Dead Voters? Probe Finds Errors In Records

By MARCEL DUFRESNE

Special to The Courant

April 20, 2008

Jane M. Drury, homemaker, mother, grandmother and great-grandmother, = spent=20 most of her life in Staten Island, N.Y, but moved to Stonington in 1992 = to live=20 with her daughter. That October, she registered to vote = there.

State=20 records show Drury never actually voted in Stonington before her death = in 2000=20 in a Groton nursing home at age 97. But the same records say she did = vote in=20 Stonington once since then =97 in a 2007 budget referendum.

That = recorded=20 vote, seven years after her death, puts Drury on a list of more than 300 = people=20 across Connecticut who appear to have voted from the grave in elections = dating=20 to 1994, a two-month investigation of voting records by journalism = students at=20 the University of Connecticut has found.

The mysterious voters = were=20 identified by matching a statewide database of 2 million registered = voters and=20 their voting histories with two separate computer lists of dead people=20 maintained by the state Department of Public Health and the federal = Social=20 Security Administration.

Following up on the matches, UConn = students=20 examined the records of nearly 100 of the suspect voters at 10 town and = city=20 halls among those with the most cases. Guilford led the state with 39, = followed=20 by West Hartford (17), Enfield (15), Stonington (13) and Norwalk=20 (11).

Some people appeared to have voted frequently after death, = the=20 research found. In Hebron, for example, records show one man voted 17 = times=20 after he died in 1992.

The investigation also identified more = than 8,500=20 people listed as dead who are still registered to vote in Connecticut, = most long=20 after their deaths. In Hamden, one woman remained a registered voter = although=20 she died in 1979.

Although the investigation found no evidence of = deliberate fraud, it uncovered numerous errors in voting and = registration=20 records kept by local registrars, and highlights weaknesses in a state = voter=20 database established in the mid-1990s, ostensibly to prevent fraud and = duplicate=20 voting.

State elections officials and others concede that the = findings=20 reveal flaws in the voting system and expose its vulnerability to = potential=20 abuse.

"I understood [dead voters] as a myth," said Joan M. = Andrews,=20 enforcement director for the State Elections Enforcement Commission, = which=20 investigates elections and campaign finance irregularities. "People make = jokes=20 about dead people voting but we never thought things like that could = actually=20 happen."

Andrews said she plans to bring the UConn findings to = the full=20 commission to determine whether further investigation is = warranted.

Trouble In Town Hall

Michael Kozik, managing = attorney for=20 the secretary of the state's office, which is responsible for the state=20 database, said he was most troubled by the thousands of dead people = still=20 registered to vote. "Clearly this is an issue that we need to look at," = he=20 said.

All but nine of Connecticut's 169 municipalities listed = dead people=20 on active voter rolls. At least 100 cases were identified in each of 28 = cities=20 and towns. In New Haven alone, 370 dead people were still registered; in = Enfield=20 there were 321; in West Haven, 310; in Hartford, 298; and in Bridgeport, = 293.

The UConn investigation also found instances of people who = were not=20 registered yet were able to vote =97 under a dead person's name =97 = along with=20 frequent discrepancies between the state data and paper voting records = at local=20 town halls.

The potential for fraud is most serious in municipal=20 elections, when officials are not required to check photo identification = at the=20 polls and voters need only sign a statement attesting to their=20 identity.

The UConn findings highlight two nagging problems = facing=20 officials trying to keep accurate voting records. First is the informal = and=20 antiquated ways many local registrars remove dead people from voter = rolls.=20 Second is the potential for clerical errors when registrars transfer = information=20 from paper voting lists to the state database, especially in the rush to = update=20 by the two-day, post-election deadline.

Many of those interviewed = blame a=20 locally based election system that has not changed much in decades. In = many=20 towns, elections and voter registration are run by part-time registrars = who are=20 paid little, have limited resources, and may resist the growing role of=20 computers in elections.

The UConn findings are disturbing but not = surprising, said Andy Sauer, executive director of Connecticut Common = Cause, a=20 public-interest group that lobbies for election reform. He blamed many = of the=20 system's problems on politically appointed local registrars using = outdated=20 methods of running elections and keeping records. Each town has two = registrars,=20 one Democrat and one Republican.

"You need to take the election=20 administration out of the hands of a politician and into the hands of a = state=20 administration," he said, a solution many in the system might consider=20 unrealistic.

Figuring Out Who Died

Jane Drury's vote in a May = 2007=20 Stonington referendum, first detected in the state data, was later = confirmed on=20 a paper check-off list at the town hall showing her name crossed off = with a=20 straight line. By all official records, Drury voted seven years after = she had=20 died.

"Somebody goofed, obviously," said Jane Gumpel, Drury's = daughter.=20 "I don't know how that could happen."

Louise Brown, a Stonington=20 registrar of voters, was one of several municipal registrars who = initially=20 expressed skepticism when informed that dead people appeared to be = voting in=20 their town. Brown was shown a list of 13 dead people who state records = indicated=20 had voted a total of 38 times in Stonington since 2001. A quick check of = several=20 names confirmed the problems, and Brown began tracking down several = suspicious=20 cases. Registrars in other towns responded similarly.

Around the = state,=20 UConn students identified a variety of explanations for why people = seemed to be=20 voting after death. The most common was clerical errors made while = checking off=20 voters' names at the polls or entering votes into the state = database.

In=20 other cases, someone with the same name, sometimes a relative who was = not=20 registered, actually cast the vote. Other times, a living person might = have been=20 mistakenly taken off the rolls when someone with the same name died. = Votes in=20 both scenarios were erroneously credited to a dead = person.

Connecticut is=20 one of many states with no formal system for notifying registrars when = someone=20 dies. Death certificates are sent to the city or town where the death = occurs,=20 not necessarily where people lived, even if they lived in the same town = all=20 their lives. When someone dies in a hospital or nursing home in another = town or=20 an adjacent state, the death certificate may never come to the = registrars'=20 attention.

Registrars tend to rely on newspaper obituaries, = relatives, or=20 word of mouth to learn of someone's death, many officials = said.

Even=20 then, registrars are advised to be extremely cautious about purging dead = people=20 from voter rolls, said George Cody, president of the Registrar of Voters = Association of Connecticut and New Canaan's Democratic registrar. They = must see=20 documentary proof, such as the death certificate or a published = obituary, before=20 removing someone's name.

A Confusion Of Names

The possibility that two or = more=20 people in the same town have the same name makes registrars especially = cautious=20 about removing a voter, Cody said.

Identical names seem to = explain the=20 case of George E. Gagnon, a retired auto body repairman who, records = indicate,=20 voted seven times in East Hartford after his death on June 15, 1991, = with all of=20 the votes coming after 2001. Gagnon, born in 1918, had moved to = neighboring=20 Manchester in 1952, long before he died.

It turns out that a = different=20 George E. Gagnon, a retired truck driver born in 1921, is very much = alive,=20 living and voting regularly in East Hartford.

"A lot of dead = people vote=20 in this state," the living Gagnon joked after finding out a different = George=20 Gagnon had been credited with his votes. "I'm still alive and if = somebody makes=20 a mistake, that's not my fault."

The explanation: Someone at the = East=20 Hartford registrars office entered the living Gagnon into the state = database=20 using the birth date on the other Gagnon's registration card, even = though the=20 latter had left town 50 years earlier. As a result, the living Gagnon = was=20 identified as dead in the computer match and the dead Gagnon appeared to = have=20 cast multiple votes in East Hartford.

While registrars are = reluctant to=20 cancel a voter's registration without physical proof of death, they are = under no=20 obligation to seek out such proof, even when it might be filed elsewhere = in=20 their own town hall. UConn students found numerous cases in which death = and=20 burial certificates on file in one town office did not find their way to = registrars in the same building.

State records show that Carmella = Vella,=20 an Italian immigrant who settled in Enfield in 1913, voted there in 2005 = and=20 2006. But Vella died across the state border in a Springfield hospital = in 1998.=20 The votes, it turned out, stem from clerical errors.

Neighboring = towns=20 are supposed to forward death certificates to a dead person's hometown, = but that=20 doesn't always happen, said Enfield's recently retired Republican = registrar,=20 Vaughan Vanderscoff.

But proof of Vella's death =97 a burial = certificate=20 for an Enfield cemetery =97 is on file in the town clerk's office down = the hall=20 from the registrar.

Vella's daughter-in-law, Barbara Vella, said = she=20 doesn't understand why her husband's mother was still = registered.

"I've=20 told them multiple times to take her off the list," Vella said. "She is=20 certainly deceased."

Kozik, of the secretary of the state's = office,=20 concedes: "There should be a better system of getting the death = information to=20 the registrars in a timely fashion. But I'm not exactly sure how we = would do=20 that."

How Mistakes Are Made

In theory, at least, there = is a=20 simple way to supply registrars with accurate death information. The = state=20 Office of Vital Records maintains a database of death certificates in=20 Connecticut =97 called the "consolidated master death file" =97 dating = to 1949. It=20 is the same database used by UConn students to identify votes attributed = to dead=20 people.

A 2005 report to Congress by the Government = Accountability Office=20 recommended that states share computerized death certificate data with = towns to=20 accurately purge dead people from voter rolls. Georgia began doing so in = 2003,=20 after a newspaper investigation revealed some 15,000 dead people still=20 registered in that state.

But Kozik, who was not familiar with = the GAO=20 recommendation, cautioned against using computer matching alone to = remove=20 voters' names.

"I would always be reluctant to have computers=20 automatically disenfranchising people to vote," Kozik said. "Before = someone is=20 actually taken off the list I would want some kind of human intervention = and=20 judgment involved."

Many votes credited to dead people appeared = to result=20 from clerical error, UConn students found.

In Southington, the = state data=20 show that 10 dead people voted, most in November 2000. But a check of = paper=20 records showed none actually had voted.

"The errors had to occur = when we=20 were typing the names of people who voted into the computer," said Carol = Sheffs,=20 deputy registrar in Southington. "Everyone is in a rush to see who = voted. You do=20 it fast and you don't have time to check over your work. That's how = mistakes are=20 made."

Local officials are well aware of the system's weaknesses, = said=20 Carole Young-Kleinfeld, a deputy registrar in Wilton and a spokeswoman = for the=20 Connecticut League of Women Voters. "The possibility of human error at = election=20 time is very real. Checkers at the polls mistakenly cross off sons = instead of=20 fathers, and vice versa, or cross off neighbors who have similar=20 names."

Kozik downplayed the significance of voter history = errors. Those=20 records are kept mostly as a convenience to political parties, which use = the=20 information to target frequent voters, he said.

But Andrews, of = the=20 elections enforcement commission, said a state record that shows who = voted and=20 who did not should be accurate. Especially troubling, she said, are = those dead=20 people whose votes are confirmed by check-off lists, which can only = result from=20 multiple errors in the process.

She described most problems as=20 "systematic" and blamed them on "a lack of uniformity" at the local = level.=20 Municipalities are supposed to follow the same state-mandated = procedures, she=20 said, "but when you go out to a town, you find that they = aren't."

Forced To Go Online

UConn students found 10 = cases in which=20 an unregistered person was able to vote under a dead person's name. In = most=20 cases, the person was a son or daughter with the same first, last and = middle=20 name, and sometimes even the same address, but not the same date of=20 birth.

One such case involved Donald R. Tichy Jr., a 1981 = graduate of the=20 University of Hartford, who died of cancer in 1992 at the age of 33. = Votes=20 attributed to Tichy in 2005 and 2006 were actually cast by his father, = Donald R.=20 Tichy Sr. The only registration card under the name of Donald Tichy = lists the=20 son's identification information, plus the suffix "Jr." added, in = pencil, after=20 the name.

"I've been voting illegally, is what you're telling = me," a=20 surprised Tichy Sr. said. He vowed to straighten out the = matter.

The=20 secretary of the state's office created the voter database in the = mid-1990s to=20 help detect fraud and duplicate registrations, years ahead of the 2002 = federal=20 Help America Vote Act that required every state to keep such a=20 database.

Participation by local registrars was voluntary at = first =97 some=20 towns resisted the move to computerization =97 but the state legislature = required=20 every town to go online by September 2003.

Calling the database = "a work=20 in progress," Andrews noted that some towns had to be forced to comply. = In=20 Norwalk, two registrars were fined $1,500 each for failing to complete = their=20 portion of the registry on time and for not using it in the November = 2004=20 election.

State officials say they have known of problems with = the=20 database since 2005, when the elections enforcement commission tested = the=20 system, looking for people who might have voted twice in the 2004 = national=20 election, Andrews said.

The agency initially was surprised to = find more=20 than 400 people who apparently voted more than once. But after spending = more=20 than a year investigating each case, the agency found no evidence of = duplicate=20 voting.

"One by one, there turned out to be some sort of = explanation for=20 them." Andrews said. The investigation found many of the same causes = identified=20 in the UConn investigation.

Some officials suggest that better = training=20 for registrars would improve the system. The secretary of the state's = office and=20 the registrars association have been conducting "a very intensive = training=20 program" for registrars for nearly two years, especially those newly = elected,=20 said Cody, the registrars association president.

But he conceded = that=20 some registrars, especially those in smaller towns, whom he described as = "woefully underpaid and understaffed," may be slow to adapt.

"We = can lead=20 them to water," Cody said, "but we can't make them drink."

Courant.com