Report to the
U. S. Election Assistance Commission
On
Best Practices to Improve Provisional Voting
Pursuant to the
HELP AMERICA VOTE ACT OF 2002
Public Law 107-252
June 28, 2006
Submitted by
The Eagleton Institute of Politics, Rutgers, The State University of New Jersey
The Moritz College of Law, The Ohio State University
2
Report to the
U. S. Election Assistance Commission
Best Practices to Improve Provisional Voting
CONTENTS
The Research Team 3
Executive Summary 5
Key Findings 6
Recommendations 9
Provisional Voting in 2004 12
Recommendations 19
Attachment 1: Data Sources 27
Appendix A: National Survey of Local Election Officials’ Experiences with
Provisional Voting
Appendix B: Relationship Between Time Allotted to Verify Provisional Ballots
and the Level of Ballots that are Verified
Appendix C: Provisional Ballot Litigation by Issue
Appendix D: Provisional Ballot Litigation by State
Appendix E: State Summaries
Appendix F: Provisional Voting Statutes by State
(Included in electronic form only)
3
The Research Team
This research report on Provisional Voting in the 2004 election is part of a broader
analysis that also includes a study of Voter Identification Requirements, a report
on which is forthcoming. Conducting the work was a consortium of The Eagleton
Institute of Politics of Rutgers, The State University of New Jersey, and The Moritz
College of Law of The Ohio State University.
The Eagleton Institute explores state and national politics through research, education, and public
service, linking the study of politics with its day-to-day practice. It focuses attention on how contemporary
political systems work, how they change, and how they might work better. Eagleton regularly undertakes
projects to enhance political understanding and involvement, often in collaboration with government
agencies, the media, non-profit groups, and other academic institutions.
The Moritz College of Law has served the citizens of Ohio and the nation since its establishment in
1891.It has played a leading role in the legal profession through countless contributions made by
graduates and faculty. Its contributions to election law have become well known through its Election Law
@ Moritz website. Election Law @ Moritz illuminates public understanding of election law and its role in
our nation's democracy.
Project Management Team
Dr. Ruth B. Mandel
Director
The Eagleton Institute of Politics
Board of Governors Professor of Politics
Principal Investigator
Chair of the Project Management Team
Edward B. Foley
Robert M. Duncan/Jones Day Designated
Professor of Law
The Moritz College of Law
Director of Election Law @ Moritz
Ingrid Reed
Director of the New Jersey Project
The Eagleton Institute of Politics
Daniel P. Tokaji
Assistant Professor of Law
The Moritz College of Law
John Weingart
Associate Director
The Eagleton Institute of Politics
Thomas M. O’Neill
Consultant
The Eagleton Institute of Politics
Project Director
Dave Andersen
Graduate Assistant
The Eagleton Institute of Politics
John Harris
Graduate Assistant
The Eagleton Institute of Politics
Donald Linky
Senior Policy Fellow
The Eagleton Institute of Politics
April Rapp
Project Coordinator
Center for Public Interest Polling
The Eagleton Institute of Politics
Sara A. Sampson
Reference Librarian,
The Moritz College of Law
Tim Vercellotti
Assistant Research Professor
Assistant Director, Center for Public Interest
Polling
The Eagleton Institute
Laura Williams
The Moritz College of Law
4
Peer Review Group
R. Michael Alvarez
Professor of Political Science
California Institute of Technology
John C. Harrison
Massee Professor of Law
University of Virginia School of Law
Martha E. Kropf
Assistant Professor Political Science
University of Missouri-Kansas City
Daniel H. Lowenstein
Professor of Law, School of Law
University of California at Los Angeles
Timothy G. O’Rourke
Dean, Fulton School of Liberal Arts
Salisbury University
Bradley Smith
Professor of Law
Capital University Law School
Tim Storey
Program Principal
National Conference of State Legislatures
Peter G. Verniero
former Attorney General, State of New Jersey
Counsel, Sills, Cummis, Epstein and Gross, PC
The Peer Review Group improved the quality of our work by critiquing drafts of our analysis, conclusions
and recommendations. While the Group as a whole and the comments of its members individually
contributed generously to the research effort, any errors of fact or weaknesses in inference are the
responsibility of the Eagleton-Moritz research team. The members of the Peer Review Group do not
necessarily share the views reflected in the policy recommendations of the report.
5
EXECUTIVE SUMMARY
Background and Methodology
This report to the United States Election Assistance Commission (EAC) presents
recommendations for best practices to improve the process of provisional voting. It is based
on research conducted by the Eagleton Institute of Politics at Rutgers, the State University of
New Jersey, and the Moritz College of Law at Ohio State University under contract to the
EAC, dated May 24, 2005.
The Help America Vote Act of 2002 (HAVA, (Public Law 107-252) authorizes the EAC (SEC.
241, 42 USC 15381) to conduct periodic studies of election administration issues. The purpose
of these studies is to promote methods for voting and administering elections, including
provisional voting, that are convenient, accessible and easy to use; that yield accurate, secure
and expeditious voting systems; that afford each registered and eligible voter an equal
opportunity to vote and to have that vote counted; and that are efficient. Section 302(a) of HAVA
required states to establish provisional balloting procedures by January 2004.
1
The process
HAVA outlined left considerable room for variation among the states, arguably including such
critical questions as who qualifies as a registered voter eligible to cast a provisional ballot that
will be counted and in what jurisdiction (precinct or larger unit) the ballot must be cast in order to
be counted.
2
The general requirement for provisional voting is that, if a registered voter appears at a polling
place to vote in an election for Federal office, but either the potential voter’s name does not
appear on the official list of eligible voters for the polling place, or an election official asserts that
the individual is not eligible to vote, that potential voter must be permitted to cast a provisional
ballot. In some states, those who should receive a provisional ballot include, in the words of the
EAC’s Election Day Survey, “first-time voters who registered by mail without identification and
cannot provide identification, as required under HAVA. . .”
3
HAVA also provides that those who
vote pursuant to a court order keeping the polls open after the established closing hour shall vote
by provisional ballot. Election administrators are required by HAVA to notify individuals of their
opportunity to cast a provisional ballot.
1
The Election Center’s National Task Force Report on Election Reform in July 2001 had described provisional ballots
as providing “voters whose registration status cannot be determined at the polls or verified at the election office the
opportunity to vote. The validity of these ballots is determined later, thus ensuring that no eligible voter is turned
away and those truly ineligible will not have their ballots counted.” It recommended “in the absence of election day
registration or other solutions to address registration questions, provisional ballots must be adopted by all
jurisdictions. “ See www.electioncenter.org
.
2
The 2004 election saw at least a dozen suits filed on the issue of whether votes cast in the wrong precinct but the
correct county should be counted. One federal circuit court decided the issue in Sandusky County Democratic Party
v. Blackwell, 387 F.3d565 (6
th
Cir. 2004), which held that votes cast outside the correct precinct did not have to be
counted. The court relied on the presumption that Congress must be clear in order to alter the state-federal balance;
thus Congress, the court concluded would have been clearer had it intended to eliminate state control over polling
location (387 F.3d at 578). An alternative argument, that HAVA’s definition of “jurisdiction” incorporates the broader
definition in the National Voting Rights Act, however, has not been settled by a higher court. But for now states do
seem to have
discretion in how they define “jurisdiction” for the purpose of counting a provisional ballot.
3
The definition of who was entitled to a provisional ballot could differ significantly among the states. In California, for
example, the Secretary of State directed counties to provide voters with the option of voting on a provisional paper
ballot if they felt uncomfortable casting votes on the paperless e-voting machines. "I don't want a voter to not vote on
Election Day because the only option before them is a touch-screen voting machine. I want that voter to have the
confidence that he or she can vote on paper and have the confidence that their vote was cast as marked," Secretary
Shelley said. See http://wired.com/news/evote/0,2645,63298,00.html
. (Our analysis revealed no differences in the
use of provisional ballots in the counties with these paperless e-voting machines.) In Ohio, long lines at some polling
places resulted in legal action directing that voters waiting in line be given provisional ballots to enable them to vote
before the polls closed. (Columbus Dispatch, November 3, 2004 .)
6
Our research began in late May 2005. It focused on six key questions raised by the EAC.
1. How did the states prepare for the onset of the HAVA provisional ballot requirement?
2. How did this vary between states that had previously had some form of provisional ballot
and those that did not?
3. How did litigation affect implementation?
4. How effective was provisional voting in enfranchising qualified voters?
5. Did state and local processes provide for consistent counting of provisional ballots?
6. Did local election officials have a clear understanding of how to implement provisional
voting?
To answer those questions, we:
1. Surveyed 400 local (mostly county) election officials to learn their views about the
administration of provisional voting and to gain insights into their experience in the 2004
election.
2. Reviewed the EAC’s Election Day Survey, news and other published reports in all 50
states to understand the local background of provisional voting and develop leads for
detailed analysis.
4
3. Analyzed statistically provisional voting data from the 2004 election to determine
associations between the use of provisional voting and such variables as states’
experience with provisional voting, use of statewide registration databases, counting out-
of-precinct ballots, and use of different approaches to voter identification.
4. Collected and reviewed the provisional voting statutes and regulations in all 50 states.
5. Analyzed litigation affecting provisional voting or growing out of disputes over provisional
voting in all states.
Our research is intended to provide EAC with a strategy to engage the states in a continuing
effort to strengthen the provisional voting process and increase the consistency with which
provisional voting is administered, particularly within a state. As EAC and the states move
forward to assess and adopt the recommendations made here, provisional voting merits
continuing observation and research. The situation is fluid. As states, particularly those states
that did not offer a provisional ballot before 2004, gain greater experience with the process and
as statewide voter databases are adopted, the provisional voting process will demand further,
research-based refinement.
KEY FINDINGS
Variation among the states
In the 2004 election, nationwide about 1.9 million votes, or 1.6% of turnout, were cast as
provisional ballots. More than 1.2 million, or just over 63%, were counted. Provisional ballots
accounted for a little more than 1% of the final vote tally. These totals obscure the wide variation
in provisional voting among the states.
5
4
Attachment 1 provides detailed information on how this study classifies the states according to the characteristics of
their provisional voting procedures. It also describes how the data used in the statistical analysis may differ from the
data in the Election Day Survey, which became available as our research was concluding.
5
HAVA allows the states considerable latitude in how to implement provisional voting, including deciding who beyond
the required categories of voters should receive provisional ballots and how to determine which provisional ballots
should be counted.
7
• Six states accounted for two-thirds of all the provisional ballots cast.
6
• The percentage of provisional ballots in the total vote varied by a factor of 1,000 -- from
a high of 7% in Alaska to Vermont’s 0.006%.
• The portion of provisional ballots cast that were counted ranged from 96% in Alaska to
6% in Delaware.
• States with voter registration databases counted, on average, 20% of the provisional
ballots cast.
• States without databases counted ballots at more than twice that rate: 44%.
7
• States that provided more time to evaluate provisional ballots counted a greater
proportion of those ballots. Those that provided less than one week counted an average
of 35.4% of their ballots, while states that permitted more than 2 weeks, counted 60.8%.
An important source of variation among states was a state’s previous experience with
provisional voting and with the fail-safe voting provision of the National Voting Rights Act.
The share of provisional ballots in the total vote was six times greater in states that had
used provisional ballots before than in states where the provisional ballot was new. In the
25 states that had some experience with provisional voting before HAVA, a higher portion
of the total vote was cast as provisional ballots and a greater percentage of the
provisional ballots cast were counted than in the 18 new to provisional balloting.
8
Part of
that difference was due to how states had implemented the National Voting Rights Act,
particularly in regard to voters who changed address within weeks of the election. Voters
in California, for example, who moved within their county must cast a provisional ballot,
the information from which is used to update the voter’s address. Other states,
Tennessee for example, found that some fail-safe voters were reluctant to vote by
provisional ballot. As a result, Tennessee abandoned provisional voting for those who
moved within counties and allows failsafe voters cast a regular ballot. Relatively fewer
provisional ballots would tend to be cast in such states.
Variation within states
Within states, too, there was little consistency among different jurisdictions. Of the 20 states for
which we have county-level provisional ballot data, the rate of counting provisional ballots varied
by as much as 90% to 100% among counties in the same state. This variation suggests that
additional factors (including the training of election judges or poll workers) beyond statewide
factors, such as experience or the existence of voter registration databases, also influence the
use of provisional ballots.
• In Ohio some counties counted provisional ballots not cast in the assigned precinct even
though the state’s policy was to count only those ballots cast in the correct precinct.
• Some counties in Washington tracked down voters who would otherwise have had their
provisional ballots rejected because they had failed to complete part of their registration
form, gave them the chance to correct those omissions, and then counted the
provisional ballot.
6
California, New York, Ohio, Arizona, Washington, and North Carolina. The appearance of Arizona, Washington and
North Carolina on this list shows that the number of provisional ballots cast depends on factors other than the size of
the population.
7
As the Carter-Baker Commission report put it, “provisional ballots were needed half as often in states with unified
databases as in states without.” Report on the Commission on Federal Election Reform, “Building Confidence in U. S.
Elections,” September 2005, p. 16.
8
See the appendix for our classification of “old” and “new” states and explanation of why the total is less than 50.
8
Resources available to administer provisional voting varied considerably among and within
states. Differences in demographics and resources result in different experiences with
provisional voting. For example, the Election Day Survey found that staffing problems appeared
to be particularly acute for jurisdictions in the lowest income and education categories. Small,
rural jurisdictions and large, urban jurisdictions tended to report higher rates of an inadequate
number of poll workers within polling places or precincts.
• Jurisdictions with lower education and income tend to report more inactive voter
registrations, lower turnout, and more provisional ballots cast.
• Jurisdictions with higher levels of income and education reported higher average
numbers of poll workers per polling place or precinct and reported lower rates of staffing
problems per precinct.
In precincts located in districts where many voters live in poverty and have low levels of income
and education, the voting process, in general, may be managed poorly. Provisional ballots
cannot be expected to work much better. In these areas, the focus should be on broader
measures to improve the overall functionality of struggling voting districts, although improving
the management of provisional balloting may help at the margin.
The lessons of litigation
Successful legal challenges highlight areas where provisional voting procedures were wanting.
A flurry of litigation occurred around the country in October 2004 concerning the so-called
“wrong precinct issue” – whether provisional ballots cast by voters in a precinct other than their
designated one would be counted for statewide races. Most courts, including the U.S. Court of
Appeals for the Sixth Circuit (the only federal appeals court to rule on the issue), rejected the
contention that HAVA requires the counting of these wrong-precinct provisional ballots. This
litigation was significant nonetheless.
• First, the Sixth Circuit decision established the precedent that voters have the right to sue
in federal court to remedy violations of HAVA.
• Second --and significantly-- the litigation clarified the right of voters to receive provisional
ballots, even though the election officials were certain they would not be counted. The
decision also defined an ancillary right – the right to be directed to the correct precinct.
There voters could cast a regular ballot that would be counted. If they insisted on casting
a provisional ballot in the wrong precinct, they would be on notice that it would be a
symbolic gesture only.
• Third, these lawsuits prompted election officials to take better care in instructing precinct
officials on how to notify voters about the need to go to the correct precinct in order to
cast a countable ballot.
States move to improve their processes
Shortly after the 2004 election, several states came to the conclusion that the administration of
their provisional voting procedures needed to be improved, and they amended their statutes.
The new legislation highlights areas of particular concern to states about their provisional voting
process.
• Florida, Indiana, Virginia, and Washington have clarified or extended the timeline to
evaluate the ballots.
9
• Colorado, New Mexico, North Carolina, and Washington have passed legislation
focused on improving the efficacy and consistency of the voting and counting process.
• Colorado, Arkansas, and North Dakota took up the issue of counting provisional ballots
cast in the wrong precinct.
The wide variation in the implementation of provisional voting among and within states suggests
that EAC can help states strengthen their processes. Research-based recommendations for
best, or at least better, practices that draw on the experience gained in the 2004 election can be
useful in states’ efforts to achieve greater consistency in the administration of provisional voting.
The important effect of experience on the administration of the provisional ballot process
indicates that the states have much they can learn from each other.
SUMMARY OF RECOMMENDATIONS FOR BEST PRACTICES
State efforts to improve the provisional voting process have been underway since the 2004
election. By recommending best practices, the EAC will offer informed advice while respecting
diversity among the states.
Take a quality-improvement approach
Defining what constitutes a successful provisional voting system is difficult. Defining quality
requires a broad perspective about how well the system works, how open it is to error
recognition and correction, and how well provisional voting processes are connected to the
registration and voter identification regimes. A first step is for states to recognize that improving
quality begins with seeing the provisional voting process as a system and taking a systems
approach to regular evaluation through standardized metrics with explicit goals for performance.
EAC can facilitate action by the states by recommending as a best practice that:
• Each state collect data systematically on the provisional voting process to permit
evaluation of its voting system and assess changes from one election to the next. The
data collected should include: provisional votes cast and counted by county; reasons
why provisional ballots were not counted, measures of variance among jurisdictions, and
time required to evaluate ballots by jurisdiction
Emphasize the importance of clarity
Above all else, the EAC should emphasize the importance of clarity in the rules by which each
state governs provisional voting. As state legislators and election officials prepare for the 2006
election, answers to the questions listed in the recommendations section of this report could be
helpful. Among those questions are:
• Does the provisional voting system distribute, collect, record, and tally provisional ballots
with sufficient accuracy to be seen as procedurally legitimate by both supporters and
opponents of the winning candidate?
• Do the procedural requirements of the system permit cost-efficient operation?
• How great is the variation in the use of provisional voting in counties or equivalent levels
of voting jurisdiction within the state? Is the variation great enough to cause concern that
the system may not be administered uniformly across the state?
Court decisions suggest areas for action
10
The court decisions following the 2004 election also suggest procedures for states to
incorporate into their procedures for provisional voting. EAC should recommend to the states
that they:
• Promulgate clear standards for evaluating provisional ballots, and provide training for the
officials who will apply those standards.
• Provide effective materials to be used by local jurisdictions in training poll workers on
such procedures as how to locate polling places for potential voters who show up at the
wrong place.
• Make clear that the only permissible requirement to obtain a provisional ballot is an
affirmation that the voter is registered in the jurisdiction and eligible to vote in an election
for federal office. Poll workers need appropriate training to understand their duty to give
such voters a provisional ballot.
Assess each stage of the provisional voting process
Beyond the procedures suggested by court decisions, states should assess each stage of the
provisional voting process. They can begin by assessing the utility and clarity of the information
for voters on their websites and by considering what information might be added to sample
ballots mailed to voters before elections. The better voters understand their rights and
obligations, the easier the system will be to manage, and the more legitimate the appearance of
the process.
Avoiding error at the polling place will allow more voters to cast a regular ballot and all others
who request it to cast a provisional ballot. Our recommendations for best practices to avoid error
at the polling place include:
• The layout and staffing of the multi-precinct polling place is important. States should
ensure that training materials distributed to every jurisdiction make poll workers familiar
with the options available to voters.
• The provisional ballot should be of a design or color sufficiently different from a regular
ballot to avoid confusion over counting and include take-away information for the voter
on the steps in the ballot evaluation process.
• Because provisional ballots offer a fail-safe, supplies of the ballots at each polling place
should be sufficient for all the potential voters likely to need them. Best practice for
states should provide guidelines (as do Connecticut and Delaware) to estimate the
supply of provisional ballots needed at each polling place.
The clarity of criteria for evaluating voter eligibility is critical to a sound process for deciding
which of the cast provisional ballots should be counted.
• State statutes or regulations should define a reasonable period for voters who lack the
HAVA-specified ID or other information bearing on their eligibility to provide it in order to
facilitate the state’s ability to verify that the person casting the provisional ballot is the
same one who registered. At least 11 states allow voters to provide ID or other
information one to 13 days after voting. Kansas allows voters to proffer their ID by
electronic means or by mail, as well as in person.
• More provisional voters have their ballots counted in those states that count ballots cast
outside the correct precinct. While HAVA arguably leaves this decision up to the states,
pointing out the effect of the narrower definition on the portion of ballots counted could
be useful to the states in deciding this question. States should be aware, however, of the
additional burden placed on the ballot-evaluation process when out-of-precinct ballots
11
are considered. And tradeoffs are involved if out-of-precinct voters are unable to vote for
the local offices that might appear on the ballot in their district of residence.
• If a state does require voters to appear at their assigned precinct, where the same
polling site serves more than one precinct, a voter’s provisional ballot should count so
long as the voter cast that ballot at the correct polling site even if at the wrong precinct
within that location. While the best practice might be for poll workers to direct the voter to
correct precinct poll workers’ advice is not always correct, and the voter should be
protect against ministerial error.
• Officials should follow a written procedure, and perhaps a checklist, to identify the reason
why a provisional ballot is rejected. Colorado’s election rules offer particularly clear
guidance to the official evaluating a provisional ballot.
In verifying provisional ballots, the time by which election officials must make their eligibility
determinations is particularly important in presidential elections because of the need to certify
electors to the Electoral College. Our research did not identify an optimum division of the five
weeks available.
• The best practice here is for states to consider the issue and make a careful decision
about how to complete all steps in the evaluation of ballots and challenges to those
determinations within the five weeks available.
After the election, timely information to voters about the disposition of their provisional ballot can
enable voters to determine if they are registered for future elections and, if not, what they need
to do to become registered.
• Best practice for the states is to establish mechanisms to ensure that voters casting
provisional ballots are informed whether they are now registered for future elections and,
if not, what they need to do to become registered.
Final observation
The detailed examination of each stage in the provisional voting process can lay the foundation
each state needs to improve its system. Efforts to improve provisional voting may be most
effective as part of a broader effort by state and local election officials to strengthen their
systems. Collecting and analyzing data about those systems will enable states to identify which
aspects of the registration and electoral system are most important in shunting voters into the
provisional ballot process. Responsible officials can then look to their registration system,
identification requirements or poll worker training as ways to reduce the need for voters to cast
their ballots provisionally.
12
Provisional Voting in 2004
In the 2004 election, nationwide about 1.9 million votes, or 1.6% of turnout, were cast as
provisional ballots. More than 1.2 million or just over 63% were counted. Provisional ballots
accounted for a little more than 1% of the final vote tally.
These totals obscure the wide variation in provisional voting among the states.
9
Six states
accounted for two-thirds of all the provisional ballots cast.
10
State by state, the percentage of
provisional ballots in the total vote varied by a factor of 1,000 -- from a high of 7% in Alaska to
Vermont’s 0.006%. The portion of provisional ballots cast that were actually counted also
displayed wide variation, ranging from 96% in Alaska to 6% in Delaware. States with voter
registration databases counted, on average, 20% of the provisional ballots cast. Those without
databases counted provisional ballots at more than twice that rate, 44%.
An important source of variation was a state’s previous experience with provisional voting. The
share of provisional ballots in the total vote was six times greater in states that had used
provisional ballots before than in states where the provisional ballot was new. In the 25 states
that had some experience with provisional voting before HAVA, a higher portion of the total vote
was cast as provisional ballots and a greater percentage of the provisional ballots cast were
counted than in the 18 new to provisional balloting.
11
• The percentage of the total vote cast as provisional ballots averaged more than 2% in
the 25 experienced states. This was 4 times the rate in states new to provisional voting,
which averaged 0.47%.
12
• The experienced states counted an average of 58% of the provisional ballots cast,
nearly double the proportion in the new states, which counted just 33% of cast
provisional ballots.
• The combined effect of these two differences was significant. In experienced states
1.53% of the total vote came from counted provisional ballots. In new states, provisional
ballots accounted for only 0.23% of the total vote.
Those voting with provisional ballots in experienced states had their ballots counted more
frequently than those in the new states. This experience effect is evidence that there is room for
improvement in provisional balloting procedures, especially in those states new to the process.
13
That conclusion gains support from the perspectives of the local election officials revealed in the
survey conducted as a part of this research. Local (mostly county level) election officials from
“experienced” states were more likely to:
9
HAVA allows the states considerable latitude in how to implement provisional voting, including deciding who beyond
the required categories of voters should receive provisional ballots and how to determine which provisional ballots
should be counted.
10
California, New York, Ohio, Arizona, Washington, and North Carolina. The appearance of Arizona, Washington and
North Carolina on this list shows that the number of provisional ballots cast depends on factors other than the size of
the population.
11
See the appendix for our classification of “old” and “new” states and explanation of why the total is less than 50.
12
To compensate for the wide differences in vote turnout among the 50 states the average figures here are
calculated as the mean of the percent cast or counted rather than from the raw numbers of ballots cast or counted.
13
Managing the provisional voting process can strain the capacity election administrators. For example, Detroit,
counted 123 of the 1,350 provisional ballots cast there in 2004. A recent study concluded that Detroit’s “ 6-day time
limit to process provisional ballots was very challenging and unrealistic. To overcome this challenge, the entire
department’s employees were mobilized to process provisional ballots.” (emphasis added.) GAO Report-05-997,
“Views of Selected Local Officials on Managing Voter Registration and Ensuring Citizens Can Vote,” September
2005.
13
• Be prepared to direct voters to their correct precincts with maps;
• Regard provisional voting as easy to implement;
• Report that provisional voting sped up and improved polling place operations
• Conclude that the provisional voting process helped officials maintain accurate
registration databases.
Officials from “new” states, on the other hand, were more likely to agree with the statement that
provisional voting created unnecessary problems for election officials and poll workers.
If experience with provisional voting does turn out to be a key variable in performance, that is
good news. As states gain experience with provisional ballots their management of the process
could become more consistent and more effective over subsequent elections. Further
information from the EAC on best practices and the need for more consistent management of
the election process could sharpen the lessons learned by experience. The EAC can facilitate
the exchange of experience among the states and can offer all states information on more
effective administration of provisional voting.
Concluding optimistically that experience will make all the difference, however, may be
unwarranted. Only if the performance of the “new” states was the result of administrative
problems stemming from inexperience will improvement be automatic as election officials move
along the learning curve. Two other possibilities exist. Our current understanding of how
provisional voting worked in 2004 is not sufficient to determine unambiguously which view is
correct.
1. “New” states may have a political culture different from “old” states. That is, underlying
features of the “new” states political system may be the reason they had not adopted
some form of provisional voting before HAVA. The “new” states may strike a different
balance among the competing objectives of ballot access, ballot security and practical
administration. They may ascribe more responsibility to the individual voter to take such
actions as registering early, finding out where the right precinct is, or re-registering after
changing address. They may value keeping control at the local level, rather than ceding
authority to state or federal directives. The training they offer poll workers about
provisional ballots may not be as frequent or effective as in other states. If the
inconsistent performance in the “new” states arises out of this kind of political culture,
improving effectiveness in the use of the provisional ballots -- as measured by intrastate
consistency in administration--- will be harder and take longer to achieve.
14
2. “Old” states may devote fewer resources to updating their registration files or databases
because they consider provisional ballots as a reasonable fail safe way for voters with
registration problems a way to cast a ballot. The adoption of statewide voter registration
databases in compliance with HAVA therefore may reduce the variation in the use of
provisional ballots among the states.
Other influences decreasing consistency among the states include:
14
Despite differing political cultures among states and the latitude HAVA provides states, the statute does, indeed
impose some degree of uniformity on issues that Congress thought essential. For example, before HAVA, took effect,
“no state gave the voter the right to find out the status of their ballot after the election. “ Now all offer that opportunity.
See Bali and Silver, “The Impact of Politics, Race and Fiscal Strains on State Electoral Reforms after Election 2000,”
manuscript, Department of Political Science, Michigan State University. Resisting HAVA’s mandates through foot-
dragging lacks any legitimate foundation in law or policy.
14
• The more rigorous the verification requirements, the smaller the percentage of
provisional ballots that were counted. Some states verified provisional ballots by
comparing the voter’s signature to a sample, some matched such identifying data as
address, birth date, or social security number, others required voters who lacked ID at
the polling place to return later with the ID to evaluate the provisional ballot, and some
required provisional voters to execute an affidavit.
15
- In the 4 states that simply matched signatures, nearly 3.5% of the total turnout
consisted of provisional ballots, and just under three-fourths of those ballots
(73%) were counted.
- In the 14 states that required voters to provide such additional information as
address or date of birth just over 1.5% of the total turnout consisted of provisional
ballots, and 55% of those ballots were counted.
- In the 14 states that required an affidavit (attesting, for example, that the voter
was legally registered and eligible to vote in the jurisdiction) just over one-half of
a percent (0.6%) of turnout came from provisional ballots, and less than one-third
of those (30%) were counted. (But note that HAVA requires all voters to certify
that they are eligible and registered in order to cast a provisional ballot, which is
functionally an affidavit. The 14 states described here used an explicit affidavit
form.)
- In the 10 states that required voters to return later with identifying documents just
under 1.5% of the total turnout came from provisional ballots, and more than half
(52%) of these were counted. Voters apparently found this requirement less
onerous than the affidavit, even though it required a separate trip to a
government office
• Voter registration databases provided information that reduced the number of provisional
ballots counted.
16
In states using provisional voting for the first time, states with
registered-voter databases counted only 20% of the ballots that were cast. States
without such databases counted more than double that rate (44%). As HAVA’s
requirement for adoption of statewide databases spreads across the country, this
variation among states is likely to narrow. Real-time access to a continually updated,
statewide list of registered voters should reduce the number of provisional ballots used
and reduce the percentage counted since most of those who receive them will be less
likely to be actually registered in the state.
• States that counted out-of-precinct ballots counted 56% of the provisional ballots cast.
States that counted only ballots cast in the proper precinct counted an average of 42%
of provisional ballots.
17
- In experienced states, the disparity was even more pronounced: just over half of
provisional ballots cast were counted in states requiring in-district ballots, while
more than two-thirds were counted in those allowing out-of-precinct ballots.
- If all states had counted out-of-precinct ballots, perhaps 290,000 more
provisional ballots would have been counted across the country.
18
15
See Table 2 in Appendix 2 for information on the verification method used in each state.
16
The Election Day Survey found that states using statewide voter registration databases reported a lower incidence
of casting provisional ballots than states without voter registration databases, suggesting that better administration of
voter registration rolls might be associated with fewer instances where voters would be required to cast a provisional
ballot due to a problem with their voter registration.
17
The Election Day Survey concluded that : “Jurisdictions with jurisdiction-wide provisional ballot acceptance
reported higher rates of provisional ballots cast, 2.09 percent of registration or 4.67 percent of ballots cast in polling
places, than those with in-precinct-only acceptance, 0.72 and 1.18 percent, respectively. Predictably, those
jurisdictions with more permissive jurisdiction-wide acceptance reported higher rates of counting provisional ballots,
71.50 percent, than other jurisdictions, 52.50 percent.”
15
• States that provide a longer the time to evaluate provisional ballots counted a higher
proportion of those ballots.
19
- Fourteen states permitted less than one week to evaluate provisional ballots, 15
states permitted between one and two weeks, and 14 states permitted greater
than two weeks
20
.
- Those states that permitted less than one week counted an average of 35.4% of
their ballots.
- States that permitted between one and two weeks counted 47.1%.
- States that permitted more than 2 weeks, counted 60.8% of the provisional
ballots cast
21
.
- The effect of allowing more time for evaluation is felt most strongly in states
where more than 1% of the overall turnout was of provisional ballots. In states
where provisional ballots were used most heavily, those that permitted less than
one week to evaluate ballots counted 58.6% while those that permitted one to
two weeks counted 65.0% of ballots, and those states that permitted greater than
three weeks verified the highest proportion of provisional ballots, at 73.8%.
Variation Within States
Not only was there little consistency among states in the use of provisional ballots, there was
also little consistency within states. This was true in both new and old states. Of the 20 states
for which we have county-level provisional ballot data, the rate of counting provisional ballots
varied by as much as 90% to 100% among counties in the same state. This suggests that
additional factors beyond statewide factors, such as verification requirements or the time
provided for ballot evaluation, also influence the provisional voting process. Reacting to the lack
of consistency within states, the Carter-Baker Commission recommended that “states, not
counties or municipalities, should establish uniform procedures for the verification and counting
of provisional ballots, and that procedure should be applied uniformly throughout the state.”
22
Electionline reported that:
• In Ohio some counties counted provisional ballots not cast in the assigned precinct even
though the state’s policy was to count only those ballots cast in the correct precinct.
• Some counties in Washington tracked down voters who would otherwise have had their
provisional ballots rejected because they had failed to complete part of their registration
form, gave them the chance to correct those omissions, and then counted the
18
This estimate is a rough approximation. States that recognize out-of-precinct ballots counted, on average, 56% of
the provisional votes cast. Applying that ratio to the 1.9 million provisional ballots cast nationwide would result in 1.1
million provisional ballots that would have been counted if all states accepted out-of-precinct votes. States that did not
recognize out-of-precinct ballots counted 42% of the provisional ballots cast, or about 813,000 ballots, for a difference
of about 290,000 votes.
19
See Appendix, Relationship Between Time Allotted to Verify Provisional Ballots and the Level of Ballots that are
Verified, David Andersen, The Eagleton Institute of Politics
20
Many thanks to Ben Shepler, of the Moritz College of Law, for assembling complete data on the time requirements
states permitted for the counting of provisional ballots.
21
43 states are included in this analysis, including Washington D.C. The 7 election-day registration states are
omitted, as is Mississippi, which never provided data on provisional ballots. North Carolina is also omitted from the
regressions, as it does not have a statewide policy on how it verifies provisional ballots.
22
Recommendation 2.3.2 of the Report of the Commission on Federal Election Reform, “Building Confidence in U.S.
Elections,” September 2005, p.16. The report also observed that, “. . .different procedures for counting provisional
ballots within and between states led to legal challenges and political protests. Had the margin of victory for the
presidential contest been narrower, the lengthy dispute that followed the 2000 election could have been repeated.”
16
provisional ballot. This would probably not have come to light except for the sharp
examination caused by the very close election for governor.
Resources available to administer provisional voting varied considerably among and within
states. The result is that differences in demographics and resources result in different
experiences with provisional voting. For example, the Election Day Survey found that:
• Jurisdictions with lower education and income tend to report more inactive voter
registrations, lower turnout, and more provisional ballots cast.
• Jurisdictions with higher levels of income and education reported higher average
numbers of poll workers per polling place or precinct and reported lower rates of staffing
problems per precinct.
• Staffing problems appeared to be particularly acute for jurisdictions in the lowest income
and education categories. Small, rural jurisdictions and large, urban jurisdictions tended
to report higher rates of an inadequate number of poll workers within polling places or
precincts.
• Predominantly non-Hispanic, Black jurisdictions reported a greater percentage of polling
places or precincts with an inadequate number of poll workers. Predominantly non-
Hispanic, Native American jurisdictions reported the second highest percentage of
staffing problems.
The conclusions to be drawn from these findings are clear. In voting districts with lower
education levels, poverty, and inadequately staffed polling places, the voting process is unlikely
to function well. More people will end up casting provisional ballots. That makes the provisional
voting process especially important in such districts. But if jurisdictions struggle with regular
voting, how well are they likely to do with the more complicated provisional balloting process? In
precincts where the voting process, in general, is managed poorly, provisional ballots cannot be
expected to work much better. In these areas, the focus should be on broader measures to
improve the overall functionality of struggling voting districts, although improving the
management of provisional balloting may help at the margin.
Effectiveness of Provisional Voting
The certainty of our conclusions about the effectiveness of provisional voting is limited because
of the complexity of the problem and a lack of important information. An ideal assessment of
how well provisional ballots worked in 2004 would require knowing the decisions of local officials
in 200,000 precincts on how to inform voters about provisional voting; their performance in
providing a provisional ballot to those qualified to receive one, and their decisions whether to
count a provisional ballot. Information needed about the eligibility or registration status of
provisional voters is also not available.
We see no automatic correlation between the quality of a state’s voting system and either the
number of provisional ballots cast or counted. Low numbers could reflect accurate statewide
voting data and good voter education. Or they could suggest that provisional ballots were not
made easily available. High numbers could be seen as signifying an effective provisional voting
system or a weak registration process. But we do know that in 2004 provisional ballots allowed
1.2 million citizens to vote, citizens who would otherwise have been turned away from the polls.
Since we do not know how many registered voters who might have voted but could not, we
cannot estimate with any precision how effective provisional voting was in 2004. The Cal Tech –
MIT Voting Technology Project, however, estimated that 4 – 6 million votes were lost in the
17
2000 presidential election for the reasons shown in Table 1 below. The estimate is an
approximation, but it may provide data good enough for a general assessment of the size of the
pool of potential voters who might have been helped by the provisional ballot process.
Estimates of Votes Lost In 2000 Presidential Election
Votes
Lost
(Millions)
Cause
1.5 – 2 Faulty equipment and confusing
ballots
1.5 – 3 Registration mix-ups
<1 Polling place operations
?
Absentee ballot administration
Table 1 Cal Tech – MIT Voting Technology Project Estimates
4 – 6 million votes are lost in presidential elections due to the causes
shown in the table. Registration mix-ups (e.g., name not on list) and polling
place operations (e.g., directed to wrong precinct) are the causes most
likely to be remedied by provisional voting.
The table shows that the universe of voters who could be helped by provisional voting might be
2.5 – 3 million voters. In 2004, about 1.2 million provisional voters were counted. A rough
estimate of the effectiveness of provisional voting in 2004, then, might be 40% to 50% (ballots
counted/votes lost)
23
. Whatever the precise figure, it seems reasonable to conclude that there
is considerable room for improvement in the administration of provisional voting.
Legislative Response
Indeed, several states
24
came to the conclusion that the administration of their provisional voting
procedures needed to be improved and amended their statutes after the 2004 election. State
legislation adopted since the election points to particular areas of concern.
• Not enough time to examine and count the provisional ballots.
Florida, Indiana, Virginia,
and Washington all have clarified or extended the timeline to evaluate the ballots. But
taking more time can prove a problem, particularly in presidential elections with the
looming deadline to certify the vote for the Electoral College.
25
23
Another interpretation of the data should be considered. The Census Bureau’s Current Population Survey (CPS)
developed the category of ”registration mix-ups” to assess the states’ registration systems. After each election the
CPS asks people if they were registered and if they voted. The CPS gives breakdowns of reasons why people did
not vote. Survey responders tend to deflect blame when answering questions about voting. In the narrow context of
provisional ballots, 'registration problems' would cover only voters who went to the polls where the determination that
they were not registered was wrong or they were registered, but in the wrong precinct. If they were in the wrong
precinct, provisional voting can help them in only 17 states. In 2004, only 6.8% of those not voting and registered
blamed registration problems, while 6.9% reported so in 2000.
24
Twelve states made statutory or regulatory changes: Arizona, Arkansas, Colorado, Florida, Georgia, Indiana,
Louisiana, Montana, New Mexico, North Carolina, Virginia and Wyoming. See Table 4 in Appendix 2.
25
The resources available to evaluate and count provisional ballots within a tight schedule may not be easily
available. The General Accounting Office reports that Detroit, where 1,350 provisional ballots were cast and 123
counted, found the 6-day time frame for processing provisional ballots “very challenging and unrealistic. To overcome
this challenge, the entire department’s employees were mobilized to process provisional ballots.” The report also
found that in Los Angeles County, “staff had to prepare duplicate ballots to remove ineligible or invalid contests when
18
• Lack of uniform rules for counting ballots and effective training
of the election officials in
interpreting and applying those rules to determine the validity of ballots. Colorado, New
Mexico, North Carolina, and Washington have all passed legislation focused on
improving the efficacy and consistency of the voting and counting process.
Litigation
Successful legal challenges to the process highlight areas where provisional voting procedures
were wanting. A flurry of litigation occurred around the country in October 2004 concerning the
so-called “wrong precinct issue” – whether provisional ballots cast by voters in a precinct other
than their designated one would be counted for statewide races. These lawsuits were largely
unsuccessful in their stated goal: most courts, including the U.S. Court of Appeals for the Sixth
Circuit (the only federal appeals court to rule on the issue), rejected the contention that HAVA
requires the counting of these wrong-precinct provisional ballots.
This litigation was significant nonetheless.
• First, the Sixth Circuit decision established the precedent that voters have the right to sue
in federal court to remedy violations of HAVA.
• Second --and significantly-- the litigation clarified the right of voters to receive provisional
ballots, even though the election officials were certain they would not be counted. The
decision also defined an ancillary right –the right to be directed to the correct precinct.
There voters could cast a regular ballot that would be counted. If they insisted on casting
a provisional ballot in the wrong precinct, they would be on notice that it would be a
symbolic gesture only.
• Third, these lawsuits prompted election officials to take better care in instructing precinct
officials on how to notify voters about the need to go to the correct precinct in order to
cast a countable ballot – although the litigation regrettably came too late to be truly
effective in this regard. In many states, on Election Day 2004, the procedures in place
for notifying voters about where to go were less than ideal, reflecting less-than-ideal
procedures for training poll workers on this point.
There was also pre-election litigation over the question whether voters who had requested an
absentee ballot were entitled to cast a provisional ballot. In both cases (one in Colorado and
one, decided on Election Day, in Ohio), the federal courts ruled that HAVA requires that these
voters receive a provisional ballot. Afterwards, it is for state officials under state law to
determine whether these provisional ballots will be counted, in part by determining if these
provisional voters already had voted by absentee ballot (in which case one ballot should be
ruled ineligible, in order to avoid double voting). These decisions confirm the basic premise that
provisional ballots should be available whenever voters believe they are entitled to them, so that
their preferences can be recorded, with a subsequent determination whether these preferences
count as valid votes.
voters cast their ballots at the wrong precinct. To overcome this challenge, staffing was increased to prepare the
duplicate ballots.” In a close, contested election, “duplicate” ballots would doubtless receive long and careful
scrutiny.” See Appendix 7, GAO, “Views of Selected Local Election Officials on Managing Voter Registration and
Ensuring Eligible Citizens Can Vote,” September 2005. (GAO Report-05-997)
19
RECOMMENDATIONS
Because every provisional ballot counted represents a voter who, if the system had worked
perfectly, should have voted by regular ballot, the advent of statewide registration databases is
likely to reduce the use provisional ballots. The one area in which such databases may not
make a difference is for those who voted by provisional ballot because they did not bring
required identification documents to the polling place. The statewide voter registration database
will facilitate verifying that ballot, but the voter will still have to vote provisionally. Beyond that
exception, even with statewide registries in every state, provisional voting will remain an
important failsafe, and voters should have confidence that the failsafe will operate correctly.
The wide variation in the implementation of provisional voting among and particularly within
states suggests that EAC can help states strengthen their processes. Research-based
recommendations for best, or at least better, practices based on the experience gained in the
2004 election can be useful in states’ efforts to achieve greater consistency in the administration
of provisional voting.
Recommendations for Best Practices
Recent legislative activity shows that state efforts to improve the provisional voting process are
underway. Those states, as well as others that have not yet begun to correct shortcomings that
became apparent in 2004, can benefit from considering the best practices described here. By
recommending best practices, the EAC will offer informed advice while respecting diversity
among the states. One way to strengthen the recommendations and build a constituency for
them would be for EAC to ask its advisory committee members to recommend as best practices
procedures that have worked in their states.
Self-evaluation of Provisional Voting –4 Key Questions
The first step to achieving greater consistency within each state is to think about provisional
voting systematically. As legislators, election officials, and citizens in the states prepare for the
2006 election, they should ask themselves these questions about their provisional voting
systems.
1. Does the provisional voting system distribute, collect, record, and tally provisional ballots
with sufficient accuracy to be seen as procedurally legitimate by both supporters and
opponents of the winning candidate? Does the tally include all votes cast by properly
registered voters who correctly completed the steps required?
2. Is the provisional voting system sufficiently robust to perform well under the pressure of
a close election when ballot evaluation will be under scrutiny and litigation looms?
3. Do the procedural requirements of the system permit cost-efficient operation? Are the
administrative demands of the system reasonably related to the staff and other resource
requirements available?
4. How great is the variation in the use of provisional voting in counties or equivalent levels
of voting jurisdiction within the state? Is the variation great enough to cause concern that
the system may not be administered uniformly across the state?
If the answers to these questions leave room for doubt about the effectiveness of the system or
some of its parts, the EAC’s recommendation of best practices should provide the starting point
for a state’s effort to improve its provisional voting system.
20
Best Practices For Each Step In The Process
We examined each step of the provisional voting process to identify specific areas where the
states should focus their attention to reduce the inconsistencies noted in our analysis. We offer
recommendations in each area appropriate to the responsibilities that HAVA assigns the EAC
for the proper functioning of the provisional voting process.
The Importance of Clarity
The EAC should emphasize above all else the importance of clarity in the rules governing every
stage of provisional voting. As the Century Foundation’s recent report observed, “Close
elections increasingly may be settled in part by the evaluating and counting of provisional
ballots. . . To avoid post election disputes over provisional ballots—disputes that will diminish
public confidence in the accuracy and legitimacy of the result-- well in advance of the election,
states should establish, announce, and publicize clear statewide standards for every aspect of
the provisional ballot process, from who is entitled to receive a provisional ballot to which ones
are counted.”
26
Litigation surrounding the 2004 election resulted in decisions that, if reflected in state statutes or
regulations and disseminated in effective training for poll workers, can increase the clarity of
provisional ballot procedures, increase predictability, and bolster confidence in the system. By
taking the following steps, states can incorporate those court rulings into their procedures.
• Promulgate, ideally by legislation, clear standards for evaluating provisional ballots, and
provide training for the officials who will apply those standards. For example, in
Washington State, the court determined that an election official’s failure in evaluating
ballots to do a complete check against all signature records is an error serious enough to
warrant re-canvassing.
27
Clear direction by regulation or statute on what records to use
in evaluating ballots could have saved precious time and effort and increased the
reliability of the provisional voting system.
• States should provide standard information resources for the training of poll workers by
local jurisdictions. Training materials might include, for example, maps or databases with
instruction on how to locate polling places for potential voters who show up at the wrong
place. Usable and useful information in the hands of poll workers can protect voters from
being penalized by ministerial errors at the polling place.
28
• State training materials provided to local jurisdictions should make clear that the only
permissible requirement to obtain a provisional ballot is an affirmation that the voter is
registered in the jurisdiction and eligible to vote in an election for federal office.
29
Recent
legislation in Arizona indicates that recommendations should emphasize HAVA’s
requirement that persons appearing at the polling place claiming to be registered voters
cannot be denied a ballot because they do not have identification with them. Poll
26
The Century Foundation, Balancing Access and Integrity, Report of the Working Group on State Implementation of
Election Reforms, July 2005.
27
See Washington State Republican Party v. King County Division of Records, 103 P3d 725, 727-728 (Wash. 2004)
28
See Panio v. Sunderland 824 N.E.2d 488, 490 (NY, 2005) See also Order, Hawkins v. Blunt, No.04-4177-CV-C-
RED (W.D. Mo. October 12, 2004). While rejecting the notion that all ballots cast in the wrong precinct should be
counted, the court ruled that provisional votes cast in the wrong precinct should be thrown out provided that the voter
had been directed to the correct precinct. This meant that provisional votes cast in the wrong precinct (and even the
wrong polling place) would count if there were no evidence that the voter had been directed to a different polling
place. The court placed a duty upon election officials to make sure the voters were in the correct locations. Note that
this question would not arise in a state that counted ballots cast in the wrong polling place but within the correct
county.
29
Sandusky County Democratic Party v. Blackwell, 387 F.3d 565, 774 (6
th
Cir. 2004)
21
workers may need appropriate training to understand their duty to give such voters a
provisional ballot.
30
A. Registration and Pre-Election Information for Voters
Providing crisp, clear information to voters before the election is important to the success of the
provisional voting process. The better voters understand their rights and obligations, the easier
the system will be to manage, and the more legitimate the appearance of the process. States
can begin by assessing the utility and clarity of the information for voters on their websites and
by considering what information might be added to sample ballots mailed to voters before
elections. Best practices in this area would include:
1. If states require identification at the time of registration, the kind of IDs required should
be stated precisely and clearly and be publicly and widely available in a form that all
voters can understand. For example, “You must bring your driver’s license. If you don’t
have a driver’s license, then you must bring an ID card with your photograph on it and
this ID card must be issued by a government agency. ”
31
2. The process to re-enfranchise felons should be clear and straightforward. To avoid
litigation over the registration status of felons, best practice should be defined as making
re-enfranchisement automatic, or no more burdensome than the process required for
any new registrant.
32
3. State or county websites for voters should offer full, clear information on boundaries of
precincts, location of polling places, requirements for identification, and other necessary
guidance that will facilitate registration and the casting of a regular ballot. An 800
number should also be provided. Models are available: the statewide databases in
Florida and Michigan provide voters with provisional voting information, registration
verification and precinct location information.
B. At the Polling Place
Avoiding error at the polling place will allow more voters to cast a regular ballot and all others
who request it to cast a provisional ballot.
1. The layout and staffing of the polling place, particularly the multi-precinct polling place is
important. Greeters, maps, and prominently posted voter information about provisional
ballots, ID requirements, and related topics can help the potential voters cast their ballot
in the right place. States should require poll workers to be familiar with the options and
provide the resources needed for them to achieve the knowledge needed to be helpful
and effective. Colorado has clear regulations on polling place requirements, including
HAVA information and voting demonstration display.
33
Many states require training of
poll workers. In some states that requirement is recent: after the 2004 election, New
Mexico adopted a requirement for poll workers to attend an “election school.”
34
A state
30
The Florida Democratic Party v. Hood, 342 F. Supp. 2d 1073, 1075-76 (N.D. Fla. 2004). The court explained that
provisional voting is designed to correct the situation that occurs when election officials do not have perfect
knowledge and when they make incorrect determinations about eligibility (the “fail-safe” notion). Denying voters
provisional ballots because of on-the-spot determinations directly contradicts this idea. Even before the cited
decision, the Florida Secretary of State’s office had determined that any voter who makes the declaration required by
federal law is entitled to vote a provisional ballot, even if the voter is in the wrong precinct.
31
Websites in 29 states describe, with varying degrees of specificity, the identification voters may need. In 18 states
voters can learn something about the precinct in which they should vote. And in 6 states (California, District of
Columbia, Kentucky, Michigan, North Carolina, and South Carolina) they can verify their registration on the website.
32
The Century Foundation, op. cit.
33
8 Colo. Code Regs. § 1505-1, Rule 7.1.
34
2005 N.M. Laws 270 page no. 4-5.
22
statutory requirement for training could facilitate uniform instruction of poll workers in
those states that do not already provide it.
2. The provisional ballot should be of a design or color sufficiently different from a regular
ballot to avoid confusion over counting, as occurred in Washington State. The ballot
might include a tear-off leaflet with information for voters such as: “Reasons Why Your
Provisional Ballot Might Not Be Counted” on one side and “What to Do if My Provisional
Ballot Is Not Counted” on the other.
3. Because provisional ballots offer a fail-safe, supplies of the ballots at each polling place
should be sufficient for all the potential voters likely to need them. In 2004, some polling
places ran out of ballots, with unknown effects on the opportunity to vote. In Middlesex
County, New Jersey, for example, on Election Day the Superior Court ordered the
county clerk to assure that sufficient provisional ballots were available at several heavily
used polling places, and it authorized the clerk “in the event additional provisional ballots
are required . . .to photocopy official provisional ballots.”
35
At least two states,
Connecticut and Delaware, provide guidelines to local election officials on how to
estimate the demand for provisional ballots. Connecticut sets the number at 1% of the
voters in the district, Delaware at 6%.
36
States that do not offer a practical method to
guide the supply of provisional ballots at polling places should consider doing so. The
guideline should take into account both the number of voters in the district and the
number of provisional ballots actually cast in recent elections.
4. To achieve the procedural clarity needed to forestall disputes, states should establish a
clear chain of custody for the handling of provisional ballots from production through
distribution, collection and, finally, evaluation. A number of states have clear procedures
for at least parts of this chain of custody. All states should examine their chain-of-
custody requirements for clarity. Illinois includes the potentially beneficial requirement
that ballots be transported by bi-partisan teams, which offers the potential to avoid some
charges of election fraud.
C. Evaluating Voter Eligibility and Counting Provisional Ballots
The clarity of criteria for evaluating voter eligibility is critical to a sound process for deciding
which of the cast provisional ballots should be counted. Public recognition of the validity of those
criteria is important to establishing the legitimacy of the system as a whole. The experience in
2004 in North Carolina, Washington, and Ohio underlines the importance of clear criteria. As the
Century Foundation report put it, “Whatever procedures the states choose [to determine if a
provisional ballot should be counted], the paramount consideration—as with all others
concerning provisional voting—is that they be clear and thus not susceptible to post-election
manipulation and litigation.”
37
Nonetheless, the Panio v. Sutherland
38
decision in New York
shows the difficulty of defining the range of administrative errors from which the provisional
voters should be held harmless. Even when the standard is “clerical error” judges can differ over
what that means exactly. Possibly a state law might be able to clarify a definition by giving
examples of clerical errors, but even then the definition is unlikely to be perfect.
35
Voting Order, November 2, 2004, Superior Court of New Jersey, Law Division, Middlesex County.
36
Connecticut: “Equal to or not less than 1% of the number of electors who are eligible to vote in any given district, or
such other number as the municipal clerk and the registrars agree is sufficient to protect voting rights. Conn. Gen.
Stat. Ann. § 9-232j.Delaware: Each County Department of Elections Office is required to provide to each election
district a number of provisional ballots equal to 6% of registered voters in that district, with a minimum allocation of 15
ballots. Additional supplies to be delivered when the supply becomes “very low.” Del.Code Ann. Tit 15 § 4948(e).
37
The Century Foundation, op. cit.
38
4 N.Y.3d 123, 824 N.E.2d 488 (N.Y. 2005) and Memorandum (LaPlante—Foley) Provisional Ballot Cases by State,
July 19, 2005.
23
.
1. State statutes or regulations should define a reasonable period for voters who lack the
HAVA-specified ID or other information bearing on their eligibility to provide it in order to
facilitate the state’s ability to verify that the person casting the provisional ballot is the
same one who registered. While there may be a concern to ensure that the individual
who returns with the ID may not be the same individual who cast the provisional ballot,
the spirit of HAVA demands that the opportunity to prove identity be provided after
Election Day. A signature match can go far in establishing that the individual who voted
and the individual returning later with identification is, in fact, the same person.
Encouraging a voter who lacks ID on Election Day to return later to help the verification
process by providing proper identification will strengthen the system and increase public
confidence in the electoral process. Our data indicate that some voters would prefer to
return with ID rather than to sign an affidavit, perhaps because of uncertainty about the
legal process involved in the affidavit. At least 11 states allow voters to provide ID or
other information one to 13 days after voting. Of particular interest is Kansas, which
allows voters to proffer their ID by electronic means or by mail, as well as in person.
39
2. More provisional ballots are counted in those states that verify ballots cast outside the
correct precinct.
40
While HAVA arguably leaves this decision up to the states, pointing
out the effect of the narrower definition on the portion of ballots counted could be useful
to the states in deciding this question. States should be aware, however, of the
additional burden placed on the ballot-evaluation process when out-of-precinct ballots
are considered. And tradeoffs are involved if out-of-precinct voters are unable to vote for
the local offices that might appear on the ballot in their district of residence. One option
for states is to involve the voters in the decision by pointing out that voters who cast their
provisional ballots in the wrong precinct may not be able to participate in the local
election. The voter could then decide to go to the correct precinct or vote provisionally
for the higher offices at the top of the ticket only.
3. Alternatively, if a state chooses to require voters to appear at their assigned precinct,
where the same polling site serves more than one precinct, a voter’s provisional ballot
should count so long as the voter cast that ballot at the correct polling site even if at the
wrong precinct within that location.
41
Ideally the voter could be directed to the correct
machine, but poll worker advice will not always be correct. One way to assess the
balance of issues here is to consider that, if a voter in a multi-precinct polling place is
sent to the wrong machine, the error is probably the poll worker’s, and the voter should
not be penalized.
39
In Kansas, the voter can provide ID to a County Election Officer any time before the County Board of Canvassers
meets to count provisional ballots. KS. ST. 25-1122(d). ID can be presented in person, OR via mail or electronic
means. Id. The Board must meet either on the Friday or Monday following a Tuesday election. Id. at 25-3104.
Deadlines in other states are: Alabama -- 5:00 P.M. on the Monday following the election AL ST § 17-10A-2(c)
(1)
Florida: until 5:00 P.M. on the third day following the election . Fla. Stat. Ann. § 101.048 (adopted after the 2004
election);Georgia—no later than 2 days after the election. GA ST § 21-2-417; 419. Illinois- 2 days to submit additional
information 10 Ill. Comp. Stat. Ann. 5/18A-15(d); Indiana— in 2004 the deadline was the close of the polls IN. ST. §.
3-11.7-5-2(a). The time period was extended to 13 days by the adoption of Indiana Code 3-11-8, Section 25,
Subsection (l); Maryland—until the meeting of the Election Board; MD ELEC LAW § 11-303. New Jersey— until the
close of business on the second day after the election 19:53C-3(i). Nevada— until 5:00 P.M. on the Friday following
the election NV ST 293.3085; New Mexico—until 7:00 P.M. on Election Day NM ADC 1.10.22 (8) (H).
40
See Andersen, op. cit, pgs. 23 – 24 for an analysis of the significant effect of counting out-of-precinct ballots. The
Election Day Survey found that, “Most notably, jurisdictions that permitted jurisdiction-wide acceptance of provisional
ballots reported higher rates of provisional ballots being cast, but also reported a much higher incidence of provisional
ballots being counted, than other jurisdictions.”
41
Chances are administrative error accounts for the voter being directed to the wrong precinct under these
circumstances.
24
4. Officials should follow a written procedure, and perhaps a checklist, to identify the
reason why a provisional ballot is rejected (e.g., check the applicable box “unregistered
voter”; “lack of signature match” “wrong precinct,” etc.) Those forms should be disclosed
publicly when completed. Colorado’s election rules offer particularly clear guidance to
the official evaluating a provisional ballot.
42
Colorado Rejection Codes (Any ballot given a rejection code shall not be counted):
RFS (Rejection federal or state) No federal or state candidates or issues to
duplicate.
RNS (Rejection not signed) Provisional Ballot Affidavit not signed.
RIN (Rejection incomplete information provided) Required information is
incomplete and the designated election official is unable to confirm voter’s
eligibility.
RNR (Rejection not registered) Voter did not register by the voter registration
deadline or by emergency registration, Colorado voter registration record
was not found, or voter was previously cancelled and has not been
reinstated pursuant to 1-2-605(10). C.R.S.
REE (Rejection envelope empty) Provisional ballot envelope is empty.
RAB (Rejection voter voted absentee) Designated election official has
confirmed that voter voted an absentee ballot.
REV (Rejection based on ballot cast in early voting) Voter voted early.
RIP (Rejection based on incorrect party) Incorrect Party in Primary Election.
RFE (Rejection felon not eligible to vote) Individual was convicted of a felony
and is either serving a sentence of confinement or detention or is on
parole.
RWC (Rejection elector not registered in county or State of Colorado) Non-
county or non-state resident; therefore voter not eligible to vote in the
county where the provisional ballot was voted.
RID (Rejection first time voter has not supplied identification upon registration
or thereafter prior to and during time voter voted) First Time Voter who
registered by mail or through a voter registration drive, is tagged as id
deficient, and did not provide id at the time of voting.
RRD (Rejection registration deficient) Voter had deficient or incomplete
registration and required information was not provided prior to or at the
time of filling in the provisional ballot envelope. Voter’s eligibility cannot
be established.
D. Verification of Provisional Ballots
1. States that use the information on the provisional ballot to permit voters who have
changed their addresses to update their registrations should adopt clear procedures on
that process and specify how the new information will be communicated between
different Boards of Elections.
2. In verifying provisional ballots, the time by which election officials must make their
eligibility determinations is particularly important in presidential elections because of the
need to certify electors to the Electoral College. States should consider in particular how
to divide the time constraints imposed in presidential election by the safe-harbor
provisions regarding certification to the Electoral College. Some part of this five-week
period will be consumed by the eligibility evaluation, but states should take care to
provide a sufficient period of time as well for challenges. If a state consumes 21 days
42
8 CCR 1505-1, at 26.5.4, adopted august 4, 2005. See also 1-2-509(3) C.R.S.
25
following the election in the eligibility evaluations, only two weeks will remain for legal
challenges to be concluded. Is that sufficient? Or should the state provide the resources
needed to complete the eligibility determinations in 10 days or two weeks, leaving three
weeks or more for legal challenges in a close election? Our research did not identify an
optimum division of the five weeks available. The prudent course here would be to
encourage states to consider the issue and then make a careful decision about how to
complete all steps in the evaluation of ballots and challenges to those determinations
within the five weeks available.
E. Post-election Information for Voters
Timely information to voters about the disposition of their provisional ballot will provide helpful
feedback and more important enable voters to determine if they are registered for future
elections and, if not, what they need to do to become registered.
1. Establish mechanisms to ensure that voters casting provisional ballots are informed
whether they are now registered for future elections and, if not, what they need to do to
become registered.
F. State Laws Governing Litigation over Provisional Voting
1. Establish special, streamlined litigation procedures for Election Day complaints that
individuals are being denied the right to cast a provisional ballot.
Broader Considerations
G. Integrity and the Appearance of Integrity
1. State laws or regulations providing for non-partisan or bi-partisan bodies to make a
public determination of the validity of provisional ballots would increase confidence in the
system.
2. To improve transparency, state laws or regulations should require the purging process
for registration to be public and with an opportunity for voters to correct an erroneous
determination that they should be purged.
3. State laws or regulation should require the evaluation process for provisional ballots to
be public, while protecting the names of those who voted provisionally.
H. Continuous Assessment of the Provisional Ballot -- Process and Performance
Defining what makes for a successful provisional voting system is difficult. The most successful
system is probably not the one with the most provisional votes cast (that could indicate
problems with the registration system). Nor is the system with the greatest number counted or
with the fewest counted necessarily superior because the evaluation process could be flawed.
Defining quality requires a broad perspective about how well the system works, how open it is to
error recognition and correction, and how well provisional voting processes are connected to the
registration and voter identification regimes. The EAC should consider engaging one of the
national quality organizations or processes, such as Six Sigma
43
or the Baldridge Quality
43
Six Sigma is a measure of quality that strives for near perfection. Six Sigma is a disciplined, data-driven approach
and methodology for eliminating defects (driving towards six standard deviations between the mean and the nearest
specification limit) in any process -- from manufacturing to transactional and from product to service.
26
process
44
to evaluate the provisional ballot process. Pending such a review, the EAC can
recommend that states take the following actions.
1. Recognize that the first step to improving quality is to see the provisional voting process
as a system and take a systems approach to regular evaluation through standardized
metrics with explicit goals for performance.
2. States should begin by collecting data systematically on the provisional voting process
so that they can evaluate their voting system and assess changes from one election to
the next. The effort should start in the 2006 election, and the data collected should
include:
-- Provisional votes cast and counted by jurisdiction, say counties, with details on
why the voter had to vote provisionally (lack of ID, not on list, challenged at
polling place, issued absentee ballot, etc) and number of ballots actually
counted in each category.
-- Reasons why provisional ballots were not counted, using categories such as
those that have been adopted by Colorado, described earlier in this report.
-- Measures of variance among jurisdictions.
-- Number of poll workers trained in administration of provisional voting by polling
place.
-- Number of jurisdictions posting information on provisional voting in the polling
place.
-- Time required to evaluate ballots by jurisdiction.
Improving understanding of the provisional voting process through analysis of detailed
information will enable state and local election officials to strengthen their systems. By collecting
and analyzing this data states can identify which aspects of the registration and electoral system
are most important in shunting voters into the provisional ballot process. Responsible officials
can then look to their registration system, identification requirements or poll worker training as a
way to reduce the need for voters to cast their ballots provisionally.
44
The Baldrige Criteria for Performance Excellence provide a systems perspective for understanding performance
management. They reflect validated, leading-edge management practices against which an organization can
measure itself. With their acceptance nationally and internationally as the model for performance excellence, the
Criteria represent a common language for communication among organizations for sharing
best practices. The
Criteria are also the basis for the Malcolm Baldrige National Quality Award process.
27
ATTACHMENT 1 – Data Sources for Classification of the States
Our research on provisional voting divided the various states into several categories to allow an assessment of how
different factors may have influenced the process of casting and counting provisional ballots. This analysis was
conducted before the release of the Election Day Study, and the categories we used may differ in some respects from
its work. The variables used to analyze a state’s use of provisional ballots were:
1. New vs. Old (states that used a provisional ballot before the 2004 election)
2. Use of a statewide database of registered voters vs. no use of a statewide database
3. Counting out-of-precinct ballots vs. not counting out-of-precinct ballots
4. Voter identification requirements
5. Method used to verify provisional ballots
6. Levels of provisional ballots cast and counted
We first assigned states within these categories based on classifications done by Electionline.org in its studies. The
Electionline data was the only published information available at the time of our research. We reviewed the
Electionline data carefully, and, in select cases, updated it with new, detailed information that had become available
after its publication. The changes we made are explained below.
--Idaho, Maine, Minnesota, New Hampshire, Wisconsin and Wyoming were excluded from our analysis. They
have election-day registration systems, and did not need to use HAVA-compliant provisional ballots.
--North Dakota does not register voters, so it also was excluded from HAVA requirements and did not use
provisional voting.
--Mississippi has not reported its provisional voting results and could not be included in our analysis, though it
was compliant in 2004.
--Pennsylvania did not report its totals for the Election Day Study, but we obtained information on Pennsylvania
and included it in our analysis.
New vs. Old States
We classified states as “new” or “old” based on the 2001 Electionline study of provisional voting,
45
but
condensed its classifications into a single dichotomous variable, new/old with all other cases excluded. The
Electionline study divided states into five categories of their use of provisional ballots in the 2000 election:
1. Use of provisional ballots (P)
2. Limited use of provisional ballots (LP)
3. Affidavit ballots (A)
4. No system in place (N)
5. Unnecessary/Not Applicable (U/NA)
We included in the list of “Old States” all states listed as using provisional ballots, limited use of provisional ballots
or affidavit ballots. States in all three categories would have been familiar with key aspects of provisional voting..
States that had no provisional voting system in place for the 2002 election, and were HAVA compliant in 2004,
were listed as “new” states, as 2004 would have been the first year in which they would be offering the option of
provisional voting. States that were listed as unnecessary or not applicable were excluded from this study, as they
were exempt from the HAVA regulations in 2004 because they either allowed same-day registration or did not
register voters.
45
This study can be found at: http://electionline.org/Portals/1/Publications/Provisional%20Voting.pdf.
28
Rhode Island is the only state categorized as an old state by Electionline that we moved into the list of new states.
Electionline’s map shows Rhode Island as a state that used provisional voting in 2000, but in the state description, it
is listed as having no system in place. We learned from the Rhode Island Board of Elections that the state had
previously permitted potential voters to sign an affidavit if they did not appear on a precinct’s list of registered
voters, but felt they were registered to vote. Based on the signed affidavit, the election official would then contact a
county official to see if the voter was on a more complete registration list. If the voter’s name was on the complete
list, that voter was permitted to cast a regular ballot. As this process did not grant the voter a provisional ballot, but
served as a different type of administrative failsafe, we concluded that Rhode Island’s first use of provisional voting
was in 2004 and, therefore, classified the state as “new” to the system of provisional balloting.
Table 1
CATEGORIZATION OF STATES -- Old and New
Old States New States HAVA Exempt or NA
Alaska Connecticut Idaho
Alabama Delaware Maine
Arkansas Georgia Minnesota
California Hawaii New Hampshire
Colorado Illinois North Dakota
DC Indiana Wisconsin
Florida Louisiana Wyoming
Iowa Massachusetts
Kansas Missouri
Kentucky Montana
Maryland Nevada
Michigan Oklahoma
Mississippi Pennsylvania
Nebraska Rhode Island
New Jersey South Dakota
New Mexico Tennessee
New York Utah
North Carolina Vermont
Ohio
Oregon
South Carolina
Texas
Virginia
Washington
West Virginia
26 18 7
29
Statewide List of Registered Voters
The Electionline preview of the 2004 Election
46
was the starting point for compiling a list of states that had a
statewide database of registered voters. That study listed 34 States that did not have their statewide database systems
complete, and 16 that did, including the District of Columbia. North Dakota does not register voters, so does not
need to compile such a database. Electionline’s criterion for concluding that a state had a statewide list was that the
state have participation from all jurisdictions in a statewide system. We added Oklahoma to the list of states with
statewide databases because we found it had met the Electionline criteria by the 2004 election, albeit too late for
inclusion in the Electionline survey.
Out-of-Precinct Ballots
We based our classification of states that allow the counting of ballots cast outside the correct precinct on the data in
the 2004 Electionline preview of the 2004 election
2
. States that evaluated ballots cast in a precinct where the voter
was not registered were categorized as “out-of-precinct.” States that invalidated such ballots were categorized as
“In-precinct only.”
Table 2
CATEGORIZATION OF STATES -- Counting Out-Of-Precinct Ballots
Out-of-Precinct In-Precinct Only HAVA EXEMPT OR NA
Alaska Alabama Idaho
Arkansas Arizona Maine
California Colorado Mississippi
Georgia Connecticut New Hampshire
Illinois
47
Delaware North Dakota
Kansas District of Columbia Wisconsin
Louisiana Florida Wyoming
Maryland Hawaii
New Mexico Indiana
North Carolina Iowa
Oregon Kentucky
Pennsylvania Massachusetts
Rhode Island Michigan
Utah Missouri
Vermont Montana
Washington Nebraska
Nevada
New Jersey
New York
Ohio
Oklahoma
South Carolina
South Dakota
Tennessee
Texas
Virginia
West Virginia
16 27 7
46
“Election Preview 2004: What’s changed, What Hasn’t and Why”. This study can be found at:
http://electionline.org/Portals/1/Publications/Election.preview.2004.report.final.update.pdf
47
In Illinois, it is not clear that all counties followed this procedure. Some counties may not have counted out-of-
precinct ballots.
30
Verification Method
We identified four different ways states assessed provisional ballots to determine if they should be counted:
signature match, match voter data, signed affidavits, and bringing back identification later. We gathered information
about these verification techniques by checking state websites and consulting journalistic accounts. We consulted
state legislation to provide further information where needed.
Table 3
CATEGORIZATION OF STATES -- Ballot Evaluation Methods
Signature
Match
Data
Match
Affidavit Return with ID NA
Alaska Alabama Connecticut Indiana Idaho
California Arizona Delaware Iowa Maine
Florida Arkansas Georgia Kansas Mississippi
Oregon Colorado Hawaii Maryland Minnesota
DC Illinois Michigan New Hampshire
Louisiana Kentucky Montana N. Carolina
*
Missouri Massachusetts New Jersey N. Dakota
Ohio Nebraska New Mexico Wisconsin
Oklahoma Nevada Texas Wyoming
Pennsylvania New York Utah
Rhode Island South Dakota
S. Carolina Tennessee
Washington Vermont
West Virginia Virginia
4 14 14 10 9
.
*
North Carolina lacked clear standards to evaluate provisional ballots and is excluded from this analysis.
31
Data Collection
To assemble our data for analysis, we began by using the data on provisional votes cast and counted reported by
Electionline. To increase the accuracy of this data, we surveyed each state’s election websites for updated data, and
for reported numbers on the county level. We then sent emails to 49 (we excluded Alaska, see below) states and the
District of Columbia, requesting updated data on the number of provisional votes cast and counted by county. We
received information from 25 states by our cut-off date of August 25, 2005.
Table 4
Updated information by State
Received Updated Data Did Not Receive
Updated Data
California Alabama
District of Columbia Alaska
48
Florida Arizona
Hawaii Arkansas
Indiana Colorado
Iowa Connecticut
Kansas Delaware
Louisiana Georgia
Maryland
49
Idaho
Missouri Illinois
Montana Kentucky
Nebraska
50
Maine
Nevada Massachusetts
New Jersey Michigan
New Mexico Minnesota
Ohio Mississippi
Oklahoma New Hampshire
Oregon New York
Pennsylvania North Carolina
Rhode Island North Dakota
South Dakota South Carolina
Tennessee Utah
Texas Vermont
Virginia Wisconsin
Washington Wyoming
West Virginia
26 States 25 States
48
Alaska was not contacted via email, as the state does not have voting districts comparable to counties in other
states and could not be matched with comparable census data.
49
Maryland reported provisional ballots that were counted per county, but not number cast.
50
Nebraska reported an incomplete list of provisional ballots cast and counted by county, but designated counties by
number, rather than by name.
32
Data Differences
The data used in this study differ from the data reported in the Election Day Study for 19 states. The Election Day
Study was not completed until well after our statistical analysis of provisional voting was finished. Where there are
differences, they are typically very small, usually fewer than 100 votes either cast or counted. Of the 9 states that
have differences of more than 100 votes cast or counted, 7 have reported their numbers directly to us and can be
considered updated data that EDS had not obtained. For one of those states, New Mexico, EDS had incomplete
data, and for another, Pennsylvania, EDS had no data at all. The data that we have collected reflects updated
numbers from the states that have changed following recounts and litigation that altered how ballots were evaluated.
Table 5
Data Differences with the Election Day Study
State EDS Numbers
Cast/Counted
Our Numbers
Cast/Counted
Differences Updated Info
from State?
51
Alabama
6,478/1,865 6560/1836 82/29 No
Alaska
23,285/22,498 23,275/22,498 10/0 No
Colorado
51,529/39,086 51,477/39,163 52/77 No
Georgia
12,893/4,489 12,893/3,839 0/650 No
Hawaii
346/25 348/25 2/0 Yes
Iowa
15,406/8,038 15,454/8,048 48/10 Yes
Kansas
45,535/32,079 45,563/31,805 28/274 Yes
Montana
688/378 653/357 35/21 Yes
Nebraska
17,421/13,788 17,003/13,298 418/490 Yes
Nevada
6,153/2,446 6,154/2,447 1/1 Yes
New Mexico
6,410/2,914 15,360/8,767 8,950/5,853 Yes
N. Carolina
77,469/50,370 77,469/42,348 0/8,022 No
Ohio
157,714/123,902 158,642/123,548 928/354 Yes
Pennsylvania
No data 53,698/26,092 53,698/26,092 Yes
Texas
35,282/7,156 36,193/7,770 911/614 Yes
Vermont
121/30 101/37 20/7 No
Virginia
4,608/728 4,609/728 1/0 Yes
Washington
92,402/73,806 86,239/69,273 6,163/4,533 Yes
Wisconsin
374/119 373/120 1/1 No
51
Data not provided by the state itself is taken from Electionline figures.
