Reply #84: two things [View All]

[OnTheOtherHand]

First of all, given the premise of the report, if the best that can be said is that their hypothesis "can't be ruled out," that is a big problem, isn't it?

Second, neither of these effects has to be very "huge," although the retrospective margin bias in some contexts is. (The biggest one I've seen was among respondents in the 1993 General Social Survey, who reported having favored Bush over Dukakis by something like a 41-point margin. Actually, I guess there is one about JFK of a similar magnitude. I wouldn't expect anything like that in a midterm exit poll.)