I am deeply concerned about Pennsylvania's interpretation of the HAVA Section 301 requirements with regard to the production of "a permanent paper record with a manual audit capacity" by voting systems that "allow the voter to correct any error before the permanent paper record is produced." I have examined all of the voting system certification reports on your Secretary of State's website and note that only your optically scanned paper systems appear to meet this dual mandate. That there is considerable debate about the Direct Recording Electronic (DRE) systems is reflected in your State Plan Element 12 section entitled "Voter Verifiable Paper Audit Trail." Although you have observed that the EAC has not mandated VVPATs in their 2005 VVSG, actually none of their guidelines are mandatory (they are all voluntary). As well, the EAC has abstained from commenting with respect to the ability to apply "uniform and nondiscriminatory standards" in states where diverse voting systems are employed that exhibit different rates of disenfranchisement (such as via high undervote rates). These are points of concern.
With regard to the semantic discussion that appears on pages 53-55 of your State Plan, the terminology has been incorrectly applied. The balloting process (whether on optically scanned paper or via a VVPAT-equipped DRE) always requires a casting action that confirms that the voter has had an opportunity to examine the paper record and has deemed it to be correct. Thus, all ballots are "verified" (in the same way that a signature on a loan contract indicates that the signer has reviewed the document, whether they actually did or not) through the casting action following the verification opportunity. Most federal and state legislative efforts in this regard have consistently used the word "verified" and not "verifiable" for this reason. I therefore recommend that the phrase "voter verified" (including in the phrase "Voter Verified Paper Audit Trail") be used in all instances in the PA HAVA Plan, rather than "voter verifiable." For additional purposes of understanding, an optically scanned ballot is a voter verified ballot, and a DRE with a VVPAT printing device also produces voter verified ballots if the printed documents are used to create the vote totals.
Pennsylvania's voting equipment examiners and the Secretary of State have already acknowledged numerous existing problems with certain of the State's DREs. Particular versions have even been belatedly decertified, after voters were disenfrancised in actual elections in the Commonwealth, or after reports of security vulnerabilities gained national attention. Unfortunately, the remaining DRE products all should be decertified by the State in order that "uniform and nondiscriminatory standards" are able to be applied, since currently only Pennsylvania's optically scanned paper systems ensure that a permanent paper record will persist that correctly reflects the voter's balloting intentions.
Each of the DRE products in Pennsylvania currently produce only an unobservable, after-the-fact paper record, generated from electronic data stored inside of the voting machine or on a removable cartridge. We have seen that if this data is deleted, damaged, or tampered with, it cannot be trusted to print records that accurately reflect each of the voters' choices. This was demonstrated when Beaver County observed their undervote rate soar by 600% following their move to the Unilect Patriot equipment from their former paper ballot system, and when Mercer County discovered that one of their Unilect voting machines only recorded 51 Presidential votes out of 289 ballots cast. In Berks County, some 40 Danaher voting machines used cartridges that were erroneously programmed by the vendor into a "training mode," resulting in 199 ballots being irretrievably lost. In each of these instances, voters were not provided with an opportunity to "correct any error before the permanent paper record is produced" since their votes vanished and were never reflected in the printouts. There is absolutely no guarantee that the existing DRE products in the State cannot and will not experience further troubles. The addition of VVPATs to the DREs would ensure that all voters will have their ballots retained and reflected in a recountable paper form.
Unfortunately, up to now, the vendor community has chosen to present the Commonwealth only VVPAT configurations that either use continuous (reel-to-reel) paper, or print identifying sequences (whether encrypted or not) on the paper records. The Commonwealth has correctly disallowed these products on the grounds that they violate voter privacy, but properly configured VVPATs do exist. Pressure by the 27 states that have now enacted VVPAT laws, some of which have implementation deadlines for the 2008 election cycle, along with certain successful lawsuits, have compelled vendors to accelerate their development processes to ensure that VVPAT solutions, that also satisfy anonymity requirements, are now being brought to market. The excuse given in the report that the Department has not yet examined a VVPAT system that it can accept, ignores the fact that the Commonwealth has also not proactively sought or worked with vendors to ensure that such equipment will be available. Similarly, the excuse that the VVPAT printers will somehow be defective is absurd, especially in a State where millions of cut-and-drop lottery tickets and ATM receipts are printed every day. Vendors have responded favorably when voting equipment was not certified, by returning to the Commonwealth with revised and improved products, so it seems highly likely that they will also want to comply with State VVPAT requirements in this regard. Pennsylvania has an obligation to all of its citizens to ensure that their votes will not vanish.
I urge that these corrections be made to the HAVA Plan and would like to offer that you can communicate with me, should you require any further explanation or comment.
Rebecca Mercuri, Ph.D.