or am I missing something?
Yes, thanks for asking.
Most everyone predicted a Hillary win. But the serious predictions were not just a selection of this candidate or that, but had odds attached to them made on the basis of polling data, with MoE’s, voter turnout models, and models of the cross-correlation of the variances in state-wide polling data. The latter attempts to model departures from randomness in the how a candidate may do, withing the polling range, in one state versus another.
As the election approached, especially in the last ten days, the polls tightened and the best model (538), while still favoring Clinton, had Trump with a 1 in 3 chance of winning. Trump’s victory facing 1 in 3 odds is an upset of sorts, but not one that calls into question the entire edifice of polling. (I’ve rolled snake eyes from time to time even thought the odds are 1/36 against.) I don’t recall seeing serious polls that gave odds beyond 20:1 - perhaps there were. At the simpler level of predicting the aggregated vote, rather than the electoral college outcome, the polls did very well.
When people opine that the outcome of this year’s election, compared to predictions shows that statistical methods of inference are fake science, that the modelers are hacks who don’t know what they are doing, or that they will never believe another poll, those opinions are not justified by what happened, and represent an alarming step toward the dark ages.
Perhaps people don’t really who vent such opinions really don’t mean them. But I think that there is a strong sense of anti-intellectualism in the Trump fans, and fear that the lasting accomplishment of Trump will be to destroy our commons sense and respect for knowledge. That will not be the end of the world - no other country is buying into that - but it will be the end of our leading edge.
