Several bloggers (Election Law, Volokh, Anthony Argyriou) have noted that certain counties reported zero undervotes in the recall portion of the special election. The idea is expressed that some of the new voting machines might be a problem. All 3 counties named used machines from the same company (Diebold). However, as California Insider points out, the same counties did report undervotes for the replacement portion of the election.
I'm going to suggest simple error in the tabulation or canvassing of the ballots in those counties, where some supervisor disregarded the non-votes as "irrelevant." Or too much work. Never assign to evil what you can assign to human error (or sloth). There is good evidence for this:
If you go and look at older ballot questions, such as the 2002 State Assembly elections you will see that certain counties reported zero undervotes in all Assembly districts within the county. These counties include San Francisco, San Mateo, Sacramento, Kern, Santa Barbara, Orange and Riverside, as well as some tiny ones. In 2002 these counties used a wide variety of voting devices.
Yet, in the same election, these counties all reported sizable non-votes for governor. This makes human error by the canvassers the most likely explanation. Some people just don't see the need to count non-votes.
Rick Hasen's correspondents have verified that your supposition is correct - the number of total voters was reduced by the 3565 people who didn't vote on the recall: http://electionlawblog.org/archives/000255.html
I've noted this in my blog, as well.
Posted by: Anthony at November 16, 2003 10:41 PM