What happens when the printed ballot face doesn't match the electronic ballot definition?

Part 4 of 4. Complete 4-part series available here.
The Sequoia AVC Advantage is an old-technology direct-recording electronic voting machine. It doesn’t have a video display; the candidate names are printed on a large sheet of paper, and voters indicate their choices by pressing buttons that are underneath the paper. A “ballot definition” file in an electronic cartridge associates candidate names with the button positions.

Clearly, it had better be the case that the candidate names on the printed paper match the candidate names in the ballot-definition file in the cartridge! Otherwise, voters will press the button for (e.g.,) Cynthia Zirkle, but the computer will record a vote for Vivian Henry, as happened in a recent election in New Jersey.

How do we know that this is what happened? As I reported to the Court in Zirkle v. Henry, the AVC Advantage prints the names of candidates, and how many votes each received, on a Results Report printout on a roll of cash-register tape. The printout reads, in this case,

    I23   Cynthia Zirkle      10
    I24   Ernest Zirkle         9
    J23   Vivian Henry        34
    J24   Mark A. Henry      33

In this election, four candidates are running for two positions in a vote-for-any-two election. Here, J23 indicates that the button at column J, row 23 on the face of the AVC advantage received 34 votes. The problem was that the poster-size printed paper covering the buttons had the name Cynthia Zirkle printed at position J23. Vivian Henry’s name was printed at position I23. That is, there was a mismatch between the printed paper and the electronic ballot-definition file. Similarly, the positions of Ernest Zirkle and Mark Henry were swapped.

Rebecca Mercuri told me that until the mid 1990s, the AVC Advantage firmware did not print the row/column numbers at all, so that mismatches like this were harder to detect.

One might think that all is well–there’s a fail-safe mechanism that can catch mistakes (or deliberate fraud) where the paper doesn’t match the electronic file. But in this election, the fail-safe mechanism did not work well at all.

First, there are almost no candidates or pollwatchers out there who know enough to look out for this kind of mismatch. In the Zirkle v. Henry election, Cynthia and Ernest Zirkle couldn’t tell from the documents available to them that the positions were switched. They and their lawyer got 28 (or more) sworn affidavits from citizens who said they voted for the Zirkles, and on that basis they got a court to permit an investigation. In any election that involved significantly more than 43 voters, it’s impractical to get sworn affidavits from everyone who voted for you. This election took place all on one voting machine; in big-time elections one would need to double-check the face of the ballot against the Results Report printout in every single precinct. This is physically possible, but it isn’t easy and independent pollwatchers are not trained to do it. In Zirkle v. Henry this came to light because certain experts got involved, but one can’t count on that in general.

Second, even in this case, the Court was uncomfortable just swapping the votes and declaring the Zirkles to be the winners of the election. That is, both the Plaintiffs (lawyers and expert witness for the Zirkles) and the Defendants (lawyers for the State of New Jersey and the County of Cumberland) stated to the court that they believed that Cynthia and Ernest Zirkle got 34 and 33 votes, respectively. Defendants Vivian and Mark Henry, representing themselves, took the position that a new election should be held.

In his ruling, the Court (Judge David Krell) said,

Based on all of the above, it is clear that the election at issue was defective and must be voided by the Court. While I do believe I have the authority to certify the Plaintiffs as the winners, I do not feel that this is the ideal result in this matter. … Accordingly, I am ordering a new election to be conducted.

If there was ever a case in which these row-and-column numbers could clearly indicate who won an election, this was it. And yet a very reasonable judge is uncomfortable using this information to declare a winner, and instead orders a new election.

Ordering a new election is not at all unreasonable, but it is important to remember that a new election can have its own problems. Citizens who came out to vote the first time may not have the time or inclination to vote again, and if so their (previous) legitimate exercise of the franchise is being devalued. Or, some who did not bother to vote the first time may take advantage of the “do-over.”

It is instructive to consider what would have happened if a similar kind of error had happened with optical-scan voting. It’s certainly possible that the position of names on the op-scan paper ballot might not match the programming of the optical-scan ballot-counter. In this case, the results would come out reversed just as they did in Zirkle v. Henry. But the Court would have simply ordered a recount, by hand, of the original paper ballots. Those ballots would have clearly showed the true result. No experts, and no do-over election, would have been necessary at all.