Richard, Philadelphia County

After a frenzied few days of collecting notarized affidavits from voters, Philadelphia volunteers managed to collect and submit nearly 300 of them to Philadelphia County's board of elections, known as the City Commissioners, under 25 P.S. § 3154 (e), before the official computation of the vote was completed, on Wednesday. Three affidavits from people who voted on election day in a given division were required in order to "recount or recanvass" that division.

The board announced on Thursday that it would recount or recanvass the vote in 75 of Philadelphia's 1686 precincts, known locally as divisions, rejecting only a small fraction of the submitted petitions for reasons it has not so far disclosed. Despite arguments from Stein campaign attorney Ilann Maazel that the programming of the machines must be examined in order to make a recount meaningful, the commissioners did not consider or even (as I recollect) explicitly acknowledge this proposal.

The first stage of the process kicked off Friday at 1pm at the cavernous Wissahickon Ave Voting Machine Warehouse, where something approaching 4000 Danaher 1242 full-face DREs stand in silent ranks like, as one observer pointed out, the terra cotta warriors of Qin Shi Huang.

Two observers were permitted to represent each political party represented in the contested races, plus one per candidate. No credentials were requested, since there were no surplus observers to compete for limited spaces — a good thing, considering that neither the observers nor the board members had any idea what a credentialing procedure should look like.

We observers were divided into three groups, each to accompany one election worker to visit the machines corresponding to the 75 divisions in question. Each machine is used in the same division for every election, and is parked in a neat yellow rectangle marked with its number, next to its division-mate. A few divisions, those with exceptionally high registration and turnout rates, are allocated one or two additional machines.

Working from a list, the election worker led us to each machine, opened the top back panel, inserted an unmarked memory cartridge and enter a code to print a results tape from the machine's redundant internal memory areas, and put this paper tape in an envelope marked with the division number. To speed things up, the workers usually worked the two machines of a given division at one time, using two cartridges.

These cartridges were described as empty, needed only to activate the machine. The same cartridges were used in all machines. The election workers were not willing to open the lower panel, under which is a PROM module that we believe should be covered by tamper-evident tape. We were able to note the machine number and the division to which it belonged. The only onboard counter is an analog counter that shows only the lifetime total of ballots cast on that machine. There were a few paper jams but all were quickly resolved. Overall the machines (mostly purchased around 2005-06) seemed to be very well maintained. Although we had made up forms to be able to quickly copy results in such a situation, we were not able to use them because all the candidates were listed in code and (at that point) we didn't have the key to the code. I felt, and others seemed to share, an unanticipated degree of reluctance to insist on any procedures that might multiply the hours we were spending on this task -- especially as it seemed increasingly pointless. The tapes are quite long and the print is small and light, they would be difficult to photograph in any useful way. In effect we were able to verify that a tape was indeed printed from each appropriate machine and put in an envelope.

After about three hours all the tapes had been collected. In the next phase, election staff brought out memory cartridges that were marked with stickers, each showing a machine number and a division number, corresponding to the 75 divisions. There were about 170 of these (again, 2+ per division). They had 75 extra machines set up to use as printers. Into each machine they put a cartridge and entered a command to print a results tape from the cartridge. They got through the cartridges in 3 batches, taking about 90 minutes in all. It is printed right on the tape whether the tape was printed from a cartridge or from internal memory. These tapes were added to their rightful envelopes. The cartridges they were printing from are said to be (and I have no reason to doubt it) the ones that are removed from the machines when the polling places close, are brought to a number of regional input centers and are inserted into proprietary readers that are connected to computers feeding the data to a central database. This database is where election night numbers come from.

And here we get to the ostensible point of this exercise: to demonstrate that the data used to generate election-night numbers and later to populate the machine-count segment of the overall vote computation has not been altered. This was to be shown by the fact that it matches the data found in the internal memory of the machines. (As I write this, I recall that the upper panels of the machines to be re-canvassed either were already open when we arrived, or were closed but had broken plastic tie-seals hanging off, whereas other machines' upper doors were still sealed. At the time I imagined that they had to open the doors to start up the machines in advance, or make sure there was enough paper tape on the roll or something. But now I must admit it seems like something to wonder about.)

The final phase of Friday's activities resembled a traditional recount a little more, except, notably, for the absence of actual ballots and therefore much of a real purpose. Three groups each proceeded as follows, under the determined if mildly befuddled gaze of the observers. Election worker A read candidate codes off a tally sheet representing one machine after another, as election worker B read the corresponding number of votes from the paper tape, first from the machine's internal memory, then from the cartridge. Election worker A dutifully filled them in one by one in two columns. We were able to watch right over their shoulders as they read and copied the numbers. It was hardly miraculous that after about two hours, not a single discrepancy was found. No one in the room evinced the slightest surprise.

The next stage of the "recount" is to start on Sunday morning at 10am, at the board's 520 N Delaware Ave location. The plan is to hand-count the various categories of paper ballots found in the 75 divisions -- likely about 1000 ballots in all -- again in three groups. Time permitting, these tallies will be added to the machine tallies and compared against the results of the official computation that was completed on Wednesday.

A couple of observations:

The various forms we brought were not suited to the task, because we had no idea beforehand what this stage of the recount would look like and what kinds of data we would be able to collect and in what order. Even when presented with data-collection opportunities we did not necessarily know whether the data might prove to be significant.

The BoE staff drove the process, impeding our ability to understand the rationale. They invented the procedures and told us on the spot what they would do and how they would do it. It was very much a "recount" by them with us allowed to observe. They did not treat us rudely, but we were not partners in any sense.

What was done had to be done, but it also served as a distraction. What is really necessary is an examination of the software on the machines as well as an examination of how the machines are programmed -- specifically where does the programming (ballot definitions) come from and how does it get from there to the machines. Although our interactions with the election workers because more casual over the course of the day Friday, and they were willing to talk about every aspect of the "recount" or re-canvass, they went completely silent whenever we edged into that topic. The commissioners themselves consistently deny that there is any chance of hacking their system, pointing out at every opportunity that -- guess what's coming, yes you got it -- that the machines are never connected to the Internet.

The extreme transparency with which the board is willing to demonstrate the output of the machines is in sharp contrast to their refusal to say anything at all about the input to the machines.

In addition to the the question of the unsealed panels, which relates to the integrity of the output of the machines, some observers have raised the question of whether we will be able to see the sign-in books from election day or find another indication of the total number of voters with which to compare the total number of votes, specifically to make sure there are not excess votes. One election staff member said he thought that the sign-in books, while outside the scope of a recount or recanvass, were nonetheless a kind of public record that could probably be examined on request.