Here are the press reports on the first big ruling in the Attorney General recount:
The AP - Court rejects Democrat's argument in contested AG's race
The Free Lance-Star - Ballots will not be rescanned
Richmond Times-Dispatch - McDonnell gets preliminary court victory
Roanoke Times - McDonnell wins panel decision on recount
Virginian Pilot - Panel's recount ruling goes McDonnell's way
Washington Post - Judges Exclude 26% of Ballots in Va. Recount
The issue was whether all of the "optical scan ballots" in the state would be rescanned. The panel ruled that only where the computer tapes were illegible or contested would the Court consider recounting of individual ballots in those jurisdictions. Optical scan ballots are but one of the several methods of balloting used in the Commonwealth.
The Court's ruling will have the effect of reducing the number of ballots that will be recounted by hand. The secretary of the State Board of Elections indicated that if the 500,000+ optical scan ballots were to be individually recounted, more than 130,000 of those would have to be recounted by hand, because "machines used in nine jurisdictions -- including Martinsville -- could not be reprogrammed to count only the votes in the attorney general's election." The Roanoke article explained: "If they could be reprogrammed, the machines could set aside for inspection any ballots in which there were write-in votes, no votes or more than one vote in the attorney general's race. Without reprogramming, recount officials would have to count each ballot by hand."
The Deeds campaign cited the undervote in Chesterfield County and Virginia Beach. I would not have thought that new votes found in Chesterfield or the Beach would be likely to decrease the margin favoring McDonnell.
The Deeds campaign based its arguments on the magnitude of the undervote. As described here, there are always fewer votes case for Attorney General than for Governor, and a spread of 5 points between the percentage of ballots recording a vote for governor and the percentage of ballots recording a vote for Attorney General is not uncommon, in the figures shown in Larry Sabato's Virginia Votes 2001, available for download here from the Center for Politics website. In other words, the existence of a 40,000 vote undervote is not surprising - the lack of such an undervote would be more surprising, if not unbelievable.
Also, I can't get out of my mind the conclusions from one commenter to this post on Rick Sincere's blog, where he concludes:
"If you examine the undervote rate -- that is, the number of people who went to the polls but didn't vote in a given race -- for the Lieutenant Governor and Attorney General you see that the average rate for optical scan systems was 2.8% for each race. The undervote rate using DRE machines, though, was 50% higher in each case: around 4.0 - 4.2%. A statistical analysis of varaince [sic] shows that the only significant factor to explain this increase in voter apathy is the technology used, NOT 'voter satisfaction' or even locality size."
The comment links to this site, with analysis of the Virginia vote, which suggests that that 99% of the voters who turned out cast a vote in the Governor's race, 97.16 % cast a vote in the AG's race, 97.01 cast a vote in the LG's race. The site has this summary of the undervote for AG, broken down by type of voting machine.