November 15, 2006
Mail Ballot Lifecycle

For all the talk of vote-by-mail, most of the public discussion centers only on the voter's experience (getting the ballot in the mail (or not), returning it on time, coping with the possibly that one's signature was rejected). Little is reported on the complex industrial operation needed to process the hundreds of thousands of mail ballots voted in a countywide King County election. The Elections staff graciously gave me a comprehensive tour of the mail ballot processing facility yesterday. Here are captioned photos that depict the lifecycle of an incoming mail ballot, from intake through tabulation --

The first wave of ballots that arrive by mail or are dropped off at polling places are presorted and batched at an outside vendor, PSI. By now, the late-arriving mailed ballots and the rest of the polling place ballots are batched in house.

This is a rack of the (correctly sealed) polling place "blue bags" full of absentee ballots waiting to be unloaded and batched.

2. A lone worker scans the bar code on every incoming ballot envelope.

3. Another worker takes the file of bar code numbers from the previous step, loads the file into the DIMS system, and assigns a batch number to the ballots. He also produces a "Wanda Errors Report", which identifies certain invalid ballots, such as duplicates from the same registered voters. This worker also starts the "batch slip", a paper slip that tracks the number of ballots in the batch and various exceptions throughout the lifecycle. The form of batch slip now in use appears to be an improvement over the form that was used during 2004.

4. Batches in trays waiting for the next stage of processing.

5. The signatures on the outer envelopes are verified against the voter registration record. Rejected envelopes are separated and voters notified to correct any correctable signature problems.

6. Citizens Elections "Oversight" Committee chairman and unrepentant mail ballot afficionado AJ Culver is shown standing as close to the signature verification process as the average citizen is allowed to get. (not close enough to see anything interesting). Official party observers are allowed to walk around and stand over the shoulders of verifiers and register any protests with supervisors. In practice, there are not emough observers to provide much coverage.

7. Verified ballot envelopes are taken to the slicing room where these frightening machines slit open the envelopes.

8. A room full of "openers" separate the ballots, the security envelopes and the outer envelopes. The openers also inspect the ballots and cull any that look like they can't be tabulated by machine and need to be "duplicated". The Openers Checklist lists the various tasks to be performed on each batch.

Close-up of openers.

9. Numerous staff perform quality control on the opened and separated ballots, accounting for all pieces and reconciling the counts.

Close-up of opening quality control

10. Dozens of "duplicators", working in teams of two, divine voter intent and "duplicate" improperly marked or otherwise defective ballots. I don't have official numbers, but a Republican observer tells me that a large number of ballots were improperly printed this year and are not machine readable, therefore the number of ballots that require duplication is expected to be higher than normal.

Close-up of duplication team

11. This is the tabulation room. Several workers, each with an Accuvote machine, feed stacks of ballots into the machine.

Close-up of tabulator

When the Accuvote encounters an unreadable ballot, it stops feeding from the stack and the worker outstacks the unreadable ballot and places it on top of his machine, so the ballot can be duplicated and rerun.

Observers are not allowed in the tabulation room (I was allowed in, only because I had an escort). Observers may watch the tabulation through a window, but it is not possible to see all that much. Tabulators are accountable only to run a specific number of ballots per batch. There are few safeguards to prevent a tabulator from running the same ballot through the machine multiple times and skipping other ballots, inadvertently or otherwise. This does happen, as Republican observer Tim Borders discovered in the post-primary recount:

KCREALS identified 13 instances during the recent Pope/LaSalata/Ottinger recount in which it was learned that some ballots had been fed through the tabulators multiple times during the initial count

Overall, the staff I met seemed knowledgeable and professional and demonstrated improvement since the 2004 debacle. (I'll withhold final judgment on this year's performance until after certification when I can examine the reported and purportedly reconciled numbers). I have serious general concerns about the time, expense and manpower required to process this quantity of ballots; about the subjective nature of certain decision points (signature verification and "duplication"); about the opportunities for potentially outcome-affecting human error (or worse) and the fact that there is so little public oversight to catch such errors. I personally don't believe this process is sufficiently reliable to deliver a precise result in a very close contest, nor would I care to trust my own ballot to such handling.

Posted by Stefan Sharkansky at November 15, 2006 12:48 PM | Email This
1. Seems like a significantly inefficient approach to ballot processing - and rather than fix the underlying problems, they add more people and more checks to try and catch errors.

A better approach would design a voting process from beginning to end that reduces the opportunity for error to occur in the first place. They need to get some operations management specialists to create an efficient processs that eliminates errors from occurring, rather than trying to catch errors created by the inefficient and error prone process.

My understanding is that King County, WA is the furthest behind of all vote counters in the nation. And with 20% of the ballot bags "unverifiable"... on that latter topic, those ballots would be considered tainted evidence inadmissable in a Court trial since the chain of handling cannot be verified. How can we count ballots whose authenticity can no longer be verified?

While I do not live in King County, KC pretty much owns the state elections. And if KC cannot accurately track and count its ballots, we all suffer.

Posted by: Ed on November 15, 2006 01:08 PM
2. Very interesting, thanks, Stefan. It helps increase confidence in the system if it is open and transparent.

It should be a concern that a ballot can be run through the machine several times. In a close race a dishonest worker could change the result. Why don't the Accuvote machines mark the ballot in some way, say a bar code, that marks it as having been read. A ballot without the bar code would mean that it hadn't been scanned.

Posted by: Obi-Wan on November 15, 2006 01:16 PM
3. And King Ron declares that this is preferable to poll tabulation? Looks like a fraud-friendly, unmonitorable, time-wasting, result-delaying jobs program to me!

Stop this madness. Change your 'absentee' status to poll voter today and get your lazy rear to your neighborhood school two or three times a year, otherwise you will not know yours was even counted.

Posted by: Concerned Citizen on November 15, 2006 01:17 PM
4. It's so much better to replace all of these union meatheads with the distributed model of polling place voting. There, each ballot can be scanned by an accuvote machine after it has been inserted by the voter. If there are undervote, overvote, or unreadable ballots, they can be fixed and addressed by the voter in real time, with real intent. And the cost is low in that all of those voting places can be churches, homes and other "free" precinct locations, staffed by volunteer or other community helpers.

And due to the largely distributed nature of polling place voting, there are not a lot of ballots in any one place making it much harder to affect the outcome of the election without carefully organized fraud.

And there's no time penalty for voting because done well, there's no reason why every person's precinct can't be within a half mile or so of their house. If one has properly researched candidates and measures in advance, it only takes but a few minutes to vote and enables all of the whiny reasons cited for the disabled or those who just like curling up in the easy chair and studying the ballot. The only difference is that one then fills out a sample ballot for use at the polling place on election day.

There's a litany of cited lies to justify all-mail-voting, but they all boil down to substituting convenience for security. Those willing to indulge a temporary convenience are opening the whole process up to more risk and cost. Wasn't it JFK, a Democrat that said "Ask not what your country can do for you?" Voting at the polls once a year seems like a pretty low bar.

But I suspect the all-mail-voting sham is mostly just driven by government union lobbies which see plenty more cushy jobs for dolts that will detract from our GNP.

Thanks for the report Stefan.

Posted by: Jeff B. on November 15, 2006 01:23 PM
5. Stefan, you said:

''I personally don't believe this process is sufficiently reliable to deliver a precise result in a very close contest, nor would I care to trust my own ballot to such handling.''

Ditto.... This post brings back (bad) memories of the 2004 machine and manual Gov-race recount epics, where I was an (R) observer for both on several weekends. I remember Tim Borders was in charge of the (R) crew at the machine recount while I was there, and he probably did as good a job on oversight as anyone could reasonably be expected to do, given the circumstances.

But the circumstances were atrocious:
I still recall at the machine recount, with markers and white-out and who knows what all flying all over the place, where in one case a KCE supervisor took a ''problem'' ballot in hand, and disappeared with it BY HIMSELF into the dim, dark recesses of the KCE ballot facility for several minutes. When he came back and handed it to the people at the tabulation tables, he said something like ''it's O.K. now''. What did he do all by himself to that ballot ??.. nobody knows.

I was astounded at the chaos, and the general lack of any effective dual-accountability custody chain during the machine recount.

And then of course at the manual recount I had the opportunity to view (from a few feet away) the infamous ''Christine Rossi'' write-in ballot. When they ended up counting that one for Gregoire, I pretty much threw up my hands.

Anyway: It sounds like things may indeed have improved some since 2000... RELATIVELY speaking, at least. But effective, comprehensive public oversight is clearly still not possible.

Posted by: Methow Ken on November 15, 2006 01:34 PM
6. Washington is NOT a "voter intent" state.

Rather, it is an "interpretation of voter intent" state.

Part of the standard report (if it isn't already) should be the number of ballots (both raw and as a percentage of the whole) requiring duplication. Process improvements could then be put in place to try to reduce the need for duplication.

The Accuvote machines should be set up to reject a ballot that has already run through the machine. If a correction was made that required the ballot to be run again, a manual override would have to be consciously done.

Again, in the processing of registration and ballots, potential spots where fraud can be introduced or simple human errors can occur should be looked at and processes, machinery, ballots, or people should be changed to reduce the potentials for contaminating the vote counts.

Also, if someone has a ballot rejected because they already sent it in, are they sent a letter informing them? What if their absentee was stolen and voted and they were forced to vote a provisional - their absentee (the fraud) would be counted and their provisional (the real vote) wouldn't - how would they ever know? How many of these situations happen every election? That should be tracked as well.

Posted by: SouthernRoots on November 15, 2006 01:47 PM
7. Ah yes, I remember the ballot in question. If I remember correctly, it wasn't "Christine Rossi," but "Chris Rossi." I could be mistaken.

Either way, it's obvious to all but the kool-aid drinking trolls that it was a biased "divination" of voter's intent. Personally I would have called in impossible to "divine."

That example and the one where a mark in the margin was "divined" to be a vote for the queen gave me all the evidence I needed to know the fix was in.

Posted by: Obi-Wan on November 15, 2006 01:48 PM
8. Thanks Stefan... this up close and personal "how things work" is fascinating!

Can I ask... was becoming a ballot counting and election mechanics expert/vanguard/aficionado in your career/hobby plans?

Posted by: Ragnar Danneskold on November 15, 2006 01:52 PM
9. //The Accuvote machines should be set up to reject a ballot that has already run through the machine. If a correction was made that required the ballot to be run again, a manual override would have to be consciously done.//

Better yet, the first time a ballot is run through a machine, it should be marked with a unique random number; a database should then list each ballot ID along with its contents. If a ballot appears to contain different markings on the different times it's read, the ballot should be sequestered for forensic analysis.

Posted by: supercat on November 15, 2006 04:20 PM
10. OT, but are those 19" flat panel monitors that the signature verifier use? Why? Why can't old 17" tubes or 15" flat panel be good enough? There is a reason why Gov't needs more money constantly.

Posted by: C. Oh on November 15, 2006 05:20 PM
11. What do they do when, as mentioned, it is discovered that some ballots were fed through multiple times?

I remember hearing a report back in '04 that an employee was caught feeding the same Gregoire-voted ballot through five times. I wonder why that doesn't happen more often.

Posted by: Michele on November 15, 2006 06:54 PM
12. This site sometimes reminds me of a review of a Jello Biafra appearance I read a decade or so ago..."'It's a conspiracy,' he summed up every few minutes."

Posted by: vic on November 15, 2006 08:15 PM
13. Hi Stefan-

Pretty good pictures. It's really hard to get a sense of the flow of the operation without seeing it firsthand. Even then, it's just a bunch of paper being moved around by people until you hear the explanations.

Next time you're at the central count (TEA), I encourage you to take a snapshot of the flowchart on the 2nd flow. It shows many of the steps used to run an election in "swim lane" format. And at a glance, it shows how mail ballot processing is much more complicated than poll ballot processing.

I'm recreating the flowchart from a bunch of snapshots I took. But my camera sucks, so it's a bunch of small pictures that I have to stitch together.

Cheers, Jason

Posted by: zappini on November 16, 2006 04:16 AM
14. Stefan, that was an excellent job of displaying the mail ballot process at King County. Yes indeed, it is complex and the staff conducting the process have to be on their toes and follow the procedures exactly in order to prevent errors. But that is true of the entire election process.

The biggest drawback to poll voting is obtaining sufficiently qualified and able poll workers. I am the Inspector of the largest precinct in the largest county in another state. I was fortunate to have an able crew for this last election but some of my contemporaries were not so fortunate. Besides, we were all in our late sixties and early seventies, and that thirteen hour day is rough on us senior citizens. Most of the younger folk are employed or otherwise engaged and not available. So, we hold down the fort.

If elections were held on the weekend, or if general election day were made a national holiday freeing up others to be available to work the polls, poll voting might work better. Just some food for thought.

Posted by: Bob on November 16, 2006 12:04 PM
Post a comment

Email Address:



Remember info?