Reference: Freedman, Pisani, and Purves,
Statistics, Fourth Edition, W. H. Norton, 2007.
In 1936, Franklin Deleno Roosevelt (FDR) was completing his first
term as president of the U. S.
He was running against Alfred Landon of Kansas.
Most election observers thought that Roosevelt would be an easy winner.
The Literary Digest magazine predicted that Landon would be
the winner by surveying 2.4 million persons.
George Gallop was just starting his survey organization.
He predicted that Roosevelt would win, by surveying only 50,000 people.
He also predicted the Digest result by surveying only 3,000 people.
The results:
The election result
62
Literary Digest's prediction
43
Gallup's prediction of Digest prediction
44
Gallup's prediction of election result
56
Think of ways to explain these results. Some possible answers:
Literary Digest chose the people to survey from phone books and
exclusive club membership lists. This biased the results. In 1936, only wealthy persons had
phones or belonged to exclusive clubs. Wealthy people were more likely to vote Republican.
Gallup's poll used scientific sampling, which uses simple random sampling: every person in the
country is equally likely to be sampled (which is easier said than done).
Gallup predicted the Literary Digest result by sampling also from phone books and
exclusive club membership lists.