by Kenneth Farrall
for Oscar Gandy, Seminar in the Political Economy of Communication and Information (COMM 734)
12/10/2003
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 2.5 License.
Introduction
- Introduction
- The History of Voting Technology Adoption from 1888 to 2003
- The Legal and Institutional Environment
- Reliability and Accuracy of DRE machines
"That love of order and obedience to the laws, which so remarkably characterize the citizens of the United States, are sure pledges of internal tranquility; and the elective franchise, if guarded as the ark of our safety, will peaceably dissipate all combinations to subvert a Constitution, dictated by the wisdom, and resting on the will of the people." --Thomas Jefferson to Benjamin Waring, 1801. ME 10:235Over the past 39 years, a growing percentage of votes in U.S. elections have been counted by computer-based electronic vote counting machines. The designs, algorithms, and mechanisms by which these votes are counted have been largely secreted from public view. It is illegal in most cases for a public citizen to examine the software source code of these vote counting machines. However, the existence of physical ballots marked by the voters themselves has always offered some degree of independent verification if the accuracy of the vote count was in doubt. Over the past few years, however, this physical evidence has been literally designed away with the introduction of Direct Record Electronic (DRE) voting machines. With this latest form of voting technology, the public is not only prevented from learning how the vote has been counted, but is also deprived of any means by which to verify the accuracy of the count. Recounts are, in fact, not even possible.
This new system of vote counting is in direct opposition to the spirit of democracy. Over the past several months there has been increasing public debate about DRE voting machines. Attention has largely focused on questionable security of the machines and associated potential for tampering and election fraud. A small but growing number of Internet news forums and web blogs have framed the problem as a vast right wing conspiracy, a scandal worse than Watergate, but little specific evidence has been offered to explain how such a conspiracy could operate within the checks and balances of the American constitutional republic and under the watchful eye of the fourth estate, the American free press. Further, no conclusive evidence has yet been presented in any major public forum proving that electronic voting machines have been used to defraud an election.
The vast majority of computer security experts agree, however, that DRE machines provide powerful new vectors for fraud and tampering. Given the long history of vote tampering and fraud in America, it would be naive to think that powerful groups with access to the vote counting process have not at least considered fixing an election. Academics, independent researchers, and journalists today have largely focused on current design laws in DRE systems, performance failures of DRE systems in recent elections, and conflicts of interest in private voting machine corporations. They have largely left unanswered two important questions:Has the use of voting technology in elections in the United States evolved in a way that has promoted the interests of certain elite groups at the expense of the general public? Is there convincing evidence that DRE machines, which have been in use to some degree since the 80s, have been used to fix elections?
The Powers that DRE: the political economy of voting technology
Sociologist C. Wright Mills was among the first to consider the problematic nature of interlocking power and influence for the functioning of democracy. In his book, "The Power Elite," Mills stressed the need to understand the increasingly important linkages between politics and economics in the evolution of power and influence:"The decisions made within the political domain determine economic activities and military programs. There is no longer, on the one hand, an economy, and, on the other hand, a political order containing a military establishment unimportant to politics and to money-making. There is a political economy linked, in a thousand ways, with military institutions and decisions. On each side of the world-split running through central Europe and around the Asiatic rimlands, there is an ever-increasing interlocking of economic, military, and political structures. If there is government intervention in the corporate economy, so is there corporate intervention in the governmental process. In the structural sense, this triangle of power is the source of the interlocking directorate that is most important for the historical structure of the present."(1)G. William Domhoff, in his work "The Powers that Be", argues that those within the ruling class exert their domination over government through the special interest process. The special interest process consists of different means by which powerful groups -- corporations, families, individuals -- obtain favors and procedural rulings that are beneficial to their interests.
"The special-interest process can be studied from two different angles. One approach starts with a specific family, corporation, industry or trade association and follows its favor-seeking operations through the particular combination of congressional committees, regulatory agencies and executive bureaucracies that must be wired in order to gain the desired governmental action. The second starts with the functioning of a given regulatory agency, congressional committee, executive department or advisory committee in order to determine how various special interests impinge upon it. Sometimes the investigator has planned the study in advance, but just as often he or she is taking advantage of an accident, scandal or leak that promises to shed new knowledge on machinations which cost the general public tens of billions of dollars each year."(2)If special interests are successfully impinging on the administration of elections then the cost to the general public is not merely economic. For democracy to function at all, special interests must not be able to influence the process of elections in a way that might favor one group of voters over another. For this reason, it is of vital importance that the research community engage in a systematic analysis of the "powers that be" in the electoral system in this country. Who are the primary actors in this space, what are their interests, and by what means do they exert their influence? In what ways, if any, are special interests able to influence outcomes that are not in the public interest?
This section provides an overview of the political economy of voting technology from two interrelated perspectives. First, what have been the key factors driving innovation in and the adoption of voting technology within election systems over time? Second, how has the legal and institutional environment affecting the use of voting technology in elections evolved over time and what events have influenced these changes?
The History of Voting Technology Adoption from 1888 to 2003
Increasingly complex forms of technology have been introduced into election systems in the United States over the past 120 years. Each time a new voting technology is introduced in order to solve problems believed to be inherent in the previous system. A consistent theme has been that each introduction of new voting technologies may solve some problems, but new, sometimes more serious problems emerge. Technology historian Edward Tenner, in his work Why Things Bite Back, explains the root cause of this problem: "the complexity of mechanical systems makes it impossible to test for all possible malfunctions and makes it inevitable that in actual use, some great flaws will appear that were hidden from designers." (1)
In the late 19th century, most voting in U.S. elections was done using simple, non-standard paper ballots. Any form of paper could be used as long as it listed the full name and elective office of the chosen candidate. The Republican and Democratic parties printed their own ballots with distinctive appearances so that party affiliated polling place monitors could observe when citizens were casting straight party line votes in their favor. This information facilitated vote buying and selling, since they could confirm if a vote purchased was properly cast. These non-standard paper ballots made vote tampering so easy that the process of buying votes developed its own lexicon: "'Colonizers' were groups of bought voters who moved en masse to turn the voting tide in doubtful wards. 'Floaters' flitted like honeybees wafting from party to party, casting ballots in response to the highest bidder. 'Repeaters' voted early and, sometimes in disguise, often."(2)
In 1888, flagrant election fraud in the state of Indiana helped Benjamin Harrison to topple President Grover Cleveland. W.W. Dudley, then treasurer of the Republican National Committee, distributed a circular on RNC stationary instructing local leaders in Indiana to assemble floaters into blocks of five with trusted men who would ensure that "none get away and all vote our ticket." (3)
Shortly after this presidential voting scandal Grover Cleveland helped implement the first new form of voting technology introduced in the US, the Australian paper ballot. The Australasia paper ballot was a standardized form listing all candidates for each elective office with boxes next to each name. The ballots were printed by the government and distributed one per voter at the polling places. The standard appearance of the ballot forms and their restricted distribution dramatically reduced the possibilities for tampering. Without the distinctive color and appearance of their party ballots it was extremely difficult for partisan monitors to clearly observe the vote process. Thus, the Australian ballot is credited with dramatically reducing vote fraud at the end of the 19th century.(4)
The Australian ballot innovation did not eliminate vote fraud however. Although it greatly reduced the ability for corrupt individuals to influence the casting of ballots themselves, ballot could still be removed, destroyed or altered by corrupt poll-workers. In addition, entrenched political machines found ways to influence the vote counting process itself. As Douglas W. Jones notes in his "A Brief Illustrated History of Voting" "a corrupt election administration can introduce bias into the count by manipulating the makeup of the tally teams and by instituting carefully crafted objective standards governing what marks on the ballot count as votes." (5)
Continuing concerns about the legitimacy of elections led Jacob H. Myers to develop the first lever voting machine, the "Myers Automatic Booth," first used in Lockport, New York in 1892. The lever voting machine allowed the voter to cast their vote behind a curtain ensuring their secrecy. The wheel and gear systems within the machine tallied the votes, preventing human error or intentional miscounting. According to Myers, his machine could "protect mechanically the voter from rascaldom, and make the process of casting the ballot perfectly plain, simple and secret." (6)
In addition to its ability to reduce fraud, the use of the lever voting machine was credited with greatly improving the efficiency and reducing the cost of elections. An entry on "Voting Machines" in the 1910 Encyclopedia Britannica is particularly illustrative:
As to the important benefits attending the use of machines, there can be mentioned accuracy both in the casting and the counting of the vote, speed in getting in returns, and economy in holding elections. The improvement in accuracy is shown by the fact that the vote for each office usually runs 99% or more of the highest possible vote that could be registered by the number of voters that have voted. Speed is shown by the fact that in the city of Buffalo, with 60,000 voters voting on election day, the complete returns, including the vote on over 100 candidates for the whole city, have been collected, tabulated and announced within 75 minutes from the closing of the polls. Economy is shown by the fact that although these machines are used but one or two days in each year, election expenses are reduced to such an extent that the machines pay for themselves in five or six elections. This is partly due to the smaller number of precincts necessary and the smaller number of election officers in each precinct and the shorter hours that they must work. The city of Buffalo has a dozen or more precincts, in each of which 800 voters or more are voted in an election day of ten hours, and in that city as many as 1041 voters have voted in one election day on one machine. (7)The lever machine had its own problems however. Although lever machines eliminated the possibility of fraud through a corrupted hand counting process, the lack of physical evidence of the voters choice reduced the value of recounts. With lever machines there is no way to verify that individual votes have been counted properly. Recounts in lever machine elections meant simply rereading the vote tally as it read on each machine. Further, the large number of moving parts in lever machines made them susceptible to tampering by maintenance workers and to general wear and tear that can result in machine jamming and lost votes. And the design of lever voting machine made it difficult to support long and complex ballots. (8)
The next major innovation in voting machine technology was implemented nearly 100 years later. In the early 1960s, Professor Joseph P. Harris of the University of California at Berkeley made several improvements and modifications to the portable IBM punch card system Port-a-Punch to create the Votomatic voting machine. The Votomatic machine uses pre-scored punch cards printed with numbers marking each voting position.
"The voter, using a stylus, punches the pre-scored locations corresponding to the choices for which he desires to vote. The hand-held stylus punches out the "chad" from the card, creating holes. Following the voting process, the numbered stub is removed from the card while the voter's choices are protected by an envelope. The card is then collected with all other voted cards and delivered to a computing system."(9)Specific information about the ballot is printed on a separate ballot label attached to the voting machine itself. When the ballot is inserted into the machine, it is completely covered except for the areas corresponding to vote choices for that election. One innovation in the design allows for ballots to include multiple pages. Each page provides access to a new row of punch positions on the punch card, allowing the votomatic machine to support more complex ballots than a lever machine without overwhelming the voter with too many choices at one time.
The Votomatic machine was the first voting technology to facilitate rapid, precise computing by digital computer. After initial success in Georgia primaries and then in the Oregon and California general elections in 1965, I.B.M. bought Harris's Votomatic machine. Three years later, IBM got out of the voting machine market as numerous problems began to surface with the technology.(10) The Votomatic machine is best known today for its role in the election chaos in Florida in 2000.
In the late 1970's, marksense or optical scan systems were introduced. Voters fill in ovals on a ballot in the same way one would mark a multiple choice exam. The first marksense system to be used in an official election was the American Information Systems "model 315 central-count ballot tabulator," used in several Nebraska counties in 1979.(11) Marksense systems have performed well compared to most other voting technologies. As of 2001, about 25% of all voting precincts used marksense systems. According to a recent Congressional Research Service report, their use is expected to continue to grow.(12)
DRE Machines
Direct Recording Electronic (DRE) voting machines eliminate the need for punch card ballots or any form of paper ballot at all. Voters using these machines indicate their choices on a touch screen and the computer within the machine records their vote. When the polls close the vote totals in each computer are uploaded to a central computer for tallying.
DRE machines are being adopted at an increasing rate. One of the primary reasons for this quick rate of adoption is the perception that DRE machines eliminate the kind of problems that plagued the US Presidential election of 2000 in the state of Florida. Because there are no punchcards or hanging chads there can be no confusion over the interpretation of individual votes. Variations on DRE technology, including Internet-based voting, are also being touted as a solution to the problem of voting access. DRE machines with audio components can be used by the blind. The Help America Vote Act requires that every voting precinct in the country have at least one DRE machine buy the year 2006.(13)
Although it is true that DRE machines eliminate the problems associated with punchcards and can be user friendly, they bring with them a host of new problems far more serious than any voting technology used to date. Specifically, DRE machines compound a problem that has been extant with computer-based voting systems since 1964.
As we have discussed earlier, computer-based voting machines since 1964 have moved the vote counting process itself from local election precinct offices to the black box of the computer itself. Although it has been claimed that computer-based ballot counting can eliminate human error, the reality has been that vote counting software frequently has bugs which can result in egregious vote counting errors. It is also possible to intentionally place malicious code in the vote counting computers which can produce incorrect totals in favor of a targeted candidate. (14)
Before the adoption of DRE machines, disputes over questionable vote totals could be at least partially addressed with recounts. Recounts have been used in two ways with computer based vote counting systems: either as an integral part of the system to ensure reliability and accuracy or as a means to address vote counting disputes. Some states have required that punch card ballots be independently recounted as a matter of course. For example, in 1984, the rules of the Maryland State Administrative Board of Election laws (SABEL) required that vote punch card ballots from two districts of Carroll County be taken to a neighboring county, Frederick, to be recounted on an independently managed system. In the process of this recount it was discovered that the vote counting machine in Carroll County had miscounted the votes, resulting in an incorrect determination that Wayne Cogswell had beaten incumbent T. Edward Lippy for Carroll County School Board. (15) Punch ballots have been recounted both by machine and by hand to resolve disputes over close or questionable election tallies on many occassions.
Because DRE technology eliminates the paper ballot entirely, there is nothing to recount. The absence of a verifiable paper trail for the voter's original vote makes most DRE systems as they are currently designed inherently inauditable. This eliminates the one critical check and balance for computer-based voting machines.
Problems with DRE systems do not end with the auditability issue. DRE systems as they are in use today have a number of important security issues that dramatically increase possible vectors for tampering and have been shown to be less reliable than paper ballots and marksense systems. For a complete discussion of DRE accuracy and reliability issues, please see the section entitled "Reliability and Accuracy of DRE machines."
Although all forms of voting technology from the Australian paper ballot to the DRE machine are currently in use in precincts throughout the country, some technologies are slowly being phased out. Lever machines are no longer produced and the use of punch cards is being phased out due to federal mandates in the Help Amercia Vote Act. Paper ballots are used in roughly three percent of all precincts nationally, mostly small towns and rural areas. Because of federal mandates to upgrade election systems, billions of dollars are now being spent nationwide on the purchase of new voting technology. Precincts upgrading their systems have a general choice between markesense and DRE systems. It is estimated that upgrades in voting machine technology required by federal law will cost between $4.2 billion and $6.5 billion nationally.(16)
Summary and Directions for Further Research
Factors that have driven innovation in and adoption of voting technology in the United States have included fraud prevention, speed and efficiency, costs, accuracy, and legal requirements. As states and localities debate the costs and benefits of marksense vs. DRE machines, it appears that there has been little consideration of the potential for paper ballots to play a continued and perhaps expanding role. Canada, for example, successfully hand counted more than 13 million paper ballots in its November 2000 elections.(17) Common notions that computer-based counting systems are more accurate simply have no basis in fact. Paper ballots have outperformed computer-based technology in voter technology comparison studies.(18) Phillip Green, Electoral Commissioner for the ACT Electoral Commission of Australia, explains why Australia continues to rely on paper ballots:
The use of paper ballots in Australian elections is one of the main ways in which we meet these principles for free and fair elections. Paper ballots are transparent: whenever a ballot is issued or counted, party scrutineers can be present. A paper ballot is watched by witnesses from the moment it is given to a voter to the moment it is counted. Security is provided by special devices printed on the ballot papers, and numbered seals placed on all ballot boxes. Paper ballots are physical things - they can be observed and touched, they can be counted, checked and rechecked. If they are changed or substitutes are made, good control measures can identify the discrepancy and prevent injustices.(19)Surprisingly, little is known about the overall cost of election administration in the United States. Expenditures on elections are small enough that they do not appear in the Census of Governments, an annual report by the U.S. Census Bureau detailing the funds state and local governments spend to support their functions. The smallest expenditure category in the report for the year 2000 was "solid waste management" ($14 billion.) Further, accounting practices for election administration differ by county. Some counties provide detailed breakdowns, while other counties provide no breakdown at all.(20)
The Caltech/MIT Voting Technology project estimates that counties spend an average of $10 per voter on elections, $1 - $2 of which is spent on voting equipment. This amounts to a national total of $150 million to $200 million annually to acquire and maintain voting equipment.(21) Upgrade requirements mandated by the HAVA will require considerably higher rates of technology spending over the next few years. A nationwide upgrade to DRE machines is estimated to cost $2.6 billion.(22) MIT researcher Stephen Ansolabehere has pointed out that electronic systems may have hidden costs over the long term:
"First, electronic devices do not last as long as mechanical devices. Second, to exploit the flexibility of design on electronics the software designs will have to change more quickly. Third, memory cards and other devices will become obsolete quickly. Finding "spare parts" in 10 or 15 years for a computer built today may be quite difficult."(23)New research should focus on explicit comparisons of long term costs between marksense, DRE, and traditional paper ballot systems to help provide local decision makers with a wider array of options. Given the low cost of election administration compared to other government functions (such as solid waste management) there needs to be more public debate over the tradeoffs involved in various voting technology choices. The examples of Canada and Australia show there may be better ways.
The Legal and Institutional Environment
Constitutional Framework
The United States Constitution outlines roles for both the state and federal governments in elections. States are given the authority to regulate the elections process itself. This includes the power to decide details of registration procedures, absentee voting requirements, establishment of polling places, and counting and certification of the vote. States are, in general, also responsible for covering the costs of elections. While election policy and regulations are decided at the state level, most states have decentralized the process so that details of election administration, including the choice of voting technologies, are done at the city or county levels.
Although states are responsible for running elections, the Constitution gives Congress the power to affect the administration of elections in many ways. For example, Article I, Section 4, Clause 1, known as the Elections Clause, provides Congress with broad authority to regulate congressional elections:
"The Times, Places, and Manner of holding Elections for Senators and Representatives, shall be prescribed in each State by the Legislature thereof; but the Congress may at any time by Law make or alter such Regulations, except as to the Places of chusing Senators."(1)Throughout the first hundred years of U.S. history, Congress rarely used its authority over states in election administration. In 1884 Congress enacted federal criminal code provisions protecting the act of voting in congressional elections from violence and intimidation. Congressional authority to enact these laws was challenged and upheld in the Supreme Court case "Ex Parte Yarbrough." Citing the Elections Clause, the court stated "it is only because the Congress of the United States, through long habit and long years of forbearance, has, in deference and respect to the States, refrained from the exercise of these powers, that they are now doubted."(2)
In 1965, Congress passed the Voting Rights Act, which forbade the states from imposing any "voting qualification or prerequisite to voting, or standard, practice, or procedure" that denied the "right of any citizen of the United States to vote on account of race or color." The Voting Rights Act extended the authority of the federal government to monitor elections to enforce voting rights. Independent researcher Lynn Landes has suggested that electronic voting machines may be in violation of the Voting Rights Act and U.S. Code, specifically Title 42 - The Public Health and Welfare, Chapter 20 - Elective Franchise, Subchapter I-A Enforcement of Voting Rights, Sec. 1973:
Sec. 1973f. - "Observers at elections; assignment; duties; reports: Whenever an examiner is serving under subchapters I-A to I-C of this title in any political subdivision, the Director of the Office of Personnel Management may assign, at the request of the Attorney General, one or more persons, who may be officers of the United States, (1) to enter and attend at any place for holding an election in such subdivision for the purpose of observing whether persons who are entitled to vote are being permitted to vote, and (2) to enter and attend at any place for tabulating the votes cast at any election held in such subdivision for the purpose of observing whether votes cast by persons entitled to vote are being properly tabulated. Such persons so assigned shall report to an examiner appointed for such political subdivision, to the Attorney General, and if the appointment of examiners has been authorized pursuant to section 1973a(a) of this title, to the court."Sec. 1973i. Prohibited acts: (a) Failure or refusal to permit casting or tabulation of vote (d) Falsification or concealment of material facts or giving of false statements in matters within jurisdiction of examiners or hearing officers; penalties. Whoever, in any matter within the jurisdiction of an examiner or hearing officer knowingly and willfully falsifies or conceals a material fact, or makes any false, fictitious, or fraudulent statements or representations, or makes or uses any false writing or document knowing the same to contain any false, fictitious, or fraudulent statement or entry, shall be fined not more than $10,000 or imprisoned not more than five years, or both.These sections of the US Code require that federal authorities be allowed to ensure that votes are being properly tabulated and prohibit the "concealment of material facts." DRE machines do not permit the observation of vote tabulation and by their nature conceal material facts. To date, no cases have been brought before the Supreme Court or the Circuit Courts addressing the constitutionality of computer-based voting systems.(3)
The Rise of Standards and Certification Bodies
In early 70's, any company could design and create a "voting system" which it could then sell to local election officials. In 1975, National Bureau of Standards and the Office of the Federal Elections produced a joint report, Effective Use of Computing Technology in Vote Tallying, which concluded that computer-related election problems were caused by a lack of appropriate technical skills at the state and local level. States could not develop or implement sophisticated standards to test voting systems hardware and software. In 1984, funds were approved by Congress to allow the FEC to develop voluntary voting machine standards. In January 1990, the FEC formally approved the Performance and Test Standards for Punchcard, Marksense and Direct recording Electronic Voting Systems. The Voting Systems Board of National Association of State Election Directors (NASED) now oversees the national testing process. The board is composed of election officials and independent technical advisors. Formal testing was initiated by NASED in 1994. Day to day operations of the NASED are conducted by the officially designated NASED secretariat, the Election Center, a non-profit organization based in Houston, Texas.(4)
Voting machine vendors who which to have there systems "qualified" for use are first expected to contact the Federal Election Commission for the latest version of the Federal Voting Systems Standards (FVSS). Once a vendor feels their system meets the FVSS, they must then contact an Independent Testing Authority (ITA) sanctioned by NASED. Hardware and software are tested by different ITAs. Currently, hardware and firmware for the voting systems are tested by Wyle Laboratories based in Huntsville, Alabama. Software is checked by one of two ITAs, either CIBER, Inc., also based in Huntsville, Alabama or SysTest Labs, LLC based in Denver, Colorado. When an ITA, through its own testing procedures, confirms that the voting systems meets FVSS, the ITA notifies NASED which then issues the voting system a "NASED qualified identification number." According to NASED "The ITAs DO NOT and WILL NOT respond to outside inquiries about the testing process for voting systems, nor will they answer questions related to a specific manufacturer or a specific voting system. They have neither the staff nor the time to explain the process to the public, the news media or jurisdictions. All such inquiries are to be directed to The Election Center on behalf of NASED." (5)
Voting systems which have been "qualified" by NASED ITAs must then be "certified" for use in their respective states. State certification requirements vary. Generally, the certification process is overseen by the Secretary of State. At times, certification is given with the assistance of an independent state voting machine examiner. According to independent researcher Bev Harris the standard test used by states in their certification process is known as the "logic and accuracy test" (L&A). The L&A is a simple test where ballots are run through the machine. If the machine counts correctly it passes the L&A. L&A tests do little to check voting machine systems against possible fraud. For example, a malicious programmer could design his vote counting code so that the fraudulent count would only be activated on election day and appear perfectly reliable on the testing date.(6) Further, malicious code placed on a smart card could be inserted into the system at any time on election day.(7) According to Arnold B. Urken, who founded the first voting machine testing lab, Election Technology Laboratories, certification of voting systems without direct inspection of the source code is meaningless. Urken refused to certify the voting machines of ES&S because they would not allow him to inspect their source code. (8)
2000 Election Debacle and the Help America Vote Act
As a result of the 2000 Presidential Election debacle, the federal government has exerted more direct authority over the administration of elections and mandated specific procedures for election reform. Among the mandates in the Help America Vote Act of 2001 are the requirement for all states to provide a process for allowing voters to cast provisional ballots by 2004 and the requirement for all states to have at least one DRE machine per county to provide enhanced access to the voting process by people with disabilities by 2006. A provisional ballot allows voters whose voter registration status is in doubt to cast a vote pending review by the local election board.(9) An additional mandate in HAVA seems to clearly require an auditable paper ballot for DRE voting machines. Specifically, HAVA requires, in section 15481, subdivision (a)(2)(B), that:
"(i) The voting system shall produce a permanent paper record with a manual audit capacity for such system.As Daryl R. Wold, former chairman of the Federal Election Commission (1998 – 2002) has said "taken together, these provisions requiring a 'paper record' that is to be used for a 'manual audit' for the 'voting system' make it apparent that HAVA requires a paper record that is seen, verified, and turned in by the voter." It appears, however, that at the State level, where HAVA is being implemented, election officials are claiming that the HAVA requirement can be satisfied by DRE machines producing a paper printout after the polls have closed.(10)
"(ii) The voting system shall provide the voter with an opportunity to change the ballot or correct any error before the permanent paper record is produced.
"(iii) The paper record . . . shall be available as an official record for any recount . . .."
The Impact of Trade Secret and Copyright Law on Electronic Voting
The software used by the major voting machine companies to count votes is proprietary and protected by trade secrets. Attempts by computer experts, independent researchers and other interested private citizens to inspect the source code of vote counting systems have been consistently rejected.
"Many court cases involving allegations of fraud were brought against vendors of electronic systems. There were no convictions. Was there ever any proof of tampering presented? No. Part of the reason for this may be that during the litigation the plaintiffs were never given access to the vote tabulating program, and hence there was no opportunity for anyone to establish evidence to either prove or disprove the allegations."(11)
When a former mayor of Boca Raton, Florida, Emil Danciu, attempted to challenge the results of a Palm Beach City council election where he ran for office in March, 2001, the voting machine vendor, Sequoia Systems, refused to allow their software to be inspected. Danciu claimed that some of the votes cast for him had been incorrectly included in totals for other candidates. Despite the ability of Danciu's consultant, electronic voting expert Rebecca Mercuri, to demonstrate flaws in the system at local election offices, Circuit Judge John Wessel denied Danciu and Mercuri's request to inspect the machines, citing Sequoia's right to maintain its trade secrets. (12)
A public record act request submitted by Jim March to the Alameda County Registrar of Voters for information regarding its use of Diebold election systems and software received the following reposes from the county:
"Please be advised that the county will not provide the information you requested...The County will not allow access or disclose any information regarding the Diebold election system as any information relating to that system is exempted from the PRA (Public Records Act)...The system provided by Diebold Election Systems Inc. ("DESI") is a proprietary system that is recognized as such in the contract between the County and DESI...
...The County contends that the official information privilege in section 1040 of the Evidence Code is applicable because the information requested was acquired by the County in confidence and the County is required to maintain its confidentiality. Any copying or disclosing of such information would violate the license agreements..." (13)
Digital Millennium Copyright Act
In the fall of 1998, Congress passed the Digital Millennium Copyright Act (DMCA), an extension of copyright protections designed to deal with problems the copyright industry saw emerging with the growing adoption of digital technologies such as the personal computer and the Internet. The DMCA, which was supported by the software and entertainment industries and opposed by academics, librarians and scientists, included provisions making it a crime to circumvent anti-piracy measures included in commercial software and stipulated that Internet service providers must remove material from their clients web sites that appears to violate copyright. Copyright holders who determine that a web site is storing or linking to information they believe to be under their copyright can notify the Internet service provider hosting the site and demand that the material be removed:
Under the notice and takedown procedure, a copyright owner submits a notification under penalty of perjury, including a list of specified elements, to the service provider’s designated agent. Failure to comply substantially with the statutory requirements means that the notification will not be considered in determining the requisite level of knowledge by the service provider. If, upon receiving a proper notification, the service provider promptly removes or blocks access to the material identified in the notification, the provider is exempt from monetary liability. In addition, the provider is protected from any liability to any person for claims based on its having taken down the material.(14)
The pull-down demand provision of the DMCA has had a direct impact on the public's access to critical information regarding electronic voting and on its ability to use the Internet as a space for organizing activism against adoption of electronic voting machines.
In late July of this year, voting technology researcher Bev Harris discovered that Diebold, a producer of electronic voting machines, had left several megabytes of company files on a company FTP site with no password protection and a prominent link pointing to the site from the company home page. The files included internal company memos and several versions of voting machine software used to operate Diebold touchscreen voting machines. Information in these files suggested that Diebold was aware of critical problems with software used in actual elections in Georgia and that the company applied patches to its software that had not been properly certified by the state, in violation of state law and federal guidelines.(15)
Bev Harris published her initial findings on her web site, blackboxvoting.org with a link to Diebold servers where the internal files were still available for public access. Diebold Inc. reacted quickly, issuing a DMCA pull-down demand to the Internet service provider hosting Bev Harris's activist web site, AIT, Inc., citing links on blackboxvoting.org to copyrighted internal memos on Diebold servers. AIT, Inc. reacted by shutting down the entire blackboxvoting.org web site. All documents, databases, and organizational files maintained at the blackboxvoting.org site were removed from public access, including personal contact information for 200 activists, and the private strategy sessions for voting machine activists. Public access to blackboxvoting.org was prevented for nearly a month after the pull-down demand. Kim Alexander of the California Voter Foundation reflected on the impact of the DMCA:"The .org site had an active discussion going on among people from all over the US and world about election security and flaws found in Diebold's voting system software. The idea that these discussion boards were confiscated is rather chilling. The discussions were password protected and not open for public view.The reaction within the Internet activist community was swift. Numerous sites, including major universities such as Swarthmore, began to host copies of the Diebold files. Diebold continued to serve pull-down request demands threatening legal action.
"In addition to the impact this action will have on the election security debate, it also may represent the first time the DMCA has been used to shut down a discussion. Although I don't know much about the DMCA, I do think it wasn't written with this kind of purpose in mind."(16)
On November 3, the Electronic Frontier Foundation (EFF) and the Center for Internet and Society Cyberlaw Clinic at Stanford Law School filed a lawsuit on behalf of the nonprofit ISP Online Policy Group (OPG) and two Swarthmore College students "to prevent Diebold’s abusive copyright claims from silencing public debate about voting, the very foundation of our democratic process." Subsequently, Congressman Dennis Kucinich included excerpts from Diebold files on his own web site and called on the House Judiciary Committee to "investigate these abuses."(17)
On December 1st, Diebold announced that it would no longer demand that "Plaintiffs or any other party cease and desist using Diebold’s email archive for noncommercial critical purposes." Further, Diebold promised it would retract earlier legal threats to ISP who had received them.(18)
Summary anmd Directions for Further Research
The U.S. Constitution gives primary responsibility and authority for running elections to the States but gives power to Congress to enact laws that stipulate requirements for the administration of fair and equitable elections. Though it appears that the proprietary nature of vote counting in direct record electronic machines may violate sections of the U.S. Code requiring polling places to allow federal monitoring of counting procedures, no cases to date have challenged the legality or constitutionality of electronic voting machines in the U.S. Supreme or Circuit Courts. Within a decade after the introduction of computer-based voting systems, the federal government began to adopt voluntary standards to assist states in the evaluation and selection of voting technology. After the presidential election of 2000, Congress passed the Help America Vote Act, which provided federal matching funds for election system upgrades and required the installation of at least one DRE machine in each voting precinct. A disputed section of HAVA seems to require that DRE machines provide an voter verifiable paper record of their vote.
Trade secret protections and copyright laws have been used by voting machine companies to prevent public inspection of their software and to stifle public discussion about the impact of their voting machine technology on democracy.
Further research should focus on understanding interpretations of section 15481 of the HAVA by federal and state election authorities. This section seems to require that DRE machines provide an auditable paper ballot. Specifically, research should search for debate relevant to this section in congressional hearings on the HAVA and within state legislatures and the public media. Also, research should focus on the progress and debate over the new bill, the "Voter Confidence and Increased Accessibility Act of 2003," sponsored by House Representative Rush Holt.(19)
Reliability and Accuracy of DRE machines
Analyses of DRE Machine reliability have focused on two areas of performance: residual votes and openness to tampering.
Residual Votes
Residual votes, or roll-off votes, are defined as votes lost in the election process for one of three reasons. First, the voter may choose more than one candidate for a particular office. This a referred to as an overvote. Second, the voter may mark the ballot in away that is unclear and uncountable. Finally, the voter may not indicate any preference at all, resulting in an undervote. 2.3% of all votes cast in presidential elections between 1988 and 2000 were residual votes. Numerous studies have shown that the percentage of residual votes in an election is highly correlated with the type of voting technology used.(1)
Software bugs, physical reliability problems and interface design issues have contributed to vote counting errors in general elections and a high rate of residual votes compared to paper based systems. A study by the Caltech/MIT voting project found that:"manually counted paper ballots have the lowest average incidence of spoiled, uncounted, and unmarked ballots, followed closely by lever machines and optically scanned ballots. Punchcard methods and systems using direct recording electronic devices (DREs) had significantly higher average rates of spoiled, uncounted, and unmarked ballots than any of the other systems. The difference in reliabilities between the best and the worst systems is approximately 1.5 percent of all ballots cast.Tampering
"....in terms of one very basic requirement -- minimizing the number of lost votes -- electronic voting does not have a very good track record. Paper systems have performed much better over the past dozen years."(2)
Recent analyses of DRE touchscreen systems produced by Diebold corporation have determined numerous possible avenues to vote tampering and fraud. In the landmark study, "Analysis of an Electronic Voting System," researchers at Johns Hopkins University identified six different potential agents of vote fraud:1) the Vote himself (using a forged smart card)Among the specific types of security flaws found by the Johns Hopkins researchers: "voters can trivially cast multiple ballots with no built-in traceability, administrative functions can be performed by regular voters, and the threats posed by insiders such as poll workers, software developers, and even janitors is [sic] even greater."(3)
2) Poll Worker (with access to storage media)
3) Poll worker (with access to network traffic)
4) Internet provider (with access to network traffic)
5) OS developer
6) voting system developer
Although the MIT report and other studies provides a comparative measure of reliability across current voting machine technologies, these studies do not measure deeper issues of accuracy. Commenting on the limits of their analysis, researchers for the Caltech/MIT Voting Technology Project said "There are errors that we cannot count. There is no way to measure whether voters accidentally cast ballots for the wrong candidate. We know of no statistically acceptable measures of fraud. And we know of no studies that attempt to measure the incidence and magnitude of errors in the counting of votes produced by transcription errors or programming errors."(4)
Given what is no know about the potential problems with electronic voting machines, current residual vote analyses are painfully inadequate measures of accuracy and reliability. Specifically, studies of voting systems to date do not actively consider the possibility of vote tampering that allows for consistency between the number of votes cast and the number of votes recorded. For example, if a computer voting system could be rigged to shave 1 out of every 10 votes from a third party candidate and add them to a targeted Republican or Democrat candidate, the number of votes cast would be equal to the number of votes counted. The voting system would show no residual votes but it would be highly inaccurate.
How does one test for voting tampering? The most obvious solution is to provide an independent parallel measure of voter preference such as the auditable paper ballot suggested by Rebecca Mercuri. In the absence of such a formal system, one possible comparative measure would be to use unofficial reports of voter preferences such as tracking or exit polls. The basic analysis would be to compare informal polls vs. official results vs. voting machine technology. If research were to show that elections using DRE machines show a much higher rate of divergence between unofficial polls and actual election results than other voting technologies, one would have strong evidence of fraud, not just machine error. Machine error would be random and would tend to average out in such an analysis. Strong deviations correlated with voting technology would be strong evidence of fraudulent human intervention. There are numerous methodological problems with such a study however. The key problems are: unofficial polls are far from a perfect measure of voter intent on election day and the level of resolution for available polling data is not equal to the level of resolution for available data on voting technologies in actual use.
Problem: Accuracy of Unofficial Polls
The methodology used in unofficial polls varies considerably. Some polls are highly unscientific and are not reliable measurements. Further, even polls that are done scientifically are measuring something different from voter preference on election day. The degree to which polls taken the night before an election synch with actual voter preference will depend in large part on the dynamics of individual elections.
Potential Solution: Standardization and Tracking Poll Selectivity
Rather than using just any polls appearing in newspapers or major online media the study should focus on either one specific national caliber pollster (Gallup, Harris, etc.) or a meta-index of reputable polling results. Major public opinion polls have been demonstrated to be highly correlated with the results of state and national elections. Researchers Desart and Holbrook found a Pearson's r correlation of .96 between candidates poll standings in 2000 and the eventual election outcome. They found correlations of .81 and .91 for 1992 and 1996 respectively.(5)
Additional predictive value has been shown to be provided by past voting behavior. Desart and Holbrook have developed a predictive model which combines poll data from the September prior to the election with data about the state's previous voting patterns to predict the outcomes of state and national elections with a high rate o accuracy. Desart and Holbrook's choice of September polling data was influenced by a desire to predict the outcomes of elections as early as possible. Although they acknowledge using October polling data would have provided a modest increase in the models's predictive accuracy they felt the gain in accuracy did not outweigh the loss in lead time for the prediction. For the purposes of the DRE accuracy measure, it is likely that polling data from the days just prior to the election will provide at least a modest improvement over the model as originally presented.
The predicted election results, V(P), will be estimated for each state contest using averaged tracking poll data from major polls conducted in early November (POLLS) and past voting behavior (PAST) for each state:
V(P) = A(POLLS) + B(PAST)
A and B refer to weighting constants (to be determined) for the poll and past behavior components. To limit the impact of dynamic contests where last minute voter swings could reduce correlation between the poll and actual voter intent on election day, the study should remove contests with rapidly changing tracking poll data.
Problem: Resolution Differences Between Voting Technology and Polling Data
Since decisions on which voting technology to implement are generally made at the county level, the ideal study would compare result predictions, official results and voting equipment at the county level. Although voting equipment data is available at the county level, quality polling data is only available at the state and national level.
Potential Solution: The DRE Index
To allow for useful comparisons at the state level, the concept of the DRE Index has been developed. The DRE index is a value between 0 and 100, representing the percentage of individual votes within the state counted by DRE machines. A state where every county offers exclusively DRE technology, like Georgia, would score a 100 on the index. A state that does not use DRE technology at all, like New Hampshire, would score a 0 on the index. The index will be a simple indication of the states total voting population that votes with DRE technology. It should be noted that the DRE composite value for each State is not constant. It changes with time. The DRE index for Georgia in 1980 was 0.
DRE Index Measurement Error
Mixed Precincts
Election Data Services (EDS) collects county-level voting equipment data from all states after each general election. Each county is classified in one to seven equipment categories: paper ballots, lever machines, punch cards, data vote, optical scanning, DRE machines, and mixed. Those counties within the mixed equipment category lack further detail as to the specific breakdown of machines within that category, introducing a level of measurement error into the index which can be calculated for each state. In addition to collecting voting equipment data, EDS collects data on the total number of ballots cast for each county and on the total number of valid votes cast in the presidential election.
Early and Absentee Voting
Early voting and absentee ballot voting often use different forms of voting technology. For example, in Los Angeles County in 2000, voters who cast ballots prior to election day used DRE machines and those who voted on election day used punch card technology(6). Data on early voting and absentee voting at the county level is generally recorded by the Secretary of State but the quality of data varies. To the extent that early voting and absentee data is unavailable at the county level a certain level of measurement error will be introduced into the DRE index. Early voting does not generally occur at a high enough rate to impact the DRE index significantly. Absentee voting, however, is becoming increasingly common. For states that do not report absentee voting at the county level it may be difficult to provide an accurate estimate of error introduced in the index.
Calculating the DRE Index: DRE(I)
Key Variables:
DREV(Cx) -- The total adjusted number of DRE votes cast for each DRE county (x). Subtract known early and absentee vote totals from the total cast ballots reported for the county (unless it is known that early votes were also cast on DRE machines.)
DREV(T) -- Total state-wide ballots cast using DRE machines. Add individual county totals: DREV(C1) + DREV(C2) + ....... + DREV(Cn), where n equals total number of DRE counties.
MCE -- Mixed county error. Add the total cast ballot totals for each county with mixed equipment.
SV(T) -- Total cast ballots for the state.
DRE Index formula:
DRE(I)= (DREV(T)/SV(T) +/- MCE/SV(T))*100
Variables for the DRE fraud study
V(Ax) - Actual state vote percentage for winning canddiate for contest x, where x is a congressional, gubenatorial, or presidential campaign for years between 1998 and 2002.
V(Px) - Predicted state vote percentage for the actual winning candidate for context x.
V(Dx) - Absolute difference between V(Ax) and V(Px).
DRE(Ix) - The DRE Index for the given state at the time of contest x.
Once data has been collected, V(Dx) will be plotted against DRE(Ix). If DRE machines are actually used to defraud elections at a higher rate than other voting technologies, there should be a significant, positive correlation between V(Dx) and DRE(Ix).
Data Sources
State-level polling data from the major pollsters for 1997 to 2003 is available at NationalJournal.com's poll track. County level equipment data and total cast ballots for each year are available at Election Data Services.
Notes:
Political Economy Introduction Notes:
1. Mills, C. Wright. (1956) The Power Elite. Oxford Press, 1956.
2. Domhoff, G. William. (1978) The Powers That Be: Processes of Ruling Class Domination in America. Vintage Books, 1978, p. 25.
Voting Technology Adoption History Notes:
1. Tenner, Edward. (1996) Why Things Bite Back. Vintage. p. 16
2. Ackerman, S.J. (1998) The Vote that Failed. Scientific American.
3. Ackerman (1988. "The Vote..."
4. Ackerman (1988). "The Vote..."
5. Jones, Douglas J. (2003) A Brief Illustrated History of Voting.
6. Hendren, John. (2000) Armed to Send Chads Into Voting Oblivion, New York Times, December 17, 2000, Final, Section 3; Page 8; Column 3; Money and Business/Financial.
7. Encyclopedia Britanica, 1910, "Voting Machines." Excerpted at http://www.cs.uiowa.edu/~jones/voting/brit11.html.
8. Dugger, Ronnie. (1988) Annals of Democracy: Counting Votes, The New Yorker, November 7, 1988
9. Saltman, Roy G. (1975). "Effective Use of Computing Technology in Vote-Tallying", p. 11. Prepared for the Office of Federal Elections, NBSIR 75-687, NBS Special Publication 500-30.
10. Dugger. (1988) "Annals..."
11. Jones. (2003) "A Brief..."
12. Fischer, Eric A. (2001) "Voting Technologies in the United States: Overview and Issues for Congress", CRS Report RL30773, updated March 21, 2001.
13. Help America Vote Act of 2002, Summary and Potential Issues/Recommendations, Produced by the National Association of Secretaries of State, Iowa.
14. Thompson, Ken. (1984). "Reflections on Trusting Trust," Communication of the ACM, Vol. 27, No. 8, August 1984, pp. 761-763.
15. Saltman, Roy G. (1988). "Accuracy, Integrity, and Security in Computerized Vote-Tallying" report for Institute for Computer Sciences and Technology National Bureau of Standards.
16. Welsh, William. (2001). High Costs Could Slow Proposed Voting Reforms Washington Technology, 01/08/01; Vol. 15 No. 19
17. Crary, David. (2000). "Canada shows U.S. how to count" AP, November 28, 2000.
18. Caltech/MIT Voting Technology Project (2001). Residual Votes Attributable to Technology
19. The Politics of the Future: The Internet and Democracy in Australia
20. Caltech/MIT Voting Technology Project. (2001) "Voting: What is, What Could Be." July, 2001.
21. Caltech/MIT Voting Technology Project. (2001) "Voting..."
22. Caltech/MIT Voting Technology Project. (2001) "Voting..."
23. Ansolabehere, Stephen (2001). Prepared Remarks for The Election Reform Task Force of the NCSL
Legal and Institutional Context Notes:
1. GAO, March 2001. "The Scope of Congressional Authority in Election Administration".
2. GAO, "The Scope..."
3. Landes, Lynn. (2003) "VOTING MACHINES: Constitutional Issues, Federal & Case Law," part of research web site.
4. FEC. 2001. "Voting Systems Performance and Test Standards: An Overview." 12/13/2001 AGENDA DOCUMENT 01-62
5. National Association of State Election Directors (2003) "General Overview for Getting a Voting System Qualified"
6. Harris, Bev. (2003) Black Box Voting. Plan Nine Publishing.
7. Hohno, T., Stubblefield, A., Rubin, A., Wallach, D. (2003) "Analysis of an Electronic Voting System," July 23, 2003.
8. Harris. (2003) Black Box Voting. p. 95.
9. Help America Vote Act of 2002, Summary and Potential Issues/Recommendations, Produced by the National Association of Secretaries of State, Iowa.
10. Wold, Darryl R. (2003). The HAVA Requirement for a Voter Verified Paper Record. Position statement. Mr. Wold served as chairman of the Federal Election Commission in Washington, D.C., during 2000, and as a Commissioner from 1998 to 2002. He is currently an attorney in private practice in Orange County, California.
11. O'Halloran, Philip M. (1996) "Pandora's Black Box." Relevance, November 1996 - Vol. III- No. V. Quotation is from Eva Waskell, the director of the Elections Project at Computer Professionals for Social Responsibility in a 1993 report entitled "Overview of Computers and Electors".
12. Bennett, George. 2002. "Touch-screen error exposed," Palm Beach Post, Wednesday, August 21, 2002
13. Copies of Jim Whites Public Records requests and Almeda Country resposnes are at his website, http://www.equalccw.com/voteprar.html
14. The Digital Millennium Copyright Act of 1998, U.S. Copyright Office Summary (Section 512(g)(1)).
16. Kim Alexander, California Voter Foundation, quoted online at http://thomasmc.com/0929c.htm
17. From Congressman Dennis J. Kucinich's Voting Rights web page.
18. Electronic Frontier Foundation Media Release "Diebold Backs Down, Won't Sue on Publication of Electronic Voting Machine Flaws," Monday, December 1, 2003.
19. H. R. 2239, text of legislation.
DRE Reliability and Accuracy Notes:
1. See Knack and Kropf, 2001. "Invalidated Ballots in the 1996 Presidential Election: A County-Level Analysis" and further references to Thomas, 1968, Steiefbold, 1965, and Mathers, 1964.
2. The Caltech/MIT Voting Technology Project. 2001. "Residual Votes Attributable to Technology".
3. Hohno, T., Stubblefield, A., Rubin, A., Wallach, D. (2003) "Analysis of an Electronic Voting System," July 23, 2003.
4. Caltech/MIT Voting Technology Project (2001). "Residual Votes Attributable to Technology", p. 8.
5. Desart, J. and Holbrook, T. (2003). Statewide Trial-Heat Polls and the 2000 Presidential Election: A Forecast Model, Social Science Quarterly, Volume 84 Issue 3 Page 561 - September 2003
6. Knack and Kropf. (2001). "Invalidated Ballots..."