Eight-five years ago this week, Literary Digest magazine was in the field with the most ambitious public opinion poll in history. At tremendous expense, the Digest mailed out millions of surveys, receiving some 2.4 million replies. Based on this data, it boldly predicted that Republican challenger Alf Landon would decisively trounce the incumbent president, Franklin D. Roosevelt.
In political science and the polling profession, 1936 is something akin to the Biblical creation story. Literary Digest was a pioneer, or should have been. It correctly foresaw a market in taking the voters' pulse before Election Day. Using its straw poll methods -- it had done the first such survey in 1916 -- the mass market magazine had predicted every presidential race since then correctly. So what went wrong in 1936 when FDR won 61% of the popular vote and Alf Landon managed to carry only Maine and Vermont?
For starters, there were huge sampling errors: Literary Digest polled its own subscribers first (they were more affluent than the average bear during the Great Depression). The other lists of names came mainly car registration and telephone directories, which, depending on the state, may or may not have been representative of the population as a whole. An even bigger problem was the response rate. To get those 2.3 million responses, Literary Digest sent out 10 million pieces of mail. The non-response bias proved fatal: Americans who disliked FDR really disliked him and were more motivated than his supporters to return their mock Landon ballot to the magazine.
One small clique of academics foresaw such problems, and came up with another model, which they called "scientific." Operating under the grand handle of the American Institute of Public Opinion, they published their predictions. Some newspapers published the two polls side-by-side. The differences were stark. Literary Digest had Landon winning 57% of the popular vote and 370 Electoral College votes. The American Institute of Public Opinion had Roosevelt at 56%, while leading in 40 states, which it called "sure" for the Democrats (and the rest too close to call).
Well. That was the end of Literary Digest -- it folded within 18 months -- but it was the beginning of scientific polling. One such pollster, in particular, was poised to take advantage. The American Institute of Public Opinion, as it happened, was a small shop run by a Columbia University college professor and some aides and grad students. That professor's name, as you may have guessed by now, was George Gallup.
Carl M. Cannon is the Washington bureau chief for RealClearPolitics. Reach him on Twitter @CarlCannon.