Reply #120: From your responses

[Febble]

to OTOH, I gather that you might be asking what size" of discrepancy I might find suspicious. I would agree with OTOH - the size might be an indicator, but not necessarily the best one. I think 2000 was corrupt and that the wrong man became president, despite the fact that exit poll accuracy was fairly good. In 1992 is it was poor, but I don't particularly suspect fraud, although it is possible. In 2004 I definitely suspected fraud, because of the size of the discrepancy, because I suspected fraud anyway after 2000, and because of DREs. But it doesn't seem to be correlated with DREs or with advantage to Bush. So now I am less convinced.

What would convince me far more than exit poll evidence would be if swing from 2000 was correlated with DREs. Or with some other attribute that we might consider, a priori, to be associated with fraud. The shift could be quite small, but we would have far more statistical power, because the analysis would not be confined to a few handfuls of NEP precincts in each state.