Wendy Weiser Before the Election Assistance Commission

The Center's Deputy Director of the Democracy program testifies regarding voter registration databases.

March 17, 2009

Download testimony [pdf]
Click here for access to the webcast.

Testimony of
Wendy R. Weiser
Director, Voting Rights and Elections Project
Deputy Director, Democracy Program
Brennan Center for Justice at NYU School of Law

Before the
U.S. Election Assistance Commission
Voter Registration Databases: Initial Discussion
on Reviewing HAVA Mandated Guidance
 

On behalf of the Brennan Center for Justice at NYU School of Law ("Brennan Center"), I thank the U.S. Election Assistance Commission ("EAC") for holding this public hearing and providing me the opportunity to submit observations and recommendations concerning statewide voter registration databases.

My name is Wendy Weiser, and I am Director of the Voting Rights and Elections Project and Deputy Director of the Democracy Program at the Brennan Center.  The Brennan Center is a nonpartisan think tank and legal advocacy organization that focuses on issues of democracy and justice.  Among other things, we seek to ensure fair and accurate voter registration and voting systems and to promote policies that maximize citizen enfranchisement and participation in elections.  We have done extensive work on the subjects of voter registration and the maintenance of voter registration lists, including conducting studies and publishing reports; providing assistance to federal and state administrative and legislative bodies with responsibility over elections; and, when necessary, litigating to compel states to comply with their obligations under federal law and the Constitution.

I.          Voter Registration Database Concerns in 2008

I was asked to discuss how statewide voter registration databases worked in the 2008 election cycle.  While the Brennan Center has not conducted a technical assessment of statewide voter registration databases, we have examined state policies regarding how those databases are used.  There were four areas of primary concern in 2008: (1) "no match, no vote" policies and other matching problems in the HAVA verification process; (2) matching voter registration data across state lines; (3) voter registration database maintenance (or purge) efforts; and (4) the brief shut down of the Social Security Administration database for maintenance.  I address each in turn and include recommendations for the EAC to help prevent future problems and improve database use and performance.  The concerns addressed in my testimony today are similar in a number of respects to those I previously raised in testimony and comments submitted to the EAC in 2005 in connection with the agency's first database guidance.  I attach those documents as an appendix to this testimony.[1]

A.        "No Match, No Vote" Policies and Poor Use of Database Matching for HAVA Verification Purposes

Perhaps the most controversial use of statewide voter registration databases in the period leading up to the 2008 elections was the "no match, no vote" policy.  "No match, no vote" refers to a practice currently in place in a small number of states whereby election officials do not register voters or purge them from the rolls if the officials cannot "match" the voters' registration information against information in the driver's license or Social Security databases, typically using electronic match processes.  The "no match, no vote" practice is rare:  in 2008, only four states (Florida, Iowa, Louisiana, and South Dakota) had a "no match, no vote" policy in place, and at least two of those states (Florida and South Dakota) used procedures to mitigate, but not eliminate, the negative impact of the policy.  Although rare, the practice is significant because it is very harmful to prospective voters.  As discussed below, because the "matching" processes used in connection with statewide voter registration databases are inherently unreliable, "no match, no vote" policies typically result in large numbers of eligible voters being left off the rolls and, as a result, being unable to vote meaningful ballots.

            1.         "No Match, No Vote" Controversies in 2008

The "no match, no vote" practice achieved notoriety in 2008 not only because it was blocking voters in Florida, Iowa, Louisiana and South Dakota, but also because there was a concerted effort in at least two other states-Ohio and Wisconsin-to force election officials to adopt a "no match, no vote" policy in the weeks immediately preceding the hotly contested election.  In both states, after election officials refused to adopt "no match, no vote" policies because of concerns over their disenfranchising effect, political operatives brought lawsuits claiming that HAVA required the states to impose those policies.  Both lawsuits were filed in September 2008, and both were decided less than three weeks before the election.[2]  Both lawsuits failed.  The Wisconsin court found on the merits that HAVA did not require election officials to link voter eligibility to a successful match, and that to do so would violate a key provision of the Voting Rights Act that prohibits vote denial on the basis of immaterial errors or omissions in paperwork.  Had these efforts succeeded, hundreds of thousands of voters would have been adversely affected.[3]

The "no match, no vote" policy was similarly controversial in Florida, where as a result of a lawsuit brought in 2007 by the Brennan Center and co-counsel, the state had suspended operation of the policy throughout much of the year.[4]  On September 8, 2008, only weeks before the voter registration deadline, the Secretary of State instructed election officials to start applying the policy to the tens of thousands of registrations that were pouring in during that period.  In an effort to minimize the disenfranchisement of eligible voters, a number of county election officials disregarded the Secretary's interpretation of Florida law and made available a procedure-albeit an imperfect one- whereby unmatched voters could remedy the problem and cast regular ballots at the polls on Election Day.  Unmatched registrants in the other counties had to follow much more onerous procedures in order to vote.  Many were unable to vote or to have their votes counted.

            2.         Matching is an Unreliable Basis to Exclude Voters 

"No match, no vote" is a harmful policy because it functions mainly to exclude eligible voters based on administrative errors.  Matching-especially the way it is employed in the election administration context-is simply not a reliable means of identifying ineligible voters.  Matches between statewide voter registration databases and other government databases regularly fail because of issues that have nothing to do with voter identity or eligibility, such as data entry errors made when inputting hand-written information from voter registration applications, inconsistent treatment of hyphenated names in different databases, or the use of a married name in one database and a maiden name in another. 

Evidence from across the nation demonstrates that, even when voters are eligible and provide complete and accurate information, these trivial discrepancies cause matches to fail at extremely high rates:

  • In Florida, in the final four weeks of the registration cycle for the 2008 general election, more than 14% (62,882) of nearly 437,638 registration records failed to match initially.[5]  The vast majority of those failed matches were later proved to be erroneous.  After follow up investigation, the State determined that nearly two-thirds of the failed matches were the result of mistakes like data entry errors.  After correcting these obvious errors, more than 22,000 registration records remained unsuccessfully matched.[6]  A significant number of those registrants later proved the accuracy of their registration information by showing county election officials or poll workers copies of their driver's licenses or Social Security cards, but more than 12,000 voter remained unmatched on Election Day.  Those who did not do so may not have received notice of the problem or may have been unable to take the time before Election Day to visit an elections office.  There is no evidence that any of the failed matches were associated with ineligible voters.
  • In mid-2008, Wisconsin initially had a non-match rate of 22%; that is, nearly one in four attempted matches failed on the first attempt.[7]  After improvements to the matching protocol, Wisconsin lowered its non-match rate, but 12% of applicants still failed the first matching attempt.[8]  A recent report from Wisconsin's elections agency concluded that more than 88,000 registrations were unmatched, and that more than 90% of these failed matches were due to issues with exact matches of names and driver's license numbers.[9]
  • In Ohio, more than 200,000 (almost one-third) of approximately 660,000 voters who registered in 2008 were not successfully matched before the November 2008 election.  A lawsuit filed on the eve of the election threatened these more than 200,000 Ohioans with disenfranchisement; the U.S. Supreme Court ultimately threw out the lawsuit on October 17, 2008.[10]
  • In Texas, in 2006, 49,271 voter registration records were not successfully matched, including 27% of all records submitted to the Social Security Administration and more than 10% sent to the state motor vehicle authority for matching.[11]  A state representative examined a random sample of 300 of these failed matches and positively determined that, in 277 cases, the matches failed because of "variations in hyphenated names, typos or maiden names."[12]
  • In the first six months of 2006, before its "no match, no vote" law was blocked by court order, Washington had a failed match rate of 16% statewide, and up to 30% in King County, which includes Seattle.[13]
  • Through April of 2006, 18% of voter registration records in Los Angeles County, California, failed to match; virtually all of the registrations in question were associated with individuals whom local election officials determined to be eligible to vote.[14]  (For this and related reasons, California abandoned its "no match, no vote" policy prior to the 2006 elections.)
  • Nearly 20% of an audit sample of 15,000 applications submitted for matching in New York City in September 2004 could not be matched; the audit of those failed matches found that most non-matches were due to typos and other data entry errors.[15]  

This evidence shows not only that no-match rates are extremely high across the country, but also that voter registration records that fail to match almost always correspond to eligible voters; virtually all non-matches can be attributed to typos and other minor errors.  Taken together, these facts demonstrate that "no match, no vote" policies disenfranchise large numbers of eligible voters for no good reason.

            3.         Problematic Matching Criteria, Especially in the HAVV Process

One of the principal reasons that no-match rates are so high and that most non-matches are erroneous is that the criteria used to perform electronic matches are overly strict and therefore bound to miss a substantial number of eligible voters.  Matching protocols that do not account for misspelled names, nicknames, variations in hyphenation, transposed numbers, and common typos, for example, will produce large numbers of "false negatives," meaning records that should have matched but did not.  The more exacting the matching criteria, the more "false negatives" a matching system will produce-and, in the context of election administration, the more eligible voters will be harmed as a result.

Despite the EAC's earlier guidance urging matching criteria that account for these kinds of errors, poorly designed and inaccurate matching protocols are still prevalent in the HAVA verification process.  This problem is most acute with respect to the process for verifying applications bearing Social Security digits administered by the American Association of Motor Vehicle Administrators ("AAMVA"). 

As you know, state election officials do not directly interact with the Social Security Administration for the purpose of verifying applications bearing Social Security digits instead of driver's license numbers.  Rather, they send that information to their motor vehicle authorities which, in turn, send that information through a process administered by AAMVA to match that information against Social Security Administration records.  The process used-the Help America Vote Verification (HAVV) process-requires an exact match of last name, first name, month of birth, year of birth, and last four digits of the Social Security number with a record in the Social Security database.  According to information technology experts, these strict match requirements do not readily account for typographical and other errors in voter registration records, and the requirement that multiple fields exactly match compounds the error rate expected for an exact match on any individual field.[16]  In November 2007, the Social Security Administration reported that of 2.3 million voter applications it had processed, nearly half-fully 44.5%-were not successfully matched;[17] record matching expert Andrew Borthwick opined that these largely represent false negatives.[18]   

The extremely high rate of match failures and false negatives renders the HAVV process largely unhelpful for election administration purposes.  Simply put, a system that fails almost half of the time provides administrators with no information about the reliability of the information on their lists.  Those administrators who wish to follow up with un-matched applicants to improve the information on their lists have a difficult time doing so when the numbers are so high.  This also has negative consequences for voters.  In the few states that have "no match, no vote" policies in place, it will result in complete disenfranchisement of the 44.5% of voters who register using Social Security digits and whose information does not match.  In other states, those voters will be unnecessarily subjected to procedures-at the very least, the requirement to show HAVA ID for un-matched first-time voters who register by mail-not required of other voters.

            4.         Use of Social Security Administration Verification Process

In October, 2008, Commissioner of Social Security Michael J. Astrue wrote to the Secretaries of State of Alabama, Georgia, Indiana, Nevada, North Carolina, and Ohio to inquire about the unusually high levels of voter registration verification requests received by the Social Security Administration ("SSA") from those states.  In five of those states, the number of verification requests exceeded the total number of registrants during that period, and in two of those states, the number of requests was more than half of the total number of registrants in the state.  As Commissioner Astrue noted, these figures raised a red flag as to whether those states were using the HAVA verification process for purposes other than those for which it was intended. 

The release of this information allowed election officials and the public to investigate and review voter registration verification procedures in those states.  In some cases, the unusually high volume of verifications was caused by errors by election officials or unnecessary automatic functions in statewide voter registration databases.  In others, the volume might reflect inappropriate uses of the verification process. 

We commend the SSA for proactively monitoring these requests to ensure that states are in full compliance with federal law and their data-sharing agreements with the SSA.  The SSA's public release of this information enabled election officials and others to identify and correct problems with the verification process.  This practice should be continued.

            5.         Recommendations for the EAC

            a.         The EAC should issue clear guidance urging states not to adopt "no match, no vote" policies.

The Brennan Center has previously submitted detailed testimony and letters to the EAC urging the agency to issue a guidance against "no match, no vote" policies and explaining in detail the legal and policy basis for that recommendation.  As noted above, those documents are attached as an appendix to my testimony.  Among other things, those documents explain why "no match, no vote" policies violate HAVA and the Voting Rights Act, and why, even if they do not, they are poor practices that unfairly disenfranchise eligible voters and impede the fair and accurate administration of elections.[19]

While the EAC's initial guidance on databases urged states to take steps to ensure that voter registration applications are "not rejected as unverifiable," it stopped short of recommending that states ensure that no voter registration application is rejected solely on the basis of a failed match.[20]  Because of the persistent problems with the matching process, we strongly recommend that the EAC take that additional step now.  Whether or not states are able to reduce their no-match rates, record matching, standing alone, is simply not a sufficient basis to determine that an application is invalid and that a voter is ineligible.  Most states recognize this fact and do not reject voter registration applications solely on that basis: virtually all states either allow un-matched voters to register and vote without further action or else ask un-matched voters to provide some form of identification at the polls-typically the forms of identification listed in section 303 of HAVA.  But EAC action is still necessary to protect voters in the few states that do impose "no match, no vote" policies as well as to stave off additional efforts-like those seen in Ohio and Wisconsin in 2008-to implement new "new match, no vote" policies.

            b.         The EAC should issue detailed guidance urging states and AAMVA to adopt more flexible and accurate matching criteria and providing technical information to ensure the use of the best known matching protocols.

We commend the EAC for recommending in its initial database guidance flexible matching rules designed to prevent match failures due to typos and other mistakes.  At this point, more detailed recommendations are warranted to help determine and promote the most effective matching protocols to be used with voter registration data.  The guidance should be designed to help states achieve greater accuracy in their matching processes and to reduce false negatives in the verification process.  It could include recommendations for which fields to use, how many characters should match, and how to use the latest and most advanced record matching techniques, for example.  In designing the guidance, the EAC would be well advised to consult with information technology and record matching experts, as well as with states that have already achieved better matching processes.  By bringing to bear research and technical guidance on data sharing, the EAC could not only improve the matching process nationwide but also help election administrators realize the benefits of economies of scale; it is inefficient for this research to be conducted by 50 separate under-funded state administrators across the country.

            c.         The EAC should sponsor a comprehensive study or survey of voter registration verification match rates nationwide.

A study of voter registration verification match rates will not only provide valuable information for policymakers considering how to use the information from matching efforts, it will also help identify which states are achieving better match rates, enabling others to emulate the most successful procedures.

            d.         The EAC should develop guidance for an improved HAVV process for verification of voter registration records containing Social Security digits.

Because the verification of voter registration records with Social Security digits is a key component of the database provisions of HAVA, the HAVV verification process is squarely within the EAC's guidance authority.  For the reasons set forth above, that process is in dire need of improvement.  Specifically, that process should not rely on exact matches of so many fields.  We understand that AAMVA uses far more flexible matching criteria for other purposes, including the verification of driver's license information, and so there is no technical barrier to improving this process.

            e.         The EAC should encourage the Social Security Administration to continue monitoring voter registration verification requests from the states and publicizing information regarding those requests.

As part of its information clearinghouse function, the EAC should work with the Social Security Administration to ensure the continued investigation and public release of information concerning the voter registration verification process with Social Security Administration records.

            f.          The EAC should promote research and issue recommendations to assist states to modernize their voter registration systems.

The matching problems associated with the verification process would not exist if states automatically registered voters at motor vehicle or other agencies, or even if those agencies electronically transmitted voter registration information.  Those records would not need to be matched against records in agency databases since they would come from those agency databases.  This would also substantially reduce the typographical errors that plague the voter rolls and make matching difficult since it would eliminate the need for election officials to decipher paper applications and to type in thousands upon thousands of registration records, often in a short period of time immediately preceding a voter registration deadline.  For a these reasons and others set forth below, we recommend that the EAC promote research and make recommendations to assist the state to move toward more automatic voter registration systems and to modernize their systems in other ways.

B.        Matching Voter Registration Data Across State Lines

Another related area of concern in 2008 was the increasing movement to match voter registration data across state lines.  There is nothing inherently problematic about the goals of interstate database matching-to help state officials detect instances of voters who have moved out of state to enable them to maintain more current and accurate voter rolls.  Nonetheless, if the interstate database matching programs are poorly designed or are used to disenfranchise or impose burdens on voters, then those match programs will cause the same kinds of problems as "no match, no vote" policies used in some inter-agency database matching programs.[21]

Over the last several years, a number of states have implemented data-sharing agreements with other states in an attempt to detect voters who are registered simultaneously in more than one state.  The goal of these programs is to use the new tool of their statewide voter registration databases to help election officials identify and remove citizens form the registration rolls when they move out of state, and protect against the possibility of voters casting ballots in more than one state. 

1.         Risks Associated With Inter-State Database Matching

There are significant risks associated with attempting to cull duplicate registrations based on the results of inter-state data "matches," because of all of the flaws with data matching outlined above.  The errors in this context are "false positives" rather than "false negatives"-that is, the errors occur when two records that apparently "match" in fact refer to two separate individuals.  It is especially difficult to reduce errors in this context.

A 2006 program involving three Southeastern states illustrates the risks.  Early that year, Kentucky, South Carolina and Tennessee conducted an inter-state data match to compare their respective lists of registered voters, and found 16,000 putative matches that they thought represented voters registered in multiple states.[22]  Kentucky determined that about half of these voters were not legitimately registered in that state, and, in April 2006, it purged 8,105 voters from its registration rolls.[23]  Unfortunately, many of these voters were legitimately registered in Kentucky, and after 250 of the cancelled voters showed up to vote in a primary election in May 2006, Kentucky's attorney general sued to have the wrongfully purged voters restored to the rolls.  After finding an error rate of about 10% in Kentucky's data match, and finding that the purge violated Kentucky's NVRA analog, the state ordered that every purged voter be restored to the rolls before the general election.[24]  Kentucky's experience highlights the risks associated with purging voters based on potentially unreliable inter-state data matches.  The error rate is likely to be much higher in other interstate matching contexts, because Kentucky, South Carolina, and Tennessee are among the few states that capture the full Social Security number in their voter registration databases, which improves the reliability of the efforts to identify duplicate records using data matching and should minimize false positives.

Another problematic example of voter purges based on inter-state data matching involved Louisiana.  In 2007, Louisiana's Secretary of State searched the voting rolls of several cities and states-including Colorado, Florida, Georgia, Tennessee, Texas and the counties for the cities of Las Vegas, New York and San Diego-for names that matched registered Louisiana voters.[25]  Based on the results of this investigation, nearly 20,000 voters were purged from the registration rolls because officials concluded they had fled New Orleans and surrounding areas after Hurricane Katrina and registered out of state. The NAACP Legal Defense Fund sued the state, and eventually about 6,700 of the purged voters, mostly from New Orleans, were restored to the rolls.  Louisiana chose to continue its controversial voter purging program in the face of public criticism, although the State Commissioner of Elections Angie LaPlace stated that no voters would be purged based on an interstate data match until after the 2008 federal elections.[26]

2.         Recent Interstate Database Matching Efforts

The need to protect against these risks is particularly important now, since the number of states using cross-state data checks is growing.  To date, in addition to the matching efforts described above involving Louisiana and Kentucky and their partners, the following states have developed matching programs:

  • Southeast: As described above, Kentucky, South Carolina and Tennessee conducted a cross-check of voter registration data in early 2006.[27]
  • Midwest: One of the earliest interstate matching efforts began when the Secretaries of State ("SOS") of Iowa, Kansas, Missouri and Nebraska signed a memorandum of understanding in December 2005.[28] By 2008, South Dakota, North Dakota and Wyoming had joined the group.[29]
  • South Central: Kansas is also part of a group of "south central" states that have created a regional cross-check program. The group includes Kansas, Arizona, Arkansas, Colorado, New Mexico, Oklahoma and Texas.[30]
  • District of Columbia: According to a report of the U.S. Election Assistance Commission, the District of Columbia has conducted cross-checks of voter registration data with Maryland and Virginia.[31]
  • West: In late January 2006, representatives of five western states-California, Idaho, Montana, Oregon and Washington-met to discuss the possibility of creating a regional data sharing agreement.[32] The states determined they were not ready to conduct any data matches, but Washington and Oregon have since developed a pilot program, discussed below, that may serve as a model for states implementing data matching across state lines.

            3.         Recommendation for the EAC

As a result of the significant risks with interstate database matching, it is important that states develop protocols to protect against inaccurate inter-state data matching and to ensure that no purges are conducted as a result without adequate voter protections, including the protections mandated by the NVRA.  The NVRA provides that no voter can be purged from the voter rolls on the ground that they moved unless either (i) the voter confirms in writing that they have changed residences or (ii) the voter fails to respond to a forwardable postcard and then fails to vote or appear to vote in the subsequent two federal elections.[33]

To assist in this process, the EAC should develop guidance to ensure that the list matching process is as accurate as possible, protecting against false positives, and that no purges are conducted without adequate procedures to protect against wrongful disenfranchisement. 

As we understand it, the pilot inter-state data matching protocol developed by Oregon and Washington in advance of the 2008 general election is consistent with federal law and provides an example from which other states can learn when conducting future inter-state database matching efforts.  After attempting to match their voter registration databases and finding more than 8,000 potential matches across both states, Oregon and Washington determined that it was not feasible to attempt any list maintenance procedure with this many voters immediately before the 2008 election.  The states opted, therefore, to conduct a pilot program involving only a limited number of border counties, in which they identified approximately 1,300 voters potentially registered on both sides of the border.[34]  Most significantly, none of these voters were automatically purged.  Instead they were sent notices saying that there was reason to believe they might be registered in both states, and requesting that the voters clarify their status. 

About half of the 1,300 voters had registered more recently in each state, and each state therefore sent notices to the approximately 650 voters with older registrations in that state.  Both Washington and Oregon received responses from about 60% of these voters,  who confirmed that they had moved and requested that their old registrations be cancelled.  The states cancelled these voters' registrations.  Neither state cancelled the registrations of any voters who did not respond; the states will conduct additional investigation in an attempt to gain further information for list maintenance purposes. 

The techniques used by Oregon and Washington-notifying affected voters, allowing the voters to confirm or clarify whether they have moved, and not automatically purging any voter who does not respond-are consistent with federal law (assuming that they do not purge the remaining voters without affording them the NVRA's protections) and could form the basis of EAC recommendations.

C.        Voter Registration Database Maintenance Problems in 2008

Statewide voter registration databases provide election officials with a valuable tool to maintain more accurate voter lists.  Because the list is statewide, officials can more easily track voters when they move to different election jurisdictions within the state.  And because it is computerized, election officials can use electronic processes to search for and identify duplicates or records corresponding to individuals who have become ineligible to vote.  At the same time, because of the relative ease of purging voter records using statewide databases, because database matching is poorly understood, and because such matching can easily lead to significant numbers of "false positives," efforts to maintain, or purge, voter registration databases pose significant risks to registered voters. 

            1.         Purge Problems in 2008 Due to List Matching

The risk of matching-based list maintenance activities, however well intentioned, is illustrated by a purge conducted in Georgia before the 2008 elections.  State election officials attempted to identify non-citizens on the voter rolls by matching the voter registration database against the motor vehicle database, which records the legal residence documents provided in order to obtain driver's licenses, flagging individuals who did not provide documentation of U.S. citizenship when they obtained their driver's licenses.  One problem with this procedure is that the motor vehicle authority does not update its records relating to legal residents when they become naturalized.  The result was a purge that unfairly caught up a significant number of naturalized citizens.  The match process was imperfect in other respects as well; according to news reports, at least some individuals who were born in the United States were affected as well.[35]

The lesson from this-as well as from other purges based on database matches that have come to light in the past[36]-is that list matching is an imperfect tool to identify records pertaining to the same individual or to determine voter ineligibility.  Safeguards are therefore necessary to ensure that eligible voters are not purged from the voter rolls, and likely disenfranchised, as a result of erroneous match attempts.  These safeguards are especially important in light of the ease with which purges can be accomplished on computer-based systems.  In one county in Mississippi, for example, a local election administrator reportedly improperly purged 10,000 voters using her home computer a week before the 2008 Mississippi primary.[37]

The extent to which these kinds of voter database purges led to problems in 2008 is not known.  Purges are typically done without any public notice or scrutiny.  Rarely do state officials make available any information about their list matching activities, let alone information sufficient for others to assess whether those activities harmed eligible voters.  Nonetheless, because there have been significant problems with virtually all of the purge efforts that have come to light in recent years, there is strong reason to believe that the problem is of a magnitude that warrants attention.

The other purge problems that came to light during the 2008 election cycle involved the failure of state officials to comply with the NVRA's purge protections, including the prohibition of systematic purges within ninety days of a federal election, the protections against erroneous purges of suspected movers, and the prohibition of purges based on a registrant's failure to vote.[38]  Lawsuits in Michigan and Colorado prevented voters in those states from being disenfranchised because of these practices.[39]  Because of the lack of public information about purges, we do not know whether and to what extent there were similar problems in other states.  While these problems do not stem from misuse or misunderstanding of voter registration databases, using voter registration databases to modernize the voter registration system could help solve them in the future.

            2.         Recommendations for the EAC

            a.         The EAC should issue guidance to ensure that voter database purging practices are as accurate as possible and include adequate safeguards for eligible voters.

In 2008, the Brennan Center published a report examining in detail the purge practices in twelve states and making recommendations to improve the purge process.[40]  We recommend that the EAC adopt these recommendations-including those for greater transparency and accountability of the purge process, for stricter matching criteria for the development of purge lists, and for fail-safes for voters-as guidance to state election officials.  These recommendations are consistent with those I previously submitted to the EAC in 2005.  If adopted, these recommendations would substantially reduce public controversies over voter database purges. 

            b.         The EAC should collect and disseminate detailed information about list maintenance activities (voter purges) in the states.

One of the greatest challenges to improving list maintenance practices is the dearth of publicly available information about those practices.  There is virtually no publicly available information on when and how purges are conducted, how many people are purged, and on what grounds.  This not only prevents members of the public from being effective watchdogs, it also prevents election officials from effectively learning from their peers' best-and worst-practices.

There are a number of steps the EAC can take to foster better public information about purge practices and their effects.  First, the agency can request more detailed information about purges from state election officials in the reports they submit pursuant to the NVRA.  Second, the agency can expand the Election Day survey to include more detailed information about purges conducted in each election cycle.  Third, the agency can sponsor studies of purge practices and purges in the states.  Fourth, the agency can issue guidance urging state officials to make their purge practices more regular and transparent.

D.        Potential Problems Relating to Maintenance of the SSA Database in 2008

            1.         Regularly scheduled maintenance of SSA database

In fall 2008, the Social Security Administration ("SSA") announced that it planned to shut down its databases for maintenance during Columbus Day weekend, from October 11-13, 2008.  While a SSA database shut down would not normally affect the administration of elections, the timing of this shut down-only three weeks before a presidential election-raised concerns that it would hamper election officials' ability to process voter registration forms on time for the November election.  In 2008, forty-one (41) states had voter registration deadlines between October 6th and October 20th, and voter registration activity typically dramatically increases as those deadlines approach.  As a result, during Columbus Day weekend, especially in a presidential election year, election officials are extremely busy attempting to process the thousands of voter registration applications submitted right before the deadline.  When the SSA database is off-line, election officials cannot process voter registrations bearing Social Security digits, since HAVA requires officials to attempt to match those registrations against the Social Security database before adding the registrants to the rolls.  Since election officials only have a short time to process these applications and do a range of other essential tasks to prepare before the elections, a database shut-down during this period risks impeding election administration.

Despite an outcry from public officials (including Senator Dianne Feinstein[41] and then-EAC Chair Rosemary Rodriguez[42]), election officials, and advocates, the SSA went ahead with its planned shut down over Columbus Day weekend.  While we have not studied the effect of this shut down on election administration, it no doubt caused unnecessary delay during this critical period for election administration.  While we recognize the importance of regular maintenance of the SSA's databases, upon which many government benefits and services depend, and while we understand that the SSA has long scheduled its routine database maintenance on Columbus Day weekend, we urge the SSA to rescheduling its future shutdowns to avoid coinciding with the busy fall months in election years.

            2.         Recommendation for the EAC

The EAC should ask the Social Security Administration to reschedule its regular database maintenance so that it does not take place in the crunch time before the voter registration deadline.

II.        Other Recommendations for the EAC to Assist States in Improving Voter Registration Database Use and Management.

A.        Need for Better Information on Voter Registration Systems

There is surprisingly little publicly available information on how statewide voter registration databases are constructed, maintained, and used, and on how they perform.   This dearth of information prevents policy-makers and members of the public from effectively assessing and improving the voter registration system.  The limited data that are available are in inconsistent formats, making it difficult to compare information across states or jurisdictions.  Given the public importance of the voter registration systems-they are both the gateway to the franchise and the primary tool of election administration-it is essential that we collect and disseminate better information about them.

The EAC can play a significant role in improving this situation both by requiring election officials to collect and report more detailed information and by fostering additional studies on voter registration systems.  In particular:

  • The EAC should require state officials to report more detailed information relating to voter registration, including information on voter registration list management and maintenance, in a standard format.  To assist in assessing performance of the voter registration system, the EAC should also require states to report more detailed information about provisional ballots, breaking down the data based on the reasons provisional ballots were cast and the reasons the ballots were or were not counted.
  • The EAC should collect and publish current information on state voter registration databases and how they are being used to register voters, verify and correct voter information, purge the voter rolls, and manage elections. 
  • The EAC should also collect and publish information on the costs of voter registration systems and each of their components.  For example, the EAC should examine the costs of processing voter registration forms, verifying applicant information, updating address information, and conducting purges of the voter rolls.

B.        Research and Guidance to Modernize the Voter Registration System

Most problems with the voter registration system have nothing to do with statewide voter registration databases, but rather with the outdated paper-based, voter-initiated system on which it is based.  In 2008, as in other election years, prospective voters experienced problems with their registrations because of errors in paperwork; election officials could not read their handwriting; their voter registration forms were not transmitted or not transmitted on time by voter registration agencies or voter registration drives; their forms were lost in the mail; their information was entered incorrectly into the voter registration database; or because election officials were unable to process their registrations on time.  Others did not make it onto the voter rolls because they were in the military and were relocated after the voter registration deadline; they moved and did not know they had to re-register; they were unaware of voter registration deadlines; or they became interested only after the deadline had passed.[43]  Election officials, in turn, had to process the deluge of voter registration forms that typically arrive right before the deadline, in the busiest period of election preparation; hire scores of temporary workers to assist with processing registrations; manually enter voter registration information in the database even though that information had previously been keyed by government workers in other agencies; deal with myriad third-party voter registration drives, some of which made mistakes; try to decipher registrants' poor handwriting so as to enable them to register; and ensure that all the registrations were processed in time to print poll books and notify the voters of their polling places.  This unnecessarily burdensome process continues to make it difficult for election officials to focus sufficient attention on the myriad other essential tasks of election administration, including poll worker training, allocation of election resources, voting system testing, and processing absentee ballot requests, among others.

These problems, as well as the other problems addressed above, could be solved by modernizing the voter registration system-building on the statewide voter registration databases now in place across the country.  Because the voter rolls are computerized, and because they are capable of sharing data with other government databases, it is now possible to move toward a system of automatic registration, in which election officials automatically add eligible voters to the rolls based on information they obtain from other government databases.  And because the databases are statewide, it is now possible to move toward a system of portable or permanent registration within states, in which election officials take steps to keep registrants' addresses up to date and in which voters do not have to re-register or change their registration addresses each time they move.  Of course, no system using automatic processes, no matter how well-designed, is fool-proof, and so these upgrades would have to be accompanied by robust safeguards to ensure that voters can correct errors or omissions in the voter registration process up through Election Day.  These three elements-automatic voter registration, permanent registration, and fail-safe procedures for voters whose information is omitted from or incorrect on the voter rolls-are the foundation of a proposal for voter registration modernization the Brennan Center released in mid-2008.[44]  Each of these elements is already in place at least in part in a number of states,[45] and ought to be expanded nation-wide.

These upgrades would substantially reduce errors in the voter rolls; reduce costs and workload for election officials; reduce duplication and waste; reduce the opportunity for voter fraud; and reduce unnecessary administrative barriers to the franchise.  They would increase the efficiency of election administration and free up essential resources for other election administration tasks. 

Because voter registration modernization makes sense, it garners significant support from election officials of all political stripes.  For example, in a recent op-ed published in Roll Call, Missouri Secretary of State Robin Carnahan and Kentucky Secretary of State Trey Greyson called for "modernization of our voter registration system" using technological innovations.[46]  It also garners support from policy-makers, the media, and the public. On March 11, 2009, the United States Senate Committee on Rules and Administration held a hearing on voter registration problems in 2008, and is likely to continue examining those problems and considering solutions.  State officials and legislators across the country are similarly considering ways they can modernize their systems.

The EAC can play a helpful role in this process.  Most significantly, the EAC can foster data-gathering and research that will assist officials and policy-makers in their efforts to upgrade their voter registration systems.  I address some research questions that can be addressed by the National Academies of Sciences and others in part III.B, below.  The research efforts that would be most helpful are those designed to provide technical assistance to states that wish to use their statewide voter registration databases to facilitate automatic or permanent registration, as well as those that wish to develop and use electronic poll books.  The EAC can also work with entities like AAMVA and the U.S. Postal Service to study the best ways of using their existing systems to facilitate voter registration updates.

III.       Recommendations For the National Academies of Sciences (NAS) Study

A.        The NAS Voter Registration Databases Committee's Interim Report

The interim report published by the NAS Committee on State Voter Registration Databases contains an enormous amount of information, distilled and made comprehensible to policymakers.  The interim report also identified several unresolved issues relating to statewide voter registration databases and highlighted policy disagreements with candor.  We believe the interim report will contribute to informed discussion of issues associated with statewide voter registration databases.

We have two primary concerns with the interim report.  First, there are several instances where the report's language could be interpreted to suggest that HAVA requires a "no match, no vote" policy and that such a policy is in place in most states.  Neither is true.  We have advised the NAS Committee of this concern, and hope that they will clarify the language in question in the final report consistent with our comments.

Second, several of the NAS Committee's conclusions and recommendations are not sufficiently supported by empirical data.  We believe that policy recommendations and conclusions by the NAS Committee should be based on empirical data and not on a small number of interviews or conventional wisdom.  The use of hard data will help avoid unnecessary controversies.  We recognize, however, that the NAS Committee may have difficulty obtaining access to hard data, because election officials either do not collect such data or do not make it publicly available.  We therefore urge the EAC to support the NAS Committee's efforts to obtain data concerning voter registration databases, including by encouraging state officials to collect and publicize such data.

B.        Recommended Expansion of the NAS Study or Additional Studies

Given the heightened interest in voter registration modernization among election officials and other policy-makers, and given the number of technical questions involved in such an effort, it would be extremely helpful if the EAC could commission studies to assist public officials in these efforts.  Since the NAS Committee has already devoted substantial time and effort studying voter registration databases, including database interoperability issues, it is well-positioned to study issues relating to voter registration modernization in an efficient manner.  

We therefore encourage the EAC to commission the NAS Committee either to expand its current study or to conduct a follow-up study on the technical issues relating to voter registration modernization.  We also encourage the EAC to commission studies on this topic from other researchers as well.

Follow-up research on voter registration modernization should focus on the best technical solutions for how to implement automatic registration, whereby eligible citizens on other government lists are automatically added to the voter rolls, and permanent registration, whereby the address information for registered voters is automatically updated based on information on other government lists.  Studies should focus more on technical questions than on policy questions.  The types of research questions that would be helpful include: whether there are technical barriers to sending personal data from particular government lists to the voter registration database and, if so, how best to overcome those barriers; whether the tool used by a majority of state motor vehicle agencies to automatically register young men for the Selective Service could be adapted to automatically register all voting-age citizens to vote; the cost of sharing data relating to voting age individuals from motor vehicle agencies, public assistance agencies, and other agencies with election agencies; whether and to what extent other state databases are or can be made interoperable with the voter database; which public agencies keep the most current address information; and what are the best ways to prevent duplicate records in the voter database if records from multiple state databases are combined into that database.  If the EAC determines that it will commission such studies, the Brennan Center would appreciate an opportunity to submit a more comprehensive list of research questions for consideration.

Thank you very much.