Allegheny County's Board of Elections does go to some lengths to prepare the Judges of Election (JoEs, for short), and also the individual pollworkers, but some of that could be done better, I think.
There is no real way to prepare for heavy turnout elections other than to just participate in them. I fear that we will not see this sort of turnout again for another four years, and that by four years from now, the vast majority of people who have this experience will have moved on to other things. Maybe even me. Anyway, I lived it, I'm thinking it, so I'd better document all this for next time, whenever and for whomever it might be.
Think of preparation for a heavy turnout election as preparation for performing CPR on a heart attack victim. You only get one chance to do it right. All the classroom training you can muster can only pale in comparison to hands-on experience.
Now that I've had a couple days to reflect on Tuesday's flow of events, I have tried to think about how I could have made that flow of people go any faster. I hate to think how many people disenfranchised themselves by either leaving the line or never joining it. Ultimately that was my responsibility, as I needed to detect and diagnose and resolve slowdowns in all those lines.
For example, I know that the R-Z line got short and stayed short early on because of an internal problem: The pollworker was skipping two major steps, neither writing down the names in the logs, nor numbering them in the books. By the time I noticed this, there was a backlog of about 130 names. I logged those names and caught up the numbering myself so that at least the numbering step could continue on by the pollworker. When I pressed additional people into service later, I had one just do that part of the bookwork. I knew it would do little to keep R-Z at speed, but it would save me hours of time at the end of the day.
One thing I learned from being part of R-Z was the speed to be gained by pre-fetching the voter cards. As voters approached the front of the line, for perhaps the first four people awaiting processing, I got their names and located their names in the can, turning their voter cards sideways. I could actually have made R-Z work faster still, except that that exacerbated my attempts to catch up the log work.
The problem with pre-fetching on R-Z is that it was not as easily transferable to A-D and E-K. First of all, those pollworkers were quite green, and just learning how to do the work at all. I rotated my L-Q pollworker through the other two lines, as she was incredibly fast, faster at the entire process solo than two people, maybe even three. By putting her on A-D for an hour, she drained that line. Then I put her on E-K for an hour and had her drain that one. Problem was, there was only one of her. That's not a training issue. Some people are just unbelievably fast.
The pre-fetching technique only would have worked well for R-Z, though, because it was the only sideways line, the only one that stayed close enough to the can that anyone could pre-fetch the cards for persons 2-4 in line. Maybe if I had had yet another table, so I could spread my pollworkers out far enough to make each line rather "J" shaped, that might have worked. In that case, I would really have needed four newbie pollworkers per line: One to do pre-fetching of cards, one to do pollbook logging, one to log voter names in one log book, one to log names in the other book. Voters would stretch out in a line as usual -- the stem of the "J" -- then first identify themselves to the pre-fetcher, who found the card in the can. Pollworker 2 has the roster, and still had to find the voter's name there to assign a sequence number, and number the tag, card and roster. Pollworker 2 handed the card to Pollworker 3, who together with 4, logged the name in the two log books. The four of them would form the base of the "J". Pre-processed voters would then queue in a second line for access to a booth. Or maybe have two pre-fetchers, each working two lines, seated between each of A-D and E-K, the other between L-Q and R-Z.
(Actually I don't think we were reliably numbering the tags, just the cards, and even if we were, we had the problem of not having unique sequence numbers on the tags. Each of lines A-D, E-K, L-Q and R-Z had simple integer identifiers. We were not prefixing them with "AD", "EK", "LQ" or "RZ", or some such. I honestly do not know if we were supposed to do that, but it wasn't getting done, in any event.)
Next was queueing for the booths themselves. Because of the mayhem at the front of the lines, people being processed from R-Z, L-Q and E-K had to work their way across those and the A-D lines to get to the line for the booths themselves. It did not help that sometimes this line was non-existent.
Despite that, we very frequently did have voters in all 13 booths. Oddly enough, though, the A-D line voters showed to be the most likely to lose their way to the booth queueing, which is strange, because they hardly needed to do more than shuffle sideways two or three feet. More than once, though, we found people from A-D wandering all over, with two-piece tear-off stub in hand, looking for where they would actually vote. I believe we found one two-piece stub on the floor, meaning we processed someone in the book, only for them to leave without actually voting. I doubt there were more than one or two of these all day. Maybe we had a half dozen from the A-D line who got lost on their way to the booths.
The booth queue sometimes did get some size to it. At its longest, it might have been 10 to 15. If I could have had all four lines processing people as fast as R-Z when it was at its skipping-two-steps fastest, maybe the 13 machines would not have been enough. I had three people putting people in booths (I called them "runners"), which seemed adequate, neither too much nor too few. I later discovered I did have a fourth PEB (the device needed to initialize a booth for each voter), and for a while could really have used it. Again, if my processing lines were faster and I had 15 booths, I could have used that fourth PEB, but then I would have needed another "runner" pollworker, and I did not have that luxury.
A big problem on Tuesday was that it took some policing to get people into the right polling place or right line. People came in the doors off of the east end of the building (nearest the merry-go-round), and seeing the lines, simply got in one. Some of these were people for Pittsburgh 4-14, which was on the west end of Posvar Hall (a.k.a. Forbes Quad). A good many did not realize that the L-Q and R-Z lines were often a lot shorter, and so stood for a half hour or more. This was difficult to monitor. Some volunteers dedicated to this job would have made things move a lot smoother.
Signage. I had two sets of large signs labeled A-D, E-K, L-Q and R-Z. One was for the desks, the other, I figure, was for identifying the end of the lines. The problem with this is that the end of the line varied in length by quite a bit. It would have been difficult to figure out where to put the sign, though we did have tripods/easels on which to mount them. Tape on the floor worked better, and for that I have the Election Protection people to thank, I think. They cut pieces of masking tape to actually spell out "A-D", etc. Very effective, I must say. Not real elegant, but I'm all for function over form, any day of the week.
A wider signage problem was what to do about external doors. Is there some way we could have put directional information on the outside doors, such that confusion would have been reduced? I don't know. Pitt's people were already sensitive to having any signage on the doors, as it was not legal to have any political signs on the doors.
The weather was not a factor today, but in November 2007, we had a windstorm that made it difficult to operate the voting site. Anything attached to a wall with tape, even inside, had a very short life. The wind that day was so strong, it was blowing one door open, making it hard to keep our paperwork in order. That day, 4-8 and 4-14 were cheek-to-jowl near the one entrance. For the two 2008 elections, 4-8 has taken a more central location near home plate, and are very likely to stay there for future setups.
Chairs. I had a staff of 12, or was supposed to; only 10 actually showed up. I deputized two later, and three more came in from the county, for a maximum of 15 at peak load around 4 or 5 p.m. At the beginning of the day, knowing we were going to be quite short on chairs, I had one of my staff drag a huge upholstered chair from an upstairs lounge. That was overkill, I realize; I was merely hoping he could find a couple with padded seats. We did make do with simple folding chairs, but it would have been nice to have something a bit more comfortable. After all, we were sitting there for 15 hours -- 13 processing voters, plus at least one more with setup and cleanup. Were it mine to decide, I would have the polls open even longer, 6a-9p, or even 6:30a-8:30p, but that's a discussion for a different day. Net effect, folding chairs are tough on the butt over that long a time.
I am glad that, before the election, I got in touch with Pitt's Facilities Maintenance people to go over the logistics of my setup. I actually did a better job before the April primary than November, but at least I had the man's name and pager number. Next time, I will have them set up 4-8's tables -- five, next time -- centered on home plate, not stuck down with 4-14's three-table setup, as I encountered when I showed up at 5:30 a.m.
Possibly the biggest unknown variable is in forecasting need. The Spring 2009 primary is not likely to be as big a draw as the 2008 presidential. We probably won't need 15 booths. We might not need a staff of 12. I just don't know. I do know that I need to write down as much now as I can recall and which is still rummaging around in my head while still fresh.
If the Spring 2009 primary looks like a sleeper, maybe I can do it with a single, J-shaped line, as described above, but putting all the cans together. One person pre-fetching from four cans, one person with the four roster books, and two persons logging the names. And one maybe two runners. And the Judge. The biggest problem there will be the re-filing of the cards in the respective cans, after logging. Normal staff is five; that's adequate for a sleeper election with a turnout in the couple-hundred range. That won't cut it for a couple-thousand turnout.
On the other end of the spectrum, thinking about November 2012, with space and manpower being unrestricted, I could have used 3 runners, 2 pre-fetchers, and 3 people apiece on 4 chunks of alphabet (roster, log book 1, log book 2), plus myself, for a total of 18 people. I would likely have needed at least three more than the 13 machines I did have. With that configuration, each voter would have been through the can-and-log processing in 30 to 45 seconds instead of 90 to 120. That would have kept the waiting lines to a minimum, and likewise self-disenfranchisement.