Database problems may disqualify legitimate voters in upcoming U.S. elections

[PHOTO: fred prouser/reuters/corbis]

The Help America Vote Act of 2002, or HAVA, has garnered most of its notoriety because it required election officials throughout the United States to replace old paper-based voting machines with controversial new electronic equipment by 2004. But there are other provisions in the law that took effect only in January 2006, and these are quietly creating their own potential for disrupting elections this November—including the 468 House and Senate contests that will determine control of Congress.

The new HAVA rules concern the databases that contain the voter rolls—and in 49 of the 50 states, if you are not on the rolls, you can’t vote. Elections have often turned on the question of who gets to vote and who does not. This time around, voter eligibility will depend in large part on the contents of a number of databases, most of which have been in existence for less than a year and some of which have not been constructed in accord with the best practices of the database industry.

The new HAVA rules sound, at first, fairly innocuous. They require each of the 50 states to maintain a statewide database of registered voters, or create one if none exists. They also call for the states to verify a voter’s registration record either with the state’s driver’s license records or with information in a federal database of all legal residents of the United States, which is maintained by the Social Security Administration.

HAVA gives the states wide latitude in reacting to database mismatches. If a voter’s registration information and the data in the other databases differ, the act does not require that a state keep registrants off the rolls. But it also doesn’t forbid the states to do so. State officials have therefore set up their own rules, which have resulted in mass purges of registrants in California, Iowa, South Dakota, Texas, and Washington, according to the Brennan Center for Justice at the New York University School of Law, in New York City.

In March, the center published an exhaustive study of the varying ways the states have responded to HAVA’s database requirements, a 300-page document titled “Making the List: Database Matching and Verification Processes for Voter Registration.” Tens of thousands of people could be disqualified in California alone, the nation’s most populous state.

The new HAVA-required databases hold the names of all registered voters, but the mismatch problem mainly concerns new voters, those who have relocated, and those who have changed their name. After an initial verification of these new and changed registrations in Los Angeles County, where one out of every four voters in the state lives, 43 percent were purged from the rolls, according to Wendy Weiser, deputy director of the Democracy Program at the Brennan Center. Statewide, about 26 percent of voter registration applications were purged. The state processed more than 6 million registrations between 2002 and 2004, the most recent years for which statistics are available.

Weiser says there are some simple reasons for the high number of mismatches, and it is not out of line with the results of any comparison of two large databases of names. They can occur when the voter omits, say, a middle initial from one list but not the other, or if a hyphen in a last name gets turned into a space, or if an apostrophe gets dropped and “O’Brien” becomes “OBrien” or “O Brien.” [See table, “The Match Game.”] When the Social Security number, which is usually assigned at birth, is used instead of the driver’s license number, mismatches are likely for women who change their surnames when they marry.

Then there’s the problem of simple database mistakes. A study by the Board of Elections in the City of New York compared 15 000 new voter registrations with the New York state motor vehicle database. It found that 20 percent of the records failed to match because of typographical errors made by election officials. Weiser says that database experts have, over the years, devised automated techniques for avoiding erroneous mismatches, such as comparing phonetic versions of names, which would solve the “O’Brien” problem, and using lists of name equivalencies, such as “John” and “Jon” or “Sarah” and “Sara.” But, she notes, many of the states didn’t employ these more sophisticated methods, including tech-savvy California, home of Silicon Valley.

California had some quirky rules that contributed to the state’s unusually high number of purged registrants, and those rules have since been tightened. Nevertheless, as of early August, 10 or 11 percent of California’s registrants were still not eligible to vote in the November election, Weiser was informally told by state officials.

The state’s rules could change again, though, because of a court case farther up the Pacific Coast. On 1 August, U.S. District Judge Ricardo Martinez, in Seattle, issued a preliminary injunction stopping Washington state election officials from purging from the rolls any voters whose election records don’t pass the state’s verification process. Although the Seattle ruling had no authority outside Washington, voting activists in California took it as a victory. California’s method of matching the voter rolls to motor vehicle and Social Security records is “very similar to the Washington state law the court put on hold,” says California state Senator Debra Bowen, a candidate in the upcoming election for secretary of state.

The confusion in California and elsewhere stems from the fact that the law requires that states “verify” the voter rolls but not that they necessarily do anything about registrants whose names or addresses can’t be verified. The tendency toward purging the rolls may have been helped along, in part, by the U.S. Department of Justice’s Civil Rights Division, the unit charged with enforcing federal election laws. Back in 2003, one of the division’s lawyers wrote a widely circulated, nonbinding opinion to officials in Maryland, saying that if the information in a registration application doesn’t match motor vehicle or Social Security records, the applicant shouldn’t be added to the voter rolls. “Congress obviously intended,” the letter says, “that…where the results indicate the registrant is not eligible, has provided inaccurate or fraudulent information, or information that cannot be verified, then the application must be denied.”

Despite the opinion, Maryland officials did not conduct mismatch purges similar to the ones in Washington and California. The necessary qualifications to be a voter, the Brennan Center report notes, have nothing to do with driver’s licenses or even Social Security records, and come down to a few simple questions. Are you 18 years old and a U.S. citizen? Are you mentally competent? And for the states that disqualify such individuals: Are you not in prison or on parole? If the answer is yes to all those questions, you’re qualified to vote.

Confusing or not, the new HAVA rules are meant to solve a real problem. Almost all elections in the United States are run by either a handful of large cities or by the nation’s 3141 counties, which have for decades maintained voter rolls without any coordination at the state level. Tracking voters when they move or die is basically impossible at the county level. Yet, despite earlier attempts by the U.S. Congress to get the 50 state governments more involved, as recently as 2000 only one state, Michigan, had a statewide database that culled duplicate records and checked them against other statewide databases.

HAVA’s answer to the problem is well illustrated in Indiana. In November 2005, the state government took voter records from Indiana’s 92 counties and linked them for the first time into a single database. It vetted the list against databases at the Bureau of Motor Vehicles (for addresses), the Department of Correction (for the names of convicted felons, who are not allowed to vote in Indiana), and the Department of Health (for identification of deceased voters). About 29 000 records were identified as probably referring to deceased voters, and 290 000 duplicate records were found.

A local company, Quest Information Systems, in Indianapolis, wrote the software for the unified database. Using a Web interface, each county can now see the voter records of other counties in the database and can update its own rolls. The database itself is stored by yet another company, Data Return, in Irving, Texas.

To determine whether a new registrant matches, say, a motor vehicle database record, the Indiana system calculates what it calls a “confidence factor.” Every matching field is assigned a point value. “The higher the point total, the higher the confidence factor,” says Scott Lee, a software engineer at Quest. “A matching last name or first name is worth 15 points each, a middle name only 5,” he says. Current and previous street addresses are each worth 10 points. Only exact matches are counted. “We only show matches above a certain point total, 50 or 55,” he says. “The final acceptance of a match is up to the county.”

That final say reflects a fundamental tension when it comes to voter eligibility, as does HAVA itself. If the law makes it too easy to register and vote, it becomes easy to cast multiple ballots, vote on behalf of someone who is deceased, or commit other voter fraud. Making it too hard to register, say by requiring people to register in person, however, disenfranchises people.

Today, the most common form of voter disenfranchisement involves laws relating to prisoners who are felons. Only two states, Maine and Vermont, allow felons to vote; only 14 allow those on probation or parole to vote; and 11 states have a lifetime ban on voting by ex-felons, according to Washington, D.C.–based The Sentencing Project.

Disenfranchisement by database made a prominent mark on Florida’s riotously controversial 2000 presidential election. The state hired an Alpharetta, Ga., company, ChoicePoint, to compare its voter rolls against a number of county felony databases within Florida and state databases from as far away as Texas. But at least 3000 ex-felons were purged from the rolls erroneously, in violation of court guidelines, according to an investigative report by journalist Greg Palast, first published in Britain’s left-liberal newspaper The Observer.

Rebecca Mercuri, a computer scientist and expert on voting technologies, notes that statewide databases, many of which didn’t exist before HAVA, have made disenfranchisement easier, both in Florida and more generally in 2000. “There’s nothing new about disenfranchisement—it’s just now done electronically,” says Mercuri.

The prospects for a chaotic election in November have diminished somewhat, according to Weiser of the Brennan Center. For example, in June, New York state passed a law that lets residents with felony convictions know of their voter eligibility status. That’s progress, Weiser says, because only if you know you are disenfranchised can you contest that status. But voter registrations are still in turmoil in a number of jurisdictions. As late as August, an ongoing squabble in Kentucky between the secretary of state and the attorney general over the status of 8105 names threatened to erupt into a lawsuit.

The states Weiser is most concerned about are the ones that are purging any and all mismatches from the voter rolls. She expects mismatched voters to be restored in Washington state because of the Seattle court case. California, too, will likely alter its practices yet again before November.

Already, in the wake of the Washington court ruling, the Pennsylvania Department of State stopped automatically rejecting database mismatches, saying that its own rules on the matter are wrong. “Neither HAVA nor Pennsylvania law requires that result,” it says. “The Department has concluded that its procedures actually are frustrating the principal purpose and intent of HAVA to ensure that eligible persons are not disenfranchised.”