Breaking the big binder's code: Part I

Figuring out how King County attempted to reconcile their ballot totals is difficult, but perhaps not impossible.

This is the first part of a hypothesis that may explain what they were trying to do after election day and presumably before the county’s election returns were certified by the canvassing board on November 17, 2004. (If my hypothesis doesn't completely fall apart, then there will be at least one more part to show whether the 348 provisional ballots that King County admitted back in January were improperly cast were included in the 660 provisional ballots that King County finally admitted on March 11 were improperly cast. I hope to figure out whether their reconciliation summary shows the total of such improperly cast provisional ballots is closer to 1008 than 660.)

This first part attempts to explain the reconciliation notes and summary for the 7 precincts that used the Bothell Regional Library as their polling place. Copies of the summary and the notes were posted at Sound Politics by Stefan Sharkansky.

Two presumptions are employed in this explanation:

• People who were issued ballots, whether provisional or regular, would ordinarily cast those ballots before departing from the polling place. Absent a notation from a polling place official who noticed someone taking a ballot out of the polling place in violation of the law, it is presumed that ballots issued were cast.
• It is likely that some ballots, both regular and provisional, were issued from the wrong precinct’s stack of unused ballots to voters who had signed a poll book. Absent evidence to the contrary, it is presumed that the number of ballots issued in the polling place should balance with the number of ballots cast in the polling place – even though the numbers of ballots issued and cast in any one or more of the 7 precincts don’t balance because of the mistaken issuance of a ballot from precinct “A” when the voter had signed the poll book for precinct “B.”

The crew of elections office personnel assigned to canvass each precinct’s ballots was trying to account for the ballots that had been issued and resolve any discrepancies between the number of ballots issued and the number of ballots inserted into the AccuVote machine at each polling place and counted during the election night tabulation.

The election night machine count of ballots inserted into the AccuVote machine at the polling place and the “add-on” ballots cast on election day and included in the tabulation provided the crew with the number of ballots cast. (Provisional ballots were, in a sense, “cast,” but they should have been in envelopes so they could be processed later in much the same way that absentee ballots are processed. Properly cast provisional ballots, then, wouldn’t be included in the election night ballot and vote count.)

The information available to the crew came from the poll books, the ballot accountability forms filled out by precinct officers, the AccuVote record of the number of ballots inserted into the machine, and the tabulation of ballots and votes on election night.

At some point, the poll books were scanned using a “wand” to capture the voter identification information contained in the bar codes next to each signature of a voter who signed the poll book before being issued a regular ballot. That scanning produced a number of signatures which would be represented on the poll book cover sheet by a canvassing crew member’s note, e.g., “WANDA 171” for Bothell Precinct 01-0254 (“BOT 01-0254”). The image of that cover sheet was posted at Sound Politics by Stefan Sharkansky.

The canvassing crew counted the signatures by hand in at least some of the poll books. For example, the cover sheet for BOT 01-0254 shows a notation which says “Signature 180hc.”

That hand count of signatures included both the signatures of people who were issued regular ballots and the signatures of people who signed in the back of the poll book as required by WAC 434-253-043 before they were issued provisional ballots.

The canvassing crew determined the total number of ballots cast by adding the number of ballots scanned by the AccuVote machine and the “add-on” ballots. [“Add-on” ballots were of two types: (1) ballots inserted into the bin below the machine when the voter decided not to insert them into the machine or decided not to correct a ballot rejected by the machine; and (2) ballots which were rejected by the machine which were then inserted again into the machine at the voter’s request while a polling place officer overrode the machine’s ability to reject ballots which the machine read as containing an “over vote” (that is, more than one candidate for an office was marked by the voter) or “blank vote” (that is, no votes at all on the ballot).]

Using BOT 01-254 to illustrate, here’s how the comparison of ballots and signatures worked:

• 180 signatures were hand counted in the poll book, including 8 signatures of people who signed before being issued provisional ballots.
• 172 signatures for regular ballots were then deduced to be in the poll book (180-8=172), and the “WANDA 171” count of signatures for regular ballots was treated as an incorrect count.
• 177 ballots had been scanned and counted by the AccuVote machine.
• 2 “add-on” ballots had been tabulated on election night.
• 179 ballots was the total of votes cast and counted on election night.
• The canvassing crew member then subtracted 172 from 179 and determined that there were 7 more ballots in the election night count than there were signatures for regular ballots.
• Since only regular ballots should have been counted on election night, the crew member noted on the poll book cover “+7” for BOT 01-0254. (That same +7 appears on the reconciliation summary in the “Plus/Minus” column.)
• The crew member saw that 4 of the 8 provisional ballots which had been issued were not present, and made a note to record that fact on the poll book cover sheet: “(4 no label).” [“No label” refers to the absence of the labeled envelope which should contain the provisional ballot. The “label” was a stick-on label containing information that the voter was required to provide so that the voter’s registration status could be determined.]
• Since the election night tabulation reported the presence of 7 more ballots than the number of regular ballots issued on election day in BOT 01-0254, it is presumed that some of those 7 ballots were the “4 no label” provisional ballots – and that the other 3 ballots were mistakenly issued from the BOT 01-0254 stack of unused ballots to voters who signed the poll books of another precinct.
• Although the source of the number is not explained in the notes, the crew member put a note on the cover sheet of the BOT 01-0254 poll book indicating that 31 “no label” provisional ballots had been inserted into the AccuVote machine at the polling place.
  This number (31) was later put into the column titled “Adjusted #” for the entire polling place in the reconciliation summary prepared and released by King County on Friday, March 11.

The way in which that “Adjusted #” of 31 was figured out wasn’t explained in any narrative report that I’ve seen, so this is my attempt to explain how it was derived from the information available to the canvassing crew.

If the canvassing crew member had gone one precinct at a time, the same result could have been reached and explained in a narrative. There were 32 more ballots tabulated on election night than there were regular ballots issued at Bothell Regional Library polling place. Of those 32 improperly cast ballots, 31 were apparently provisional ballots issued at that polling place and inserted into the AccuVote machine rather than into signed and labeled envelopes for later verification of each voter’s eligibility to cast a ballot. The source of 1 of the 32 ballots cannot be explained with certainty, but it was probably an absentee ballot which someone inserted into the AccuVote machine rather than submitting the ballot in a signed envelope as required by law.

Here is a precinct by precinct explanation of how I believe the available information could be used to arrive at the conclusions reflected in the canvassing crew’s notes and the summary of their work for the polling place at the Bothell Regional Library:

• BOT 01-0253 had 3 “no label” ballots and had 3 more ballots in the count than the number of regular ballots issued. It appears, then, that all 3 of the “no label” provisional ballots were improperly inserted into the AccuVote machine. For this precinct, the “Adjusted #” to account for the presence of those 3 extra ballots should have been “-3,” leaving a zero “Net Difference.”
• BOT 01-0254 had 4 “no label” ballots and had 7 more ballots in the count than the number of regular ballots issued. It appears, then, that “-4” should have been put in the “Adjusted #” column to represent the 4 “no label” ballots which presumably were inserted into the AccuVote machine, leaving a “Net Difference” of +3 unexplained ballots for this precinct. Note that BOT 01-0259 could explain those 3 extra ballots. BOT 01-259 had 2 “no label” ballots and also had a shortage of 1 regular ballot in the count – for a total of 3 ballots that seem to have disappeared from that precinct. It is presumed that “crossover” errors resulted in the issuance of 3 ballots from the BOT 01-0254 stack of ballots to people who signed the BOT 01-0259 poll book. The “Adjusted #” column should then read, “-4-3,” leaving a “Net Difference” of zero for this precinct.
• BOT 01-0259 had 2 “no label” ballots and had 1 less ballot in the count than the number of regular ballots issued. This precinct, then, appears to have had 3 ballots disappear after issuance. It seems probable that the apparent disappearance is the result of a “crossover error” in which people signed the BOT 01-0259 poll book, but were issued ballots from the BOT 01-0254 stack of ballots. That would leave this precinct short by 3 ballots and the other precinct with an extra 3 ballots in its count – exactly what appears in the numbers. The “Adjusted #” should have been +1 to reflect the “crossover” issuance of a regular ballot from the wrong stack, leaving a “Net Difference” of zero for this precinct. [Somewhere in the notes should have been an explanation that the 2 “no label” provisional ballots noted in BOT 01-0259 had not actually disappeared, but were instead ballots taken from the BOT 01-0254 stack. The way the form is set up, those other 2 ballots wouldn’t be included in the “Adjusted #” column for BOT 01-0259, since only the apparent shortage of 1 ballot in the tabulation is being explained by the “Adjusted #” column.]
• BOT 01-0260 had 2 “no label” ballots and had 2 more ballots in the count than the number of regular ballots issued. Presumably, the 2 "no label" provisional ballots were inserted into the AccuVote machine. A “-2” should have been put in the “Adjusted #” column, making the “Net Difference” for this precinct zero.
• BOT 01-0262 had no provisional ballots which turned up missing, but did have 1 more ballot in the count than the number of regular ballots issued. There aren’t any other “no label” ballots from other precincts which could explain the presence of this extra ballot. It shows up in the summary for the polling place in the “Net Difference” column. The presence of this extra ballot cannot be explained, but it could have come from one of two sources. It could have been a ballot issued to someone at the polling place without having that person sign the poll book. I don’t know if the canvassing crew had all the unused ballots from this precinct available to them, so even if they were inclined to count those ballots to verify the number of ballots issued, I don’t know if they could have done so. Since the AccuVote machine would accept an absentee ballot coded to identify it as from BOT 01-0262, this extra ballot might have been an absentee ballot that some unknown person inserted into the AccuVote machine rather than into a sealed and signed envelope as required by law. The source of this extra ballot cannot be explained with any certainty, so it is included along with 215 other King County ballots in the “Net Difference” column of the summary.
• BOT 01-3271 had 10 “no label” ballots and 16 more ballots in the count than there were regular ballots issued. The 10 “no label” ballots could explain part of the extra ballots in the election night count, but not all 16. Note that in BOT 01-3398, there were 6 “no label” ballots that weren’t needed to explain the number of extra ballots in that precinct’s count. It appears that “crossover” errors occurred in issuing provisional ballots, so that these two precincts can balance each other out. The “Adjusted #” should be “-10-6,” based on the presumption that the 10 “no label” ballots issued by BOT 01-3271 were inserted into the machine and that the other 6 extra ballots in the election night count came from the mistaken issuance of ballots from the BOT 01-3271 stack to people who signed the poll book for BOT 01-3398. The “Net Difference” should be zero for this precinct.
• BOT 01-3398 had 10 “no label” ballots and 4 more ballots in the count than there were regular ballots issued. The 4 extra ballots in the count presumably resulted from inserting 4 of the 10 “no label” ballots into the machine – leaving 6 “no label” ballots that were signed for in this precinct’s poll book, but not apparently in the possession of the polling place officers. See above – those 6 were probably mistakenly issued from the stack of ballots coded for BOT 01-3271. The “Adjusted #” for this precinct should be “-4” based on the presumption that 4 “no label” provisional ballots were issued and improperly inserted into the AccuVote machine, leaving a “Net Difference” of zero for this precinct.

This hypothesis would explain how the numbers of ballots issued and ballots counted were reconciled, but note that ballots which were improperly inserted into the AccuVote machine rather than into signed and labeled envelopes for later verification of the voter’s eligibility to vote weren’t transformed into legitimate ballots by this process of figuring out where they came from.

What this type of reconciliation accomplishes is to provide some assurance that the presence of too many ballots in one precinct’s count or too few ballots in another precinct’s count didn’t result from forged ballots brought into the polling place and "stuffed" into the AccuVote machine or from the "ditching" of ballots. It shows that (with the exception of one ballot) the ballots cast at the polling place were apparently issued at the polling place, and that ballots cast were subsequently included in the vote tabulation.

Unfortunately, there is no indication that Dean Logan’s reports to the other canvassing board members informed them that the adjustments made during this reconciliation process only made it appear that there was a slight variance between ballots issued and ballots cast in each polling place – and they were not being informed of the magnitude of the variance between ballots legitimately cast and ballots counted.