Guest post by Max Hailperin
Dan McGrath, the executive director of Minnesota Majority, is the key advocate of Minnesota's proposed constitutional amendment regarding voting. Dan worked tirelessly in 2011 to shepherd the so-called "21st Century Voting Act" through the legislature, only to have it stopped by Gov. Dayton's veto pen.
That set the stage for the amendment Dan has been selling us ever since. The amendment fight is incomprehensible without trying to understand Dan, difficult though that is. You might as well study the fall of the Romanovs without trying to understand Rasputin.
For that reason, people sat up and took notice when Dan recently began saying that under the amendment, Election-Day registrants could still use all existing proofs of residence including vouching. That was a remarkable turnabout for the man who in February, testifying to the Senate Local Government and Elections Committee about the proposed amendment, said "Combining Election Day registration with vouching is ridiculous." Lest anyone misunderstand him, he followed up with a blog post saying "The truth is Voter ID will not eliminate Election Day registration – only vouching."
This turnabout on vouching came not in isolation, but together with claims that photo IDs would not need current addresses and that provisional ballots would only be used for voters without ID, not for registrants whose eligibility couldn't be verified. These assertions contrast with a report Dan presented as the legislature took up the amendment, in which he called for "provisional ballots so that Election Day registrants who lack proper identification including their current address in the precinct can be subjected to the same scrutiny as all other voters, including verification of their address by [Postal Verification Card (PVC)] or other means before their ballot is counted."
Dan's demand for "the same scrutiny" lay behind the amendment's requirement that all voters "be subject to substantially equivalent ... eligibility verification" before their ballots are counted. His startling new positions seem to suggest that this clause of the amendment could be implemented in a way that falls far short of its motivations. For example, the legislature could stop checking residences entirely, leaving pre-registered voters as unverified as Election-Day registrants. That doesn't seem politically realistic but might reflect Dan's newly defensive situation. Instead of trumpeting the amendment's virtues, he's minimizing its faults. Until this week, I thought that was all that was going on.
On Monday, Dan spoke at an informational forum on my college campus. He repeated his new positions on vouching and ID-card addresses. Notably, he did so only seconds after he had pointed to the large number of undeliverable PVCs as a sign that Election-Day registrants might have used false addresses, which he cited as a motivation for passing the amendment. I couldn't help but wonder, if this is still the problem he is trying to solve, how could he reconcile that with continuing to allow all the same proofs of residence, including vouching? I put the question to him after the forum.
Dan's answer to my question was quite illuminating as to how his new positions are grounded in something much deeper than mere speculation by a man forced to play defense. Indeed, his new positions are grounded in the same fears and hopes that have driven him since the conception of the "21st Century Voting Act."
Read the rest below the fold
First, Dan explained to me why an ID requirement is his top priority, even if it couldn't be coupled with eligibility verification. Of course, the amendment itself would demand eligibility verification; Dan's hopes explain how he anticipates compliance. However, he lead off not with those hopes, but with the fears that make him wiling to shortchange eligibility verification if necessary to make an ID amendment fly.
You might think the only question about undeliverable PVCs is whether they reflect voters who moved after the election, as Beth Fraser testified at the February hearing, or falsified addresses. (Beth is Director of Governmental Affairs for the Office of the Secretary of State.) Dan considered undeliverable PVCs to be highly suspicious if the Post Office had no forwarding address; Beth explained that one-third of those who move don't provide the Post Office with a change of address.
In fact, though, Dan thinks those undeliverable PVCs might represent something even more sinister than falsified addresses. They might represent voter identities that were created "out of whole cloth." There could be some criminal mastermind who is registering and voting repeatedly under a whole list of fictitious identities without ever getting caught, indeed without the County Attorneys even being aware that this crime spree is happening under their noses.
If one accepts this premise, then an ID, even without address verification, might be enough to stop false-address voting. Once authorities knew for sure who is voting, they could easily check for multiple voting; anyone who considered using more than one address would be deterred by the certainty of being caught. Deterrence could even prevent a scenario in which hundreds of voters each cast a single vote in a competitive district rather than their real, uncompetitive district. In this latter scenario, each voter would risk being caught for falsifying their address. Contrast that, Dan suggests, with the impossibility of ever prosecuting someone whose name is as unknown as their address.
Dan ties this possibility of fictitious identities to the practice of vouching; he says that at present, with no photo ID requirement, vouching is de facto establishing identity as well as residence. This presents the first weakness in his argument. He is trying to sketch a scenario where a crime would be committed with no one who could be held accountable, yet he wrecks that by bringing a voucher into the picture. The voucher must already be a registered voter or a designated residential facility employee.
As such, even if the vouched-for registrant turns turns out to be a phantom, the voucher could be located and prosecuted if they are a knowing accomplice. Thus, our untraceable phantom voter is apparently not only going to need to lie to the election judges but also to con some gullible, naive registered voter into vouching for a stranger. A good con artist might be able to pull that off, but this extra step increases the chance of detection -- and the phantom voter has never once been caught.
A second weakness in Dan's nightmare scenario is that the PVC is only one of the checks done after the election. The Election-Day registrant's identifying information is also matched against the databases of the Department of Public Safety and the Social Security Administration. If Micky Mouse registers at 123 Main Street with a birthday of April 1st, the County Auditor will get more than just a PVC delivery failure -- they'll also get the information that no Mr. Mouse is known to exist with that birthdate. (If the registrant made up an identifying number, that will be yet one more fact that will fail to check out.)
This combination of circumstances would be much more of a red flag than a mere undeliverable card. Maybe a sleepy County Auditor wouldn't notice, but once again the chance of detection has gone up -- and no County Auditor or County Attorney has ever reported these signs of a phantom voter. Likewise, if the criminal instead usurps some real individual's identity, other signs could surface -- and they haven't.
In addition to these possibilities for detection within the election system, there are many more mundane risks, like running into an old friend at an inopportune moment or bragging indiscreetly while drunk. Suppose, though, our master criminal is so good at avoiding all these risks they can pull off the phantom vote with a 99% chance of remaining undetected. Is that good enough? The real problem is that even a slight risk, if taken repeatedly, grows exponentially. Take the 99% chance of success, for example. If repeated enough times to overcome Senator Franken's 312-vote margin, the overall probability of remaining undetected becomes 0.99 raised to the 313th power: only about a 4% chance. Even such a slick operator as to have a 99% chance of success turns into a dead duck with a 96% chance of detection as a result of the need for repetition.
This leaves us with two possibilities. Either our elections aren't rife with phantom voters, or the crime is being perpetrated by someone of truly phenomenal abilities, an absolute master of deception to the point of seeming invisibility. Dan's fear is that such a person exists. Who is this person? Dan himself provided a clue in his February testimony.
Earlier testifiers had described the difficulties they faced getting identification cards. For example, an oxygen-dependent woman described going from government office to government office in search of supporting documents, dragging her tank along and hoping to get the process done before she ran out of oxygen. Dan scoffed at these hardships, saying "It's almost like the testifiers think there are armies of ninjas that they have to hand-to-hand combat their way through to the DPS station in order to get their free identification card." Ninjas, huh? Maybe he's on to something.
Even a cursory examination of Wikipedia reveals that ninjas were not so much fighters as cunning saboteurs, known above all for the ability to pass undetected into the enemy's castle. They were such masters of false identity that they were said to have the art of invisibility. In short, it is Dan himself who fears ninjas -- ninjas who are sneaking into polling places as though they were castles in feudal Japan.
On the other hand, Dan recognizes that he can't entirely give up on verifying eligibility. If nothing else, he persuaded the legislature to mandate it in the proposed amendment. If he now thinks Election-Day registrants can continue presenting proofs of residency he has deemed unreliable (such as sworn testimony of personal knowledge, the gold standard of the courtroom), how does he expect eligibility to be verified at the polling place? (We must take him at his word that he doesn't expect massive provisional balloting, the alternative listed in his January report.) The answer lies in his hopes, as expressed in the "21st Century Voting Act."
Dan is quite fond of the phrase "21st-century," which to him connotes the use of cutting-edge computer technology. The 21st century began with 2001, so HAL is a good symbol for Dan's technological hopes. Like HAL, Dan's imagined electronic poll books would have all relevant information available at a moment's notice. Equally importantly, we could rely upon the entire fleet of computers, knowing none "has ever made a mistake or distorted information." Indeed, the electronic poll books would be "by any practical definition of the words, foolproof and incapable of error." This is essential, because unlike the computers used to scan ballots, these would stand between a prospective voter and the ability to vote.
Specifically, Dan pointed out to me that the PVC is only one of the checks done on registrations and that if it were omitted and the rest of the checks done in the polling place, then that could be "substantially equivalent." His stated goal is to bring the same verification technology used by County Auditors into the polling places.
This dream has only three flaws: the non-PVC checks don't suffice to verify addresses, they aren't accomplished in real-time even in the Auditors' offices, and extending online computer access to the precincts is a recipe for disaster second only to HAL's "open the pod bay doors" moment. Indeed, speaking as a computer scientist, if I were commissioned to realize Dan's vision, I'd quote HAL itself: "I'm afraid I can't do that."
Most of the verification checks on registration data don't include the place of residence. For example, verification against the databases of the Department of Public Safety (DPS) and Social Security Administration check only the name, date of birth, and identifying numbers. Nor does the state legislature have much ability to overcome this limit. Paragraph 205(r)(8)(D) of the Social Security Act imposes it with regard to the federal data. Within the state databases, address mismatches often happen because the voter registration data is more current.
In fact, only one of the external databases compared against registration data provides address information: the National Change of Address data provided by the United States Postal Service. Rather than being a comprehensive list of addresses, it provides automated updating of addresses for those who have recently moved and have submitted forwarding information to the Postal Service.
Still, for all the limitations these databases have, if they are instantly accessible in a County Auditor's office, why not perform the same checking on Election-Day registrants? The first huge problem with this logic is that its premise is false. The databases are not, in fact, instantly accessible. They are checked overnight in a very 20th-century batch fashion. Each night, the Secretary of State's office ships registration data to the DPS and a National Change of Address service provider and receives back the matching results. Each morning, the County Auditor receives a report listing questionable matches in need of manual review. As with the scope of the matching, this process is not easily changeable for legal as well as technical reasons. The Social Security Administration and Postal Service are understandably very concerned about preventing inappropriate disclosure of their data. As a result, they impose stringent restrictions on how matching takes place.
Last but not least, there is the question of what kind of technology deployment is feasible in thousands of polling places scattered around the state, popping up for a single day only and staffed by volunteers. This was the aspect of the 21st-Century Voting Act I was intensively involved with during the 2011 session. As HAL itself reminds us, 21st-century technology can go horribly wrong when it fails. Indeed, we have no need to turn to fiction or even leave the realm of electronic poll books to see the potential for disaster. As described in an assessment report by consultants, Denver's ill-considered deployment of electronic poll books in 2006 "led to unacceptably long waiting times for voters and an abandonment rate estimated at 18,000-20,000 voters (approximately 20% of the anticipated physical turnout on Election Day)."
As originally introduced, the 21st-Century Voting Act called for each precinct to be connected to the statewide system so that registrants and voters could be immediately checked. In particular, this system was required to check that no identity was used in more than one precinct. As I pointed out in my testimony, this precaution was unnecessary given the bill's stringent requirements for proof of residence within the precinct. As I also pointed out, the language in the bill had been lifted from Colorado's statutory and regulatory requirements for electronic poll books used in counties that had switched from precinct-based voting to "vote centers." In the vote center model, a county offers a smaller number of large polling places and allows each voter to freely choose among them. For example, a voter could vote near their place of employment rather than their home.
In this context, checking for double voting becomes important. However, the technical burden is also reduced because there are so many fewer polling places. Moreover, the requirement was only for countywide communication, not statewide. I warned that a statewide system reaching the level of individual precincts was completely unprecedented. "One consequence is needless expense. But an even greater concern is the risk of election-day chaos. There is no better recipe for technology adoption to go wrong than inappropriate requirements that demand needless complexity and interconnection of components."
After my intervention, the bill was amended so that counties could individually decide whether to use the electronic poll books and so that those that opted in would not be required to connect the polling places to a central server in an "online" fashion. Instead, each polling place could operate independently, as is current practice. Data would be loaded onto the electronic poll books in advance of the election and results would be uploaded after the election was over. This "offline" process would greatly reduce the risks of disaster.
For those counties that opted not to use electronic poll books, the final version of the 21st-Century Voting Act required that each polling place have a printed list of ineligible individuals (felons) to consult when processing Election-Day registrations. This means counties would have a choice between a capital expenditure for computers or an operating expense for printed lists and the election judges to look through them. The ineligible voter list is feasible from a legal and technical standpoint. Indeed, Wisconsin has used such lists since 2006.
Dan has pointed to this final less-ambitious version of the 21st-Century Voting Act as an example of how the proposed constitutional amendment could be implemented with a combination of offline computers and paper lists. However, when I lobbied for that combination in 2011, we were under no constitutional mandate to provide "substantially equivalent ... eligibility verification," and indeed the bill (had Gov. Dayton signed it) would have achieved no such thing. Checking a list of felons is one very plausible approach toward solving a real problem our state faces. (The Task Force on Election Integrity has suggested others.) However, it could not duplicate the checking of external databases done each night. Nor could it even duplicate the one verification check currently performed in real time at the County Auditors' offices, and which an online system could bring to the polling places (at great expense and risk).
Suppose real-time checking of databases were possible. Could it be achieved offline by simply giving each polling place its own copy of the data, much as they would have their own lists of felons? Recall that one of the barriers to real-time checking is the security concerns of the data stewards. For the Secretary of State's office to check Social Security records, they need to send the data to the Department of Public Safety, which must have negotiated an agreement with the Social Security Administration that includes "safeguards to assure the maintenance of the confidentiality of any applicable information". I doubt that negotiation would go far if Minnesota asked to distribute thousands of copies around the state in both digital and printed form.
Lastly, consider the one check that Auditors do perform instantly. The system ensures that each voter is only registered at one residence; entering a new address automatically invalidates any prior address. This is a crucial verification step even given the use of PVCs, because many Minnesotans have more than one address at which they can receive mail; only one of those addresses can be their residence for voting. The 21st-Century Voting Act as it was finally passed did not perform this check in the polling place for Election-Day registrants. Instead, it imposed a very stringent requirement for proof of residence. Moreover, this bill did not abandon the current notion that some eligibility verification can be done after a ballot is cast.
Dan has recently been making three claims: (1) that all existing proofs of residence, such as vouching, will remain available; (2) that polling places will not need online computer connections; and (3) that Election-Day registrants will be able to actually cast ballots, not just submit provisional ballots. The proposed amendment itself provides an additional requirement: (4) that all voters be subject to substantially equivalent eligibility verification before their ballots are counted. Political reality impose a final requirement: (5) that we not drop our existing verification that each voter is only voting in a single precinct where they legally reside.
My conclusion is that these five are mutually inconsistent. Dan might hope for enabling legislation that achieves all of them, but that hope has no basis in reality. HAL is science fiction, not science fact.
Professor Max Hailperin, of Gustavus Adolphus College, has examined new technologies in voting nationwide. His written testimony was included in hearings on HF 210, Mary Kiffmeyer's photo ID (and much more) bill that was vetoed last year by Governor Dayton.
Images: Dan McGrath, by Ken Avidor (above); HAL (below)