Strengthening Presidential Polls
Test 9 - Section 2 - Question 6
Wait, you mean to tell me that presidential election polls don't always perfectly predict the winner? (I'm looking at you, 2016 Nate Silver.)
Of course there are all sorts of ways that polls can get it wrong. And I don't actually hate Nate Silver. Dude is smart. And he's the first to acknowledge that biases are extremely hard to avoid. Nate's modern predictions would always be probabilistic, never determinative. This question asks us to explain why Landon was leading in the 1936 telephone poll, but Roosevelt won handily. Why didn't the poll get it right?
A) This doesn't explain anything. It seems like interviewers should avoid revealing their own political affiliations when possible, doesn't it? If they did conceal their affiliations this would only make the poll more accurate, not less.
B) The first half of this answer also seems like it would make the poll more accurate. Why should we survey people that aren't eligible to vote? The second half of this answer is nonsensical, given the first half.
C) Bingo. If the survey only represented rich folk who could afford this newfangled telephone machine, then it could have been biased in favor of a candidate (Landon) that rich folks (presumably) liked more. Poor folks without telephones (presumably) liked Roosevelt better, so the survey didn't accurately predict the winner. This is a perfect, commonsense answer.
D) Why do political affiliations matter? The survey asked who folks were going to vote for. The folks said Landon. But in the actual election, Roosevelt crushed. How does "the survey didn't ask for political affiliations" explain why Roosevelt won?
E) Why are reasons important? The survey results showed that folks were gonna vote for Landon. But that's not what happened. I don't see how "the survey didn't ask for reasons why people were going to vote for whom they were going to vote" does anything to explain the surprising election result.
The answer is C because if only rich folks could afford phones, then this phone survey could have been biased.