Inside Story

Politics vs. pollsters

Peter Horner, editor
peter.horner@mail.informs.org

On the morning of Nov. 8, Election Day, one analytics-oriented website indicated Hillary Clinton had a 99 percent chance of garnering the 270 electoral votes needed to win the U.S. presidential election. A few days earlier, statistics guru Nate Silver of “FiveThirtyEight” fame gave Clinton a 75 percent chance of winning the election. Meanwhile, in the weeks leading up to the election, virtually every credible “scientific” poll showed little or no viable path through the Electoral College to the White House for Donald Trump.

How could so many pollsters and political pundits get it so wrong?

That’s a question a lot of people – with the possible exception of quantitative historian Allan Lichtman – have been asking since the results of Nov. 8 stunned the world. In mid-September, Lichtman, a history professor at American University in Washington, D.C., publicly predicted a Trump triumph despite the fact that Clinton held a substantial lead at the time in most national polls.

So then the question becomes, how did Lichtman get it right? The answer: by accident.

As chronicled in his book (“Predicting the Next President: The Keys to the White House”) and in the mainstream media, as well as in the pages of OR/MS Today for the last 20 years, Lichtman bases his presidential predictions on a “13 Keys” model – a series of 13 true/false statements concerned for the most part with the incumbent administration. If six or more of the statements are false, the incumbent party candidate, in this case Hillary Clinton, loses. The Keys model focuses on incumbent governance rather than the perceived strengths or weaknesses of the current candidates and their campaigns.

So where did Lichtman go astray?

Well, his model comes with a caveat: It is designed to predict the winner of the national vote, not the Electoral College vote. Until this year, the model correctly predicted the winner of the national vote in every presidential election since 1984, including Al Gore’s “win” in 2000. At the time of this writing, however, Clinton held about a two million lead in the 2016 national vote. So, due to misinterpreting one of the keys, Lichtman apparently got the national vote wrong but correctly predicted the next president.

Doug Samuelson introduced Lichtman to operations research and OR/MS Today readers in 1996 (“Unlocking the Door to the White House,”), and Samuelson has been writing about the “13 Keys” every four years since. A prolific contributor to OR/MS Today on a variety of topics, Samuelson is perhaps best known to readers as the author of the “ORacle”– this issue marks the completion of the 30th year of the column.

In the July-August 2016 issue of Analytics magazine, Samuelson gave a preview of this year’s election based on the 13 Keys model. At the time, three of the keys had yet to be determined, so Lichtman held off on making a prediction. For this issue of OR/MS Today, Samuelson again touches base with Lichtman for a post mortem on the election, predictions and lessons learned during this most unusual and unpredictable of presidential campaigns.
For more on the story, including the advent of political “tribalism” and a nod to Marshall McLuhan who famously stated that the “medium is the message” and who in 1964 predicted the multimedia mess of an election we witnessed this year, see “The election that confounded everybody” (page 38).
Clearly, political pollsters, pundits and prognosticators – and those who do the analytics and number crunching in support of all of the above – have a lot to learn between now and 2020.