In the past year, concerns about the accuracy and integrity of computerized elections have entered the general consciousness and become accepted as serious. Issues that I addressed in the March 1, 2006, edition of the Spectator have since been written about in the national media, and the momentum has grown for legislative solutions to be found at the federal level. A new Congress is getting under way, and decisions will be made that will profoundly affect the way Americans cast and count their votes.
While computerized voting has been touted as a way to make elections easier and the results more reliable, an increasing number of voters, poll workers, and election officials have concluded that the process in 2006 was more difficult—not easier—and confidence in the tallies has been undermined. Many activists and legislators now question both the wisdom of relying on software to record votes, and the degree to which our elections depend on computerized voting systems and the manufacturers that sell them.
PRIMARY PROBLEMS—Throughout the 2006 primary season, state after state experienced problems stemming directly from computerized voting. Programming errors led to the wrong candidates being declared victorious; the delivery of uncertified software to the polls led to investigations and penalties in several states; machine failures led to long lines of frustrated voters; and malfunctioning electronic poll books created confusion.
In March, in Illinois, candidates from both parties challenged the results after a host of malfunctions occurred; these included a touch-screen voting machine that “blew up like an M80”; machines showing votes that hadn’t been cast; and machines not working at all. In Cook County, final vote totals were delayed for over a week, leading officials to deny further payment to Sequoia Voting Systems, the vendor that had supplied their equipment.
In the Texas primaries, counties across the state became mired in programming problems, ballot issues, and administrative difficulties that led to delays in reporting and questions about the accuracy of the vote. Things got particularly bad in Tarrant County, where Hart Intercivic eSlate’s voting machines mysteriously added over 100,000 votes to the total.
In Indiana, the delivery of uncertified software for the primaries led to investigations and fines being imposed on voting machine manufacturer Election Systems and Software (ES&S). The primary and subsequent run-off election in Arkansas was described by one election official as “a royal mess” that also led to an investigation of ES&S.
Memory card failures in the Diebold machines used in Ohio’s primaries in May caused the process to decline into chaos. A report issued after the primary concluded that “relying on this [Diebold] system in its present state should be viewed as a calculated risk. . . . the outcome may be an acceptable election, but there is a heightened risk of unacceptable cost.”
In September, problems with electronic poll books, and the failure of election officials to provide the “smart cards” required for voting on Diebold touch-screen systems, delayed the opening of many polling places in Maryland. Thousands were turned away or forced to vote provisionally on makeshift ballots, leading to several legal challenges.
While election train wrecks were occurring in state after state, a series of independent studies confirmed the security vulnerabilities of electronic voting. A widely reported study conducted by researchers at Princeton University showed that touch-screen voting machines made by one particular company could be hacked quickly and undetectably. A report prepared by Carnegie Mellon University computer scientists showed that even basic security features of another company’s voting machine could also be compromised. And in September, an independ-ent review commissioned by the New York State Board of Elections showed that national testing laboratories had inadequately tested software found in as much as 90 percent of the machines that were used to count votes across the country.
AN ERROR-FILLED ELECTION—Predictions of a meltdown, in the weeks before the general election, drew considerable media attention. Many were understandably worried that America was voting on unreliable machines stuffed with undertested and error-prone software. While the direst of predictions were not realized on November 7, the range and severity of the problems that did occur are a clear indication that something must be done to ensure a solid outcome before the next federal election cycle begins.
Widespread problems were reported with electronic equipment in New Jersey, including Essex County, where machines broke down or failed to start at all in over half the towns and cities. Confusion about emergency paper ballots resulted in many voters being turned away without an opportunity to vote. In Cuyahoga County, Ohio, over 100 machine failures led to at least 43 of the county’s 573 polling places either failing to open on time or being unable to use voting equipment. Though many polling places were kept open late, there is no way to determine how many voters were unable to vote as a result of the failures.
In Pennsylvania, thousands of machines failed on Election Day, leading to a call from the state’s Republican Party to impound equipment in 27 counties. In Colorado, the poorly designed and untested software running electronic poll books in Denver overloaded, resulting in voters’ waiting for hours, with an estimated 20,000 leaving without casting a vote. Voting machines failed at thousands of polling places in over half the states—including entire counties in Indiana and Pennsylvania. The problems caused such severe delays in eight states that the voting hours had to be extended.
Control of the United States Senate depended on the results from Virginia, where the reported margin of victory in a senatorial race was three-tenths of one percent, and a recount was impossible because there was no way to recover voter intent from the electronic tallies.
Voters in 16 states reported that their votes had been recorded for the wrong candidate or did not appear on review screens at the end of the voting process. The latter problem led to the most publicized meltdown, one that could potentially have a great impact on policy in the new Congress—the lost votes in Sarasota County, Florida.
FLORIDA, AGAIN—During early voting in Sarasota County, hundreds of voters reported that their votes in the House race between Vern Buchanan and Christine Jennings simply disappeared from the voting-machine screen. While the election director acknowledged these reports in a memo before the election, she refused to investigate or take the machines out of service. The final results showed that over 13 percent of the county’s voters (over 18,000) had not registered any vote in that race. Buchanan’s razor-thin, 369-vote margin of victory over Jennings has led to legal action.
Perhaps most revealing is the fact that the Congressional “undervote” among Sarasota County voters who voted absentee—on paper ballots—was less than 3 percent, similar to the rates among all voters in the neighboring counties of the 13th District, who voted entirely on paper ballots. The official results would suggest that of the voters who used touch-screen machines in Sarasota County, more than 16 percent declined to state a preference in one of the most hotly contested Congressional races in the country.
While unconvincing explanations have been proposed to shift blame from the machines to the voters, it seems evident that the votes of many thousands were simply never recorded. There are hundreds of reports from voters who noticed that no vote for either Congressional candidate appeared on the review screen at the end of the voting process. There were presumably thousands more who didn’t notice the same.
The results of minimal official testing and audits have supported the winner’s claim that one in six of the voters in Sarasota County simply chose not to vote for either candidate. But experts and many Sarasota voters remained unconvinced. The transparency of the election process received a further blow when a circuit court judge, in upholding the “trade secret” protection of the voting system software, ruled that the commercial interests of the voting machine manufacturer superseded the voters’ right to know that their votes were counted accurately. The ruling has been appealed.
In the meantime, the debate has shifted from Florida to Washington, D.C. In December, Christine Jennings filed a formal contest of the election, which will be referred to the Committee on House Administration for investigation. Whatever the ultimate outcome, the experience should serve to focus the larger debate about the appropriate use of computerized equipment in the election process.
PAPER TRAILS OR PAPER BALLOTS—In December, a subcommittee of the panel established to draft new voting-system standards, working together with staff from the National Institute of Standards and Technology, recommended that those new standards should require voting to be “software independent.” The overwhelming sense of the computer science community is that an election process in which the correctness of the results depends solely on software is unacceptable, because there is simply no way to guarantee that the results are correct.
While so-called DRE voting systems (for “direct recording electronic”) have been in use in some jurisdictions in the country since the 1980s, their use has increased significantly since 2000. The Help America Vote Act (HAVA) in 2002 set in motion the most dramatic change in voting technology in the nation’s history, with more than a third of American voters using new electronic equipment in 2006, and many states and counties choosing to implement paperless DRE systems. These changes have been accompanied by a growing awareness and concern among citizens about the accuracy and integrity of electronic voting in general and paperless DREs in particular.
Facing what seemed to be an inevitable wave of touch-screen machines, computer scientists and election reform activists initially responded that these machines needed to produce a simultaneous paper record that the voter could examine before casting a vote electronically. This “voter verified paper audit trail” (VVPAT) would be retained and used in audits to verify, at least to some degree, the accuracy of the electronic tallies.
After initial resistance to the idea of the VVPAT, the four largest voting machine manufacturers eventually developed printer attachments to their pre-existing equipment. These printers were designed in the absence of standards or guidance and, not surprisingly, have proven to be disappointing at best. Many use thermal paper that cannot withstand multiple counts, and most utilize a reel-to-reel mechanism, similar to cash register tapes, that make auditing and recounting cumbersome. Many of the printed records are difficult to read or require the voter to constantly shift focus from the screen to the paper record. In practice the printers have proven to be prone to paper jams and other malfunctions that compromise their value as a check on electronic tabulation.
Many of the advocates who initially supported the addition of VVPAT printers to DRE voting machines are now questioning the wisdom of the solution. Their reassessment is based not only on the inadequacy of the current generation of printers, but rather on a more fundamental rejection of the practice of direct electronic recording of votes.
There seems to be little doubt that federal legislation requiring the minimum safeguard of a voter verified paper-audit trail and mandatory hand-counted audits will pass in the coming months, but given the experience in Sarasota County, many question whether this is an adequate response.
Thousands of Sarasota County voters apparently failed to notice that their vote did not appear on the review screen, and there is no compelling argument that those or other voters would have noticed the missing vote on a VVPAT printer. It is essential that any legislation requiring VVPAT also require that voters be notified of its significance and warned to verify their paper ballots before casting their votes. If VVPAT printers are to be used, they should be required to produce a durable paper ballot capable of withstanding multiple recounts and be usable by disabled voters as well.
Whenever electronic voting machines are used, the initial count is generated with software, whether that count is derived from marks on paper or electronic impulses transmitted from a touch-screen to create a wholly digital record. A voter verified paper ballot, at least theoretically, provides a “software independent” record that serves as an independent means of verification for use in audits and recounts. (Paper ballots marked by the voter, of course, serve that function in situations in which ballots are counted by scanners.) But would a voter verified paper record be a reliable record of voter intent, and, given the considerable problems experienced with paper jams and other printer failures, would the primacy of the paper ballot hold up in court?
A BETTER WAY TO VOTE—Many states have already come to the conclusion that a paper ballot voting system, with ballots either counted by hand or with optical scanners, is not only more accurate and reliable but it is also significantly less expensive. Innovative ballot-marking devices and other systems have allowed 17 entire states and jurisdictions in another 16 states to retain paper ballot systems while still providing voters with disabilities and language minority voters with the opportunity to cast their votes privately and independently. Other states seem certain to follow the lead of New Mexico and Connecticut, where initial plans to purchase touch-screen voting machines were abandoned in favor of paper ballot systems. Federal legislation should help states and counties that want to make the transition to paper ballot systems.
Rep. Rush Holt (D-NJ) has announced his intention of introducing an enhanced version of his Voter Confidence and Increased Accessibility Act, which enjoyed the co-sponsorship of a majority of the members of Congress in the last session. Holt says that the new bill will encourage the adoption of paper ballot systems and establish strict standards for verifiability, accessibility, and security for all voting systems.
While not calling for the elimination of DREs, Holt’s legislation would address a wide range of issues from ballot verification and mandatory random audits to public disclosure of voting system software to prohibiting conflicts of interest between the voting machine manufacturers and the laboratories that test their equipment to federal standards. Senator Dianne Feinstein (D-CA) has indicated that she will introduce similar, if not identical, language. These issues will also be addressed in bills to be introduced by Senator Hillary Clinton and others.
In its new session, Congress can focus attention on our nation’s election process through its investigation of the Sarasota County election contest, and through oversight hearings. Now is the time for meaningful change so that in 2008 our votes can be cast with confidence and counted properly.