Reply #2: huh?

[OnTheOtherHand]

Your formula says that a 3-point error in Utah is impossible with 56% turnout rate because it would require 122% Bush voters among refusers?

Something has gone wrong here. As I understand your logic, you would have 798 responders and 627 refusers to get the 56% response rate, totalling 1425 attempted interviews. About 559 of the 798 UT responders were Bush voters (70.1%). 1039 of the 1425 attempts should be Bush voters (72.9%). So 480 of the 627 refusers should be Bush voters, which is 76.6%, not 122%. Somehow, your table is way too sensitive to small changes in response rate.

(Something else is wrong even before we start. A 3-point error in a 798-person sample can't be "impossible" even if there were 100% response rate, because there is sampling error.)