yes, i can present an obvious source of the error in the modeling, otoh, and that's actually with black voters.

now, i'm just hypothesizing this; i'll leave it to somebody else to look it up. but, black turnout in 2008 was through the roof, and he got more than 99% of it. i can't imagine any modelling that would have given any candidate 99% of anything, or at least not anywhere west of moscow.

and, that might be an ironic part of the answer - it could be that a part of the reason the wilder effect didn't show up with obama is because he overpowered it with exceedingly strong black support.

you'd have to look at the old models and crunch the numbers and figure it out, but i'd guess you can get a substantive amount of the discrepancy that way, without even looking at whether white voters lied to pollsters or not.