“Election procedures are set in the statute.
You don’t have to be a person who got a PhD from Harvard, that’s for sure.”
–Mayor Menino, November 10th, 2006, on poll worker qualifications.
This past Tuesday, I was the election warden, aka the Chief, for a Boston precinct. I didn’t announce it ahead of time, because I wanted the realistic experience of an average warden, in an average precinct. I put my knowledge of computer science and crypto aside, went to the mandated 2 hours of new-warden training, read the manual over and over and over again over the weekend, and managed the precinct as best and as cleanly as I could. It was, in a word, a fantastic experience.
The Process, Summarized
In Boston, the election process runs as follows in theory:
- Alice, the voter, arrives at the polls, and is asked for her current address, last and first name.
- The clerk finds Alice on the check-in list. If she is properly registered with no special circumstances, she is given a paper ballot and a privacy sleeve. The clerk reads Alice’s name and address out loud so that the check-out desk and any observers can hear.
- Alice proceeds to an isolation booth, where she makes her choices by filling in the appropriate bubbles on the paper. Write-ins are handled by filling in a “write-in” bubble and then writing in the name in the appropriate space.
- Alice places her ballot inside the privacy sleeve and leaves the isolation booth.
- Alice proceeds to the check-out desk, where a police officer checks her off the check-out list, and a poll worker directs her to the optical-scanner voting machine.
- Alice extracts her ballot from the privacy sleeve and scans it into the machine. The ballot may be spit out if there is an undervote/overvote situation, and the poll worker may, with the voter’s consent, override the rejection by pressing a button upon rescanning.
The Warden’s Job
As warden, my job was to manage the overall process at my polling location, oversee poll workers and assign them various tasks, and make judgment calls anytime the situation called for it. During the day, this involved, in large part, handling any complication so that poll workers could handle the normal flow of voters smoothly. As expected, there were plenty of complications.
The first thing you learn as a poll worker is that there is no way to follow the process as perfectly as the theory indicates. Some voters don’t use the privacy sleeve. Sometimes the check-out poll worker is away for a few minutes, and the police officer ends up helping the voter with her ballot (we try to not let that happen so that voters don’t get intimidated, but….). Sometimes the check-in poll worker helps the voter find her address on the list when a clerical error gets in the way, or sometimes the poll worker completes the voter’s sentence at identification time:
Alice: 123 Spring Street.
Poll Worker: Alice Jones?
With that in mind, here’s what I observed.
Opening of the Polls
I arrived at the polling location at 5:45am. All poll workers were there on time by 6am (except for one who had arranged a starting time of 8am.) According to our instructions, we were to immediately prepare the polls to let voters begin casting votes at 7am sharp.
The secure materials—blank ballots, voting machine with sealed memory card, sealed absentee ballots, event log, precinct cell phone—are supposed to be delivered by the police officer by 6:15am. Our police officer arrived at 6:05am, but did not have the voting machine with him. He was not aware that he was supposed to have it, as it was his first time helping run an election.
The non-secure material—informational signs, registration forms, pens, voting booths, the large plastic (empty) ballot box, residence affirmation forms, provisional ballot envelopes and affirmations, etc…— are supposed to be pre-delivered to the location the night before. We could not find any of it.
I called the central office to let them know about the missing materials and the missing voting machine. They informed me that “someone was on their way.”
We received all the missing materials at 6:47am (13 minutes before the mandated poll opening time): it turns out they were all locked in the janitor closet next door, but the janitor was not around to unlock them. An official from the Boston Election Office figured this out, found the janitor, and delivered the missing materials, including the voting machine. In the ensuing rush, I was not able to determine if the voting machine was locked in the janitor closet along with the non-secure material, or if it was delivered specially by the Boston Election Office.
We rushed to put up all legal signs, ones at eye level for people standing up, ones at eye level for people in wheelchairs, ones in English, ones in Spanish, ones in Chinese. We cleared the handicapped access path (36 inches wide), propped open doors, and put up signs to indicate the meandering path through the building basement to lead voters to our location.
In view of other poll workers, I verified that the ballot box was empty, turned on the voting machine, checked and signed the zero-count printout (a long register-tape printout with all races and candidate names total, all indicating zero.) I asked two workers to set up the check-in and check-out lists, locked the ballot-box/optical scanner in place, and, before I had time to give poll workers my detailed instructions, a voter showed up. It was 7:02am. We were late. I declared the polls open and chose to finish up the remaining setup in parallel (as suggested by the manual.)
Fortunately for us, there was no 7am rush.
As I secured the critical voting materials—extra ballots, etc.—, the police officer asked for the ballot-box and voting-machine keys for safekeeping, as per his printed instructions. My instructions conflicted, telling me to store the keys in the silver election box. I chose to go along with my instructions rather than the police officer’s. In the end, this was a fortuitous decision: the police officer was replaced at 4pm, and it’s fairly likely we would have forgotten the key in his pocket, leaving us in serious trouble if we’d had to deal with any voting-machine issue, e.g. a jammed ballot.
The Two-Page (or is it Four-Page?) Ballot
In Boston, the paper ballot this November was 2 sheets long, each double-sided. Thus, four pages. I expected this to be extremely confusing to voters, in particular because no one knew, and the local news never mentioned it. That said, it turned out to be not quite as bad as I feared: the second sheet was ballot questions only, which many voters didn’t care to read anyways. The casting process for two sheets, however, was quite messy, in large part due to the privacy sleeve. (More on this in the “Optical Scan Voting Machine” section below.)
The two-sheet ballot certainly complicated our job as poll workers. Early on, we mistakenly gave 3 sheets to one voter (which we caught at scanning time), and thus chose to subsequently check ballot sheet pairs twice, once ahead of time and once at ballot handout time. Ballot spoiling became interesting: voters who make a mistake generally need to replace only one of the two sheets they have. However, to keep the ballot number count fairly sane, we were told to always spoil ballot sheets in pairs. Thus, the process for spoiling involved getting a complete new ballot, giving the voter the sheet they requested, and spoiling their sheet plus the new blank version of the other sheet.
As the voters began to show up, we started to see special registration cases: some entries on the voting list were marked with a notice requiring an ID check, a special residence affirmation form, etc… (the details are a bit tedious.) In every such case we encountered, we were able to resolve the situation according to the process defined in our manual, allowing the voter to cast a normal, unchallenged ballot. Most voters were quite grateful that, though they had not kept their registration up-to-date, we were able to accept their votes.
Things got a bit ugly, though, if the voter was not listed at all. We encountered this situation quite a few times (approximately 50 people—and our precinct only had 215 votes total.) In these cases, our only option was to call the central office to figure out the situation. At one point, we had to stay on the line with the central office for about 15 minutes straight as new unlisted voters kept coming in and we just kept asking the central office for their information. During rush hours (8am, lunch, 4pm, and 7pm), we were often disconnected while on hold waiting for a central office representative. However, through diligent redialing, the process was fast enough that no voter fled before we had found some resolution.
A couple of times, the central office notified us that the voter was indeed listed with our precinct, but at a different address, usually because the voters were college students who forgot they had changed dorms since the last election.
A number of times, the central office notified us that the voter was listed at a different precinct. In this case, we were strongly encouraged to tell the voter to go to their official precinct and vote there. Near the end of the voting day, we raced to call the central office in case such a situation presented itself. Except for a couple of last minute cases, voters seemed to have enough time to travel to their official precinct.
In other cases, the central office notified us that the voter was simply not registered, usually because, after moving and not re-registering for a while, they were dropped from the rolls. My observation was that, typically, voters who had moved twice without re-registering were unlisted. In this case, we encouraged them to register them for the next election, but were unable to let them cast a normal ballot.
However, in all cases where the voter was not listed with us, a choice of last-resort presented itself: the provisional ballot.
The Provisional Ballot allows any voter to assert his or her right to vote and cast a ballot. The ballot is not counted immediately—in our case it isn’t scanned by the voting machine. Instead, the ballot is sealed in its own individual envelope and set aside. The voter’s eligibility is eventually checked by the central office, and the ballot can thus be counted after further investigation.
So what’s wrong with that? Well, in most cases, we know ahead of time that the ballot won’t be counted. If a voter votes provisional and is not on the list, then there’s almost no chance his vote will be counted—the central office would have to have made a registration mistake which they can detect and correct, which is quite rare. The other case is much worse: if a voter from one precinct votes provisional in a different precinct, then the vote will definitely not count. So, for convenience, a voter may opt to vote provisional instead of going to his official precinct, thereby ensuring that his vote won’t count.
Yet poll workers are not well trained in this. They think voting provisional is a good last option. I had to ask the question specifically in training, in particular regarding voters who show up at the wrong precinct: I was surprised to find out that such provisional ballots never count. So voters who vote provisional are under the impression that their vote has a good chance of counting, but really it’s just a way to make voters feel better. As one voter told me at 7:55pm (5 minutes before close of election) because she couldn’t make it to her official precinct: “I want to at least feel like I’ve voted.” “Yes,” I told her, “that’s exactly what the provisional ballot is for.” The irony was a bit much for me, but she seemed happy with that answer.
The Optical-Scan Voting Machine
The optical scanning machine worked reasonably well throughout the day, though I was surprised with the number of times that a ballot was rejected only to be accepted upon rescanning. The security freak in me kept wondering how well it was reading ballots overall if it rejected so many on the first try.
I was fairly pleased with the undervote/overvote functionality. When a blank ballot is inserted, it is rejected as “blank vote.” When a ballot is inserted with too many answers for a given race, it is rejected as “overvote” with details of the race where the overvote happened. The voter is then offered the choice to cast the ballot as is or to go fix their ballot (in the case of an overvote, this requires a ballot spoil.) If the voter wishes to cast the ballot as is, a poll worker has to push a specific override button (using a pencil through a single hole in the front panel) while rescanning the ballot.
On the operational end, by taking the time to tell the voter that it was okay to re-do their vote, we salvaged most of the overvotes and some of the undervotes. In some cases, though, voters simply didn’t care about the second ballot sheet, which contained only Questions 2 and 3 and no candidate races.
One thing that bothered me a bit was the lack of information on undervotes. Given that, even with strict instructions and posted information sheets, voters often used checkmarks that didn’t fill the oval, it was unclear how many of the races registered during a successful ballot scan, other than “at least one.” It would be nice to know if 9 out of 10 races registered as blank because of poorly filled ovals.
Overall, I was fairly pleased with the physical security setup of the machine. With the scanner in full view of all election officials, and with no physical ports or buttons accessible to the voter once the machine was locked in place, it was quite easy to ensure that voters only scanned in their two ballot sheets and nothing else. In fact, the time we mistakenly gave a voter three ballot sheets, we caught it at scanning time. (We’re fairly certain this didn’t happen again because we counted the number of sheets of each kind, and the numbers matched.)
A Residual Vote, Courtesy of Yours Truly
At around 10am, a ballot jammed in the machine. I asked poll workers to observe as I opened the front panel, pulled the scanner out a few inches from its plastic ballot box/stand, and proceeded to remove the jammed ballot and drop it in the ballot box. That was a mistake: I had misread the diagnostic screen, which was instructing me to rescan the ballot (albeit in a cryptic way: “returned ballot jammed in printer.”)
It was too late to do anything about this at that point, and I had to really repress my desire to come up with wacky ways to fix the problem: I imagined opening up the ballot box, retrieving the top ballot, and scanning it, but there was no guarantee that this top ballot would really be the last one I dropped. And so we had to give up on this one sheet, or half-ballot. Interestingly, this made our machine count an odd number for the rest of the day, instead of the expected even number given the two-sheet ballot. Later in the day, one astute teenager, who was assisting his father, noticed the odd count and questioned us about it.
In theory, with the paper ballots available, this half-ballot wasn’t really lost: it could have been rescanned at any time. But, in practice, because counts and even recounts hardly ever resort to rescanning the paper, the half-ballot was likely effectively lost. We logged as much information as we could in the log book so that a recount might indeed be justified in rescanning the ballots. But, in the end, we closed up the day with 214.5 scanned ballots, instead of the expected 215 according to our check-in list.
That’s a half of a residual vote (a vote that is “lost”), courtesy of the Warden.
Closing of the Polls
At 8pm, I declared the polls closed. The police officer secured the entrance, while one last voter who was already in line finished casting his ballot. Two observers, one from a candidate’s campaign and the other from a press organization, had arrived and were anxiously awaiting the results. The last voter asked if he could stay and watch. “Sure.” I warned everyone that this was my first time closing the polls and, thus, it would be slow and overly methodical.
I had one poll worker read out the exact check-list to me while another poll worker verified that the steps were being read correctly. Scan any unscanned ballots (we had none). Insert the Ender Card into the scanner, watch it print the totals and announce them in a clear and loud voice (that took a good 5 minutes with the number of races.) Print out 3 copies of the totals, post one, leave one on the scanner, and deposit one in the results envelope. Reconcile the check-in and check-out lists. Count the number of spoiled ballots. Count the number of provisional ballots.
Then, we had to deal with the actual ballots. In full view of other poll workers, I opened the ballot box, and took out the non-write-in ballots (write-ins gets separated automatically into their own bin). Four poll workers quickly scanned through them to make sure they did not contain write-ins (sometimes they can fall into the wrong bin.) They didn’t. Then I assigned three poll workers to counting write-in votes. This took a good 15 minutes, and with only 13 write-in ballots: it’s tough to count write-ins.
I then assigned my remaining poll workers to count the unscanned, unused ballots, while I sealed the cast, scanned ballots into official envelopes and signed the envelopes. In the end, we added up all the numbers on the official record: apart from my one-half residual vote, the count was correct. The check-in list was also correct, though the check-out list was off by one (one forgotten checkmark.)
We put everything away in the designated boxes, in particular secure materials in the silver box. We handed the secure materials, the scanner, and the results envelope to the police officer, who immediately took them to City Hall. We left the precinct-count printout tape on the wall for all to see, and called it a night. It was 9:15pm, and we were all exhausted.
Write-Ins, Spaghetti Westerns, and Hand Counting
During the write-in count, one poll worker asked if I could discern one of the written-in names, as she could not. Maybe because I was pre-disposed to write-in jokes, I immediately saw that the name was “Clint Eastwood.” That caused a good laugh. Combined with the one write-in that said “none of the above” for a number of races, this raised doubts about the value of the write-in. Of course, there is a significant reason for write-ins: to let anyone run even if weird procedures leave that person off the ballot list. But the complications are significant, and we shouldn’t under-estimate them.
The other point that was abundantly clear from the write-in counting process is just how hard it is it to hand count. It takes forever. It’s confusing. Mistakes are made and have to be triple-checked. Why? Because there are 12 races on the ballot! Reading checkmarks off a page is prone to significant error. Hand counting of ballots in any significant number is completely impractical.
Experience and Force of Habit
A few times during the day, poll workers who had worked at this location told me how things were done the year before. Most of the time, this was extremely helpful: they reminded me about details of the process that were missing from training. Sometimes, this was unhelpful: some poll workers remembered the room set up a certain way in the past, others a different way, and they argued about which way was better for a little while. I made the call, and in the process made one poll worker fairly unhappy.
Interestingly, at 3pm, I had the chance to see another precinct’s operations: my own voting precinct, which I visited during a 30 minute break. Their process was quite different: check-out happened after ballot scanning which almost certainly led them to highly disparate check-in and check-out lists, as voters tend to race out after having scanned in their ballot. Since the check-in, check-out, scan process is clearly explained in the manual, I suppose someone at that precinct had, by force of habit, decided that things were always done this way, and thus should continue to be done this way.
So experience is helpful, but force of habit can be problematic. Wardens should listen to experienced poll workers, but should also come in and set expectations early in the day on controversial issues, so that poll workers can adjust.
- Re-Train Poll Workers on Provisional Ballot
This is a big issue. We should strive to make sure that voters know that provisional ballots, in certain cases, are truly useless. Poll workers should not think that provisional is “better than nothing.” In many well-defined cases, it is exactly nothing, and sometimes it is a placebo that prevents the voter from casting a vote that actually counts!
- Establish University Student Registration Systems
University students are significantly disenfranchised by the voter registration system: they change dorms and don’t think about how that affects their voter registration, or they register in their home state and don’t realize that they can’t vote in their school state. Universities should help by checking their students’ registration status: after all, universities already know their students’ secret info (name, birthdate, address). They could easily monitor their voter registration status, and notify them when they go inactive.
- Provide Electronic Poll Books at High-Traffic Locations
High-traffic locations should have one electronic poll book so they can direct voters to the right location without having to call the central office. Alternatively, a simple computer that can access a web site with the same information given the voter’s name and birthdate (effectively the same interface that Boston already offers on its web site for voters to check their precinct location.)
- Get Rid of the Privacy Sleeve: Make Ballot Single-Sided
The privacy sleeve was more confusing than helpful. When it came time to scan the ballot, it was so confusing that it generally resulted in the poll worker having to take the ballot cards out of the sleeve and hand them back to the voter for scanning. In other words, privacy sleeve = less privacy. Get rid of it. Make the ballot single-sided, even if it means more sheets. Instruct voters to turn the ballot over when scanning. Everyone understands that concept. In fact, everyone tried to do that upon scanning their ballot, only to realize they had an unanswered question on the back of the ballot card.
- More Explicit Undervote Notification/Rejection
Any undervote should cause a ballot rejection with a precise error message: “3 undervoted races.” This would cause more poll worker intervention to override the notification, but it would also make everyone far more confident that votes were correctly scanned.
Awe and Inspiration
I should mention that the whole day left me in awe of the effort made by hundreds, thousands of people to help their fellow citizens vote. The Boston Election phone bank answered dozens of calls from our little precinct alone. We had ballots in English, Spanish, and Chinese. We also had a translation poster with a dozen languages so that a voter could point to their language on the poster, at which point we would call the interpreter phone bank and let the voter speak with an official interpreter. One Vietnamese man was able to vote this way at our precinct, even though it took a solid 20 minutes.
This is hard stuff. Getting election equipment and materials out to hundreds of precincts, manning the phones, providing the translation, the signs, the handicapped access… it was beautiful and awe-inspiring. Anyone who’s ever criticized the system should be a poll worker for a full day. This is a job that deserves hands-on appreciation. I sincerely hope I can help out as a poll worker in every election in the future, wherever I may be.