CRS Report for Congress
Voting Technologies in the United States:
Overview and Issues for Congress
Updated March 21, 2001
Eric A. Fischer
Senior Specialist in Science and Technology
Resources, Science, and Industry Division

Congressional Research Service ˜ The Library of Congress

Voting Technologies in the United States:
Overview and Issues for Congress
The 2000 presidential election raised questions about whether changes are
needed in the voting technologies used in the United States, and what should be the
federal role. Elections are administered by states and localities through approximately
10,000 jurisdictions at the county level or below. The main federal agency involved
is the Office of Election Administration, which is part of the Federal Election
Commission and assists state and local election officials. In 1990, the FEC released
voluntary standards for computer-based voting systems. The standards were
developed in response to congressional direction and have been adopted in whole or
in part by thirty-two states. Updated standards are in development.
Currently, five different kinds of voting technologies are used: hand-counted
paper ballots, mechanical lever machines, computer punchcards (Votomatic and
Datavote), marksense forms (also called optical scan), and direct recording electronic
systems (DRE). The last three systems are computer-based. All systems except lever
machines and DRE use document ballots on which the voter records choices.
Punchcard systems are the most common, used by about one-third of registered
voters, with marksense systems used by about one-quarter. In all but a few states,
more than one kind of technology is currently in use.
For some of the technologies, concerns have been raised about ballot design,
voter errors, and counting accuracy. Effective ballot design involves balancing the
fairness and clarity of presentation, as well as goals such as promoting completion of
the entire ballot by voters. Different technologies place different constraints on ballot
design. The three basic kinds of error are overvote, undervote (not necessarily an
error), and unintended choice. Technologies differ in how they help voters prevent or
correct such errors, and consequently, the incidence varies with the technology
employed. It may also depend on the condition of equipment and the demographics
of the population. Vote counting involves issues such as the accuracy of the counting
methodology, its speed, its integrity and security, and recounting where necessary.
Those depend on many factors, including the characteristics of the technology used,
the design and condition of the equipment and software, and human behavior.
Reports on the accuracy of different systems vary. Questions have also been raised
about the impacts of remote voting, including absentee and mail-in balloting. Another
form of remote voting currently in development is Internet voting, which so far has
been used only on a very limited basis. The overall prevalence of remote voting is
increasing, raising concerns in particular about potential compromises to ballot
secrecy. A central issue is what role the federal government should play in addressing
the concerns that have been raised about voting systems. Current debate centers on
several questions: What is the extent of congressional authority to regulate voting
technology and procedures? Should national standards be voluntary or mandatory?
Should the scope of the standards be broadened to include ballot design, counting
procedures, and other aspects of election management? Should a uniform technology
be adopted nationally or on a state-by-state basis? Should federal funding be made
available to states or local election jurisdictions for upgrading voting systems?
Significant legislative activity is expected in the 107th Congress (see the CRS Election
Reform Electronic Briefing Book).

Kinds of Voting Technology.......................................1
Paper Ballots...........................................2
Lever Machines.........................................3
Punchcards ............................................. 3
Marksense Forms........................................4
Electronic Voting........................................4
Remote Voting.................................................5
Federal Agency Activities.........................................6
Issues ........................................................6
Ballot Design...............................................6
Voting Errors..............................................8
Counting ................................................. 10
Accessibility ............................................... 14
Standardization ............................................ 14
Upgrading Voting Systems...................................17
List of Tables
Table 1. Types of Voting Technologies Used in the United States, 1998......2

Voting Technologies in the United States:
Overview and Issues for Congress
The Presidential election of the year 2000 has raised several issues about the
voting technologies used in the United States. This report provides an overview of
the technologies used and the issues raised, including a brief history and
characteristics of the different methods, the design of ballots, kinds and sources of
voter error, vote counting, and standards. It discusses whether changes may be
needed, what those changes might be, what is the federal role, and legislative
initiatives in the 106th and 107th Congresses.
Kinds of Voting Technology
Elections in the United States are administered at the state and local level, and
the federal government does not set mandatory standards for voting technologies.
There are approximately 10,000 election jurisdictions that administer major federal
elections at the county level or below. In most states, elections are run at the county
level. However, in some, they are administered by townships or other county
subdivisions.1 Currently, five different kinds of voting technologies are used: paper
ballots, lever machines, punchcards, marksense forms, and electronic systems (see
Table 1). Votes in federal elections are categorical — a voter chooses only one
candidate per office. Some countries, however, use other approaches, such as ordinal2
voting, where voters rank candidates. Despite the wide range of alternative voting

1As of 1988, of the approximately 3,140 counties and independent cities in the United States,

2,870 in 41 states conducted state and federal elections. In nine states (Connecticut, Maine,

Massachusetts, Michigan, Minnesota, New Hampshire, Rhode Island, Vermont, and
Wisconsin), 7,610 minor civil divisions, which are county subdivisions such as townships, ran
those elections (Roy G. Saltman, Accuracy, Integrity, and Security in Computerized Vote-
Tallying, National Bureau of Standards Special Publication 500-158, August 1988, 48; no
more recent sources on minor civil divisions were available for this report, but county data are
also listed in Election Data Services, “1998 Voting Equipment Study Report,”
[], n.d.; however, some observers
say that because of consolidation of responsibilities, the total number of election jurisdictions
is approximately 6,800 for federal elections rather than 10,000). Delaware has a state-run
2Categorical voting is used in a first-past-the-post system, in which the candidate with the
plurality of votes wins, or in a two-round system, in which a run-off election is held if no
candidate receives a majority of the vote. Ordinal or preferential voting provides a way of
ensuring that the winning candidate has a majority, not just a plurality, of the votes cast
without holding a run-off election. This system is used in Australia. For more information,

technologies that are now used in the United States, the use of categorical voting has
remained a basic feature.
Table 1. Types of Voting Technologies Used in the United
States, 1998
Computer-NumberPercentage UsingMethod
Voting MethodDocumentBallot?Assistedof
Tabulation?CountiesVoters aPrecincts
Paper ballotYesNo4101.6%2.9%
Lever machineNoNo48018.6%21.8%
Punch Cards:
Votomatic typeYesYes57831.0%33.4%
Datavote Yes Yes 57 3.3% 4.0%
Marksense Yes Yes 1,217 27.3% 24.7%
Electronic No Yes 257 9.1% 7.3%
Mixed system b––1419.1%5.9%
Source: Numbers are from Election Data Services, “1998 Voting Equipment Study Report,”
[], n.d. a
Registered voters.b
A mixture involving more than one kind of voting technology used in a county. Numbers listed
do not reflect the actual percentages of precincts or voters using mixed systems, but rather the
percentages of voters registered in and precincts located in counties that Election Data Services has
identified as using mixed systems.
Paper Ballots. For the first 100 years of the Republic, only one voting
technology was available — paper ballots.3 The first major technological change to
those ballots came with the invention of the Australian secret ballot in 1856. Prior to
its adoption, U.S. voters obtained printed ballots containing the names of the
candidates for whom they wished to vote and placed those ballots in the ballot box.
Such a ballot, called a prox or ticket, was printed by each political party or other

2 (...continued)
see Administration and Cost of Elections Project, Way of Voting,
[], 21 September 1998.
3All voting technologies using document ballots use paper or cardstock, but the term paper
ballot generally refers to those that are designed to be read by humans rather than machines.
The main alternative to paper ballots before the 1890s was voice vote, which was used in
some states as late as 1888 (S.J. Ackerman, “The Vote that Failed,” Smithsonian Magazine,
November 1998, 36).

faction that had candidates in the election.4 A version of the ticket ballot is still used
in a few countries.
The basic ballot used in the United States is the Australian or mark choice ballot,
which lists the names of all candidates. Voters mark their choices on the ballot. The
most common form of this ballot type lists candidates by office (the office group or
Massachusetts ballot). One variant gives voters the option of choosing an entire slate
with a single mark (the party column or Indiana ballot), or choosing a slate with
exceptions. The Australian ballot was adopted in the United States beginning in the
1880s, after a series of scandals involving vote-buying and other problems. Paper
ballots are still used in about 3% of precincts, mostly in less populous areas. The5
percentage of voters using paper ballots has declined by half since 1992. All voting
technologies used in federal elections employ ballots with the basic characteristics of
the Australian ballot — all possible choices are displayed for all offices, the voter
marks choices through some mechanism, and those choices are secret. Also, only
ballots provided by the relevant election office can be used, and controls ensure that
a voter submits only the ballot that he or she was issued.
Lever Machines. The next technological advance in voting came with the
introduction of the lever voting machine in 1892. In this technology, there is no
document ballot. A voter enters the voting booth and chooses candidates listed on
a posted ballot by pulling a lever for each candidate choice. The votes are recorded
by advances in a counting mechanism that are made when the voter leaves the booth.
The lever machine therefore eliminates the need to count ballots manually. Instead,
poll workers read the numbers recorded by the counters. Because there is no
document ballot, recounts and audits are limited to review of totals recorded by each
machine. Write-in votes, however, must be recorded on separate document ballots.
About 22% of precincts currently use lever machines. That percentage has declined
substantially since 1992 and is expected to continue to decrease, because the machines
are no longer manufactured, although parts are still available.
Punchcards. The first technological approach utilizing computers to count
votes was the punchcard system, first used in 1964. In this system, the voter records
choices by punching holes in appropriate locations on a paper computer card that is
later fed into a computer reader to record the vote. The piece of card that is punched
out is called a chad. The computer card serves as the document ballot on which the
votes are recorded. As with other document ballots, punchcards can be manually
recounted and audited.
There are two basic types of punchcard system. In one, numbered boxes are
printed on the card, with each box corresponding to a particular ballot choice. The
choices corresponding to those numbered boxes are indicated to the voter in a posted

4 S.J. Ackerman, “The Vote that Failed,” Smithsonian Magazine, November 1998, 36,38;
Richard Reinhardt, “Tapeworm Tickets and Shoulder Strikers,” American West 3 (1966):

34–41, 85–88.

5Data on the use of voting technologies in 1992 cited here and for other systems is from Kevin
J. Coleman, Thomas H. Neale, and Joseph E. Cantor, The Election Process in the United
States, CRS Report 95-800, 6 July 1995, 69.

ballot in the form of a booklet attached to a voting machine, with the appropriate
places to punch indicated for each candidate or other ballot choice. A voter slips the
card into the “throat” of the voting machine, where it rests on a set of rubber strips
under the ballot book. A simple stylus is used to punch out the chad for the box(es)
corresponding to the candidate(s) chosen for each race or other item on the ballot.
Turning a page in the booklet exposes another set of boxes on the card,
corresponding to another set of ballot choices. The card may be prescored —
partially cut through — at the locations within each box where a hole can be punched.
The Votomatic6 system used in Palm Beach County, Florida, and elsewhere is an
example of this system. Write-in choices are not placed on the card itself but are
written elsewhere, such as on the envelope in which the card is placed. If the card
used is not prescored, a special, spring-loaded stylus is used to punch the holes.
About 33% of precincts use this type of system. It is therefore the most widely used
voting technology at present. The number of voters using the system has declined
since 1992, and that decline is expected to continue.
In the other kind of punchcard system, called Datavote, voters punch holes next
to the names of candidates or other ballot choices that are printed on the cards
themselves — there is no ballot book. The voter places the ballot card in a voting
apparatus that has a stapler-like punching mechanism on a slide. The cards are not
prescored. Write-in votes can be placed directly on the card. About 4% of precincts
use the Datavote system, and usage of this system has also declined.
Marksense Forms. This technology, also called optical scan, has been used
for decades in scoring standardized tests. It first became available for use in voting
in the 1980s. In this system, a voter using a paper form and an appropriate writing7
instrument fills in a box or oval or completes an arrow corresponding to each
candidate choice. The completed ballot is then read by a computerized device that
senses and records the marks. Write-in votes can be placed directly on the ballot.
About 25% of precincts use marksense voting systems. The percentage of voters
using this technology has almost doubled since 1992, and that increase is likely to
Electronic Voting. This technology was first introduced in the 1970s.8 Called
direct recording electronic (DRE) technology, this system is somewhat analogous to
(although more sophisticated than) the lever voting machine. Rather than marking a
paper ballot, the voter chooses candidates from a posted ballot. Depending on the
equipment used, the ballot may be printed and posted on the voting machine, or it may
be displayed on a computer screen. Voters make their choices by pushing a button,
touching the screen, or using a similar device. The voter submits those choices before

6 Originally a brand name, the term Votomatic has passed into general usage to refer to this
kind of technology, of which there are now several different models by different
7For example, some systems have required the use of a #2 pencil, while others can read ink.
8Peter A. Schocket, Neil R. Heighberger, and Clyde Brown, “The Effect of Voting
Technology on Voting Behavior in a Simulated Multi-Candidate City Council Election: A
Political Experiment of Ballot Transparency,” Western Political Quarterly, 45 (June 1992):


leaving the booth, for example by pushing a “vote” button, and the votes are directly
stored in a computer memory device such as a removable disk or nonvolatile memory
circuit. If the voting equipment has a keyboard, write-in votes can be recorded
electronically. Otherwise, they must be recorded separately on a document. About
7% of precincts use DRE voting systems. As with marksense systems, the percentage
of voters using DRE has almost doubled since 1992 and is expected to continue to
One form of electronic voting currently in development is Internet voting, in
which voters make their choices online. Internet voting differs from DRE systems in
several ways. First, it is often done using a general purpose personal computer rather
than a custom-designed voting machine, although such machines can also be used.
Second, results are not accumulated at the polling place but are sent to the tabulating
computer when cast. Third, results (ballots or counts) are not sent over a direct
modem connection or physically transported to the central tabulator, but are sent over
the Internet. Those features make Internet voting a promising technology in some
ways but pose special challenges for ensuring authentication, secrecy, and security in
the voting process.9 The use of Internet voting is currently limited to demonstration
projects. For example, for the November 2000 election, voters in several counties in
California cast nonbinding votes online, from online voting machines placed in central
locations.10 In the same election, 84 overseas military personnel cast their actual votes
via the Internet through a small pilot project run by the Federal Voter Assistance
Program (FVAP).11
Remote Voting
The voting technologies described above were designed primarily for use at
designated polling places. The systems using document ballots — those that are
paper- or card-based — permit remote voting via absentee or other mail-in balloting.
Lever-machine and DRE systems cannot accommodate remote balloting, so in those
cases a document-based alternative must be used. However, Internet voting could
change that. If used in the polling place, it would be analogous to DRE. However,
it could also provide the possibility of voting from home or another location. If used
in that way, it would be analogous to mail-in balloting.

9See Kevin Coleman and Richard M. Nunno, Internet Voting: Issues and Legislation, CRS
Report RS20639, 16 January 2000. A recent report sponsored by the National Science
Foundation has cautioned against the use of Internet voting until problems associated with
security and other issues are resolved (Internet Policy Institute, Report of the National
Workshop on Internet Voting: Issues and Research Agenda, March 2001,
10California Secretary of State, Online Voting Demonstrations,
[], 21 November 2000.
11This program is run by the Department of Defense. See Federal Voter Assistance Program
(FVAP), [], 15 November 2000.

Remote voting can be very convenient for the voter and therefore may increase
turnout. While states vary in the circumstances for which they permit mail-in ballots,
an increasing percentage of voters use such remote balloting. For example, Oregon
conducted its November 2000 election entirely by mail-in ballot. In California, the
percentage of ballots cast by absentees has increased steadily, from 3% in 1962 to
25% in 1998.12 The Uniformed and Overseas Citizens Absentee Voting Act of 1986
(P.L. 99-410) contains provisions to improve absentee voting for U.S. citizens and13
military living abroad.
Federal Agency Activities
The Office of Election Administration of the Federal Election Commission assists
state and local election officials on matters related to election administration.
Activities of other agencies are limited. The National Science Foundation has
undertaken a study of Internet voting. The Department of Defense, through its
Federal Voting Assistance Program office, and in cooperation with several states,
undertook a test, mentioned above, of Internet voting during the November 2000
federal election.14 The National Institute of Standards and Technology does not have
any current activities in voting technologies, although its predecessor, the National
Bureau of Standards, had undertaken studies of this matter.15
The presidential election of the year 2000 has drawn attention to several issues
relating to the voting technologies currently in use. Issues discussed below include
the following: ballot design, prevention and correction of voter errors, vote counting,
standardization, and upgrading voting systems.

12Data are from a table listing historical absentee ballot use in California (California Secretary
of State, Casting a Ballot, [], 2 November


13See Kevin J. Coleman, The Uniformed and Overseas Citizens Absentee Voting Act:
Background and Issues for the 107th Congress, CRS Report RS20764, 21 February 2001.
14See Coleman and Nunno, Internet Voting, 6, for details.
15Roy G. Saltman, Accuracy, Integrity, and Security in Computerized Vote-Tallying,
National Bureau of Standards Special Publication 500-158, August 1988; this report is still
widely cited. See also Roy G. Saltman, Computer Science and Technology: Effective Use
of Computing Technology in Vote-Tallying, National Bureau of Standards Special
Publication 500-30, April 1978.

Ballot Design
Ballot design16 is not a simple art, and standards vary from state to state.
Concerns raised in the presidential election about the design of ballots, in particular
those used in the Florida counties of Palm Beach and Duval, have led some observers
to call for improvements in ballot design. A central question is, how can a ballot best
be designed to ensure both a fair and clear presentation of choices? Issues of fairness
include factors such as the order of presentation of candidates and the use of uniform
typefaces. Those are often addressed by state law or regulation. For example, some
states require alphabetical presentation of candidates, and some require that the order
of presentation be rotated for different parties from one election to another or even
among precincts during the same election. Another question is how to minimize
ballot fatigue, also called voter fall-off, or roll-off, which refers to a voter completing
only the first part of a ballot. The length, complexity, and manner of presentation of
the ballot (which depends in part on the voting technology used) may all have some
effect on the prevalence of this well-documented phenomenon, and even on the
percentage of voters who complete the first ballot item. Roll-off may even vary
depending on whether the ballot items are all placed on one page or several pages,17
and the position of an item on the page.
The issue of clarity is perhaps more difficult, as it requires the ballot designer to
weigh potentially conflicting design goals. For example, election officials may wish
to enhance the readability of a ballot by using a fairly large typeface. If, however,
there is a long list of candidates for an office, that may require splitting the list
between two separate pages of a ballot book, as was done in Duval County, or
placing them on facing pages — the so-called butterfly ballot used in Palm Beach
County. Both approaches could, however, contribute to errors by voters. With the
former approach, some voters may inadvertently vote for candidates on both pages
(thereby invalidating their vote for that office), and with the latter some may
inadvertently vote for a different candidate than the one intended, or punch two
adjacent holes. The kinds of errors cited for the butterfly ballot particularly can be a
problem if the ballot is not designed to unambiguously guide the voter to the
appropriate hole to punch, or if there is any misalignment of the ballot in the voting
machine that makes it difficult to discern the appropriate hole. The feature of some
marksense ballots that requires voters to complete an arrow pointing to the candidate
of choice is an example of a design that might serve to substantially reduce the risk
of such ambiguity. Similarly, a lever machine may be designed so that pulling a lever
points it at the name of the candidate. Such a feature is not possible with punchcard
systems. The example therefore illustrates that different voting technologies place

16As used here, the term ballot design refers to both document and posted ballots — that is,
it refers not only to the ballots themselves, but also to the presentation of candidate choices.
That includes the voting booklets used in Votomatic and similar punchcard systems, and the
identifying information placed next to levers or other choosing devices in lever and electronic
voting machines.
17R. Darcy and Anne Schneider, “Confusing Ballots, Roll-Off, and the Black Vote,” Western
Political Quarterly 42, No. 3 (1989): 347–364; Stephen M. Nichols and Gregory A. Strizek,
“Electronic Voting Machines and Ballot Roll-Off,” American Politics Quarterly 23, No. 3
(1995): 300–318.

different constraints on the way a ballot can be designed to improve its clarity and
ease of use.
There are a number of views on ballot design. One is that no broad
improvements are necessary, because it is the responsibility of the voter to mark the
ballot properly and to get help if there is a problem. Another approach would be the
development of a standard national ballot for federal elections. Such a common ballot
would eliminate the risk of problems that might arise from local differences in designs
and is discussed further below in the section on standardization. A third alternative
would be the development of uniform voluntary standards for ballot design that are
based on well-established design principles widely used in other interactive
applications, such as Internet Web sites. Such standards might also include guidelines
for using established procedures to test ballots for usability.18
Voting Errors
There are three basic kinds of error that a voter might make: overvote,
undervote, and unintended choice. An overvote is a vote for more candidates for a
particular office than is permitted, such as voting for two candidates for President,
and is usually considered an error. An overvote on a ballot item invalidates the vote
for that item. An undervote is a vote for fewer than permitted, such as voting for no
candidate for President. An undervote may or may not be an error — a voter might,
on the one hand, have tried to vote for a candidate but was unsuccessful in marking
the ballot unambiguously, or might, on the other hand, have chosen not to vote for
any candidate.19 An unintended choice is inadvertently voting for a candidate other
than the one intended. The use of secret balloting precludes determining whether an
observed overvote, undervote, or counted vote is an error or intentional. In this
report, the combination of overvotes plus undervotes will be referred to as roll-off.20
Voting technologies differ in how they help a voter prevent or correct errors, and21
consequently, the incidence varies to some extent with the technology employed.
Lever machines can prevent overvoting through the use of interlocking mechanisms
that prevent a voter from pulling a lever for more than one candidate for a given
office. Electronic systems can prevent overvoting through an electronic equivalent
of such a mechanism. Some marksense systems can reduce overvoting by permitting

18There appears to have been little research to date on ballot design with respect to usability.
See, however, Susan King Roth, “The Unconsidered Ballot,” Visible Language 28, No. 4
(1994): 48–67; and Susan King Roth, “Disenfranchised by Design: Voting Systems and the
Election Process,” Information Design Journal 9 (1998): 1–8, available at
19Undervotes associated with ballot fatigue are an example of the latter.
20Other terms used for this include fall-off, no-vote, uncounted vote, residual vote, and even
undervote. Undervotes are also sometimes called blank ballots and overvotes, spoiled ballots
(although ballots can be spoiled in other ways, such as by having marks that could potentially
be identifying). This proliferation of terms can be potentially confusing.
21See, for example, George B. Mather, Lost Votes: Effects of Methods of Voting on Voter
Participation, (Iowa City: University of Iowa, 1986).

a ballot to be checked by the tabulator (sometimes called a “smart ballot box”) before
submission and indicating if there is an overvote; the voter can then be given a new
No system can prevent undervoting, but electronic systems can potentially
reduce them by indicating via a light or other mechanism the offices for which a voter
has not yet cast a vote,22 or by guiding the voter through a multipage ballot
electronically, thereby reducing the risk of inadvertent page-skipping. Paper or
marksense ballots can easily be mismarked, for example by a voter circling the name
of a candidate rather than marking the appropriate box. Such a mismark would likely
be read by machine as an undervote. That kind of mismark is probably less likely with
a punchcard ballot, since it is marked with a stylus, not a pen or pencil. Marksense
systems that use smart ballot boxes can potentially reduce undervotes if the tabulator
is set to check for them, but that may be impractical if the incidence of ballot roll-off
is high.
Unintended choices also cannot be prevented. However, how often they occur
depends in part on the clarity of the ballot design and functioning of the voting
equipment. For example, concerns have been expressed that misalignment of
punchcard ballots in voting machines may lead to mistaken choices. Some
touchscreen electronic systems can potentially reduce the risk of unintended choice
by allowing the voter to review a summary of the choices made before submitting the
The incidence of errors may also depend on the condition of voting equipment
and the demographics of the voting population. For example, a faulty lever or
mechanical counter on a lever machine or a problem with a circuit on an electronic
system may cause a failure to record a vote. A malfunctioning stylus or deteriorated
equipment in a punchcard voting machine might increase the risk that a punch would
be incomplete. Also, some studies have found that inexperienced and elderly voters
tend to make more errors with punchcard than with other systems.23 There is even
some evidence that the height of a voter can affect the frequency of ballot roll-off with
lever machines.24
Voting technologies also vary with respect to how voters can review their
choices and correct any errors they might have made. With both lever and electronic
systems, voters can review their choices and make changes before they leave the
voting booth, although they have no way of checking to ensure that their choices were
accurately recorded by the voting machine. With other systems, voters who wish to
make a change must leave the booth and obtain a new ballot. With marksense
systems where tabulation is done at the precinct, ballots may be checked by the
tabulator for some kinds of error before being submitted. However, marksense
systems where tabulation is done at a central location do not permit such machine-

22Stephen M. Nichols and Gregory A. Strizek, “Electronic Voting Machines and Ballot Roll-
Off,” American Politics Quarterly 23, No. 3 (1995): 300–318.
23Schocket, “Effect of Voting Technology.”
24 Susan King Roth, “The Unconsidered Ballot,” Visible Language 28, No. 4 (1994): 48–67.

assisted error correction. With paper ballots and punchcards, voters must check the
ballots for errors, although some precinct-based card readers can also be programmed
to check for errors. Some observers contend that such error-checking does in fact
reduce the incidence of overvotes and spoiled ballots.
Differences between the Votomatic and Datavote systems illustrate some of the
error-handling trade-offs involved in ballot design. A Votomatic ballot usually
requires only a single card. If a voter wants to check the ballot to make sure that a
vote was not miscast, he or she has to find the hole that was punched out, find the
corresponding number, and check that against the number in the ballot book. That
must be done for each ballot item, a complex process. A Datavote ballot may require
several cards and may need to be marked on both sides. That raises the possibility
that a voter might miss a card or fail to vote on both sides. However, because the
names of candidates are printed on the cards, a voter can more easily check a
Datavote ballot for errors.
The counting of votes after they are cast involves several issues, including the
accuracy of the counting methodology, its speed, and its integrity and security.
Counting may be done either in individual precincts or at a central location within the
election jurisdiction, or both. It may be performed by a machine, or by human
inspection, or both. Paper ballots are counted manually, usually at the precinct. With
lever machines and electronic systems, there are no document ballots. Counts are
taken at the precinct from lever machines. With marksense or electronic systems,
counts can be performed electronically at the precinct and sent electronically via
modem or by other means to a central location. With punchcard systems, ballots are
usually taken to a central location for machine counting and are tabulated by precinct.
However, they may also be counted at the precinct in some cases.
The accuracy of a vote count depends on many factors, including the
characteristics of the technology used, the design and condition of the equipment and
software, and human behavior. For example, since paper ballots are counted
manually, the accuracy of the count depends on the performance of the people doing
the counts. Lever machines reduce some kinds of human error, but problems with
counts may occur as a result of malfunctioning machines or from errors made by the
poll workers who read them. Punchcards and marksense forms are read by machine,
reducing some kinds of human error. Other problems may arise from software or
hardware errors, or from the ballots themselves. For example, with prescored cards,
voters may not completely remove chad when punching the cards. Such incompletely
removed (hanging or swinging) chad, or even detached chad that is loose in the
counting device, may block punched holes and be read by the counting machine as
undervotes. With marksense systems, voters might make ambiguous marks that a
tabulating machine may read differently depending on factors such as the alignment
of the ballot sheet when it is fed into the machine. Since DRE systems record each
vote electronically as it is cast, there is no risk of human error in counting, and voters
cannot make ambiguous or unreadable choices. Problems might still arise from other
sources such as software or hardware failure, however.

Reports on the accuracy of different systems vary. For example, some have
claimed that punchcard readers can have an error rate as low as 1 vote out of each
10,000 counted under ideal test conditions.25 Error rates as high as 1 in 100 have
been reported from prior elections, and some experts believe that Votomatic
punchcard systems using prescored cards may be the least accurate of the available
technologies.26 However, estimates from actual elections are based on roll-off (ballots
for which votes were not counted) and cannot distinguish errors that occur because
of inherent limitations of the technology from roll-off resulting from mistakes (voter
error) or intentional actions by voters.27 Assessment of the accuracy of a particular
voting technology should also take into account other factors that might affect the
observed roll-off, such as population size or other demographic variables. Voting
systems that reduce or eliminate overvoting (lever machines, DREs, and precinct-
tabulated marksense or punchcard systems) would be expected to produce lower total
roll-off rates than those that do not (paper ballots and centrally tabulated systems).
Some results support that expectation, but others do not.28 Also, the accuracy of a

25Saltman, Accuracy, Integrity, and Security in Computerized Vote-Tallying, 5, recommended
a maximum error rate of 1 in 100,000.
26Ford Fessenden and Christopher Drew, “Alas, Vote-Count Machines Are Only Human,”
The New York Times, Friday, 17 November 2000, Sec A, 1, 25. The error rate of a particular
technology can be of concern especially in a very close election, where the margin of victory
is very small. For example, the final margin of 537 votes out of almost 6 million cast in the

2000 presidential election in Florida was equivalent to about 1 out of every 11,000 votes cast,

which suggests that an ideal error rate might best be considerably smaller than 1 in 10,000.
27 For example, about 180,000 ballots, or 3%, of all those cast in Florida did not have any
vote counted for President (that roll-off includes overvotes, undervotes, and otherwise
uncounted ballots) (Governor's Select Task Force on Election Procedures, Standards and
Technology, Revitalizing Democracy in Florida [final report of the task force], 1 March
2001, 32, []). The rate in counties using
punchcards was 3.8%, whereas in counties using marksense systems it was 1.3% (Pam Iorio,
President, Florida State Association of Supervisors of Elections, “Election Reform in the
Aftermath of Florida’s 2000 Presidential Election,” Remarks to the Governor’s Select Task
Force on Election Procedures, Standards and Technology, Tallahassee, Florida, 8 January

2001). There is no way to determine the degree to which machine error, voter error,

intentional voter action, and other factors each contributed to those numbers.
28For example, in Florida, counties using precinct-tabulated marksense forms had much lower
average roll-off for the presidential race (<1%) than counties using centrally tabulated
marksense forms (6%) or punchcards (4%) in the November 2000 election (Governor's Task
Force, Revitalizing Democracy in Florida, 31–32). Those differences were highly significant
statistically. In contrast, in the neighboring state of Georgia, the average roll-off was 4–5%
for each of the technologies most widely used (lever machines, punchcards, and precinct- or
centrally tabulated marksense forms) (Georgia Secretary of State Cathy Cox, The 2000
Election: A Wake-Up Call for Reform and Change: Report to the Governor and Members
of the General Assembly, January 2001, 7, available online at the Secretary of State's web site
[]. In addition, a recent
preliminary analysis of voting systems used across the United States found that DRE systems
had higher roll-off than marksense systems, lever machines, or even paper ballots. However,
the study did not analyze precinct- versus central-tabulated systems and did not attempt to

system in a given election may depend as well on the particular design and condition
of the voting and counting equipment and the degree to which technical procedures
and specifications are followed by the election administrators.
Pre- and postelection tests are widely performed on voting-machine systems to
check for accuracy and also to guard against tampering. In addition, manual recounts
may be routinely performed on a small percentage of ballots as a check on the validity
of the machine count. Accurate operational tests are most difficult with electronic
and lever-machine systems, where there is no ballot document and the count is
recorded at the voting booth. A thorough test would require hundreds of simulated
votes to be placed on each machine.
Voting technologies may also affect recounts. With lever machines and, in many
cases, DRE systems, recounts are limited to checking the vote totals recorded by each
machine. Some observers consider that an advantage because it limits the potential
for human or machine error to affect the recount. Others consider it a disadvantage,
because it does not allow for a ballot-by-ballot paper audit trail.29 With punchcard
and marksense systems, machine recounts may not produce fully repeatable results —
tallies may vary slightly if recounts are repeated.30 Whether hand recounts are more
accurate than machine counts has been the subject of considerable debate. Some
observers claim that machine tabulators may miss valid votes, misidentifying them as
undervotes or overvotes, and that manual counting can detect more accurately the
voter’s intent. Others assert that manual counting is less objective and that voters
should be required to follow instructions for properly marking ballots. State laws
vary with respect to when manual recounts are appropriate and what standards are to
be used.
All current technologies except paper ballots can produce large counts rapidly
once polls are closed. Systems in which ballots are counted electronically as they are
submitted in the precinct can probably produce the most rapid results. With
Votomatic systems, accuracy may be increased if the cards are manually inspected to31
remove loose chad before counting, but that will sacrifice some speed.
Security requirements and measures also vary among the technologies used.
Document ballots require security measures and controls from the initial printing of

28 (...continued)
determine the causes of the observed results (R. Michael Alvarez and others, “A Preliminary
Assessment of the Reliability of Existing Voting Equipment,” 1 February 2001,
29DREs, however, can record a ballot-by-ballot audit trail electronically.
30This issue has been known for many years with respect to votomatic-type punchcard systems
(see, for example, Saltman, Accuracy, Integrity, and Security in Computerized Vote-Tallying,
110). A major source of the variability is loose or hanging chad, which may or may not block
punchholes when fed through the reader. However, marksense readers may also produce some
variability, for example if some ballots are marked in a way that is close to the readability
limits of the tabulator.
31Fessenden and Drew, “Vote-Counting Machines,” 25.

the ballots through counting and storing them. However, the ballots can serve as a
basis for an audit trail, which is not available for lever machines. DRE systems can
produce electronic audit trails by storing anonymized electronic ballots for each voter,
or by printing a document with ballot choices that the voter deposits in a ballot box
but which is retained solely for audit or recount purposes.32 Experts differ on the
importance of such a paper audit trail for ensuring the security and integrity of the
voting process. Special measures and controls have also been developed for both
hardware and software used in computer-based systems.
Some observers believe that the very diversity and decentralization of the voting
systems used in the United States enhance the integrity of the voting process by
making widespread tampering more difficult. Not only are there five different basic
kinds of voting technology in use, but dozens of different models are used. That
makes tampering much less feasible than if a single system were used. Furthermore,
election jurisdictions differ in how they configure their voting equipment to meet the
requirements of the particular election being administered. That includes how ballots
are arranged and, therefore, how votes are recorded.33 That variability makes
widespread tampering with software or computerized counting machines much more
One important aspect of security is maintaining ballot secrecy. This is a concern
particularly for remote voting and has increased as that form of voting has become
more common. Ballot secrecy is widely considered a crucial mechanism for
preventing vote tampering and fraud. The rapid, widespread adoption of the
Australian ballot in the United States at the end of the 19th century came about
because of scandals resulting in part from the fact that the ticket ballot was not
secret.34 Modern polling-place voting ensures that voters cast secret ballots in two
ways. First, voter identification and ballot casting are performed in two separate
steps. Second, ballots are cast in private. Remote voting, such as by mail or the
Internet, can increase voter participation but can potentially create problems for ballot
secrecy. Two basic aspects of ballot secrecy are first, that once a ballot is cast, it
cannot be traced by a second party to an individual voter, and second, that a voter
cannot demonstrate to others how he or she voted. Failure to maintain those aspects
of ballot secrecy is of concern because it can facilitate corruption. While controls can
be put in place that can greatly reduce the risk of traceability, it is not clear how to35
prevent remote voters from proving how they voted.

32At least one system also provides the capability of electronically checking the printed ballot
against the electronically recorded one. However, a DRE system that also provides a printed
ballot is likely to be expensive, as each voting booth must include a printer as well as the
voting machine.
33Ballot configurations may also vary within jurisdictions. For example, Hillsborough
County, Florida had 82 different versions of the ballot in the November 2000 election, to
accommodate different intracounty races (Iorio, “Election Reform”).
34Ackerman, “The Vote that Failed.”
35For example, in casting absentee ballots, it is common practice that the ballot is placed in
an unsigned envelope that is then placed inside the mailing envelope, which is signed. Election

Federal law sets some requirements for elections with respect to accessibility.
The Voting Rights Act of 1965, as amended, requires that ballots, other relevant
materials, and assistance be provided in the language of citizens from non–English-
speaking minorities who constitute a sufficiently large population in a state or political36
subdivision (42 U.S.C. 1973aa-1a). In some jurisdictions, materials and assistance
will need to be provided in several languages. Assistance must also be allowed for
voters who are blind, disabled, or unable to read or write (42 U.S.C. 1973aa-6). The
Voting Accessibility for the Elderly and Handicapped Act of 1984 (42 U.S.C. 1973ee)
requires that election jurisdictions make available accessible polling places and aid to
elderly and disabled voters. Some states may also have additional requirements for
accessibility and assistance.
Voting technologies differ in how easily they can accommodate those
requirements. DRE systems provide the greatest flexibility. For example, with touch-
screen systems, each machine can be programmed to display ballots in any required
language. Electronic systems can also be designed to accommodate wheelchair-
bound voters and to provide auditory assistance via headphones for blind voters.
With other technologies, special ballots may need to be printed and individual
assistance provided.
States, not the federal government, regulate the voting technologies they use.
However, in response to concerns raised in the 1970s and 1980s about the then
largely unregulated voting-technology industry,37 Congress directed the Federal
Election Commission to develop voluntary standards for computer-based voting
systems.38 The standards, developed in collaboration with the National Association
of State Election Directors and approved in 1990, had been adopted in whole or in
part by 32 states as of November 2000.39 They were developed for both hardware

35 (...continued)
officials can then check to ensure the identity and eligibility of the voter without seeing how
the person voted. However, a voter could still photocopy the ballot or show it to another
person before mailing it, thereby proving how he or she voted. Also, the ballot may not be
filled out in private.
36For a discussion of the Voting Rights Act, see Garrine P. Laney, The Voting Rights Act of

1965, as Amended: Its History and Current Issues, CRS Report 95-896, 14 February 2001.

37See, for example, Saltman, Accuracy, Integrity, and Security in Computerized Vote-
38Federal Election Commission, Frequently Asked Questions about Voting System Standards,
[], 10 July 1988.
39Specifically, those states have adopted either the “standards or the testing of systems against
the standards by independent testing authorities (ITAs) designated by the National Association
of State Election Directors,” (FEC, FAQs about Standards). Standards testing has not been

and software and include functional and documentation requirements, performance
characteristics, and testing procedures for punchcard, marksense, and DRE systems.
Updated standards are expected next year.
One focus of current debate is whether a need exists for more standardization.
Some have suggested that mandatory, rather than voluntary, standards be adopted
nationwide for existing voting technologies. Others have proposed adoption of a
single voting technology nationwide, and still others creation of a standard national
ballot for presidential elections. The adoption of mandatory national standards might
provide more uniformity in the performance of different kinds of voting technology.
Such standards might also be expanded in scope from the current set, which is limited
to computer-based systems and does not address issues such as to what extent a
system reduces voter error. New standards might cover all technologies and other
aspects of election administration, such as election management and ballot design.
However, such standards are unlikely to anticipate all circumstances under which a
technology is actually used, and their adoption by states may not resolve problems
caused by such circumstances or by other differences that the standards do not
address. 40
Adoption of a standard national ballot would likely require adoption of a single
technology as well, since different types of voting systems require different ballot
designs. Either of those two approaches would make the development of standards
easier and would eliminate the risk of problems arising from local differences in voting
technology. For example, in a close statewide election in which different election
jurisdictions use different technologies, some observers have expressed concern that
the outcome of the election could be affected by differences in the accuracy of the
systems used. The economies of scale associated with national adoption of a single41
technology could reduce per-unit cost and improve efficiencies in other ways.
Concerns have been raised about adopting a uniform national technology or a
standard national ballot. The extent of congressional authority to impose such a
system on local election jurisdictions, as well as its desirability, is a subject of some
debate. While the U.S. Constitution gives Congress authority to regulate
congressional elections,42 election jurisdictions must hold state and local elections as

39 (...continued)
applied to all such systems, however; for example, some older systems already in use when
the standards were adopted and were “grandfathered in.”
40For example, Florida is one of the 31 states listed by the FEC as adopting the standards or
tests using them. Yet, the use of those standards did not prevent the kinds of problems widely
cited in Florida with respect to the November 2000 election. Among those problems, some
were arguably caused by differences in the voting technologies used, and others occurred with
respect to aspects of election administration, such as ballot design, not covered by the current
41 While many other countries use a standard national system, in many of those cases, paper
ballots and manual counting are used (see Note 32).
42 Article I gives Congress authority to regulate congressional elections, but it is less clear that

well. Elections are often consolidated, with ballots listing federal, state, and local
candidates, and other ballot items such as referenda. A standard national technology
might reduce the flexibility of local governments to respond to local circumstances,
and converting to such a system would probably be expensive. Jurisdictions would
be unlikely to adopt a technology for use solely in federal elections if it could also be
used for local elections. Some observers also argue that adoption of a common
system could greatly reduce the market for new voting technologies, thereby reducing43
economic incentives to develop further advances. Others claim that the very
diversity of the voting systems used in the United States makes systematic tampering
more difficult, as discussed in the previous section of this report.
Some observers have proposed that, rather than developing national standards,
each state should be encouraged to adopt a uniform voting technology for the entire
state. Currently, almost all states use more than one voting technology, and a few use
all five discussed in this report. This proposed approach might ease some of the
concerns associated with adopting a single technology, but the potential benefits
might also be reduced. Some observers have argued that a uniform statewide system
might be necessary to avoid future election challenges based on violation of the equal
protection clause of the U.S. Constitution. For example, concerns have been raised
that voting technologies thought by some observers to be particularly prone to voter
error, such as Votomatic punchcards, may be disproportionately used by minority or
poor populations. However, a recent analysis found little support for that assertion
with respect to the nationwide distribution of voting equipment.44
Other observers recommend modernization rather than adoption of uniform
systems, arguing that many improvements are best left to the election jurisdictions
themselves. The National Association of Secretaries of State established a National
Election Standards Task Force that drafted a resolution, adopted in February 2001,
calling on state and local governments to work toward modernizing the voting

42 (...continued)
Article II provides such authority for selection of the presidential electors. For discussion of
this issue, see Kenneth R. Thomas, Congressional Authority to Standardize National
Election Procedures, CRS Report RL30747, 27 November 2000. See also General
Accounting Office, The Scope of Congressional Authority in election Administration, GAO-

01-470, 13 March 2001.

43The Canadian system provides an interesting example. The Canada Elections Act
[] provides for a uniform voting system for
national elections. That system has codified the use of a single technology — paper ballots.
Therefore, there is no significant incentive to develop or adopt new technologies for use in
national elections. Nevertheless, some jurisdictions may use other technologies for local
elections. For example, in a recent election for mayor and certain other offices, residents of
Toronto voted using either a marksense form or touchscreen DRE system (“Toronto Vote
2000 Frequently Asked Questions,” [], 17
October 2000). That election was held two weeks before the national election on November

27, 2000.

44Stephen Knack and Martha Kropf, Who Uses Inferior Voting Technology?, January 2001,
(available from the authors).

process, including voting technology, but calling for uniform standards and
procedures within states only for recounts and contested elections.45
The National Association of Counties and the National Association of County
Recorders, Election Officials, and Clerks established a National Commission on
Election Standards and Reform in November 2000. The Election Center, an
association of election and voter registration officials, has established an Elections
Reform Task Force to review concerns about election systems and recommend
changes. Both groups are currently deliberating.
Several states have also examined the issue of modernization and adoption of
standards. On December 14, 2000, Governor Bush of Florida established by
executive order the bipartisan Select Task Force on Election Procedures, Standards,
and Technology.46 The task force examined several issues associated with election
administration and has issued its recommendations. Several other states have
established task forces or are otherwise examining voting technologies and
procedures.47 A privately funded National Commission on Federal Election Reform,
cochaired by Presidents Carter and Ford, is also examining a wide range of issues
relating to voting technology and election administration.48
Upgrading Voting Systems
A decision to upgrade a voting system inevitably involves trade-offs among
different goals, such as cost-efficiency, speed, and accuracy. Paper ballot technology
is the most labor-intensive alternative and may cost several times as much to operate
as computer-based systems, but it requires little technological infrastructure and
comparatively low maintenance costs. Lever machines are no longer made, although
replacement parts are available; maintenance, storage, and transportation costs can
be high. Costs of other systems vary depending on the vendor and features. Given
the variety of systems available, the varying needs and circumstances of different
election jurisdictions, and the uncertain effect of economic factors such as economies
of scale, accurate estimates of the costs of the different technologies could not be
devised for this report. However, punchcard systems have often been considered
among the most economical and efficient, especially for jurisdictions with large

45National Association of Secretaries of State, “Election Reform Resolution,
[], 6 February 2001.
46Governor Jeb Bush, Executive Order 00-349, 14 December 2000. See also “Select
Elections Task Force,” [], n.d.
47Among them are Arizona, California, Colorado, Georgia, Indiana, Iowa, Kentucky,
Maryland, Minnesota, Missouri, New York, Ohio, Oregon, Pennsylvania, Washington,
Wisconsin, and the District of Columbia (Karen Pierog, “States Shocked into Voting Reform
after Florida,” Reuters, 22 January 2001; see also National Association of Secretaries of
State; “State Election Reform Information,” 13 March 2001,
48“Former Presidents Carter and Ford to Oversee Bipartisan National Commission
on Federal Election Reform,” News Release, The Miller Center of Public Affairs, 30 January

2001 [].

populations. DRE systems have often been considered the most expensive (except
perhaps for lever machines), but they are also arguably the most adaptable, with the
greatest potential for speed and accuracy. In any case, replacing an existing system
will usually require a substantial capital expenditure by the relevant jurisdiction.
Figures cited are often a million dollars or more for a jurisdiction, depending on the
kind of system and population size, among other factors. Upgrading may therefore
be considered a low priority compared to other needs, such as schools and roads.
Some also argue that the nature of the market for voting technologies affects their
development and cost. The market is relatively small and is highly regulated, and that
might result in higher costs and less incentive for industry research on voting systems.
Some states have already committed to or are expected to upgrade their voting
systems. For example, the Florida Election Task Force has recommended that
marksense systems be leased for use throughout the state in the 2002 elections.49 The
Florida Secretary of State has reportedly recommended in addition the statewide50
adoption of a DRE system by 2004. The District of Columbia is replacing its
Datavote system with precinct-tabulated marksense technology.51
Currently, no federal programs specifically provide funding for upgrading voting
systems. Recent estimates are that a nationwide upgrade would cost from $2–9
billion.52 However, substantial improvements could probably be made at considerably
lower cost by focusing, for example, on upgrading those systems that have the highest
documented rates of roll-off or other problems. Should Congress decide that federal
assistance is necessary, there are at least two options for providing funding:
discretionary grants, in which funding would be based on the specific proposals
submitted by the relevant jurisdiction; or block or formula grants, in which the amount
of funding would be tied to population or some other relevant factor. Congress could
designate particular eligibility requirements, such as adherence to national standards
or that only certain kinds of voting systems would be eligible for replacement under
the program. Such a funding program might stimulate research by private industry

49Governor's Select Task Force on Election Procedures, Standards and Technology,
Revitalizing Democracy in Florida [final report of the task force], 1 March 2001, 30,
[]. Leasing is estimated to cost $20 million.
50Mark Silva, “Elections Chief Offers Electronic Vote Plan,” Miami Herald, 21 February

2001. The estimated cost is $200 million.

51See [] for an illustration of the system, which
was reported to cost $1 million to purchase (Jim Willard, “How Well Intentioned
Requirements Paved the Road to Election Hell,” News & Analysis @ Govcon, 26 Dec. 2001
[ t/news/article.asp?DocID={98A020A3-DB49-11D4-A76
52See, for example, Alan C. Miller and Nick Anderson, “Voting Reforms Join Race for
Funding,” Los Angeles Times, 13 December 2000, Part A, 1. The basis for the estimate was
not provided, however. Total costs would depend in part on the technology adopted. For
example, adoption of DRE systems could cost several times as much as adoption of marksense
technology. According to Doug Lewis, Executive Director of The Election Center, recent
estimates are $3.5–5 billion to replace all current voting equipment with new technology that
provides full accessibility for persons with disabilities (testimony before Senate Committee
on Rules and Administration, Hearing on Election Reform, 14 March 2001).

on further improvements in voting systems. In addition, Congress might consider
federal research in this area. Such research might be performed by the National
Institutes of Standards and Technology, or by the National Science Foundation, or
Two private efforts have been announced to develop improved voting
technologies. On December 14, 2000, the California Institute of Technology and the
Massachusetts Institute of Technology announced a joint effort to develop an
“easy-to-use, reliable, affordable and secure” voting machine for use in U.S.
elections.53 On January 11, 2001, three major information technology companies —
Unisys, Dell, and Microsoft — announced a joint effort to provide an integrated54
system for the management of elections.

53“Caltech and MIT Join Forces to Create Reliable, Uniform Voting System,” Press Release,
[], 20 December 2000.
54Michael Orey, “Unisys, Dell and Microsoft To Jointly Create Electronic Voting System,”
The Wall Street Journal, 11 January 2001, sec. B, 1, 6. See also “Bye-Bye, Chad,” Unisys
Press Release, [], 11 January