ABSTRACT
The other part of the answer lies in the self-selection of respondents. Only about 24% (2,400,000) of the people responded, that is, they self-selected to respond while others did not. Considerable research on public opinion polling since that time strongly suggests that in response to mailed polls, people from higher so-
using about 56,000 respondents, correctly predicted that Roosevelt would win. These other polls were correct in calling the election because they avoided serious sources of bias such as those in the Literary Digest poll. As you may know, pollsters today conduct national surveys and get rather accurate results using only about 1,500 respondents. This is possible by concentrating on getting a good cross section of the voting population rather than using a large sample. So the lesson is clear: Having a sample that is free of bias is much more important than having a large sample. More concretely, asking millions of the wrong people gives you a wrong answer millions of times over.
