the error is common and widespread across academic disciplines; it's called generalizing the specific. here's how this works. nate comes up with a pretty good model for the 2008 election - and it was a pretty good model. then, he assumes that this model is universally applicable, perhaps with just minor tweaks. so, he should be able to apply it to every other election. this is how you do science, after all, right?
but, understanding elections is not like building physical laws. the basis of science is repeatably demonstrable experiments. if you drop an apple on monday, it falls. if you drop it on tuesday, it falls. if you drop it next friday, it falls. it's always the same. that's physics. well, that's newtonian physics, anyways. but, what if you were to change the gravitational constant sometime in the middle of the week? well, you can't - not when you're doing physics, anyways. they're constants.
when you're analyzing elections, the constants are going to fluctuate dramatically year over year or even month over month. you could wake up to a single event that will throw everything out of balance. or, demographics could shift midway through the cycle. that is where nate's error really is: he's assuming that his 2008 model is a universal truth, rather than a good analysis of a specific election.
because he's made this bad assumption, he then needs to defend it. see, you're misinterpreting this as some kind of bias. it's really just his ego trying to uphold the universality of his model, based on the flawed assumption that you can have a universal theory of elections in the first place.
trump's win is largely a function of his opponents' errors and the weakness of the field. i don't think anybody really saw it coming from a distance, but when you saw the way things unfolded, it became obvious what was happening. there simply wasn't an electable candidate in the field. i don't know if you blame bush for being lacklustre or if you blame the banks for taking too long to figure it out. i don't know if you blame rubio for being transparently false or if you blame cruz for splitting the vote. but, he won by controlling the centre of the republican spectrum in the vacuum of a moderate candidate. i still think that kasich would have cleaned his clock, one-on-one, from an early point - if he had comparable media coverage. but, it just upheld what we know about open competition: it's not the best candidate that wins, but the one that is the most violent. if he had to fight against a strong, centrist candidate over a smaller field then he would have probably lost every state. so, i don't think he was wrong; i actually said something very similar, and i do stand by it. i just think trump's opponents were particularly incompetent, in ways that were hard to predict from a distance.
what he should have done with the democratic primary was build a new model with new information. reality is that there's lots of blacks that voted for obama because he was black, and it was reasonable to build a model around it. but, nobody is voting for clinton because obama is black. people may be voting for clinton because she's methodist, because she's female, because she's a hawk, because she's more conservative, because she'll tax them less - or, because they're old. that's the biggest takeaway, right. clinton wins when the voters are older, and loses when they're younger. had he crunched those numbers and built a new model, with new data, it would have probably been pretty good.
but, universalizing the specific means you don't do that. it means you assume the apple falls the same way every time. and, it means you have to defend your work against scrutiny.
https://www.youtube.com/watch?v=gP7fxNdAs5E