Yesterday, I went to Orange County to watch them conduct the one-percent manual recount. It was a valuable learning experience, but after awhile you realize you are watching groups of people count paper. Under California law, all local election officials are required to conduct a one-percent manual recount after every election. The recount requires that one-percent of all precincts in the jurisdiction be recounted. However, all races must be recounted, which often means that specific local races in additional precincts must be counted as well. The manual recount provision in effect now in California requires that, in jurisdictions using electronic voting, the recount be conducted using the voter-verified paper audit trail (VVPAT).

Because so few recounts have been conducted to present using VVPAT, observing the Orange County experience allows us to learn about the issues associated with conducting VVPAT recounts. In the Orange County one-percent recount, there were eight teams of four conducting the count and it was expected that it would take between four and five days to complete the count. (The turnout in the County was approximately 27 percent).

Pre-Recount Canvass and Ballot Accounting

The first step in the recount is the initial canvass and ballot accounting. In Orange County, there are two documents that are central to this process: the “Combined Roster-Index Ballot Statement” (CRIBS) and the “Printed Turnout, Unofficial” (PTU).

The key to the reconciliation process is to examine the number of signatures in the poll roster and compare this to the number of machine access codes voted, which is equivalent to the number of ballots issued. This access code data is on the CRIBS form and also on the PTU, which lists every precinct by number and the number of ballots cast by precinct. The formula for the accounting process at this point is:

Number of Signatures = Number of Codes Voted.

Recount Process

Once the signatures and ballots cast have been reconciled, the County did a random sample of both precincts and races. A computer program generates the random selection for the one-percent precinct count. Then, the races not covered by the one-percent random count are selected to ensure that all races are covered by the recount.

Although Orange County has done a one-percent recount before, this is the first time that they have counted the paper audit trails. The process for doing the recount was as follows:

1. Generate a list of precincts and races to be recounted.

2. Find the bag containing the sealed VVPAT for the polling place in question. (In an election with consolidated precincts, the polling place has to be located all VVPATs examined to find only the votes for the precinct in question, as discussed below).

3. Remove all VVPAT printers from the bag.

4. Assign the VVPAT for the precinct in question to a counting team.

5. Have the team take each VVPAT, unseal it, unscrew the back, and open it.

6. Advance the VVPAT printer tape until the end of the voted tape is located. Cut the tape.

7. Reseal empty container and place it back into the precinct bag. The “unvoted” tape is still kept with the bag in case any issues arise.

8. Unroll all of the ballots from the tape and cut each ballot. (This part of the process could potentially be automated, with the correct technology, but at present, cutting the VVPATs into individual ballots makes sense).

9. Separate the ballots by precincts if the polling place had consolidated precincts.

10. Separate the “ballot accepted” from the “ballot rejected.”

11. Place all “ballot rejected” ballots into a separate envelope with the precinct number. Seal this envelope.

12. In a primary election, sort the ballots by party.

13. Stack the ballots and then, for each race, call the candidate on the ballot. (In a primary election, you first select a party and then, for each race, call the candidate on the ballot). If the voter made no selection, you call the no selections for accounting purposes.

14. The person calling the ballot then hands the called ballot to the observer, who verifies the call.

15. The observer then sorts the called ballots into piles for each candidate (including a “no selection” pile).

16. Two individuals separately tally the called ballot.

17. Once all ballots for a race have been called, the team then compares the tally totals on the two tally sheets.

18. If these two numbers are the same, the team then compares the tally with the number of VVPAT ballots in the piles for each candidate.

19. If these two totals are the same, the team then verifies the tally against the “unofficial results,” which list the vote total recorded by the voting machines by precinct by candidate for early voting and Election Day voting. (Note: the one-percent count of VVPATs only verifies Election Day totals).

If, at the end of this process the totals do not match and a re-check by the team cannot determine the cause of the problem, the senior election staff take the VVPAT ballots, along with the roster and the electronic ballot storage card, and re-examine the precinct’s totals.

Team Dynamics

As was noted implicitly above, the counting of the VVPATs is done by teams of four:

• One person calls the candidate for the race being counted.
• One person checks that the correct name was counted and piles the ballots by candidate.
• Two people independently tally (using “tick” marks on a tally sheet) each time a candidates name is called.

Across the eight teams, there was variation in how effectively each team worked together. The two most effective teams had the following characteristics:

• The team members had worked together before. In one case, the workers represented the members of a precinct Election Day poll worker team that also works pre- and post-election. In the other case, the individuals have worked a post-election audit and manual recount previously and all like working together.

• Someone was in charge on the team. Although there is no official “leader” on each team, one individual needs to be someone to whom everyone will listen and knows how to handle various situations that arise.

• They work efficiently. The team members have to develop a rhythm to the work and move efficiently through the process. The euphemism “Slow and steady wins the race,” is apropos here.

• They knew what they were doing and why they were doing it. The effective teams had workers who understood exactly what they were doing, how to follow the process, and the underlying rationale for why the recount was being conducted.

Interestingly, the teams that were most ineffective were those that were comprised of multiple individuals who work as the lead person in a polling place. This might seem counter-intuitive; it would be reasonable to think that a team of lead polling place people would include the best people possible. However, the problem that arises is t
hat these teams have too many people who wanted to be in charge and not enough people who are willing to follow the lead. As a result, these teams often became bogged down in debates over minutiae about how to handle problems.