Reply #24: let me see if I understand your question

[OnTheOtherHand]

I think this parallels what some of us have been talking about in e-mails.

One argument would be that in the absence of vote suppression, turnout in heavily black and Democratic urban precincts should be expected to equal turnout everywhere else -- so we can notionally set turnout in these precincts at the county mean, or the state mean, or whatever, and see how many votes were stolen.

The counterargument would be that maybe, even given an effective Get Out The Vote operation, turnout in these precincts is likely to be lower than in other places.

One side would say that long lines are direct evidence that (attempted) turnout was high. The other would say that we don't know how long the lines were everywhere, and that if a precinct with historical 50% turnout adds some new registered voters and then experiences long lines, we still can't assume that the increased turnout would hit some average.

And, you would propose that instead of trying to sort this out just by staring at Ohio figures, we try to gain additional perspective by looking elsewhere.

The problem with LA and NYC is that neither is in a swing state, although I have no info on the intensity of the GOTV efforts there. Philly might be best. Clearly there won't be a perfect "control," because wherever there is GOTV, there may be suppression of various forms. Cincinnati could serve as a baseline of sorts. I reckoned that its turnout was about 69% (of registered voters -- a risky standard of comparison because some counties purge their lists more stringently than others, and because I don't think Cincinnati has any schools as enormous as Ohio State, so I am inclined to expect extra stale registration lists in parts of Columbus).

Scatterplotting turnout by % Kerry, I find that turnout in Cincinnati is everywhere higher (on average by % Kerry, that is!!) than in Columbus; that turnout in both cities drops markedly as % Kerry increases; but that the gradient is steeper for Columbus. Just eyeballing the graph (this is not at all a rigorous modeling exercise!!), turnout in Cinti is about 75% at 40% Kerry, about 65% at 90% Kerry. Turnout in Columbus is about 60% at 40% Kerry, about 45% at 90% Kerry.

Sigh -- there is an entire literature, including Febble's, that I'm not addressing at all, as well as some other people's work in progress. Just a little raw empiricism to stir into the mix. I will try to do more -- juggling too many things right now.