“A forecaster should almost never ignore data, especially when she is studying rare events like recessions or presidential elections, about which there isn’t very much data to begin with. Ignoring data is often a tip-off that the forecaster is overconfident, or is overfitting her model—that she is interested in showing off rather than trying to be accurate.”
― Nate Silver, The Signal and the Noise: Why So Many Predictions Fail-but Some Don't
“Results Today Point to Close November Race: This Year's May Be the Sharpest Battle Since 1916; 1936 Election Seen as Closest in Years,” declared the Washington Post* in July 1936 regarding the battle between Republican Alf Landon and the incumbent Democrat Franklin Delano Roosevelt for the presidency.
The article summarized results from recent polls that showed American voters seemingly split or giving Landon a lead in the race.
One of the surveys that would come to this conclusion was conducted by the Literary Digest**. Among the most esteemed magazines of the time, the Literary Digest had a history of accurately predicting the winners of presidential elections going back to 1920.
The mailing list for the publication’s 1936 straw vote poll was culled from automobile registrations and telephones books, resulting in about 10 million names. Approximately 2.4 million recipients returned the mock ballots. This is a HUMUNGOUS sample for such a survey – the Literary Digest poll remains one of the largest and most expensive polls ever conducted.
This massive sample size instilled a sense of confidence for the magazine and many of the American people in the accuracy of the results, which indicated Landon would get 57% of the vote and Roosevelt 43%.
In the actual election, Roosevelt took 62% of the popular vote against 38% for Landon.
This was the highest popular-vote percentage won at the time and the second highest won in U.S. history – the highest percentage of the popular vote went to Lyndon Johnson in 1964.
“In the greatest demonstration of public confidence and appreciation ever awarded a President, the voters of the United States of America swept Franklin Delano Roosevelt back into office by the startling total of a possible 523 electoral college votes to a meager 8 for Republican candidate, Alfred M. Landon,” reported the Atlanta Daily World* two days after the remarkable election. “The Democrats were amazed at the magnitude of their own victory.”
Roosevelt took a record 98.5% of the electoral votes available, winning the race by the greatest electoral landslide since the beginning of the current two-party system in the 1850s. Roosevelt carried every state except Maine and Vermont.
The only candidate to surpass Roosevelt’s victory was Ronald Reagan in the 1984 election when there were 7 more electoral votes available to contest.
As a result of this election, some experts predicted the demise of the Republican Party, as many voters held them responsible for causing the Great Depression. However, Republicans would make a strong comeback in the 1938 congressional elections, although they were not able to win back the presidency until 1952.
The sampling error was 19% in the Literary Digest poll, the largest ever in a major public opinion poll.
Over the decades, scholars have speculated and hypothesized about where the Literary Digest went wrong. One problem many experts had with the survey was that it didn’t actually sample the American public at all. It sampled only the car- and telephone-owning American public, a more well-to-do segment of the population that wasn’t inclusive of lower income voters who might seem more likely to support Roosevelt.
(For insightful analysis of the factors that contributed to the disastrous Literary Digest poll, check out Peverill Squire’s article “Why the 1936 Literary Digest Poll Failed,” published in Public Opinion Quarterly and available on the ProQuest Central database.)
That same election year, George Gallup, an advertising executive who founded the American Institute of Public Opinion, initiated the scientific polling method that is still used today in the prediction of election results. The Daily Boston Globe* visited the offices of the American Institute of Public Opinion just after the 1936 election to report on this new methodology in the article “How Gallup Got His Figures: Poll that Predicted Roosevelt Election is Based on Harvard Professor’s Formula for Sampling Voters.”
Gallup’s poll not only predicted that Roosevelt would win the election – based on a sample of 50,000 people – he also predicted that the error in the Literary Digest results. The Gallup Poll would become an integral part of future presidential elections, and remains one of the most prominent election polling organizations.
*Available from ProQuest Historical Newspapers™
**Indexed in Periodicals Index Online