helene_t, on 2016-November-10, 06:39, said:
A few random thoughts:
- The pollers didn't really get it wrong. As 538 notes, if you move 1% of the valid votes from Trump to Clinton (which is within the normal error margin of polls), Clinton would have won comfortably and the polls would have gotten all states bar NC right.
- The narrative that floods my facebook news feed is all kinds of generalizations of Americans, white Americans, old Americans, Working class Americans, American men. But within all these segments there is room for considerable disagreement. About half would vote for Trump and about half for Clinton, somewhat more in some segments than in others. Largely as expected. The differences between segments might in some ways be a bit bigger than usual. Anyway, a 1% deviation from the forecasts, while hugely significant in terms of political consequences, doesn't really justify a completely reversed narrative about the zeitgeist or of the mental health of Americans in general.
- For someone like me who never watches TV and very rarely clicks on a political video link on the internet, it feels incomprehensible that so many would consider Trump more trustworthy than Clinton. But maybe it is understandable considering that most voters will make there assessment largely based on TV. I have only seen Hillary a couple of times on video so it doesn't have that much influence on my opinion about her, but the little I have seen looks like fake smiles. The kind of facial expression that wouldn't pass a lie detector. If I place myself in the shoes of someone who suspects NYT of being about as biased as Fox News, and spends a lot more time watching politics on TV than reading about politics, then it is maybe not so surprising.
I think "The pollers didn't really get it wrong." is maybe technically right in some sense, but basically wrong. The polls weren't rigged, they reported what they were told, but this doesn't mean that they didn't get it wrong. This is different from, say, the World Series. After game 4, with Cleveland ahead 3 games to 1, no doubt the chances on a Cleveland win would have been rated as far higher than the 71% that 538 was giving Clinton on Tuesday morning. Chicago still won, but as the series evolved it was not settles whether Fowler would or would not hit a home run, it was only settles that he would try. The election is different. By the time of the late polls, most people had chosen how to vote. It seems unlikely that random changes of intent between, say, a Saturday poll and a Tuesday vote explains what happened. The voters had chosen who they would vote for, they did it, the polls failed to predict it. It is reasonable to refer to this as the polls getting it wrong. It need not mean that pollsters were stupid or incompetent, but I do think it means they got it wrong.