"Landon, 1,293,669 Roosevelt, 972,897"

[timrollpickering]

For those wanting a good saying and link to respond to people who argue large self-response rates are key:

Not as well known a headline as "Dewey defeats Truman" but this is the other great miscall of 20th century US elections. However there isn't so much detail available on the web despite it being the obvious rejoinder to anyone who points to sample sizes alone as a sign of accuracy.

Anthony Wells has posted the details about the Literary Digest survey that got the 1936 Presidential election spectacularly wrong despite nearly 2.4 million responses. Meanwhile George Gallup made his name with at most 40,000 responses and a good call, getting the best publicity (in July he also predicted the Literary Digest result to within a point). However even he wasn't the most accurate.

Size alone is not enough – the tale of the Literary Digest

(This was also the election that busted the saying "As Maine goes, so goes the nation" which, contrary to assumption, actually meant that the early September governor & congressional elections there were often seen as a predictor of which party would get the Presidency. In fact Alf Landon carried only Maine and Vermont.)

[maxque]

The issue is not self-response, it's than they surveyed their readers.

Obviously, especially given the depression, their readers were wealthier than the average elector (and the Republican Party was the party of the industrial Northern industrial elites, then, while Democrats was the party of the agricultural Southern working-class).

It's like if the Guardian surveyed 2 millions of their readers and were surprised it didn't give the right result for the 2020 election.

[marksenior]

The first British opinion poll was not a success . BIPO who became Gallup forecast a Conservative win in the 1939 Batley and Morley by election by 50.7 to 49.3% The result was in fact a Labour win by 55.4 to 44.6%
Most pollsters in 1970 forecast a Labour win in the 1970 GE Gallup by 7.0% and Marplan by 9.6% , in fact the Conservatives won by 2.3%

[timrollpickering]

They actually surveyed more than their readers but the contact lists they used were also biased in the same direction. And the study linked to in the article suggests a bias in the response rates as well - had everyone replied or done so in equal proportion the headline would have predicted a Roosevelt victory but Landon voters were more likely to return their cards.

[john07]

Feb 14, 2017 22:51:09 GMT timrollpickering said:

For those wanting a good saying and link to respond to people who argue large self-response rates are key:

Not as well known a headline as "Dewey defeats Truman" but this is the other great miscall of 20th century US elections. However there isn't so much detail available on the web despite it being the obvious rejoinder to anyone who points to sample sizes alone as a sign of accuracy.

Anthony Wells has posted the details about the Literary Digest survey that got the 1936 Presidential election spectacularly wrong despite nearly 2.4 million responses. Meanwhile George Gallup made his name with at most 40,000 responses and a good call, getting the best publicity (in July he also predicted the Literary Digest result to within a point). However even he wasn't the most accurate.

Size alone is not enough – the tale of the Literary Digest

(This was also the election that busted the saying "As Maine goes, so goes the nation" which, contrary to assumption, actually meant that the early September governor & congressional elections there were often seen as a predictor of which party would get the Presidency. In fact Alf Landon carried only Maine and Vermont.)

The 1936 poll derived its responses from two sources:

1. Subscribers to the Literary Digest

2. Bell telephone customers

That is hardly likely to have produced a representative sample of voters in 1930s USA. It correctly predicted how upper income voters might behave.

It is covered well in How to lie with statistics by Darrell Huff.

As for "As Maine goes, so goes the nation", this was amended to "As goes Maine, so goes Vermont".