I have sent the following email to Keith Olbermann (who is on vacation).
Dear Mr. Olbermann,
You recently wrote here about:
"Two conflicting scholarly studies on the variance between the national exit polling and the presidential election results..."
My read of these studies is a bit different than yours. I do not think they are opposing. Each reaches some useful conclusions and they do not actually conflict.
(more below the fold -- and take the poll!)
The University of Pennsylvania scholar
Steven F. Freeman concludes that the variance is way outside the margin of error -- and statistically so improbable in the aggregate that some explanation is required. (The paper does not provide one, and says a conclusion that there was vote tampering is premature.)
The second paper (from the Caltech/MIT voting project) does not contradict anything in the U of PA paper as far as I can see. It debunks only one idea: that touch screen electronic voting machines were responsible for the unpredicted variance in Bush's favor. This is a narrow conclusion, and it's true. It was optical scan and punched card precincts that showed a variance in Bush's favor. If there was vote tampering, it was with the old fashioned equipment -- or rather, in the tallying process and perhaps the machines used to tally the results from those precincts -- because the touch screen machine precincts don't seem to be where the variance comes from.
Taken together, these two papers reinforce each other (if read carefully). The U of PA paper is the more significant because its conclusion is so important -- namely, that the conflict between the exit polls and the official results is real, and needs an explanation. But the Caltech/MIT paper also makes a useful contribution -- by showing that the variance cannot be explained by someone having hacked the touch screen programming. Therefore the explanation must be elsewhere. But we already knew that -- the U of PA paper clearly identifies optical scan and punch card precincts as contributing massively to the variance.
Finally, I refer you to a third analysis, this time of North Carolina. It's not a scholarly paper but it is interesting legwork nonetheless: http://www.dailykos.com/story/2004/11/13/9743/8602. What is interesting here is the comparison of absentee vs. polling place results. Apparently in NC about 30% of the vote was absentee, and it showed no variance (it matches the exit polls closely). All the variance (and there was a lot!) is concentrated in the polling place results. What does that tell us? Again, the mere fact that the variance isn't universal is suspicious. And it tells us that if there was vote fraud, it was somewhere in the tallying process on election day, and did not affect the absentee vote counting process. This conclusion also does not conflict with the two papers you wrote about.
The good news here seems to be that if there was fraud, it must have been in the paper ballot precincts, or somewhere downstream of there, when they were tallied. The hard thing to believe is that anyone would have dared attempt that on a massive scale. Because an honest recount should tell us what the voters actually said -- those paper ballots are not going away. We are in for some interesting times ahead.
Thanks for your pursuit of this story. I suspect you are right that it will be THE story all too soon, but you jumped on it with admirable initiative before anyone else in TV news, as far as I can see. May it remain "your story" as it plays out.