Presidential Advisory Commission on Election Integrity September 12, 2017, Meeting Materials

I’ve reviewed the materials released in advance of the September 12, 2017, meeting of the Presidential Advisory Commission on Election Integrity (PACEI), submitted by testifiers and members. What follows is my critical summary, organized to match the agenda for the meeting. I did my review a couple days before the meeting, so if there are any last minute additions, those may be missing.

Panel One: Historical Election Turnout Statistics and the Effects of Election Integrity Issues on Voter Confidence

Dr. Andrew Smith, Associate Professor of Political Science, University of New Hampshire, provided slides for his presentation. He shows that turnout varies substantially between states and over time and that there are lots of factors influencing this, such that attributing changes to any particular cause is difficult. Nonetheless, he considers public confidence important, not because of its effect on turnout but because it ensures that the public will accept the results of elections and go along with the outcomes. However, there are “tradeoffs between stricter voting laws to increase trust and barriers that may discourage voting.”

Kimball Brace, President, Election Data Services, Inc., provided slides for his presentation. He warns of the complexities hidden in two seemingly simple questions: (1) how the number of registered voters compares with the number of voting-age citizens, and (2) what fraction turn out to vote. Some of the difficulties are definitional; individual jurisdictions make different choices how to define important terms. Other difficulties result from estimation errors. In particular, the number of voting-age citizens residing in any jurisdiction at any time is necessarily an estimate. Small jurisdictions are more apt to have proportionately large estimation errors, and Brace reminds the commissioners that most jurisdictions are small. As an example of how these factors combine, Brace gives several reasons why a jurisdiction’s number of registered voters might exceed its number of voting-age citizens — not just the obvious one of “deadwood,” but also the inclusion of military and overseas citizens and pre-registered 16 and 17 year olds, as well as population estimation error.

Dr. John Lott, President, Crime Prevention Research Center and Author, Evidence of Voter Fraud and the Impact that Regulations to Reduce Fraud Have on Voter Participation Rates (2006), provided both the paper mentioned in his title and slides for his presentation. That presentation is divided into two parts. The shorter first part proposes checking voters with the same system used for gun buyers, the National Instant Criminal Background Check System (NICS). The longer second part summarizes the 2006 paper.

Lott’s first part is based on the following syllogism: any verification system imposing low enough burden on gun rights also would impose low enough burden on voting rights; NICS imposes low enough burden on gun rights, according to Democrats; therefore, NICS should be accepted by Democrats as imposing low enough burden on voting rights. The essential question about this syllogism is not its logic but Lott’s rhetorical purpose in offering it. He does not agree with the premise that NICS has low burden, nor does he offer any independent reason why people like himself, who see NICS as high burden, ought to nonetheless favor its use for voting. Therefore, it does not appear that he intends the syllogism to be used in its forward direction as a reason for adopting NICS in the voting context. Rather, it appears he is advancing the logically equivalent contrapositive: because NICS is overly burdensome for voting, Democrats ought to agree it is overly burdensome for buying guns. If this is his point, he is wasting the commissioners’ time; they have no mandate to make recommendations about gun laws.

Lott’s second part, like the paper it is based on, has as its core his analysis of county-level data from July 1996 to July 2006. He claims this analysis shows no overall impact of photo ID laws on turnout but an increase in turnout from their adoption in counties identified by the American Center for Voting Rights as particularly prone to fraud. However, these conclusions are undermined by the statistical naivety of his analysis, which may explain why the paper remains unpublished. Dr. Michael McDonald, Associate Professor of Political Science, University of Florida, has provided a concise summary of that naivety in the form of a Twitter thread. For a layperson not interested in the methodological details, a telling point may be that Lott’s same analysis suggests one can suppress turnout by 12 percentage points simply by putting a business-regulation initiative on the ballot. (Lott does not include this result from the paper in his presentation to the commission.)

Panel Two: Current Election Integrity Issues Affecting Public Confidence

Donald Palmer, Former Secretary, Virginia State Board of Elections, provided a written version of his testimony. He summarizes and endorses pervious commissions’ recommendations regarding the maintenance of accurate lists of registered voters. For example, he advocates that states offer online voter registration and participate in interstate comparisons. Beyond the existing recommendations, he suggests the commission consider recommending new federal mandates to ensure uniform implementation of best practices for list maintenance. For example, states could be mandated to participate in interstate exchanges and to provide information about each new registrant to the election official where that registrant previously resided.

Robert Popper, Director, Election Integrity Project, Judicial Watch, provided a written version of his testimony. He has been involved in litigation claiming that states are not abiding by the list-maintenance requirements of the National Voter Registration Act (NVRA). Although the existence of that litigation suggests that these claims are contested, he zealously advocates for his position. For example, he ignores Brace’s cautions (summarized above) regarding the complex reasons why registrations may exceed population estimates and regarding the diversity of jurisdictions, many of them small. He also shows considerable frustration with legal challenges that have been mounted against more aggressive list-maintenance activities, again presenting these cases in one-sided fashion. For example, he expresses astonishment that a court would apply the NVRA’s prohibition on registration purges in the 90 days preceding an election to the purging of registrations by noncitizens. After all, “noncitizens were never eligible to vote.” Notably absent is any mention of the risk that a citizen’s registration might accidentally be purged in the course of purging noncitizens.

Ken Block, President, Simpatico Software Systems, provided slides for his presentation as well as a related report from the Government Accountability Institute on duplicate voting, which he helped prepare. Although he devotes a portion of his presentation to the claim that a “fake voter” can, in theory, cast a ballot, the more controversial part is his claim to have found 8,471 cases in which he is highly confident an individual cast more than one ballot in the 2016 general election.

This claim rests on matching data released by 21 states. That released data does not even include all the data available to the states themselves; for example, it may omit exact birthdates and the portion of the Social Security Numbers (SSNs) that the states record. [Update: Block pointed out a portion of the GAI report I missed: although states were only asked to provide full birthdates “when available,” for any that didn’t, “GAI did not use its data.”] Yet Block’s presentation indicates that “only pairs of votes where the social security numbers matched are counted as high-confidence matches.” Likewise, the report makes a big deal out of the statement that “The probability of correctly matching two records with same name, birth date and SS# is close to 100%.” (In other words, if all that information were available, the match would be highly confident.)

How is this discrepancy bridged? Block’s team took the limited data released by the state and extended it with additional data (such as full SSN) by using commercial data sources. Only after this extension did they match it between states. However, they neglect to point out that the extension process is itself a form of matching, subject to its own rate of false matches. When a state’s released data is matched with commercial sources, SSNs are not available to guide that matching.

Here is an example. Are the John Doe born in 1953 who voted in one state and the John Doe born in 1953 who voted in another state the same person? Maybe, maybe not. Suppose commercial data sources allow an educated guess to be made that the first John Doe was born on April 1, 1953, and has SSN 111–22–3333. Suppose that those same sources likewise allow an educated guess to be made that the second John Doe was also born on April 1, 1953, with SSN 111–22–3333. Are the John Does born on April 1, 1953, and April 1, 1953, with SSNs 111–22–3333 and 111–22–3333 the same? Yes, surely. But are the two voters the same? Maybe, maybe not — all depending on whether the educated guesses were correct. [Update: as noted above, the April 1 portion would have been in the source data. This doesn’t change the point that the SSNs that provide the sure match are not from the states.]

[Added: also, even if the two John Does are the same, one or the other voter history record may be in error. Election administrators know there is some rate of errors in posting the history from the pollbooks. Sometimes, someone is shown in the computerized history as having voted, but when you go to look at the actual pollbook, you see it was the person above or below who actually signed in.]

Hans von Spakovsky, Senior Legal Fellow, Heritage Foundation and Member, PACEI, provided slides for his presentation. He makes a case for the existence of election fraud and registration list inaccuracies and suggests that these problems not merely exist, but are substantial. In part, he does this by reiterating the same data as Popper and Block. However, he also presents other sources of data, including the broad overview of many different forms of election fraud, over a long time period, that the Heritage Foundation has assembled. The most novel part of the presentation, which has gained considerable attention from the press, focuses on data from New Hampshire’s 2016 general election regarding voters who registered using out-of-state driver’s licenses and did not subsequently obtain a New Hampshire license or motor vehicle registration. To support that portion of the presentation, the meeting materials include a data request from the Speaker of the New Hampshire House, the responses from the Departments of State and Safety, and the Speaker’s report of those responses.

As the departments’ responses explicitly state, these people may have been legal New Hampshire voters. They may not have continued to drive in New Hampshire and so may not be subject to any requirement to get a New Hampshire driver’s license. Or, they may have driven but been exempt from that requirement based on the differing definitions of domicile and residence used in the election and motor vehicle laws. Even if neither of these circumstances apply to an individual, it would generally suggest a violation of the motor vehicle licensure law, not voter fraud.

Panel Three: Electronic Voting Systems and Election Integrity — A Primer

Dr. Andrew Appel, Professor of Computer Science, Princeton University, provided a written version of his testimony with slides embedded. Appel starts with a broad focus, looking at the full problem of ensuring accurate, secret voting without needing to trust others. This leads him into such diverse topics as the role of graphical design in helping voters accurately record their intent and the potential for tampering with computer systems. His bottom-line conclusion is that elections should be conducted using “voter-verified paper ballots, random audits before results are certified, and transparency in reporting.”

Dr. Ronald Rivest, Professor of Computer Science, Massachusetts Institute of Technology, provided slides for his presentation. Those slides leave a great deal to the oral portion of the presentation, but it appears that his recommendations are the same as Dr. Appel’s. He expands on the audit component, explaining the power of comparison risk-limiting audits, which compare randomly chosen paper ballots with how those votes were recorded. Even with far fewer ballots examined than in a full recount, if no problems are discovered in the audit, it provides a low risk that any undiscovered problems would affect the election’s outcome.

Harri Hursti, Co-Founder of Nordic Innovation Labs, provided slides for his presentation. He offers an attacker’s perspective on the vulnerability of election-related computer systems to tampering. The short version would be “be afraid, be very afraid” (unless, per Appel and Rivest, you can check election results without relying on computer systems). He points out three facts that challenge conventional wisdom: attackers may be aiming for “distrust and chaos” rather than a particular election outcome; attackers can achieve large-scale results using the Internet even if election systems are not connected to the Internet during elections; and the tools for doing so are readily available at little cost. The bulk of the presentation is occupied by eye-opening examples of what is possible. For example, something that appears to be a USB storage device may in fact take control of the computer into which it is plugged. Control over central election-management systems can be leveraged to attack disconnected vote tabulators. And even something as innocuous as a bar code may exert unexpected control when scanned.

Discussion and Other Business — All Members

In addition to Hans von Spakovsky, who is testifying on Panel Two, four other commission members submitted materials. Secretary Bill Gardner is hosting the meeting and his presentation slides look like they are intended for his welcoming remarks, whereas the other three are presumably slated for the “discussion and other business” agenda item. Beyond laying out the importance of confidence and turnout, Gardner’s slides make the point that “if it ain’t broke, don’t fix it.”

J. Christian Adams, President and General Counsel, Public Interest Legal Foundation, submitted five items. One is a paper from the foundation entitled “Garden State Gotcha: How Opponents of Citizenship Verification for Voting are Putting New Jersey’s Noncitizens at Risk of Deportation.” The other four are reports of court cases in which he participated as counsel: Bellitto v. Snipes (2016), Am. Civ. Rights Union v. Martinez-Rivera, Bellitto v. Snipes (2017), and Voter Integrity Project NC, Inc. v. Wake Cnty. Bd. of Elections. The court cases all involve claims that election officials did not conduct the list maintenance activities required by the NVRA, while the foundation’s paper points to the potential for the NVRA’s “motor voter” provision to facilitate noncitizens’ accidental registration to vote, in absence of citizenship verification.

Matthew Dunlap, Maine Secretary of State, submitted a collection of news summaries regarding Bates College, including in particular one that summarized news stories about an attempt to suppress student voting in the November 2016 election with an anonymous “legal advisory” flier claiming that those who chose to register and vote locally would need to pay to change their driver’s licenses and motor vehicle registrations.

Alan King, Judge, Probate Court of Jefferson County, Alabama, submitted a “Statement of Issues/Recommendations” with supporting attachments. (According to a published report, “King won’t be attending the … meeting … because of a scheduling conflict.”) The statement has drawn attention for its remark that “this commission, and we as a people, should be expanding the rights of our citizens to vote, instead of arguably looking for ways to keep people from voting.” However, it also contains ten specific recommendations, spanning a wide range of topics: federal funding for voting equipment, federal criminalization of voting suppression, reliance on statistical experts, reaffirmation of state responsibility for elections, federal felony-level criminalization of double registration, legislation treating election cooperation with a foreign government as treason, federal felony-level criminalization of “hacking” election systems, security and non-partisan governance for interstate registration data exchanges, engagement with “hacking” experts, and focusing on alleged Russian “hacking” of the 2016 election, with full disclosure of the findings.

All ten of these recommendations are general topics for the commission to work on and would require greater specificity before they could be adopted. However, the level of detail Judge King provides varies. For example, with regard to federal funding, he specifies a particular dollar amount, whereas with regard to the criminalization of double registration, he leaves open such important questions as what a workable intent standard would be.

About the author: Max Hailperin is Professor Emeritus of Mathematics, Computer Science, and Statistics at Gustavus Adolphus College; he earned his Ph.D at Stanford and S.B at the Massachusetts Institute of Technology. In 2014, he was awarded the National Association of Secretaries of State (NASS) Medallion “in recognition of his service and contributions to election-related technology and legislation.” He was appointed by Governor Mark Dayton to the Electronic Roster Task Force in 2013.