Policy Brief on Inaccurate Purges of the Voter Rolls

September 12, 2006



* Ill-conceived state policies concerning new statewide voter registration databases are driving eligible voters off of the rolls, through no fault of their own.
* Purging based on database matching can be unreliable.
* Purges can also be discriminatory.
* Purging requires matching standards appropriate for the task - and adequate procedural safeguards in addition to matching.
* Legitimate voters also deserve common-sense technological protection.

Ill-conceived state policies concerning new statewide voter registration databases are driving eligible voters off of the rolls, through no fault of their own. Across the country, states are implementing the Help America Vote Act of 2002 ("HAVA"), creating new statewide computerized databases of registered voters. The technology has the potential to improve the registration process substantially. However, it also has the potential to be quite dangerous - in particular, when officials unduly rely on the ability to “match” information from one source to another. As states maintain their voter rolls, most will attempt to match voter registration records to lists of individuals who may be ineligible. These may be lists of deceased citizens, persons with disqualifying felony convictions, voters in other states, and the like. Often, however, an ostensibly “matching” record only tells a portion of the relevant story - and leads to eligible voters being20purged illegally.

Purging based on database matching can be unreliable. Many attempts to match information will yield predictable errors. The infamous Florida purges of 2000 were caused in part by bad matching standards. Any Florida voter was purged from the rolls if his name shared 80 percent of the letters of a name in a nationwide felon database; a California felon named John Michaelson would cause an eligible Floridian named John Michaels to be disenfranchised. These purges were wildly inaccurate: over half of the purged voters who appealed their removal from the rolls were deemed eligible. A 2005 attempt to find duplicate entries on the New Jersey rolls was similarly flawed; in seeking duplicates, it ignored middle names and suffixes, capturing eligible voters like “J.T. Kearns, Jr.” and “J.T. Kearns, Sr.”

Even when names and birthdates match precisely across lists, it is premature to assume that an individual is registered twice and therefore subject to being purged. Elementary statistics students are often surprised to learn that among just 23 individuals, it is more likely than not that two will share a birthday. Similar statistics show that for most reasonably common names, it is extremely likely that at least two people with the same name will share the same date of birth. Such purported “matches” may not represent the same person20at all.

Finally, in addition to the dangers above, even a technically perfect “match” may not yield enough information about the voter’s eligibility to ensure that a purge is reliable. Potential purges must always be reviewed for mistakes due to common clerical errors, incomplete information, or faulty assumptions. Ostensibly duplicate entries may in fact be explained by a data entry error. A person with an ostensibly disenfranchising conviction may in fact have had her voting rights legally restored. An individual ostensibly living at an invalid business address may in fact be living at an unusual - but legitimate - residence.

Purges can also be discriminatory. Many purges occur out of the public spotlight. In such conditions, criteria can be selected that cause predictable disparities among the voters removed from the rolls. In 2004, for example, Florida attempted to match its voter rolls to lists of persons disqualified due to conviction. This time, the state apparently tightened its criteria, requiring more exacting matches of the information on both lists - including the race of the voter. However, because Florida voting records recognized Hispanic individuals as a distinct racial category but Florida corrections records did not, only a handful of records of Hispanic voters were “matched” based on race. The result: approximately 22,000 African-American voters were slated to be purged, compared to approximately 63 Hispanic voters. In addition to the blatant racial disparity, given the dominant partisan voting patterns of African-Americans and Hispanics in Florida, the perceived electoral impact of the purge generated widespread outrage.

Purging requires matching standards appropriate for the task - and adequate procedural safeguards in addition to matching. Because potential purges suffer from the predictable problems above, they must be conducted pursuant to transparent procedures that protect eligible voters. This means that a voter should be flagged as ineligible or removed from the rolls only when the state is certain that the voter in question is the voter whose eligibility is suspect - and then only pursuant to the procedures laid out in the National Voter Registration Act ("NVRA"). There should be redundant checks before a record is determined to be a duplicate or to belong to an ineligible voter; voters should be notified, with an opportunity to correct mistakes before any removal occurs; and most purges should be prohibited within 90 days of an election.

Legitimate voters also deserve common-sense technological protection. Many states are currently constructing for the first time the large statewide voter registration databases that facilitate both registration and purging. In addition to comprising the official list of voters - and thereby determining whether any individual is ultimately able to vote - the databases will also contain a substantial amount of private personal information. States must therefore implement common-sense technological protections for these enormous systems, such as requiring that a log of all database transactions be maintained, in order to track and remedy improper access. These databases must also be protected by layers of access and authorization, to ensure that only authorized transactions are made and only by authorized people.


* Florida. In 2004, the Brennan Center discovered and drew national attention to Florida’s flawed attempt to purge the voter rolls using error-laden and discriminatory lists. The extensive publicity, fueled by original Brennan Center research, caused the state to scrap its plans to purge many thousands of eligible voters.

* New Jersey. In 2005, a list of purported fraudulent votes and duplicate registration entries was delivered to the state Attorney General, with a demand that the voter rolls be purged. Together with a prominent political scientist, the Brennan Center demonstrated the flaws in the matching process used to generate the list, and proved that the vast majority of allegations in fact yielded no reason for concern.

* Kentucky. In 2006, the Kentucky State Board of Elections attempted to match names on its registration database against lists of voters in Tennessee and South Carolina, and purged 8,000 voters as a result of the match - without notifying the voters, and in violation of specific provisions of federal law. Based in part on Brennan Center research, the Kentucky Attorney General filed a lawsuit to undo the unlawful purge; the case is still pending in the state trial court.

Last Updated September 2006