Reply #110: maybe, but quite possibly the opposite is true [View All]

Data from individual precincts are likely to be very noisy indeed; one would hope that by pooling precincts, one would get a better idea of what is going on in a particular state. When we are discussing what we believe to be 6- or 8- or 10-point discrepancies at the state level, there is no a priori reason to expect them to wash out in state-by-state "noise." But precinct-level analysis is especially useful for variables that cut across state levels. For instance, a lot of us sort of expected the WPEs to be a lot higher in touchscreen precincts, but they aren't.

Sure, the exit polls are unreliable for several major things, including state projections, which is why the pollsters resist calling states without having some official counts to look at. When I say that I have pretty much zero confidence in the reliability of exit polls as fraud detectors, I don't mean that they convey zero information, but that they can't be relied upon in themselves.

I can't "rule out" much of anything. Heck, for all I know the Warren County Board of Elections was conducting a blood sacrifice to corrupt the Edison/Mitofsky database. But there is no obvious inconsistency between what we knew about the exit polls on election night and what we know about them now. They showed Kerry ahead then, and they show Kerry ahead now.