Washblog

We do not need to buy any more of the existing tabulators.

Laird Hall, our elections' computer guy, told the King County Council during testimony that KC only owns 40 central count tabulator units (AccuVote OS optical scanner plus AccuFeed ballot feeder).

A few weeks ago, Barbara Whitt showed the documentary "Commander in Thief". She invited Bill Huennekens, the vote by mail project manager, raincity calling, and myself to handle Q&A afterwards. Huennekens also stated that we only have 40 units.

In truth, we own 60 (SIXTY) units. The Jones/Lazarus review of the tabulator business case states on page 14 that King County currently owns 60 functioning (usable) units.

Why would Hall and Huennekens lie about something that is so easily fact-checked? It boggles the mind. I expect my opponents to vigorously argue their case. But lying is just too much. Our elections officials have asked us to trust them. They're actually a little bit hurt when their integrity is questioned. But when they lie about something so banal, don't you wonder just a little bit what else they're lying about?

I'd also like to know the penalty for lying to the council. (Too bad people aren't sworn in to testify.)

Faster Election Results

We counted 152,249 ballots on election day using just 16 units for the 2000 general election.

King County's historical turnout data

That's ~9,500 ballots per unit per day. 60 central count units is more than enough. With 60 units, we could count ~570,000 ballots per day.

I estimate that ~330,000 (roughly 1/3rd) of all returned mail ballots will be ready to tabulate on election day. Huennekens/REALS state that ~600,000 ballots will be ready to count on election day.

My estimate relies on historical data from the Sec of State and REALS (e.g. Logan's April 2006 report). The rule of thumb is that for general elections, 1/3rd of mail ballots will be ready to tabulate on election day. (The fraction is bigger for low-turnout primary and special elections, because those are the motivated regular voters who generally return their ballots.)

I believe REALS's estimate must factor in divine action by the Ballot Fairy, who will conjure the missing ballots on election day. (I'm looking forward to seeing that. I want to believe. But having not seen the Ballot Fairy, I remain skeptical.)

I covered the "fallacy of early results" before:

Reckless Plan to Buy Diebold's New Ballot Counters

Ballots counted per unit per hour would be a better metric. Alas, I haven't done the leg work to figure that out. The batch accountability spreadsheets indicate a current average of 12.5 ballots per minute. But tabulators are not counting the whole time. Batches of ballots have to be fetched, loaded, unloaded, returned to the cage, etc. And I don't yet know the whole cycle time.

Procedures

I don't know why central count tabulator is slower today than in the past. For the general 2004, 233,254 mail ballots were counted using 40 machines. That's ~5,800 ballots per unit per day. Just 61% of the throughput 6 years earlier.

One theory is the "stop the line" procedures are the cause. When a ballot isn't processed correctly, the processing of that batch is stopped and a supervisor comes over to see if it can resolved. Versus outstacking the ballot (for duplication) then and there, allowing processing to continue.

One could argue we need these new procedures to prevent the problems from 2004 from recurring. But as council member Gossett, our elections officials, and others, have pointed out: tabulation was not the problem in 2004.

One could also argue that "stop the line" is a part of some total quality improvement initiative (that "six sigma" stuff). Having been a QA manager, I learned the hard way that all those quality assurance notions are rules of thumb, not rigid doctrine. Improvement requires a balance, to optimize the entire system, not just some individual aspect. (Also known as the "theory of constraints".)

Another theory is the tabulation in 2006 was slower because of the (still unacknowledged) ballot printing problems Diebold inflicted on our county. Diebold bought new printers in 2006 and apparently had to rediscover how to apply ink to paper (in register). That's why we had to manually duplicate an unprecedented 16% of mail ballots (at taxpayer expense).

Conclusion

We don't need new tabulation equipment to handle the increased number of mail ballots anticipated with vote by mail. Ever.

King County Elections just make stuff up to bolster their argument for blowing nearly $5m on all new crap.