By Jim Fox, Lead Democratic Observer
As Democratic Observer I attended both most of the Election Day Counting and today’s Hand Count Audits. Per state law, ballots were pre-processed (checked for validity and readability) as they were received prior to Election Day.
On Election Day some 28,000 “clean” ballots were machine read (but not totaled until 8:00 pm). Three (Sequoia 400-C) industrial quality counting machines are in a secure basement room. There is no network (Internet, etc.) connection to the outside. Access is pass controlled, and as always workers never work alone. The machines had a Secretary of State “Logic and Accuracy Test” August 12th. Those results (still in the machine) were printed, counters zero-ed, and an empty Initial Report printed and approved. Every activity is logged on a master machine log.
Staff Operators ran batches of (typically 100-150) ballots through the machines (max 400/minute). Thanks to new silicon treated paper, there were fewer feed-problems and the machines stay cleaner than previously. Ballots that are misread are diverted to an “Outstack” and ballots with Write-Ins to a 3rd bin. Misreads are reviewed for problems and then either re-read, or diverted back to the Checking Team.
If a jam or misfeed occurs, the count can be stopped, rejected and re-run. When the batch is complete, the count is accepted. As everywhere, Log-In and Log-Out counts are strictly accounted. “Lost” or damaged ballots are quickly found and accounted.
During machine scanning, results are saved in machine/run files. Randomly during the day Party Observers called for six hand-count audit batches, 2 from each machine. These were set aside. Normally once counted, batches are sealed in boxes for archiving per state/county procedures. All pre-processed ballots were scanned by 6:30 pm and the machines secured. Observers from one or both parties were present most of the day.
After the polls closed at 8:00 pm, we returned to retrieve totals. Raw totals are printed and a copy of raw-text data is transferred by flash-drive to the county data network computers upstairs, where the Preliminary First Results were published about 8:45 pm. Some 25% of registered voter’s ballots were counted by Election Eve.
Meanwhile, on Election Day many new ballots were received, from drop-boxes around the county. As received, they began pre-processing through the multi-step log-in, signature-check, opening, checking, and if necessary cleaning or duplication. In Whatcom County this tightly structured procedure helps ensure that “voter intent” is tallied correctly, and that counting is trouble-free. Other states do differently.
For the next 2-weeks ballots will arrive by mail and be counted for final certification by the 3-member Canvassing Board on September 3rd. Roughly half of expected ballots have been counted.
This morning, six audit batches from yesterday’s machine-counts were hand counted. We had 2 experienced staff Counters, with 2 Supervisors and 7 Observers. Each batch of 100-150 ballots, was first sorted into piles for each of the possible (10) Governor Candidates (or blank). Counters exchanged piles for a check of their sort. Then individual piles were counted, logged, exchanged, recounted and logged. The process was repeated for ballots with Write-ins (which must be tallied manually later). Totals were then compared with the machine-count.
For five batches the (1-machine and 2 human) counts agreed. In one batch, one ballot with all votes circled was immediately spotted by the Counter. While “intent” was clear, it could not be machine-read correctly, and should have been “duplicated” as part of the checking pre-process. Everyone agreed with what happened; it was not a machine or programming count error; it was a pre-checking process error. One formal request for Hand-Count Escalation was noted. The batch will be returned to Checking for pre-processing and re-count, the other batches will be archived.
In general we in Whatcom County have a very good system. We have good machines, good materials, increasingly good procedures; and exceptionally dedicated staff at every level working as hard as possible to ensure that “voter intent” is tallied properly. How the inevitable exceptions to the standard exceptions are handled and resolved is the test of any real-world data-processing system. Election tallies are by nature a complex process, complicated with human variables. However, within the inherent limitations of human-marked, folded and mailed paper ballots, our processing is relatively secure, very-traceable, and accurate.
Observing is an interesting, necessary and enlightening activity. I encourage everyone to volunteer for training as an approved Observer sometime.