Archive: The Last Jelly Bean: An Election Observer's Diary

In November of 2004, I had the privilege of serving as an election observer in the general election. Below is an account of that experience.

Early October: I’m on the phone with Hawai’i County Clerk Al Konishi when he brings up an intriguing proposal. “Did Gretchen [then-Journal Editor Gretchen Kelly] mention to you that we need a member of the press to act as Poll Watcher?” he asks. “I think you’d be ideal....”

It turns out that the reason I would be ideal not because of my journalistic brilliance, but because of the Journal’s deadline schedule. On Election Night, the reporters from the island’s two dailies are likely to be busy putting together last-minute stories about who won and who lost--and whoever would be in the actual vote counting area would be under virtual quarantine, lest partial results be leaked prematurely. But as a Journal correspondent, I can be more leisurely about my deadlines. Al and I agree that if I serve as an observer, I can write about the whole experience, so long as I write after the election is over and the results are announced.

I later learn the story behind the invitation. Two activists, Bill Eger and Milo Clark, had discovered a state law that called for a member of the press to be on the Election Observer team. That spot on the team had previously been filled by a retired journalist who recently had passed away. Eger and Clark pointed out the lack of a press observer to elections officials, including Konishi, whose domain includes the County Elections Office.

I call a staff member at the County Elections Commission, and get some details about my duties. At each precinct, there will be poll watchers assigned by each registered party (to the extent that the parties can find volunteers to fill those positions) to ensure no irregularities during voting. But I’ll be assigned to a team the Counting Center in Hilo.

This year, the county is actually using two different voting machine systems. A company called Election Systems and Software (ES&S) manufactured the two machines that will handle the paper ballots: the M-100, into which voters feed the ballots after they leave the polling booths; and the M550, which tallies the ballots again, once they arrive at the County Building in Hilo. The other system is a new, totally electronic one, mandated by a new federal law that was passed in the wake of the 2000 election controversies. Each precinct is supposed to have at least one or two of these new electronic voting booths, in order to make voting easier for those with physical impairments.

But controversy already swirls around the new system. The Journal has been getting e-mails and letters decrying the fact that the E-slates don’t leave a “paper trail” in the form of ballots, perhaps making the electronic results more vulnerable to hacking and election fraud. In March of this year, Mother Jones magazine reported that Walden “Wally” O’Dell, CEO of Diebold, which manufactures some of the new electronic voting machines, was a heavy financial backer of President Bush’s campaign and had written a letter pledging pledging to help “Ohio deliver its electoral votes to the President." and (See the Viewpoint on p. for more on this controversy.) The Mother Jones article noted that ES&L also had strong ties to the Republican Party: former company chair Chuck Hagel won election to the Senate in an upset race counted mostly by ES&S machines.

The Hawai’i State Elections Office has chosen an all-electronic voting system manufactured not by Diebold, but by a rival manufacturer, Hart Intercivic. Clark is watching the Hart machines like the proverbial hawk. During the primary, he discovered that the machines had electronically issued several Green Party ballots, despite the fact that there were no Green Party races to vote on in Hawai’i County.

Eger and Clark have also been busy in other areas. They’ve noted that two election-day teams--the “Manual Audit” and “Poll Book Audit” teams had no observers during the primary. They’ve been going through the voter registration books in a couple of precincts, and found 90 cases where, they claim, the same street addresses appeared in two different districts--an indication that voters may have been placed in the wrong precincts.

The two activists’ prodding may be helping to fix some of those gaps--such as getting an official Election Observer from the Press.

Part of my job, I’m told by a county elections official, will be to help test the voting machines before they’re used. To do so, I will need to fill out a set of dummy test ballots, which I can pick up at the County Elections Office on October 4 On the evening of October 11, the observer team will test the ES&S M-100 machines and Hart electronic eSlates. On October 30, we’ll be testing the ES&S M550 machines.

On October 4, I drive into Hilo from Volcano to pick up my sample ballot packet. But when I get to the Elections office, the packet isn’t there. I’m told to go home, and they’ll contact me when they arrive.

A day or two later, I get a call from the County Elections office, telling me that the State Elections Office will be sending me the packet directly, via certified mail.

Monday, October 11: The packet of ballots still hasn’t arrived. I call the County Elections office. An apologetic official there tells me I can use the packet of another observer who has dropped out. She asks me to come in an hour early, so I can fill the ballots out.

But as oft happens with journalists, fate intervenes, in the form of two telephone calls as I’m heading out the door. When I finally arrive at the Elections Office, my hour lead has shrunk to less than ten minutes.

My packet is waiting for me. I’m sent upstairs to the Counting Center, a vacant office area that formerlyhad housed the Department of Public Works before its mover to Aupuni Center. About half of the cavernous space is now filled with the black chassis of M-100s. Several tables are filled with suitcase-like eSlate terminals stand nearby.

Sue Irvine, the leader of the observer team, introduces me around. A state elections official explains to me about the sample ballots. “What you want is to create an easily recognizable pattern,” he says. “For instance, I can mark the first candidate in the first race, then the second in the second, then the first in the third, then the second in the forth....”

I need to create two “decks” or sets of sample ballots. I have up to 50 blank ballots to use in the endeavor, but the official notes that if I choose to mark all 50, I’m going to be here all night. I decide to make two decks of ten, and to use a slightly more complicated pattern: the first candidate in the first race, the second in the second, the third in the third, the fourth in the fourth, then back to the first candidate in the fifth. But many races only have two candidates. I do the one-two-three-four pattern in the multiple-candidate races, and alternate between one and two on the two-party races.

Meanwhile, the veteran Poll Watchers are already at work, feeding their test decks into the M-100s. I start “voting” frantically. I spoil a couple of ballots, grab a couple more....

Finally, I’m done. I’m lead to the first set of M-100s that I’m supposed to check, and shown how operate it. First, the machine must run a “zero count” printout to prove that there are no votes already sitting in the memory. Then I feed my test ballots through. The commands are punched into the machine to “close poll” then print the results. A cash-register-like tape spits out of a slot.

The pattern on the tape looks nothing like I’d expected it to look, though there does seem to be a pattern.

Now I have to sit down with my test ballots, to make sure that each candidate got as many votes as my 10 fictitious voters gave him.

Good lord, I thought. They’ve entrusted the tally of democracy to an English major.

But the numbers check out. Now I go to the next machine, “vote” the same test deck, and compare the printout to my first printout. They match, except for the one vote that I changed. The same for the next two machines. Then I switch decks for the next four.

My second deck, I discover, is even further off my expectations. I’d done it in even more of a hurry, and left a couple of races blank, and somehow only done nine ballots. That’s all right, as it turns out: it’s just a different pattern.

Next I’m led to the eSlates. Unlike the notorious Diebold machines, this is not a touch-screen system--as I discover, when I touch the numbers on the screen and nothing happens. There’s a little dial that the voter must turn, to move a highlight bar up and down the screen. An election official, or maybe it’s a Hart employee--I’m a little vague on everybody’s affiliations--shows me how to command the machine get a zero count and open the poll. Then, reading the results off one of my test decks, I twirl the dial to highlight the same candidates and each constitutional choice, then press a button to enter it, duplicating the results on the test ballot. This way, there’s at least a paper ballot to compare to the machine’s results in this pre-test, even though there won’t be any such ballots when the real thing happens.

The eSlate is actually a pretty nifty piece of machinery. The screens are easy to read, and there’s an audio headset for the visually impaired. And the eSlate does have one big advantage: it’s easier to fix a mistake or change my mind. After I’ve electronically “marked” the last constitutional issue, a screen shows up with a summary of my votes. If I don’t like one of my choices, I can backtrack and change it. Then, when I’m satisfied, I hit a big button marked “VOTE,” and my ballot officially enters the machine’s memory bank.

After the machines have been tested, each is put in a closed box with a wire-and-plastic seal that can only be broken by cutting the wire. They’ll stay sealed until Election Day. Each seal is individually numbered, and the number is logged when it is used. When the boxes arrive at the polling place, the seal numbers will be checked again, to make sure that they haven’t been tampered with.

Our test ballots and printouts are also sealed up in a big rectangular steel box known a “can.” (The metal ballot boxes in which the votes will be transported, officially referred to as Voted Ballot Containers or VBCs, are also called “cans.”)

We’re done for the night.

Saturday, October 30: the last minute phone-callers have struck again. I arrive 5 minutes late, and find a briefing session already underway. I take a seat and start taking notes--partly for the future story, but partly because I’m still the new kid on the block, trying to figure out what is going on.

Milo Clark is commenting on the lack of poll-watchers at the walk-in absentee balloting stations.

“We cast a heavy light on the mail-in ballots,” he comments. “We don’t cast any light on the walk-ins.”

The discussion moves on to the issue of the Hart machines. Clark presses a Hart representative about how the the machines’ accuracy can be audited. The company man explains that while the machines can’t produce identical copies of the paper ballots, they can produce “cast vote records.” Clark and he get into a discussion of the difference between the two, and how they can be reconciled.

The discussion is generally constructive--people bringing up minor glitches and talking about how to solve them.

After the briefing session, we adjourn to a back room, where two M-550s reside: high-speed paper-ballot scanners that will be used to read the “AB-mails,” or mail-in absentee ballots. The seal is broken on the Poll Watcher’s “can,” and we all bring in our test decks in to be read. The machines whiz through each deck at the rate of 550 ballots per minute, and spit out more readout tapes for us to compare to our originals. Again, my results match.

Afterwards, the machine’s memory chips are to be pulled and presented to Sue Irvine, who will seal them in the Observers’ Can until Election Day.

Election Day. I arrive on time, at 8 a.m., but the big room is already a hive of activity. The M-100 and eSlate machines are gone, carted off to the polls. Most of the tables that had held the eSlates have been pulled together to form a single space, piled high with thousands of AB Mail envelopes.

The election teams gather for a short briefing.

“Everybody knows that their cell phones should be off, or in the car, or at home?” we’re asked.

For the first time, the true complexity of a modern election begins to become apparent. An election isn’t just a simple counting of votes. It’s a highly complex operation, with specialized teams, security measures, and built-in redundancies, almost all designed to ensure one thing: public confidence that the vote was as accurate and unbiased as possible.

The Counting Center alone has 13 different teams, from the Receiving Team, which receives and inventories the “cans” of marked ballots as they arrive from the precincts, to the Results Distribution Team, whose job is to “prepare and disseminate election results to the media and general public.

Each Observer is assigned to watch one of the other teams. My assignment is the Poll Book Audit Team, whose duties don’t begin until fairly late in the process, so I have some time to check in on most of the other teams.

Activity shifts from one part of the room to other parts as the day passes. At the beginning of the day, most the action centers around the absentee ballots. A dozen or so women, mostly middle aged or better, has one of the more thankless tasks: opening that enormous pile of envelopes. But they seem amazingly dextrous at it, slicing hundreds of envelopes with nary a finger opened in the process, while engaging in cheery conversation, as if this were an old-fashioned quilting bee.

“Most of these ladies have been at it for years,” a veteran Watcher tells me.

In the M-550 room, the high-speed vote counter is not operating so fast this morning. The chief culprit seems to be not the machine itself, but the mail-in envelopes, which are slightly smaller than they should have been. Many ballots have been folded not in three, but in five or six to fit the envelope. As a result, the machine has a problem feeding some ballots, which then have to be extracted and fed through again; if that doesn’t work, they’re sent to the Duplication Team, which hand-fills-out exact duplicate ballots.

I’m briefly pressed into service for another duty: if a ballot unreadable because something has been over-voted, or a mark is too faint for the machine to read, it has to be hand-examined by a team of three observers, to determine if it is a “true overvote” or just a reading problem. Some of the problem ballots turn out to be erasures, with just enough ink left to fool the machine; in others, the voter tried to change his or her mind by ‘X’ing out a previous choice. If all three observers decide that the ballot is not a true over-vote, then it goes to the Duplication team.

Therein lies the only controversy in which I become involved today. We’re somewhat thrown by the fact that two ballots in a row have been both X’d out and initialed--with the same initials on each ballot. I find that coincidence a little puzzling. So does Irvine, who stops in to check on us at this moment. She says to call it an overvote, because she isn’t certain about the voter’s intent --which means that the voter is disenfranchised in that particular race.

I reluctantly agree that I’m a little spooked by the initials. But one of the other observers then says that we must count all other X’d out votes as over-votes as well, to which I don’t agree. Irvine steps in again, and changes the team, bringing in two official representatives from the Republicans and Democrats.

Meanwhile, the Duplication Team, whose work is sporadic, has started a jigsaw puzzle at their table. As the day grinds on, the absentee ballots get finished and the afternoon wait for the precinct ballots begins, members from other teams join in on the jigsaw construction. The puzzle, which shows an American Eagle, is polished off fairly quickly. But a second puzzle, depicting hundreds of jelly beans, proves to be a “buggah.”

Then the precinct “cans” start to come in, and activity erupts in whole new sections of the room. The Hart team sits at a bank of three Dell laptops loaded with proprietary software programs that will be used to download the data disks from the voting machines, tally them and send them to Honolulu. The Hart employees are somewhat bedeviled by Milo Clark, who wants compatible results from a precinct so that he can get to work with the Manual Audit Team--which is charged with manually counting ballots from sample precincts and checking them with the electronic tally from both machine systems.

I’m finally called to duty with the Poll Book Audit Team, whose job is count the names of voters that have been checked off in the poll books as they came to vote, and compare those numbers with the number of votes recorded in the two machine systems.

The team consists, I’m told, mostly of people with book-keeping and accounting experience. They count the names on each page, then enter the totals in old-fashioned calculators. If totals don’t match up by more than five or six names, they count again. I check on their checking by counting names after they’ve counted.

At Clark’s suggestion, I also examine the Poll Logs, which are kept by each Precinct Captain to record any problems that arise. And there are problems, of all the sorts that happen when mostly-amateur human beings engage in anything this complex. One polling place opens late, for instance, because of confusion over who has the keys to the building. In another, an over-zealous Democratic Poll Watcher has to be evicted because he can’t resist talking to the voters.

In only one precinct do I find an excessive number of incidents involving the eSlates--mostly voters starting to vote on the machines, but then giving up because of the wait. I wonder if it’s not because some poll worker has been over-enthusiastic about pushing people to try out the new machines.

Most of the accounts of problems include the note: “Called Election Center.” The Election Center is downstairs, where a bank of elections officials trouble-shoots the entire operation.

As the evening wears on, the results start to come in. Ironically, at the Counting Center, we only see the posted results for Hawai’i County. For a while, we have a small television to watch the statewide and national results, but as the evening wears on, somebody takes the TV home.

As midnight approaches and operations wind down, more attention focuses on the infernal crossword puzzle. Irvine vows not to go home until the last piece is in place. Finally, a bit of purple gumdrop finds a home. Election night is over.

Afterword. Milo Clark still isn’t satisfied. He thinks the eSlate system has a couple of major weaknesses.

“Can these machines be manipulated? Yes. There's two ways,” he believes. One is the early programming stage:“Whoever does the programming on the Hart MBB cartridges--there's where somebody needs to check and double checked and triple checked.” The other is the Hart system’s use of off-the-shelf Dell computers and Microsoft software--both of which, he maintains, are easily hacked--to transmit the results to Honolulu.

But even he admits that he saw no signs of deliberate tampering in Hawai’i’s results--only flaws that needed to be fixed.

Al Konishi is more optimistic. “When you look at all that goes on here--how could anything this complex be manipulated?” he says.

And he has a point, to some extent. The Election System isn’t just a system; it’s an evolutionary process. Since elections were first invented, people have found ingenious ways to cheat. But the cheaters have to find their way past thousands of other people with an interest in keeping the elections honest: people like Clark, who point out the problems, and others who invent new ways to check on the system--poll book audits, manual audits, ballot machine checks--and thousands more who volunteer to make them work. Democracy continues to work, so long as there are those thousands of poll volunteers working the system, staying until the last jelly bean falls into place.