Countering Objections to New York's BMD BillMay 20, 2022
Part of
the Voting and Elections web pages, http://homepage.cs.uiowa.edu/~dwjones/voting/
|
On April 28, 2022, Davidoff Hutcher & Citron, LLP issued a Memo in Opposition to S.309B (Myrie) / A.1115C (Paulin), on behalf of their client Election Systems and Software. S.309B / A.1115C is bill that was introduced in the New York state legislature earlier in 2022. As I had a hand in drafting this legislation, I feel it is appropriate for me to address the points raised in this memo.
Before addressing the memo, it is appropriate to briefly summarize the legislation. In New York, voters vote on paper ballots that are counted by machines commonly described as ballot scanners.
In order to meet the needs of voters with disabilities, the state has long provided ballot marking devices (BMDs), machines that provide an electronic interface to allow voters to use a touch screen, audio output or pushbuttons to formulate their vote which the machine then marks on the paper ballot for them.
This combination of technologies is used in many other states.
For a long time, the most widely used BMD in the New York was the Automark, a machine that marked on ballots identical to those used by voters who hand mark their ballots. See U.S. Patent 7,080,779, issued in 2006.
There are, however, several variations on BMDs that have, in recent years, been agressively sold by several voting system vendors.
First, there are BMDs that print a brief summary ballot. A summary ballot lists only the candidates the voter selected, omitting all others, so that ballots submitted by BMD users look completely different from ballots submitted by those who hand-mark their ballots. In a polling place where only a small fraction of the voters use the BMD, their ballots are easily distinguished from others, leading to a real reduction in those voter's right to a secret ballot. Therefore, the proposed legislation would eliminate the use of summary ballots.
Second, the BMDs that print summary ballots print them with both English text and a bar code or similar computer readable code. When the summary ballot is inserted in the ballot scanner, the scanner reads the bar code and ignores the English text. The English text is there so that the voter can verify that the ballot actually represents their selection, but voters cannot verify that the bar code corresponds to the English text. Therefore, the proposed legislation bans the use of such bar codes.
Finally, some of the newer BMDs on the market offer combined ballot marking and ballot tabulation, so that the same machine serves as a BMD and as ballot tabulating machine. The economic incentive to do so is obvious, but if the paper passes by the ballot marking mechanism on its way into the ballot tabulator, the voter has no way to be certain that the ballot has not been altered after the voter approved the ballot. Therefore, the proposed legislation bans machines that combine marking and tabulation.
In drafting this legislation, we had to be careful about the term scanner. Most BMDs incorporate scanners in their mechanisms for several reasons. If the BMD marks on regular pre-printed ballots, it needs to verify the orientation and ballot style of the ballot. After the BMD has printed a vote, it should verify what it printed to be sure that it is not, for example, out of ink. Finally, BMDs serving the needs of blind voters must be able to read back the votes they have marked to offer the voter some assurance that the ballot was correctly marked. Therefore, the legislation does not ban BMDs that incorporate scanners, but rather, bans the BMD from retaining any information about the ballots it marked after completing work on behalf of a voter.
In drafting this legislation, we also had to be careful about over-stating the ban on bar codes. Since the dawn of optical-scan ballot tabulators, ballots have been marked with what are essentially bar codes that code for things like ballot style, precinct number, and election ID. Our goal was to limit the use of bar codes for conveying voter selections while permitting these other uses.
The memo of opposition alleges that the proposed legislation "would intentionally ban the use of the company's ExpressVote XL," while it "would ... permit the use of the same type of technology provided by Dominion Voting Systems, an ES&S competitor."
The ban on bar coded summary ballots in the proposed legislation is not simply a response to ExpressVote XL. The ES&S ExpressVote BMD generates a bar-coded summary ballot following a pattern that dates back decades. The first proposal I am aware of for a BMD using such a ballot format was made by Julien Anno and Texas Instruments in U.S. Patent 5,189,288, issued in 1993. Sanford Morganstein and Populex Corporation received U.S. Patent 7,284,700 in 2007 for a variation on this idea, and Populex machines were certified to the U.S. Election Assistance Commission's Voting System Guidelines and successfully used in some U.S. jurisdictions.
I and others involved in elections had several extended discussions with Morganstein about the use of bar codes in the Populex system. In those discussions, we discussed all of the reasons that a voter might not trust voter verification of the English text on a summary ballot when it was the bar code that was being counted as a vote. The ban on bar codes in the proposed legislation owes as much to that discussion as it does to machines currently on the market.
The memo of opposition alleges that the proposed legislation would "nullify the voting system certification process that was carefully established to enact laws according to the Help America Voting Act (HAVA) ..."
Here, I speak as a former voting system examiner for the State of Iowa, a former member of the Technical Guidelines Development Committee that helps set the Voluntary Voting System Guidelines to which HAVA compliant voting systems are tested, and as one of the people who helped draft the legislation New York enacted in response to HAVA.
First, while the voting system certification process conducted under HAVA has a degree of rigor, those who drafted HAVA never intended the Voluntary Voting System Guidelines to be the last word. States have always been free to set additional requirements, and many do. There is no sense in which setting such requirements nullifies HAVA, nor does passing new requirements make any kind of mockery of previously enacted requirements. New York's requirement for full-face ballot display is far more stringent than anything HAVA has required. Colorado has banned bar-coded summary ballots, and many states have banned direct-recording electronic (DRE) voting machines even though, when HAVA was drafted, the authors seem to have thought DRE machines were going to dominate future elections.
The memo of opposition states that "All currently available voting voting machines ... use both machine-readable and human-readable ballot data points."
This objection is a bit disingenuous. The "machine-readable ballot data points" that are universal are things like ballot-style, election ID and precinct numbers coded on the ballot. Some ballots have fiducial marks, that is, alignment marks to help the scanner determine the size and orientation of the ballot. These are not points of contention here!
Furthermore, we are aware of occasional problems with these other data points. For example, in November 2008, some print runs of ES&S balots for Madison and Carroll Counties in Indiana could not be tabulated on ES&S scanners because of incorrectly printed coding on the ballots.
The big difference between incorrect printing of bar codes for ballot style, election ID or precinct, on the one hand, and misprinting of the bar code representing a voter's vote is that misprinting the former generally causes the ballot to be rejected, forcing manual inspection and immediately bringing the problem to light. Misprinting the coded information representing a voter's vote could easily lead to accepting the ballot and an incorrect vote count.
In the case of pre-printed codes on ballots, as occurred in the two Indiana counties cited, simple testing by the county before the election could have detected these problems before the election. The counties would have avoided embarrasment and delayed election results had they done this. In contrast, the testing required to detect errors in the bar codes used to record candidate selections on summary ballots are significantly more complex.
There is no point, here, in distinguishing between deliberate hacking of the ballot printing, on the one hand, and accidental misprinting. The cases cited in Indiana were honest mistakes, not attempted election hacking.
It is possible to imagine misprinted ballot style or precinct information causing a ballot to be incorrectly tabulated instead of being rejected, but the proposed legislation does not address that issue. I would recommend that carefully constructed misprints be included in any tests of ballot tabulators.
It is worth noting that the use of bar codes or similar non-human readable codes on ballots is unnecessary because optical character recognition (OCR) could be used to directly read the human readable content of the ballot. OCR is a relatively difficult problem when trying to read handwriting or text printed in arbitrary fonts, but text printed on a ballot (or on a summary ballot) is printed in a font that is completely controlled by the voting system. Fonts such as OCRA, developed in 1968, would work fairly well on ballots, and newer OCR fonts are even more legible.
The memo of opposition complains about the provision to "allow county boards to use voting machines that they have already purchased which violates the prohibitions of the proposed legislation."
This provision is a matter of fiscal responsibility. Forcing the immediate junking of existing equipment when tighter standards are enacted is extremely expensive. This is why, when new consumer safety standards are enacted, they generally apply only to new purchases. Only rarely do we require manufacturers to issue product recalls.
A reasonable timetable for a jurisdiction to purchase a new voting system involves approximately two years between the initial investigation of the systems currently on the market and the first deployment of the new system. This gives the adequate time for the jurisdiction to create a well crafted request for quotes, for vendors to respond with quotes, for the jurisdiction to evaluate the quotes, for vendors to manufacture the machines, and for the jurisdiction to train staff and pollworkers in the use of the new machinery.
An accelerated schedule for replacing existing machines has numerous risks. It is likely to reduce the number of vendors willing to offer quotes, it is likely to lead to inflated quotes to cover the costs of rushed manufacturing, it is likely to lead to quality control problems and poor pollworker training. The net result is both a high cost to the taxpayers and a high risk of serious problems during voting system deployment. The initial rollout of ES&S iVotronic voting systems in Miami in the early 2000's clearly illustrates these risks. The botched acquisition of Microvote MV464 voting machines by Montgomery County, Pennsylvania in the 1990 illustrates these risks even better.
Legislation to allow continued use of existing machines while forbidding replacement with non-conforming machines would work for trucks, but it would not work for voting machinery. The problem is, voting machines are purchased as part of a system that includes an election management system that runs at the jurisdiction's election office, central tabulators for absentee allots, ballot marking devices for the precincts, and ballot tabulators or the precincts. If a voting machine breaks, you cannot replace just that machine with one that is not compatible. You either replace that machine with an identical replacement or you replace the entire system.
The final point in the memo asserts that the legislation is "fueled by false claims and ulterior motives."
I cannot speak for all proponents of the bill, but I resent the implication that I am a liar or that I have some kind of ulterior motive. I have been involved with election technology since 1994, and I know personally many of the senior employees both Election Systems and Software and Dominion. I respect both companies, but I am also very aware of their shortcomings.
I have mentioned the Automark BMD above. That was a machine that ES&S sold under its own brand for years. As our efforts to draft the proposed legislation progressed, I became more and more impressed by the wisdom of Eugene Cummings, the inventor of that machine. We ended up using the Automark as a benchmark, being careful to check that nothing in our draft would threaten continued use of machines like it.
Today, when fully 1/3 of the electorate may believe that there was massive fraud in the election of 2020, merely conducting an honest election is not sufficient. A large fraction of the electorate does not trust the people who build election machinery or the people who conduct elections. To some on both the left and even more on the right, there is a deep-seated fear that voting machine manufacturers and election officials are somehow controlled by the deep state or some other nefarious, anti-democracy force.
Responding to this distrust with requests that people must trust some authority will not work. Too many voters will not trust any expert, any testing laboratory, any voting system vendor or any election administrator.
As a result, it is essential that we conduct our elections as transparently as possible. If a voter cannot see that the voting machine is incapable of altering a voted ballot, then we should not be using that machine. If a voter cannot see that a bar code correctly represents that voter's selections, then we should not use bar codes.