Election study finds widespread ballot-counting problems

By THOMAS HARGROVE
Scripps Howard News Service
December 20, 2004

- A review of election results in a 10-county sampling revealed more than 12,000 ballots that failed to record a vote for president, almost one in every 10 ballots cast. The unofficial audit by Scripps Howard News Service uncovered malfunctioning voting machines, improperly designed ballots and poor accounting procedures around the nation.

The review of certified election returns led authorities to restore 662 votes for president in Louisiana and West Virginia that had been miscounted in easily detectable errors made by local officials.

But most of the ballots discovered missing in the study will remain lost.

"I'm so upset over this that I can't sleep," said Sandy Campbell, clerk of Pike County, Ark., upon learning that a damaged optical scanning machine permanently lost nearly 700 votes. "We had no idea this had happened. But I'll know what to look for in the future. We'll try never to let this happen again."

The study - part of a yearlong project examining errors in America's election practices - checked the accuracy of the Nov. 2 election by comparing official results for president against the reported number of ballots cast in more than 2,400 counties nationwide.

Ten counties with some of the nation's worst voting record discrepancies were ed in the project. Local election officials were asked why their vote tallies didn't match their ballot counts.

All but one county official admitted they did not make this important crosscheck before reporting results that, most now concede, contained significant errors. State officials also failed to notice the discrepancies.

"This speaks to the need for states to be more thorough," said DeForest Soaries, chairman of the U.S. Election Assistance Commission, after seeing results of the Scripps Howard study. "Every state is accountable to ensure their procedures and practices deliver election results with integrity."

Four of the 10 counties contacted in this study have amended their official vote certifications.

"We're so glad we found this. I don't know how it happened. Thank you so much for calling us," said Registrar of Voters Bettye Moore in Madison Parish, La., after finding errors that misplaced 419 votes for Democrat John Kerry and 15 lost votes for President Bush.

Richard Kirby, clerk for Calhoun County in rural West Virginia, said he accidentally erased 229 votes for Bush. He had earlier indicated the incumbent had received just one vote.

"I messed up when transferring the data (onto official reporting forms). It was so obvious that it should have been caught but it wasn't," Kirby said. "I will send a revised certification report to the secretary of state. In the future, I'll just check myself better and have somebody else check my work."

Two counties in Indiana made significant mistakes when counting the number of voters who went to the polls last month.

"I can't believe I did this!" said Jackson County Clerk Sarah Benter upon discovering a 1,664-vote error in her ballot count. "I'm going to have to get with our state Election Division to get this fixed."

The absentee vote was accidentally counted twice in White County, Ind., causing a 992-vote error in the ballot count. "You are just way too observant for me," Election Clerk Mary Reid said. "I need to get this fixed right now."

Other problems uncovered in this study found much larger numbers of miscounted votes that cannot be corrected or recovered. The largest discrepancy occurred in Carteret County, N.C., where a programming error in a 10-year-old touch-screen voting machine erased 4,439 ballots.

"It was the most elementary error," said Edward Pond, chairman of the Carteret County Elections Board. "The machine was set so that it had a vote capacity of only 3,005 votes. If I hadn't been there watching the tally, I don't know when we would have noticed this."

Pond said his board plans to "send a letter of apology to each of these voters" and to invite them to a new election Jan. 11 to vote for North Carolina's agriculture commissioner, which is still undecided because of the error.

The Scripps Howard study found a significant discrepancy in the vote count in Burke County, N.C., which also uses dated touch-screen machines. Burke County voters cast 34,604 ballots, but only 30,762 votes for president were recorded. Less than 89 percent of the voters recorded a presidential preference.

Election experts say any time 2 percent or more of ballots cast in a major race fail to record a vote, those numbers should be investigated.

"These are the figures that I have," said Greer Suttlemyer, director of elections in Burke County.

Scripps Howard notified the North Carolina State Board of Elections of the discrepancy, and an audit team was sent to the county last week under the direction of the board's general counsel, Don Wright.

The attorney concluded the electronic ballot was poorly designed. It displayed the option to vote for president on the screen where voters could the "straight-party voting option."

That design is particularly confusing, Wright said, because under North Carolina law a straight-party vote to all candidates from one party does not include the vote for president.

"I can only make the assumption that voters thought the straight-party option would include the other office that was on the same screen," Wright said.

Voting machines again came under suspicion in two Arkansas counties where optical scanners apparently falsely reported that hundreds of voters ed two or more presidential candidates, a disqualifying error called "overvoting."

"Basically, we had a mechanical failure," said Pike County Clerk Campbell. The scanner disqualified 692 of the county's 4,083 voters - more than one in six voters. The machine also disqualified 433 votes in the U.S. Senate race for the same reason.

"We sent a technician to Pike County to check," said Meghan McCormick, spokeswoman for Omaha-based Election Systems and Software Inc., which manufactured Pike County's optical scan machine. "There was a scratch on Sensor 'A' that has already been repaired. The scratch probably occurred during the election, but we'll never be certain."

A similar optical scanner in Crittenden County, Ark., reported that 1,853 of the county's 17,284 voters had ed more than one presidential candidate. Another 131 ballots were counted as having no checkmarks for president.

About one in every eight ballots cast in Crittenden County failed to register a choice for president.

"I'm surprised by this. But that ballot was very confusing," said Scott Ferguson, a full-time radiologist and part-time chairman of the Crittenden County Election Commission. "I think it was a grouping problem."

Crittenden County's presidential ballot contained several unusual design elements, including presidential slates that provided ions next to both presidential and vice presidential candidates.

Ferguson admitted,"I'm not sure what would happen" if voters darkened the oval next to both President Bush's name and the oval next to his running mate, Vice President Dick Cheney.

"The vice presidential bubbles were not programmed," said Lou Ann Driver, the county's data processing technician who wrote the tabulation routines for the scanner. "If people voted for Bush and Cheney, well, that wasn't supposed to count as an overvote."

Driver said county workers tried to put the presidential and vice presidential candidates on one line, but the names were too long to fit in the available space.

Regardless of the cause of overvotes in Crittenden County, there was no justification for conducting a recount, according to Tim Humphries, staff counsel to Arkansas Secretary of State Charlie Daniels.

"There is no authority under the Arkansas Code for a recount after the election has been certified," Humphries said. "The law says if there is to be a recount, it must happen before then."

He said Arkansas state employees "do not compare the presidential vote against the total vote. Hopefully, the county election commissions do that."

But most of the local election officials contacted for this story - and in dozens of other interviews conducted throughout the year - said they were unaware of discrepancies between vote tallies and the number of ballots cast. Instead, county officials often claim large discrepancies are the deliberate choice of many voters to ignore major races like president or governor.

"Most of them just didn't vote," Kirby said. "It's weird but it happens." But he double checked the vote in his West Virginia county and called back a day later to confess: "It was an error. I know, because I made the error."

Officials in two of the 10 counties ed in this study refused to review ballots or check for missing votes.

"I don't have time to go through all of this, although it does sound like it ought to be gotten to the bottom of," said County Clerk John Jones of Worth County, Mo.

Only 1,132 of his 1,326 ballots registered a presidential vote, which means nearly 15 percent of Worth County voters appear to have ignored the presidential race. "Sir, I cannot record votes for president that are not there," Jones said later.

Only 2,063 of the 2,283 ballots cast in Adams County, Idaho, registered a vote for president. Officially, only 90 percent of the voters in this rural county had an opinion in the presidential race.

"If you can convince the secretary of state of Idaho to order a recount and to reimburse my county for it, then we'll do a recount. But I'm not going to do it," said Adams County Clerk Michael Fisk. "All of that stuff has been put away. What difference does it make? The election is over."

Fisk said Scripps Howard's request that he review he election tallies angered him.

"You are not going to assert that any votes are missing in Adams County. I'm not going to listen to that stuff. Goodbye," With that, Fisk hung up the telephone.

Back in Washington in his commission office four blocks from the White House, Soaries said it's important for election reform advocates to understand the severe limitations many local governments face.

"They have limited resources, varying standards. So why do people get surprised when we get results like these?" Soaries said while holding pages of data from the Scripps Howard study.

"We don't want to attack people who don't have the resources or who've not had the training necessary. We have to tell the truth, but we don't want to beat up on these people."