The Poll That Killed a Magazine
The Literary Digest had a method that looked unbeatable: mail out millions of postcards and count what comes back. It correctly predicted the winner of five straight presidential elections, from Woodrow Wilson in 1916 through Franklin Roosevelt in 1932.
Then in 1936 it surveyed 2.3 million voters and concluded that Alf Landon would win. Roosevelt was re-elected in a landslide.
George Gallup worked out what had gone wrong, and the answer had nothing to do with sample size. The first problem was participation bias: Landon's supporters were more enthusiastic about returning their postcards. The second was the list itself, which targeted a more affluent audience than the country as a whole, and therefore one more likely to lean Republican. At the same time, Gallup, Archibald Crossley and Elmo Roper ran far smaller but scientifically grounded surveys, and all three called it correctly. The Literary Digest soon went out of business.
What a Poll Actually Claims
An opinion poll, often just called a survey, is a research survey of public opinion taken from a particular sample. The person who runs one is a pollster.
The design is where the honesty lives. A poll is built to represent the opinions of a whole population by asking a series of questions of a small group, then extrapolating to generalities, either as ratios or within confidence intervals that state how much uncertainty is attached.
That is why the 1936 disaster matters so much. It demonstrated that a sample of 2.3 million people chosen badly is worth less than a sample of a few thousand chosen well. Size flatters, structure decides. Every poll you read today is built on the lesson that a magazine learned the hard way and did not survive.
Before the Science, the Straw Vote
Polling is older than the science behind it. The first known example of an opinion poll was a tally of voter preferences published before the 1824 presidential election, by the Raleigh Star and North Carolina State Gazette and the Wilmington American Watchman and Delaware Advertiser.
It showed Andrew Jackson leading John Quincy Adams by 335 votes to 169. Jackson went on to win both that state and the national popular vote, which gave the method early credibility.
After that, such straw votes gradually became more popular, though they stayed local, usually citywide affairs. The leap to national scale came with The Literary Digest's 1916 survey, which was launched partly as a circulation-raising exercise. Polling arrived, in other words, as a marketing stunt that happened to work.
How Polling Went Global
The idea travelled fast. In September 1938, after meeting Gallup, Jean Stoetzel founded IFOP in Paris, the first European survey institute. He began political polling in the summer of 1939.
Gallup launched a British subsidiary that produced one of polling's great vindications. In the 1945 general election it was almost alone in correctly predicting a Labour victory, when virtually every other commentator expected the Conservatives under the wartime leader Winston Churchill to win.
After the war, the Allied occupation powers helped create survey institutes across the Western occupation zones of Germany in 1947 and 1948, in order to better steer denazification. By the 1950s, various forms of polling had spread to most democracies. There is a commercial thread running through all of it: the historian Jackson Lears argues that advertisers, led by Gallup, who was a vice president at the agency Young and Rubicam, rehabilitated the idea of consumer sovereignty by inventing scientific public opinion polls and making them the centrepiece of market research.
