Four years ago, Clinton supporters watched in anguish as the surprise results came in. (Photo: Matt Rourke/AP)
You stayed up all night in 2016, watching the numbers shift, stomach in knots. Or maybe you woke the next morning, jabbed your phone, and saw: He’d actually done it. In horror (or glee), you exclaimed, “But the pollsters swore Hillary would take it!”
Ever since, we have vowed not to trust the numbers—only to find ourselves nearing this U.S. presidential election, devouring each estimate of who’s up and down. Poll after poll indicates that Trump is losing to Biden. Can we believe them?
In an ongoing Persuasion feature, “Ask an Expert,” we seek insight into something we’ve been wondering. In this case, we phoned up the historian Allan Lichtman of American University, who has spent decades promoting an alternative system of forecasting winners, a list of 13 “keys,” developed by studying all presidential races since 1860.
His model poses a series of yes/no questions on everything from the health of the economy, to the presence of a third-party campaign, to the administration’s success or failure in foreign affairs. Lichtman’s model has predicted every president since 1984, with the exception of Al Gore in 2000, when it guessed the winner of the popular vote, though not that of the electoral college.
Persuasion: The pollsters say this time is different. Is it?
Lichtman: This time is different with respect to the polls, simply because the pollsters have made some effort to correct some of the sampling errors last time. And Joe Biden’s lead is much larger and much more consistent than Hillary Clinton’s lead—there’s been very little change, remarkably, in this election. But I always tell people to take polls with a big grain of salt because they are snapshots; they’re not predictors. They were abused as predictors, but they only tell you what’s happening at the moment, and that can change. The other thing is that the real error-margin is much larger than what they tell you. You know the polls say, for example, “Our margin of error is plus or minus 3%”? Well, that’s pure statistical error. That’s the error you would get if you drew a sample of green balls out of a huge jar of green and red balls. It does not take into account errors that can be produced by deceptive responses, or by errors in estimating likely voters since you don’t know for sure who is going to vote.
Persuasion: So why does polling have such a grip on how we understand elections?
Lichtman: One, for the media, it is very easy to write a story about the polls—you don’t even have to get out of bed in the morning; just read the polls, and write your story. Two, the polls create drama: what’s happening today, how are things going to change tomorrow. They give this impression that elections are like horse races, with candidates leaping ahead and falling behind. Finally, they give an impression of precision. Polls presume to give you these exact numbers, when in fact the numbers could be quite different from the poll estimate.
Persuasion: Does our obsession with polling warp the political process?
Lichtman: Very much. My 13 “keys” to the White House have correctly forecast the outcome of all American presidential elections since I predicted in April 1982—2½ years ahead of time—that Ronald Reagan would be re-elected in the midst of what was then the worst recession since the Great Depression. What makes “the keys” different is they probe the underlying dynamics of how elections really work. They look at the key factors of incumbent strength and performance based on the insight that American presidential elections are essentially votes up or down on the record of the party holding the White House. It is governing, not campaigning, that counts.
Persuasion: Only two of your “keys” focus on the personality of candidates—specifically, their charisma. Do we fixate too much on the individual, when we should be looking at the broader picture?
Lichtman: We do tend to obsessively focus on the candidates and on the campaign, and not look at the dynamics of the election nearly enough. And there’s a reason for that. In 1961, Dwight Eisenhower, in his farewell address, warned of the grip of a “military-industrial complex.” Well, today we have a political-industrial complex, with its own “iron triangle.” At one point [of the triangle]: all the pollsters, the consultants, the handlers, the ad men, who make huge amounts of money based upon the false notion that elections are horse races decided by who’s up and who’s down on a daily basis, and what the candidates are doing and saying. Then, there is the media, which makes its money covering the election day-by-day, and telling you who’s up and down. On the third point of the triangle are the candidates, who are afraid to go against the media and the consultants, the pollsters and the ad men. That’s why we have these narrow, attack-driven, soundbite-driven campaigns. If you believe “the keys”—that it’s governing, not campaigning, that counts—we could have a totally different kind of campaign. According to “the keys,” you should campaign to build a mandate to govern. Give us your vision. Tell us specifically what bills you’ll introduce in your first hundred days, what executive orders you’ll promulgate, what kinds of persons you’ll put in your cabinet, in the judiciary, in the White House. If you do that, you’ll have established a basis for governing, which will help your party win the next election. And even if you lose, you will at least have left your mark on the country, and maybe inspired the next generation. Who remembers anything that recent losing candidates have said? Hillary Clinton, Mitt Romney, John McCain, John Kerry, Al Gore—no one remembers anything they said.
Persuasion: You are predicting a Biden win this time. According to your predictive model, what does this Democratic candidate have that Hillary Clinton lacked?
Lichtman: When I predicted Trump’s win in 2016, it wasn’t because “keys” were in Trump’s favor. It was because there were enough “keys” out against the incumbent Democrats. Under my 13 “keys,” it takes six or more negative ones to count out the White House party. And the Democrats had exactly six against them [in 2016], which is why I said any Republican will beat any Democrat because, based on “the keys,” this was a change election. At the end of 2019, Trump was down four “keys”—two short of a predicted defeat. But then we were hit with the pandemic and the cries for social and racial justice. Instead of dealing substantively with these challenges, he reverted to his challenger playbook and thought he could talk his way out. Of course, that doesn’t work. The result was he lost three more “keys”— the short-term economic “key” gauged by an election-year recession; the long-term economic “key” because of the negative growth; and the social-unrest “key” because of what’s raging across the land. He is now down seven “keys”—one more than is needed to predict that he will become the first sitting president since George H.W. Bush to lose a re-election bid. Never in the history of the United States has the party holding the White House suffered such a sudden and dramatic reversal of fortune in just a matter of a few months. And Trump has no one to blame but himself.
[Interview condensed and edited for clarity.]
Lichtman says of polls, “they are snapshots; they’re not predictors. They were abused as predictors.” That’s half trivial and half right. Obviously, they are snapshots, like his 13 keys. So they cannot speak to make predictions. But just as he uses his keys to predict, so we can use polls.
The problem lies, not with the polls, but with the predictors who abuse them. I took graduate-level stats, so here’s how not to abuse them.
First, any meaningful prediction should not say Biden will win, or Trump will win. No one can know that. An honest prediction says something like Clinton has a 75% chance of winning. The same goes for Lichtman and his 13 keys developed with a Russian earthquake expert. As much as I would love for him to be an all-seeing guru, he’s just like any other predictor. There’s some chance he’ll be right and some chance he won’t.
If you are going to play dice, there’s no use in being told: “it will come up 1 or 2” No one knows. But you will make better bets if you know: “there's one chance in 3 it will come up 1 or 2.” The problem is: how to get from the polling results to Biden’s chance of winning.
It’s possible to find that chance pretty accurately — with a lot of hard work. Lichtman fails to note this possibility. He’s right that the reported accuracy of single polls fails to account for various biases. Many of them can be taken into account by looking at the pollster’s past performance. And then you need to take account of the electoral college … and the chance that if polls are too optimistic in one state they will also be too optimistic in a similar state. I could go on, but that’s what Nate Silver does, and he does it brilliantly.
But even then, he tells you there are some things the polls can’t capture, but human judgment can. So yes, Nate Silver knows that Lichtman’s judgment and that of others should be factored it. Everyone has their favorite guru, but how can we aggregate their combined intelligence. As Nate and many others will tell you — look to the betting markets. When I bet $100 on Biden, as much as I want him to win, I wouldn’t do that if I thought he’d lose. Same with all the other bettors. Well, it’s a bit more complicated. We also take account of the odds. And because of that, the odds end up reflecting the collective wisdom, such as it is, of the bettors.
Right now one betting market (Predictit.org) gives Biden about a 65% chance. And Nate has said if he himself were predicting (and not just using his 10,000 polls) he would factor that in. I personally think the prediction from Nate’s 10,00 polls is more accurate by, perhaps, twice. And the weighted average of Nate’s polls (giving Biden 89%) and the betting market’s 65% is that Biden has an 81% chance of winning. Scary. So don’t think it’s a sure thing and stay home.
Now please note, that is NOT a prediction that Biden will win. It is a statement that he has a 19% chance of losing. It’s too bad Lichtman hasn’t figured out the odds that he will be right. Then we could make use of his obviously excellent intuition.
If Persuasion is going to argue for the ideals of the Enlightenment, science, and all that, perhaps we should look to experts like Nate Silver rather than “experts” who abandon all the math and science in the field they are discussing.