Galluping Away With Data

Prior to the 1890 census, there was an overwhelming amount of basic information about tens of millions of Americans that had to be sorted every ten years. Every census saw longer and longer periods of counting, with the 1880 census only being fully tabulated by 1887.

Then, Herman Hollerith came up with a punch-card based machine—let’s call it a proto-computer—that was able to keep up with the increasing demand, and improvements in process and better machines kept up. By 1930, the population of the US had doubled and the census bureau was asking more questions, but the work was completed within three years instead of seven.

In other words, America now understood how to count its people, and it was becoming routine. Boring, even.

People wanted to know what those folks thought, not just whether they were men or what age they were. But how could you ask every American detailed questions? You really couldn’t—it just wasn’t practical.

What you needed was an ample sample. George Gallup set out to collect a calibrated set of answers that would give you an idea of what the huddled and not-so-huddled masses really thought of America, its politicians and policies, and life in general.

In 1935, he founded the American Institute of Public Opinion, where he began to publish some of his results. Then, in 1936, Gallup made headlines by correctly predicting the winner of the Presidential election, in spite of Literary Digest picking Landon as the winner (spoiler: it was FDR).

This was really significant. Literary Digest had a huge following, and an even huger mailing list. It sent out about ten million sample Presidential ballots and received millions of them back, and Landon was the clear winner. Gallup, by contrast, used a much smaller sample set that was much more carefully selected to represent the voting population in the US, not just subscribers to a magazine.

People really began to pay attention to Gallup, and they started to understand and believe in his methods. Gradually, they improved and became standardized and even institutionalized.

Asking a small group of the right people the right questions turned out to be more effective than asking a group of ten million people the wrong questions.