This post is in reponse to another DU'ers question: "What was your take on the Berkeley analysis?"
I will make some comments about the Berkeley (Hout) paper below - but first I want to express the opinion that the purpose of the statistical analyses is to provide just enough doubt to cause a complete audit of all voting records. If we stay in 'academic debate mode' long enough then we may eventually come to a consensus about the right way to do the analysis and some agreement about what the results mean. We don't have the time to debate. I have become fond of the "Prove that my vote counts NOW!" calls posted elsewhere on DU. We don't have to have evidence of fraud - all we need to have is concern - and it is 'their' job to show that fraud did not happen.
The main point I wish could be corrected in the Berkeley paper (and in other posts and news stories is the claim that we can't audit the vote. We can! Bev Harris and BBV are auditing right now. Even if E-Vote machines don't print voter verified paper receipts they do report to poll booths and the poll booths print 'poll tapes' at the end of the day. The poll tapes are then public records and (at least in Florida) are signed by lots of officials. If hackers planted software to add or shift votes at the level of the individual machines then the poll tapes from Evote precincts are useless records. BUT if they hacked at the level of the central tabulators that add up all of the votes, then the original poll tapes won't match the reports printed by the central tabulators. We can audit the poll tapes and the poll books (which should show the number of people who signed in to vote) in the Evote counties. In the opscan and punch-card counties we can recount the individual votes. I think that they hacked the vote by a wide enough margin so that people would not call for a recount because it doesn't seem as if a recount would put Kerry ahead. I think we have to counter this belief and do the 'forensic analysis' of all available materials in all suspicious states.
Comments about the Berkeley (Hout) and U Penn (Freeman) studies: One reasonable critic of both analyses is Marc Blumenthal at
http://www.mysterpollster.com Blumenthal's 11/23 post reports on critiques of Hout's paper. The critics were very harsh - particularly Michael McDonald and Andrew Gelman. The upshot of the criticisms was that the critics believe that Hout's entire result is due to only two outliers: Broward County and Palm Beach County. This suggests to the critics that there was almost certainly not any systemic fraud. If there were, it would have showed up in more than just two counties (they say). They argue that the Jewish vote swung Republican in those counties (they offer one anecdote, but no evidence to support this hypothesis). Blumenthal notes, however, that anyone who wants to continue investigating possible fraud in Florida anyway should focus on Broward and Palm Beach. Sounds okay to me, but I would still insist on recounting the whole state.
Blumenthal's 11/19 post seems to indicate that he is convinced that Stephen Freeman's methodology and conclusion is fair -- that the probability of 10 out of 11 swing states going to Kerry in the exit polls and to Bush in the vote tabulation is very, very low indeed -- absent an error in the exit polls *or* vote tabulation. Of course, Blumenthal concludes that the error is in the exit polls (and quotes Morin's WP article in his 11/22 post to this effect). Blumenthal's conclusions make an excellent point: All Freeman's paper (and other analysis like it) can do is to indicate that the swing from exit polls to vote results is extremely unlikely to have occurred by chance. The ONLY way to decide why this is so is to EITHER have complete data from Edison-Mitofsky so that we can see exactly what happened with the polls OR to recount the votes. I say let's recount the votes!