*Literary Digest* (most popular news magazine of its day) poll...
*LD* had been doing preference poll for several presidential elections without biases manifesting themselves in incorrect projection. 1936 projection was one indication of class-polarizing politics of time since poll method was same. Send opinion cards to 10 million prospective voters from mailing lists with names selected from automobile registrations, civic clubs, telephone books, magazine subscribers. Sampling bias generally excluded working class (both employed & unemplyed). Plus, only about 2 million people responded, hence, non-response bias.
George Gallup made his reputation in '36 when he predicted that *LD* projection was wrong. He did two polls that year, a random sample that projected FDR would received 56% of votes cast *and* smaller version of magazine 'straw poll' (so-named from 1920s movie theaters that offered moviegoers opportunity to 'vote' for candidates by tossing drink straws into trash cans with said names...real point was to sell drinks) that projected election result similar to *LD* in which Landon would receive about 55% of votes cast. If memory serves, FDR received about 62% of votes cast.
*Literary Digest* ceased publication a couple of years later. Despite magazine's prior polling success, it couldn't overcome '36 fiasco. But then, this is magazine that had predicted at turn of 20th century that 'horseless carriage' would never supplant bicycle... Michael Hoover