to reach a “random” sample that included enough Trump voters, or some Trump voters didn’t feel comfortable saying they were voting for Trump. These so-called “shy Trumpers” have proved to be a real problem for pollsters.
In politics though, as in physics, for every action there is often an equal, and opposite, reaction. If Republicans have “shy” Trump voters who are cowed into refusing to admit they support Trump, well then, Democrat pollsters and consultants have to have “shy” Harris voters to help explain underwhelming polling.
Some campaign strategists have begun to claim that domineering husbands/partners are forcing some women to “hide” their intentions to vote for Harris. “I don’t have direct evidence of this, but in this highly contentious environment, it’s not implausible for there to be a significant percentage of women voters who are not just hiding their intended Harris vote from their significant others, but also from pollsters,” Mark Putnam, a Democratic consultant, told the Wall Street Journal. It’s also “not implausible” that pollsters and party consultants are inventing the “shy Harris” voter to cover their asses if their polls, or campaign strategies, are wrong.
Some pro-Democratic groups are so afraid of men bullying their woman partners into voting for Trump that they have rolled out a series of ads, featuring the actress Julia Roberts, reminding women that their votes are secret. “You can vote any way you want, and no one will ever know,’’ Roberts assures them, as two women are shown marking their ballots for Harris.
Mainstream pollsters though seem unconvinced that enough women are being bullied into supporting Trump to skew polling results. The problem for Democrats, as Putnam admits, is that there’s “no direct evidence of this,” while there is evidence for the “shy” Trump voter and the failure of polls to reach a representative sample of all voters that correctly predicts Trumps actual support among the electorate. “The polls underestimated Donald J. Trump in 2016. They underestimated him again in 2020,” the New York Times’ Cohn wrote yesterday, “So can we trust the polls this time?” No, says Cohn, “at least if you mean by ‘trust’ what I think you do. You can’t safely assume that the candidate leading in the polls is going to win.”
However, it wasn’t only “shy” Trump supporters that may have led polls to underestimate Trump’s support in 2016 and 2020. More politically-engaged voters (those who follow politics more closely) with more formal education, for example, college graduates, are much more likely to agree to participate in polls in the first place. Those voters tend to be Democrats or at least lean that way. That leads to overestimating support for the Democratic candidate while underestimating Trump’s.
Trump does better among the less politically-engaged blue-collar voter without a college education. So if you’re not including these voters in your poll, your poll is going to be off. The higher quality polls are struggling for ways to find these voters and encourage them to participate in their surveys—some are even paying them to answer their questions. Without representative samples that include all likely voters, polls won’t be accurate.
“The case for pessimism on accuracy is straightforward,” writes Nate Cohn, “There’s no reason to believe that pollsters can reach enough less engaged and less educated voters, and there’s every reason to believe Mr. Trump still excels among them.”
Does this mean that the polls are still underestimating Trump’s support, and that he will coast to an easy victory? Not so fast. “The close polls don’t tell the whole story,” cautions data scientist Lenny Bronner, writing in the Washington Post. “It’s not unlikely that if we see a polling error of a typical size, one candidate — Donald Trump or Kamala Harris — could run away with the election.” Speaking of running away, I’ve already begun to see some folks on social media saying they’re leaving the country if their candidate doesn’t win. The upside there is that might help solve the housing crisis that Kamala Harris has been talking about.
Eli McKown-Dawson, an elections analyst for Nate Silver’s Substack “Silver Bulletin,” agrees with Bronner. “Donald Trump has a 0.3-point lead in Pennsylvania, while Harris has small leads in Michigan (D +1.1) and Wisconsin (D +0.9),” McKown-Dawson writes, but “that doesn’t mean the actual outcome will be all that close. If the polls are totally accurate we’re in for a nail-biter on Tuesday night. But a systematic polling error is always possible…And because things are so close, even an average polling error would upend the state of the race.”
I’ve been asked by a few people who I think will win tomorrow. My guess, I guess, is a good, or as bad, as anyone’s. I have no crystal ball, but if there is, in fact, “systematic” polling error, it’s likely that error will accrue to Trump’s favor. “If we applied the 2020 polling error to this year’s polls,” Bronner noted, “Trump would win six of the seven most critical states.” Lest Trump supporters get too cocky though, Bronner cautions that, “If a 2012-style polling error were to happen this year, Harris would win six of the seven swing states.”
“Fixing” the polls has proved an elusive goal—one that even election analysts working for the Washington Post and New York Times acknowledge. “It’s possible,” wrote the New York Times’ Cohn, "the polls could badly underestimate Mr. Trump once again.”
Here’s what I’m thinking. If pollsters haven’t yet corrected or updated their polling methodologies, they are likely to have underestimated the Trump vote. And given how close the race is in all of the battleground states, even slightly underestimating Trump’s support might be enough to swing some, or all, of those his way.
It seems that Trump has momentum in Nevada, Georgia and North Carolina. Harris seems strongest in Michigan and Wisconsin. And that puts the road to the White House squarely on Pennsylvania. Almost every poll has Pennsylvania a dead heat—so any lurking underestimate of Trump’s support might swing Pennsylvania his way. If he wins Pennsylvania it’s almost certain he’s President again.
I can hear my left-leaning and Trump-hating friends collective groan, but Harris might still come out on top in Pennsylvania—and/or Nevada or North Carolina. If Harris wins Michigan and Wisconsin, and sneaks by in PA, she’ll be the one living on Pennsylvania Avenue. So I think that Michigan and Wisconsin are “must-haves” for her. Trump needs Georgia, and either North Carolina or Nevada, or both, along with Pennsylvania.
It looks like the Keystone State is the key to tomorrow’s election.