This is for those who want to know the New York Times polling methodology ( as you showed in the link you provided, thans ):
The New York Times and Kaiser have produced a sample in Arkansas that reports they voted in 2012 for Romney over Obama—by one point. But Romney carried Arkansas in 2012 by 24 points. Similarly, the Kentucky sample is +3 Romney when reality was +23. The Louisiana sample is +3 Obama in a state Obama lost by 17, and the North Carolina sample is +7 Obama in a state he lost by 3.
The whole point of question 12 is to provide a reality test for the sample. That’s why they ask that question—we know what happened in 2012, so the only thing to be learned by asking the 2012 question of the sample is to ensure that it’s a reasonably accurate snapshot of voters in the state. Of course there’ll always be some variance between reality and the sample’s report of its vote a year and a half ago—but not a 23 point variance.
A reputable news organization would have looked at question 12 and thrown the poll out. But then again, it was the New York Times.