The history of pollsters blowing it decades before Trump vs. Biden

The typical method for a straw poll in the 1930s went like this: A newspaper or magazine printed a sample ballot in its pages, and readers would fill it out and send it in. Based on all of the responses, the newspaper would make a prediction.

Gallup, who had pursued a PhD in psychology and worked in ad research, thought straw polls were nonsense. Really, you were just surveying the type of people who read the newspaper, cut something out and mailed it. Not exactly representative of the electorate.

Rather than measure the opinion of a large number of the same type of person, Gallup developed a system of "quota sampling" - surveying a small cross-section of Americans who mirror the demographics of the entire population - to get a supposedly more accurate measure. In the United States, he calculated that could be done with 3,000 people from different regions and of different ages, races, educational backgrounds, etc.

That's how Gallup ushered in the modern era of polling - a method that failed to predict Donald Trump's victory over Hillary Clinton in 2016 and got the strength of the president's support against Democratic challenger Joe Biden wrong again this week.

While the nation waits for an outcome in the agonizingly close 2020 election, it's worth examining how we came to rely on polls.

As historian Jill Lepore explained in the New Yorker in 2015, the word "poll" used to mean "head," as in, the thing being counted when voting "involved assembling (all in favor of Smith stand here, all in favor of Jones over there)." The term "straw poll" evolved from an old expression about throwing hay into the air to see which way the wind was blowing, according to William Safire.

In the mid-1930s, Gallup got an important ally to help prove his theory. In her memoir "Personal History," Washington Post publisher Katharine Graham wrote about her father, then-publisher Eugene Meyer, taking interest in his new polling method. At the time, Gallup's "polls weren't taken very seriously," she wrote, but "ever the logical thinker, and having always put a premium on the importance of research, my father signed the first contract with Gallup and ran his polls on the front page."

By the 1936 election between President Franklin D. Roosevelt and his Republican challenger, Kansas governor Alf Landon, the best-known straw poll was conducted by the magazine Literary Digest. Its method wasn't exactly scientific. The digest would randomly select millions of addresses in the phone book and car registration data, and then send postcards to those addresses asking for whom the recipient planned to vote.

Gallup spotted a big problem with this method: by only polling people with cars and phones, they had no input from the poor, who couldn't afford either. And poorer voters largely broke for Roosevelt.

Using his quota sampling system and in-person interviews, he predicted that Roosevelt would win reelection. Not only that, he predicted what the Literary Digest would predict - a win for Landon with 56% of the vote - and explained why the magazine would be wrong.

Roosevelt was reelected, and Gallup prevailed; though, in a bit of foreshadowing, he was way off on the margin of victory.

Literary Digest was also wrong about its prediction of 370 electoral votes for Landon. He got eight. The magazine went out of business soon afterward.

Twelve years later, Gallup was the head of a polling empire with offices in 11 countries. His political polls were carried in more than 100 newspapers, and he had spawned competitors, primarily pollsters Elmo Roper and Archibald Crossley.

In May of 1948, the pollster was treated to a gushing profile in Time Magazine, which described him as a "big, friendly teddybear of a man" who "loves children and animals" and whose "fondest dream is that Congress will someday abolish the Electoral College."

In the last poll published before the election, Gallup had Dewey beating Truman by five percentage points. So sure was he of Dewey's victory, the firm had stopped polling two weeks before the election and simply ignored the 14% of voters who said they were undecided.

What happened, of course, is best encapsulated by the famous photo of Harry S. Truman, grinning broadly, and holding up a copy of the Chicago Daily Tribune erroneously headlined "Dewey Defeats Truman." In fact, Truman beat Dewey by 4½ percentage points.

Katharine Graham, by then a controlling owner of The Post with her husband Phil Graham, described in her memoir how they handled the embarrassment of having predicted Truman's loss: "When it became clear that Truman had indeed fooled the pundits and pulled off a political miracle, Phil . . . sent off a telegram to the president, which he printed on Page 1 of the morning-after paper:

"You are hereby invited to attend a 'crow banquet' to which this newspaper proposes to invite newspaper editorial writes, political reporters and editors, including our own along with pollsters . . . The main course will consist of breast of tough old crow en glace (You will eat turkey.) Dress for guest of honor, white tie; for others - sack cloth."

Truman responded magnanimously: "I have no desire to crow over anybody or to see anybody eating crow . . . We should all get together now and make a country in which everybody can eat turkey whenever he pleases."