A Day As Observer
by Milo Clark

October 4, 2004   

 
(Swans - October 4, 2004)   My question: "Did I see too little and learn too much?"

Hawaii's primary election on September 18, 2004 found us in the Hawaii County Counting Room as Official Observers, "the eyes and ears of the public."

Following late July telephone conversations with Dwayne Yoshina, Hawaii's Chief Election Official, regarding registration questions here, he asked if we would like to be Observers for the 2004 elections. "Gladly" was our answer.

My wife, Lee Maniscalco, and I began our involvement this year with the Democratic Party district four precinct caucus on February 24th. Contrary to past years, there was an eager and enthusiastic crowd motivated largely by Dennis Kucinich's progressive agenda.

Toward the end of June, we took the class to become Deputy Voter Registrars (DVRs) and away we went! Routinely, we spend part of each Sunday at the Makuu Farmers' Market, a community event which attracts hundreds every week. We get to talk to a lot of people and to hear their views on politics and government.

In processing voter affidavits, we run into the quirks of rural life and the varieties of residential addresses out here. The directory of streets and possible addresses given to us as DVRs has problems. We learned how to bolster descriptions to get people registered. We learned about limitations, too.

Tracing the people on our half-mile of rural lane, we found about half misclassified in some respect. We found that we have been in the wrong precinct since we registered in 1999. Some people who didn't live on or own property on our lane were included. Another sample using identical addresses found people who lived at the same address and shared a last name were also misclassified into wrong precincts.

Trying to clarify these situations ran us into the County Election Office attitude that problems were the voter's fault, not theirs. Enter Mr. Yoshina. After we had cited the relevant law to him, he wrote a letter to Mr. Al Konishi, Hawaii County Chief Election Officer by virtue of being County Clerk.

For a while, Mr. Konishi pretended we didn't exist. Eventually, a meeting was assembled on a minor point. The meeting included a Deputy Hawaii Attorney General, a Democratic Party National Committeewoman, Mr. Yoshina, Mr. Konishi, a League of Women Voters representative, a representative of the county Planning department, my co-chair from the District Four Voter Task Force, Bill Eger and myself.

Mr. Konishi danced about on minutiae and trivia for a couple hours. Mr. Yoshina had his opinions. County Election Office staff proved us wrong in their estimation. We repeatedly made the point that a sample is an indicator. To "correct" a symptom deals little with the larger problems involved. On the positive side, we got Mr. Konishi's attention. I'll call that a start.

Mr. Konishi subsequently found areas of common interest about which we could talk in non-threatening ways. I call that a good start.

In preparation for the Primary Election, Observers meet at least twice to cast sample ballots as a means of testing the machines used at precinct levels.

In 2004 Hawaii elections, two types of balloting processes will be used. The Election Software and Systems (ES&S) has been the balloting system used in Hawaii since 1998. They replaced an IBM 1081 punch card voting system. Both ES&S and IBM use paper ballot forms suitable for their processing systems. The second system, being somewhat hastily introduced, is the Hart InterCivic eSlate Direct Recording Electronic (DRE) system. The Hart system is completely electronic, although not one of the "touch screen" DRE systems as such. The Hart eSlate uses a or dial and push buttons.

At the first Observer sample balloting session, all 71 plus backup ES&S M100 precinct counting machines were present. Each was tested in several ways. Starting and summary tapes, PCMIA data cartridges, along with the physical ballots used for testing by Observers, were sealed and padlocked to protect them from possible intrusions, etc.

Although promised, no Hart eSlates were available. A personable young man of Japanese ancestry from Hart's Colorado plant extolled the wonders of the Hart system. All blow and no show, in other words. Also, public demonstrations of the Hart machines scheduled for Big Island locations were canceled and not rescheduled. Score one for the personable young man. Take away several points for no show and canceled demonstrations. Not a good start for Hart.

While Hart eSlates were shown to precinct workers during short training sessions, Observers were not to see them until the second testing session. Then a bank of nine machines were lined up on which to run the same test balloting patterns earlier cast on ES&S M100 machines. The Hart eSlates throw out feet of adding machines-sized tape showing results. Of systems import is the fact that all ES&S machines to be used in Hawaii islands' 71 precincts plus backup were tested. Only nine of the Hart eSlates were tested. Unfortunately, we didn't pick that difference up at the time.

In a separate session, Mr. Yoshina arranged for Hart executives from Colorado and Austin, TX offices as well as a technician to show off their machines. At that session, we thoroughly grilled the executives and put a sample of machines through their paces. There we learned that Hart has a triple redundancy system running identical data and summary data through three electronically separated processes. One of those processes called SERVO also has the capacity to print out each individual ballot cast using a Hart eSlate. Although no printer was present for us to see ballots printed on paper, we did see the screens showing ballots on a laptop. We were promised a sample Hart cartridge and samples of SERVO printouts, a promise as yet unkept. Score one for SERVO. Take away a couple more for reneging on promised cartridge and printed sample reports.

With background in place, we can now move to primary election day in Hawaii County. Lee and I arrived, as instructed, before 8:30 am for what would be a long day. In a previous Observer meeting, Lee was assigned as Observer to the ballot duplication team. I was assigned to the poll book audit team.

The day starts with processing absentee ballots of which there were over 4,000. Absentee ballot envelopes have been date and time stamped on arrival. Names and signatures are said to be compared to registration records, although we did not see this step in process.

Absentee ballots arrive in a cover envelope within which there is a sealed yellow envelope containing the actual ballot in ES&S format. The ballots are mailed out in a larger envelope for which the ballot has been folded. The return envelope is one size smaller and the yellow envelope is two sizes smaller than the envelope for which the ballot is folded. To put a completed absentee ballot in the proper yellow envelope to fit within the return envelope, the ballot needs to be refolded or the yellow envelope flap has to be closed so that the glue can stick to the enclosed ballot. Doesn't take much observation to conclude that the mail form absentee ballots will be either folded in complicated patterns or that glue may be stuck to them. Either case means that the ES&S M550 ballot counting machines will have problems handling them.

Picture then, a large table surrounded by women who first sort the return envelopes by precincts, count and double count them and enter numbers on report forms. Once sorted, counted and double-counted the absentee ballot return envelopes are opened. Some go to a slitting machine. Others are opened with handheld slitters. The slitting machine, unless very carefully tended and envelopes battered down sharply, will slit some of the yellow envelopes and some of the ballots. The hand slitters can also damage contents. That this damaged relatively few of the total was a commendation to the women doing the work.

The return envelopes are inspected and reinspected to make sure each contains a yellow ballot envelope. Then, the yellow envelopes are separated from the return envelopes, sorted and counted. The white envelopes are carefully inspected and reinspected to see that no yellow envelopes escaped. The white return envelopes are put into big metal boxes to be sealed and padlocked for storage.

Next, the yellow ballot envelopes are opened, ballots separated, given a preliminary unfolding and straightening before being put in white boxes for transport to the ES&S M550 machine room and processing. As you would now expect, the yellow envelopes are inspected and reinspected to assure that no ballots are missed. The yellow envelopes also go into big metal boxes to be sealed and padlocked for storage.

In the ES&S M550 machine room, the ballots are backfolded, smoothed, straightened and rolled as well as stacked under heavy weighted boxes to flatten them enough to make the M550 counting machine happy.

When stacks of ballots reworked and reworked as best as possible are fed through the M550, the machine stops about every 5 or 6 ballots to kick one out or to jam one or more in the machine's innards on and on and on for several hours. The very experienced machine operator uses every trick he has learned over the years but jams are much too frequent. Absentee ballot design and the related envelopes used to handle them are primary causes of many, too many difficulties.

As the ballot format is confusing in organization and layout, too many ballots are rejected for voting more than one party (cross) or more than one candidate for an office (over). We don't have hanging chads in Hawaii, we have crosses and overs. Whatever the name, far too many make mistakes and lose their votes. Overall, 9,559 ballots in the state end up as defective, that is, their party candidate ions are uncounted.

In Hawaii County, only one County Council race was close enough so that defectives could have influenced the outcome, but that candidate decided not to challenge the results. The Honolulu mayoral race, however, was decided by a difference less than the state defective ballot total unrelated, though, in fact.

Working with Observer Todd Belt, a political science professor at UH-Hilo, we went through hundreds and hundreds of ballots checking crosses and overs and defectives (illegible marks, crossed out ions, etc.).

Some defective ballots will show a voter intent clearly through crossing out mistakes or other indicators of intent. In those situations, if Observers of both major parties agree, a new ballot can be created by the duplicate ballot team and processed.

As the back side of the ballots held the non-partisan offices, although the party side of the ballot was rejected as cross or over or defective, the non-partisan votes can be counted assuming, of course, that these choices were not overs or defective for other reasons. In any case, ES&S people can shut off three sorting parameters and count these ballots.

All absentee ballots need to be processed, summarized and results sent up line to State levels before polls close at 6:00 pm. Given the problems with absentee ballots as received, the ES&S team barely squeaked in before the deadline.

I also roamed around the counting room looking at what was going on or not going on in other stations. As nothing will happen until polls close and ballots and precinct level ES&S M100 counting machine PCMIA cards are received, Observers wait patiently.

The Hart representative sat behind his desk going through his files, checking his programs and watching his laptops. We get acquainted and establish a rapport. He too will have little to do until polls close and his equivalents of the ES&S PCMIA cartridges begin to come in along with the Hart machines from precincts.

Earlier, Susan Irvine, Observer Chairperson, changed my assignment from the poll book audit to the manual audit team. No one is assigned to replace me on the poll book audit team. When the manual audit team arrives, I go over to the other end of the room to watch and to learn what they do.

On getting organized, the manual audit team, mostly experienced workers on this project, first a total of seventeen precincts to audit. One large, one medium and one small; one urban and one rural and, randomly, twelve others.

Parallel processes now work to, one, bring in the ballots in sealed brown cardboard boxes for deposit in the ballot storage room and two, generate ES&S printed reports by precinct as available.

After collecting the precinct's ballots from the ballot storage room, the manual audit team looks at the ballot and s one office to audit. The ballots are fanned out around the table and sorted: candidate, other than candidate or short or over or defective one way or another. Sorted, the ballots are piled into tens before entering the numbers into a form designed to count by tens. Slashes are entered for each ten counted and the leftover ballots are noted as "+ X." The total of slashes by ten plus the leftover are entered in the total space. This method of counting and related report format date from then olden days of hand counted paper ballots.

Introduction of Hart results in Hart format will not fit this method. That actuality will not be known until quite late in the night when Hart results for the Manual Audit Team to process arrive.

Once printed results come in, if there is a difference in count or classification, the Manual Audit Team recounts and recounts until printed results and their count agrees. In cases where problems involve classification of cross or over or defective, sometimes the team will back into agreement with the printed results.

Problems arise, also, when ballots get through the precinct ES&S M100 or otherwise are included in a precinct's ballots. Classification of these ballots causes most of the disagreements in precinct numbers. These classifications pop up again and again when attempting to reconcile totals into summary levels.

I spent most of my time for most of the evening and into the early morning at the Manual Audit Team table. From time to time, another Observer comes by for a minute or two and then wanders off again. I have a sense that she has also been assigned to the Manual Audit Team but Susan Irvine said nothing to me about her role.

I checked in at the ES&S M550 room, I watched what was going on at the Hart station and talked more with the Hart representative, John Gartrell, from the Houston office. I noticed that the Observer assigned to the Hart desk was a woman who, in earlier conversations, displayed little or no knowledge about electronics, computers or voting machines in general. She spent most of the day sitting at a desk in a corner of the room.

Late in the evening as the Manual Audit Team finished the ES&S system precincts, Susan Irvine came over with the woman I sensed may also be assigned to the Manual Audit Team. Susan called Merle, the Team chair, away to a table in back of where I was sitting. I was not invited to join this session which turned out, in retrospect, to be crucial to dealing with Hart results. Susan brought with her, I discovered later, a set of envelopes which contained Hart summaries by the seventeen precincts to be audited along with printed versions of the Hart ballots cast on the precinct eSlates.

The Manual Audit Team had no prior instruction or information about handling Hart results. There had been no initial preparation to print out Hart facsimile ballots for use by the Manual Audit Team. John Gartrell, the Hart representative, had to scrounge a laser printer over the objection of the Hawaii Election Office program manager. A printer cable was taken from another office so that Gartrell could hook it up and print.

The Hart SERVO program gives two options for printing facsimile ballots. It can print four or five per "8.5 x 11" page or one per page. SERVO shows only the actual ions made by each voter for each ballot cast. In contrast to the ES&S system's ballot where a candidate shows up in the same place on each ballot, Hart just makes a list of candidates chosen by a voter. To find a candidate on the ES&S ballot, look at the same place on each ballot to see if there is a mark. Once a mark is seen or not seen, ballots are stacked in piles accordingly. To audit Hart "ballots," the report form used for ES&S ballot audit and report turn out to be useless. The Manual Audit Team ended up reading through the Hart ballot printouts and making tallies on the table top cover. Fortunately, there were few ballots cast by Hart in the tested precincts.

Seeing this problem, I went to John Gartrell and described the situation. Hart facsimile ballot printouts ganged on a page are awkward to work with. He told me that California, which requires 10% of ballots be manually audited, chose the one ballot per page alternative. I visualized mountains of paper printouts. I took him back to the Manual Audit Team table to show him the problem.

There is a simple solution which, apparently occurred to no one. Provide the Manual Audit team with tally sheets to use for Hart facsimile ballots which can be summarized and entered into the report forms. Lacking tally sheets, the team scribbled tally marks on the paper covering the table they worked at.

At no time during the day, evening and morning did I notice an Observer associated with the Poll Book Audit team. From time to time, I would spend a few minutes watching what was going on and having brief conversations with the team chairman. He showed me a few pages from some books and told me what they did. Each member of his team had a calculator on which they entered page totals as reviewed. They had no formal procedure as I observed related to the yellow poll book which also records precinct chair's comments during the day. Later, I learned that both sets of poll books are sent to Pearl City election office for processing. Processing, I was also told, collected the totals by precinct up to state levels. Yellow book comments, etc., were transcribed. Both reports, I understand now, are for internal use only, not made public.

The Poll Book entries and totals are a key check point on election processes. With adjustments for spoiled ballots and such, poll book totals must very closely approximate the number of ballots cast by precinct. I do not yet understand why this information is not available.

In the early morning confusions, Susan Irvine said that both the Manual Audit Team and Poll Book Team were lacking data or submittals necessary to complete their work. At that time, this information was due at approximately 5:00 am. She suggested we go home and come back around 9:00 am Sunday morning to complete this work and to lock down all remaining ballots, reports, papers, etc., due to go into storage for the requisite 22 months.

We returned around 8:30 am to find both Manual Audit Team and Poll Book Audit Team tables cleared, related report forms and Hart results envelopes gone, no one present and no answers available as to what happened. As Observer of the Manual Audit Team, I voiced my objection and concern.

Later, Al Konishi (County Clerk and Chief Election Officer), Susan Irvine (Observer chair), George Matsuo (staff Counting Room Coordinator), myself and, from time to time, Lee Maniscalco spent considerable time attempting to reconcile the ES&S results with added Hart results. Al made several speakerphone contacts with Scott Nago (State Counting Room chief) in Pearl City on Oahu. Scott said he would send by fax a 50-page report with relevant data. He said he would bring a hard copy of that report which we could look at on Monday afternoon prior to the Observer Debriefing session planned for 5:00 pm. Al Konishi said we should get together to go over this new information with Scott.

Lee and I arrived about 4:00 on Monday afternoon. The Counting Room was locked. In the Election Office we were told it would be opened about 5:00 for the Observer Debriefing. No information was offered about where Scott Nago or Dwayne Yoshina or Al Konishi may be or if any messages had been left. Our antennae go up sharply and search for input.

At the Observer Debriefing session, questions about shutting down both the Manual Audit Team and Poll Book Audit Team were deflected. No acknowledgment of data reconciliation problems was forthcoming in other than general terms. When I asked Scott about availability of Hart results and precinct level results, I was told maybe by the end of the week on the website.

On Wednesday, while in the County Election Office, I talked with Al Konishi about our Sunday experience. He had his notes available and we talked some more. He put in a call to Scott Nago and said he would let me know what came up.

By e-mail, Al Konishi asked for a meeting on Monday to go over results which Scott promised for then. He later forwarded a report from Scott containing Hart results for the state by precinct. This report covered about 50% of state Hart ballots cast, around 6,500, and consumed 2,005 pages.

Obviously, some serious problems need work.