Washington Post has a neat op-ed piece about how a totally fictional parody pundit got the election more right this time than the “data journalists,” and it is surprisingly informative and insightful.
The truth is that statistics, by itself, is NOT science and the use of (big) data even less so. As I often repeat, statistics can serve science if it focused on estimating and quantifying how wrong our theories are, under what circumstances. But if the goal is limited to just predictions, without theorizing about underlying processes, it is either right or wrong, dichotomously. There is nothing to be learned from them.
This was, and will be a strange election. Everyone will be wrong, by huge margins. This is not something anyone should be ashamed of, as long as we can use how and where our theories are wrong to learn about how electoral process works and what was/is/will be different in 2016 that we can generalize to other contexts. From the looks of it, it won’t happen. There will be a lot of trashtalk and no one will want to dig through the data to learn. That would be abuse of science. That is how, in the example used by Vladimir Vapnik (and he got the example from Popper, I thnk) advanced astrology has come to incorporate hyper advanced math and plentiful data, but is still junk.