The SacBee gives us some follow-up information to the decision to tally votes centrally in Sacramento:
mproper maintenance of some of Sacramento County’s voting machines – and the tint of the Feb. 5 ballots – were to blame for malfunctions that sidelined vote-counting scanners and delayed results of last month’s presidential primary, according to the county’s top election official.
The problems have been corrected and the scanners are expected to be used in the June election, Registrar of Voters Jill LaVine said in a report to the county Board of Supervisors.
…
Because of the malfunction, all ballots had to be counted in the election department’s central command in south Sacramento – instead of some being processed as usual at the precincts.
…
During its investigation, the county said that the vendor that supplies and maintains the scanners, Elections Systems & Software, conducted improper recalibration and preventive maintenance on the machines in December.
My take:
- I’m glad to see that there was some sort of quality testing that occurred so that problems were noticed.
- I’m surprised that the issues were universal at all precincts. For a problem of that scale you would expect that someone (ES&S – the vendor?) would be contractually responsible for their mistakes.
- I’m not very comfortable with the resolution which was to use central tabulation. Central tabulation requires that there are observers of the counting and a strong chain of custody for the ballots. Precinct level counting is more robust in this regard. Perhaps a fall-back to manual precinct level counting might have been better.
All About Voting 
Precinct counting can be more robust, but also more susceptible to ballot manipulation due to less observers and, in general, smaller numbers of volunteers around.
Barring something like Punchscan or Scantegrity — I’d like to see both, where the scanner makes a copy, and one box gets delivered while the other gets counted locally, then results are compared.