Following up on this post, Justin Levitt of the Brennan Center sent the following very interesting message to the election law list, reprinted here with permission:

Rick’s report of Weinschenk v. Missouri makes a very important point about the overall balance of potential harms involved in voter ID laws. Another element of his posting, however, deserves a brief clarification, especially in the context of reports on any irregularities that may occur in the upcoming election.
Rick accurately quotes the dissent’s characterization of available evidence on fraud. Regrettably, however, Judge Limbaugh’s characterization was itself inaccurate, and incorrectly bolsters a report often cited as proof that fraud occurs.
In the snippet that was disseminated, the dissent claims that “A subsequent report from then Secretary of State Matt Blunt noted, as even the plaintiffs have acknowledged here, that 79 voters registered from vacant lots, 45 people voted twice, and 14 votes were cast by the ‘dead.'”
On page 9 of the report itself (http://bond.senate.gov/mandate.pdf), Secretary Blunt was more precise:
“Based on information provided by the City and County Boards of Election Commissioners (the “City Board” and “County Board”, respectively), it is highly probable that twenty-three (23) people voted more than once in the November 7 election, and it is likely that an additional forty-five (45) persons voted twice.
“Based on information provided by the Missouri Department of Health, fourteen (14) persons in St. Louis City and County who were reported as deceased before the November 7 election nevertheless are recorded as having voted in the election.
“Based on information provided by the City Board, it appears that seventy-nine (79) voters who were registered from vacant lots in St. Louis City voted in the November 7 election.”
(emphasis added).
Later in the report, Secretary Blunt revealed the basis for these three assessments. The first two do not necessarily reveal fraud. And the last has been conclusively disproved.
For the double-voters, investigators used voting history entries in St. Louis City and County computer databases to determine whether individuals were listed as voting twice. 23 voters who were apparently listed in both databases as voting in the same election matched name, date of birth, and social security number; 45 voters matched names and date of birth. Although I do not have ready access to the actual list of compared names, if the names are even relatively common, it would not be surprising to find 45 pairs of distinct voters whose name and date of birth match, out of more than 613,000 votes cast in St. Louis City and County in 2000. And though it is, of course, possible that the voters whose name, date of birth, and social security number were actually double voters, before determining that such a conclusion is “highly probable,” I’d want some rough approximation of the rate of clerical errors — which have been mistaken for double votes before. (See, e.g., http://www.jsonline.com/story/?id=350183)
For the deceased voters, investigators compared a list of individuals who died in St. Louis City and County from 1990-2000 against the same voter databases, and found 14 voters with the same names, dates of birth, and social security numbers. Again, it is possible that these votes were fraudulent. But before so concluding, in addition to examining the clerical error rate, it would be useful to double-check to see whether any of the voters on the deceased list actually died after the election. (See, e.g., Marcia Myers, Election Theft Ruled Out, Baltimore Sun, Aug. 24, 1995, at 1A.)
Finally, for the 79 voters apparently registered from vacant lots, the St. Louis City Board of Election Commissioners provided a report of addresses “verified to be vacant by the Board” and the names and voting history of individuals registered to vote at those addresses. Exemplary follow-up investigation by the St. Louis Post-Dispatch found that the supposedly vacant lots were, in fact, not vacant; indeed, some apparently contained buildings more than 50 years old. As a Demos report by Lori Minnite and David Callahan reported, “Errors in the city’s property records and methods for classifying vacant a multi-parcel address if only one of the parcels at the address is vacant” apparently accounted for the errors. (See, e.g., http://www.demos.org/pubs/EDR_-_Securing_the_Vote.pdf at 49 & n.88).