i want to tie what i said about ohio back to the 538 article, because despite coming short on expectations, i think i beat them pretty solidly on the math.
they were claiming - based on demographics and polling - that ohio should be more like michigan, and illinois and missouri should be bigger clinton wins. it was in the form of "more black people, therefore more clinton" and "ohio is white, so it should be more like michigan".
i said - no way. check the voting history. which states are liberal? which are conservative?
and, i think the results solidly debunk all of those arguments. based on those kinds of arguments, the white working class state of ohio should have solidly backed sanders and missouri should have backed clinton by the ten points they claimed.
i made one error - i was off by about ten points all around. i called missouri and illinois for sanders by ten and ohio a functional draw - she won ohio by a little less than ten, and missouri and illinois were a draw.
so, i got the pattern right, at least - i'm just off by ten points. the aggregates and demographic modelling didn't even get the pattern right.
why was i off by ten points? turnout. either in terms of myself exaggerating turnout, in terms of turnout being less than it could have been or in terms of the stuffing counteracting it is not clear. but, it's reduced to an error in projected turnout nonetheless. and, that shifted the results in all three states by about ten points.
--
i want to be clear on that point, as it's not trivial, right.
i looked at the polling and said "ok. but they're underpolling independents. so, i'm going to shift the results."
i did that in michigan, and it got me closer - but it wasn't enough of a shift. tonight, this kind of thinking continued to beat the models, but it was too much of a shift - about 10%. consistently.