Phoenix New Times
Ballot Box Breakdowns
An independent voting-technology expert finds serious problems in the Maricopa County Elections Department
Originally published by Phoenix New Times 2006-01-12
By John Dougherty
An independent voting-technology expert has discovered widespread problems within the Maricopa County Elections Department that raise serious questions over the ability of voting officials in the nation's fourth-most-populous county to conduct fair and accurate elections.
"Any election where the margin of victory is under 2 percent could be called into question," says University of Iowa computer science professor Douglas Jones.
Jones is one of the nation's top experts on voting-machine technology. He discovered the irregularities during an inspection of the county's vote-tabulation machinery late last month.
"These problems," Jones says in an interview, "suggest a systemic problem with election administration" in Maricopa County and a failure by the state to properly oversee the county's handling of elections.
Jones, who testified before Congress on voting-machine technology following the 2000 presidential election, was inspecting the county's tabulation machines to determine why 489 votes inexplicably appeared during a September 2004 recount of a state legislative race that changed the outcome of the election.
Jones says the appearance of the votes during the District 20 recount on September 21, 2004, appears to have been caused by either failure of the county's voting machines to accurately read ballots or illegal vote tampering. Jones says the only way to determine what happened beyond this is to visually inspect the District 20 ballots.
"Someone should take a look at the real ballots," he says.
An examination of the District 20 ballots, Jones says, would also provide the public with crucial information about the overall fairness and accuracy of elections in Maricopa County. Visual inspection would allow researchers to determine what type of writing instruments voters used to mark early ballots in the race. Early ballots accounted for about half of the votes cast in the 2004 primary and general elections.
Jones says evidence he has uncovered reveals that the county's written instructions to voters using early ballots to use a black ballpoint pen and draw a single line to make their vote creates a 1 in 12 chance that the vote will not be counted.
"We already have found something that is not good news for the elections department," Jones says. "They should have been giving better advice."
Jones' examination of the voting machines has also revealed other serious problems with the way Maricopa County conducts elections, including:
* The county has failed to uniformly calibrate its optical scanning machines used to count ballots, which appears to be a violation of federal election law. At least two of the six Optech 4-C optical scanners Jones tested are far less likely to detect votes cast with black and blue ballpoint pens than the others.
* The optical scanners are improperly calibrated to be ultra-sensitive to pencil marks and will detect even tiny specks of lead and smudge marks as votes, leading to the possibility of unintended votes being counted.
* Election officials appear to lack fundamental knowledge of how their election machinery operates and are making public statements about how to mark ballots that are contrary to the actual behavior of their equipment.
Jones is also critical of Secretary of State Jan Brewer, whose office approves voting machines used by Arizona counties.
"This reflects poorly on the state's process of approving these machines," Jones says.
Neither County Recorder Helen Purcell, who oversees the Maricopa County Elections Department, nor Brewer responded to New Times' requests for interviews to discuss Jones' findings.
But in an interview published last July in the Arizona Capitol Times, Brewer said Maricopa County's optical scanning machines have worked without any problems -- which Jones' report contradicts.
Maricopa County has used its current vote-tabulation equipment since 1995. The county has known since at least 2002 that the machines did not consistently read votes. In that year, a recount of a legislative race found more than 100 new votes, but that election did not generate public scrutiny because the recount did not change its outcome.
There have been at least two major elections in the past six years decided by 2 percent margins or less.
Jones' suggestion that the District 20 ballots be subject to visual inspection is expected to be strongly opposed by key Republican leaders, including Purcell, Maricopa County Attorney Andrew Thomas, Maricopa County Treasurer David Schweikert, Speaker of the House Jim Weiers and Senate President Ken Bennett.
Purcell, Thomas and Schweikert have already rejected a public records request submitted in October by New Times to have an expert examine the ballots. Weiers and Bennett have taken steps to block an investigation of the District 20 recount by Republican Senator Jack Harper, chairman of the Government Accountability and Reform Committee, whose mission is to investigate such irregularities affecting state government.
Last spring, Thomas' office investigated the recount flap at the request of local Maricopa County Republican Party leadership, alerted to significant problems in the election by a New Times column ("Election Eve Nightmare," October 14, 2004).
Despite finding serious problems within the elections department, including evidence of voter-machine malfunction during the recount and witness tampering, Thomas closed the investigation saying there was no evidence of criminal wrongdoing. Thomas has said he will oppose in court any private or state-funded effort to conduct a recount of the District 20 ballots.
Thomas is taking this stand despite the fact that looking at the votes again cannot change the outcome of the election. No official recount is legally possible in the race, since the election has been certified by a judge.
Also, Jones says an unofficial recount is not necessary to determine what happened. Instead, he says, a scientifically random selection of about 1,000 ballots pulled from the 17,000 cast in the District 20 Republican primary would be sufficient to determine whether the voting machines failed to accurately count votes.
Or whether someone tampered with the ballots and added votes between the primary and the recount.
Jones' recommendation for a visual inspection of the District 20 ballots will be included in a report that was scheduled to be released on January 12 to Harper, who planned to hold legislative hearings to determine what caused the votes to appear during the District 20 recount.
New Times obtained an advance copy of Jones' report because the paper agreed to pay Jones to test the accuracy of the county's voting machines.
Jones inspected the county's machines on December 20, after Harper issued a legislative subpoena to the Maricopa County Elections Department demanding that the machines be made available to an independent expert. Senate President Bennett refused to authorize Harper to use Senate funds to pay the costs related to Jones' inspection, telling the committee chairman to find a private source.
Harper turned to New Times, which agreed in late November to cover the cost of Jones coming in.
Harper also issued a subpoena to County Treasurer David Schweikert to release the District 20 ballots, which are required by law to be held in the treasurer's vault for two years following the primary. But Schweikert, acting on the recommendation of County Attorney Thomas, refused to release the ballots to Jones.
The District 20 recount has been marred by controversy and unusual, if not suspicious, actions since September 2004, when Anton Orlich defeated John McComish by four votes in the Republican primary for a seat in the state House of Representatives. The narrow margin of victory automatically triggered a recount in which the 489 additional votes suddenly appeared. McComish won the recount by 13 votes.
Nearly all of the new votes that appeared in the District 20 recount came from early ballots that voters cast at home and mailed to the county elections department. The county used Optech 4-C scanners to count early ballots in the primary and used the same type of machine to conduct the recount. But the Optech 4-C used in the recount, known as Machine No. 5, detected an 18 percent increase in votes from early ballots.
Orlich filed a lawsuit seeking to block the recount, claiming the elections department mishandled the ballots between the primary and the recount and that it appeared one of the county's voting machines failed to accurately count votes.
Maricopa County Superior Court Judge Eddward Ballinger upheld the results of the recount after an extremely hostile and unusual hearing in which a key witness who worked for the manufacturer of the election machines failed to appear in court despite receiving a subpoena.
Thomas' investigation later determined that the witness, Tina Polich, was instructed to "lay low" during the hearing by county officials and her employer, Election Systems and Software Incorporated, who did not want her to testify about the possibility that the voting machines malfunctioned.
According to County Attorney Thomas' investigation, it appears that several county officials deliberately misled Judge Ballinger during the court hearing when they told him that they didn't know how to contact Polich.
Judge Ballinger expressed serious concern about Polich's failure to appear at the hearing and about the county's ability to accurately count early ballots.
Nevertheless, he went ahead and certified the recount as valid.
Elections department director Karen Osborne has repeatedly said in published statements and in sworn court testimony before Judge Ballinger that the reason for the sudden appearance of the votes was that voters casting mail-in ballots used marking instruments that were difficult for the optical scanners to read.
Osborne said glitter pens, gel pens and black felt-tipped pens were particularly troublesome for the county's Optech 4-C scanning machines.
Jones' tests on the voting machines found that the opposite of what Osborne claimed under oath was true.
"The sensitivity test showed that these machines are extraordinarily sensitive to black pencil, to ink from a Jelly Roll brand blue glitter pen and to black ink from a Sanford Sharpie Extra Fine Point pen. With these pens and pencils, the Optech 4-C scanner would pick up and count as a vote even a single dot," Jones' report states.
Jones says in the interview that "the blue glitter pen was more reliably scanned than any pen I tried."
The marking utensils best read by the machine used for the early ballots were No. 2 pencils and black ballpoint pens, Osborne has said repeatedly.
Jones, however, found serious problems with both of these instruments. Jones discovered that even tiny specks of lead from No. 2 pencils could be counted erroneously as votes.
"It's not a good idea to adjust machines to be so sensitive that they pick up fly specks of lead as votes," Jones says.
Votes cast with black ballpoint pens were least likely to be detected, Jones says.
The county "shouldn't have given instructions to voters that led the voters to make marks that [voting officials] can't guarantee will be read," Jones says. "My measurements show they really can't guarantee [the machines] will read a Bic black ballpoint."
Jones also criticized the county's written directions on ballots instructing voters to draw a single line between arrows marked on the ballots when casting a vote. Single lines drawn with blue and black pens, he says, are not dark enough for the county's voting machines to consistently detect.
"Requesting a dark mark instead of a single line would encourage voters to make marks that would be far more likely to be counted," Jones' report states.
Not only has the county been providing incorrect instructions on how to mark early ballots, its Optech 4-C scanners used to count early ballots are not calibrated equally, Jones says.
During his tests, Jones found that two of the six scanners he examined were less sensitive to marks made by blue and black ballpoint pens than the other four machines.
The county's incorrect instructions to voters combined with the inconsistent calibration of scanners could have resulted in the sudden increase in votes detected between the District 20 primary and the recount, he says.
If a sufficient number of ballots in District 20 were cast where voters used black and blue ink pens to draw single lines, and if those ballots were run through a less sensitive Optech 4-C scanning machine during the primary, Jones says the votes may have gone undetected.
And, he says, if those same ballots were then run through a more sensitive Optech 4-C machine during the recount, the uncounted votes may have suddenly appeared.
"An examination of a random sample of the actual . . . Republican early voting ballots from the election would allow this hypothesis to be confirmed or ruled out," Jones states in the report.
"If the examination shows that a sufficient percent of the ballots were marked with a single stroke using a ballpoint pen, as opposed to use of pencil or votes cast with a deeply scribbled mark, this hypothesis would become more likely."
Jones also says visual inspection of the ballots would very likely provide conclusive evidence of whether the ballots were tampered with between the primary and the recount.
Osborne testified during the September 23, 2004, recount hearing before Judge Ballinger that the District 20 ballots were handled and sorted without first notifying the candidates, who are entitled by law to be present during that endeavor.
This unsupervised handling "may have created an opportunity . . . to surreptitiously add marks to ballots," Jones' report states. "I want to emphasize that I do not allege that any such ballot alteration occurred, only that the opportunity may have existed."
Jones says the characteristics of the District 20 race made it ripe for tampering.
Voters were asked to vote for two of the five Republican candidates running for seats in the state House. In some instances, voters either voted for no one or only one of the candidates. Such ballots are called "undervotes."
It would have been possible during the handling of District 20 ballots between the primary and the recount, Jones says, for somebody to have sought out the undervoted ballots and added new votes to them.
Jones also notes that nearly all of the votes that suddenly appeared in the recount were on ballots that were counted as undervotes in the primary.
So as not to attract attention to any tampering, Jones says, hundreds of ballots would have had to be altered and all five candidates would have needed to receive additional votes. The recount showed that the 489 new votes were spread across all five candidates.
"If someone [were] in a hurry to surreptitiously alter large numbers of ballots, I would expect them to mark those ballots all with the same pen or pencil and to make no serious effort to match the style of marks made by the original voter," Jones states.
"Therefore, there is a reasonable chance that an examination of a sample of the . . . Republican early-voting ballots would reveal the presence of surreptitiously made marks, so long as this sample is indeed a representative random sample."
Osborne said in an October 2004 interview that she doesn't believe tampering caused the increase in votes in the recount.
"In my heart of hearts, I don't believe anyone tampered with these ballots at all," she said. Then, she acknowledged, "anything is possible."
©2005 New Times, Inc. All rights reserved.