Interesting post here from Sean Trende on how gingerly the pollsters and analysts can treat anything like actual analysis. And believe you me: they do, in fact, worry themselves sick on the subject. I was struck by this bit in particular:
…this fear of getting it wrong is probably creating a similar herding effect among analysts. In late 2014, Nate Cohn of the New York Times could claim that he wasn’t aware of a single theory for why polls would be biased toward Democrats in 2014. Given how Republicans outperformed the polls in that election, there should have been one (and in fairness, FiveThirtyEight subsequently published a piece that included just such a theory).
Part of the reason the theory didn’t exist might be that journalists and political scientists tend to be left-of-center, so they subconsciously resisted creating hypotheses that favored Republicans. But even people on the right shied away from constructing a pro-Republican electoral theory. A likely explanation for that hesitancy is that, after the unskewed polls debacle of 2012, few wanted to risk suggesting that the polls would be biased toward Democrats, and chance suffering the humiliation that would follow if they were proved wrong. There is safety in numbers for analysts as well.