Deanron's proposal to force everybody to vote by mail would increase the number of voted mail ballots per election by somewhere between 40% to 60%. If an election the size of November 2004 were held entirely by mail, King County would have to process another 330,000 mail ballots on top of the 570,000 that Deanron failed to process correctly in 2004. That would mean accounting for 1.8 million pieces of paper -- 900,000 ballots and 900,000 envelopes. How well would we expect Deanron to perform this task should his mail-only proposal pass the council? Not well at all.

Ballot accounting was a total mess in 2004. Now that I've finally received what purports to be a complete audit trail of absentee ballot processing, I can see that the accounting was even worse than I realized, and certainly worse than has been reported in the press. And it wasn't only 2004. After all the public scrutiny and (admittedly Mickey Mouse) blue ribbon panels, the mail ballot accounting in the fall of 2005 was still a total mess. Deanron has never specifically acknowledged how inadequate the system is, what specifically went wrong in 2004 and 2005, and what specific steps he has taken to correct the problems for future elections . [I invite King County Elections to respond to any and every statement in this post].

Follows are gory detail and specific examples (with data records and photographs!) of just how worthless the ballot accounting really is, and why no sane person should trust it to ensure that their vote is counted properly.

First, I want to offer overdue thanks and acknowledgement to everybody who dedicated hours of their own time to help me wade through absentee and provisional ballot envelopes last year -- Greg Andrews, Glenn Avery, Toby O'Brien, Dana McWilliams, Doug Timpe, Bob Edelman, Steve Haffner, Troy Beardslee and Mary Lane, not to mention the background experts who helped me make sense of it all. Without you, none of this would have come out! (If I inadvertently forgot somebody, let me know ASAP)--

Now for absentee ballot processing in a nutshell. Every ballot envelope is identified with an av_election_id number that is unique per absentee voter per election. It is printed on the return ballot envelope in digits and barcode just below the voter's address

There should be a database record for every ballot issued, identified by av_election_id. In some cases when a voter needs to have a duplicate ballot issued (if they spoiled the first one, for example) a second ballot will be issued with the same av_election_id number. In other cases (e.g the voter changes address), the old av_election_id number is "suspended" and a new ballot is issued with a new av_election_id number. When ballots return, they are first processed by a vendor who sorts them into batches of roughly 200-300 envelopes. The database is supposed to be updated with the "date returned" and the batch number for every record by av_election_id. However, ballots which have been "suspended" but are returned do not show up in the database as having been returned. They are supposed to be set aside as "Wanda rejects". [see Nicole Way deposition p. 29]. If the voter's signature checks out, the envelope is opened and the ballot is removed and tabulated (possibly "duplicated" if improperly marked). If there are no problems with an envelope or ballot the ballot is tabulated with its original batch. If a problem is discovered with the sealed envelope, such as missing or mismatched signature, the envelope is challenged, removed from the original batch and set aside in case the problem can be fixed (e.g. signature affidavit) by certification. The database record has a "challenged" field which is set to a code depending on the problem (e.g. NSOF for "no signature on file", or SIG for signature mismatch). A suspended ballot has the challenge code "SUSPD". If a signature problem is cured, the challenge code is deleted from the database and the envelope is placed in a new batch, called a "rebatch", with a new batch number. In general, if a database record has a non-empty "date returned" field, a non-zero batch number and no challenged code, that should correspond with a ballot that was counted and a voter that was credited. Unfortunately, it doesn't all match up. One of the big problems is that the database records (at least the ones that have been released to me) are not updated to show the new rebatch number when a signature problem has been corrected.

Here are a few specific examples of things that went wrong and for which there is no way to account.

1) Suspended ballots were counted [photo] See the audit trail here Scroll down to the "Absentee Ballots" table and look at the entries where election_id 1204 and then the corresponding entries in the Absentee Ballot Transactions table. This was the voter's first ballot which was suspended. The handwritten notes indicate that it came back with batch 329 (which would have been on Oct. 21. It was counted and there's no corresponding data record that it was even returned. There is, however, a data record showing that the voter's reissued ballot, with a different av_election_id number, was returned and counted on Nov. 4. I didn't have enough data at the time to know to ask to see the batch it was supposedly in, so I can't say whether or not the voter's second ballot was also counted. But either a voter had two ballots counted, or the data records purport that a ballot that was counted when it actually wasn't and that a ballot wasn't returned when it was both returned and counted. There is no accounting for returned, suspended ballots and no accounting controls or audit trail as a check whether suspended ballots are improperly tabulated.

2. Two ballots counted for the same voter [photo] [audit trail records]. This is not like most of the other double voters where people were registered twice (the ballot accounting part of the process wouldn't be expected to catch those). This is a case where the voter's ballot was reissued with the same av_election_id number. (Not suspended). The record shows that a second ballot was issued. Presumably the system is supposed to reject a second such ballot. But there is no record that the second ballot was returned let alone counted.

3. "Not in Batch" (I). The person who verifies signatures gets both the stack of batched envelopes and an ordered list of the av_election_id numbers of the envelopes in the batch. If the list contains an av_election_id number for which there is no corresponding envelope, the corresponding database record is given the challenge code "N-I-B", which means "not in batch". About 600 ballots in November 2004 were coded as N-I-B. But we found a number of the corresponding envelopes, open and empty sitting in either the original batches where they were supposed to be, or in different batches. [photo] [audit trail records]

It's possible that the envelopes were initially misplaced and later found, but they were never accounted for. (These by the way, are examples where the ballot was counted, but the voter not credited, explaining some but not all of the appearance of "more votes than voters" in mail ballots).

4. "Not in Batch" (II). Not all the N-I-B absentees were found. Or if they were, they were kept hidden. For several months in 2005 I was scratching my head over 93 voters who appeared to have returned an absentee ballot but were credited provisional. I asked for both ballot envelopes for each voter, They told me all they could find were 25 or so of the provisionals. The rest of the provisionals showed up sitting together at the top of a box that they tried to prevent me from seeing. (It also contained a lot of the fatal pends). The correspnding absentee ballots never showed up. I specifically mentioned these in my motion in Snohomish County Court, but Deanron's lawyer insisted that they don't exist. Here's a photo of one of these provisional ballot envelopes. Here are the audit trail records. On October 20 the absentee ballot was challenged as N-I-B, but on November 16 somebody decided that this provisional ballot should be counted How could they have been certain that these N-I-B ballots did not exist even though a lot of the other ones were merely misplaced? (all the other provisional/absentee ballot pairs in this category have a similar pattern in the audit trail).

I stress that these are only a few examples of what fell apart in 2004. We found varying numbers of similar examples for each category and only examined less than 5% of the total number of ballots returned. In general, King County ended up with roughly 1,000 more mail ballots tabulated than mail voters identified and for good reason -- for all intents and purposes there were are no viable controls to ensure that every valid ballot was tabulated and no invalid ballots were tabulated. This post was long enough. I'll follow up with specific problems from 2005 in the next post.