Lost in a Gallup

This content has been archived. It may no longer be accurate or relevant.

From The Wall Street Journal:

Griping about polling goes back a long time, even to the days before George Gallup published the first random-sample opinion poll in October 1935—as many years away from us in 2020 as that first poll was from the Compromise of 1850. And truth to tell, it doesn’t seem intuitively obvious that the responses of a randomly chosen group of 800 people should come reasonably close, in 19 cases out of 20, to those you’d get if you could interview everyone in a nation of 209 million adults. Even sharp math students don’t always know much about statistics and probability. So the griping goes on.

Some of it reflects a misunderstanding of what polling is. It’s not prediction: Polls are a snapshot taken at a point in time, not a movie preview of what you’ll see later. That fundamental point is often lost or at least misplaced by W.Joseph Campbell in “Lost in a Gallup: Polling Failure in U.S. Presidential Elections,” an otherwise fast-moving narrative history of some attempts to gauge public opinion amid electoral politics. “Election polls are not always accurate prophecies,” Mr. Campbell writes early on. He notes that “polling failures tend to produce broadly similar effects—surprise, anger, bewilderment and frustration at their failing to provide the American public with accurate clues about the most consequential of all U.S. elections.” Surprise, anger, bewilderment, frustration: This sounds like the response to the result of the 2016 election in the city where Mr. Campbell teaches, Washington (which voted 91% for Hillary Clinton and 4% for Donald Trump).

But Mr. Campbell’s gaze goes far beyond the Beltway and back further in history than the astonishing election night four years ago. He is well aware that the national polls in 2016 were close to the results; the pre-election Real Clear Politics average showed Hillary Clinton ahead by 3.3%, close to her 2.1% plurality in the popular vote. Polls in some states were further off. Still, Nate Silver’s FiveThirtyEight gave Donald Trump a 29% chance of winning, and 29% chances happen about one-third of the time. Mr. Campbell quotes RCP’s Sean Trende saying, rightly, that 2016 “wasn’t a failure of the polls. . . . It was a failure of punditry.”

The subject of “Lost in a Gallup” is not so much election polling as its effects on political journalism. Mr. Campbell, a prolific author and a communications professor at American University, admits up front that he is not concerned with “jargon and the opaque methodological arcana that pollsters and polling experts are keen to invoke.” The book is a history of mistakes and overcompensating for mistakes. Polling pioneers Gallup, Elmo Roper and Alexander Crossley, after bragging how closely the past three elections matched their poll numbers, all showed Thomas Dewey leading Harry Truman in 1948. Having got that wrong, they fudged their results to project a close race in 1952. Wrong again!

. . . .

Mr. Campbell devotes much attention, justifiably, to the 1980 election. For months, polls showed a close race between incumbent Jimmy Carter and elderly (age 69) challenger Ronald Reagan. But when the exit polls—invented by polling innovator Warren Mitofsky, also the inventor of random digit-dialing phone interviewing—showed Reagan well ahead, NBC projected his victory, to almost everyone’s astonishment.

But were the polls actually wrong? The author quotes the Carter and Reagan pollsters, Patrick Caddell and Richard Wirthlin, saying that opinion shifted strongly to Reagan after the candidates’ single debate seven days before the election and after Mr. Carter’s return to Washington the next weekend to tend to the Iran hostage crisis. Both pollsters told me the same thing back in the 1980s. Their story makes sense. Reagan’s “are you better off than you were four years ago?” debate line (stolen, though no one then realized it, from Franklin Roosevelt’s 1934 pre-election fireside chat) worked in his favor, and Mr. Carter’s job rating, buoyed upward all year by his efforts to free the hostages, was liable to collapse when he failed.

Link to the rest at The Wall Street Journal (PG apologizes for the paywall, but hasn’t figured out a way around it.)